multi-phase packed-bed thermogravimetric analyzer

135
Multi-Phase Packed-Bed Thermogravimetric Analyzer by Erik G. Jarrett (Under the direction of K.C. Das) Abstract A bench scale (571 mL capacity) packed-bed thermogravimetric analyzer (PB-TGA) with a condensate measurement unit was designed, built, characterized, and tested to study both the solid decomposition and condensate formation characteristics during the pyrolysis of biomass at a constant heating rate. This attempts to satisfy limitations of current analytical systems such as high sensitivity to particle homogeneity, the limited ability to observe bio-oil formation characteristics, and limitations in carrier gas based mass transfer. Weight changes in both the solid phase and condensate were measured from 25 C to 600 C at 3 C min with a nitrogen carrier gas flowrate of 0.25 L min . The PB-TGA was tested and compared to a commercially available analytical TGA using a cellulosic powder and pine-pellets. The PB- TGA experiments exhibited increased char yield and thermal lag, more symmetric and lower magnitude differential thermogravimetric (DTG) curves, and differences in reaction kinetic parameters. Index words: TGA; thermal analysis; development; condensate; char; cellulose; pine-pellets; pyrolysis; theses (academic)

Upload: others

Post on 08-Jan-2022

9 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Multi-Phase Packed-Bed

Thermogravimetric Analyzer

by

Erik G. Jarrett

(Under the direction of K.C. Das)

Abstract

A bench scale (571 mL capacity) packed-bed thermogravimetric analyzer (PB-TGA) with

a condensate measurement unit was designed, built, characterized, and tested to study both

the solid decomposition and condensate formation characteristics during the pyrolysis of

biomass at a constant heating rate. This attempts to satisfy limitations of current analytical

systems such as high sensitivity to particle homogeneity, the limited ability to observe bio-oil

formation characteristics, and limitations in carrier gas based mass transfer. Weight changes

in both the solid phase and condensate were measured from ∽25◦C to 600◦C at 3◦Cmin

with

a nitrogen carrier gas flowrate of 0.25 Lmin

. The PB-TGA was tested and compared to a

commercially available analytical TGA using a cellulosic powder and pine-pellets. The PB-

TGA experiments exhibited increased char yield and thermal lag, more symmetric and lower

magnitude differential thermogravimetric (DTG) curves, and differences in reaction kinetic

parameters.

Index words: TGA; thermal analysis; development; condensate; char; cellulose;pine-pellets; pyrolysis; theses (academic)

Page 2: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Multi-Phase Packed-Bed

Thermogravimetric Analyzer

by

Erik G. Jarrett

B.S. The University of Georgia, 2005

A Thesis Submitted to the Graduate Faculty

of The University of Georgia in Partial Fulfillment

of the

Requirements for the Degree

Master of Science

Athens, Georgia

2008

Page 3: Multi-Phase Packed-Bed Thermogravimetric Analyzer

c© 2008

Erik G. Jarrett

All Rights Reserved

Page 4: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Multi-Phase Packed-Bed

Thermogravimetric Analyzer

by

Erik G. Jarrett

Approved:

Major Professor: K.C. Das

Committee: Mark Eiteman

James Kastner

Sudhagar Mani

Electronic Version Approved:

Maureen Grasso

Dean of the Graduate School

The University of Georgia

December 2008

Page 5: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Dedication

I would like to dedicate this thesis to the Ramsey sports center (for keeping my blood flowing),

cows (for the yogurt and cheese I ate), REI and the people of REI (for supplying me with

tools for distraction), my cool Lamy fountain pens, all the millipedes that died (annual

migration from the front door to the bathroom), Kashi and FiberOne (for regularity), little

red beans (without which I would probably would have evaporated), lots of electricity, and

all the trees that went into my paper.

iv

Page 6: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Acknowledgments

Dr. Das, you have put up with a lot and given a great deal of patience. In doing so, you have

given me the opportunity of a lifetime. You let me take my idea, fly with it, and finish it

despite all of my distractions. Thank you so much for taking me as student, supporting my

ideas and aspirations, supporting me financially, and supporting the project. Did I mention

patience? Thank you Dr. Das.

To the rest of my committee, Dr. Kastner, Dr. Eiteman, and Dr. Mani, thank you so

much for taking the time to critically review this work. It is not easy sort through all the

words and numbers that evidently make sense in my mind. It requires patience, time, and

an eye for detail. Thank you for helping me complete my goals!

Roger, thank you so much for your help. You helped do a lot of the initial algae TGA

runs, helped structure my time, read my thesis; and, you have been a great person to go to

for advice. You have helped quite a lot in this project and I am truly thankful for you being

around. I am most thankful, however, for your death cake input and consumption!

To Dr. Kisaalita, Dr. Keidekkar, Steve McDonald, and Pat. All four of you provided

valuable input, whether it be technical or moral advice, that has help move this project

along. Thank you!

Dear Anjali, you have passed through many “um, this may take longer to finish...” mo-

ments. You have also passed through a great deal of “Erik freaking out”, “Erik confused”,

and “Erik distracted” moments. You have supported me in every way and really helped push

me along. You are a great motivator, cook, fitness motivator, and are my adventure enabler.

Thanks Anjali, I love you so much!

v

Page 7: Multi-Phase Packed-Bed Thermogravimetric Analyzer

vi

Parents: Thanks for editorial and proofreading help! Reader, please know that Mom and

Dad are super editors! Both of you have given great support, advice, and encouragement. You

also, of course, helped make me. Thank you raising me to have an inquisitive and creative

mind (reader, if I don’t then don’t burst my bubble!). Thanks for all your love. I love you

guys!

Chris and Karen: You have given great support! Thanks so much for the logistical help,

providing wonderful coffee, giving ergonomic chairs, providing annoy-a-tron fun, keeping my

feet warm with toed socks (and other types), providing me with a babble receptacle, being a

healthy living inspiration, providing a multi-meter, printer, good books, and phone support

etc.. Thanks! I love you two too to 2 II!

To all of the Philiposes; Beena, Prasad, Ashu, Alex (Appu), and of course Anjali again.

Thank you for putting together the wonderful daughter that I married (that would mostly be

Beena and Prasad), being so supportive, making such good food, and of course for making

me a part of your family! I love all of you, thank you!

Friends: Jesus, Brenda, and Ravi, you three encouraged me at the beginning, introduced

me to graduate life, and shared my first graduate coffee. Brenda, you have always had a way

of motivating me and keeping me on track. Thank you so much. Jena, Jarrod, and Singh,

you helped a great deal, given great advice, given support, and been great friends! Alex,

Brandon, and David thanks for putting up with the “zoned out where’s Erik” Erik. Dinene,

thanks for being a wonderful friend who I know will always be there. Wherever you are, I

love you buddy!

Finally, Dusty, I hope your life is full of mental stimulation, friendly humans, adventures

in the forest, and good food. Thanks for listening to my likely incomprehensible babble.

This thesis typeset using LYX and LATEX 2ε on Linux. Dr. Michael Covington worked

hard to develop and maintain UGA’s thesis LATEX 2ε style for years. Thank you much.

Page 8: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Table of Contents

Page

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x

Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

Chapter

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Development of a bench sized packed-bed thermogravimetric an-

alyzer for both solid decomposition and condensate formation . 14

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2 Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.3 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . 24

2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3 Comparison of a novel bench size packed-bed solid and liquid

phase TGA and micro-TGA . . . . . . . . . . . . . . . . . . . . . . . . 48

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

vii

Page 9: Multi-Phase Packed-Bed Thermogravimetric Analyzer

viii

3.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . 57

3.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.1 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . 75

Appendix

A Solid Phase load cell conditioning circuit . . . . . . . . . . . . . . 77

B R-Code for developing the calibration model . . . . . . . . . . . . 78

C R-Code for kinetics analysis . . . . . . . . . . . . . . . . . . . . . . . 86

D Example Procedures for a blank run . . . . . . . . . . . . . . . . . . 96

D.1 Customized Experimental Section . . . . . . . . . . . . . . . . 97

D.2 Notes During Run (Record timer time and furnace PID

temp) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

D.3 Experiment Procedures . . . . . . . . . . . . . . . . . . . . . . 101

D.4 Opening and Cleaning Procedures . . . . . . . . . . . . . . . 106

E Example Procedures for a Biomass run . . . . . . . . . . . . . . . . 110

E.1 Customized Experimental Section . . . . . . . . . . . . . . . . 111

E.2 Experiment Procedure Section . . . . . . . . . . . . . . . . . 115

E.3 Opening and Cleaning Procedures . . . . . . . . . . . . . . . 120

Page 10: Multi-Phase Packed-Bed Thermogravimetric Analyzer

List of Figures

2.1 PB-TGA flow diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.2 Section view of the solid-phase decomposition system. . . . . . . . . . . . . . 20

2.3 Picture of the condensate system. . . . . . . . . . . . . . . . . . . . . . . . . 23

2.4 Calibration model residuals where residuals are the gram difference between

the predicted weight and the actual applied weight. . . . . . . . . . . . . . . 32

2.5 Blank run time profiles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.6 Uncorrected cellulose time profiles. . . . . . . . . . . . . . . . . . . . . . . . 36

2.7 Blank run corrected cellulose time profiles. . . . . . . . . . . . . . . . . . . . 37

2.8 Blank run corrected temperature based TGA and DTG profiles of cellulose

runs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.9 Blank run corrected TGA profiles of cellulose runs. . . . . . . . . . . . . . . 43

3.1 Flow chart of the PB-TGA. . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.2 Cellulose profile showing temperature and weights as a function of time. . . . 58

3.3 Pine Pellet profile showing temperature and weights as a function of time. . 59

3.4 Cellulose TGA and DTG values as a function of mean temperature. . . . . . 62

3.5 Pine Pellet TGA and DTG values as a function of mean temperature. . . . . 63

A.1 Schematic diagram of the loadcell conditioning circuit contained within the

balance mechanism enclosure. . . . . . . . . . . . . . . . . . . . . . . . . . . 77

ix

Page 11: Multi-Phase Packed-Bed Thermogravimetric Analyzer

List of Tables

1.1 Functions f(x) as described by Guo and Lua (Guo2001). . . . . . . . . . . . 7

2.1 Weight calibration experimental design. . . . . . . . . . . . . . . . . . . . . . 25

2.2 Feedstock Characteristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.3 Calibration regression model statistics. . . . . . . . . . . . . . . . . . . . . . 30

2.4 Maximum errors observed during the entire temperature program. . . . . . . 38

2.5 Corrected cellulose product yields. . . . . . . . . . . . . . . . . . . . . . . . . 39

2.6 Mean (n=4) cellulose DTG Peak 1 values and locations . . . . . . . . . . . . 41

2.7 Mean (n=4) cellulose DTG Peak 2 values (DratioD◦C

= 1◦C

) and locations (tem-

perature and time). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.1 Feedstock Characteristics. Values are in percent with 95% confidence intervals

(α = 0.05, n=3). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.2 Product yields after entire temperature program. Weights are percent normal-

ized to the initial sample weight. (α = 0.05, n=4). . . . . . . . . . . . . . . . 57

3.3 Maximum errors observed during the entire temperature program. Values are

the 95% confidence intervals (α = 0.05, n=4) of the normalized mean weight. 60

3.4 Temperature DTG (dWnormalizeddTemperature

) peak locations. Values are the change in

normalized weight over the change in mean temperature (n=4) of the sample

and furnace/reference temperatures. . . . . . . . . . . . . . . . . . . . . . . 64

3.5 Cellulose Kinetics and R2 values based on the X conversion (X = Wo−WWo−Wf

)

data between X=0.1 and X=0.8 (α=0.05, n=4). . . . . . . . . . . . . . . . . 67

3.6 Pine-Pellet Kinetics and R2 values based on the X conversion (X = Wo−WWo−Wf

)

data between X=0.1 and X=0.8 (α=0.05, n=4). . . . . . . . . . . . . . . . . 67

x

Page 12: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Nomenclature

DSC Differential Scanning Calorimetry

TGA Thermogravimetric Analysis

DTG Differential Thermal Analysis

micro-TGA Commercially available analytical-scale TGA device

PB-TGA Newly developed Packed-Bed TGA device

A(

1s

)

Pre-exponential factor (a.k.a., frequency factor) in the Arrhenius equation

E(

Jmol

)

Activation energy in the Arrhenius equation

R The universal gas constant (8.314472 JK ·mol

)

n Reaction order when in the Arrhenius equation, or number of replications

T Temperature in either Kelvin or ◦C

t Time in seconds, unless otherwise stated

W Current sample weight

W o Initial sample weight

W f Final sample weight

X Fractional conversion of the sample-based on weight

λ The Box-Cox variance stabilization factor

q=dTdt

Heating rate in ◦C per minute

R2 Regression goodness-of-fit

xi

Page 13: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Chapter 1

Introduction

With increasing demand for energy and decreasing relative fossil fuel availability, there is a

growing need to find and optimize alternative energy sources. Biomass derived energy sources,

such as from simple wood fires and charcoal, have existed for a very long time and provide one

alternative [2]. When properly utilized, they have the benefit of being renewable and carbon

neutral. Examples of biomass sources include agricultural wastes, energy crops (e.g., rapeseed

and algae), and sewage solids [12]. However, only recently has there been a recognized need

to find a sustainable method for utilizing biomass-based energy, with much effort focusing

on maintaining a sustainable system. Biomass can often be burned directly but can also be

converted into more efficient energy products that are cleaner and more efficient to transport.

Conversion technologies include biochemical processes such as fermentation and anaerobic

digestion (e.g., ethanol, and bio-gas), chemical refining and extraction, liquefaction [4], and

thermal conversions such as pyrolysis, gasification, burning, and various combinations thereof

[9, 12, 13].

In pyrolysis, a biomass material is heated to a high temperature (e.g., 600◦C) in a non-

oxidizing environment such as nitrogen. Instead of burning, other reactions may take place

including the thermally induced cracking of molecules. Some of the products are volatile and

exist as a vapor or gas, while others remain as a solid product. A portion of the vapor can

be collected and condensed into a material known as bio-oil. The solid product that is left

behind is char (e.g., charcoal). Char and bio-oil have the potential to be used as energy dense

products that are easier to transport than raw biomass [12, 13]. These can either be further

1

Page 14: Multi-Phase Packed-Bed Thermogravimetric Analyzer

2

processed such as by refining or gasification, or used in energy extraction processes, such as

combustion [30].

The actual reaction schema during pyrolysis can involve many different and complex re-

actions. Some compounds may volatilize and immediately be carried away by the carrier gas

in a single step. However, many compounds will undergo multiple and possibly interacting

reactions before volatilizing [6]. Even when in a carrier gas, or after condensation, additional

thermal cracking and/or other reactions may occur. A number of researchers have attempted

to simplify and model the reactions involved, however this area has not yet reached maturity

[6, 5, 33, 18, 17, 20, 21, 1, 3]. Basic pyrolysis parameters, such as heating rate, peak tem-

perature, vapor residence time, particle size, water content, and gas type, can be adjusted

to favor different products, such as char, bio-oil, or gas.

Pyrolysis processes can roughly be categorized as very slow (<0.6◦Cmin

), slow (0.6◦Cmin

to 120◦Cmin

), fast (100’s◦Cmin

), and flash pyrolysis (up to 6x107 ◦Cmin

) [8, 28]. One of the best

examples of very slow (e.g., 0.4◦Cmin

) pyrolysis is charcoal production in which charcoal yield

is maximized by slow heating rates [2, 9]. As heating rate is increased, bio-oil and gas are

favored products and char yield is lower [8, 9, 21].

Particle size can also affect the products. Small particles can experience low individual

thermal gradients and low intra-particle gas residence time (time spent within the particle)

[1, 21]. Large particles, such as wood chips, and wood pellets, can not heat as quickly and

become heating rate limited. Also, large particles may be subject to more secondary reactions

due to increased intra-particle residence time, some of which contribute to increased char

yields [21].

Carrier gas characteristics play an important role. Often nitrogen is used in pyrolysis since

it is non-oxidizing and readily available, but this is not always the case. In very slow pyrolysis

charcoal pit kilns, the feedstock effectively removes most of the oxygen via combustion and

the carrier gas is a complex mixture of compounds retained within the kiln. In this case,

the vapor residence time is very high since charcoal production is favored [2]. On the other

Page 15: Multi-Phase Packed-Bed Thermogravimetric Analyzer

3

extreme (i.e., flash pyrolysis) the vapor residence time may be very low since there is often

a very high relative carrier gas flow rate [28]. In other systems, especially when used with

other conversion technologies, other gases may be used, such as steam, hydrogen [30], and

carbon dioxide [22].

Biomass feedstocks are typically composed of cellulose (a polysaccharide), hemicellulose

(variable composition polysaccharides), lignin, and extractable material [31]. The pyrolytic

behaviors of these are different under similar conditions. For example, at 50◦Cmin

, Raveendran

et al. [31] reported cellulose, hemicellulose, and lignin decomposition peaks at about 401◦C,

300◦C, and 421◦C, respectively, when using isolated components. At a slower heating rate of

10◦Cmin

, Yang et al. [34] found distinctive peaks between 315 ◦C to 400◦C for cellulose, 220◦C

to 315◦C for hemicellulose, and 160◦C to 900◦C for lignin. In untreated biomass, such as pine

pellets, wood chips, and algae, analysis may be complicated since materials are combined so

that the decomposition peaks overlap [18, 14].

These pyrolytic characteristics are based on highly controlled and somewhat idealized

analytical scale pyrolysis reactors called thermogravimetric analyzers (TGA’s). Larger scale

research, pilot, and production pyrolysis systems can exist in a variety of complex reactor

configurations that influence pyrolysis. Examples include continuous fluidized-bed reactors,

continuous conveyor reactors, packed-bed reactors, and batch reactors with the sample simply

surrounded by the carrier gas [8, 26]. Each combination of parameters in pyrolysis favors a

different outcome, not only in yields, but also the chemical characteristics of the products and

the energy required to make them. The optimal set depends on what is available, economical,

and efficient. It is important to have techniques available to critically evaluate possibilities

so as to pick the best solution to a given energy problem.

To make matters more complicated, there are many factors other than composition that

influence pyrolysis. These may be feedstock related, such as particle size [20, 15], particle heat

transfer characteristics [6], and tissue structure. They may be operating condition related,

Page 16: Multi-Phase Packed-Bed Thermogravimetric Analyzer

4

such as reactor type (i.e., packed-bed, fluidized-bed, etc.) [9], operating temperature, heating

rate [3, 33], thermal-gradients [27], carrier gas [30], and carrier gas flow rates [2].

All of these feedstock and operational parameters directly relate not only to the produc-

tion and chemical composition of different products, but also to the operation and energy

balance of the whole process, an important concern when pyrolysis is employed for biomass

based renewable energy. Thus, it is important to have analytical techniques that can be used

to efficiently characterize pyrolysis systems. This information can then be used in feasibility

assessments, energy balances [11], model development, and other techniques so that efficient

and optimized biomass conversion systems can be developed [5].

One of the basic ways to study pyrolysis is by evaluating the feedstock’s internal de-

composition behavior. Analytical techniques may involve an analysis of pyrolysis products,

such as yield, detailed chemical composition analysis, energy content, and optimal fuel prop-

erties. Additionally, close observation of events during the pyrolysis reaction is required to

gain a good understanding of decomposition behavior and reaction kinetics. Two common

techniques involve differential scanning calorimetry (DSC) and thermogravimetric analysis

(TGA). In DSC, the energy required to maintain the sample at a temperature set-point

is recorded, thus providing information about exothermic or endothermic reactions [25]. In

TGA, a sample is heated at a constant heating rate in a non-oxidizing atmosphere while

being continuously weighed to provide a curve of weight versus temperature. Any loss of

material from the sample, such as by volatilization, is recorded as a weight change [32]. For

example, at 100◦C a sample of water would require significant energy to maintain the heating

rate since there is a phase change from liquid water to gaseous water. Taking the differential

of the TGA profile (DTG) gives the weight change per unit time or unit temperature, which

can be used to identify peak locations and weight loss reaction onset and ending locations.

The difference between the initial and final sample weight gives the final char yield as well

as the yield of combined condensable and non-condensable volatile material.

Page 17: Multi-Phase Packed-Bed Thermogravimetric Analyzer

5

In small analytical DSC and TGA systems, a sample may range from 3 mg to 30 mg

while in larger systems this may be as high as 0.5 g to 4 g. The samples are typically placed

in crucibles with or without a lid. Individually and combined, both DSC and TGA can

provide phase change information (i.e., solid to liquid or gas), energetics information, and

reaction kinetics. These can be associated with time-dependent and temperature-dependent

chemical analysis with other down-stream analytical instruments, such as mass spectrome-

ters, gas chromatographs, and Fourier Transform Infrared spectrometers (FT-IR) to measure

characteristics of the evolved gases [7, 10, 14, 16, 29].

In TGA analysis, there are several methods for extracting kinetic information, two of

which are isothermal and dynamic analysis. In isothermal analysis the sample is heated

and held at the analysis temperature until all decomposition is complete and there is no

significant mass loss. The sample is then heated to the next temperature and the process

is repeated until the desired temperature range is completed. These curves are processed

to give a series of kinetic parameters at each temperature. In dynamic analysis, the sample

is heated at a constant rate to a target temperature, and the ensuing TG curve is used

to determine the kinetics [10]. In either method, basic information includes product yields,

major reaction peaks and magnitudes, and peak onset and ending times.

Analytical-scale TGA analysis is typically performed to approximate an idealized reactor

and to look at intrinsic pyrolysis reaction characteristics. The small sample sizes help approx-

imate this by: (a) reducing thermal lag, an effect that can result in erroneous kinetic analysis

[27]; and (b) reducing vapor residence time. Even so, pyrolysis of biomass feedstocks involves

many different compounds and reactions. Hence, it is important to realize that TGA derived

biomass pyrolysis kinetics represent the total sum of pyrolysis reactions that are occurring

and that they are derived empirically. In some cases it may be better to refer to these as

simply model parameters.

Kinetic parameters typically refer to at least the pre-exponential factor (a.k.a., frequency

factor) A (1s), and the exponent term (a.k.a., activation energy) E ( J

mol) in the Arrhenius

Page 18: Multi-Phase Packed-Bed Thermogravimetric Analyzer

6

(Equation 1.1) or a similar equation. There are a number of methods for estimating these

values that take into account a variety of assumptions. For example, many basic methods

assume that samples are not heat and mass transfer limited and that the data represent

individual consecutive reactions. Guar and Reed [19] review a number of methods for deter-

mining these parameters. One way is to perform the transformation in Equation 1.3 and to

estimate the coefficients using regression analysis.

dX

dt= Ae

−E

RT f(X) (1.1)

X =Wo − W

Wo − Wf

(1.2)

ln

(

dX

dt

)

= ln (A) −E

RT+ ln (f(x)) (1.3)

f(X) = (1 − X)n (1.4)

The other terms in Equation 1.1 are the universal gas constant R (8.314472 JK ·mol

), ab-

solute temperature T (Kelvin), and time t (seconds). X is the fractional reaction conversion

based on weight (Equation 1.2) and is expressed in terms of the initial sample weight (Wo),

the current measured weight (W ), and the final sample weight (Wf ). The function f(x) is a

transformation based on the X conversion and can be represented by a number of different

models based on reaction mechanisms. A simple example is the nth order model shown in

Equation 1.4 where n is the order number of the reaction [21]. Guo and Lua [21] list a number

of other f(x) functions that have been used for decomposition research (Table 1.1).

