process control improvement primer - greg mcmillan deminar

Interactive Opportunity Interactive Opportunity AssessmentAssessmentInteractive Opportunity Interactive Opportunity AssessmentAssessment

Demo and Seminar (Deminar) Series for Web Labs –

Process Control Improvement PrimerProcess Control Improvement PrimerSept 8, 2010

Sponsored by Emerson, Experitec, and MynahCreated by

Greg McMillan and Jack Ahlerswww.processcontrollab.com Website - Charlie Schliesser (csdesignco.com)

[File Name or Event]Emerson Confidential27-Jun-01, Slide 2 Slide 2

Welcome Welcome Gregory K. McMillan

– Greg is a retired Senior Fellow from Solutia/Monsanto and an ISA Fellow. Presently, Greg contracts as a consultant in DeltaV R&D via CDI Process & Industrial. Greg received the ISA “Kermit Fischer Environmental” Award for pH control in 1991, the Control Magazine “Engineer of the Year” Award for the Process Industry in 1994, was inducted into the Control “Process Automation Hall of Fame” in 2001, was honored by InTech Magazine in 2003 as one of the most influential innovators in automation, and received the ISA “Life Achievement Award” in 2010. Greg is the author of numerous books on process control, his most recent being Essentials of Modern Measurements and Final Elements for the Process Industry. Greg has been the monthly “Control Talk” columnist for Control magazine since 2002. Greg’s expertise is available on the web site: http://www.modelingandcontrol.com/


““Top Ten Things You Don’t Want to Hear During Startup”Top Ten Things You Don’t Want to Hear During Startup”Courtesy of Hunter Vegas (October 2010 Control Talk)Courtesy of Hunter Vegas (October 2010 Control Talk)

““Top Ten Things You Don’t Want to Hear During Startup”Top Ten Things You Don’t Want to Hear During Startup”Courtesy of Hunter Vegas (October 2010 Control Talk)Courtesy of Hunter Vegas (October 2010 Control Talk)

(10) We never really could figure out what the old system was doing. (9) Do I have a system backup?!? I thought YOU were making

backups! (8) They want to make our startup into a reality show. (7) The displays are fine and dandy but where are the panel boards? (6) We have changed our mind – we want the old system back. (5) Can you reprogram it so the wrong valve still works? (4) Didn’t you get the revised batch sheets? (3) Is a blue screen bad?? (2) What is that burning smell?

And the Number 1 thing you don’t want to hear:


““Top Ten Things You Don’t Want to Top Ten Things You Don’t Want to Hear During Startup” During Startup”Courtesy of Hunter Vegas (October 2010 Control Talk)Courtesy of Hunter Vegas (October 2010 Control Talk)

““Top Ten Things You Don’t Want to Top Ten Things You Don’t Want to Hear During Startup” During Startup”Courtesy of Hunter Vegas (October 2010 Control Talk)Courtesy of Hunter Vegas (October 2010 Control Talk)

(1) We are out of coffee!


IntroductionIntroduction There is no clear picture of what is the potential source

and size of a process control improvement Practical process control knowledge is detailed,

fragmented, and experience driven This seminar will attempt to provide a unified approach

and understanding of the impact of the PID, final control element (e.g. valve or variable speed drive), process, disturbance, and measurement on loop performance


Unifying ConceptsUnifying Concepts “It is all about management of change”

– 90% of process control improvements involve the following concepts: – Delay– Speed– Gain– Sensitivity-Resolution– Backlash-Deadband– Nonlinearity– Noise– Oscillations– Resonance– Attenuation– Optimum

Delay, speed, and gain are the most prevalent limiting concepts


DelayDelay “Without deadtime I would be out of a job” Fundamentals

– A more descriptive name would be total loop deadtime. The loop deadtime is the amount of time for the start of a change to completely circle the control loop and end up at the point of origin. For example, an unmeasured disturbance cannot be corrected until the change is seen and the correction arrives in the process at the same point as the disturbance.

– While process deadtime offers a continuous train of values whereas digital devices and analyzers offer non continuous data values at discrete intervals, these delays add a phase shift and increase the ultimate period (decrease natural frequency) like process deadtime.

Goals– Minimize delay (the loop cannot do anything until it sees and enacts change)

Sources– Pure delay from deadtimes and discontinuous updates

• Piping, duct, plug flow reactor, conveyor, extruder, spin-line, and sheet transportation delays• Digital devices - scan, update, reporting, and execution times (0.5T)• Analyzers - sample processing and analysis cycle time (1.5T)• Sensitivity-resolution limits• Backlash-deadband

– Equivalent delay from lags• Mixing • Column trays • Heat transfer surfaces• Thermowells• Electrodes • Transmitter damping • Signal filters


Speed (Rate of Change)Speed (Rate of Change) “Speed kills - (high speed processes and disturbances and low speed

control systems can kill performance)” Fundamentals

– The rate of change in 4 deadtime intervals is most important. By the end of 4 deadtimes, the control loop should have completed most of its correction. Thus, the short cut tuning method (Deminar #6) is consistent with performance objectives.

