Q: Is Model-Less Predictive Control Any Good?
I receive InTech/ISA Review, and recently I read an article on multivariable predictive control using models by Allan Kern, and I do not agree with his analysis on model-less strategies. I have implemented many multivariable predictive controllers on complex plants (SMOC2/Shell license) where both the ROI and the performance was excellent in the long run. Obviously competent process control engineers had to review and adjust, if necessary, the model's parameters, but the control algorithm is very robust, and the needs for adjustment are not so frequent even with non-linear processes.
Typically the control algorithm is robust using state-variable theory (state estimators, Kalman filters, observers) to arrive at black and gray models. This means that for good performance more (not fewer) models are needed, and the models-less views, as told by Kern are incorrect.
What do you think of model-less use in predictive control?
See Allan Kern's answer in the comments at the bottom of this article and read "Take the path to model-less multivariable control" to learn more.
A: You are right. In order to control/optimize a process, one must understand it.
A: I agree with your assessment. In our book Advanced Control Foundation we provide many examples where MPC has produced amazing results. These results would not be possible without an explicit knowledge of the process; i.e., control being based on the process model.
Q: Referring to Equation 8.6(2) in your handbook Process Control and Optimization, page 1575, my application involves a DP type flow element and the following operating conditions:
Measured flow (F) 100 kg/s
Measured pressure (P) 100 barg
Absolute pressure (Po) 1.01325 bar
Reference pressure (PR) 101.01325
Actual temperature (T) 500 °C (773 .15 ºK)
Ref. temperature (TR) 813.15 ºK
Density (ρ) 30.13869861 [P = 98.98675bar (00barg) T = 500 °C]
Specific gravity (SG) 25.01136814
(The density of air at NTP = 1.205 kg/m3 and density/1.205)
Reference density (ρ R) 28.18237295 [P = 98.98675 bar a (100barg) T = 540 °C]
Ref. Spec. gravity (SGR) 23.38786137
(The density of air at NTP = 1.205 kg/m3 and ref density/1.205
Compressibility (X) 0.920440391 [P = 98.98675 bar (100barg) T=500 °C]
Ref. compressibility (XR) 0.935913474 [P = 98.98675bar (100barg) T = 540 °C]
Actual steam quality (Q) 100 (steam temperature is above saturation, quality is 100%)
Ref. steam quality (QR) 100 (steam temperature is above saturation, quality is 100%)
I am interested in calculating the compensated flow (FC) in kg/, but depending on which equation, I use the results are different.
If I use specific gravity (calculated from measured steam pressure and temperature) for flow compensation, then
(#1) FC = F*ρ /ρ R = 100*1.069417 = 106.9417 kg/s
If I use steam pressure and temperature for flow compensation, then
(#2) FC = F*(P+P0)/PR*TR/(T+T0) = 100*1.051736 = 105.1736 kg/s
If I use steam pressure, temperature, compressibility (calculated from measured steam pressure and temperature) and quality(100%) for flow compensation, then
(#3) FC = (P+P0)/RP*RT/(T+T0)*X/XR*QR/Q = 100*1.034348 = 103.4348 kg/s
When steam pressure OR steam temperature decreases, the steam density increases and compressibility drops. Decreasing in steam pressure OR increasing steam temperature, reduces steam density and increases compressibility.
In my example the temperature 40 °C dropped from design, therefore density and specific gravity increased, whereas compressibility decreased. For testing purposes I have modified Equation 8.6(2) of page 1575.
If I modify the equation and use the one below, then:
(#4) FC = (P+P0)/RP*RT/(T+T0)*XR/X*QR/Q = 100*1.069417 = 106.9417 kg/s
In this case the result of equation with specific gravity compensation (#1) and the modified equation (#4) with pressure, temperature & compressibility is exactly matching.
Please advise if equation 8.6 (2) in your handbook needs to be corrected OR my modified equation is wrong.
A: Usually the problem is that the compressibility and/or quality (X & Q) terms are not accurately known. I believe that Equation 8.6 (2) is correct, but I am asking Drs. Cheng, Meeker and Liu (or any other expert) to double-check and let you know.
A: You are making the assumption that steam follows the ideal gas law. Since it doesn't, this might explain the discrepancies you have noted (I have not tried to check your calculations, but it would not surprise me).
In the old days we had to use multipliers/dividers to do P/T compensation like you have. Modern control systems have significantly superior capabilities. You can use steam look- up tables and directly use density and compressibility in the compensation equation.
I have not looked under the hood of the recent DCS but I strongly suspect that they already have standard built-in compensation blocks for steam.
Q: Rotary Valve Torque Calculation
How can I calculate the torque of a rotary valve without using manufacturer data? Is there any general rule for the torque calculation of rotary valves? I need this for actuator sizing. Can you provide me with a good reference on this subject?
A: Don't try to do this without using the valve manufacturer's data because the torque values are critically dependent on the valve design, the seal material, the friction factor of the ball to seal, process pressure, frequency of operation, dryness or lubricity of process fluid, etc. And the maximum allowable stem torque depends on stem material and design. The calculation is hard enough if you use the manufacturer's data, and the risk is high that without it, you might undersize or oversize the actuator.
A: Fisher Control Valve Handbook (http://www.pacontrol.com/download/Emerson- Control-Valve-HandBook.pdf ), Rotary Actuator Sizing (pp 132) has a discussion including “Torque Equations” and two practical tables for easy parameters' selection.
Rotary Actuator Applications Guide (http://www.parker.com/literature/Literature%20Files/euro_cylinder/v4/htr-app_1230-uk.pdf ), has a special section dealing with “Calculating Torque Requirements” (p.5,) as well as some examples.
Rotary Actuator Sizing, Valdisk and Valdisk 150, (http://www.flowserveperformance.com/performhelp/sizing_selection:valtek:rotary_actuator_sizing), Steps 3-6 are dealing with some forms of torque calculations.
If these are not enough, feel free to contact me directly.
Q: Pulse Input Symbols
What symbols are used for pulse inputs to PLC/DCSs in flowsheets or other schematic?
Hiten A. Dalal PE
A: There is no standard symbol for PLC / DCS schematics. Each vendor provides a generic function block to treat the signals. By the way you have not mentioned what physical parameter is being measured with the pulses.
A: There is no specific symbol used for pulse input or, for that matter, for any analog or digital input. ISA standard 5.1 is the document where these things are defined. I have looked at the most recent edition, which I think is 2009, and there is no provision to show the technology used to make a measurement. For example, there are many ways in which a flow can be measured: differential pressure, turbine, Coriolis, magnetic, ultrasonic, vortex shedding, target, etc. The flow diagram only shows these as an FT or flow transmitter.
Your pulse input must represent something with a pulse count, pulse duration or pulse frequency. Whatever it is, it's only the technology used to measure something like flow, level, temperature, whatever. The symbol is xT where x is the transmitted property. You are permitted to supply additional notation (words or illustration) on a P&I Diagram if you want to identify the type of transmitter.
A: In ANSI/ISA-5.1-2009, on page 36 shows the symbols to use whether its field, DCS, PLC or other, which should be column A or C. On page 46 you will find the interconnect signals, and you would be using # 6, and on page 70 and 71 you would use the binary symbol of #15 or #18, for the input symbol.
You need to list in the legend sheet the symbols used.
Also you could use the symbols on page 38 and 39 #1 or #7 with the corresponding notation of page 39.