When a feedstock is analyzed in dynamic mode with several discrete heating rates, Equa-

tions 1.1 and 1.2 can be converted into Equation 1.3. At a specific temperature during the

reaction, the weight and temperature data for each discrete heating rate are transformed

into three data sets; ln(dXdt

), 1T, and ln(f(X)). The linear regression yields of these data

Page 19: Multi-Phase Packed-Bed Thermogravimetric Analyzer

7

Table 1.1: Functions f(x) as described by Guo and Lua [21].No. Model f(x)

Sigmoid Rate Equations

1 Avrami-Erofe’ev 2 · (1 − X) · [−ln (1 − X)]1/2

2 Avrami-Erofe’ev 3 · (1 − X) · [−ln (1 − X)]2/3

3 Avrami-Erofe’ev 4 · (1 − X) · [−ln (1 − X)]3/4

Geometric Models4 Contracting area 2 · (1 − X)1/2

5 Contracting volume 3 · (1 − X)2/3

Diffusion Mechanisms

6 One-dimensional (2 ·X)−1

7 Two-dimensional [−ln (1 − X)]−1

8 Three-dimensional (3/2) · (1 − X)2/3 ·[

1 − (1 − X)1/3]

−1

9 Ginstling-Brounshtein (3/2) ·[

(1 − X)−1/3 − 1]

−1

Order of Reaction10 nth Order (1 − X)n

yield an equation that can be solved for A, E, and other parameters in f(x), at that par-

ticular temperature [10]. The process is repeated at different temperatures throughout the

reaction resulting in a set of kinetic parameters at each temperature. Additional regression

diagnostics can be used to assess the validity of the model.

It is also possible to analyze the data using a single heating rate (Equation 1.5). This

can be substituted into Equation 1.1 via Equations 1.5-1.7 to give Equation 1.8. Linear

regression is performed on the ln(dXdT

), and 1T

data. A, E, and n can be obtained from

regression coefficients [10].

q =dT

dt(1.5)

dXdTq

= Ae−E

RT f(X) (1.6)

Page 20: Multi-Phase Packed-Bed Thermogravimetric Analyzer

8

dX

dT=

A

qe

−E

RT f(X) (1.7)

ln

(

dX

dT

)

= ln

(

A

q

)

−E

RT+ ln (f(x)) (1.8)

There is much discussion about which models, methods, f(x) transformations, and re-

action selection protocols are the best under particular conditions. There may be countless

numbers of reactions taking place, but the combined effect of all reactions is observed. This

overall reaction kinetic information can be used to evaluate different biomasses, create mod-

els, and compare operating conditions to help find an optimized solution for converting a

given biomass into energy products [21, 23, 24, 32, 33].

Although many existing TGA systems can provide important thermogravimetric and re-

action kinetics information, there are several important limitations. Most analytical TGA

techniques apply for just the solid-phase decomposition, utilize small samples sizes, use small

crucible type reactors, and use a carrier gas to maintain a constant atmosphere around the

crucible (rather than passing the carrier gas through the sample at a constant rate). While

this minimizes thermal-gradients in the sample and can provide detailed information about

an idealized system (important in understanding fundamental reaction mechanisms), it is

much different from production or pilot-sized larger-scale pyrolysis systems, where sample

thermal gradients may be much greater, mass transfer may have different limitations, and

particles may be bigger. Additionally, non-homogeneity of natural materials is a major con-

cern in small samples. Finally, a detailed knowledge of bio-oil formation characteristics is

important since it is being collected as an energy product, but typically there is too little to

collect via condensation in a small analytical TGA system.

These limitations become striking when compared to larger scale pyrolysis systems for

production of charcoal and bio-oil. In these cases, carrier gas may pass directly through a

bed of feedstock; reactors may be larger; feedstock may experience a non-uniform heating

profile; bio-oil condensation is important; and different particle morphologies (i.e., powder

Page 21: Multi-Phase Packed-Bed Thermogravimetric Analyzer

9

vs. pellets) may be used. The differences between what is available in analytical-scale ther-

mogravimetric analysis devices, the conditions found in larger scale pyrolysis reactors, and

the need to understand bio-oil formation during pyrolysis suggest the need for a new device.

It should provide thermogravimetric information about both the solid and the condensable

materials (excluding low temperature condensables, such as nitrogen and carbon dioxide)

and produce a sufficient quantity of char and bio-oil for further analysis, while mimicking a

reactor configuration and allowing for different particle types.

The overall goal of this research is to overcome these limitations by completing two ma-

jor objectives. The first, described in detail in chapter 2, is to design, build and test a TGA

device in which thermogravimetric information is obtained for both the solid phase decom-

position and condensate (bio-oil) formation during pyrolysis. The feedstock being analyzed

is in a packed-bed configuration. This new device is henceforth referred to as the Packed-Bed

Thermogravimetric Analyzer (PB-TGA). The second objective, described in chapter 3, is to

compare the newly developed TGA device, the PB-TGA, to a small, commercially available

analytical scale TGA device (henceforth referred to as the “micro-TGA”).

1.1 References

[1] Antal, M. J. J. (1995). Cellulose pyrolysis kinetics: The current state of knowledge.

Industrial and Engineering Chemical Research 34, 703–717.

[2] Antal, M. J. J. and G. Morten (2003). The art, science, and technology of charcoal

production. Industrial and Engineering Chemical Research 42, 1619–1640.

[3] Antal, M. J. J., G. Varhegyi, and E. Jakab (1998). Cellulose pyrolysis kinetics: Revisited.

Industrial and Engineering Chemical Research 37, 1267–1275.

[4] Aresta, M., A. Dibenedetto, and G. Barberio (2005). Utilization of macro-algae for

enhanced CO2 fixation and biofuel production: Development of a computing software for

Page 22: Multi-Phase Packed-Bed Thermogravimetric Analyzer

10

an lca study. Fuel Processing Technology 86, 1679–1693.

[5] Babu, B. and A. Chaurasia (2003). Modeling, simulation and estimation of optimum

parameters in pyrolysis of biomass. Energy Conversion and Management 44, 2135–2158.

[6] Babu, B. and A. Chaurasia (2004). Heat transfer and kinetics in the pyrolysis of shrinking

biomass particle. Chemical Engineering Science 59, 1999–2012.

[7] Baker, R. R., S. Coburn, C. Liu, and J. Tetteh (2005). Pyrolysis of saccharide tobacco

ingredients: a TGA-FTIR investigation. Journal of Analytical and Applied Pyrolysis 74,

171–180.

[8] Bridgwater, A. and G. Grassi (1991). Biomass pyrolysis liquids upgrading and utilization,

Chapter A review of biomass pyrolysis and pyrolysis technologies, pp. 11–92. London:

Elsevier Science.

[9] Bridgwater, A. and G. Peacocke (2000). Fast pyrolysis processes for biomass. Renewable

and Sustainable Energy Reviews 4, 1–73.

[10] Capart, R., L. Khezami, and A. K. Burnham (2004). Assessment of various kinetic

models for the pyrolysis of a microgranular cellulose. Thermochimica Acta 417, 79–89.

[11] Daugaard, D. E. (2003). Enthalpy for pyrolysis for several types of biomass. Energy

and Fuels 17, 934–939.

[12] Demirbas, A. (2001a). Biomass resource facilities and biomass conversion processing for

fuels and chemicals. Energy Conversion and Management 42, 1357–1378.

[13] Demirbas, M. and M. Balat (2006). Recent advances on the production and utilization

trends of bio-fuels. Energy Conversion and Management (47), 2371–2381.

[14] Díaz, C. J. G. (2006). Understanding Biomass Pyrolysis Kinetics: Improved Model-

ing Based on Comprehensive Thermokinetic Analysis. Ph.d. dissertation, Departament

d’Enginyeria Química.

Page 23: Multi-Phase Packed-Bed Thermogravimetric Analyzer

11

[15] Encinar, J., F. Beltrán, A. Ramiro, and J. González (1996). Pyrolysis of two agricultural

residues: olive and grape bagasse. influence of particle size and temperature. Biomass and

Bioenergy 11 (5), 397–409.

[16] Galvagno, S., S. Casu, M. Martino, E. Di Palma, and S. Portofino (2007). Thermal

and kinetic study of tyre waste pyrolysis via TG-FTIR-MS analysis. Journal of Thermal

Analysis and Calorimetry 88, 507–514.

[17] Garcia-Nuñez, J. A. (2005). Determination of kinetic constants and thermal model

during the pyrolysis of palm oil mill solid wastes. Ph. D. thesis, University of Georgia.

[18] Garcia-Nuñez, J. A., M. Garcia-Perez, and K. Das (2008). Determination of kinetic

parameters of thermal degradation of palm oil mill by-products using thermogravimetric

analysis and differential scanning calorimetry. Transactions of the ASABE 51 (2).

[19] Guar, S. and T. B. Reed (1998). Thermal Data for Natural and Synthetic Fuels. New

York: Marcel Dekker, Inc.

[20] Guo, J. and A. Lua (2000). Kinetic study on pyrolysis of extracted oil palm fiber.

Journal of Thermal Analysis and Calorimetry 59, 763–774.

[21] Guo, J. and A. Lua (2001). Kinetic study on pyrolytic process of oil-palm solid waste

using two-step consecutive reaction model. Biomass and Bioenergy 20 (3), 223–233.

[22] Kim, Y. J., M. I. Kim, C. H. Yun, J. Y. Chang, C. R. Park, and M. Inagaki (2004).

Camparative study of carbon dioxide and nitrogen atmospheric effects on the chemcial

structure changes during pyrolysis of phenol-formaldehyde spheres. Journal of Colloid

and Interface Science 274, 55–562.

[23] Leung, D. and C. Wang (1998). Kinetic study of scrap tyre pyrolysis and combustion.

Journal of Analytical and Applied Pyrolysis 45, 153–169.

Page 24: Multi-Phase Packed-Bed Thermogravimetric Analyzer

12

[24] Mangut, V., E. Sabio, J. Gañán, J. González, A. Ramiro, C. González, S. Román, and

A. Al-Kassir (2006). Thermalgravimetric study of the pyrolysis of biomass residues from

tomato processing industry. Fuel Processing Technology 87, 109–115.

[25] McNaughton, J. and C. Mortimer (1975). Differential Scanning Calorimetry, Volume 10

of PRS: Physical Chemistry Series 2. London: Butterworths, London.

[26] Meier, D. and O. Faix (1999). State of the art of applied fast pyrolysis of lignocellulosic

materials. Bioresource Technology 68, 71–77.

[27] Narayan, R. and M. J. J. Antal (1996). Thermal lag, fusion, and the compensation effect

during biomass pyrolysis. Industrial and Engineering Chemical Research 35, 1711–1721.

[28] Onay, O. and O. M. Kockar (2003). Technical note: Slow, fast and flash pyrolysis of

rapeseed. Renewable Energy 28, 2417–2433.

[29] Pappa, A., S. Kyriakou, K. Mikedi, N. Tzamtzis, and M. Statheropoulos (2004). Design

considerations and an example of applications of an in-house made TG-MS interface.

Journal of Thermal Analysis and Calorimetry 78, 415–426.

[30] Pindoria, R. V., A. Megaritis, R. C. Messenböck, D. R. Dugwell, and R. Kandlyoti

(1998). Short communications: Comparison of the pyrolysis and gasification of biomass:

effect of reacting gas atmosphere and pressure and eucalyptus wood. Fuel 77 (11), 1247–

1251.

[31] Raveendran, K., A. Ganesh, and K. C. Khilar (1996). Pyrolysis characteristics of

biomass and biomass componenets. Fuel 75 (8), 987–998.

[32] Stenseng, M., A. Jensen, and K. Dam-Johansen (2001). Investigation of biomass py-

rolysis by thermogravimetric analysis and differential scanning calorimetry. Journal of

Analytical and Applied Pyrolysis 58-59, 765–780.

Page 25: Multi-Phase Packed-Bed Thermogravimetric Analyzer

13

[33] Várhegyi, G., M. J. J. Antal, E. Jakab, and P. Szabó (1997). Kinetic modeling of

biomass pyrolysis. Journal of Analytical and Applied Pyrolysis 42, 73–87.

[34] Yang, H., R. Yan, H. Chen, D. H. Lee, and C. Zheng (2007). Characteristics of hemi-

cellulose, cellulose and lignin pyrolysis. Fuel 86 (12-13).

Page 26: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Chapter 2

Development of a bench sized packed-bed

thermogravimetric analyzer for both solid

decomposition and condensate formation1

1Jarrett, Erik and Das, K.C. (2008). To be submitted to the Journal of Analytical and AppliedPyrolysis.

14

Page 27: Multi-Phase Packed-Bed Thermogravimetric Analyzer

15

Abstract

Thermogravimetric analysis (TGA) is a method for evaluating the pyrolysis decompo-

sition of biomass by measuring the weight of the sample as a function of temperature.

Characteristics that can be measured include solid product yields and reaction kinetics, such

as activation energy and frequency factor. Currently, most TGA devices measure the weight

change of only the solid-phase in crucibles of milligram capacity. This does not provide infor-

mation about the formation of condensable vapor, a critical element in understanding and

optimizing bio-oil formation. In addition, a carrier gas often surrounds the sample but is not

forced through it at a constant rate. To overcome these limitations, a new TGA device was

developed in which nitrogen is metered through a maximum 300 g and/or 571 mL packed-bed

of biomass. Weight changes are determined as the biomass is heated from ∼25◦C to 600◦C

at 3◦Cmin

, thus providing a solids decomposition TGA curve of weight vs. temperature. The

exhaust is mechanically isolated from a condensing unit that is also weighed throughout the

experiment, thus providing a condensate TGA curve of weight vs. temperature. The system

was developed and is described here in four steps: (a) mechanical and electrical design, (b)

weight calibration, (c) blank run effects, and (d) verification using cellulose as a feedstock.

Keywords: TGA; development; thermal analysis; pyrolysis; bio-oil; charcoal

Page 28: Multi-Phase Packed-Bed Thermogravimetric Analyzer

16

2.1 Introduction

With the increasing demand for energy and the decreasing fossil fuel reserves, there is a

growing need to find and optimize energy sources. Biomass derived energy sources, such as

simple wood fires and charcoal, have existed for a very long time and provide one alternative

[2]. When properly utilized they have the benefit of being renewable and carbon neutral.

Technologies, such as anaerobic digestion, chemical refining and extraction, liquefaction [4],

pyrolysis, and gasification [7, 9, 10] can be used to convert the biomass into efficient energy

products.

In pyrolysis, a biomass material is heated to a high temperature (e.g., 600◦C) in a non-

oxidizing environment, such as nitrogen. Instead of burning, thermal cracking and other

secondary reactions take place that cause some of the material to volatilize into the sur-

rounding gas. This vapor is removed and condensed to yield a complex product often called

“bio-oil.” The solid product that is left behind is char (charcoal). Char and bio-oil have the

potential to be used as energy dense products that are cleaner burning and easier to transport

than raw biomass [9, 10].

Parameters that influence pyrolysis conversion include heating rate, peak temperature,

vapor residence time, feedstock type and size, water content, and gas type. These can be

adjusted to favor different products, such as char, bio-oil, or gas. Processes can roughly be

categorized as slow (0.6◦Cmin

to 120◦Cmin

), fast (100’s◦Cmin

), and flash pyrolysis (up to 6x107 ◦Cmin

)

[6, 20] with higher heating rates generally favoring bio-oil and gas formation over char for-

mation [6, 7, 15]. Particle size can also affect the products. Small particles experience a low

thermal gradient and low intraparticle gas residence time (time spent within the particle)

[1], while large particles, such as wood chips, wood pellets, and poultry litter, can not heat

as quickly and may be subject to more secondary reactions due to increased residence time.

Finally, there are many different types of reactors, such as continuous fluidized-bed reac-

tors, continuous conveyor reactors, packed-bed reactors, and batch reactors with the sample

simply surrounded by the carrier gas [6, 18].

Page 29: Multi-Phase Packed-Bed Thermogravimetric Analyzer

17

Each combination of parameters in pyrolysis favors a different outcome, not only in yields,

but also the chemical characteristics of the products and the energy required to make them.

The optimal set depends on what is available, economical, and efficient. It is important to

have techniques available to critically evaluate possibilities so as to pick the best solution to

a given energy problem.

One of the common ways to study pyrolysis is by thermogravimetric analysis (TGA). In

TGA a small sample, typically 3 mg to 30 mg (0.5 g to 4 g in a few larger scale TGA devices),

of biomass is placed in a crucible with or without a lid. This is heated at a constant heating

rate in a non-oxidizing atmosphere while being continuously weighed to provide a curve of

weight versus temperature. Any loss of material from the sample, such as by volatilization,

is recorded as a weight change [22]. The TGA can be operated in-line with other analytical

techniques such as mass spectrometry, gas chromatography, and Fourier Transform Infrared

spectroscopy (FT-IR) to measure characteristics of the evolved gases [5, 8, 11, 12, 21].

Traditional TGA analysis is a quick and relatively easy way to evaluate the effects of

pyrolysis operating parameters in a highly idealized environment. For example, a simple

subtraction of the final sample weight from the initial weight gives the combined production

of gases and vapors, while the final weight provides a measure of the char produced. Taking

the differential of the TGA profile (DTG) gives the weight change per unit time or unit

temperature and can be used to identify reaction peaks, and weight loss reaction onset and

endings. Depending on which kinetics model is used, either the time or temperature based

DTG can be used to calculate kinetic parameters for each thermal decomposition step, such

as activation energy, frequency factor, and reaction order [14, 15]. All of this information can

be used to evaluate different biomasses, create models, and compare operating conditions

to help find an optimized solution for converting a given biomass into energy products

[15, 16, 17, 22, 23].

Although these existing TGA systems provide important thermogravimetric and reaction

kinetics information, there are several important limitations. First, the small sample crucibles

Page 30: Multi-Phase Packed-Bed Thermogravimetric Analyzer

18

may or may not be open to the carrier gas, thus limiting control of the carrier gas contact

with the sample. Second, only the solid phase decomposition characteristics are measured.

This means that any gas and condensable liquids, such as bio-oil, are lost. Although in-

line instruments can be used to chemically analyze the exhaust, there is typically too little

condensable material to collect via condensation. This may be a severe limitation if the

major concern is to analyze the bio-oil formation characteristics. Third, the small crucibles

mean that the feedstock particles must be small enough both to fit in the unit and provide a

homogeneous sample. This limits the ability to measure the effects of varying particle sizes.

These limitations become striking when compared to larger scale pyrolysis systems used

for charcoal and bio-oil production. In these cases the carrier gas may pass directly through a

bed of feedstock, the sample may be much larger, the feedstock may experience a non-uniform

heating profile, bio-oil condensation is important, and different particle morphologies (i.e.,

powder vs. pellets) may be used.

These differences between what is available in thermogravimetric analysis devices and

conditions found in larger scale pyrolysis reactors suggest a need for a new device. Such

a device should provide thermogravimetric information about both the solid and the con-

densable material (excluding low temperature condensables, such as nitrogen and carbon

dioxide), and can produce a sufficient quantity of char and bio-oil for further analysis, while

mimicking a reactor configuration and allowing for different particle types.

The objective of this project was to design, build and test such a TGA device in which

real-time thermogravimetric information could be obtained for both solids decomposition

and condensate formation (bio-oil) for a packed-bed biomass feedstock.

2.2 Development

Figure 2.1 shows a flow chart representation of the TGA device. The solid phase decom-

position system consists of a carrier gas preheater, a heated packed-bed reactor, and an

Page 31: Multi-Phase Packed-Bed Thermogravimetric Analyzer

19

Figure 2.1: PB-TGA flow diagram. Preheated carrier gas is fed into a packed-bed reactorheated by a muffle furnace. The exhaust is cooled by nitrogen in the mixer so it can passthrough a flexible silicone tube and into the condenser. The reactor balance weighs thepreheater, reactor, and mixer. The condenser balance weighs the condensate. Thermocouples(not pictured) are located in the reactor balance, preheater inlet and outlet, reactor inletand center, furnace, and mixer exit. Measurements are either recorder by a computer, or byhand.

exhaust cooler. These hang from a balance that, when calibrated and compensated, gives

the weight of the sample in the furnace. Exhaust gas flows through a flexible line anchored at

its midpoint and into a temperature controlled condenser system. This is also continuously

weighed to provide data on the amount of condensate formed. Thermocouples are placed at

key points. Temperature control is accomplished with PID-Controllers (Proportional, Inte-

gral, Differential), and temperatures and weights are logged with a computer-based system.

When a constant heating rate is used, the resulting weight and temperature relationships

constitute the TGA curves for the solid (biomass to char) and liquid (bio-oil) products.

2.2.1 Solid Phase Decomposition System

Figure 2.2 shows a detailed cross-section of the solid phase decomposition system. The bal-

ance mechanism and muffle furnace are anchored to an aluminum frame resting on four

Page 32: Multi-Phase Packed-Bed Thermogravimetric Analyzer

20

Load CellConditioning

Circuit

Thermal-wellEntrance

Load Cell Pivot

Pre-Heater

Baffle Plate

Counter BalanceLocation

Connecting Rods

Tem

per

ature

Con

troll

edE

ncl

osu

re (

Coole

r)

Muffle FurnaceAnchored to Frame

Sanitary FlangeFitting with Graphite Gasket

Reactor (n=10.44cm,H=6.67cm)Usable Volume = 571 ml

Carrier GasInlet

Baffle Plate

Exhaust and QuenchMixing LocationQuench Gas Inlet

Exhaust Gas Connection

to Condenser(Figure 3)

ExhaustThermocouple

Furnace Thermocouple

Located in FurnaceCavity

Reactor EntranceThermocouple

Pre-heaterThermocouple #1

Pre-heaterThermocouple #2

SampleThermocouple

Aluminum PlateAnchoredto Frame

Figure 2.2: Section view of the solid-phase decomposition system. The preheater, reactor,exhaust, and quenching system hang from the balance mechanisms and are balanced using acounter-weight. The frame (not pictured) holds the balance mechanism and muffle furnace.

vibration damping feet. A counter-weight is used such that the rest of the system balances

from a pivot point in the balance mechanism. All connecting lines, such as thermocouple

wires (flexible, small gauge, and coiled), and tubing (flexible Tygon or silicone) are anchored

to the frame. All heated metal parts were constructed of stainless steel and all thermocouples

are K-type (Omega Engineering, Stamford, CT, USA), unless otherwise noted.