Goals– Make control systems faster and make processes and disturbances slower

Sources– Control system

• PID tuning settings (gain, reset, and rate)• Slewing rate of control valves and velocity limits of variable speed drives

– Disturbances• Steps - Batch operations, on-off control, manual actions, SIS, startups, and shutdowns• Oscillations - limit cycles, interactions, and excessively fast PID tuning• Ramps - reset action in PID

– Process• Mixing in volumes due to agitation, boiling, mass transfer, diffusion, and migration


GainGain “All is lost if nothing is gained” Fundamentals

– Gain is the change in output for a change in input to any part of the control system. Thus there is a gain for the PID, valve, disturbance, process, and measurement. Knowing the disturbance gain (e.g. change in manipulated flow per change in disturbance) is important for sizing valves and feedforward control.

Goals– Maximize control system gains (maximize control system reaction to change) and

minimize process and disturbance gains (minimize process reaction to change). Sources

– PID controller gain – Inferential measurements (e.g. temperature change for composition change in

distillation column) – Slope of control valve or variable speed drive installed characteristic (inherent

characteristic & system loss curve)– Measurement calibration (100% / span). Important where accuracy is % of span– Process design– Attenuation by volumes (can be estimated)– Attenuation by PID (transfer of variability from controlled to manipulated variables)


Sensitivity-Resolution Sensitivity-Resolution “You cannot control what you cannot see” Fundamentals

– Minimum change measured or manipulated - once past sensitivity limit full change is seen or used but resolution limit will quantize the change (stair step where the step size is the resolution limit). Both will cause a limit cycle if there is an integrator in the process or control system.

Goals– Improve sensitivity and resolution

Sources– In measurements, minimum change detected and communicated (e.g. sensor

threshold and wireless update trigger level) and quantized change (A/D & D/A)

– Minimum change that can be manipulated (e.g. valve stick-slip sensitivity and speed resolution)


Backlash-DeadbandBacklash-Deadband “No problem if you don’t ever change direction” Fundamentals

– Minimum change measured or manipulated once the direction is changed - once past backlash-deadband limit full change is seen or used. Both will cause a limit cycle if there are 2 or more integrators in the process or control system.

Goals– Minimize backlash and deadband

Sources– Pneumatic instrument flappers, links, and levers (hopefully these are long gone)

– Rotary valve and damper links, connections, and shaft windup

– Variable speed drive setup parameter to eliminate hunting and chasing noise


NonlinearityNonlinearity “Not a problem if the process is constant, but then again if the process

is constant, you do not need a control system” Fundamentals

– While normally associated with a process gain that is not constant, in a broader concept, a nonlinear system occurs if a gain, time constant, or delay changes anywhere in the loop. All process control systems are nonlinear to some degree.

Goals– Minimize nonlinearity

Sources– Control valve and variable speed drive installed characteristics (flat at high flows)

– Process transportation delays (inversely proportional to flow)

– Digital and analyzer delays (loop delay depends upon when change arrives in discontinuous data value update interval)

– Inferred measurement (conductivity or temperature vs. composition plot is a curve)

– Logarithmic relationship (glass pH electrode and zirconium oxide oxygen probe)

– Process time constants (proportional to volume and density)


NoiseNoise “The best thing you can do is not react to noise” Fundamentals

– Extraneous fluctuations in measured or manipulated variables Goals

– Minimize size and frequency of noise and do not transfer noise to process Sources

– Bubbles

– Concentration and temperature non-uniformity from imperfect mixing

– Electromagnetic interference (EMI)

– Ground loops

– Interferences (e.g. sodium ion on pH electrode)

– Velocity profile non-uniformity

– Velocity impact on pressure sensors


OscillationsOscillations “Oscillations are best kept in control theory textbooks” Fundamentals

– Sine wave, square wave, and saw-tooth periodic disturbances perpetually upset a system and can get amplified by resonance.

