Structure of discussion
Integration issues
• Control systems technology • Field wiring • Other application issues in closed loop control systems
Basic process control • • • •
Controllers Processes Measurement devices Actuators
- selecting controlled and manipulated variables - control performance measures for common input changes - other practical issues: filtering, PID algorithm, output processing.
• Integration issues • Empirical model building • PID controller tuning
1
Controller diagram of a typical control loop Actuator System F2
T1
T2
2
Components and signals of a typical control loop
1. Control systems technology
F1
• Overall control system objectives • Further reading • Case study: Measurement and control of hazardous materials. • Tutorial questions
F1
F2
T1
T2 Thermowell
3-15 psig
T
Sensor System Controller
TC
Air
I/P
4-20 ma
D/A
Thermocouple millivolt signal
T TT
Reference: http://www.che.ttu.edu/pcoc/software/ppt.htm
Operator Console
3
Tsp
DCS Control Computer
A/D
4-20 ma
Transmitter
4
Controllers • Pneumatic analogue controllers
Analogue control systems are built around the concept of a control panel
• Electronic analogue controllers
Technology: electronic, pneumatic, hydraulic
• Computer controllers - Distributed Control Systems (DCS) - Programmable controllers - Fieldbus technology. A survey of industry reported in Control Engineering magazine in July 2005 (http://www.controleng.com/article/CA622640.html) stated • 13% used pneumatic loop controllers • 40% used electronic single loop controllers • 21% used electronic multi-loop controllers • 25% used PC based control.
5
The panel incorporates a variety of components to support operator access, data recording and alarms
Reference: Process Control Special Short Course 2006 - Fisher-Rosemount Systems
6
Panel wiring is often a challenge to maintain …
7
8
Pneumatic controllers • Introduced in the 1920’s. • Installed in the field next to the valve. • Use bellows, baffles, and nozzles with an air supply to implement PID action. • Provided automatic control and replaced manual control for many loops. • Transmitter type pneumatic controllers began to replace field mounted controllers in the late 1930’s. • Controller located in control room with pneumatic transmission from sensors to control room and back to the valve. • Allowed operators to address a number of controllers from a centralized control room.
Pneumatic controller installation F1
F2
T1
T2 Thermowell
3-15 psig
T
Mid 1940’s pneumatic controller Reference: www.peci.org/library/PECI_ControlOverview1_100 2.pdf
1940’s control room. Each circle is a pneumatic PID controller.
Air
Tsp
Pneumatic Controller
Thermocouple millivolt signal 3-15 psig
Air Transmitter
9
Some comments on pneumatic technology • Pneumatic transmitters, controllers, indicators, recorders and panel mounted instruments are cheaper than their electronic counterparts. • Established technology. • More maintenance required then electronic technology (mechanical systems, air system – compressors etc). • Rated as 'passive devices' by the National Electrical Code (USA) and are intrinsically safe. • In the case of a power outage, pneumatic devices continue to run and record process conditions. • In conclusion, pneumatics remain in service, and new installations occur, due to - Real or perceived fear of introducing electricity into a control scheme. - Continued ease of maintaining and monitoring an existing system. - Prohibitive cost of changeover to newer control technology (4-20 mA) - Convenience of a readily available air supply.
• Technical implementation – see Ogata, K. (2002). Modern Control Engineering, Prentice-Hall. 11 Reference: Johnson, D. (1999). Pneumatic control: not dead yet, Control Engineering, July. http://www.controleng.com/article/CA192589.html
10
Electronic analogue controllers • Became common in the late 1950’s. • Replaced the pneumatic tubing with wires. • Used resistors, capacitors, and transistors based amplifiers to implement PID action. • Out sold pneumatic controllers by 1970. • Allowed for advanced PID control: ratio, feedforward etc.
F1
F2
T1
T2 Thermowell
3-15 psig
T Air
I/P
4-20 ma Thermocouple millivolt signal Tsp
Electronic Analog Controller
4-20 ma
Transmitter
12
Computer controllers
Distributed control systems (DCS)
• Originally based upon a mainframe digital computer. • Offered the ability to use data storage and retrieval, alarm functions, and process optimization. • First installed on a refinery in 1959. • Had reliability limitations. Video Display Unit
Alarming Functions
Printer
Supervisory Control Computer
Analog Control Subsytem
Interfacing Hardware
Data Storage Acquisition System
...