Page 33: Multi-Phase Packed-Bed Thermogravimetric Analyzer

21

Nitrogen gas is metered through two mass flow controllers (Model UFC 8100 and 1660

1 Lmin

, UNIT Instruments, San Jose, CA, USA ). The first line passes into a preheater

that maintains its exit gas temperature at the furnace temperature set-point using a PID-

controller (Model CN96221TR-C2, Omega Engineering, Stamford, CT, USA ) and preheater

thermocouple #1. This heated gas passes into the reactor gas entrance tube that houses

the reactor temperature thermal-wells. This tube has an aluminum plate on the top that

is an attachment point for the counter-weight and is connected to the balance pivot via

four connecting rods. On the bottom it has a 4 in sanitary-type flange fitting (#50485k225,

McMaster-Carr Supply Co., Atlanta, GA, USA) that mates with the top of the reactor. The

second nitrogen line is connected to the quench gas inlet line to cool the exit gas enough for

it to safely pass into the silicone decoupling line between the reactor and condensate system.

The exhaust line also has a sanitary-type flange fitting that mates with the bottom of the

reactor.

The cylindrical reactor has a usable volume of 571 mL and has sanitary flange fittings

welded to the top and the bottom. These mate with the entrance and exit gas lines using

Grafoil gaskets (OD 11.89 cm, ID 9.73 cm, 0.24 cm thick, Mercer Gasket and Shim, Bellmawr,

NJ, USA) and two three-section clamps (Tri-Clover #13MHHS 4in-S, Alfa Laval, Lund,

Sweden). The biomass sample inside the reactor is retained by five stainless steel mesh

screens (mesh sizes from top to bottom: No.80x80, No.325x325, No.400x400, No.400x400,

and No.80x80). There are two thermal-wells containing thermocouples, one in the initial

center of the biomass, and the other at the entrance of the reactor. The reactor hangs inside

a PID controlled (Model CN96221TR-C2, Omega Engineering, Stamford, CT, USA ) muffle

furnace (modified model 550-14, Fisher Scientific, Waltham, MA, USA ) using the furnace

cavity temperature as feedback and the same temperature profile as the preheater.

The balance mechanism consists of the enclosure, load cell mount, reactor to pivot cou-

pling, and the conditioning circuit. The hinged, gasketed, and electrically shielded 5 cm thick

polystyrene enclosure covers the top and sides of the balance mechanism (not pictured in

Page 34: Multi-Phase Packed-Bed Thermogravimetric Analyzer

22

Figure 2.2). Between the balance and the aluminum plate is a layer of acrylic and ceramic

thermal insulation. Four small holes are present for the connecting rods. In addition to pro-

viding electrical shielding, the enclosure maintains the load cells and conditioning circuitry

at a temperature set-point using a peltier cooling device and PID-controller (Model CN75,

Omega Engineering, Stamford, CT, USA, ).

The reactor and associated components hang from a single pivot point that rests on a

lubricated, spring steel disc that in turn rests on a square aluminum plate connected to all

four load cells via vibration damping bushings. Thus, the weight of the reactor assembly is

translated to a point load evenly distributed between the load cells.

2.2.2 Condensate Handling System

The cooled exhaust carrier gas from the solid phase decomposition system passes through a

flexible silicone tube (63.3 cm long) anchored at its midpoint for mechanical isolation, and

into the condensing unit (Figure 2.3). The condensing unit sits on a 0.1 g resolution electronic

balance (Model RD6RM, Ohaus, Pine Brook, NJ, USA) whose output is continuously logged.

The first collection flask catches any material that has condensed both upstream and in the

three reflux condensers. The exhaust passes up through the reflux condensers, down through

a Graham condenser, and through a final flask that catches the remaining condensate. The

exit gas passes through an anchored flexible tube and is vented outside. The condensers

are maintained at a constant 20◦C using a circulating ethylene-glycol-water bath and two

anchored connecting lines. When tared and corrected to remove the effects of the condenser

system itself, this system allows the weight of the condensate to be measured continuously

throughout the experiment.

2.2.3 Electronics

The furnace, carrier gas preheater, and balance temperature are controlled by PID-controllers

using solid-state relays. These are housed in an mechanically ventilated enclosure.

Page 35: Multi-Phase Packed-Bed Thermogravimetric Analyzer

23

Figure 2.3: Picture of the condensate system. Exhaust from the solid-phase system passesthrough the flexible silicone line which is anchored at its midpoint for mechanical isolation.The exhaust enters the first collection flask, goes up through three reflux condensers, downthrough a Graham condenser into the last collection flask, and out through a flexible exhaustline. A circulating chiller maintains the condenser temperature and a balance continuouslyweighs the condenser system.

PID-controller temperatures are logged manually every ∼30 minutes throughout the ex-

periment, while all other temperatures and weights are logged using a computerized data

acquisition system (NI Lab-View v.8.2 with data acquisition card model NI PCI-MIO-16E-4

Dev1, National Instruments, Austin, TX., USA). An electrically and thermally insulated

terminal box provides a common point for making the connections and contains a precision

cold junction compensation temperature sensor. All thermocouples are calibrated using a

calibrator (Model CL3512A, Omega Engineering, Stamford, CT, USA). The data acquisi-

tion system reads 1200 data points at 600 Hz and provides a moving average of 19200 data

points (equivalent to 32 seconds) logged into a text file.

Page 36: Multi-Phase Packed-Bed Thermogravimetric Analyzer

24

The load cell signals are translated into a voltage output using a custom built circuit

(Figure A.1). A 120 V AC to ±15 V DC linear power supply (Model SLD-12-1010-12T,

Solar, Rosemont, IL) housed in a mechanically ventilated enclosure provides the power for

two temperature compensated ±10 V DC voltage regulators (generic LM317 and LM337

types) located in the insulated balance assembly. This, in turn, powers OP213 op-amps used

for summation, amplification, and filtering of the load cell output, as well as providing power

for the load cell excitation voltage regulator (another LM317 and LM337 based ±5 V power

supply).

The positive and negative outputs from the load cells pass into two OP213 unity gain

inverting amplifiers, whose outputs pass into a first-order, low-pass filter with a gain of 10

and a low-pass corner frequency of 15.9 Hz. This signal then passes through a final first-order,

low-pass unity-gain filter with a corner frequency of 0.53 Hz. The final voltage is proportional

to the weight of the reactor and is logged using the data acquisition system. A calibration

model translates the voltage output to grams.

2.3 Materials and Methods

Once built, the packed-bed TGA solid-phase decomposition system was calibrated using a

blank run correction method developed to correct for blank run effects. Biomass runs were

performed using a cellulose substrate.

2.3.1 Calibration

The voltage signal from the load cell conditioning circuit needs to be converted to grams as

needed for TGA measurements. Ideally this relationship would be linear, but in reality the

circuit is also sensitive to temperature changes in the conditioning circuitry (i.e., resistance

heating, environmental changes, etc.). Thus, weight (W) is a function of both voltage output

(V) and circuit temperature (Tcircuit). It was necessary to perform an experiment to develop

Page 37: Multi-Phase Packed-Bed Thermogravimetric Analyzer

25

Table 2.1: Weight calibration experimental design. Numbers 1-4 indicate the randomizedapplication order within each row (circuit temperature) and replication. The applicationorder of the circuit temperature was 26.4◦C, 26◦C, and 25.6◦C.

Replication 1 Replication 2 Replication 3Temp Order 0 169.3 332.1 503.1 0 169.3 332.1 503.1 0 169.3 332.1 503.1◦C n g g g g g g g g g g g g

25.6 3rd 3 1 4 2 4 2 1 3 4 3 2 126 2nd 1 2 4 3 1 4 3 2 4 3 2 1

26.4 1st 4 2 3 1 2 3 1 4 4 2 1 3

this model so that the weight of the reactor could be measured and the accuracy and precision

of the measuring system could be quantified.

This experiment was performed by hanging four weights (0 g, 169.3 g, 332.2 g, and 503.1 g)

sequentially on the empty reactor while maintaining the load cell conditioning circuitry at

three different fixed temperature (25.6◦C, 26◦C, and 26.4◦C). The weights are discrete values,

but the actual temperatures drift around the set-point due to normal variations. These ranges

were picked since they represent the maximum values expected. The four weight and three

temperature levels yielded 12 different condition combinations which were replicated three

times. The weight applications were randomized within the temperature applications, which

were not randomized (Table 2.1). For example, the load cell circuitry temperature was first

set to 26.4◦C. Then, during replication one, the 503.1 g weight was first hung on the reactor,

followed by the 169.3 g weight, and so on. Voltage and temperature data were collected for

five minutes for each condition with two minutes between each weight application and one

hour between temperature changes, to allow for equilibrium to be reached.

After the experiment, the voltage, temperature, and applied weight information for each

condition was combined. The software packages Gnumeric (V.1.8.2, GNU open source), the R

statistical environment (V.2.6.2, The R Foundation for Statistical Computing, Open Source),

and the Maxima algebra system (V.5.13.0, Open Source) were used for analyzing and de-

Page 38: Multi-Phase Packed-Bed Thermogravimetric Analyzer

26

veloping the calibration equation. The model was verified at least once before each of five

subsequent blank runs in which the whole system was left as-is (i.e., not cleaned, taken apart,

or modified) between runs. Before each blank run the same weights used for the calibration

(0 g, 169.3 g, 332.2 g, and 503.1 g) were hung, in that order, from the empty reactor. Un-

like the calibration experiment, data were only collected for two minutes per condition and

allowed to stabilize until no visible movement was detected. The error between the actual

applied weight and the weight predicted by the calibration equation was calculated as well

as compared to the first predicted weight.

2.3.2 Blank Runs

During an experiment, three error inducing factors exist: (a) heating rate and temperatures,

(b) carrier gas flow, and (c) formation of air bubbles in the condensing unit. Additionally,

changes in temperature (a) change gas density, (b) induce updrafts around the reactor, (c)

expand metal parts, and (d) change the pliability of tubes and electrical insulation. All of

these potential effects may cause error in the weights being measured.

In order to correct for as many of these errors as possible, a series of experiments was

performed with the reactor empty and set up exactly as with biomass samples. The data

then reflected the system’s effects on the measurements without the biomass sample effects.

This set of data was subtracted from actual biomass runs to give a set of data representing

only the characteristics of the biomass.

Five blank-run experiments were conducted with nitrogen as the carrier and quench gas

at a flow rate of 0.25 Lmin

each. The room was maintained as close to 23◦C as possible, with

the balance circuit set to 26◦C, and the chilled liquid circulator for the condensers set to

20◦C. Each experiment began with the furnace, reactor, and preheater stabilized at room

temperature and purged with nitrogen. The furnace and preheater temperature were heated

at a rate of 3◦Cmin

until attaining 600◦C, where the temperature was maintained for 45 minutes

Page 39: Multi-Phase Packed-Bed Thermogravimetric Analyzer

27

before cooling down. Before each experiment, the load cell output and condenser tare weights

were recorded for 10 minutes, averaged, and used as the tare weights.

During the experiments, the load cell voltage output, and temperatures (balance mech-

anism, cold-junction-compensation, preheater, reactor entrance, sample center, and exhaust

outlet) were logged automatically and continuously by the computer. The furnace and pre-

heater PID-controller temperatures were logged manually. The reactor and condenser weight

readings were averaged again over 10 minutes just before the end of the experiment. After

each run, the voltage information was transformed using the calibration model.

2.3.3 Biomass Runs

Testing the device with an actual feedstock not only provides example TGA and DTG in-

formation for both the solid and liquid phases in a packed-bed reactor but also allows the

device, calibration, and blank run corrections to be tested under more complex, realistic con-

ditions. Cellulose filter packing media (Sigma C8002 α-Cellulose, Sigma-Aldrich, St. Louis,

MO) were used for the feedstock. Feedstock analysis is located in Table 2.2.

For each experiment, 180 g of cellulose were dried to 4.57%±0.05% (α = 0.05, n=4)

moisture content at 110◦C in an oven (Fisher IsoTemp). Of this, approximetly 144 g were

placed inside the reactor within 5 minutes of removal from the drying furnace. Then the

reactor was assembled, the chiller was set to 20◦C, the balance weighing system (circuitry

and mechanism) was set to 26.0◦C, and the room temperature was set to 23◦C. Nitrogen

was used as a carrier and quench carrier gas. Flow rates were 0.25 Lmin

each (0.5 Lmin

total at

the exhaust).

After the temperatures stabilized, the verification procedure was repeated to insure that

the weight measurement system had not significantly changed, such as by damage to the

load cells. The balance output was averaged for 10 minutes with the condenser coupling

connected, thus providing the tare value of the reactor. The condenser balance was also

tared.

Page 40: Multi-Phase Packed-Bed Thermogravimetric Analyzer

28

Table 2.2: Feedstock Characteristics. Values are in percent with 95% confidence intervals(α = 0.05, n=3)

Parameter Cellulose

Cellulosea 68.60±1.15Hemicellulosea 7.89±1.05

Lignina 0.00±0.00Extractivesa 23.51

Cb 44.28±0.59Hb 5.69±0.04Nb 0.42±0.06Sb 0.14±0.10Ob 49.46±0.59

Ashbc 0.00±0.00Fixed Carbonbc 11.67±0.55

Volatilesbc 88.64±0.62Moisturebc 4.88±0.03

HHV (MJkg

)bc 16.42±0.24aTested at the University of Georgia (UGA) AESL Feed and Environmental Quality Laboratoryusing a ANKOM200 fiber analyzer.bTested on location with a LECO CHNS-932 and VTF-900 Ultimate analyzer using ASTM standardsD5291 and D3176cTested on location with a LECO TGA 701 Proximate analyzer using ASTM standards D5142 and

E1131. Higher heating value (HHV) based on ASTM D3180. Values measured using an oven-dry

basis.

The temperature profile used for the blank runs was also used for these biomass runs

with the system increasing at 3◦Cmin

from room temperature to 600◦C and held at 600◦C for

45 minutes. After this, the furnace and preheater were turned off and allowed to cool down

overnight with a standby nitrogen flow rate of 0.05 Lmin

each. The nitrogen flow remained at

the standby rate until disassembly of the system.

The solid (char) yield was calculated by the weight difference of the reactor core before

and after the experiment using a laboratory balance. The condensate yield was measured

using the condensate balance data and later corrected using the blank run data.

Page 41: Multi-Phase Packed-Bed Thermogravimetric Analyzer

29

Load cell output data collected during the experiment were transformed into weight

measurements using the calibration model and method. These data, along with the condenser

weights, were corrected by subtracting the average blank run data from each matching time

point. Averages of the biomass experiment runs were calculated after all transformations,

such as normalization and finding the derivative of the TGA curve (DTG).

2.4 Results

2.4.1 Calibration

Voltage output from the load cell conditioning circuit decreased both with increasing weight

and, to a smaller degree, the temperature of the balance mechanism. The actual load cell

conditioning circuit temperature varied between 25.8◦C and 27.3◦C for all three temperature

set-points due to natural oscillations and limitations of the PID temperature controller. Two

sets of data stood out as major outliers and were removed during the regression analysis:

(a) replication #1 at 25.6◦C and 169.3 g, and (b) replication #3 at 26.4◦C and 169.3 g. The

outlier data in the first replication appeared to have been due to a major experimental

error (e.g., wrong weight application). Temperature fluctuations in the laboratory may have

caused the outliers in the third replication.

A number of regression model forms of Equation 2.1 (W=weight, Tc= Circuit temper-

ature, V=Voltage output from load cells) were tested, such as linear, quadratic, cubic, and

combinations thereof. The cubic relationship (Equation 2.2) yielded the best fit, least signif-

icant trends in the residuals (|PredictedWeight−ActualWeight|), and least deviation from

the assumption of error normality (A, B, C, D, E, F are regression coefficients and I is the

intercept).

V = f(W,Tc) + error (2.1)

V 2 − 1

−2= A ·W + B ·W + C ·W + D ·Tc + E ·Tc2 + F ·Tc3 + I (2.2)

Page 42: Multi-Phase Packed-Bed Thermogravimetric Analyzer

30

Table 2.3: Calibration regression model statistics. All coefficients (Co.) have a greater than0.001 significance.

Co. Value Std.Error t value Prob.(>|t|)

I 1.300 2.46×10−1 5.28 1.35×10−7

A 2.176×10−4 2.87×10−8 7589.41 <2.22×10−16

B -5.287×10−8 1.50×10−10 -352.31 <2.22×10−16

C 1.493×10−11 1.96×10−13 76.28 <2.22×10−16

D -1.781×10−1 2.78×10−2 -6.40 1.71×10−10

E 6.932×10−3 1.05×10−3 6.61 4.14×10−11

F -8.978×10−5 1.32×10−5 -6.82 1.04×10−11

R2=1.000; adjR2=1.000; df=4902

The regression procedure uses the assumption that the error is a normal distribution

with equal variances. However, analysis of the untransformed (i.e., just Voltage, not V 2−1

−2)

response variable in Equation 2.2 indicated a possible deviation from normality. The Box-

Cox method Neter et al. [19] was used to determine a variance stabilization factor (λ) of

-2. The transformed response of V −1−2

increased the linearity of the model and reduced the

deviation from normality. Table 2.3 shows the value, standard error, t-value, and probability

for each coefficient of the regression equation. All of the coefficients approached a p<0.001

significance or smaller with 4902 degrees of freedom.

In order to predict weights based on voltage and temperature, Equation 2.2 needed to be

transformed so weight was predicted as a function of voltage and temperature. Through a

series of substitutions the root-finding method outlined in algorithm 1 was used to calculate

the weight (W) of the reactor for each measurement as a function of measured temperature

(T), voltage (V), and regression coefficients (A, B, C).

Figure 2.4 shows a series of residual graphs showing gram error between the predicted

and actual weights versus temperature (Figure 2.4a), actual weight (Figure 2.4.b), and time

(Figure 2.4.c). In Figure 2.4.a there is a disturbance between 26.2◦C and 26.4◦C that may

have been due to unaccounted temperature fluctuations. In Figure 2.4.b and there appears

Page 43: Multi-Phase Packed-Bed Thermogravimetric Analyzer

31

Algorithm 1 Empirical transformation used to calculate weight (W) from measured tem-perature (T), voltage (V), and the regression coefficients (A, B, C, D, E, F, I). Variables Q,S, and R are intermediate variables.

Q =3 ·A ·C − B2

9 ·A2(2.3)

G = D ·T + E ·T 2 + F ·T 3 + I +(−V )−2 − 1

2(2.4)

R =9 ·A ·B ·C − 27 ·A2 ∗ G − 2 ·B3

54 ·A3(2.5)

S =3

R +√

Q3 + R2 (2.6)

T =3

R −√

Q3 + R2 (2.7)

W = S + T −B

3 ·A(2.8)

to be a trend of increasing variance and mean error with weight and may be due to heavier

sample weights causing movement of the hanging solid-phase decomposition system and thus

changing the influence of attachments (i.e., tubes and wires). The precision of the reactor

(solid phase) weights predicted by the calibration model was 0.0100 g±0.0065 g (α = 0.05,

n=4945) (Figure 2.4.a).

The error in the verification experiments made before each blank run was calculated by

taring the weight with the currently measured empty reactor weight (Equation 2.9) where E

is the current error, Wp is the predicted weight, Wt is the tare weight, and W is the actual

applied weight. Additionally, the error was also calculated by taring the weight with the first

verifications of measured empty reactor weight, thus giving an idea of change in error over

time when compared to the first verification before any blank runs.

%E =(Wc − Wt) − W

500 g(2.9)

Page 44: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 45: Multi-Phase Packed-Bed Thermogravimetric Analyzer

33

No significant relationship was observed between the error in the mean and the time

between each verification (correlation of R2 = 0.269) when tared to the measured empty

reactor weight from that day. There is a relationship between the error in the mean and the

time between each verification (α = 0.05, R2 = 0.984) when tared to the initial empty reactor

weight with the mean error decreasing at a rate of about -0.00138 %Errorh

(-0.0069 gh). This

could be due to relaxing effects of connecting elements, such as wire insulation and tubes. The

mean error between the predicted and applied weights was 0.040%±0.022% (0.198 g±0.112 g)

with α=0.05, four applied weights, and five verifications (n=20).

During verification, there was not a significant difference (α = 0.95) between the error

associated with the 169.3 g and 332.2 g weights, nor between the 169.3 g and 503.1 g weights;

but there is a difference between the 332.2 g and 503.1 g weight measurements. The 169.3 g,

and especially the 503.1 g, weight measurements roughly follow the trend in Figure 2.4.b by

deviating increasingly from a mean of 0 g with increasing applied weight.

These results indicate the importance of performing initial tare measurements and veri-

fications before each experiment.

2.4.2 Blank Runs

During the blank runs, the measured reactor and condenser weights decreased significantly

as a function of time (Figure 2.5.a) instead of remaining at 0 g. The average offset at the end

of the each experiment (i.e., final yield) was -24.257 g±0.034 g (α = 0.05, σ = 0.671 g, n=5)

for the reactor and -0.705 g±0.017 g (α = 0.05, σ = 0.333 g, n=5) for the condenser. Figure

2.5.b shows the error in the mean weights of all the blank runs. The maximum standard

deviation of the solid weight was 1.766 g. Assuming a normal distribution of error, 95% of

the data points fell within an error range of ±3.4609 g (1.96 × σ, n=5) at that point. The

maximum 95% confidence interval of the mean at that point was ±0.278 g (α=0.05, n=5).

Page 46: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 47: Multi-Phase Packed-Bed Thermogravimetric Analyzer

35

The maximum standard deviation of the condenser weight was 0.375 g (n=5). Assuming

a normal distribution of error, 95% of the data points fell within an error range of ±0.7350 g

(1.96 × σ, n=5) at that point. The maximum 95% confidence interval of the mean at that

point was ±0.059 g (α=0.05, n=5).

The sample temperature lagged behind the furnace temperature (Figure 2.5.a) with a

maximum temperature difference of about -66.6◦C at 53.6 minutes into the run (furnace

temperature 181.0◦C). Since the preheater was implemented to maintain the reactor inlet

temperature the same as the furnace temperature, this lag suggests the preheater was not

working properly. Also, as the furnace temperature approached 600◦C, the preheater gas

exit temperature began lagging behind the furnace set-point. The preheater could likely not

supply enough energy to the carrier gas at high temperatures and/or the preheated gas was

cooled by the reactor entrance tube, thermal-wells, and balance hardware.

The deviation from the mean temperatures is shown in Figure 2.5.c. In general, tempera-

tures stayed within ±1.5◦C or less. The spike near 190 minutes is an artifact due to a missed

reading.

The average weight and time data points developed here were used as a correction curve

for each blank run.

2.4.3 Biomass Runs

Figure 2.6 shows the uncorrected average normalized weight, temperature, and time profile

of the biomass runs. Figure 2.7 shows the same data with the blank run curve subtracted.

The highest rate of decomposition was between 100 and 150 minutes. Two thermal events

were observed in the difference in temperature between the furnace and sample temperature,

one during the main reaction at about 125 minutes and the other at about 150 minutes.

The maximum errors in the corrected data for both the solid-phase and the condensate

(Table 2.4) are similar to those in the uncorrected data and occurred at 117.87 minutes

for the solid and 123.40 minutes for the liquid. Assuming α=0.05 and four replications

Page 48: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 49: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 50: Multi-Phase Packed-Bed Thermogravimetric Analyzer

38

Table 2.4: Maximum errors observed during the entire temperature program. Values arethe 95% confidence intervals (α = 0.05, n=4) of the normalized mean weight (mean initialsample weight of 143.95g).

