Powder Technology 269 (2015) 178–184

Contents lists available at ScienceDirect

Powder Technology

j ourna l homepage: www.e lsev ie r .com/ locate /powtec

Validation of an inline particle probe in a high-shear mixer for particlesize determination

V. Wenzel ⁎, H. NirschlKarlsruhe Institute of Technology, Institute for Mechanical Process Engineering and Mechanics, D-76128 Karlsruhe, Federal Republic of Germany

⁎ Corresponding author. Tel.: +49 721 608 44139; fax:E-mail address: valentin.wenzel@kit.edu (V. Wenzel).

http://dx.doi.org/10.1016/j.powtec.2014.09.0030032-5910/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 22 May 2014Received in revised form 2 September 2014Accepted 3 September 2014Available online 16 September 2014

Keywords:AgglomerationHigh shearMixingParticle probeInline measurement

This paper deals with agglomeration processes in a high-shear mixer and the integration of an inline particleprobe. The purpose is to determine particle size distribution continuously in real-time. High-shear mixers rotatewith a very high mixing tool speed. As a result, not only a mixing effect, but also a granulating effect (under cer-tain conditions) is obtained. This granulating effect is to be studied extensively using an inline particle probe. Inthe first instance, it is decisive to integrate the probe and find settings tomeasure and to represent thewhole par-ticle size distribution in themixing vessel. This articlewill highlight some of the findings obtained from the com-prehensive parameter study. Additionally, the technical implementation of the inline particle probe in a high-shear mixer will be described.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

More process analysis technologies (PAT) are not only claimed in theGuidance for Industry [1] of the Food and Drug Administration (FDA),but are rather in the interest of all industries defining products bytheir physical properties. One goal is the real-time analysis and docu-mentation of manufacturing processes. Results can be used to monitorquality parameters, set target values, and identify and eliminate productdefects. The implementation of process analytical technologies bears agreat potential for process optimization and understanding as well asenormous cost-saving opportunities. It is important not only to identifyfundamental properties, but also to determine the technical behaviourin manufacturing processes.

For determination of process-relevant parameters, four classifica-tions of process analysis are made: offline, atline, inline, and online.Offline and atline analytics are determined by manual or semi-automatic sampling followed by discontinuous evaluation in centrallaboratories (offline) or in specific analyzers near to the productionplace (atline). In both cases, the product properties may change duringcharacterization. On the other hand, online and inline procedurespermit continuous process control without manual sampling. Thismeans that a continuous correlation between received informationand the features of theprocess or product is possible. For online analysis,

+49 721 608 42403.

in most cases, the sample is measured in a bypass. In the case of inlinemeasurements, the measuring cell is located directly in the productstream so that sampling is eliminated completely.

The particle size distribution of a product is crucial tomany of its fea-tures and to subsequent processing. Hence, there is an increasing needfor its real-time determination. With this information, it is possible toreact directly to changes in the product quality. Many investigations todetermine the particle size continuously with an inline particle probehave already been carried out in fluidised beds [2,3] and grinding pro-cesses [4,5]. Real-time monitoring and control of agglomerations in ahigh-shear mixer are of importance to science and manufacture. Inthis way, more information is obtained about the process and aboutthe operating parameters that influence agglomeration. In general,many basic investigations have focused on wet granulation processes[6–10]. More detailed studies consider binder properties [11–15] orgrowth kinetics [16,17]. Nevertheless, there is still a lack of understand-ing. This article illustrates the possibility to determine the particle sizeinline and the interaction between particle stress and different processparameters.

2. Instruments and materials

2.1. High-shear mixer

All experiments were carried out on an Eirich type high-shear mixerR02manufactured in Hardheim, Germany. Themixer consists of a rotat-ing mixing vessel, a rapidly rotating eccentric mixing tool, and a wall

Table 1Agitator—technical details.