Goals– Minimize source and attenuate by controller tuning and process design

Sources– Limit cycles from sensitivity-resolution and backlash-deadband

– On-off control (common for sump level control by switches)

– Aggressive tuning (common for reactor temperature control)

– Excessive reset action (common for level and other integrating processes)


ResonanceResonance “Don’t make things worse than they already are” Fundamentals

– Oscillation period close to ultimate period can be amplified by feedback control. Goals

– Make oscillation period slower or control loop faster Sources

– Control loops in series with similar loop deadtimes (e.g. multiple stage pH control)

– Control loops in series with similar tuning and valve sticktion and backlash

– Day to night ambient changes to slow loops (e.g. column temperature control)


AttenuationAttenuation “If you had a blend tank big enough you would not need control” Fundamentals

– Attenuation increases as the volume of the blend tank increases and the ultimate period of the control loop decreases.

Goals– Maximize attenuation by increasing volume and mixing and making loops faster

Sources– Mixed volume size and degree of mixing

– Control loop speed


OptimumOptimum “Most setpoints are not at their optimum” Fundamentals

– The primary loop setpoint is offset from the optimum temperature, pressure, or concentration.

Goals– Minimize offset from optimum

Sources (of non-optimum operation)– Process variability

– Measurement error

– Sensitivity-resolution

– Backlash-deadband

– Lack of process knowledge

– Process nonlinearity (e.g. catalyst degradation and production rate changes)

– Operator preference (e.g. sweet spots)

– Incorrect SIS settings


Time (seconds)

% Controlled Variable (CV) or

% Controller Output (CO)

CO

CV

o p2

Kp = CV CO

CV

CO

CV

Self-regulating processopen loop

negative feedback time constant

Self-regulating process gain (%/%)

Response to change in controller output with controller in manual

observed total loopdeadtime

Self-Regulating Process Open Loop Response

Self-Regulating Process Open Loop Response

oor

Maximum speedin 4 deadtimes

is critical speed


Integrating Process Open Loop ResponseIntegrating Process

Open Loop Response

Time (seconds)o

Ki = { [ CV2 t2 ] CV1 t1 ] } CO

CO

ramp rate isCV1 t1

ramp rate isCV2 t2

CO

CV

Integrating process gain (%/sec/%)

Response to change in controller output with controller in manual% Controlled Variable (CV)

or% Controller Output (CO)

observed total loopdeadtime


is critical speed


Runaway Process Open Loop Response

Runaway Process Open Loop Response

Response to change in controller output with controller in manual

o ’p2

Noise Band

Acceleration

CV

CO

CV

Kp = CV CO Runaway process gain (%/%)

% Controlled Variable (CV) or

% Controller Output (CO)

Time (seconds)observed total loopdeadtime

runaway processopen loop

positive feedback time constant

For safety reasons, tests are terminated after 4 deadtimes

’oor


is critical speed


Loop Block Diagram(First Order Approximation)

Loop Block Diagram(First Order Approximation)

p1 p2 p2 Kpvp1

c1 m2 m2 m1 m1Kcvcc2

Kc Ti Td

Valve Process

Controller Measurement

Kmvvv

KLLL

Load Upset

CV

CO

MVPV

PID

Delay Lag

Delay Delay Delay

Delay

Delay

Delay

Lag Lag Lag

LagLagLag

Lag

Gain

Gain

Gain

Gain

LocalSet Point

DV

First Order Approximation: ov p1 p2 m1 m2 c

vp1m1m2c1 c2

%

%

%

Delay => Dead TimeLag =>Time Constant

Ki = Kmv(Kpv / p2 ) Kcv

100% / span


CV change in controlled variable (%) CO change in controller output (%) Kc controller gain (dimensionless) Ki integrating process gain (%/sec/% or 1/sec) Kp process gain (dimensionless) also known as open loop gain MV manipulated variable (engineering units) PV process variable (engineering units) t change in time (sec) ts sample time (sec) ototal loop dead time (sec) ffilter time constant (sec) mmeasurement time constant (sec) p2primary (large) self-regulating process time constant (sec) ’p2primary (large) runaway process time constant (sec) p1secondary (small) process time constant (sec) Ti integral (reset) time setting (sec/repeat) Td derivative (rate) time setting (sec) To oscillation period (sec) Lambda (closed loop time constant or arrest time) (sec) fLambda factor (ratio of closed to open loop time constant or arrest time)

Nomenclature Nomenclature


Impact of Fast and Slow DisturbancesImpact of Fast and Slow Disturbances Objective – Show the effect of disturbance speed Activities:

– For Single Self-Regulating Loop:• Review fast upset test (primary upset lag and reset time = 6 seconds)

• Increase primary upset lag to 60 seconds

• After about 5 minutes review slow load upset test results


Practical Limit to Loop PerformancePractical Limit to Loop Performance

ocp

x EKK

E

)1(

1

ocp

fxii E

KK

tTE

Peak error decreases as the controller gain increases but is essentially the open loop error for systems when total deadtime >> process time constant