Device and Phase Cellulose

PB-TGA Solid Phase 1.11%±0.19%@ 117.8 min

PB-TGA Condensate 0.76%±0.13%@ 123.4 min

(n=4), the maximum normalized solid-phase error was 1.11%±0.19% (1.592 g±0.280 g) and

0.76%±0.13% (1.089 g±0.192 g) for condensate.

The average final weights shown in Table 2.5 were calculated by comparing the av-

erage of weights measured during the last 10 minutes of the experiment and weight of

the reactor after the experiment. The final solid (char) weight measured by the reactor

balance was 23.11%±0.02% normalized (33.27 g±0.03 g). In comparison, the actual weight

of the char measured after the experiment using an analytical laboratory balance was

22.99%±0.03% normalized (33.09 g±0.04 g). These are statistically different with an error of

0.497%±1.441%. The average corrected condensate weight was 53.27%±0.02% normalized

(76.68 g±0.02 g). Thus, the average yields were 23.11% char, 53.27% condensate, and 23.63%

other (gas and un-condensed vapor).

In comparison, Capart et. al. Capart et al. [8] obtained 14.15% char when using a heating

rate of 3.2◦Cmin

, micro-crystalline cellulose, and a micro-TGA device with a much smaller

sample. The slightly lower heating rate in the current research may contribute to the lower

char yield [3, 23], but it is more likely due to increased secondary reactions due to the

increased vapor-particle contact time in the packed-bed TGA.

Figures 2.8.a, 2.8.b, and 2.8.c show the blank run corrected TGA and DTG curves as

a function of temperature (sample-center, furnace, and average). The derivatives were con-

structed using the format in Equation 2.10 and had a window size of 1 minute (32 data

Page 51: Multi-Phase Packed-Bed Thermogravimetric Analyzer

39

Table 2.5: Corrected cellulose product yields(α=0.05, n=4).Solid Liquid

PB-TGA 23.11%±0.02% 53.27%±0.02%(33.27 g)±0.03 g) (76.68 g)±0.03 g)

Actual 22.99%±0.03%(33.09 g)±0.04 g)

Error 0.497%±1.44%∗Based on an initial sample weight of 143.95 g±0.43 g

points). There are clear differences between the different temperature references that are

likely due to the thermal lag observable in Figures 2.6 and 2.7. The TG data in reference to

the sample-center temperature is shifted to the left since most of the sample, being heated

primarily by the furnace, is hotter than the center. Thus, pyrolysis reactions may be more

advanced in the material contacting the reactor wall. Likewise, the furnace temperature is

shifted to the right [13]. The average of these two temperatures is used as a reference in an

effort to correct for this lag and to approximate the overall temperature of the sample.

DTG =dW

dT=

∆W

∆T=

W−0.5min − W+0.5min

T−0.5min − T+0.5min

(2.10)

There were two resolvable peaks: peak 1 (Table 2.6) occurs at an average of 124.85 minutes

in the reactor and 123.67 minutes in the condenser with a difference of 1.18 minutes; and

peak 2 (Table 2.7) occurs at an average of 153.19 minutes in the reactor and 151.95 minutes

in the condenser with a difference of 1.24 minutes. This suggests that condenser weight

measurement occurs after the reactor weight measurement by an average of 1.21 minutes,

assuming no significant mass transfer lag.

Page 52: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 53: Multi-Phase Packed-Bed Thermogravimetric Analyzer

41

Table 2.6: Cellulose DTG Peak 1 values ((DratioD◦C

= 1◦C

)) and locations (temperature andtime).

Temperature referenceSample Furnace Average

Solid-phase -0.0126 1◦C

-0.0092 1◦C

-0.0105 1◦C

279.94◦C 395.97◦C 338.78◦C124.07min 125.34min 125.14min

Condensate 0.0105 1◦C

0.0075 1◦C

0.0086 1◦C

279.4◦C 390.16◦C 335.38◦C123.80min 123.40min 123.80min

Peak 1 was distinguished by a white, opaque vapor in the condenser and most likely

corresponds primarily to cellulose decomposition. The peak, occurring at an average 338.78◦C

in the solid-phase and 335.38◦C in the condensate, is similar to values reported by Guar and

Reed [13]. They found cellulose decomposition peak locations at 340.3◦C (Avicel PH 102

cellulose), 351.1◦C (Baker analyzed cellulose), and 325.1◦C (Acetobacter xylinum cellulose)

using a micro-TGA device, a 10◦Cmin

heating rate, and samples less than 10 mg.

Peak 1 also corresponds to the first thermal event in Figures 2.6 and 2.7 at about 125

minutes. The increase of the sample-center temperature relative to the furnace temperature

indicates an exothermic event, a characteristic reported in some of the cellulose analyses

performed by Guar and Reed [13]. This change in temperature lag compresses the data into

the sharp peak observed in Figure 2.8.a. At the end of peak 1 the condenser unit cleared

quickly and almost completely. Shortly thereafter, and corresponding to peak 2, a semi-

transparent brown vapor filled the condensers.

Unlike peak 1, peak 2 does not appear to correspond to standard decomposition charac-

teristics of cellulose, hemicellulose, or lignin. Although the mass change during peak 2 was

below the noise level, it was well defined by a semi-transparent brown vapor and by the

second thermal event observable in Figures 2.6 and 2.7 at about 150 minutes. The sharp

change in the temperature difference between the sample-center and furnace temperature

Page 54: Multi-Phase Packed-Bed Thermogravimetric Analyzer

42

Table 2.7: Cellulose DTG Peak 2 values (DratioD◦C

= 1◦C

) and locations (temperature and time).Peak 2 is not present when the furnace temperature is used.

Temperature referenceSample∗ Average

Solid-phase -0.055 1◦C

-0.0017 1◦C

448.75◦C 461.89◦C155.40min 150.97min

Condensate 0.147 1◦C

0.0008 1◦C

448.72◦C 459.81◦C155.10min 148.80min

∗Approximate highest negative (for reactor) and positive (for condenser) values smaller than ∞.dratiod◦C oscillated sharply between positive and negative values.

compressed the data in Figure 2.8.a so peak 2 is resolvable in the DTG. Figure 2.8.b shows

no such peak since the reference furnace temperature increases at a relatively constant 3◦Cmin

.

From this information it appears that the reaction(s) occurring during peak 2 have a large

exothermic component with minimal mass change. Since peaks 1 and 2 are so clearly defined

it would be relatively easy and interesting to compare the char, oil, and reactor before and

after each peak by halting the experiment at the appropriate temperatures.

Another way to present the corrected cellulose data is as time dependent TGA and DTG

curves (Figure 2.9). Instead of presenting the TGA and DTG as a percent or change in

percent of the initial weight, they are presented in terms of the fractional conversion “X”

(Equation 3.2). The second derivative is also shown in Figure 2.9 to help illustrate the

location of inflection points. In micro-TGA analysis with small samples and low heating

rates the time and temperature TGA and DTG curves are similar since there is very little

temperature lag. In this case, however, the shape of the TGA and DTG curves are quite

different. This time dependent and “X” conversion representation of the TG data is often

used as an input for kinetic models.

Page 55: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 56: Multi-Phase Packed-Bed Thermogravimetric Analyzer

44

last point is especially important since bio-oil is often a key product and this provides a way

to correlate TG information to bio-oil characteristics.

The biggest limitation was low solid-phase weight sensitivity and susceptibility to noise

that required an extensive calibration procedure, electrical shielding, and temperature con-

trol. This could be solved by using an off-the-shelf lab balance that possesses calibration and

temperature compensation features. Also, the preheater should be either redesigned to track

the furnace temperature and help reduce the thermal lag, or removed.

The large thermal lag in the sample can either be seen as an advantage or disadvantage.

While this can be a complicating factor compared to an idealized micro-TGA with a small

thermal lag, it is a characteristic reflected in larger-scale real-world systems.

Some changes could be made to expand the capabilities of the system, such as adding

additional sample temperature sensors to help characterize the thermal-lag, using a frac-

tionation condenser system to separate condensable material by condensation temperature,

and using an in-line system such as gas chromatography, mass spectrometry, and/or Fourier

Transform Infrared Spectroscopy (FT-IR), for characterizing the exhaust.

Below are research questions that could be easily investigated with this system, or with

some modification:

1. What are the TGA based reaction kinetics (i.e., activation energy and frequency factor)

for the solid phase, condensate, and (by difference) of the non-condensable gas phase?

2. What effect does feedstock morphology (i.e., pellets, powder, chips) and bed depth

have on the TGA kinetics, including secondary reactions?

3. What effects do different operating parameters, such as heating rate, gas type (e.g.,

helium, nitrogen, carbon dioxide), flow rate, and condenser temperature have on TGA

kinetics?

Page 57: Multi-Phase Packed-Bed Thermogravimetric Analyzer

45

2.6 References

[1] Antal, M. J. J. (1995). Cellulose pyrolysis kinetics: The current state of knowledge.

Industrial and Engineering Chemical Research 34, 703–717.

[2] Antal, M. J. J. and G. Morten (2003). The art, science, and technology of charcoal

production. Industrial and Engineering Chemical Research 42, 1619–1640.

[3] Antal, M. J. J., G. Varhegyi, and E. Jakab (1998). Cellulose pyrolysis kinetics: Revisited.

Industrial and Engineering Chemical Research 37, 1267–1275.

[4] Aresta, M., A. Dibenedetto, and G. Barberio (2005). Utilization of macro-algae for

enhanced CO2 fixation and biofuel production: Development of a computing software for

an lca study. Fuel Processing Technology 86, 1679–1693.

[5] Baker, R. R., S. Coburn, C. Liu, and J. Tetteh (2005). Pyrolysis of saccharide tobacco

ingredients: a TGA-FTIR investigation. Journal of Analytical and Applied Pyrolysis 74,

171–180.

[6] Bridgwater, A. and G. Grassi (1991). Biomass pyrolysis liquids upgrading and utilization,

Chapter A review of biomass pyrolysis and pyrolysis technologies, pp. 11–92. London:

Elsevier Science.

[7] Bridgwater, A. and G. Peacocke (2000). Fast pyrolysis processes for biomass. Renewable

and Sustainable Energy Reviews 4, 1–73.

[8] Capart, R., L. Khezami, and A. K. Burnham (2004). Assessment of various kinetic models

for the pyrolysis of a microgranular cellulose. Thermochimica Acta 417, 79–89.

[9] Demirbas, A. (2001a). Biomass resource facilities and biomass conversion processing for

fuels and chemicals. Energy Conversion and Management 42, 1357–1378.

[10] Demirbas, M. and M. Balat (2006). Recent advances on the production and utilization

trends of bio-fuels. Energy Conversion and Management (47), 2371–2381.

Page 58: Multi-Phase Packed-Bed Thermogravimetric Analyzer

46

[11] Díaz, C. J. G. (2006). Understanding Biomass Pyrolysis Kinetics: Improved Model-

ing Based on Comprehensive Thermokinetic Analysis. Ph.d. dissertation, Departament

d’Enginyeria Química.

[12] Galvagno, S., S. Casu, M. Martino, E. Di Palma, and S. Portofino (2007). Thermal

and kinetic study of tyre waste pyrolysis via TG-FTIR-MS analysis. Journal of Thermal

Analysis and Calorimetry 88, 507–514.

[13] Guar, S. and T. B. Reed (1998). Thermal Data for Natural and Synthetic Fuels. New

York: Marcel Dekker, Inc.

[14] Guo, J. and A. Lua (2000). Kinetic study on pyrolysis of extracted oil palm fiber.

Journal of Thermal Analysis and Calorimetry 59, 763–774.

[15] Guo, J. and A. Lua (2001). Kinetic study on pyrolytic process of oil-palm solid waste

using two-step consecutive reaction model. Biomass and Bioenergy 20 (3), 223–233.

[16] Leung, D. and C. Wang (1998). Kinetic study of scrap tyre pyrolysis and combustion.

Journal of Analytical and Applied Pyrolysis 45, 153–169.

[17] Mangut, V., E. Sabio, J. Gañán, J. González, A. Ramiro, C. González, S. Román, and

A. Al-Kassir (2006). Thermalgravimetric study of the pyrolysis of biomass residues from

tomato processing industry. Fuel Processing Technology 87, 109–115.

[18] Meier, D. and O. Faix (1999). State of the art of applied fast pyrolysis of lignocellulosic

materials. Bioresource Technology 68, 71–77.

[19] Neter, J., M. H. Kutner, C. J. Nachtsheim, and W. Wasserman (Eds.) (1999). Applied

Linear Statistical Models (4th edition ed.). New York: McGraw-Hill International.

[20] Onay, O. and O. M. Kockar (2003). Technical note: Slow, fast and flash pyrolysis of

rapeseed. Renewable Energy 28, 2417–2433.

Page 59: Multi-Phase Packed-Bed Thermogravimetric Analyzer

47

[21] Pappa, A., S. Kyriakou, K. Mikedi, N. Tzamtzis, and M. Statheropoulos (2004). Design

considerations and an example of applications of an in-house made TG-MS interface.

Journal of Thermal Analysis and Calorimetry 78, 415–426.

[22] Stenseng, M., A. Jensen, and K. Dam-Johansen (2001). Investigation of biomass py-

rolysis by thermogravimetric analysis and differential scanning calorimetry. Journal of

Analytical and Applied Pyrolysis 58-59, 765–780.

[23] Várhegyi, G., M. J. J. Antal, E. Jakab, and P. Szabó (1997). Kinetic modeling of

biomass pyrolysis. Journal of Analytical and Applied Pyrolysis 42, 73–87.

Page 60: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Chapter 3

Comparison of a novel bench size packed-bed solid

and liquid phase TGA and micro-TGA1

1Jarrett, Erik and Das, K.C. (2008). To be submitted to the Journal of Analytical and AppliedPyrolysis.

48

Page 61: Multi-Phase Packed-Bed Thermogravimetric Analyzer

49

Abstract

A bench scale (571 mL) packed-bed thermogravimetric analyzer (PB-TGA) with a con-

densate measurement unit was designed and built to study both solid decomposition and

condensate formation during pyrolysis. The PB-TGA and a traditional micro-TGA were

compared using cellulose powder and pine pellet feedstocks. Both devices heated the sam-

ples at 3◦Cmin

with a nitrogen carrier gas. Samples in the micro-TGA were ∼10 mg contained

in a small, open crucible surrounded by the carrier gas. The PB-TGA contained 160 g to

250 g (depending on feedstock) of sample with the carrier gas forced at 0.25 Lmin

through

the sample bed. Unlike the micro-TGA, the PB-TGA measured not only the weight loss of

the solid phase, but also the weight gain of the condensable liquids (bio-oil) as a function

of temperature. The two reactors were compared by product yields, error between repli-

cations, temperature dependent differential thermogravimetric analysis (DTG) peaks, time

dependent DTG peaks, and major kinetic parameters. The greatest differences were in the

product yields, the presence of a significant thermal lag in the PB-TGA, the shapes of the

TGA and DTG curves, and the kinetics parameters.

Keywords: TGA; packed-bed; kinetics; cellulose; pine-pellets; pyrolysis

Page 62: Multi-Phase Packed-Bed Thermogravimetric Analyzer

50

3.1 Introduction

Pyrolysis, a process in which biomass is heated to a high temperature in a non-oxidizing

environment, has been employed to produce energy products that are potentially cleaner and

more efficient to transport than raw biomass. Instead of burning, molecules thermally crack

and either volatilize, or take part in further reactions before volatilizing and/or becoming

char. A portion of the vapor can be collected and condensed into a material known as bio-oil.

The resulting char, bio-oil, and gas yields and chemical characteristics can be controlled by

modifying operating parameters, such as biomass type and reactor conditions [2, 12, 20, 9].

Thermogravimetric analysis (TGA), a technique in which the weight loss of a sample

is measured as a function of temperature, is often used to evaluate the effect of different

operating parameters on pyrolysis [23]. Although existing TGA instruments have proved to

be useful in pyrolysis research there are a number of limitations that prompted the creation

of a new TGA device.

During pyrolysis, weight loss is reflected by the volatilization of compounds that are

both immediately carried away by a carrier gas in a single-step reaction and those that

undergo multiple, possibly interacting, reactions before volatilizing [5]. The characteristics

of these reactions are governed by a variety of factors. These may be feedstock-related, such

as composition, particle size [16, 11], particle heat transfer characteristics [5], and tissue

structure. They may also be operating-condition related, such as reactor type (i.e., packed-

bed, fluidized-bed, etc.) Bridgwater and Peacocke [6], operating temperature, heating rate

[3, 24], thermal gradients [18], carrier gas Adebanjo [1], Pindoria et al. [19], and carrier gas

flow rates [2].

All of these many feedstock and operational parameters directly relate not only to just

the production and chemical composition of different products, but also to the operation

and energy balance of the whole process. This is of particular concern when pyrolysis is

employed for biomass-based renewable energy. Thus, it is important to have analytical tech-

niques, such as TGA analysis, that can be used to efficiently characterize pyrolysis systems.

Page 63: Multi-Phase Packed-Bed Thermogravimetric Analyzer

51

This information can then be used in feasibility assessments, energy balances [8], model de-

velopment, and other techniques so that efficient and optimized biomass conversion systems

can be developed [4].

TGA analysis provides a method to easily test the effects of many operating conditions

on weight related changes during pyrolysis in an a idealized environment. In small analytical

systems, a sample may range from 3 mg to 30 mg while in larger systems this may be as

high as 0.5 g to 4 g. The sample is contained in an open or closed crucible subjected to a

pre-programmed heating profile and an inert sweep gas, such as nitrogen.

There are several methods for extracting reaction information, two of which are isothermal

and dynamic analysis. In isothermal analysis, the sample is heated and held at the analysis

temperature until all decomposition is complete and there is no significant mass loss. The

sample is then heated to the next temperature, and the process is repeated until the desired

temperature range is completed. In dynamic analysis, the sample is heated at a constant

rate to a target temperature and the resulting TG curve is used to determine the kinetics

Capart et al. [7]. In either method, basic information includes product yields, major reaction

peaks, magnitudes, and peak onset and ending times. Often, TGA systems are coupled

with downstream analytical instruments, such as Fourier-Transform Infrared Spectroscopy

(FTIR), mass-spectrometry [10, 25, 13], and chromatography.

Additional analysis can be performed on the TG data to determine overall reaction

kinetics information for specific reactions/peaks in the thermogravimetric (TG) data using

the Arrhenius equation, of which there are a number of forms Guar and Reed [15]. One

common dynamic analysis model is shown in Equation 3.1. A (1s) is the pre-exponential

factor, or frequency factor, E ( Jmol

) is activation energy, R (8.314472 JK ·mol

) is the universal

gas constant, T (Kelvin) is absolute temperature, and t (seconds) is time. X is the fractional

reaction conversion based on weight (Equation 3.2) and is expressed in terms of the initial

sample weight (Wo), the current measured weight (W ), and the final sample weight (Wf ).

The function of f(x) is a transformation based on the X conversion and can be represented

Page 64: Multi-Phase Packed-Bed Thermogravimetric Analyzer

52

by a number of different models based on reaction mechanisms. Some common ones are

outlined and evaluated by Guo and Lua [16] and Capart et al. [7]. A simple example is the

nth order model shown in Equation 3.3 where n is the order number of the reaction Guo and

Lua [17].

dX

dt= Ae

−E

RT f(X) (3.1)

X =Wo − W

Wo − Wf

(3.2)

f(X) = (1 − X)n (3.3)

When the sample is analyzed in dynamic mode with one heating rate, Equation 3.4 can

be used as a substitution, along with Equations 3.2 and 3.3, in Equation 3.1 to give Equation

3.5. Linear regression is performed on the ln(dXdT

), 1T, and ln(1−X) data. A, E, and n (unless

restricted to a particular value) can be obtained from the regression coefficients [7].

q =dT

dt(3.4)

ln

(

dX

dT

)

= ln

(

A

q

)

−E

RT+ n · ln(1 − X) (3.5)

Much information can be gained from typical TGA systems, but some of their charac-

teristics can also be seen as limitations when attempting to understand pyrolysis in large

reactors that are far from ideal and in which bio-oil is a desired product. Larger reactors

may experience large thermal gradients, use a variety of feedstock particle sizes, and produce

bio-oil as an energy product. It is important to have a method to collect TG data in these

system in order to better understand pyrolysis characteristics and how they relate to the

more idealized conditions found in typical TGA systems.

Recently a new TGA device was developed in an attempt to satisfy these limitations in

which relatively larger samples (∼571 mL) are analyzed in a packed-bed configuration with

Page 65: Multi-Phase Packed-Bed Thermogravimetric Analyzer

53

the carrier gas forced at a constant rate through the sample. The weight change of both

the solid phase and the condensable liquid (bio-oil) is measured while providing a reactor

configuration more comparable to some larger scale reactors. The objective of this project was

to compare a small analytical scale TGA (micro-TGA) to the new larger scale packed-bed,

solid and liquid phase TGA (PB-TGA).

3.2 Experimental

A powdered cellulose filter packing medium (Sigma C8002 α-Cellulose, Sigma-Aldrich, St.

Louis, MO, USA) and pine pellets (Equine bedding from Southern Shavings Co., Cher-

ryville, NC, USA) were used as feedstocks. Cellulose was chosen given its frequent use in

related pyrolysis research. Pine pellets were chosen as a representative woody biomass with

cellulose, hemicellulose, and lignin. Table 3.1 shows the composition of each feedstock. The

α-cellulose was composed of 68.60% cellulose, 7.89% hemicellulose, no lignin, and 23.51%

extractable material. Pine pellets were composed of 47.43% cellulose, 14.44% hemicellulose,

24.62% lignin, and 12.50% extractable material. Extractives were not removed from either

feedstock before any of the TGA experiments.

The cellulose powder was used in both the micro-TGA and the PB-TGA without mod-

ification, but the pine pellets were ground (Cyclotec 1093 Sample Mill, Foss-Tecator, Eden

Prairie, MN, USA) into a powder for use in the micro-TGA because of size constraints. Both

feedstocks were dried (Fisher IsoTemp, Fisher Scientific, Waltham, MA, USA) at 110◦C un-

til reaching a steady-state 4.57%±0.05% (α=0.05, n=4) moisture content for cellulose and

7.85%±0.18% (α=0.05, n=3 ) for pine pellets.

The micro-TGA device was a Mettler Toledo TGA/SDTA851 (Mettler Toledo, Inc.,

Columbus, OH, USA) with an automatic loading robot. Open-top alumina crucibles con-

tained the ∼10 mg samples. The time, sample weight, sample temperature, and furnace/ref-

erence temperature were logged every 15 seconds.

Page 66: Multi-Phase Packed-Bed Thermogravimetric Analyzer

54

Table 3.1: Feedstock Characteristics. Values are in percent with 95% confidence intervals(α = 0.05, n=3).