Unit Star agitator Pin agitator

Diameter of the mixing tool [mm] 135 125Impact body [−] 8 6Striking surface [mm2] 20586 10672Impact body surface [mm2] 22130 3104Mass [g] 475 875

179V. Wenzel, H. Nirschl / Powder Technology 269 (2015) 178–184

scraper. The mixing vessel and the agitator can be operated in thecounter- and co-current flow modes. As a result of the mixing vessel'srotation, the test product is transported into the high-shear zone,i.e., towards the mixing tool. The wall scraper is a static tool whichprevents deposit build-up on the vessel wall. Products adhering to thevessel wall are detached. Two mixing tools were studied. Fig. 1 showsthe star agitator (left) and the pin agitator (right). Due to their differentgeometries, they also differ in other aspects. The star agitator exerts a ra-dial mixing effect with cutting and impact load. In addition, it fluidisesthe mixing vessel content. The pin agitator, by contrast, exerts an axialmixing effectwith a high shear loadon the productwhich keeps thema-terial more compact. Other features are listed in Table 1 and [18].

2.2. Determination of particle size and particle size distribution

2.2.1. Inline particle probeThe investigated method to determine the particle size is a fibre-

optical spatial filtering technique, which is well-studied and still inte-grated, e.g., in fluidised bed processes [2,3,18–20]. The “Inline ParticleProbe 70” has been designed for inline particle size determination in in-dustrial production facilities by ParsumGmbH, Chemnitz, Germany. Thesystem determines simultaneously the time and the velocity of particlespassing the probe's sensing volume. It consists of a periodic arrange-ment of waveguides in the form of a lattice with a characteristic latticeconstant. With an optical-fibre spatial-frequency filter anemometer(“ab“differential lattice), particle velocity vp is determined. For this pur-pose, moving particles are projected onto the optical lattice. Then, burstdetection/frequency analysis of the produced signal is carried out. Asupplementary optical channel (“c” monofibre) allows the particle di-ameter xp to be determined from the pulse width tp. The statistical par-ticle diameter xp is measured directly as the chord length in thedirection of the spatial filter axis. The mode and function of the particleprobe are described in [2,18–21] in greater detail. The measurementrange extends from 50 μm to 6 mm (without additional dispersingunit). Due to the high solid concentration in the process, it is necessaryto use an additional dispersing unit, which serves to disperse and diluteparticle flow in the sensing volume. Since the entrance window of thedispersing unit has a nozzle with 2.5 mm, the maximum measurableparticle size is reduced (from 6 mm) to 2 mm.

2.2.2. Retsch Camsizer®To assess the agglomerates and validate the inline particle probe, an

automatic image evaluation system by Retsch & Technology is applied.The Camsizer® uses the principle of digital image processing. It is qual-ified for dry analysis of powders and granulates. Themeasuring range ofparticles is between 30 μm and 30 mm. The granulate curve can be de-termined in a non-destructive manner. Digital image processing yields

Fig. 1. High-shear mixing tools for the Eirich-type high-shear mixer. Star agitator (left);pin agitator (right).

the following characteristics: Median value x50, particle size x10, andparticle size x90.

2.3. Material

The agglomeration experiments are carried out using cohesive clayand feldspar. These two powders are mixed in a ratio of 3:2 (kaoliniticclay:feldspar). The properties of the test product are listed in Table 2.To trigger the agglomeration process, the binder liquid is added via adropping funnel directly into the mixing zone. During each mixing ex-periment, purified water is used as binder liquid.

3. Experimental procedure, test results, and interpretation

3.1. Pilot test

To ensure that the probe works properly and a comparison withother particle size measuring systems is possible, some preliminarytests have to be made. The particle sizes and particle shapes of thesmall glass marbles (x = 0.75…1 mm) and dry quartz sand (x =0.15…1 mm) used are similar to those of the agglomerates examinedat a later stage. Most pilot tests are performed in a simple test setupand not in the high-shear mixer itself, because the mechanical stresscaused by glass marbles, in particular, is too high. In the pilot testsetup, the particles fall freely through a pipeline and directly throughthe probemeasurement volume. As a result, particle flow is very homo-geneous and constant. For the glass marbles, the comparison of theinline particle probe, the Laser Diffraction Sensor (Helos by Sympatec)and the Camsizer has revealed a very goodmatch of the cumulative dis-tribution. The results for the dry quartz sand are shown in Fig. 2. Basedon the higher mechanical stress of the quartz sand, it was possible tocalibrate the particle size directly in themixer, too. The values obtainedat an agitator speed of 500 rpm (IPP-70 dynamic) are shown. The resultsagree with the experiments performed outside of the mixer (IPP-70static). The observed differences are due to the different measuringmethods. As a result, the Inline Particle Probe and the Camsizer havethe chord length of the particle. The diameter of the Laser DiffractionSensor is equivalent to that of the particles. These differences arise be-cause quartz sand is an irregularly shaped product. It is very importantto realize that the results obtained using the inline particle probe are in-between those obtained by the two established methods. Further