13
• Introduced in the late 1970’s. • Less expensive per loop for large plants and less expensive to expand. • Facilitates the use of advanced control. • The data storage and trending capability of a DCS greatly facilitates troubleshooting of control problems i.e. the sources of process upsets can be tracked down through the process by trending a group of process measurements until the source of the process upset is located.
System Consoles
Host Computer
Data Storage Unit
PLC
Data Highway (Shared Communication Facilities)
Local Console
Local Control Unit
Local ..............Control Unit
Local Console
Process Transmitters and Actuators 14
Many features of today’s control systems may be traced to early DCS technology
Distributed Control System (DCS) • Allowed control rooms to be centralised and control distributed throughout the plant • Built around custom hardware and interfaces 15
16
DCS operator interface
DCS controller cabinet
Early DCS operator interfaces were expensive to install and to maintain, as physical components were often custom designed and constructed.
• Contains controllers, I/O cards and termination panels • Physical layout and size is limited by available device and manufacturing technology • Redundant communications, controllers and I/O are supported.
17
18
Programmable logic controllers (PLC’s)
DCS – typical I/O supported
• PLC’s are used for discrete and continuous control. • Discrete control is used for startup and shutdown and batch sequencing operations. • Ladder logic is used to program PLC’s.
19
• Advantages of PLC over DCS: Withstands harsh operating environments better, faster cycle time are possible, easier to maintain due to modular nature, and lower cost, for small and medium sized applications. • Advantage of DCS over PLC: Lower cost per loop for applications involving a large number of control 20 loops.
2. Field Wiring
Fieldbus technology • Based upon smart valves and smart sensors. • Uses data highway to replace wires. • Can mix different vendors of sensors, transmitters and control valves.
Field wiring may be terminated to the junction boxes or panels and from there wired to the controller termination. There is often a requirement to interface intelligent devices e.g. PLC’s into the control system.
21
Local panels
22
Overview
• In some cases, local control is still required to allow field adjustments. • Typically, this is required in batch processing.
23
• The control system interfaces to the process through field devices (for measurement). • Our ability to control a process is limited by the accuracy of measurement devices and the resolution and deadband associated with actuators. • Incorrect selection of field devices directly impacts the quality of control. • Set-up of input and output blocks in a control systems requires some knowledge of the field devices used. 24
Field wiring – 2 wire devices
Field wiring – 4 wire devices
• Power for the device is supplied by the twisted pair. • Transmitters provide current proportional to measurement. • Current to the valve is changed by the controller. Based on actuator, transmitter or positioner setup, valve may either open or close with increasing current.
• Analysers and some types of flow and level transmitters are 4 wire devices. • Auxiliary power is required e.g. 220 V a.c. • Input to the controller should be electrically isolated from the device power and ground. 25
26
Field wiring
– Foundation Fieldbus devices
Field wiring – HART devices • HART protocol supports 2-way digital communications for measurement and control devices. • Allows remote process variable checking, parameter setting and diagnostics. • Communications signal is superimposed on top of the 4-20 mA signal from 2-wire or 4-wire transmitters.
• Devices may be connected from a single twisted pair. • All data is accessed by digital communications. • Measurement, calculations and control may be moved to the fieldbus device. 27
28
3. Other application issues in closed loop control systems
• In reality, many elements in the control loop affect safety, reliability, accuracy, dynamics (I.e. speed of response) and cost. Engineers need to understand the details. • For example: inputing a step change to the process (without feedback control):
In P&ID diagrams, sensor location, process variable measured, connection to the final control element and location of final control elements are typically shown.
Reference: Marlin, T.E. (2000). Process control, 2nd edition, McGraw-Hill, Chapters 7 and 12. 29
30
A step change of 1% of full scale deflection is made at the manual station. The output of the process (temperature) and the display output (at the manual station) are monitored.
