Você está na página 1de 82

PROCESS DYNAMICS AND CONTROL

Paul erban Agachi Department of Chemical Engineering, University Babe-Bolyai, Roumania Keywords Process control, steady state, dynamics, process engineering, control loop, transducer, controller, control valve, advanced process control Contents 1. Introduction 1.1. History of Process Control 1.2. Fundamentals in Process Control 1.2.1. What is a chemical process system? 1.2.2. Manual and automatic process control 1.2.3. Steady state and dynamics of the control systems 1.2.4. Mathematical modeling and simulation of the processes 1.2.5. System dynamics 2. Analysis and synthesis of the feedback control systems 2.1. The controlled process 2.2. Transducers and measuring systems 2.3. Controllers 2.4. Control valves and actuators 2.5. Synthesis of the control system 3. Advanced control systems 3.1. Cascade and feed-forward control systems 3.2. Advanced Process Control 4. Automation of complex processes. Summary Process control is the automatic operation of a process, namely the process operation without the human intervention. At origin, all manufacturing processes were operated manually, not existing restrictions to impose a certain quality of the product, or a certain deliverable quantity at a certain time, or, more recent, an environmental constraint. The mass production and the increased demands of quality of the products imposed the replacement of the manual control with automatic process control. The globalization enhanced the competition among the producers, meaning that the company with the most flexible production, the lower production costs and the best quality of the product is the winner on the market. Process control is now met everywhere, from home appliances to the heavy chemical, metallurgical industry or spacecraft. And this, due to the commodity in operation, the precision in execution, the permanent supervision of the process conditions and the savings in energy, raw materials and human labor. 1. Introduction Process control, namely automatic process control, is the operation of a process without the human intervention. Process control can be addressed to the manufacturing processes (e.g. chemical, metallurgical, waste or potable water treatment, natural gas or oil exploitation and transportation etc.), from which it stemmed, but also to transportation or household processes. 1

1.1. History of Process Control As one may see, the development of the process control is strongly related to the manufacturing processes. These are traced in the ancient times of the humanity, starting with the metal, fabric and pottery production. The industrial manufacturing and actually the engineering were the innovations of the XVIIIth century, during the Industrial Revolution. The population increased sensibly during this century, the population being inclined to consume more and better. The consumerism led to an escalation in demand, both with regard to quantity and quality, for food, clothing, footwear, housing, transportation, which stimulated the production of construction materials, textiles, chemicals etc. Each global conflict, after its end (e.g. first and second World Wars), induced the same behavior and the same reaction on behalf the society and production companies. The production became mass production with huge quantities of products delivered at deadlines and with a certain quality expectation. The mass production, under these circumstances, could not be anymore controlled manually because of the expectations. In the meantime, in the second half of the XXth century, the environment became important, fact which imposed the environmental constraints on the manufacturers. Top on that, the globalization of the economy amplified the competition among world companies and only those capable to reduce costs and to respect the environment, resisted on the market. Together with these facts, an important impact on the development of the technology and especially computing facilities had the war and space race between the major world powers: the US and USSR. All these sequences led to a tremendous development of the control equipment, techniques as a part of process control development. Process control was seen as a major tool for development and complying with the constraints. Milestones in the modern history of control are C. Drebbels contribution in inventing the first temperature control device for a furnace, around 1624, D. Papin invention of the first safety valve for his steam engine pressure regulator in 1704, E. Lees first controlled positioning system for a wind mill in 1745 (Figure 1), T. Polzunovs first level controller for his steam engine (1765), J. Watts fly ball governor in 1768 pressure regulator for his steam engine (Figure 2). The first obviously advanced combination between process engineering and process control was H. Jacquards loom in 1801 (Figure 3) which stored the model of the carpet on punched cards.