Table 2Test product properties.

Unit Kaolinitic clay Feldspar Mixture

Colour [−] Grey White Grey

Particle diameter x10 [μm] 1.09 1.49 2.29Particle diameter x50 [μm] 4.41 12.26 6.30Particle diameter x90 [μm] 22.40 41.05 32.16Specific surface SV [m2/cm3] 2.27 1.43 1.79Solid density ρS [g/cm3] 2.66 2.62 2.64bulk density ρbulk [kg/m3] 696.5

Fig. 2. Comparison of measurement methods for dry quartz sand.Fig. 3. Particle size distribution at different acquisition parameters, nA = 1000 rpm.

Table 3Acquisition parameters.

Acquisitionparameter

Particlebuffer

Loading Probability ofresidence

Experimentalorder

Probeposition

A 50 000 70% OFF I 3B 50 000 20% OFF I 3C 50 000 30% ON II 3D 10 000 30% ON II 3

180 V. Wenzel, H. Nirschl / Powder Technology 269 (2015) 178–184

investigations to validate the inline particle probe in a high-shearmixerin this paper are divided into three parts:

• Description of the inline particle probe measuring system• Differences in themeasuringmethods andmeasurement specifications• Variation of other parameters

3.2. Description of the system

The experimental procedure comprises three parts, i.e., preparationof the mixing test, the mixing experiment itself, and post-processing.

To prepare the experiments, the residual moisture (base products),the initial weight (base product), and the amount of water needed forthe agglomeration process are determined. Execution of the experimentincludes the filling of the mixer vessel, start of data acquisition (particleprobe), and beginning with themixing and agglomeration procedure inthe counter-flowmode. Attempts to describe the drymixing of clay andfeldspar have shown that 15 s are sufficient to achieve a goodmixing ef-ficiency. After these 15 s of dry powder mixing, the binder liquid isadded. Depending on the amount of water, this takes up to 20 s. The ag-glomeration process starting next takes 120 to up to 540 s.

After agglomeration, the high-shear mixer and the data acquisitionare stopped. The post-processing steps include determination ofthe residual moisture (agglomerates), sample separation, sieving, andCamsizer® analysis. To determine powder and agglomerate moisture,a rapidly measuring moisture meter by Sartorius MA 30, Göttingen,Germany, is applied.

After the mixing test, the agglomerates produced in the mixing ves-sel are separated using a rotating sample separator PT 100 supplied byRetsch GmbH. In this way, the total product quantity between 2 and3 kg is reduced to a fewhundred grammes per experiment. After sampleseparation, one assay is sieved. Sieving (50–2800 μm) is necessary, be-cause the metering capacities of both particle size measuring systemsare completely different. Under certain conditions, agglomerates biggerthan 2800 μmare generated. The Camsizer®measurement (not sieved)yields information about the whole content of the vessel. From the dis-tributions reflecting themetering capacity of the probe (sieved), gener-al agreement between themeasurement results can be determined andit can be found out whether the probe distributions are realistic.

Fig. 3 shows several particle size distributions for an agitator speedof nA = 1000 rpm depending on different probe settings, referred toas acquisition parameters. The specified measurement range of theprobe is marked in grey. For this figure, the same experiment wasdone four times. The only difference consists in some probe acquisitionparameters and the data acquisition moment (Table 3). The cumulativedistribution measured additionally by the Camsizer® is visualized, too.