Example: Consider the open loop dynamic responses of two process and instrumentation systems (i.e. responses without the feedback controller). The systems are identical, except for the process; they involve electronic transmission, a pneumatic valve, a first order plus deadtime process and a thermocouple in a thermowell.
Case A
31
Case B
• Clearly, what is displayed is not always what is occurring. • For Case A, the sensor and final control elements contribute significant dynamics. • Controller design based on the process alone would not be adequate 32 for Case A.
3.1 Selecting controlled and manipulated variables • Feedback control provides a connection between controlled and manipulated variables. • The engineer must decide what measurement to control and what valve to adjust. • Take a continuous flow chemical reactor example.
There are 7 categories of control objectives (below). Later, we will consider some of these in more detail. For the example, consider the product quality objective; the engineer decides that the most important process variable associated with product quality is the concentration of reactant A in the reactor effluent.
33
Now, the manipulated variable has to be selected. There are six ‘input’ variables that would affect the measured variable selected (concentration of reactant A in effluent). They are • Solvent feed temperature • Solvent flow rate • Feed composition before mixing • Flow of pure A • Flow of coolant • Coolant inlet temperature
34
This leaves two adjustable variables: • Flow of pure A • Flow of coolant. These variables are candidates for selection as the manipulated variable. The final choice of manipulated variable will depend on which variable best fulfils the following criteria: 1. 2. 3.
However, some of these variables cannot be adjusted easily i.e. they are due to changes in other plant units and/or environment outside the plant. They are classed as disturbances; in this application, they are •Solvent feed temperature •Solvent flow rate •Feed composition before mixing 35 •Coolant inlet temperature
4.
5.
Is there a causal relationship between controlled variable and proposed manipulated variable (required) – yes, in both cases Is there a valve available to adjust the proposed manipulated variable (required) - yes, in both cases Can the proposed manipulated variable be adjusted without upsetting the rest of the plant (desired) - yes, in both cases If the manipulated variable is chosen, can the largest disturbance be compensated when the feedback controller is designed (desired) – first identify (perhaps from operating data) the largest of the four disturbances (e.g. it could be feed composition before mixing). Answer: yes, in both cases To which manipulated variable does the controlled variable react quickest (desired). Theoretical modelling or experimental work would 36 be needed to answer this.
3.2 Control performance measures for common input changes
For the continuous flow chemical reactor example:
37
Set point change (servo response)
38
Disturbance (regulator) response
39
40
Disturbance response – stochastic system Often, a process experiences a continual stream of small and large disturbances, so that the process is never at an exact steady state.
41
Tutorial question
42
3.3 Other practical issues
Comment on the quality of control for the four responses shown.
43
44
3.3.1 Input processing - filtering
45
46
47
48
Perfect filter
Practical filter
• Analog (anti-aliasing) filter – time constant is typically small (few tenths of a second). • (Optional) digital filter – also has a small time constant. Typically, it is built by the engineer for each application.
49
Guidelines to reduce the effect of noise on feedback
50
3.3.2 PID algorithm
1. Reduce the amplification of noise by the control algorithm by putting the derivative time constant equal to zero. 2. Select a small filter time constant compared to the dynamics of the process being controlled. This prevents the filter from degrading the performance of the resulting closed loop control system. One rule of thumb: filter time constant < 5% of (process time delay + process time constant). 3. Reduce the noise effects on the manipulated variable by selecting the filter time constant > 5/(noise frequency). Often, step 2 takes priority over step 3.
51
52
53
Another alternative is proportional 54 band (PB).
Integral (reset) windup All actuators have physical limitations e.g. a control valve cannot be more than fully open or fully closed, a motor has limited velocity etc. This has severe consequences for control.
55
When the actuator saturates, the feedback loop is broken, as the saturating element is then not influenced by the output signal from the controller (in the saturation region). 56
Integral windup (continued)
Integral windup (continued) If the controller (or the process) contains an integrator, the non-zero error will continue to be integrated and the output of the controller will wind-up to a very large value (so-called “reset windup”). When the actuator de-saturates, it may take a long time for the system to recover; the actuator could also bounce several times between high and low values before the system recovers. SIMULATION: Filenames - ex_windup.mdl, exa_windup.mdl. In these simulations, 1 0.036 1 G p (s) = G c (s) = 0.271 + = 0.27 + s(s + 1) s 7.5s Saturation at a level of ± 0.1 is included. The process output (controlled variable), the control signal (manipulated variable) and the output of the integrator are all monitored.