Parameter Cellulose Pine Pellets

Cellulosea 68.60±1.15 47.43±0.34Hemicellulosea 7.89±1.05 14.44±1.64

Lignina 0.00±0.00 24.62±0.48Extractivesa 23.51 13.50

Cb 44.28±0.59 50.91±0.59Hb 5.69±0.04 5.58±0.13Nb 0.42±0.06 0.54±0.09Sb 0.14±0.10 0.10±0.01Ob 49.46±0.59 40.27±1.02

Ashbc 0.00±0.00 2.59±0.00Fixed Carbonbc 11.67±0.55 18.54±0.51

Volatilesbc 88.64±0.62 78.87±0.43Moisturebc 4.88±0.03 7.00±0.01

HHV (MJkg

)bd 16.42±0.24 19.01±0.57aTested at the University of Georgia (UGA) AESL Feed and Environmental Quality Laboratory.bTested at the UGA Biological and Agricultural Engineering Bio-Conversion Laboratory.cMeasured using an oven-dry basisdHigher Heating Value

The PB-TGA was a newly developed device used to obtain TGA curves for both the solid-

phase (char) and the condensate (bio-oil). Figure 3.1 shows a flow chart of major components

in the device. Feedstock was contained within a 571 mL, vertically oriented, cylindrical stain-

less steel vessel, packed bed reactor heatedby a modified muffle furnace (Model 550-14, Fisher

Scientific, Waltham, MA, USA). Nitrogen passed through the packed-bed at 0.25 Lmin

and was

quickly cooled after exiting the reactor with another stream of 0.25 Lmin

nitrogen. The pre-

heater and reactor hung from a custom built balance mechanism that measured weight loss

in the feedstock during heating.

The exhaust from the the solid-phase TGA system passed through a mechanical decou-

pling device and into a condenser assembly, which was maintained at 20◦C. The condensers

Page 67: Multi-Phase Packed-Bed Thermogravimetric Analyzer

55

Figure 3.1: Flow chart of the PB-TGA.

were mounted on an electronic balance (Model RD6RM, Ohaus, Pine Brook, NJ, USA) that

continuousy logged the condensate weight.

The PB-TGA data were logged using a computer based LabView data acquisition system

(National Instruments, Austin, TX., USA) at 600 Hz for 1200 data points. The data were

averaged using a moving average with a window of 32 seconds. Solid-phase weight information

was transformed into a usable weight format using a calibration model and a blank run

correction method. Major data recorded included the solid weight, condensate weight, sample

temperature, and furnace temperature.

3.2.1 Procedure

The micro-TGA was used to analyze four samples (replicates) of cellulose powder and four

samples of pine pellet powder in a completely randomized order and in one series. Before

each experiment the micro-TGA was purged with a nitrogen carrier gas for at least 8 hours.

Each crucible was tared, and blank runs were performed to automatically correct for crucible

weight and blank run effects during the actual experiment.

Page 68: Multi-Phase Packed-Bed Thermogravimetric Analyzer

56

Each tared crucible was filled with as close to 10 mg of feedstock as possible. These were

not pre-dried, since the samples would inevitably absorb moisture from the atmosphere while

waiting to be run. During each experiment the sample was initially cooled to 20◦C and then

heated to 600◦C at a rate of 3◦Cmin

. The sample was then held at 600◦C until the end of the

experiment 45 minutes later. The furnace was cooled down, the finished sample removed,

and a new sample loaded for the next experiment.

The PB-TGA was also used to analyze four samples (replicates) of pine pellets and four

samples of cellulose powder in that order. The samples were not randomized between each

experiment due to the logistical constraints of the device. The reactor was filled to capacity

with the dried feedstock (144 g for cellulose and 250 g for pine pellets) and inserted into the

nitrogen purged solid-phase TGA system within five minutes of removal from the drying

furnace. The nitrogen carrier and cooling gases were switched to 0.25 Lmin

and the system

leak tested at 25 psi.

Three key steps were performed before doing the actual experiment; (a) balancing, (b)

weight checking, and (c) taring. The system first needed to be balanced using a counter-

balance and by adjusting connection lines so there was no physical contact between the

hanging reactor assembly and anchored furnace and frame. The solid-phase weighing system

was checked by hanging different standard weights on the reactor and comparing the output

with the known weight. Finally, the reactor was weighed for 10 minutes and the output

averaged to provide a PB-TGA solid-phase tare value.

The sample was heated from room temperature (∼22◦C) to 600◦C at a rate of 3◦Cmin

. Like

the micro-TGA, the sample was then held at 600◦C until the end of the experiment 45 min-

utes later. Raw data for each experiment (e.g., for replications each of micro-TGA:cellulose,

micro-TGA:pine-pellets, PB-TGA:cellulose, and PB-TGA:pine-pellets) were compiled into

individual Gnumeric (V.1.8.2, GNU open source) software worksheets for post-experiment

analysis and processed in R (V.2.6.2, The R Foundation for Statistical Computing, Open

Source) for kinetics analysis. PB-TGA data were adjusted and corrected using the calibra-

Page 69: Multi-Phase Packed-Bed Thermogravimetric Analyzer

57

Table 3.2: Product yields after entire temperature program. Weights are percent normalizedto the initial sample weight. (α = 0.05, n=4).

Device and Phase Cellulose Pine Pellets

micro-TGA Solid Phase 14.63%±0.02% 23.87%±0.11%PB-TGA Solid Phase 23.11%±0.02% 29.61%±0.02%PB-TGA Condensate 53.27%±0.02% 49.21%±0.03%

tion model and blank run correction curve. Since the PB-TGA samples were already dried

at 110◦C, the PB-TGA results were assumed to be of a dried feedstock. The micro-TGA

samples, on the other hand, were not initially dried. Because of this, the micro-TGA data

for each experiment were offset using the average weight over 0.5 minutes (3 data points) at

110◦C.

3.3 Results and discussion

3.3.1 Time profile and errors involved

Figures 3.2 (cellulose) and 3.3 (pine pellets) show the mean profiles of weights and tem-

peratures as a function of time for both the micro-TGA and PB-TGA. The weights were

normalized to the initial sample weight. The micro-TGA only had one weight profile, since

only the solid-phase was being measured, and the difference between the furnace and sample

temperature approached zero, indicating a low thermal gradient in the sample. At the end

of the micro-TGA experiments, 14.63% cellulose char and 23.87% pine pellet char remained

(Table 3.2).

The micro-TGA cellulose yields are similar to the 14.15% char found by Capart et al.

[7] when using a heating rate of 3.2◦Cmin

, micro-crystalline cellulose, and a slightly smaller

sample.

Page 70: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 71: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 72: Multi-Phase Packed-Bed Thermogravimetric Analyzer

60

Table 3.3: Maximum errors observed during the entire temperature program. Values are the95% confidence intervals (α = 0.05, n=4) of the normalized mean weight.

Device and Phase Cellulose Pine Pellets

micro-TGA ±0.87% ±0.53%Solid Phase @ 105.8min @ 108.3minPB-TGA ±1.11% ±3.37%

Solid Phase @ 117.8min 132.1minPB-TGA ±0.13% ±0.29%

Condensate @ 123.4min 132.4min

The PB-TGA has two weight profiles, one for the solid phase (decreasing in weight)

and the other for the condensate (increasing in weight). Like the micro-TGA, the furnace

temperature curve increases at a steady 3◦Cmin

. However, unlike the micro-TGA, there is a

significant visual difference between the furnace and center sample temperature curves (i.e.,

∼80◦C), the presence of which indicates a large thermal gradient within the sample. Since

the carrier gas pre-heater was designed to minimize a thermal gradient, this difference also

indicates the pre-heater did not work as well as intended.

The shape of the sample-center and temperature difference curves indicates two thermal

events between 100 minutes and 160 minutes that cannot be explained by a steady state

thermal gradient. These correspond to the center and end of the major region of weight

loss/gain and indicate possible endothermic and/or exothermic reactions. The event from

140 minutes to 150 minutes (∼300◦C to 450◦C sample temperature) may correspond to an

exothermic reaction.

The PB-TGA reactor had a significantly higher char yield than the micro-TGA with

23.11% (8.48% higher) and 29.61% (5.74% higher) remaining for cellulose and pine pellets

respectively (Table 3.2). This higher solid yield is likely due to the increased potential for

secondary reactions in the packed-bed compared to the small sample size in the micro-TGA.

The condensate yield was 53.27% for cellulose and 49.21% for pine pellets in the PB-TGA.

Page 73: Multi-Phase Packed-Bed Thermogravimetric Analyzer

61

Table 3.3 provides the maximum 95% confidence intervals of the mean of four replications

for each biomass and device. The maximum solid-phase errors are higher than the condensate

errors in the PB-TGA. The solid-phase error could be improved by using either a more

sensitive solid-phase balance circuit with a higher signal to noise ratio or an off-the-shelf

precision balance. However, the error in the PB-TGA condensate data was lower than for

the micro-TGA.

3.3.2 Temperature dependent differential thermogravimetric anal-

ysis (DTG)

Figures 3.4 (cellulose) and 3.5 (pine pellets) present the weight data in terms of temperature,

rather that time. Instead of the normalized weight presented in Figures 3.2 and 3.3 (useful

for quick visualization), weights have been transformed into the fractional reaction conver-

sion mentioned in Equation 3.2 (useful for kinetics analysis). The temperature derivative

(Temperature DTG = dWeightdTemperature

) is the average of all four replications in each condition

(device and feedstock). Temperatures used in both the derivative and x-axis are the means of

the sample and furnace/reference temperatures and are assumed to represent average tem-

peratures in the sample. This assumption is likely valid in the micro-TGA; since very little

difference occurred between temperatures, and the sample is quite small. In the PB-TGA

the temperature varies much more, as shown in the temperature lag of Figures 3.2 and 3.3. A

more accurate method would be to calculate the mean based on several points between and

including the sample center and reactor wall. The temperature based DTG peak locations

and magnitudes are listed in Table 3.4.

There are several important differences between the micro-TGA results (Figures 3.4.a and

3.5.a) and the PB-TGA results (Figures 3.4.b and 3.5.b). First, the PB-TGA DTG curves are

much more symmetrical with a lower magnitude. The micro-TGA cellulose curve is negatively

skewed, a likely simultaneous result of cellulose, hemicellulose, and extractives pyrolysis. In

the micro-TGA pine pellet experiments the DTG curve appears to be a convolution of

Page 74: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 75: Multi-Phase Packed-Bed Thermogravimetric Analyzer
Page 76: Multi-Phase Packed-Bed Thermogravimetric Analyzer

64

Table 3.4: Temperature DTG (dWnormalizeddTemperature

) peak locations. Values are the change in normal-

ized weight over the change in mean temperature (n=4) of the sample and furnace/referencetemperatures.

Device and Phase Cellulose Pine PelletsdXdT

= 0.0213 X◦C

0.0138 X◦C

micro-TGA X = 0.673 0.692Solid Phase T = 336.97◦C 343.85◦C

t = 105.75min 108.00mindXdT

= 0.0140 X◦C

0.0078 X◦C

PB-TGA X = 0.441 0.497Solid Phase T = 335.80◦C 349.36◦C

t = 123.97min 127.60mindXdT

= 0.0163 X◦C

0.0082 X◦C

PB-TGA X = 0.435 0.489Condensate T = 335.38◦C 348.29◦C

t = 123.80min 127.44min

three curves for processes happening simultaneously, likely corresponding to hemicellulose

(∼300◦C), cellulose (main peak at ∼349◦C), and lignin (shoulder at ∼400◦C). In the PB-

TGA data there do not appear to be any skew or visually apparent convoluted curves.

This anomaly could be due to thermal lag effects within the PB-TGA sample that indicate

a radial variation in reaction progress. In essence, the reactions resulting in weight loss may

occur at different times (temporal offset) depending on radial position in the reactor. Material

close to the reactor wall may initially be at a higher temperature and experiences weight

loss reactions soonest, resulting in widening of the peak. In addition, the DTG peaks may

be reduced due to a greater propensity for more secondary reactions to occur, some of which

result in char formation.

The main peak of the cellulose temperature DTG in the PB-TGA approaches a relatively

much sharper peak than in the micro-TGA, a trend that is reversed in the pine pellet

experiments. The sharp point corresponds to the change in thermal gradient observed at

∼120◦C in the temperature difference curve of Figure 3.2.a. The decrease in the gradient,

Page 77: Multi-Phase Packed-Bed Thermogravimetric Analyzer

65

perhaps due to an exothermic reaction, causes the mean sample and furnace temperatures

to increase by more than 3◦Cmin

, thereby compressing the data points around the observed

peak. This thermal event does not appear to be present with pine pellets, so no sharp peak

is observed.

Another major difference between the micro-TGA and PB-TGA temperature DTG curves

is the presence of additional, albeit small, peaks in the PB-TGA solid decomposition and

condensate curves. The PB-TGA main peaks were characterized by a very dense and opaque

white vapor observed during experimental runs. The vapor cleared away almost completely

as the main peak ended. However, within 10 minutes and continuing until the end of the

experiment, a brown and slightly less opaque vapor flowed to the condenser. The beginning

of this second stage coincided with the second small peak in both the PB-TGA cellulose (at

about 462◦C and 151 minutes solid-phase) and pine pellet (at about 447◦C and 145 minutes)

experiments. These temperatures are close to the range of documented lignin decomposition

peaks Garcia-Nuñez et al. [14], Singh et al. [22], Raveendran et al. [21]. No well-defined

corresponding peak was observed in the micro-TGA cellulose data.

Although there is some weight loss during these second peaks in the PB-TGA, the peaks

are mostly characterized by an exothermic process that increases the center sample tempera-

ture faster than the 3◦Cmin

heating rate. These increases might not show up on the micro-TGA

because: (a) there was a low signal to noise ratio, (b) the reaction was not occurring, or (c)

a larger thermal gradient (more material) might have been needed to help emphasize the

temperature differences.

Despite these important differences, the main peak locations occur at close to the same

point. With cellulose, the micro-TGA DTG peak occurs at 336.97◦C (Table 3.4) and occurs

in the PB-TGA at 335.80◦C for the solid phase and 335.38◦C for the condensate. In contrast,

the micro-TGA DTG peak for pine pellets occurs ∼5◦C lower than for the PB-TGA. Both

the solid and condensate peaks in the PB-TGA occurred at similar points indicating a

minimal lag between when solid weight loss and condensate weight gain were recorded. Also,

Page 78: Multi-Phase Packed-Bed Thermogravimetric Analyzer

66

neither cellulose nor pine-pellet solid phase TG curves reach an asymptote by the end of the

experiment, indicating decomposition was still occurring. However, since the condensate in

the PB-TGA does appear to reach an asymptote, most of the decomposition products were

non-condensable gases at higher temperatures.

In comparison to these findings, Capart et al. [7] using a micro-crystalline cellulose pow-

der, found peak locations at 332◦C for cellulose at a heating rate of 3.2◦Cmin

.

3.3.3 Kinetics

Since the experiments were performed at one heating rate in dynamic mode, the kinetic

parameters were calculated using the nthorder model substituted into Equation 3.5. The

parameters A, E, and n were solved for from the regression coefficients using the micro-TGA

and PB-TGA data corresponding to X values between 0.1 and 0.8. The 0.8 value was picked

since the regression R2 values dropped significantly when using higher values.

This kinetic analysis method was selected since it uses common techniques and provides

a standard method of comparing the micro-TGA to the PB-TGA. However, this procedure

should be taken only as a proof of concept since at least a number of assumptions are violated,

most notably that there is one reaction occurring under a constant heating rate. While the

micro-TGA samples experience a close to constant heating rate, the convolved appearance of

the pine pellet DTG (Figure 3.5.a) likely violates the assumption of one reaction. Although

there is not an obvious convolution of multiple reactions in the PB-TGA DTG curves (with

the exception of the small peak between 400◦C and 500◦C), the thermal lag indicates the

sample is not experiencing a constant heating rate. In addition, the use of linearization in

regression likely imparts bias in the model coefficients.

Because of these problems the estimated kinetic parameters should be thought of as

model parameters that estimate characteristics of the system and serve as an example of a

need for developing new models that take these into account and for testing the scalability

of models that already exist.

Page 79: Multi-Phase Packed-Bed Thermogravimetric Analyzer

67

Table 3.5: Cellulose Kinetics and R2 values based on the X conversion (X = Wo−WWo−Wf

) data

between X=0.1 and X=0.8 (α=0.05, n=4).Device and Phase Value

A = 1.09±0.06×107 1s

micro-TGA E = 114.697±0.928 kJmol

Solid Phase n = 0.465±0.023R2 = 0.999±0.000A = 1.05±4.08×1014 1

s

PB-TGA E = 160.623±80.690 kJmol

Solid Phase n = 2.437±1.680R2 = 0.908±0.171

Table 3.6: Pine-Pellet Kinetics and R2 values based on the X conversion (X = Wo−WWo−Wf

) data

between X=0.1 and X=0.8 (α=0.05, n=4).Device and Phase Value

A = 76.19±45.581s

micro-TGA E = 58.522±2.494 kJmol

Solid Phase n = 0.264±0.047R2 = 0.995±0.001A = 5.95±22.25×107 1

s

PB-TGA E = 97.118±38.094 kJmol

Solid Phase n = 2.059±1.133R2 = 0.957±0.069

The kinetic analysis data are presented in Table 3.5 (cellulose) and Table 3.6 (pine pel-

lets). There were differences in errors, pre-exponential factors, reaction orders, and activation

energies. The errors associated with the mean micro-TGA kinetics were much smaller and

the R2 values higher than for the PB-TGA kinetics.

The pre-exponential values varied widely and were very sensitive to change in the data

range selected for analysis. In all the analyses, with the exception of the micro-TGA kinetics,

the error range in the mean pre-exponential factors was greater than the mean value. In the

Page 80: Multi-Phase Packed-Bed Thermogravimetric Analyzer

68

PB-TGA experiments, the mean pre-exponential factors are in fact not statistically different

from zero.

The reaction order of cellulose was 0.47 in the micro-TGA and 2.44 in the PB-TGA.

The reaction order of pine pellets was 0.26 in the micro-TGA and 2.06 in the PB-TGA. The

PB-TGA reaction order values were not statistically different and were inclusive of “2”.

In both pine pellet and cellulose experiments the activation energies were lower in the

micro-TGA than the PB-TGA with activation energies of both phases in agreement in the

PB-TGA.

Again, this kinetics analysis is a proof of concept for making rough comparisons between

the PB-TGA and the micro-TGA. Future researchers should consider using techniques that

take into account the highly non-ideal characteristics of the PB-TGA. One possible way

would be to use the idealized micro-TGA to determine intristic reaction characteristics and

use finite element analysis to develop scalable models of the PB-TGA. In addition, it may

also be possible to derive kinetic information from the condensate TGA data.

3.3.4 Conclusion

A commerically available micro-thermogravimetric analyzer (micro-TGA) was compared to a

newly developed packed-bed thermogravimetric analyzer (PB-TGA). While the micro-TGA

provides a highly idealized environment for collecting intrinsic pyrolysis data for solid-phase

decomposition, it is dissimilar in several ways to larger, real-world reactors that have much

larger sample sizes and in which collection of bio-oil is important. The PB-TGA has a much

larger sample size and a controllable carrier gas flow rate, allows collection of TG information

based on condensable material, and produces enough condensable material for downstream

analysis. This is similar to some larger real-world systems, but is far from idealized.

The PB-TGA solid phase decomposition system performed well, although signal noise

levels and errors in the means were significantly higher than in the micro-TGA. This could

be improved by using a more costly, off-the-shelf precision laboratory balance. Char yields

Page 81: Multi-Phase Packed-Bed Thermogravimetric Analyzer

69

in the PB-TGA were significantly higher than in the micro-TGA. This trend correlates to

increasing sample weight in data collected by Capart et al. [7], although this similarity may

be confounded by other effects, such as differences in gas residence time.

The condensate TGA system of the PB-TGA appeared to work well and provided a

significant amount of bio-oil that could easily be used for further analysis. The PB-TGA

condensate formation system actually yielded mean condensate weight errors smaller than

those for the solid weight measured in the micro-TGA. This difference was unexpected and

may indicate benefits of using a well designed condensate-based TGA system in other, per-

haps larger, systems where weighing the solid phase would be unfeasible.

Thermal lag in the PB-TGA was much greater than for the micro-TGA and may have

caused data compression in the PB-TGA temperature-dependent DTG peaks due to thermal

(exothermic or endothermic) events. This lag complicates understanding of the PB-TGA

pyrolysis and is usually not desirable in micro-TGA analysis. It is, however, a very real-

world effect present in larger sample sizes, such as might be found in larger scale reactors,

and may be usable as a type of differential thermal analysis (DTA) technique and help in

identifying reactions.

Kinetic analysis indicated significant differences in the kinetics calculated from the PB-

TGA and micro-TGA. Although this is a proof of concept operation this should provide

some motivation for using the PB-TGA as a potential tool to build more robust kinetics

models that can be scaled up to larger systems.

Although micro-TGA systems work well in striving for idealized solid-phase decompo-

sition characteristics, the new PB-TGA shows potential for providing information on more

realistic feedstock systems. For example, the PB-TGA could be modeled using finite element

modeling techniques that implement micro-TGA derived models. In addition, models could

be developed that incorporate condensate formation and real-time testing before a full-scale

system is designed. Although some refinement is desirable, such as decreasing noise, using

a fractionalized condensing system, and measuring temperatures at multiple sample loca-

Page 82: Multi-Phase Packed-Bed Thermogravimetric Analyzer

70

tions, this new system shows promise as a tool to be used in concert with existing analytical

micro-TGA systems.

Page 83: Multi-Phase Packed-Bed Thermogravimetric Analyzer

71

3.4 References

[1] Adebanjo, A. (2005). Production of Fuels and Chemicals from Biomass-Derived Oil and

Lard. Ph. D. thesis, University of Saskatchewan.

[2] Antal, M. J. J. and G. Morten (2003). The art, science, and technology of charcoal

production. Industrial and Engineering Chemical Research 42, 1619–1640.

[3] Antal, M. J. J., G. Varhegyi, and E. Jakab (1998). Cellulose pyrolysis kinetics: Revisited.

Industrial and Engineering Chemical Research 37, 1267–1275.

[4] Babu, B. and A. Chaurasia (2003). Modeling, simulation and estimation of optimum

parameters in pyrolysis of biomass. Energy Conversion and Management 44, 2135–2158.

[5] Babu, B. and A. Chaurasia (2004). Heat transfer and kinetics in the pyrolysis of shrinking

biomass particle. Chemical Engineering Science 59, 1999–2012.

[6] Bridgwater, A. and G. Peacocke (2000). Fast pyrolysis processes for biomass. Renewable

and Sustainable Energy Reviews 4, 1–73.

[7] Capart, R., L. Khezami, and A. K. Burnham (2004). Assessment of various kinetic models

for the pyrolysis of a microgranular cellulose. Thermochimica Acta 417, 79–89.

[8] Daugaard, D. E. (2003). Enthalpy for pyrolysis for several types of biomass. Energy and

Fuels 17, 934–939.

[9] Demirbas, A. (2004). Combustion characteristics of different biomass fuels. Progress in

Energy and Combustion Science 30, 219–230.

[10] Díaz, C. J. G. (2006). Understanding Biomass Pyrolysis Kinetics: Improved Model-

ing Based on Comprehensive Thermokinetic Analysis. Ph.d. dissertation, Departament

d’Enginyeria Química.