The first experiments for probe validation in the mixer are performedwith the standard software parameters (acquisition parameter A). Asis obvious, there is a large difference between the IPP-70S (acquisitionparameter A) and the Camsizer®. According to Fig. 3, it can be observedthat particles are measured although their size is larger than the rangestated by the manufacturer. It shows that the factory specifications arevery conservative and quite larger particles can be detected. Even athigher agitator speeds, a difference between the measurements of theinline particle probe with standard settings and the Camsizer® isnoted. Exceeding the measuring range, the values do not fit well. Thedistribution is much tighter and agglomerates are smaller. The remain-ing acquisition parameters (B, C, andD) are discussed below. For furtherinquiries, it is important to find out the reason for the inconsistency.Therefore, a deeper understanding of the probe setting is needed andthe dynamics of the system has to be coordinated.

3.3. Differences in the measuring methods

To analyse the behaviour of particle size distribution at variable soft-ware parameters, loading, probability of residence, and theparticle buff-er are examined. Loading describes the particle concentration in thesensing volume. If it is too high, it may lead to coincidence effects. As aresult, the measured distribution functions are shifted towards largerparticles, since larger particles apparently have been found. If the limitvalue “maximum loading” is exceeded, data recording (in the particlering buffer) is stopped. The limitation of loading helps to identify the pe-riods inwhich the particle flow is too concentrated for a directmeasure-ment. The probability of residence is a correction module taking intoaccount the influence of particle size. Moreover, small particles maybe concealed by larger ones within the measurement volume and,thus, are “not seen”. In the case of many closely distributed products,the influence on the measured distributions is small. In the case ofvery widely distributed products, use of this module (“ON”) may pre-vent under-representation of the proportion of finer particles. The par-ticle buffer is a parameter to specify the number of valid particles used

181V. Wenzel, H. Nirschl / Powder Technology 269 (2015) 178–184

in computing distributions. During measurement, the particles are de-posited continuously in an appropriately sized loop memory, with the“oldest” particles being replaced by new ones. This allows a continuouscalculation of particle size distributions. A low value in this loopmemo-ry yields more volatile results. A larger ring buffer, by contrast, causessmoothing of the measurement, together with a longer reaction timeto rapid changes in the process. The statistical reliability of measure-ment increases with larger values. The particle number in the buffercan range from 100 to 200,000 [22].

In the course of one experiment, especially in the case of tests ofshort duration, the particle buffer is not filled completely. As a result,the particle size distribution is influenced by incorrectly measured par-ticles at the beginning (dry mixing process). At the early stages, there isonly the dry base product. The particle size of this base product is muchsmaller than the metering capacity of the IPP-70; this means that theprobe does not provide reliable data. The easiest way to reduce thequantity of “incorrectly measured particles” is to start data acquisitionafter dry powder mixing (data acquisition starts with adding binderliquid—experimental order II). Continuous particle size determinationwith the particle probe focuses on the agglomerate size rather than onthe primary particle. The advantage of this new experimental order isthat not many particles are accumulated in the particle buffer duringdry powder mixing and the first seconds of water addition. The follow-ing figure exhibits the influence of some acquisition parameters and,hence, the best choice of these parameters.

For in-depth understanding, it is interesting to study two acquisitionparameters (here A and C) and the corresponding particle size distribu-tion inmore detail. Fig. 4 depicts the behaviour of the particle size (x10,3;x50,3; x90,3) vs. time and the registered particle number in the sensingvolume. The reduction of the particle size distribution to the threevalues (x10,3; x50,3; x90,3) allows a clearer representation of the timecourse with respect to the mean value x50,3 and the width of the distri-bution (analogous to the representations in [19,20,23]). The relevantcumulative distributions (after the processing time) are shown inFig. 3. The main differences in the test conditions of Fig. 4 are the max-imal loading and the experimental order (see acquisition parameters Aand C). In Fig. 4 (left), data acquisition starts immediatelywith dry pow-der mixing. Consequently, large amounts of dry powder pass the sens-ing volume. These primary particles are out of the probe measurementrange and are probably measured incorrectly, but will be added to theparticle buffer. The agglomeration time amounts to 210 s, which is notlong enough to exchange the whole particle buffer (counted particlesafter 210 s b 50 000) so that the result is influenced by “incorrectlymea-sured” particles. Moreover, the gradient of the particle number is muchsmaller in Fig. 4 (left) than in Fig. 4 (right). The particle number de-scribes the agglomerates counted and measured up to this time. Basedon this information, the particle size distribution is made. The cumula-tive number of particles, however, could not exceed the value of the