57
Integral windup (continued)
58
Anti-windup: conditional integration Integral action is switched off when actuator saturates and switched on again when it desaturates. One implementation:
With saturation and windup Without saturation
controlled variable •Manipulated variable saturates immediately step is applied; feedback is broken. •Integrator output continues to increase (positive error). •Integrator output starts to decrease when controlled variable > set point, but manipulated variable remains saturated. •Manipulated variable finally decreases at t=14s.
manipulated variable
integrator output
59
60
Anti-windup – tracking
Anti-windup – tracking
controlled variable
With windup With anti-windup
manipulated variable
• Anti-windup parameter (typically a gain) has no effect when the actuator does not saturate. • When the actuator saturates, the output of the integral term is recalculated so that its new value gives an integral term output at the saturation limit.
integrator output
Reference: Airikka, P. (2003). The PID controller: algorithm and implementation, IEE Computing and Control Engineering Magazine, Vol. 14, 61 No. 6, pp. 6-11.
62
3.3.3 Output processing
Note limits on MV Failure modes
63
64
Tutorial questions
65
66
67
68
4. Overall control system objectives 4.1 Economics The effect of a process parameter on the final product depends on the plant limitations and the plant operating conditions. Production issues fall into two categories: • Achieving global production maximum for a given operating condition. • When production maximum occurs at an upper or lower bound for a given operating condition.
Global production maximum • Production is greatest when the “variation band” is reduced to zero and the process parameter is maintained at a value corresponding to maximum production. • The plant design conditions may be used as a guide in establishing setpoints for best operation.
69
Production maximum at limit
70
Example – Pressure control in an ammonia plant
• Here, maximum production is obtained by maintaining the process parameter at a limit determined by some plant limitation. • How close to the limit at which the process can be operated depends on the quality of the control. • Production improvement is obtained by operating closer to product specification or operating limit.
71
72
4.2 Safety and equipment protection • The control system must provide safe operation – Alarms, safety constraint control, start-up and shutdown.
• A control system must be able to “absorb” a variety of disturbances and keep the process in a good operating region: – Thunderstorms, feed composition upsets, temporary loss of utilities (e.g., steam supply), day to night variation in the ambient conditions
Safety example 1: Pressure control in a boiler • Boiler “draft” control is used to maintain negative pressure in a boiler. • If the pressure were to go positive, then hot gases from the boiler could blow back through access ports on operating people.
73
Safety example 2: Standpipe level control
74
5. Further reading
• Equipment protection. • Standpipe level must be maintained to avoid loss of liquid flow to the pump. • Loss of flow for an extended period of time may damage the pump.
75
• Marlin, T.E. (2000). Process Control, McGraw-Hill, Chapter 2 – Control objectives and benefits (chemical engineering approach) • Seborg, D.E., Edgar, T.F. and Mellichamp, D.A. (2004). Process Dynamics and Control, 2nd Edition, Wiley, Chapters 1 and 10 (chemical engineering approach). • Shinskey, F.G. (1994). Feedback controllers for the process industries, McGraw-Hill, Chapter 1 (control engineering approach) • Goodwin, G.C., Graebe, S.F. and Salgado, M.E. (2001). Control system design, Prentice-Hall, Chapter 1. • Gruhn, P. (2002). Designing instrumentation and control for process safety, Chemical Processing, May; available at http://www.chemicalprocessing.com/articles/2002/259.html 76
6. Case study: Measurement and Control of Hazardous Materials
Objectives • Introduce the Regulatory Environment • Outline CEMS (Continuous Emission Monitoring Systems) • Insight on Analysers • Outline New Philosophy – Predictive Emissions Monitoring Systems (PEMS)
Acknowledgements: This part of the presentation was prepared with the assistance of Jesus Caballo, Schering Plough (Avondale) Ltd., Rathdrum, Co. Wicklow. Schering Plough is a manufacturer of active pharmaceutical ingredients. 77
78
Towards EMS (Emission Monitoring Systems). Why?
