26 Nov 12 Process Linearity, Integral Windup, CBE 491 / CBE 433 Linearity, Windup, & PID 26 Nov 12 Process Linearity, Integral Windup, PID Controllers 30 Mar 07 2 Apr 08 27 Mar 09

Process Linearity Test the Heat Exchanger process linearity by: Starting Loop Pro trainer Set %CO to 80% Make steps down (say 10% down) to the %CO Measure the response Calculate the process gain

K = -1.09 K = -0.69 K = 0.-45 K = -0.33 K = -0.26 K = -0.15 Adaptive Control ?

Integral (Reset) Windup “Windup” can occur if integral action present Most modern controllers have anti-windup protection If doesn’t have windup protection, set to manual when reach point of saturation, then switch back to auto, when drops below sat. level IE: LoopPro Trainer, select Heat Exchanger Set %CO to 90%; SP to 126; Kc to 1 %/deg C; Tau I to 1.0 min Set Integral with Anti-Reset Windup ON Change Set Point to 120 deg. C. (~10 min); then change back to 126 deg. C Repeat with controller at ON: (Integral with Windup)

Integral (Reset) Windup

In-Class PID Controller Exercise Tune the Heat Exchanger for a PID Controller: Use the built in IMC, and choose Moderately Aggressive Start Loop Pro trainer Tune at the initial %CO and exit temperature Compare PI with PID Compare PID with PID with filter

Advanced control schemes 26 Nov 12 Cascade Control: Ch 9 CBE 491 / CBE 433 Advanced control schemes 26 Nov 12 Cascade Control: Ch 9 30 Mar 07 2 Apr 08 27 Mar 09

Improve Feedback Control Disturbance must be measured before action taken ~ 80% of control strategies are simple FB control Reacts to disturbances that were not expected We’ll look at: Cascade Control (Master – Slave) Ratio Control Feed Forward

Cascade Control Control w/ multiple loops Used to better reject specific disturbances + - Take slow process: Split into 2 “processes” that can measure intermediate variable? + + - - Gp2 must be quicker responding than GP1. Inner (2nd-dary) loop faster than primary loop Outer loop is primary loop

Material Dryer Example MT % moisture MC steam Heat Exchger air blower T + -

Separate Gp into 2 blocks MT % moisture MC TT TC steam Heat Exchger air blower T + + - -

A cascade is comprised of two normal PID controllers The secondary loop is nested inside the primary loop.

Problem Solving Exercise: Heat Exchanger TC Single feedback loop. Suppose known there will be steam pressure fluctuations… steam TT Hot water Heat Exchger T Design cascade system that measures (uses) the steam pressure in the HX shell. PT steam TT Hot water Heat Exchger T

Temperature Control of a Well-Mixed Reactor (CSTR) Ti Responds quicker to Ti changes than coolant temperature changes.

Temperature Control of a Well-Mixed Reactor (CSTR) Use Cascade Control to improve control. Ti If Tout (jacket) changes it is sensed and controlled before “seen” by primary T sensor. Secondary Loop Measures Tout (jacket) Faster loop SP by output primary loop Primary Loop: Measures controlled var. SP by operator

Cascade Control Benefits: Disturbances in secondary loop corrected by 2ndary loop controller Flowrate loops are frequently cascaded with another control loop Improves regulatory control, but doesn’t affect set point tracking Can address different disturbances, as long as they impact the secondary loop before it significantly impacts the primary (outer loop). Challenges: Secondary loop must be faster than primary loop Bit more complex to tune Requires additional sensor and controller

Cascade Control Examples Distillation Columns Cascade Control Examples Objective: Regulate temperature (composition) at top and bottom of column

Heat Exchanger Furnace TP out Objective: Keep TP out at the set point T2 out Objective: Keep T2 out at the set point

In-Class Exercise: Cascade Control System Design What affects flowrate? Valve position Height of liquid P (delta P across valve) Design a cascade system to control level (note overhead P can’t be controlled)

In-Class Exercise: Cascade Control System Design Does this design reject P changes in the overhead vapor space?

Tuning a Cascade System Both controllers in manual Secondary controller set as P-only (could be PI, but this might slow sys) Tune secondary controller for set point tracking Check secondary loop for satisfactory set point tracking performance Leave secondary controller in Auto Tune primary controller for disturbance rejection (PI or PID) Both controllers in Auto now Verify acceptable performance

In-Class Exercise: Tuning Cascade Controllers Select Jacketed Reactor Set T cooling inlet at 46 oC (normal operation temperature; sometimes it drops to 40 oC) Set output of controller at 50%. Desired Tout set point is 86 oC (this is steady state temperature) Tune the single loop PI control Criteria: IMC aggressive tuning Use doublet test with +/- 5 %CO Test your tuning with disturbance from 46 oC to 40 oC

In-Class Exercise: Tuning Cascade Controllers Select Cascade Jacketed Reactor Set T cooling inlet at 46 oC (again) Set output of controller (secondary) at 50%. Desired Tout set point is 86 oC (as before) Note the secondary outlet temperature (69 oC) is the SP of the secondary controller Tune the secondary loop; use 5 %CO doublet open loop Criteria: ITAE for set point tracking (P only) Use doublet test with +/- 5 %CO Test your tuning with 3 oC setpoint changes Tune the primary loop for PI control; make 3 oC set point changes (2nd-dary controller) Note: MV = sp signal; and PV = T out of reactor Criteria: IAE for aggressive tuning (PI) Implement and with both controllers in Auto… change disturbance from 46 to 40 oC. How does response compare to single PI feedback loop?