Page 84: Multi-Phase Packed-Bed Thermogravimetric Analyzer

72

[11] Encinar, J., F. Beltrán, A. Ramiro, and J. González (1996). Pyrolysis of two agricultural

residues: olive and grape bagasse. influence of particle size and temperature. Biomass and

Bioenergy 11 (5), 397–409.

[12] Fuwape, J. and S. O. Akindele (1997). Biomass yield and energy value of some fast-

growing multipurpose trees in Nigeria. Biomass and Bioenergy 12 (2), 101–106.

[13] Galvagno, S., S. Casu, M. Martino, E. Di Palma, and S. Portofino (2007). Thermal

and kinetic study of tyre waste pyrolysis via TG-FTIR-MS analysis. Journal of Thermal

Analysis and Calorimetry 88, 507–514.

[14] Garcia-Nuñez, J. A., M. Garcia-Perez, and K. Das (2008). Determination of kinetic

parameters of thermal degradation of palm oil mill by-products using thermogravimetric

analysis and differential scanning calorimetry. Transactions of the ASABE 51 (2).

[15] Guar, S. and T. B. Reed (1998). Thermal Data for Natural and Synthetic Fuels. New

York: Marcel Dekker, Inc.

[16] Guo, J. and A. Lua (2000). Kinetic study on pyrolysis of extracted oil palm fiber.

Journal of Thermal Analysis and Calorimetry 59, 763–774.

[17] Guo, J. and A. Lua (2001). Kinetic study on pyrolytic process of oil-palm solid waste

using two-step consecutive reaction model. Biomass and Bioenergy 20 (3), 223–233.

[18] Narayan, R. and M. J. J. Antal (1996). Thermal lag, fusion, and the compensation effect

during biomass pyrolysis. Industrial and Engineering Chemical Research 35, 1711–1721.

[19] Pindoria, R. V., A. Megaritis, R. C. Messenböck, D. R. Dugwell, and R. Kandlyoti

(1998). Short communications: Comparison of the pyrolysis and gasification of biomass:

effect of reacting gas atmosphere and pressure and eucalyptus wood. Fuel 77 (11), 1247–

1251.

Page 85: Multi-Phase Packed-Bed Thermogravimetric Analyzer

73

[20] Raveendran, K. and A. Ganesh (1996). Heating value of biomass and biomass pyrolysis

products. Fuel 75 (15), 1715–1720.

[21] Raveendran, K., A. Ganesh, and K. C. Khilar (1996). Pyrolysis characteristics of

biomass and biomass componenets. Fuel 75 (8), 987–998.

[22] Singh, K., K. Das, R. Mark, and W. John (2007, May 21-23). Determination of com-

position of cellulose and lignin mixtures using thermogravimetric analysis (TGA). In

Proceedings of ICON 5: 15th North American Waste-to-Energy Conference. ASME.

[23] Stenseng, M., A. Jensen, and K. Dam-Johansen (2001). Investigation of biomass py-

rolysis by thermogravimetric analysis and differential scanning calorimetry. Journal of

Analytical and Applied Pyrolysis 58-59, 765–780.

[24] Várhegyi, G., M. J. J. Antal, E. Jakab, and P. Szabó (1997). Kinetic modeling of

biomass pyrolysis. Journal of Analytical and Applied Pyrolysis 42, 73–87.

[25] Yang, H., R. Yan, H. Chen, D. H. Lee, and C. Zheng (2007). Characteristics of hemi-

cellulose, cellulose and lignin pyrolysis. Fuel 86 (12-13).

Page 86: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Chapter 4

Conclusions

The two objectives of this research were were: (a) to develop a packed-bed solid and liq-

uid phase thermogravimetric analyzer (PB-TGA), and (b) to compare the PB-TGA to a

commercially available analytical TGA (micro-TGA).

Completion of the first objective resulted in the successful development of a packed-bed

pyrolysis reactor coupled to a vapor condensation system. This entire system, when coupled

with the supporting electronics and computer based data acquisition system (DAQ), en-

abled collection of solid phase and condensate weight information, temperatures, and time.

Thus, thermogravimetric (TG), temperature-based differential thermogravimetric (temper-

ature DTG), and time-based differential thermogravimetric (time DTG) curves could be

developed for both solid phase decomposition and condensate formation during pyrolysis.

Transforming the PB-TGA from just a mechanical and electrical device into an analytical

tool required three major steps; calibration development (to transform the output of the

load cells into a solid-phase mass measurement), blank run correction (to correct for time

dependent effects), and testing with biomass.

Two potentially correctable issues were identified: (a) solid-phase weight measurement

accuracy and electrical noise susceptibility, and (b) failure of the pre-heater to maintain the

reactor entrance temperature at the set-point. A third issue, thermal lag between the furnace

and sample center temperatures, can either be seen as a shortcoming or used an example of

a real-world issue in large sample sizes.

74

Page 87: Multi-Phase Packed-Bed Thermogravimetric Analyzer

75

In the second objective, the completed and characterized PB-TGA was compared to a

micro-TGA by testing two feedstocks, cellulose powder and pine pellets, in each device in four

replications and in experimental conditions identical to those in the blank runs. Criteria used

for comparison included (a) product yields, (b) subjective comparisons of thermal profiles,

(c) temperature based DTG peak locations, and (d) kinetic parameters. Differences found

were higher char yields in the PB-TGA than in the micro-TGA, better accuracy in the micro-

TGA when comparing the solid-phase decomposition data, more symmetrical DTG peaks in

the PB-TGA, a clearly defined second reaction in the PB-TGA, and a significant thermal

lag between the furnace and the sample center temperature in the PB-TGA. Many of these

may be attributable to differences in vapor residence time and heat transfer limitations.

The PB-TGA overcomes several limitations of the micro-TGA system. First, the PB-

TGA allows TG analysis to be done on larger samples in a packed-bed. This means different

particle morphologies can be tested, different bed depths can be used, and carrier gas flow rate

can be controlled. Second, the PB-TGA allows TG data to be collected for the condensables,

enough of which is collected for later analysis. This is especially useful when doing research

on bio-oil formation. Existing analytical-scale TGA systems are widely used in material

characterization, but the PB-TGA provides a new tool that is especially catered to pyrolysis

energy research. The PB-TGA shows promise a tool that bridges the gap between highly

idealized micro-TGA systems, and larger real-world applications.

4.1 Recommendations

There are a number of improvements that should be made to the PB-TGA. The most im-

portant ones are to improve the the solid-phase weighing system by implementing a com-

merically available laboratory balance, and to either improve the pre-heater or remove it.

Recommended features that would add to functionality would be (a) a fractionalized con-

Page 88: Multi-Phase Packed-Bed Thermogravimetric Analyzer

76

densing system, (b) multiple sample temperature measurement locations, and (c) an easier

to use user interface with more streamlined operation.

The features that the PB-TGA do and could provide open the door to some interesting

new research questions:

1. What effect does feedstock morphology and bed depth have on PB-TGA kinetics,

including secondary reactions?

2. What effects do different operating parameters, such as heating rate, gas type, flow

rate, and condenser have on PB-TGA kinetics?

3. Why is there a clearly defined second reaction in the PB-TGA, but not the micro-TGA?

4. Can idealized micro-TGA based kinetic models be used to model the behavior of the

PB-TGA using a finite element modeling approach?

5. Can better kinetic models be developed that incorporate condensate formation TG

information?

6. How does the PB-TGA compare to different types of larger scale pyrolysis systems?

7. Can the condensate based TG method used in the PB-TGA be used to predict overall

reactor kinetics in larger scale systems where it is not feasible to weigh the solid phase?

There are undoubtedly other possible questions.This new PB-TGA device shows potential

for contributing to the understanding of pyrolysis and, more importantly, the overall issue

to trying develop more efficient and sustainable methods for managing energy.

Page 89: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Appendix A

Solid Phase load cell conditioning circuit

Figure A.1: Schematic diagram of the loadcell conditioning circuit contained within thebalance mechanism enclosure. The lower left circuit is fed by a linear 12VDC power supplyand provides the power for the op-amps and the load cell excitation source. The rest of thecircuit is used for summing, filtering, and amplifying the load cell output into a signal fedinto a computer-based data acquisition system (DAQ).

77

Page 90: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Appendix B

R-Code for developing the calibration model

78

Page 91: Multi-Phase Packed-Bed Thermogravimetric Analyzer

79

Listing B.1: (Calibration model development) R-statistics script for analyzing and creating

the cubic regression formula used to convert the load cell output (mV) into a temperature

corrected weight (g) in the PB-TGA solid phase decomposition system.

1 #0. Put your data . t x t f i l e ( see be low a b i t ) and t h i s s c r i p t inthe same f o l d e r ) .

2 #1. Open R3 #2. check working d i r e c t o r y wi th getwd () . This shou ld be the one

wi th your data4 #3. change working d i r e c t o r y to the one wi th your data wi th setwd

( d i r ) where d i r=your d i r e c t o r y path5 #3.5 OR i f us ing the GUI, change working d i r wi th the f i l e op t ion

.6 #4. invoke s c r i p t wi th source ( ’ c a l i b r a t i o n . r ’ )7 #5. c a l l f unc t i on s us ing ana lyze ( )89 #This f unc t i on s i s des i gned to

10 #The inputed f i l e i s " data . t x t " data f i l e wi th the header and11 #foo t e r removed .12 #13 ana lyze<−function ( )14 {1516 close . screen ( a l l=TRUE)17 dev . of f ( )18 rm( l i s t=l s ( a l l=TRUE) ) #I n i t i a l i z e system : Remove a l l o b j e c t s19 rm( l i s t=l s ( ) )20 graphics . of f ( ) #Shutdown a l l g raph i c s d e v i c e s to c l e a r

memory2122 #Invoke L i b r a r i e s23 l ibrary (MASS) #For doing box cox p l o t2425 #Read in data . Format o f data must be t ha t which i s s p e c i f i e d in

the header o f t h i s s c r i p t .26 data=read . table ( ’ d a t a f u l l . txt ’ , header=TRUE)27 fdata=data . frame (data )28 bdata=data . frame (data )2930 #Determine index va l u e s f o r we igh t d i f f e r e n t l e v e l s31 weightL1 <− which( fdata$actua lwe ight == 0)32 weightL2 <− which( fdata$actua lwe ight == 169 .3 )33 weightL3 <− which( fdata$actua lwe ight == 332 .2 )

Page 92: Multi-Phase Packed-Bed Thermogravimetric Analyzer

80

34 weightL4 <− which( fdata$actua lwe ight == 503 .1 )35 tempL1 <− which( fdata$settemp == 25 . 6 )36 tempL2 <− which( fdata$settemp == 26)37 tempL3 <− which( fdata$settemp == 26 . 4 )3839 pdf ( "DataPlotLcVSTemp . pdf " , width=12, he ight =12)40 sp l i t . screen (c ( 2 , 2 ) )4142 #P lo t t i n g data f o r v i s u a l i n s p e c t i on#43 #Visua l Ana lys i s o f raw data at weightL1 ( l c = f ( weightL1 and

coo l e r temperature ) )44 screen (1 )45 plot ( fdata$ coo le r tempcat [ weightL1 ] , fdata$ l coutput [ weightL1 ] , x lab="

Cooler Temp ( deg C) " , ylab="LC (mV)" ,main=" l c vs . temp atWeightL1" , pch=20)

4647 #Visua l Ana lys i s o f raw data at weightL2 ( l c = f ( weightL2 and

coo l e r temperature ) )48 screen (2 )49 plot ( fdata$ coo le r tempcat [ weightL2 ] , fdata$ l coutput [ weightL2 ] , x lab="

Cooler Temp ( deg C) " , ylab="LC (mV)" ,main=" l c vs . temp atWeightL2" , pch=20)

5051 #Visua l Ana lys i s o f raw data at weightL3 ( l c = f ( weightL3 and

coo l e r temperature ) )52 screen (3 )53 plot ( fdata$ coo le r tempcat [ weightL3 ] , fdata$ l coutput [ weightL3 ] , x lab="

Cooler Temp ( deg C) " , ylab="LC (mV)" ,main=" l c vs . temp atWeightL3" , pch=20)

5455 #Visua l Ana lys i s o f raw data at weightL4 ( l c = f ( weightL4 and

coo l e r temperature ) )56 screen (4 )57 plot ( fdata$ coo le r tempcat [ weightL4 ] , fdata$ l coutput [ weightL4 ] , x lab="

Cooler Temp ( deg C) " , ylab="LC (mV)" ,main=" l c vs . temp atWeightL4" , pch=20)

58 identify ( fdata$ coo le r tempcat [ weightL4 ] , fdata$ l coutput [ weightL4 ] )59 close . screen ( a l l=TRUE)60 dev . of f ( )61 graphics . of f ( )6263 pdf ( "DataPlotLcVSWeight . pdf " , width=12, he ight =12)64 sp l i t . screen (c ( 2 , 2 ) )65

Page 93: Multi-Phase Packed-Bed Thermogravimetric Analyzer

81

66 #Visua l Ana lys i s o f raw data at tempL1 ( l c = f ( we igh t and coo l e rtemperatureL1 ) )

67 screen (1 )68 plot ( fdata$actua lwe ight [ tempL1 ] , fdata$ l coutput [ tempL1 ] , xlab="

Weight ( grams ) " , ylab="LC (mV)" ,main=" l c vs . Weight at tempL1" ,pch=20)

6970 #Visua l Ana lys i s o f raw data at tempL2 ( l c = f ( we igh t and coo l e r

temperatureL2 ) )71 screen (2 )72 plot ( fdata$actua lwe ight [ tempL2 ] , fdata$ l coutput [ tempL2 ] , xlab="

Weight ( grams ) " , ylab="LC (mV)" ,main=" l c vs . Weight at tempL2" ,pch=20)

7374 #Visua l Ana lys i s o f raw data at tempL3 ( l c = f ( we igh t and coo l e r

temperatureL3 ) )75 screen (3 )76 plot ( fdata$actua lwe ight [ tempL3 ] , fdata$ l coutput [ tempL3 ] , xlab="

Weight ( grams ) " , ylab="LC (mV)" ,main=" l c vs . Weight at tempL3" ,pch=20)

77 close . screen ( a l l=TRUE)78 dev . of f ( )79 graphics . of f ( )808182 #Regress ion o f a l l data83 g=lm( fdata$ l coutput ~ fdata$actua lwe ight + fdata$actua lwe ight2 +

fdata$actua lwe ight3 + fdata$ coo le r tempcat + fdata$coo lertempcat2 + fdata$coo lertempcat3 )

84 gs=summary( g )85 gp=print ( gs , d i g i t s =6)86 #Ca l cu l a t e the con f idence i n t e r v a l s o f the two main indep .

v a r i a b l e s87 gsc=con f i n t ( g , l e v e l =0.95)88 print ( gsc , d i g i t s =6)89 #Write the r e s i d u a l s to a f i l e f o r f u r t h e r p roce s s ing i f d e s i r ed90 #wr i t e ( r e s i dua l s , f i l e ="r e s i d u a l s . t x t " , sep="\n" , append=FALSE)9192 pdf ( " Regre s s i onUnStab i l i z ed . pdf " , width=12, he ight =12)93 sp l i t . screen (c ( 3 , 2 ) )949596 #Test f o r o u t l i e r s and l e a v e rag e po in t s

Page 94: Multi-Phase Packed-Bed Thermogravimetric Analyzer

82

97 #Cooks d i s t ance (measure o f i n f l u en c e which t a k e s in t oaccount the r e s i d u a l e f f e c t and the l e v e r a g e ) − Removel a r g e s t one and redo r e g r e s s i on . Ingenera l , v a l u e sg r ea t e r than 1 may meri t c l o s e r examination .

98 screen (1 )99 cook=cooks . distance ( g )

100 plot ( cook , ylab="Cooks Distance " , main="Cooks Distance " , pch=20)

101 segments ( 1 : 4947 , 0 , 1 : 4 947 , cook ) #draw drop l i n e s from datato x−ax i s

102 # i d e n t i f y ( cook ) #c l i c k po in t to i d e n t i f y time o f po in t103104 #Plot o f l e v e r a g e s (Look f o r h igh l e a v e rag e po in t s )105 #Consider po in t s t ha t are (2∗parameters )/(number o f

o b s e r va t i on ) . Observat ion va l u e s above t h e s e po in t sshou ld be i n v e s t i g a t e s .

106 screen (2 )107 a=model . matrix ( g )108 l ev=hat ( a )109 plot ( lev , ylab="Leverages " ,main=" Index p l o t o f Leverages " ,

pch=20)110 abline (h=2∗6/4947) #This i s the c u t o f f v a l u e s111 # i d e n t i f y ( l e v )112113 #Studen t i z ed Res idua l s ( Ou t l i e r t e s t )114 #Consider po in t s l a r g e r than 2 in a b s o l u t e va lue as

p o s s i b l e o u t l i e r s .115 screen (3 )116 stud=g$ r e s/ ( gs$ s i g∗sqrt(1− l e v ) )117 plot ( stud , ylab=" Student ized Res idua l s " ,main=" Student ized

Res idua l s " , pch=20)118 abline (h=0)119 abline (h=2)120 abline (h=−2)121 # i d e n t i f y ( s tud )122123 #Write the s t u d en t i z e d r e s i d u a l s to a f i l e i f d e s i r ed124 #s r e s i d u a l s = stud125 #wr i t e ( s r e s i d ua l s , f i l e ="s r e s i d u a l s . t x t " , sep="\n" , append=

FALSE)126127 #Q−Q used to i l l u s t r a t e d e v i a t i on from Normal i l t y128 screen (4 )

Page 95: Multi-Phase Packed-Bed Thermogravimetric Analyzer

83

129 qqnorm( g$ res , main="Untransformed Normal Q−Q" , xlab="Theo r e t i c a l Quant i l e s " , ylab=" Res idua l s " , pch=20)

130 abline (h=2∗6/4947)131 qqline ( g$ r e s )132133 #Box−Cox Use to i d e n t i f y lambda va lue at maximum log−

l i k e l y h o o d . Note t ha t the response v a r i a b l e must bep o s i t i v e , which i s a problem when the we igh t i s zero .I w i l l redo the r e g r e s s i on ana l y s i s and add a s e t va lueto the response v a r i a b l e s .

134 screen (5 )135 bdata$ l coutput=fdata$ l coutput∗(−1)136 g=lm( bdata$ l coutput ~ fdata$actua lwe ight + fdata$

actua lwe ight2 + fdata$actua lwe ight3 + fdata$coo le r tempcat + fdata$coo lertempcat2 + fdata$coo lertempcat3 )

137 boxcox (g , p l o t i t=T, lambda=seq (−5 ,5 ,by=0.1) )138 close . screen ( a l l=TRUE)139 dev . of f ( )140 graphics . of f ( )141142 #####Changing the r e g r e s s i on equat ion . STABILIZING FOR VARIANCE! !143 # pdf (" second boxcox . pd f ")144 g=lm ( ( ( bdata$ l coutput ^(−2)−1)/(−2) ) ~ fdata$actua lwe ight +

fdata$actua lwe ight2 + fdata$actua lwe ight3 + fdata$coo le r tempcat + fdata$coo lertempcat2 + fdata$coo lertempcat3 )

145 boxcox (g , p l o t i t=T, lambda=seq (−2 ,3 ,by=0.1) )146 # dev . o f f ( )147148 gs=summary( g )149 gp=print ( gs , d i g i t s =6)150 #Ca l cu l a t e the con f idence i n t e r v a l s o f the two main indep .

v a r i a b l e s151 gsc=con f i n t ( g , l e v e l =0.95)152 print ( gsc , d i g i t s =6)153154 #Write the s t a b i l i z e d r e s i d u a l s out to a f i l e i f d e s i r e d155 #re s i d u a l s = stud156 #wr i t e ( s r e s i d ua l s , f i l e ="s t a b r e s i d u a l s . t x t " , sep="\n" , append=FALSE)157158159 pdf ( " Reg r e s s i o nS t ab i l i z e d . pdf " , width=12, he ight =12)160 sp l i t . screen (c ( 3 , 2 ) )

Page 96: Multi-Phase Packed-Bed Thermogravimetric Analyzer

84

161162163 #Test f o r o u t l i e r s and l e a v e rag e po in t s164 #Cooks d i s t ance (measure o f i n f l u en c e which t a k e s in t o

account the r e s i d u a l e f f e c t and the l e v e r a g e ) − Removel a r g e s t one and redo r e g r e s s i on . Ingenera l , v a l u e sg r ea t e r than 1 may meri t c l o s e r examination .

165 screen (1 )166 cook=cooks . distance ( g )167 plot ( cook , ylab="Cooks Distance " , main="Cooks Distance " , pch

=20)168 segments ( 1 : 4947 , 0 , 1 : 4 947 , cook ) #draw drop l i n e s from data

to x−ax i s169 # i d e n t i f y ( cook ) #c l i c k po in t to i d e n t i f y time o f po in t170171 #Plot o f l e v e r a g e s (Look f o r h igh l e a v e rag e po in t s )172 #Consider po in t s t ha t are (2∗parameters )/(number o f

o b s e r va t i on ) . Observat ion va l u e s above t h e s e po in t sshou ld be i n v e s t i g a t e s .

173 screen (2 )174 a=model . matrix ( g )175 l ev=hat ( a )176 plot ( lev , ylab="Leverages " ,main=" Index p l o t o f Leverages " ,

pch=20)177 abline (h=2∗6/4947) #This i s the c u t o f f v a l u e s178 # i d e n t i f y ( l e v )179180 #Studen t i z ed Res idua l s ( Ou t l i e r t e s t )181 #Consider po in t s l a r g e r than 2 in a b s o l u t e va lue as

p o s s i b l e o u t l i e r s .182 screen (3 )183 stud=g$ r e s/ ( gs$ s i g∗sqrt(1− l e v ) )184 plot ( stud , ylab=" Student ized Res idua l s " ,main=" Student ized

Res idua l s " , pch=20)185 abline (h=0)186 abline (h=2)187 abline (h=−2)188 # i d e n t i f y ( s tud )189190 #Write the s t u d en t i z e d r e s i d u a l s to a f i l e i f d e s i r ed191 #s r e s i d u a l s = stud192 #wr i t e ( s r e s i d ua l s , f i l e ="s r e s i d u a l s . t x t " , sep="\n" , append=

FALSE)193

Page 97: Multi-Phase Packed-Bed Thermogravimetric Analyzer

85

194 #Q−Q used to i l l u s t r a t e d e v i a t i on from Normal i l t y195 screen (4 )196 qqnorm( g$ res , main="Untransformed Normal Q−Q" , xlab="

Theo r e t i c a l Quant i l e s " , ylab=" Res idua l s " , pch=20)197 abline (h=2∗6/4947)198 qqline ( g$ r e s )199 close . screen ( a l l=TRUE)200 dev . of f ( )201 graphics . of f ( )202203 gs=summary( g )204 gp=print ( gs , d i g i t s =6)205 #Ca l cu l a t e the con f idence i n t e r v a l s o f the two main indep .

v a r i a b l e s206 gsc=con f i n t ( g , l e v e l =0.95)207 print ( gsc , d i g i t s =6)208209 }

Page 98: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Appendix C

R-Code for kinetics analysis

86

Page 99: Multi-Phase Packed-Bed Thermogravimetric Analyzer

87

Listing C.1: (micro-TGA and PB-TGA kinetics analysis.) R-statistics script for performing

kinetics analysis on both the micro-TGA and the PB-TGA solid phase and condensate data.