Fig. 4. Particle size distribution, nA = 1000 rpm, acquisition

particle buffer. As is evident from Fig. 4, fewer particles per rate aremea-sured at acquisition parameter A. As the loading is 70%, these particleshave been identified (with highest probability) as being too large. Forfurther systemoptimisation, the particle buffer is reduced to 10000 par-ticles. This makes the system more dynamic and it can respond tochanges in particle size distribution more quickly.

Based on these findings, the following investigations are carried out.

3.4. Variation of other parameters

After having understood the inline particle probe as well as the soft-ware parameters' interaction, other variables are investigated based onthe parameters shown in Table 4. The experimental order correspondsto the second step, where data acquisition (particle probe) starts afterthe drymixing process. The influence of moisture content on agglomer-ation, growth rate of the agglomerates, rotating speeds of the agitator,probe position,mixing tools, and filler loadingwere examined. Thefillerloading is defined as the ratio betweenweighed powder portion and themaximumvessel volume (10 l) using the bulk density of themixture ac-cording to Table 2. Before modifying the parameters that might influ-ence mixing agglomeration, it is necessary to determine the mass-related moisture xW (=agglomerate moisture or residual moisture) atwhich themixture agglomerates and stable agglomerates exist. The ag-glomerate moisture xW describes the amount of water used for the ag-glomeration. It is defined in Eq. (1) as

xW ¼ mliquid

mliquid þmsolidð1Þ

The moisture content significantly influences the adhesive forces ofthe liquid, which change granulate distribution and the median value[24]. The median value x50 and the scattering parameter x90/x10 areshown in Fig. 5. It was found out that all the powder will agglomeratewhen the solid moisture reaches a value of xW ≥ 12.5%. The point ofcomplete agglomeration is a point of immediate change located be-tween the unstable point of xW = 12.5% and xW = 13.0% solid mois-ture. These observations during the tests confirm the results shown inFig. 5. It can be observed that themedian value and scattering parameterbetween amass-relatedmoisture xW=11% and xW= 12.5% differ con-siderably. This can be explained by the still dry product (more or less)powder (x50 = 6.3 μm). This does not correspond to the measuringrange and, hence, no reliable data are obtained. At a moisture contentof xW ≥ 12.5%, the median values and scattering parameters matchvery well. High moisture xW N 12.5% causes the agglomerate size to in-crease. In [23], it is shown that the rotating speeds of the agitator have abig influence on thewidth of the particle size distribution and themedi-an value x50. Therefore, it is necessary to know from which agitatorspeed the probe represents the whole content in the mixing vessel.

parameter A (left), and acquisition parameter C (right).

Table 4Constant process and software parameters.

Unit

Particle in ring [−] 10 000Loading [%] 30Probability of residence [−] OnPurge air internal [l/min] 20Purge air external [l/min] 3Vessel speed [rpm] 50Agglomeration time [s] 210Experimental order [−] II

Fig. 6. Particle size distribution as a function of the rotating speed (nA = 1000, 4000 rpm)under constant test conditions: moisture content xW = 12.5%, filling level φ = 35%, andmixing tool: Pin agitator.

182 V. Wenzel, H. Nirschl / Powder Technology 269 (2015) 178–184

Moreover, it is of interest whether the test reading of the probe can becompared with the results of the Camsizer®. For this purpose, severalexperiments were performed at a variable number of revolutions. Theresults for two (nA = 1000, 4000 rpm) rotational frequencies areshown in Fig. 6. First, it is revealed that with increasing impeller speedthe particle size decreases. This is identical with the results pointedout in [24–27] for binder liquids with low viscosities. The findingsshow the values measured by the inline particle probe and twoCamsizer® particle size distributions. Whereas one represents the dis-tributions in the whole vessel (not sieved), the other (sieved) is a cut-out of the distributions that reflect the metering capacity of the probe.The sieved distribution reveals the agreement of the measurement re-sults. In both cases (Fig. 6), there is a good agreement between thesieved and the inline distribution. This indicates that the probe displaysthe particle size, according to its range of measurement, independentlyof the agitator speed. However, the not sieved and the inline particlesize distribution depend on the rotating speeds. As already mentioned,the particle size distribution is very wide at a low agitator speed.Based on these experiments, it can be stated that the inline measure-ment requires an agitator speed of nA ≥ 2000 rpm to represent thewhole vessel content.

