B OOK D ESCRIPTION
1.4. C ONTROL T YPES
There can be different types of control action sought. These are as follows;
1.4.1. Feedback Control
Examples of feedback control was shown in Example 1-2.9, Example 1.2.11. During feedback control the output variable is measured. The measured signal is compared against a set point using comparators. The error signal forms the input into the controller. Control action depends on the air. For example, as will be discussed in the later chapter, a P only controller is made in a manner that the control action is directly proportional to the error. In Chapter 5.0, feedback control of single-input or manipulated variable and single output or measured variable is discussed in greater detail. The feedback control structure is arrived at by selection of manipulated variable and increased or decreased to control the measured output variable. The desired value of the measured process output is also called as setpoint.
Later control gain and control time constant would be decreased. In a similar manner, the process also can be characterized with a process gain and process time constant. The gain is the rate of change of the output to the rate of change of input. This can provide a measure of the sensitivity of the process or controller. The gain can be positive or negative. Negative gains leads to inverse response. This is one kind of instability as will be discussed in chapter 3.0 and chapter 4.0. A control algorithm can be constructed and control parameters can be identified. These control parameters can be tuned in order to obtain better performance. The feedback control can further be of different types. These are as follows;
a. On-off Controller
b. P, Proportional only Controller c. PI, Proportional Integral Controller d. PD, Proportional Derivative Controller
e. PID, Proportional, Integral and Derivative Controller
1.4.2. Feedforward Control
An example of feedforward control is discussed in Example 1-2.9. Vasodilation and evaporative cooling/sweat formation happens immediately after a change in surroundings.
This kind of control action is different from the control action described in feedback control
Introduction 25 above. The disturbance variable is measured and control action is taken proactively. An estimate is made on the effect of the control action on the output variable. Output variable is not measured directly. Therefore the control action can work well when there are reliable models that can relate the process output to the manipulated variables. Further the measured disturbance need to be the only aberration in the scheme of things.
1.4.3. Hybrid Feed forward and Feedback Control
The feedback and feedforward schemes can be hybridized. In such cases the engineer has more things on his hands to adjust. (see Example 5.2).
1.4.4. Internal Model Control, IMC
IMC, internal model control action uses the knowledge about the process that is being controlled. This is different from the black-box approach of feedback control of single output variable by tweaking the inpuit variable based on measurements of output variable. The model developed for the mixing tank heated by the hot fluid in the jacket in Figure 7.4 can be used to design an internal model controller. A good knowledge of when the process is stable and when the process is underdamped oscillatory unstable can lead to better control action.
Control action of unstable systems may result in unsatisfactory results. In general the model-based controller can be added as shown in Figure 7.8. Filters can be added to make the controller more realizable. The transfer function of the output has to be proper. The Laplace transform expression of the transfer function can be represented by;
( ) ( ) P s
Q s (14)
When the order of the polynomial of Q(s) in the denominator of Eq. (14) is greater than the order of polynomial P(s) in the numerator then the transfer function is said to be proper.
When the order of the polynomial in the numerator P(s) is the same as the order of the polynomial in the denominator Q(s) then the transfer function is said to be semi-proper.
1.4.5. Ratio Control
Ratio controllers are used where ratio of the reactant mixture is of increased significance.
For example when flue gas is needed the ratio of air to CO2, is controlled in such a manner that the heat of reactions from the Boudard reaction and heat of reaction from the oxidation reaction “cancel” each other rendering the reactor at a adiabatic state. One or more valves are used in a split range control. More on ratio control is discussed in section 7.1. Ratio control can be used in distillation columns where the reflux ratio is an important design variable.
1.4.6. Statistical Process Control, SPC
SPC is discussed in greater detail in section 7.2. SPC is used to control the output variable within certain specifications. Control charts are used to better understand the errors between measured and set point values. Attributable causes are assigned to variations. Noise is delineated from a disturbance. Control action is taken according to a 3 sigma or a 6 sigma limit. The statistical measures of mean and variance are only used. This approach falls intermediate between feedback control of a process in a black-box and an IMC, internal model control where the process models are clearly known.
1.4.7. Estimation and Control of Polymerization Reactors
The high viscosities and exothermic nature of the polymerization reactions make the control of polymerization reactors an ardous task. The control objectives need to be specified.
The process dynamics is non linear. Some of the control strategies that were discussed when the process can conform to a prototypical first order or prototypical second order process for linear systems cannot be applied to systems that are governed by equations that are nonlinear.