Advanced control schemes CBE 491 / CBE 433 Advanced control schemes 26 Nov 12 Ratio Control: Ch 10 30 Mar 07 2 Apr 08 27 Mar 09

Ratio Control Special type of feed forward control A B Blending/Reaction/Flocculation A and B must be in certain ratio to each other

Ratio Control Possible control system: B A FC FY FC FY FT FT A B What if one stream could not be controlled? i.e., suppose stream A was “wild”; or it came from an upstream process and couldn’t be controlled.

Ratio Control Possible cascade control systems: A B A B “wild” stream FT Desired Ratio FY FC FT B “wild” stream A FT Desired Ratio This unit multiplies A by the desired ratio; so output = FY FC FT B

Ratio Control Uses: Constant ratio between feed flowrate and steam in reboiler of distillation column Constant reflux ratio Ratio of reactants entering reactor Ratio for blending two streams Flocculent addition dependent on feed stream Purge stream ratio Fuel/air ratio in burner Neutralization/pH

In-Class Exercise: Furnace Air/Fuel Ratio Furnace Air/Fuel Ratio model disturbance: liquid flowrate “wild” stream: air flowrate ratioed stream: fuel flowrate Minimum Air/Fuel Ratio 10/1 Fuel-rich undesired (enviro, econ, safety) If air fails; fuel is shut down Check TC tuning to disturbance & SP changes. PV Desired 2 – 5% excess O2 Disturbance var. TC Dependent MV TC output Ratio set point Independent MV

Advanced control schemes 26 Nov 12 Feed Forward Control: Ch 11 CBE 491 / CBE 433 Advanced control schemes 26 Nov 12 Feed Forward Control: Ch 11 30 Mar 07 2 Apr 08 27 Mar 09

FF must be done with FB control! Feed Forward Control steam Suppose qi is primary disturbance TC TT Heat Exchanger ? What is a drawback to this feedback control loop? ? Is there a potentially better way? steam Heat Exchanger What if Ti changes? FF Draw block diagram for this feedback control loop FT TT FF must be done with FB control!

Feed Forward and Feedback Control TC steam FF TY TY FT TT Heat Exchanger Block diagram: + + + + + -

Feed Forward Control + + + + + - Response to MFF No change; perfect compensation!

Feed Forward Control + + + + + - Examine FFC T.F. + + For “perfect” FF control: + +

Feed Forward Control: FFC Identification Set by traditional means: Model fit to FOPDT equation: FF Gain Lead/lag unit Dead time compensator Often ignored; if set term to 1 Accounts for time differences in 2 legs { FFC ss } steady state FF control { FFC dyn } dynamic FF control

Feed Forward Control: FFC Identification How to determine FOPDT models : With Gc disconnected: Step change COFB, say 5% Fit C(s) response to FOPDT + + Still in open loop: Step change Q, say 5 gpm Fit C(s) response to FOPDT lead time lag time

Lead/Lag or Dynamic Compensator Look at effect of these two to step change in input Output or response Final Change from: Magnitude of step change, Initial response by the lead/lag, Exponential decay from lag,

+ - Feed Forward Control Rule of Thumb: if lead-lag won’t help much; use FFCss (p 389) In text: pp 393-395, useful comments if implementing FFC + - 1. Compensates for disturbances before they affect the process 1. Requires measurement or estimation of the disturbance 2. Can improve the reliability of the feedback controller by reducing the deviation from set point 2. Does not compensate for unmeasured disturbances 3. Offers advantages for slow processes or processes with large deadtime. 3. Linear based correction; only as good as the models; performance decreases with nonlinear processes. No improvement using FFC with set point changes.

In-Class PS Exercise: Feed Forward Control What is the Gm, and what is the GD? Determine FCC Tune PI controller to aggressive IMC For disturbance: Tjacket in 50oC – 60oC – 50oC Test PI Controller Test PI + FFCss only Test PI + FFC full

In-Class PS Exercise: Feed Forward Control PI only PI + FFCss only PI + full FFC

CBE 491 / CBE 433 30 Mar 07 2 Apr 08 27 Mar 09

Problem Solving Exercise: Heat Exchanger TC PC PT steam TT Hot water Heat Exchger T Draw the block diagram: what is the primary and what is the secondary loop? + + - -