As the installation space is very small, there is only one position tofixthe probe in the vessel. For this reason, the orientation of the entrancewindow is varied. As specified by the manufacturer and according tofirst reports of experience, a probe position (entrance window) of 90°relative to the main product flow has proved to be good in fluidised

Fig. 5. Influence of mass-relatedmoisture xW on the powder/agglomerate as a function of themlevel φ = 35%, rotating speed of agitator nA = 4000 rpm, and mixing tool: Star agitator.

beds. Three different probe positions (shown in Fig. 7) were evaluated.Two main product streams can be distinguished in the vessel. The firststream is from the agitator and the second from the wall scraper. Atthe first probe position, the entrance window is aligned directly to theagitator stream. Consequently, the second flow is located at an angleof 90°. At probe position two, it is exactly the opposite: The entrancewindow is directly aligned to the wall scraper stream and the firststream is oriented at 90° against the probe. At probe position three,only thewall scraper stream is important. The influence of the probe po-sition is examined as a function of filler loading and agitator speed; seeFig. 8. The diagrams show the median value x50 versus filler loading foreach probe position. At an agitator speed of nA = 4000 rpm, it is of noimportance whether the sample is sieved or not (see explanations “ag-itator rotating speed”). It is evident from Fig. 8 that there is a very goodagreement between the Camsizer® and the probe for position 2. Obvi-ously, the combination of agitator stream and wall scraper stream is

edian value x50 and the scattering parameter x90/x10 under constant test conditions: filling

Fig. 7. Probe positions: first position= the entrancewindow is oriented towards the agitator (left), second position= the entrancewindow is oriented towards thewall scraper (centre),and third position = the entrance window is oriented 90° relative to the main product flow from the wall scraper (right).

Fig. 8.Median value x50 as a function of filler loading for probe positions 1, 2, and 3. Conditions: xW = 12–13%, nA = 4000 rpm, star agitator, sample preparation: Sieved.

183V. Wenzel, H. Nirschl / Powder Technology 269 (2015) 178–184

so suitable that the whole vessel content is represented. At position 1,the entrance window was blocked very often. It was assumed that thewater added to the mixing zone reached the probe and the materialclotted near to the entrance window. In the case of probe position 3,the small undertow generated by the difference between internal andexternal purge air is not enough to attract the major agglomerates.This might explain the differences between the Camsizer® and the par-ticle probe at position 3. In general, it is expected that the filler loadinghas an influence on the results. With increasing filling quantity, the fre-quency and the probability of a particle experiencing stress decrease.According to Fig. 8, no large differences are found for the consideredquantities in the mixing vessel. In the context of filler loading and con-tinuous particle size determination it is important, however, that the

Fig. 9.Median value x50 and scattering parameter x90/x10 as a function of mixing tool andspeed.

filler level is high enough for an acceptable amount of particles passingthemeasurement volumeduring the agglomeration process. Otherwise,filler loading should not be too high, as this might cause problems withthepurge air. The probe should be able to “breathe”. The positioning andconfiguration of the inline particle probe are investigated in a fluid bedgranulator by Roßteuscher-Carl et al. [19] too. They also observed that,depending on the probe positioning, different agreements with anoffline methodology exist. For a successful start-up of new materialsthey recommend an appropriate validation. Fig. 9 shows how themixing tools influence themedian value x50 and the scattering parame-ter x90/x10. Fig. 10 depicts the corresponding particle size distributionsfor two different agitator speeds. From both diagrams (Figs. 9 and 10)it can be seen that at low agitator speeds, the distribution is much

Fig. 10. Particle size distribution depending on the rotating speed and the mixing tool.