The returned values are activation energy, frequency factor, reaction order, and goodness of

fit.

1 #Erik J a r r e t t2 #200808283 #Ti t l e : TGA Batch Kine t i c s Process ing4 #I n s t a l l s y s t em f i t , caTools , Matrix56 ##1s t : Determine the i n i t i a l and f i n a l convers ion r a t i o boundar ies

from a spreadshee t7 ##2nd : Determine the averag ing window necessary f o r 1 minute ! Odd

number .8 ##3rd : Do the f o l l ow i n g .9 #Invoke : source ( ’ prog2 . r ’ )

10 #Run : k i n e t i c s ( ’ f i l ename i . e . d a t a p c e l l . t x t ’ , convers ion r a t i o ( i . e. 0 .05) beg in ing boundary , convers ion r a t i o ( i . e . 0 .75) f i n a lboundary , averag ing window ( the number o f rows covered by 1minute , 0 f o r s o l i d or 1 f o r l i q u i d . 7 po in t s f o r mTGA, 31 f o rPB−TGA) )

11 #Returned va lue i s a l i s t o f rep1 , rep2 , rep3 , and rep4 IFuncommented

1213 ##Do eve r y t h in g14 k i n e t i c s<−function ( f i l ename , convstart , convend ,window , phase ) {1516 #Clean the data and re turn a l i s t wi th 4 s e t s o f data , one f o r

each rep .17 print ( "Cleaning : I n i t a t i n g " )18 datac leaned<−datac ( f i l ename ,window , phase )19 print ( "Cleaning : F in i shed " )2021 print ( "Evaluat ion : I n i t i a t i n g " )2223 for ( i in 1 : 4 ) {24 da t a c l e an ed s i ng l e<−as . data . frame ( datac leaned [ i ] , ncol=5) #

Convert l i s t o f one rep in da tac l eaned to a dataframe25 da t a c l e an ed s i ng l e<−as . matrix ( da t a c l e an ed s i ng l e )26 i f ( i==1) rep1<−datae ( da tac l e aneds ing l e , i , convstart ,

convend ,window , phase )27 i f ( i==2) rep2<−datae ( da tac l e aneds ing l e , i , convstart ,

convend ,window , phase )

Page 100: Multi-Phase Packed-Bed Thermogravimetric Analyzer

88

28 i f ( i==3) rep3<−datae ( da tac l e aneds ing l e , i , convstart ,convend ,window , phase )

29 i f ( i==4) rep4<−datae ( da tac l e aneds ing l e , i , convstart ,convend ,window , phase )

30 print (c ( "Evaluat ion : Rep" , i , " eva luated " ) ,quote=FALSE)3132 }3334 print ( " F in i sh ing : Model summaries" )35 print ( "Rep1" )36 print (summary( rep1 ) )37 print ( "Rep2" )38 print (summary( rep2 ) )39 print ( "Rep3" )40 print (summary( rep3 ) )41 print ( "Rep4" )42 print (summary( rep4 ) )4344 print ( " F in i sh ing : Model averages " )4546 f r e q<−array ( 1 : 4 )47 f r e q [ 1 ]<−0 .05∗exp( rep1$ c o e f f [ 1 ] ) #0.05 deg/ sec i s 3

deg/min48 f r e q [ 2 ]<−0 .05∗exp( rep2$ c o e f f [ 1 ] )49 f r e q [ 3 ]<−0 .05∗exp( rep3$ c o e f f [ 1 ] )50 f r e q [ 4 ]<−0 .05∗exp( rep4$ c o e f f [ 1 ] )51 freqmean<−mean( f r e q [ 1 : 4 ] )52 f r eq sd<−sd ( f r e q [ 1 : 4 ] )5354 order<−array ( 1 : 4 )55 order [ 1 ]<−rep1$ c o e f f [ 2 ]56 order [ 2 ]<−rep2$ c o e f f [ 2 ]57 order [ 3 ]<−rep3$ c o e f f [ 2 ]58 order [ 4 ]<−rep4$ c o e f f [ 2 ]59 ordermean<−mean(order [ 1 : 4 ] )60 ordersd<−sd (order [ 1 : 4 ] )6162 energy<−array ( 1 : 4 )63 energy [ 1 ]<−(−1)∗rep1$ c o e f f [ 3 ]64 energy [ 2 ]<−(−1)∗rep2$ c o e f f [ 3 ]65 energy [ 3 ]<−(−1)∗rep3$ c o e f f [ 3 ]66 energy [ 4 ]<−(−1)∗rep4$ c o e f f [ 3 ]67 energymean<−(mean( energy [ 1 : 4 ] ) )68 energysd<−sd ( energy [ 1 : 4 ] )

Page 101: Multi-Phase Packed-Bed Thermogravimetric Analyzer

89

6970 rep1<−summary( rep1 )71 rep2<−summary( rep2 )72 rep3<−summary( rep3 )73 rep4<−summary( rep4 )7475 r sq r<−array ( 1 : 4 )76 r sq r [ 1 ]<−rep1$ o l s . r . squared77 r sq r [ 2 ]<−rep2$ o l s . r . squared78 r sq r [ 3 ]<−rep3$ o l s . r . squared79 r sq r [ 4 ]<−rep4$ o l s . r . squared80 rsqrmean<−(mean( r s q r [ 1 : 4 ] ) )81 r sq r sd<−sd ( r s q r [ 1 : 4 ] )8283 print ( " K ine t i c s Ana lys i s : " )84 print (c ( "Frequency f a c t o r=" , freqmean , " sd=" , f r eq sd ) ,quote=

FALSE)85 print (c ( "Reaction Order=" , ordermean , " sd=" , ordersd ) ,quote=

FALSE)86 print (c ( " Act ivat ion Energy=" , energymean , " sd=" , energysd ) ,

quote=FALSE)87 print (c ( "R Squared value=" , rsqrmean , " sd=" , r sq r sd ) ,quote=

FALSE)88 print (c ( "Filename= " , f i l ename ) ,quote=FALSE)89 print (c ( "Window= " ,window) ,quote=FALSE)90 print (c ( "Phase= " , phase ) ,quote=FALSE)9192 #models<− l i s t ( rep1 , rep2 , rep3 , rep4 )93 #return ( models )9495 }9697 ##Func : Import , c l ean the data98 datac<−function ( f i l ename ,window , phase )99 {

100 ##Clear v a r i a b l e s101 #rm( l i s t=l s ( a l l=TRUE) ) #I n i t i a l i z e system : Remove a l l

o b j e c t s102 graphics . of f ( ) #Shutdown a l l g raph i c s d e v i c e s to

c l e a r memory103104 l ibrary ( caTools ) #Has the t o o l f o r doing the moving

average . Test t h i s .105

Page 102: Multi-Phase Packed-Bed Thermogravimetric Analyzer

90

106 #Read data in from data . t x t . This must have headers andf o o t e r s removed

107 rawdata=read . table ( f i l e=fi lename , header=TRUE)108 fdata=data . frame ( rawdata )109110 #Bui ld matrix o f data111 mdata=as . matrix ( fdata )112 dims=dim(mdata )113114 i f ( phase==1){ #Separate matrix in t o d i f f e r e n t reps FOR

Liqu id115 data1<−as . matrix (mdata [ , c ( 1 , 5 , 9 , 13 , 21 ) ] )116 data2<−as . matrix (mdata [ , c (2 , 6 , 10 , 14 , 22 ) ] )117 data3<−as . matrix (mdata [ , c (3 , 7 , 11 , 15 , 23 ) ] )118 data4<−as . matrix (mdata [ , c (4 , 8 , 12 , 16 , 24 ) ] )119 }120 else {121 #Separate matrix in t o d i f f e r e n t reps FOR SOLID122 data1<−as . matrix (mdata [ , c ( 1 , 5 , 9 , 13 , 17 ) ] )123 data2<−as . matrix (mdata [ , c (2 , 6 , 10 , 14 , 18 ) ] )124 data3<−as . matrix (mdata [ , c (3 , 7 , 11 , 15 , 19 ) ] )125 data4<−as . matrix (mdata [ , c (4 , 8 , 12 , 16 , 20 ) ] )126 }127128 print ( "Cleaning : Data s p l i t " )129130 #Average a l l data131 avewin<−window #Averaging window . Put as argument f o r

datac l a t e r .132 data1<−datac . ave ( data1 , avewin )133 data2<−datac . ave ( data2 , avewin )134 data3<−datac . ave ( data3 , avewin )135 data4<−datac . ave ( data4 , avewin )136137 print ( "Cleaning : Data Averaged" )138139 #Dele te d u p l i c a t e weights , r e f temperatures , and sample

temperatures .140 for ( j in 5 : 3 )141 {142 data1<−datac . dup ( data1 , j )143 data2<−datac . dup ( data2 , j )144 data3<−datac . dup ( data3 , j )145 data4<−datac . dup ( data4 , j )

Page 103: Multi-Phase Packed-Bed Thermogravimetric Analyzer

91

146 }147148 print ( "Cleaning : Dupl i cate weight removed" )149150 datatemp<−l i s t ( data1 , data2 , data3 , data4 ) #Returns data as a l i s t .

Use a<−datac ( ) . a [1]= data1151 return ( datatemp )152 }153154 ##########################Cleaning f unc t i on s155156 ##Average data po in t s157 datac . ave<−function (data , avewin )158 {159 for ( i in 1 : 5 ) {160 data [ , i ]<−runmean (data [ , i ] , avewin , a l g=" exact " , endru le="NA"

)161 }162163 # dims<−dim( data )164 dims<−dim(data )165 rows<−dims [ 1 ] #

I n i t i a l i z e rows to maximum166 i=1167 while ( i<=rows ) {168 i f ( i s .na(data [ i , 5 ] ) ) {169 data<−data[−c ( i ) , ]170 rows<−rows−1 #Reduce

number o f rows l e f t by 1171 }172 else173 i<−i+1174 }175 return (data )176 }177178 ##Dele te d up l i c a t e e n t r i e s179 datac . dup<−function (data , column )180 {181 dims<−dim(data )182 rows<−dims [ 1 ] #

I n i t i a l i z e rows to maximum183 i=1 #

I n i t i a l i z e index counter

Page 104: Multi-Phase Packed-Bed Thermogravimetric Analyzer

92

184 while ( i<rows ) #Continueu n t i l rows run out

185 {186 i f (data [ i , column]==data [ i +1,column ] ) #Compare

f o r consecu t i v e d up l i c a t e numbers187 {188 data<−data[−c ( i +1) , ] #Remove

dup l i c a t e data from datarep1189 rows<−rows−1 #Reduce

number o f rows l e f t by 1190 i f ( i >1) i<−i−1 #Make

index repea t l a s t p o s i t i o n i f g r ea t e rthan 1

191 }192 else193 i<−i+1 #Increment

row194 }195 return (data )196 }197198 ##Dele te po in t s t h a t go up !199 #I may or may not need to do t h i s .200201 ###########################Evaluat ion f unc t i on s202 datae<−function (data , rep , convstart , convend ,window , phase ) {203204 dims<−dim(data )205206 #Convert temperatures to k e l v i n and minutes to seconds207 data [ , 3 ]<−data [ , 3 ]+273 .15208 data [ , 4 ]<−data [ , 4 ]+273 .15209 data [ , 1 ]<−data [ , 1 ] ∗60210211 #Average the very end o f the we igh t . #3 i s 4 to average .

Make t h i s changab le ! The output i s in in the o r i g i n a lwe igh t input

212 #window=31213 wf ina l<−mean(data [ ( dims [1]−window) : dims [ 1 ] , 5 ] ) ∗data [ 1 , 2 ]214 databig<−cbind (data , w f i na l )215 #pr in t ( w f i na l )216217 i f ( phase==1){218 #Find X fo r LIQUID PHASE

Page 105: Multi-Phase Packed-Bed Thermogravimetric Analyzer

93

219 xconv<−(0−databig [ , 5 ] ∗databig [ , 2 ] ) /(0−databig [ , 6 ] )220 databig<−cbind ( databig , xconv )221 }222 else {223 #Find X fo r SOLID PHASE224 xconv<−( databig [ ,2 ] − databig [ , 5 ] ∗databig [ , 2 ] ) / (

databig [ ,2 ] − databig [ , 6 ] )225 databig<−cbind ( databig , xconv )226 }227228 #Find X average f o r i d e n t i f y i n g boundary ies . Make

changab le ! !229 #Change averag ing window by 20230 meanwin<−window∗5231 xconvav<−runmean ( databig [ , 7 ] , meanwin , a l g=" exact " , endru le="

NA" )232 databig<−cbind ( databig , xconvav )233234 #Find average temperature235 tave<−( databig [ , 3 ]+ databig [ , 4 ] ) /2236 databig<−cbind ( databig , tave )237238 #Find dxdT . THATS dX over dTEMPERATURE in Kelv in ! ! Use

only odd window s i z e s .239 #window<−5 #Window s i z e i s the t o t a l span o f data

points , i n c l u d i n g the cen te r . Thus , (window−1)/2 i sthe number o f po in t s on each s i d e

240 dxdt<−array ( "NA" , dims [ 1 ] ) #Create array to f i l l241 for ( i in ( (window−1)/2+1) : ( dims [1 ] −(window−1)/2) ) { #

Go from center beg in ing to cen te r end .242 #The ba s i c equat ion form i s dx/dt=(Xconv+(win−1)/2

− Xconv−(win−1)/2)/(Time+(win−1)/2 − Time−(win−1)/2)

243 dxdt [ i ]<−( databig [ ( i +(window−1)/2) ,7]− databig [ ( i −(window−1)/2) , 7 ] )/ ( databig [ ( i +(window−1)/2) ,9]−databig [ ( i −(window−1)/2) , 9 ] )

244 }245 dxdt<−as . numeric ( dxdt )246 databig<−cbind ( databig , dxdt )247248 #Remove NA dxdt249 databig<−na . omit ( databig )250251 #Se l e c t data in boundary

Page 106: Multi-Phase Packed-Bed Thermogravimetric Analyzer

94

252 databig<−subset ( databig , databig [ , 8 ] > convs ta r t )253 databig<−subset ( databig , databig [ , 8 ] < convend )254255 ##Remove nega t i v e numbers HERE256 #datab i g<−da tab i g [−c ( which ( da t a b i g [ ,7]<=0) ) , ]257 databig<−subset ( databig , databig [ , 7 ] > 0)258 databig<−subset ( databig , databig [ , 1 0 ] > 0)259260 #Find ln o f dxdT261 l o g s t u f f<−as . numeric ( databig [ , 1 0 ] )262 databig<−cbind ( databig , " lndxdr "=log ( l o g s t u f f , exp (1 ) ) )263264 #Find 1/RT265 gasconst<−8.314472 #Units J/(K∗mol ) . Remember J=1kg∗

m^2/s^2266 onert<−1/ ( gasconst∗as . numeric ( databig [ , 9 ] ) )267 databig<−cbind ( databig , onert )268269 ##Regress ion s e c t i on270271 #Get t ing the model r e l a t i o n s h i p put t o g e t h e r272 databig<−data . frame ( databig )273 l ibrary ( s y s t em f i t ) #A system fo r f i t t i n g systems o f

equat ions , but i t can a l s o do 1 and cons t ra in ther e l a t i o n s h i p s

274 a<−databig$ lndxdr #a=response v a r i a b l e275 b<−log(1−databig$xconv , exp (1 ) ) #b=second term276 c<−databig$onert #C=th i r d term277 formula<−a ~ b + c #A standard R formula entry278279 #Reaction order r e s t r i c t i o n . I f you don ’ t use i t , j u s t

l e a v e i t .280 Rrest r <− matrix ( 0 , 1 , 3 ) #Rows are coe f f , and columns are

c o e f f L a b e l and Re s t r i c t i on . See s y s t em f i t PAPER/Ar t i c l e f o r he l p .

281 Rrest r [ 1 , 2 ]<−1 #Mul t i p l y by number to a c t i v a t e t ha tc o e f f i c i e n t . Set to 1 f o r f i x i n g r eac t i on order . Setto 0 f o r not .

282 q r e s t r<−1 #The sum tha t the a c t i v a t e d c o e f f i c i e n t smust add up to . Set to 1 to f i x r eac t i on order to 1

283284 #Model : Pick between cons t ra ined r eac t i on order and

unconstra ined r eac t i on order

Page 107: Multi-Phase Packed-Bed Thermogravimetric Analyzer

95

285 #model<−s y s t em f i t ( formula , r e s t r i c t . matrix=Rrestr , r e s t r i c t .rhs=q r e s t r ) #This one f o r cons t ra ined r eac t i on order

286 model<−s y s t em f i t ( formula ) #This one f o runconstra ined r eac t i on order

287 #b1 i s i n t e r c ep t , b2 i s l o g b l a b l a , and b3 i s oner t . Wewant to r e s t r i c t b2=1

288289 residuals<−residuals . s y s t em f i t (model)290291 #x11 ()292 #p l o t ( ( da t a b i g$ tave −273.15) , da t a b i g$dxdt )293 #x11 ()294 #p l o t ( ( da t a b i g$ tave −273.15) , r e s i d u a l s$eq1 )295296 return (model)297298 }

Page 108: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Appendix D

Example Procedures for a blank run1

1Procedures: 20080328_1 Baseline Experiment 4 of 5 (2 April 2008 Rev.1 by Erik Jarrett)

96

Page 109: Multi-Phase Packed-Bed Thermogravimetric Analyzer

97

D.0.1 Purpose

The purpose of this experiment is to perform five baseline experiments. This should provide

the information necessary to do a baseline analysis and correction.

D.0.2 Summary

The reactor will have as close to 0 grams of material inside. Use a flow rate of 1/4, 1/4 L/min

for both carrier and quench. The heating rate is 3 deg C / min and does not change from

ambient temperature.

D.1 Customized Experimental Section

D.1.1 Independent Variables

Sample and Drying

1. Biomass type: None

2. Dry Biomass target weight: 0 g

3. Biomass drying temperature: NA

4. Biomass drying furnace: NA

Experiment

1. Carrier and quench gas: Nitrogen

2. Bubble trap volume (keep same flask): 300 ml

3. Room temperature setpoint: 73 deg F

4. Load Cell Instrument cooler setpoint: Variable around 26 deg C

Page 110: Multi-Phase Packed-Bed Thermogravimetric Analyzer

98

5. Standby flow rate: Carrier flow rate: 0.05 L/min, Quench flow rate 0.05 L/min or OFF

and sealed

6. Experiment step flow rates: Carrier flow rate: 0.25 L /min; Quench flow rate: 0.25

L/min

7. Experiment step chiller temperature: 20 deg C only when duing baseline run.

8. Experiment PID manual recording interval: 20 minutes minimum during baseline run

Experiment temperature profile

1. Start at room temperature

2. Ramp at 3 deg C / min from 20 deg C to 600 deg C (3:13.2 hrs)

3. Soak 45 min (from 3:13.2 hr to 3:58.2 hr).

4. End program and return to room temperature

Virtual Instrument and Signal Processing

1. Virtual Instrument Name: AAFinalProject9.vi

2. File name for testing: test.txt

3. File name for run data: 20080328_1experiment.txt (many files)

4. Calibration from file: NO CALIBRATION

5. Calibration equation: NO CALIBRATION THAT IS MEANINGFUL YET

D.1.2 Dependent

Drying Period

1. Initial drying pan weight: NA g

Page 111: Multi-Phase Packed-Bed Thermogravimetric Analyzer

99

2. Initial sample weight (in tared pan; target 280g): 0 g

3. Date and time of sample insert: NA pm

4. Final Total sample weight (final + drying pan): NA g

5. CALC Final sample weight (Final Total - Initial drying pan): NA g

6. CALC Moisture loss: (1-FinalSampleWeight/InitialSampleWeight)*100: NA %

7. Date and time of sample removal: NA pm

8. Elapsed time (from timer, or subtraction): hr min

Reactor and Biomass Weights (same as previous experiment)

1. Reactor tare weight (SCREENING, RETAINING RING, NO GASKETS): [593.0 g

from *25 run]

2. Sample weight (measured in tared reactor): [0 g from *25 run]

3. VI calibrated, non-offset or tared value, AVERAGED FOR 10 minutes (TEFLON

TUBE DISCONNECTED): -12.52 g (OLD CAL), -0.834894 mV, 26.741 DegC

4. VI calibrated, non-corrected value (TEFLON TUBE CONNECTED). This is the offset

value: 21.37 g (OLD CAL) -0.836233 mV, 26.725 DegC

5. VI final solid product mass value, AVERAGED FOR 10 minutes (TEFLON TUBE

CONNECTED):

6. VI final solid product mass value, AVERAGED FOR 10 minutes (TEFLON TUBE

DISCONNECTED): Points taked next day: -12.19 g (old VI), -0.834734 mV, 26.12 deg

C

7. VI final solid product mass negating effect of teflon tube=#6 + (#4 - #3)=g+(g-g):

Page 112: Multi-Phase Packed-Bed Thermogravimetric Analyzer

100

8. Final reactor weight with solid product, (SCREENING, RETAINING RING, NO GAS-

KETS): 592.72 g

9. Final solid product mass value (#8-#1):

10. Final solid product mass value (from dumping into tarred storage container): NA g

Pre-Run Leak Test

Leak test results: Pressure=25psi.

1. Carrier setpoint 100%, flow: % DID NOT DO

Condenser and liquid weights

1. Initial Condenser tare weight (All lines disconnected, 20 DEG C STABILIZED ONLY

AND BUBBLES EMPTIED!): 5451.7 g

2. Same as #1, but with coolant lines attached: 5560.7 g

3. Same as #1, but with coolant lines AND exhaust bubbler attached: 5566.7 g

4. Initial Condenser tare weight (ALL LINES ATTACHED):5578.1 g

5. Final tared Condenser weight (ALL LINES ATTACHED, IMMEDIATLY AFTER

RUN, AVERAGED 10 min see VI): See VI

6. Final tared Condenser weight (ALL LINES ATTACHED): See VI

7. Final tared Condenser weight (Same as 6, but without coupling):

8. Final tared Condenser weight (Same as 6, but without exhaust to bubbler line AND

coupling lines): -26.2 g ??

9. Final tared Condenser weight (Same as 6, but without exhaust to bubbler line AND

coupling lines AND Coolant lines): -159.8 g

Page 113: Multi-Phase Packed-Bed Thermogravimetric Analyzer

101

10. Final Liquid weight ((3-1)+6): -18.7 g

11. Effect of Teflon line on final weight (#5-#8): g

12. Final Liquid weight (POURED INTO BOTTLE): 0 g

D.2 Notes During Run (Record timer time and

furnace PID temp)

1. Woops forgot to tare again at the begining.

D.3 Experiment Procedures

D.3.1 Replace Reactor I

1. Engage virtual instrument and signal processing virtual instrument in section III.B.

Set instrument to test instrument. Check output and leave running.