Figure 1

Figure 2

Figure 3

The first publication in the field of control systems was elaborated by J.C. Maxwell in 1868 and approached a theoretical analysis of the stability of Watts fly ball governor (1868). The next papers on the subject of automatic control appeared only in the first half of the XXth century (1922 - 1934) and have to be noted the first in the field of control in chemical/process engineering. 2

A major innovation represented G. Philbricks Automatic control analyzer Polyphemus, actually the first analogue computer (1937-1938) just before the World War II. The World War II brought extremely important innovations in the field of automatic control: automatic rudder steering, automatic gun positioning systems, automatic pilot of V1 and V2 etc. Immediately after the World War II, the field of process control exploded. It was helped by the construction of the mega computers ENIAC (1946) and UNIVAC (1951), Shockleys patent on transistor (1950) and Feynmans premonition regarding nanospace expressed at the American Physics Society Meeting in Caltech Theres plenty of room at the bottom (1959). The new frontier of challenging the outer space launched by the US president John Fitzgerald Kennedy produced the portable computer which influenced thoroughly the more recent history of Process Control. The first process control computer system in a chemical plant was installed in 1964 by Standard Oil California and IBM at El Segundo refinery in a FCC Unit, under the name 1710 Control System. Thus, there were practically three stages in Process Control development: - the first of Measurement and Control Devices (World War I mid XXth century), dominated by manual control and operation of the industrial processes; there were used sometimes field mechanical controllers; - the second of Classical Feedback Control (50s 70s), when most of the processes were supervised and controlled automatically in the simplest way feedback control; the control equipment was either electronic or pneumatic, rarely hydraulic; - the third stage of Computer Assisted Control (70s present) dominated, as main feature, by the control systems involving a micro or minicomputer; most processes are controllable and they need classical control systems, but around 20% of them from process industry are less controllable and need Advanced Control techniques and equipment. 1.2. Fundamentals in Process Control 1.2.1. What is a chemical process system? A chemical system is the ensemble of the physical-chemical processes in evolution and the technological procedures for their realization. Thus, a chemical/process engineering system comprises the processes, the technological equipment where the processes take place, the measurement and control devices and their intercommunication. Any system is composed of parts linked among them and defined through their characterization (input, output, state) variables (Figure 4 and Figure 5).

Figure 4 3

Figure 5 The output variables are those of interest. That is, for a distillation column, the output variables are the head and bottom production expressed by the distillate flow (D) and bottom flow (B) and their purity (xD and xB); for a heat exchanger, it is interesting the outflow temperature (To). The input variables are those influencing the output variables. Some of the input variables are the so called disturbances which cannot be influenced by the human operator or the control system and some are the manipulating variables, possible to be modified by the operator or the control system For the distillation process, the disturbances could be the feed flow (F), the feed flow composition (xF), atmospheric pressure (patm) in the case of the atmospheric distillation or the exterior temperature (Toext). Disturbances occur all the time due to different causes: weather change, change of the raw materials, change of the quality of the fuels, ageing of the technological equipment etc. These mean that practically there is no steady state in the functioning of any system. The manipulating variables are the heat supplied to the column (Q), the reflux flow (R). For the heat exchanger, the disturbances are the input temperature of the heated flow (Toi), the input temperature of the heating agent (Toiag); the manipulated variable is the agent flow (Fag). The state variables are those describing the state of the process. Sometimes the state variables are identical with output variables. In the case of the distillation process, state variables are the tray concentrations and temperatures (xi, yi, Toi). The influence of the input variables upon the state or output variables is transmitted on the transfer paths described mathematically by transfer functions. The transfer functions express the dynamics of an action of a certain input upon an output. The transfer functions derive from the differential equations describing mathematically the dynamics of behavior and are expressed in the complex space. They show how much the input is amplified in the output and how fast is the response of the output at a certain input. Equation (1) is the differential equation describing the influence of the agent temperature of the heating process in Figure 5, upon the output temperature T . 4