184 V. Wenzel, H. Nirschl / Powder Technology 269 (2015) 178–184

wider for the pin agitator (x90/x10 high). Themedian value is larger thanfor the star agitator. With increasing energy input, the difference in themedian value is reduced. In the end, themedian value is nearly identical.The scattering parameter is smaller for the star agitator. The medianvalue is nearly constant for the star agitator.

4. Conclusions

It was demonstrated by the investigations performed that the inlineparticle probe is suited for measurements in a high-shear mixing pro-cess. Identification of the best conditions has been very intensive andmany experiments have been needed, because a lot of parameters andsettings interact with each other. After testing the inline particleprobe, the corresponding software and software parameters were ex-amined. Important software parameters of the IPP-70S are the “particlebuffer” and “loading”. The start of data acquisition and processing pa-rameters (e.g. agitator speed, probe orientation) are considered, too.The effect of the agitator rotating speeds on the particle size showsthat with increasing agitator velocity, the particle size decreases. Itwas found out by our study that a probe position directly in the mainproduct stream from the wall scraper is recommended to achieve agood agreement between the measurement results and the actual con-tent of the vessel. Of course, the measurement system reaches its limitswhen the combination of product and binder is gluey or a lowmechan-ical stress leads to big agglomerates (N2800 μm). Taking into account allfindings obtained and reported above, it can be stated that real-timeparticle size determination by means of particle probe IPP-70S in ahigh-shear mixer is possible for the products considered.

Symbols used

DA

[mm] Diameter of mixing tool (agitator) Q3 [%] Cumulative distribution nA [rpm] Agitator speed nV [rpm] Vessel speed SV [m2/cm3] Specific surface tp [s] Pulse width vp [m/s] Particle velocity x10 [mm] Particle size of the granulate size distribution at Q3 = 10% x50 [mm] Particle size of the granulate size distribution at Q3 = 50% x90 [mm] Particle size of the granulate size distribution at Q3 = 90% x90/x10 [−] Scattering parameter xp [mm] Particle diameter xW [%] Mass-related (agglomerate) moisture

Greek symbols

ρS [kg/m3] Solid density φ [%] Mass-related filling level

Acknowledgements

The authors owe special thanks to their industrial project partners“Maschinenfabrik Gustav Eirich GmbH & Co KG” and “ParsumGesellschaft für Partikel-, Strömungs- und Umweltmeßtechnik mbH”as well as to BMWi (German Federal Ministry of Economics) for finan-cial support within the framework of ZIM (Central Innovation ProgramMedium-Sized Businesses) (KF2066603KI9).

References

[1] Guidance for Industry PAT—A Framework for Innovative Pharmaceutical Develop-ment, Manufacturing, and Quality Assurance, in: F.a.D.A. U.S. Department ofHealth and Human Services (Ed.), 2004.

[2] S. Schmidt-Lehr, H.-U. Moritz, K.C. Jürgens, Online-Kontrolle der Partikelgrößewährend der Wirbelschichtgranulation, Pharm. Ind. 69 (2007) 478–484.

[3] T. Närvänen, T. Lipsanen, O. Antikainen, H. Räikkönen, J. Heinämäki, J. Yliruusi,Gaining fluid bed process understanding by in-line particle size analysis, J. Pharm.Sci. 98 (2009) 1110–1117.

[4] Nara Machinery Co. Ltd., Granulate vergleichmäßigen, Schüttgut, 162010. 172–713.[5] Frewitt Fabrique de Machines SA, Über den Mahlprozess immer im Bild, Schüttgut,

162010. 118–119.[6] P. Holm, T. Schaefer, H.G. Kristensen, Granulation in high-speed mixers Part VI.

Effects of process conditions on power consumption and granule growth, PowderTechnol. 43 (1985) 225–233.

[7] G. Imanidis, Untersuchung über die Agglomerationskinetik und die elektrischeLeistungsaufnahme beim Granulierprozess im Schnellmischer, Universität Basel,Basel, 1986.

[8] P. Knight, Challenges in granulation technology, Powder Technol. 140 (2004)156–162.