Measurement of polymer structure is by no means a done deal. Estimation techniques such as Kalman filter and Weiner filter are needed to obtain parameters that are difficult to measure.
1.4.8. Neural Networks
ANNs, artificial neural networks may be used in control of distillation columns where the number of output variables is in the hundreds and the relation to the input variables are governed by equations that are nonlinear. ANNs comprise of a number of computing elements which resemble neurons and synapses of a human brain in a network. NNs can be used to approximate any mathematical function to a desired level of accuracy. In supervised learning, a network is given an input along with its desired output. On the other hand, a network in unsupervised learning is given only an input. After each presentation of an input, the performance is measured to tell how the network is doing. A network is expected to self-organize information by using the performance measure as guidance.
1.4.9. Multivariable Process Control
Most control schemes discussed is single variable input and single variable output, SISO systems. MIMO systems are multiple variable input and multiple variable output systems.
Interaction effects of variables in MIMO systems are often important. A disturbance at any input causes a response in some or all of the outputs. Control and stability analysis of MIMO systems are more tedious compared with analysis of SISO systems. The control system can be decoupled so that control of some output variables may be affected. Cross controllers [Marlin, 2000] in addition to principal controllers may be used to accomplish the control objectives.
The number of controllers needed for the operation increases ponentially with the number of
Introduction 27 inputs and outputs. 4 controllers are needed for a system of two inputs and two outputs. 9 controllers are needed for systems of 3 inputs and 3 outputs. The characteristic equation for a MIMI system can be examined for stability from its roots.
1.4.10. Adaptive Control
Rather than a black box approach to control action taken or use of a process model adaptive control is based on observations made at discrete intervals of time, updation of model based on the observations and then control action taken.
1.12. G
LOSSARYBatch Process – Product is made from raw materials batch by batch. The raw materials and catalyst are charged into the reaction vessel. The reaction vessel is operated at the selected temperature and pressure. The product is removed after the reaction and separated from the unreacted raw materials. The vessel may be stirred at the set RPM, revolutions per minute.
Continuous Process – Product is recovered at certain lb/hr from the process plant. Raw materials are fed continuously into the reactors. Reactors are operated at the set temperature, pressure, degree of intensity (agitator RPM). The product is separated from the unreacted raw materials in a suitable unit operation such as devolatilizer, distillation column, centrifuge, etc.
Industrial Controls Market – This includes all the elements of the control process: (i) Measurement; (ii) Comparisons; (iii) Computations and; (iv) Corrective Actions. Sensors, comparators, desktop computers dedicated to control action and actuators, signal transmitters form the industrial controls market. Detectors, transducers, transmitters and controllers are all part of the industrial controls market.
Process Dynamics – Study of the response of the different unit operations in the process to a step change or some other perturbation given to the system.
PLC – Programmable Logic Controller. A computer dedicated to automation of electromechanical processes.
Automation – Operation of a device or process with minimal human input.
Runaway Reaction – Conversion of the reactant increases in an uncontrolled manner.
Heat is given out in exothermic reactions. The heat generated, if not removed, increases the reaction temperature. Often reaction rate doubles with every 10 0C increase in temperature.
This further increases the reaction rate followed by more heat generation. Lack of heat removal causes a runaway condition.
Transient Study – Study of physical quantities that have not reached steady state yet.
Block Diagrams - Boxes are used to denote functions and lines are used to show the relations between the functions.
Data Acquisition – Enables measurements at different time intervals from the equipment to be converted to digital format to be stored in the computer. A/D analog to digital converter can be used to convert transducer voltages to digital values in the desktop computer.
Robot – Programmable, automatic machine that is multi-purpose and manipulatable.
Free Radical Polymerization – Method of polymerization where the initiation, propagation and termination reactions take place by the free radical mechanism. Free radicals are formed from initiators and monomer and propagating species is a free radical.
Supply Chain Robotics – Use of robots in the distribution and supply of goods
PW, Present Worth – Money value of a stream of cash payments of receipts and costs projected in the future in terms of today‟s dollars include capital, interest, tax and tariff, labor, utilities, materials, overhead, revenue from sales, salvage value over the life of the plant.
Artificial Intelligence – Synthesize intelligence using computer software programming and execution.
Kinematics – Science of motion without worry about the forces that causes it.
3 Arm Mechanical Manipulator Device with three rigid lengths and two joints. The