+T =

ag(1)

where V is the volume of the fluid inside the exchanger, is the density of the heated fluid, cp is the heat capacity of the heated fluid, KT is the heat transfer coefficient and AT is the heat transfer area of the exchanger. The corresponding transfer function on the transfer path Toag to To is expressed by equation (2) where K is the gain of the transfer function and T is the time constant. ( ) where K= (2)

The gain shows how much an increase in Toag is felt by To; the time constant shows which is the delay of the input action of increasing Toag on the output To. The higher the heat transfer given by KT and ATis, the stronger the reaction on the path is; the larger the inner volume of the heat exchanger is, the larger the time constant is, meaning a slower response. Example: for a transfer function between Fag and To with a gain factor K= 0.5 ( ) and a time constant T= 10 min at an increase of the agent flow with 0.3 , the temperature will increase with (Figure 6).

Figure 6

1.2.2. Manual and automatic process control Taking into consideration the same simple process of heat transfer from Figure 5 one may consider the way a process operator is trying to maintain the temperature stable at a value given by the technological process (Figure 7). Since a disturbance occurs, say a decrease of the heating agent 5

Figure 7

temperature during the winter, the operator watching the process observes the change and manipulates the valve opening it in a step way, thus increasing the heating agent flow and the heat flow to the process. The temperature will increase after a long while, the delay being given by the time constant of the process (Figure 7 a). An experienced operator, knowing the dynamics of the process, opens abruptly the valve (Figure 7 b), gives a thermal shock to the process, hurrying in this way the change of temperature. After a while, the operator decreases the agent flow to an intermediate value and smoothly reaches the desired output temperature. This is a way of controlling the temperature, similar to the automatic control (PID). The drawbacks of the manual control are obvious: - It needs continuous attention in supervising the parameters whether on night or day shift; if there are hundreds of parameters to be supervised, it is very difficult to operate appropriately the process; - For a good operation of the process, it needs high level of understanding the process (including its steady state and dynamics) on behalf of the operator; These considerations support the introduction, wherever possible, of the automatic control systems. 6

Figure 8shows the usual control solution for the heat exchanger. Actually, the control system copies the actions of the operator: a. the observation of the temperature; b. the calculation of the deviation from the prescribed value; c. the estimation of the change needed in the manipulated variable (in operators mind); d. the actual change of the heating flow until the set point is reached.

Figure 8. The devices performing all the above mentioned tasks are: the transducer for action a; the controller for actins b and c; the control valve (final control element) for action d. The block diagram for a general feedback system is given in Figure 9.

Figure 9 The significance of the notations are the following: Process - the controlled process, namely the transfer path manipulating variable - controlled variable; T (transducer) - the device which measures the controlled parameter and transforms it into a signal compatible with the other signals in the loop; S (summing point or comparator) - the device which compares the set point with the signal of the transducer and calculates the error; C (controller) - the device which calculates the control signal, according to an embedded control algorithm and the sign and magnitude of the error; AD(actuating device or final control element) - the device which modifies the manipulating variable, to change the process output. 7