[9] M. Usteri, Untersuchung über das Agglomerationsverhalten pharamazeutischerHilfsstoffe im Schnellmischer, Universität Basel, Basel, 1988.

[10] M. Usteri, H. Leuenberger, Agglomeration of binary mixtures in a high-speed mixer,Int. J. Pharm. 55 (1989) 135–141.

[11] T.M. Chitu, D. Oulahna, M. Hemati, Wet granulation in laboratory scale high shearmixers: effect of binder properties, Powder Technol. 206 (2011) 25–33.

[12] A. Johansen, T. Schæfer, Effects of interactions between powder particle size andbinder viscosity on agglomerate growth mechanisms in a high shear mixer, Eur. J.Pharm. Sci. 12 (2001) 297–309.

[13] S.T. Keningley, P.C. Knight, A.D. Marson, An investigation into the effects of binderviscosity on agglomeration behaviour, Powder Technol. 91 (1997) 95–103.

[14] G.I. Tardos, M.I. Khan, P.R. Mort, Critical parameters and limiting conditions in bind-er granulation of fine powders, Powder Technol. 94 (1997) 245–258.

[15] S.M. Iveson, J.D. Litster, Fundamental studies of granule consolidation part 2: quan-tifying the effects of particle and binder properties, Powder Technol. 99 (1998)243–250.

[16] S.M. Iveson, J.D. Litster, K. Hapgood, B.J. Ennis, Nucleation, growth and breakage phe-nomena in agitated wet granulation processes: a review, Powder Technol. 117(2001) 3–39.

[17] A. Realpe, C. Velázquez, Growth kinetics and mechanism of wet granulation in alaboratory-scale high shear mixer: effect of initial polydispersity of particle size,Chem. Eng. Sci. 63 (2008) 1602–1611.

[18] D. Petrak, S. Dietrich, G. Eckardt, M. Köhler, In-line particle sizing for real-time pro-cess control by fibre-optical spatial filtering technique (SFT), Adv. Powder Technol.22 (2011) 203–208.

[19] K. Roßteuscher-Carl, S. Fricke, M.C. Hacker, M. Schulz-Siegmund, In-line monitoringof particle size in a fluid bed granulator: investigations concerning positioning andconfiguration of the sensor, Int. J. Pharm. 466 (2014) 31–37.

[20] P. Dieter, D. Stefan, E. Günter, K. Michael, In-line particle sizing for real-time processcontrol by fibre-optical spatial filtering technique (SFT), Adv. Powder Technol. 22(2011) 203–208.

[21] A. Bluma, T. Höpfner, P. Lindner, C. Rehbock, S. Beutel, T. Scheper, ProzessanalytischeIn-situ-Sonden zur Inline-Charakterisierung dispergierter Partikel, Chem. Ing. Tech.82 (2010) 517–524.

[22] Parsum GmbH, Bedienungsanleitung In-line Particle Probe IPP 70.[23] B. Daumann, X. Sun, H. Anlauf, S. Gerl, H. Nirschl, Mixing agglomeration in a high-

shear mixer with a stirred mixing vessel, Chem. Eng. Technol. 33 (2010) 321–326.[24] T.M. Chitu, D. Oulahna, M. Hemati, Wet granulation in laboratory-scale high shear

mixers: effect of chopper presence, design and impeller speed, Powder Technol.206 (2011) 34–43.

[25] J.S. Ramaker, M.A. Jelgersma, P. Vonk, N.W.F. Kossen, Scale-down of a high-shearpelletisation process: flow profile and growth kinetics, Int. J. Pharm. 166 (1998)89–97.

[26] T. Schæfer, B. Taagegaard, L.J. Thomsen, H. Gjelstrup Kristensen, Melt pelletization ina high shear mixer. IV. Effects of process variables in a laboratory scale mixer, Eur. J.Pharm. Sci. 1 (1993) 125–131.

[27] C. Mangwandi, M.J. Adams, M.J. Hounslow, A.D. Salman, Effect of impeller speed onmechanical and dissolution properties of high-shear granules, Chem. Eng. J. 164(2010) 305–315.