Integrated error decreases as the controller gain increases and reset time decreases but is essentially the open loop error multiplied by the reset time plus signal delays and lags for systems when total deadtime >> process time constant

Peak and integrated errors cannot be better than ultimate limit - The errors predictedby these equations for the PIDPlus and deadtime compensators cannot be better

than the ultimate limit set by the loop deadtime and process time constant


Ultimate Limit to Loop PerformanceUltimate Limit to Loop Performance

opo

ox EE

)(

opo

oi EE

)(

2

Peak error is proportional to the ratio of loop deadtime to 63% response time

Integrated error is proportional to the ratio of loop deadtime squared to 63% response time

For a sensor lag (e.g. electrode or thermowell lag) or signal filter that is much largerthan the process time constant, the unfiltered actual process variable error can be

found from the equation for attenuation


Disturbance Speed and AttenuationDisturbance Speed and Attenuation

oL EeE Lo )1( /

f

oof

TAA

2*

Effect of load disturbance lag (L) can be estimated by replacing the open loop error with the exponential response of the disturbance during the loop deadtime

The attenuation of oscillations van be estimated from the expression of the Bode plot

equation for the attenuation of oscillations slower than the break frequency where (f) is the filter time constant, electrode or thermowell lag, or a mixed volume residence time


Implied Deadtime from Slow TuningImplied Deadtime from Slow Tuning

)(5.0 oi

Slow tuning (large Lambda) creates an implied deadtime where the loop performsabout the same as a loop with fast tuning and an actual deadtime equal to the

implied deadtime (i)


Effect of Implied Deadtime onAllowable Digital or Analyzer Delay

Effect of Implied Deadtime onAllowable Digital or Analyzer Delay

In this self-regulating process the original process delay (dead time) was 10 sec. Lambda was 20 sec and the sample time was set at 0, 5, 10, 20, 30, and 80 sec (Loops 1 - 6)

The loop integrated error increased slightly by 1%*sec for a sample time of 10 sec which corresponded to atotal deadtime (original process deadtime + 1/2 sample time) equal to the implied deadtime of 15 seconds.

http://www.modelingandcontrol.com/repository/AdvancedApplicationNote005.pdf

sample time = 0 sec

sample time = 5 sec

sample time = 10 sec




Effect depends on tuning, which leads to miss-guided generalities based on process dynamics


Fastest Practical PID Tuning Settings(Practical Limit to Loop Performance) Fastest Practical PID Tuning Settings(Practical Limit to Loop Performance)

op

pc KK

24.0oiT 2 1d pT

For runaway processes:

For self-regulating processes:

oic KK

15.0

oiT 4 1d pT

oic KK

16.0

oiT 40 1d 2 pT

For integrating processes:

op

pc KK

2'6.0

oic KK

14.0 short cut tuning method (near integrator approximation)

short cut tuning method (near integrator approximation)


Effect of Tuning Speed on Oscillatory Disturbance

Effect of Tuning Speed on Oscillatory Disturbance

1

UltimatePeriod

1

1FasterTuning

Log of Ratio ofclosed loop amplitudeto open loop amplitude

Log of ratio ofdisturbance periodto ultimate period

no attenuationof disturbances

resonance (amplification) of disturbances

amplitude ratio isproportional to ratio ofbreak frequency lag to

disturbance period

1

no better than manual worse than manual improving control


Visit Visit http://www.processcontrollab.com/ to Create Valuable New Skillsto Create Valuable New SkillsVisit Visit http://www.processcontrollab.com/ to Create Valuable New Skillsto Create Valuable New Skills

Free State of the Art Virtual Plant Independent Interactive Study Learn in 10 minutes rather than 10 years Online Performance Metrics Standard Operator Graphics & Historian Control Room Type Environment No Modeling Expertise Needed No Configuration Expertise Needed Rapid Risk-Free Plant Experimentation Deeper Understanding of Concepts Process Control Improvement Demos Sample Lessons (Recorded Deminars)


Help Us Improve These Deminars!Help Us Improve These Deminars!

WouldYouRecommend.Us/105679s21/


Join Us Oct 13, Wednesday 10:00 am CDTJoin Us Oct 13, Wednesday 10:00 am CDT

PID Deadtime Compensation PID Deadtime Compensation (How to setup and tune a PID for deadtime compensation)

Look for a recording of Today’s Deminar later Look for a recording of Today’s Deminar later this week at:this week at:

www.ModelingAndControl.com

www.EmersonProcessXperts.com


QUESTIONS? QUESTIONS? QUESTIONS? QUESTIONS?

process control improvement primer - greg mcmillan deminar

Education

control loop

process control improvements

ph control

feedforward control

process deadtime

process industry

low speed control systems

final control element