Background
• Regulatory pressures Greenhouse Gases : Major Climate Change
– EU commitments to Kyoto Protocol • Reduction to 8% below 1990 levels • Emissions Trading
• If emissions continue, by 2100,
– Irish EPA –Integrated Pollution Prevention and Control (IPPC) license
– Global T will rise between 1.4 and 5.8 °C – Sea level will rise between 9 and 98 cm
• Social pressures
• Consequences:
• Unsafe and costly manual sampling
– Extinction or geographic shift of species – Changes in rainfall / extreme weather events
Reference: Environmental Protection Agency (2004). “Ireland Environment 2004: The State of the Environment”, Chapter 8. Available from http://www.epa.ie/NewsCe ntre/ReportsPublications/Ir elandsEnvironment2004/
• Major threats: – Acid rain – Urban air pollution – Toxic air emissions 79
80
In summary, the requirements are:
All images courtesy of Schering Plough except Flame ionisation All images courtesy of Schering Plough except Flame ionisation detectors (FID) and Paramagnetic Analysers: Thermomagnetical detectors (FID) and Paramagnetic Analysers: Thermomagnetical Measuring Principle [from Worthington, B. (2000). “Continuous Measuring Principle [from Worthington, B. (2000). “Continuous emissions advance with regulations monitoring technology”, Control emissions advance with regulations monitoring technology”, Control Engineering, February] Engineering, February]
Detectors and Analysers
– Inherent Safety – Preventive and protective measures
Nondispersive infrared analyzers (NDIR) Flame ionization detector (FID)
– Measure, monitor and report
UV analyzer
Can this be done? Yes !
CO
• New, economical and powerful equipment exists. • Measurement of multiple gases • Platform approach and application-specific • Interface to control system and computer networks » display of results » manual interactive operation » self-diagnosis and automatic fault recognition
SO2,NOx O2
Hydrocarbons
An analyser for each use
Paramagnetic analyzers Hazardous Pollutants Toxic waste
Hazardous Pollutants
Gas Chromatography Exotic Pollutants
81
82 Mass Spectrometer
Emission Monitoring Systems • CEMS: Continuous Emissions Monitoring System Equipment and programmes to analyze emissions, quantify the amount of specific compounds emitted and process the information for reporting
Fourier Transform Infrared Analyser (FTIR)
CEMS Technology Key requirements: –Reliable instruments for gas analysis –Adequate sampling and sample conditioning unit –Measuring and control systems
• PEMS: Predictive Emissions Monitoring System Predict the concentration and emission rate of a contaminant based on correlation(s) with other monitored parameters
83
84
Modular System Concept
TYPICAL CEMS Interface
with Fourier Transform Emission Analyser (FTIR) To Cope with: • Higher demands and stricter limitations • Growing number of pollutants to be measured • Processes have to be monitored exactly Advantage: • Ethernet interface -
Information obtained from PC; Remote operation and diagnosis
85
TYPICAL CEMS Interface - continued
86
Predictive Emissions Monitoring System (PEMS) Technology … predicts the concentration and emission rate of a contaminant based on correlation(s) with other monitored parameters Steps: – Establish the parameter to be measured – Obtain a model based on prior knowledge of • Process • Emissions
– Validate the model with data from analyser Basis: – Measure parameters instead of pollutants – If predicted levels remain under control, then there is confidence that actual emissions will also remain under control 87
88
PEMS Technology - continued
PEMS Technology - continued
89
90
PEMS Control System
PEMS – Pros and Cons
Example: Steam Furnace
• Pros: – Instruments at combustion end (less hostile media) – Better accuracy (less maintenance and corrosion) – Sensors in 'handier' location for calibration/ maintenance
Elements to be controlled: NOX , CO Elements measured: • Firing rate - Affects NOX • Fuel gas density • Furnace air preheat temperature • Excess of oxygen - Affects CO • Stack temperature • Inlet air humidity and temperature
• Challange: – Obtaining the model to accurately predict conditions for every process variable
91
92
7. Tutorial Questions
Conclusion CEMS can – Satisfy regulatory requirements – Boast production efficiencies – Decrease operating costs
Which of these element dynamics affect closed loop system performance ?