The system is named feedback control system, because it measures the controlled variable (xe) with T (xr - the reaction variable), feedbacks it to the entrance of the system, where it is compared with the desired value (xi or xset - input or set point) inside the S; S calculates the deviation of the controlled variable from the set point ( error; = xset - xr); the error is supplied to C and the controller calculates according to its algorithm (P, PI, PID or others) the control variable xc; the AD transforms xc in the manipulated variable xm which is a mass or energy flow which, modified, is supposed to bring the deviation/error to 0. Of course, the automatic control systems require a better preparation of the personnel, increased initial costs (about 10% - 15% additional to the costs of the equipment), a good organization of the plant including metrology, maintenance, a good understanding of the economics of additional areas like optimization, heat integration, control. 1.2.3. Steady state and dynamics of the control systems. The time functioning of a system in steady state implies the time invariability of its parameters. Thus, the period of functioning of the heat exchanger outside the interval t0 - t1 (Figure 7b) represents a steady state. From the moment t0, the process modifies its working parameters entering in the so called dynamic state (until t1). Starting with t1, the process calms down to a steady state. Usually the dynamic state of a process is not desirable, because the process can become uncontrollable. But sometimes, especially in the batch processes case, the dynamic behavior is desired and conducted in such a way to obtain the maximum of production, for example, in the minimum time. Unfortunately the ever appearing disturbances offer short periods of stationarity for a process affecting seriously its technical and economic performances. For example, in the functioning of the distillation column (Figure 4), the stability of the composition at the top of the column is disturbed almost permanently by: changes of the composition of the feed flow (changed raw material, changes in the functioning conditions of the previous stages), variations of temperature at the bottom of the column (due to the changes in the steam quality), variations of the feed flow (clogging of the pipelines, ageing of the centrifugal pump), the outside pressure change (variable weather fronts). The permanent presence of disturbances imposes the automatic control of a process. From the point of view of stability, the process can be self-regulating or non-self-regulating. A process is self-regulated if, at the occurrence of a bounded disturbance, goes from a steady state to another steady state via a dynamic, transitory period (Figure 10a). If the process, at the occurrence of a bounded disturbance, goes from a steady state to an unstable regime (Figure 10b), it is non-selfregulated. This is the case of the exothermic reactors.

Figure 10

Figure 11

There are cases, when a stable by nature process becomes unstable due to a badly designed control system (Figure 11). This is a cause which sometimes makes the control loops to be set on manual operation. The performance of the control system is judged both in steady state and in dynamic one. Thus, if a control system is well designed and tuned, responds to a step increase perturbation with a damped oscillation with the following elements distinguishable (Figure 12): 1 overshoot; are the amplitudes of the second and third oscillation; - the steady state error; - transient time (duration of the transient process).

Figure 12. 9

One can say a process control system behaves properly if during the dynamic state: (3) Meaning that the overshoot does not exceed a technologically imposed value e.g. 10% of the setpoint value; (4) Meaning that the transient time is shorter that a technologically imposed value e.g. for a drying process it is important that the deviations from the setpoint to be shorter than 10 min. ( ) (5)
1

Where is the damping ratio and it is important to be smaller than an imposed value; it is considered usually a good behavior if the 5 meaning that the oscillations respect the rule of quarter decay ratio, (Figure 13).

Figure 13 The performance in the steady state is the steady state error, that is the difference between the settled controlled variable and its set point. (6) Meaning the steady state error should be smaller than an imposed value; technologically, is acceptable if it is under 2% of the set point value.

1.2.4. Mathematical modeling and simulation of the processes The mathematical model of a process represents the mathematical relationship between the output variables and the input and state variables. In the case of the process form Figure 5, the mathematical model is an expression in its most general form: ( ) (7) wheret is the variable time. If the model describes the steady state, t is missing, the equation being an algebraic one. If the model describes the dynamic state, equation (7) is a differential one in a form similar to the differential equation (1).In the case the process has several output variables, the model takes the form of a system of algebraic or differential equations or combined.

10

The models can be theoretical, analytical ones based on the conservation, thermodynamic and kinetic equations; or empirical ones in form of mathematical regressions, based on experimental values, processed statistically. For more complex processes, the models are combined containing both analytical and empirical parts. It is extremely important that all terms of the model to be dimensionally consistent, that is to be expressed in the same Measurement Unit System. The use of the mathematical models is extended from the activity of design and operation to those of simulation and control. The simulation is the representation of the reality based on the results from running a mathematical model of that reality. The process simulators can be used either for training the personnel in a plant, for designing the plant or its control system. To describe shortly the mathematical modeling, let us consider a Continuous Stirred Tank Reactor (CSTR) in which neither its liquid content (holdup), nor its compositions and temperature are constant (Figure 14).