2. Place lower gasket on lower reactor cap.

3. Leave stopper on top of thermowell opening

4. Purge carrier and quench gas for 15 seconds. Turn on carrier and quench gas to 1L/min

(100%).

D.3.2 Replace Reactor II Sample Insertion

PUT GLOVES ON AND USE PAPER TOWEL TO HOLD HOT SAMPLE PAN

1. Start countdown timer for 5 minutes (time to insert sample).

2. Remove drying sample from oven.

Page 114: Multi-Phase Packed-Bed Thermogravimetric Analyzer

102

3. Zero balance and weight sample pan + biomass (total weight). Calculate the difference

later.

4. Tare reactor

5. Pour sample into reactor and weigh. Record as initial sample weight.

6. Record time, date, and elapsed time.

7. Wrap clamps around reactor to use as handle.

8. Insert reactor.

9. Rest on lower gasket and slip top clamp around top reactor flange just to hold it.

10. Position lower gasket and secure by tightening screw thumb tight

11. Lower top flange with baffle and stap.

12. Put gasket in place

13. Put gas tank cap on thermocouple inlet to hold down top flange,

14. Slip top clamp around flanges and tighten.

15. Use large hex (or wrench) to tighten as much as possible. Only leave 5 mm ish between

the two arms. If it is much different than that, take off and reevaluate. The gasket is

likely not seated correctly.

16. Place thermocouple plate aligned so front is facing front. Tighten down nut with two

wrenches

17. Leak test: Make sure pressure is at 25 psi and caps are in place.

18. Turn carrier and quench gas to 100% (1 L/min) and record.

Page 115: Multi-Phase Packed-Bed Thermogravimetric Analyzer

103

D.3.3 Reassemble Heating Unit

1. Remove preheater strap

2. Attach thermocouple wires (if they were disconnected before)

3. Lift balance assembly and lower into position over balance plate.

4. Tilt back cooler and make sure pin is in position

5. Double check thermocouple connections (are the matched?).

6. Tighten wing nuts to lift reactor. Recheck balance pin is in position.

7. Close cooler and tighten down strap.

8. Check baffles.

9. Attach counter balance.

10. Remove cardboard

11. Blow out dust. Top hole, bottom hole, inside, repeat.

12. Hang loose thermocouple wires if needed.

13. Balance unit: XY, rotational, ohause balance level

D.3.4 Check Condenser Unit

1. Make sure condenser unit is assembled, and bubbling unit is in position. Make sure

bubbling unit has set amount of water.

2. Remove coolant lines

3. Record condenser weights

4. Attach coolant and exhaust line

Page 116: Multi-Phase Packed-Bed Thermogravimetric Analyzer

104

5. Record condenser weights

6. Attach teflon line to hose.

7. Finish Condenser weighing

8. Balance everything: XY, rotational, ohause balance level

D.3.5 Prepare for Final Experiment

1. Make sure connections are tight

2. Turn flow to experimental flow rates

3. Make sure coolant is flowing

4. Make sure outside exhuast line is clear and in position

5. Set-up VI and PID spreadsheet. Position windows

6. Set data output to initialweight output file.

7. Check data output

8. Check PID programs. Make sure they are the same and match the program indicated in

section III.A. Do not engage, but set to run so all one has to do is exit the programing

mode for it to start.

9. Check balance: XY, rotational, ohause level

10. Record weight for 20 minutes and average. Record weight as tare weight.

D.3.6 Final Step Experiment

1. Reset VI to experimental data output.

2. Tare ohause balance and record from within VI.

Page 117: Multi-Phase Packed-Bed Thermogravimetric Analyzer

105

3. Tare load cell output from with in VI. To do so, put in offset value, and then enter

biomass weight. Load cell output should now reflect biomass weight.

4. Check balance: XY, rotation, level

5. Record start temperatures in PID spreadsheet

6. Plug in preheater and furnace ground fault interupters. Test and reset check.

7. Start PID controllers

8. Start timer and VI at the same time.

9. Do not change display during experiment and be sure to disrupt the mouse so the system

does not lock.

D.3.7 During run

1. Record PID preheater and furnace temperatures every 10 minutes or if needed.

2. Clean area, prep cleaning solutions, vacuum cleaner, etc.

3. Find bottle for storing liquid and bag for charcoal. Label and record tar weights on

the containers.

4. Prepare biomass for drying for next run (see section XIV)

5. Prepare procedures and data files for next run (30 min

6. Analyze previous days run (120 min) (see section XIII)

D.3.8 Analysis from previous day

1. Update procedures

2. Update data into spreadsheet

3. Write short report

Page 118: Multi-Phase Packed-Bed Thermogravimetric Analyzer

106

D.3.9 Dry Biomass / Prep

1. Set drying oven to setpoint temperature

2. Measure weight of drying pan (tare)

3. Place target amount of sample in pan and distribut

4. Put sample pan in oven (when sp is reached) and record date and time. Start timer.

5. Set chiller to setpoint and clamp off flow to condensers, but leave bypass flow fully

open.

D.3.10 Terminate Experiment

1. Unplug preheater and furnace when program has ended. Do this carefully. Stop program

and reset to test. Save and quit.

2. Turn down carrier and quench gases to standby flows

3. Let system cool on its own to about 400 degrees C, then open furnace.

4. Open condenser bypass line and pinch off condenser coolant line.

5. Make sure chimney is ok. If rain expected, remove.

6. CHECK THAT FURNACE AND PREHEATER ARE UNPLUGGED!!

7. LEAVE GAS RUNNING THROUGH BOTH LINES UNTIL COMPLETELY COOL.

LET COOL COMPLETELY OVERNIGHT AND PROCEED WITH CLEANING THE

NEXT DAY!

D.4 Opening and Cleaning Procedures

WARNING: LABCOAT, RESPIRATOR, GOGGLES, RAG, GLOVES, and LAMP RE-

QUIRED. SETUP BEFORE HAND

Page 119: Multi-Phase Packed-Bed Thermogravimetric Analyzer

107

D.4.1 Pre-Clean

1. Record condenser and reactor weights with teflon line attached.

2. Position bucket to catch loose material

3. Record condenser weights with teflon line removed only from reactor

4. Record condenser weights with teflon line removed only from reactor, exhaust un-

screwed, and coolant disconnected

5. Transfer condenser unit to cleaning area.

D.4.2 Disengage balance mechanism

1. Insert cardboard under reactor

2. Mark balance rod top wing nuts if necessary

3. Lower upper baffle washers, raise upper baffle plate

4. Remove thermocouple wires

5. Lower wing nuts on opposing corners and remove all.

6. Remove counter balance

7. Remove cooler strap

8. Lift off cooler and balance and slide backwards until protruding bolt can lock in 8020

slot. Unit should balance.

D.4.3 Remove reactor core

1. Remove balance plate

2. Tilt cooler back and twist off thermocouple plate. Lay to the side.

Page 120: Multi-Phase Packed-Bed Thermogravimetric Analyzer

108

3. Plug thermocouple inlet with stopper

4. Use strap to lift preheater.

5. Spray a bit of WD40 on reactor clamp screws

6. Loosen top and bottom reactor screws

7. Completely loosen bottom reactor plate followed by top reactor plate

8. Move preaheater assembly out of position by clamping upper baffle and using strap.

9. Remove reactor with top clamp wrapped around reactor.

10. Carefully remove the gasket material from from the reactor flanges with a razor blade

if needed. Do not disrupt biomass.

11. Use SS brush to remove gasket material. Vacuum away.

12. Record reactor weight in section IV.A.

13. Transfer char in reactor to previously labeled bag. Record new weight, subtract from

tar, to give solid weight.

14. Completely Scrape gasket material from top and bottom flanges of inside furnace and

reactor. Vacuum away.

15. Replace screening

D.4.4 Clean reactor

1. Make sure quench gas line is removed

2. Make sure quench inlet is capped.

3. Make sure standby carrier gas is on.

Page 121: Multi-Phase Packed-Bed Thermogravimetric Analyzer

109

4. Make sure stopper is in thermocouple inlet.

5. Make sure bucket is in position.

6. Flush with degreaser.

7. Cap exit and fill with a little acetone and rest with degreaser. Let sit until condensers

are cleaned.

8. Remove caps and flush out until clear with acetone and water.

9. Plug bottom flange hole with compressed air fitting and turn on. Let interior dry.

D.4.5 Clean and Replace Condensers

Use metal dishrack

1. Clean bubble flask and associated exhaust line

2. Remove and clean exhauast tube.

3. Remove and clean small fittings.

4. Remove and clean condensers.

5. Pour oil into sample bottle and record weights.

6. Clean flasks

7. Put vacuum grease on fittings

8. Reassemble condenser with special care to have inlet fitting pointing in line with con-

densers (to the left)

9. Insert teflon line inlet so drips dont drip but slide down the surface of the flask.

10. Place on balance and connect coolant lines.

11. Flush coolant by opening valves and tilting unit.

Page 122: Multi-Phase Packed-Bed Thermogravimetric Analyzer

Appendix E

Example Procedures for a Biomass run1

1Procedures: 20080420_1 Fourth Cellulose Biomass Run (20 April 2008 Rev.1 by ErikJarrett)

110

Page 123: Multi-Phase Packed-Bed Thermogravimetric Analyzer

111

E.0.6 Purpose

The purpose of this run is to perform the fourth and last alpha-cellulose run.

E.0.7 Summary

The reactor will have as close to 145 grams of material inside. Use a flow rate of 1/4, 1/4

L/min for both carrier and quench. The heating rate is 3 deg C / min and does not change

from ambient temperature.

E.1 Customized Experimental Section

E.1.1 Independent Variables

Sample and Drying

1. Biomass type: alpha-cellulose.

2. Dry Biomass target weight: 145 g

3. Biomass drying temperature: 110 Deg C

4. Biomass drying furnace: Inside 700 lab

Experiment

1. Carrier and quench gas: Nitrogen

2. Bubble trap volume (keep same flask): 300 ml

3. Room temperature setpoint: 73 deg F

4. Load Cell Instrument cooler setpoint: Variable around 26 deg C

5. Standby flow rate: Carrier flow rate: 0.05 L/min, Quench flow rate 0.05 L/min or OFF

and sealed

Page 124: Multi-Phase Packed-Bed Thermogravimetric Analyzer

112

6. Experiment step flow rates: Carrier flow rate: 0.25 L /min; Quench flow rate: 0.25

L/min

7. Experiment step chiller temperature: 5 deg C only when duing baseline run.

8. Experiment PID manual recording interval: 10 minutes during baseline run

Experiment temperature profile

1. Start at room temperature

2. Ramp at 3 deg C / min from 20 deg C to 600 deg C (3.22 hrs)

3. Soak 45 min (3.72 hrs total). 30 minutes for final experiment and 10 minutes for final

no teflon average

4. End program and return to room temperature

Virtual Instrument and Signal Processing

1. Virtual Instrument Name: AAFinalProject9.vi

2. File name for testing: test.txt

3. File name for run data: 20080420_1experiment.txt (many files)

4. Calibration from file: NO CALIBRATION

5. Calibration equation: NO CALIBRATION THAT IS MEANINGFUL YET

E.1.2 Dependent

Drying Period

1. Initial drying pan weight: g

2. Initial sample weight (in tared pan; target 280g): g (4 scoops of 175 ml)

Page 125: Multi-Phase Packed-Bed Thermogravimetric Analyzer

113

3. Date and time of sample insert: @pm

4. Final Total sample weight (final + drying pan): g

5. CALC Final sample weight (Final Total - Initial drying pan): g

6. CALC Moisture loss: (1-FinalSampleWeight/InitialSampleWeight)*100: %

7. Date and time of sample removal: @ pm

8. Elapsed time (from timer, or subtraction): hr min

Reactor and Biomass Weights (same as previous experiment)

1. Clean clamps and nuts: g

2. Same as 1 but with anti-seize: g

3. Weight of anti-seize: g

4. Reactor tare weight (SCREENING, RETAINING RING, NO GASKETS): g

5. Sample weight (measured in tared reactor): g

6. VI calibrated, non-offset or tared value, AVERAGED FOR 10 minutes (TEFLON

TUBE DISCONNECTED): g (OLD), mV, DegC

7. VI calibrated, non-corrected value (TEFLON TUBE CONNECTED). This is the offset

value: g (OLD), mV, DegC)

8. VI final solid product mass value, AVERAGED FOR 10 minutes (TEFLON TUBE

CONNECTED): SEE VI g

9. VI final solid product mass value, AVERAGED FOR 10 minutes (TEFLON TUBE

DISCONNECTED): g (OLD), mV, deg c

Page 126: Multi-Phase Packed-Bed Thermogravimetric Analyzer

114

10. VI final solid product mass negating effect of teflon tube=#6 + (#4 - #3)=g+(g-g)=

g

11. Dirty clamps and nuts: g

12. Weight change in clamps: g

13. Final reactor weight with solid product, (SCREENING, RETAINING RING, NO GAS-

KETS): g

14. Final solid product mass value(#12-#4): g

15. Final solid product mass value (from dumping into tarred storage container): g

Pre-Run Leak Test

Leak test results: Pressure=25psi.

1. Carrier setpoint 100%, flow: %

Condenser and liquid weights

1. Initial Condenser tare weight (nothing connected): g

2. Same as 1, but with coolant: g (weight of coolant line: g)

3. Same as 2, but with exhaust: g (weight of exhaust line: g)

4. Same as 3, but with coupling (this is the tare weight): g (weight of coupling: g)

5. Final tared Condenser weight (ALL LINES ATTACHED, IMMEDIATLY AFTER

RUN, AVERAGED 10 minsee VI): g

6. Final tared Cond. weight (next day or whenever, single point): g (next day)

7. Same as 6, but no coupling: g (weight of coupling: g)

Page 127: Multi-Phase Packed-Bed Thermogravimetric Analyzer

115

8. Same as 7, but no exhaust: g (weight of exhaust: g)

9. Same as 8, but no coolant lines: g ( g nothing) (weight of coolant lines: g)

10. Change in coupling (f-i): g

11. Change in exhaust line (f-i): g

12. Change in coolant lines (f-i): g

13. Final Liquid weight (6-2): g

14. Final Liquid weight (POURED INTO BOTTLE): g

E.1.3 Notes During Run (Record timer time and furnace PID temp)

E.2 Experiment Procedure Section

E.2.1 Replace Reactor I

1. Engage virtual instrument and signal processing virtual instrument in section III.B.

Set instrument to test instrument. Check output and leave running.

2. Place lower gasket on lower reactor cap.

3. Leave stopper on top of thermowell opening

4. Purge carrier and quench gas for 15 seconds. Turn on carrier and quench gas to 1L/min

(100%).

E.2.2 Replace Reactor II Sample Insertion

PUT GLOVES ON AND USE PAPER TOWEL TO HOLD HOT SAMPLE PAN

1. Start countdown timer for 5 minutes (time to insert sample).

Page 128: Multi-Phase Packed-Bed Thermogravimetric Analyzer

116

2. Remove drying sample from oven.

3. Zero balance and weight sample pan + biomass (total weight). Calculate the difference

later.

4. Tare reactor

5. Pour sample into reactor and weigh. Record as initial sample weight.

6. Record time, date, and elapsed time.

7. Wrap clamps around reactor to use as handle.

8. Insert reactor.

9. Rest on lower gasket and slip top clamp around top reactor flange just to hold it.

10. Position lower gasket and secure by tightening screw thumb tight

11. Lower top flange with baffle and stap.

12. Put gasket in place

13. Put gas tank cap on thermocouple inlet to hold down top flange,

14. Slip top clamp around flanges and tighten.

15. Use large hex (or wrench) to tighten as much as possible. Only leave 5 mm ish between

the two arms. If it is much different than that, take off and reevaluate. The gasket is

likely not seated correctly.

16. Place thermocouple plate aligned so front is facing front. Tighten down nut with two

wrenches

17. Leak test: Make sure pressure is at 25 psi and caps are in place.

18. Turn carrier and quench gas to 100% (1 L/min) and record.

Page 129: Multi-Phase Packed-Bed Thermogravimetric Analyzer

117

E.2.3 Reassemble Heating Unit

1. Remove preheater strap

2. Attach thermocouple wires (if they were disconnected before)

3. Lift balance assembly and lower into position over balance plate.

4. Tilt back cooler and make sure pin is in position

5. Double check thermocouple connections (are the matched?).

6. Tighten wing nuts to lift reactor. Recheck balance pin is in position.

7. Close cooler and tighten down strap.

8. Check baffles.

9. Attach counter balance.

10. Remove cardboard

11. Blow out dust. Top hole, bottom hole, inside, repeat.

12. Hang loose thermocouple wires if needed.

13. Balance unit: XY, rotational, ohause balance level

E.2.4 Check Condenser Unit

1. Make sure condenser unit is assembled, and bubbling unit is in position. Make sure

bubbling unit has set amount of water.

2. Remove coolant lines

3. Record condenser weights

4. Attach coolant and exhaust line

Page 130: Multi-Phase Packed-Bed Thermogravimetric Analyzer

118

5. Record condenser weights

6. Attach teflon line to hose.

7. Finish Condenser weighing

8. Balance everything: XY, rotational, ohause balance level

E.2.5 Prepare for Final Experiment

1. Make sure connections are tight

2. Turn flow to experimental flow rates

3. Make sure coolant is flowing

4. Make sure outside exhuast line is clear and in position

5. Set-up VI and PID spreadsheet. Position windows

6. Set data output to initialweight output file.

7. Check data output

8. Check PID programs. Make sure they are the same and match the program indicated in

section III.A. Do not engage, but set to run so all one has to do is exit the programing

mode for it to start.

9. Check balance: XY, rotational, ohause level

10. Record weight for 20 minutes and average. Record weight as tare weight.

E.2.6 Final Step Experiment

1. Reset VI to experimental data output.

2. Tare ohause balance and record from within VI.

Page 131: Multi-Phase Packed-Bed Thermogravimetric Analyzer

119

3. Tare load cell output from with in VI. To do so, put in offset value, and then enter

biomass weight. Load cell output should now reflect biomass weight.

4. Check balance: XY, rotation, level

5. Record start temperatures in PID spreadsheet

6. Plug in preheater and furnace ground fault interupters. Test and reset check.

7. Start PID controllers

8. Start timer and VI at the same time.

9. Do not change display during experiment and be sure to disrupt the mouse so the system

does not lock.

E.2.7 During run

1. Record PID preheater and furnace temperatures every 10 minutes or if needed.

2. Clean area, prep cleaning solutions, vacuum cleaner, etc.

3. Find bottle for storing liquid and bag for charcoal. Label and record tar weights on

the containers.

4. Prepare biomass for drying for next run (see section XIV)

5. Prepare procedures and data files for next run (30 min

6. Analyze previous days run (120 min) (see section XIII)

E.2.8 Analysis from previous day

1. Update procedures

2. Update data into spreadsheet

3. Write short report

Page 132: Multi-Phase Packed-Bed Thermogravimetric Analyzer

120

E.2.9 Dry Biomass / Prep

1. Set drying oven to setpoint temperature

2. Measure weight of drying pan (tare)

3. Place target amount of sample in pan and distribut

4. Put sample pan in oven (when sp is reached) and record date and time. Start timer.

5. Set chiller to setpoint and clamp off flow to condensers, but leave bypass flow fully

open.

E.2.10 Terminate Experiment

1. Unplug preheater and furnace when program has ended. Do this carefully. Stop program

and reset to test. Save and quit.

2. Turn down carrier and quench gases to standby flows

3. Let system cool on its own to about 400 degrees C, then open furnace.

4. Open condenser bypass line and pinch off condenser coolant line.

5. Make sure chimney is ok. If rain expected, remove.

6. CHECK THAT FURNACE AND PREHEATER ARE UNPLUGGED!!

7. LEAVE GAS RUNNING THROUGH BOTH LINES UNTIL COMPLETELY COOL.

LET COOL COMPLETELY OVERNIGHT AND PROCEED WITH CLEANING THE

NEXT DAY!

E.3 Opening and Cleaning Procedures

WARNING: LABCOAT, RESPIRATOR, GOGGLES, RAG, GLOVES, and LAMP RE-

QUIRED. SETUP BEFORE HAND

Page 133: Multi-Phase Packed-Bed Thermogravimetric Analyzer

121

E.3.1 Pre-Clean

1. Record condenser and reactor weights with teflon line attached.

2. Position bucket to catch loose material

3. Record condenser weights with teflon line removed only from reactor

4. Record condenser weights with teflon line removed only from reactor, exhaust un-

screwed, and coolant disconnected

5. Transfer condenser unit to cleaning area.

E.3.2 Disengage balance mechanism

1. Insert cardboard under reactor

2. Mark balance rod top wing nuts if necessary

3. Lower upper baffle washers, raise upper baffle plate

4. Remove thermocouple wires

5. Lower wing nuts on opposing corners and remove all.

6. Remove counter balance

7. Remove cooler strap

8. Lift off cooler and balance and slide backwards until protruding bolt can lock in 8020

slot. Unit should balance.

E.3.3 Remove reactor core

1. Remove balance plate

2. Tilt cooler back and twist off thermocouple plate. Lay to the side.

Page 134: Multi-Phase Packed-Bed Thermogravimetric Analyzer

122

3. Plug thermocouple inlet with stopper

4. Use strap to lift preheater.

5. Spray a bit of WD40 on reactor clamp screws

6. Loosen top and bottom reactor screws

7. Completely loosen bottom reactor plate followed by top reactor plate

8. Move preaheater assembly out of position by clamping upper baffle and using strap.

9. Remove reactor with top clamp wrapped around reactor.

10. Carefully remove the gasket material from from the reactor flanges with a razor blade

if needed. Do not disrupt biomass.

11. Use SS brush to remove gasket material. Vacuum away.

12. Record reactor weight in section IV.A.

13. Transfer char in reactor to previously labeled bag. Record new weight, subtract from

tar, to give solid weight.

14. Completely Scrape gasket material from top and bottom flanges of inside furnace and

reactor. Vacuum away.

15. Replace screening

E.3.4 Clean reactor

1. Make sure quench gas line is removed

2. Make sure quench inlet is capped.

3. Make sure standby carrier gas is on.

Page 135: Multi-Phase Packed-Bed Thermogravimetric Analyzer

123

4. Make sure stopper is in thermocouple inlet.

5. Make sure bucket is in position.

6. Flush with degreaser.

7. Cap exit and fill with a little acetone and rest with degreaser. Let sit until condensers

are cleaned.

8. Remove caps and flush out until clear with acetone and water.

9. Plug bottom flange hole with compressed air fitting and turn on. Let interior dry.

E.3.5 Clean and Replace Condensers

Use metal dishrack

1. Clean bubble flask and associated exhaust line

2. Remove and clean exhauast tube.

3. Remove and clean small fittings.

4. Remove and clean condensers.

5. Pour oil into sample bottle and record weights.

6. Clean flasks

7. Put vacuum grease on fittings

8. Reassemble condenser with special care to have inlet fitting pointing in line with con-

densers (to the left)

9. Insert teflon line inlet so drips dont drip but slide down the surface of the flask.

10. Place on balance and connect coolant lines.

11. Flush coolant by opening valves and tilting unit.