CEMS and PEMS technology is available … now engineers must ensure correct implementation. 93
• Every element in the closed loop affects the control performance, at least in theory.
• • • • • • •
Electrical transmission Signal conversion Pneumatic transmission Process Sensor Controller calculation Valve
Question
Answer
• Electrical Transmission/Signal conversion These elements are very fast and would not significantly affect the performance
• Pneumatic Transmission This element would be fast when the pneumatic line is short and in this situation, would not significantly affect the performance. • Process This element is usually the slowest element in the feedback loop and limits the performance.
94
The behaviour of the controlled variable, CV, is important because...
• Sensor This element could be fast (compared with the process dynamics) in many cases, but a sensor can introduce substantial dynamics, especially when a sample extraction or analysis is performed by the sensor.
• Controller calculation This element is usually fast (compared with the process dynamics), and the engineer must ensure that the digital computer can execute the calculation fast enough.
• Valve
This element is usually fast (compared with the process dynamics), but a valve can introduce substantial dynamics, especially for applications with 95 very fast processes.
• It was selected to achieve one or more of the seven control objectives • The sensor measures this value • Large fluctuations create messy trend plots. 96
Question
Answer • It was selected to achieve one or more of the seven control objectives Yes, this is the main purpose of feedback control. One of the key control objectives is to achieve the desired behaviour of the controlled variable.
• The sensor measures this value No, sensors within plants measure many different variables, including monitoring the process performance. Therefore, many sensors that are not used for control are located in a plant.
The behaviour of the manipulated variable, MV, is important because...
• Large fluctuations create messy trend plots. No. We like to see smooth trend plots, but engineers must look beyond the appearance of the plot. For most controlled variables, significant variation (deviation from set point) results in safety, product quality, or profit degradation.
• • • •
Large fluctuations might damage equipment High frequency fluctuations might wear out the valve Large fluctuations might cause disturbances elsewhere in the plant The signal transmission cannot keep up with fast fluctuations
97
98
Question
Answer • Large fluctuations might damage equipment Yes. For example, large changes in the manipulated flow could cause a pressure surge in a distillation column.
• High frequency fluctuations might wear out the valve Yes, if the manipulated variable were constantly opening and closing, the valve would be likely to wear out.
• Large fluctuations might cause disturbances elsewhere in the plant Yes. In such a case, the disturbance would occur in the flow that is being manipulated and could be propagated to the source of this flow.
• The signal transmission cannot keep up with fast fluctuations No, electrical transmissions would always be able to keep up with the fluctuations (why ?)
99
Which of the following criteria are required for a manipulated variable? (Distinguish criteria that are (i) required, (ii) desirable, and (ii) not relevant.) • • • • •
The ability to compensate for large disturbances Automated valve to influence the selected controlled variable Fast speed of response Causal relationship between the valve and the controlled variable The ability to adjust the manipulated variable rapidly and with little 100 upset to the remainder of the plant.
• The ability to compensate for large disturbances
Question
Answer
No, the manipulated variable does not have to be able to compensate for large disturbances. This property is not required but is always desirable
Five approaches to control are ‘no control’, ‘on-off’, ‘manual’, ‘automatic’ and ‘emergency’. Which of the five approaches would you apply to each of the following variables?
• Automated valve to influence the selected controlled variable Yes, the manipulated variable must have an automated valve to influence the selected flow. If this was not the case the controller could not adjust the system.
• Fast speed of response No. Although speed is always desirable, the manipulated variable does not have to have a fast speed of response.
• Causal relationship between the valve and the controlled variable Yes, there must be a causal relationship between the manipulated variable and the controlled variable i.e. if we change the value of the manipulated variable, there will be a subsequent change in the controlled variable.
• • • •
Temperature in a garage Temperature in a oil storage tank that could freeze in winter Temperature of soup being heated on a stove Temperature of a CSTR (continuous stirred tank reactor) making pharmaceuticals • Temperature in a closed vessel with an exothermic chemical reaction.