Figure 14 The chemical reaction in the CSTR is first order reaction with the rate constant of the reaction. The reaction is exothermic, with the heat of reaction are the input flow, molar input concentration of component A, input temperature of the reactant and input temperature of the cooling agent. , , , are the output flow, output molar concentrations of both components, temperature of the reactor and outflow and temperature of the cooling jacket and agent outflow. is the volume of the mass of reaction, are the heat transfer coefficient and heat transfer area respectively. There are three conservation fundamental laws which are used in the mathematical modeling: Mass conservation law which expresses the accumulation either of the total mass in the reactor (Equation 8), or the composition variation inside the reactor (mass conservation on components) (Equation 9). 11

(8)

}{

+={

}(9)

Energy conservation law which expresses the accumulation of heat in the reactor (Equation 10) { { } } { }=2 3 (10)

Impulse conservation law expressing the velocity of movement in the system: rate of change of the movement quantity of a system is equal to the total force applied to the system. (Equations 11 and 12): { }=* = ( ) + or (11) (12)

ij

Thus, the equations (8) (12) take the following form for the CSTR in the example: Total mass conservation ( ) Mass component conservation ( ) for component A and ( Energy conservation ( ( Impulse conservation for the outflow pipeline ( ) where h is the height of the liquid column in the reactor, Ap the cross area of the pipeline, coefficient of the pipeline and v the velocity of the liquid in the pipeline. ) ) ( ( )for the inner mass )for the jacket ) for component B

(13) (14) (15) (16) (17)

(18) the friction

12

Mass conservation equations are used to determine the variation in time of the total mass and its components in the reactor, energy equations to calculate the temperatures and impulse equations for determining the flows. There are systems with lumped parameters, where all parameters are constant relative to the place they are measured (e.g. CSTR has in all its mass of reaction the same density, concentration, temperature) and with distributed parameters where the parameters vary with point of measurement (e.g. inside a long pipeline for gas transportation, the pressure differs due to the pressure drop). The behavior of the lumped parameters systems is described by ordinary differential equations and of the distributed parameter systems by partial differential equations. The simulation is the process of imitating the reality using a model. The model is solved and its solutions represent the variations induced in the process by the inputs, or construction constants. The simulators are used in industry to train the operators before operating a process. 1.2.5. Systems dynamics Mathematical models describe the behavior of the systems in steady and dynamic state. The operation or control of a process is done always in dynamic state. Solving the differential equations implies changing the input variables of the systems. In order to judge and to compare the behavior of a large variety of systems, it is necessary to have standard input signals. Standard input signals Step input The step function is defined as ( ) 2 3 (19)

0 time is the time when the abrupt change occurred. It is one of the most important test functions for the behavior of the control systems. A step function can be the abrupt opening of a valve with 25% flow increase, or the abrupt switching of feed from an empty chlorine tank to another one in a process of manufacturing chlorinated solvents (Figure 15). Usually the changes in the processes are not so abrupt, but one prefers to test the control system to such a change in order to have sufficient margin for a sure and good behavior.

Figure 15

13

Ramp input The ramp function is defined as ( ) 2 3 (20)

Where K is a proportional factor having as significance the slope of the signal. The ramp inputs appear when the process is drifting. Figure 16 is representing a ramp function, corresponding to a slow but steady change of the input (e.g. opening a valve with 5%/min). Sometimes the ramp signal is used in tuning the controller parameters.

Figure 16 Sinusoidal input The sinusoidal function (Figure 17) is defined as ( ) 2 3 (21)

Very often, the processes are subjected to sinusoidal changes (e.g. diurnal change of temperature). But according to Fourier series, any periodical signal can be decomposed in a series of sinusoidal functions (Figure 18), the most important being the fundamental (largest period of oscillation) and several secondary harmonics.

Figure 17 14

Figure 18 Thus, a periodic disturbance like a series of rectangular non uniform steps (e.g. the fluctuation of water flow during the washing of recipients) can be considerably disturbing when the fundamental frequency of the signal is close to the own natural oscillation frequency of the system, producing a phenomenon of oscillation. Impulse input The impulse function or Dirac delta function is defined as ( ) { }with

(22)

and is shown in Figure 19. This test function is used sometimes in determining the residence time of equipment. It is possibly to be obtained physically through an injection of material in the reaction system.