• The ability to adjust the manipulated variable rapidly and with little upset to the remainder of the plant. No, the manipulated variable is not required to be able to be adjusted quickly and with little 101 upsets to the remainder of the plant. This behaviour is however desirable.
• Temperature in a garage No control is necessary for the temperature in a garage.
Answer
• Temperature in a oil storage tank that could freeze in winter On/Off control. A temperature sensor would indicate to the control system when the oil temperature was too low, and the system would engage the heater. When the temperature had risen to the appropriate level, the sensor would signal the system to turn off the heater. In this situation, it is not important to maintain a specific temperature, just maintain the temperature within upper and lower limits.
102
Question Make all the necessary equipment changes and additions to be able to control the fluid level in the draining tank in the figure.
• Temperature of soup being heated on a stove Manual control. The burner would be turned on high until the soup boiled at which point the burner would be lowered to a simmer. Once the soup was ready, the burner would be turned off.
• Temperature of a CSTR making pharmaceuticals For producing pharmaceuticals, a high level of control is required to assure product quality. The CSTR would have been designed with an optimum temperature and automatic control would be the best choice to maintain this temperature.
• Temperature in a closed vessel with an exothermic chemical reaction Emergency control. When the temperature within the closed vessel exceeds the maximum allowable temperature, then the control system vents the pressure or cools the tank.
103
104
Answer To control the liquid level in the draining tank three equipment changes are necessary. • First, we have to measure the level; select a pressure difference sensor for this purpose. • Second, a control valve is needed to provide a variable resistance in the outlet pipe from the tank. • Third, a calculating device is needed to perform the controller calculation. This would often be a computer with communication with the sensor and valve.
Question For the two closed-loop control responses, rank which is better for each of the following performance criteria: Stability, Return to set point (i.e. zero steady state error), IAE (integral square error), Rise time, Settling time, Damping coefficient, Max |SPCV| and MAX MV overshoot.
Note: A pump is not required unless the liquid level in the tank is not sufficient to provide the desired flow rate. 105
A comparison of the criteria are given in the following table. The preferred response is in green.
Answer
Case A
Case B
Stable
Yes
Yes
Returns controlled variable to its set point
Yes
Yes
smaller
bigger
approx. 30
approx. 25 (faster)
Performance Criteria
IAE (integral abs. value of error) Rise time Settling time Damping ratio Max. CV deviation from set point MV overshoot from its final value
approx. 30 (smaller)
100+
approx. 0 (smaller)
approx. 1/2
106
Question Consider the situation in which a control loop has been functioning well, then the failure position of the valve was switched. What would have to be changed in the computer algorithm?
Not applicable to set point changes responses. This is relevant for disturbance responses! approx. 10/25 (smaller)
approx. 10/15
In general, CASE A provides better performance and would be preferred in most industrial applications. It provides better performance in all items except rise time. 107
• Nothing should be changed. • Need to re-do the tuning procedure (reaction curve and calculations). • The controller gain should be reduced to prevent oscillations. • The controller sense switch should be changed.
108
Answer • Nothing should be changed No, if nothing were changed, the control loop would become unstable!
• Need to re-do the tuning procedure (reaction curve and calculations) No, there is a much easier solution. However, repeating the entire tuning procedure would result in the correct answer.
• The controller gain should be reduced to prevent oscillations No. This would still result in closed loop instability (why ?)
• The controller sense switch should be changed Yes, the sign of "delta CV"/"delta controller output" has changed. Therefore, we must change the sign of the controller gain, which is the effect of changing the controller sense switch.
Question
True or false ?
• All control valves used for pressure control should be fail open to prevent pressure build-up. • If we average many readings from a sensor with excellent reproducibility, we will obtain a value very close to the true process variable. • Measured values should be filtered to smooth sensor noise before being displayed on trend plots. • In the data from a feedback PID control loop, the poor performance is due to (integral) reset windup.
Controlled variable
Valve position
109
• All control valves used for pressure control should be fail open to prevent pressure build-up.
Answer
False. The safest position could be either open or closed. Generally, for flows into a vessel, the safest is closed; for flows out of a vessel, the safest is open! The entire process must be analyzed before defining the safest condition.
110
Question When the PID controller in the equation is placed in service, which of the following are performed as part of the initialization?
• If we average many readings from a sensor with excellent reproducibility, we will obtain a value very close to the true process variable. False, a sensor with good reproducibility can have an average value different from the true value. In some cases, this difference could be significant.
• Measured values should be filtered to smooth sensor noise before being displayed on trend plots. False. People need to see the true value from the sensor without delay or filtering. They diagnose problems with equipment from the measurements, and the higher frequency variation could help them in this diagnosis.
• In the data from a feedback PID control loop, the poor performance is due to (integral) reset windup. True. The performance is not good because the valve saturates i.e. reaches a physical limitation of fully closed, at times during the transient. When the valve is completely closed, feedback 111 control is no longer possible, and integral windup results.
• Set the error to zero • Set the integral error to zero • Calculate the bias (I) so that the manipulated variable does not change • Calculate only the proportional mode for the first execution; then, calculate all modes thereafter.
112
• Set the error to zero
Question
Answer
Why do we design a version of the PID controller with a filter on the derivative mode only ?
No, the error is calculated according to E = SP - CV at every execution.
• Set the integral error to zero Yes, the integral error is automatically zero, since t = 0. However, an "old" value may be stored from the last time that the controller was in operation, so "clearing" the stored value to zero is good practice.
• Calculate the bias (I) so that the manipulated variable does not change. Yes, this approach will result in "bumpless transfer", so that the valve does not change at the first instant, and the controller performs feedback control thereafter.
• Calculate only the proportional mode for the first execution; then, calculate all modes thereafter.
• • • •
No, this will result in the manipulated variable (and the valve opening) "jumping" when the controller is turned on.
There is no theoretical basis, its purely based on experience. The design provides zero steady-state offset We never design the controller as described It reduces excessive variation in the derivative due to high frequency noise without slowing the proportional and integral modes.
113
• There is no theoretical basis, its purely based on experience
114
Question
Answer
Consider a situation in which the control loop has been functioning well; then, we increase the sensor range from 0-100 to 0-200 C. What should be changed in the commercial controller?
No, there is a good reason. Think about the affects of noise on each mode.
• The design provides zero steady-state offset No, does the derivative mode ever affect offset at steady-state?
• We never design the controller as described No, you need a break to come back to the topic refreshed !
• It reduces excessive variation in the derivative due to high frequency noise without slowing the proportional and integral modes. Yes, high frequency noise has an especially bad effect on the derivative mode; therefore, we filter it. The proportional mode is affected less by such noise.
115
• • • •
Must repeat the process reaction curve experiment. Reduce or eliminate the derivative mode. Increase the scaled controller gain by a factor of 2. Do not change anything.
116
• Must repeat the process reaction curve experiment
Answer
No, we do not have to do all the work over again!However, repeating the entire tuning procedure would result in the correct tuning.
Question
For the process below, a heating medium flows through the exchanger coils, and the valve is fail closed. What is the correct sense for the controller ?
• Reduce or eliminate the derivative mode. No, you really need a break !
• Increase the scaled controller gain by a factor of 2. Yes, the scaled error is reduced by a factor of 2. Therefore, we have to increase the controller gain (Kc) by a factor of 2, so that the feedback loop is unchanged.
• Do not change anything. No.
• Must repeat the process reaction curve experiment
117
Answer
No, reconsider the commercial form of the PID controller.
• Either sense would work OK. No, does it make a difference whether you turn the steering wheel to the right or left when driving a car?
• The sense would be direct acting (increase-increase). Yes, let's think about an example. If the temperature is above its set point: Error = SP - T < 0, therefore desired change in valve is to close. Since the valve is fail closed, signal to valve (MV) < 0 Since E x Kc = MV, the requirement is (negative number) x (Kc) = (negative number) This requires that Kc is positive, or direct acting sense.
• The sense would be reverse acting (increase-decrease). No. 119
• • • •
Must repeat the process reaction curve experiment Either sense would work OK. The sense would be direct acting (increase-increase). The sense would be reverse acting (increase-decrease).
118