Figure 19 15

Laplace transform and the algebra of transfer functions The Laplace transform of a function f(t) is defined as ( ) , ( ) ( ) (23)

Where ( ) is the time function subjected to differentiation and has to be continuous for the integration interval, is the Laplace operator, ( ) is the Laplace transform, s is the Laplace transform complex variable. It is useful because it translates a function from time domain in the complex domain with the property of transforming the differential/ integral equations in algebraic ones, much easier to be solved. Transfer functions The translation of the differential equations in s complex space have a result not only the transformation of the linear differential or integral equations in algebraic ones, but also the transfer function describing synthetically the reaction of the process on a certain transfer path, in terms of gain and delay. The transfer function is defined as the ratio between the Laplace transforms of the output, Xe(s) and input variables, Xi(s) (Figure 20).

Figure 20 The transfer function is then ( )


( ) ( )

(24) ( )(25) , ( )-

The properties of the Laplace functions are: , ( )- Linearity , ( )- Distributivity , ( ) ( )-

, ( )-

( )

( )

(26)

The transformation of the differential or integral equations in algebraic ones are based on the Real Differentiation and Real Integration theorems with the result given by equations (27) and (28): 0 0
( )

( ) ( ) 16

(27) (28)

( ) 1

Thus, a differential equation of the first order, linear (first order element)
( )

( )

( )

(29)

Applying the Laplace transform to Equation 29, based on the definition (24) and properties (25) and (26), one is obtained, 0
( )

, ( )-

, ( )-

(30)

And further on, ( ) ( ) ( ) (31)

giving finally the transfer function of the first order element: ( )


( ) ( )

(32)

The algebra of the transfer functions Let us consider the case of two elements in a series (Figure 21), described by their own transfer functions Y1(s) and Y2(s)

Figure 21 The transfer function of the whole system is derived: ( )


( ) ( ) ( ) ( ) ( )

( ) ( )

(33)

Generalizing, for n elements in a series, ( ) ( ) (34)

Considering two elements in parallel (Figure 22),

17

Figure 22

the transfer function derived is: ( )


( ) ( ) ( ) ( ) ( ) ( ) ( )

( )

( )

(35)

and generalizing for n elements in parallel, ( ) (s) (36)

The importance of these series or parallel structures is given by the fact that in majority of the processes, there are several elements either linked in series or in parallel. For example, on the transfer path, Fag in the heat exchanger case (Figure 5) there are several elements in a series: the heating jacket, the inner metal walls of the heated tubes in a shell, the heated fluid (Figure 23).

18

Figure 23 Considering a negative feedback system (Figure 24)

Figure 24 The transfer function derived is: ( ) ( )


( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ), ( ) ( ) ( )( ), ( ) ( ) ( ) ( )-

( )

( ) ( ) ( )Therefore (37)

19

Frequency response analysis Frequency response analysis is an alternative way to transfer function analysis to interpret the dynamics of a system. The importance of the frequency response analysis derives from the fact that the majority of the disturbances are repetitive and thus, they can be decomposed in a series of sinusoidal functions according to Fourier theory. As a consequence, the fundamental can seriously influence the stability of a controlled system. This is the reason that, through frequency response analysis, the own frequency of oscillation can be determined and ways to remove the system from that undesired situation to be found. If in a differential equation the operator is replaced by the operator , the result is an algebraic equation. The solution of this equation is the representation in the complex space of the dependant variable modified through the application of a sinusoidal signal (Equation 21). If the period of a complete oscillation is P, the correspondent frequency can be expressed in several forms in radian/time unit in deg/time unit in 1/time unit The time unit is chosen function of the rapidity of the system. The range of oscillation frequencies for the chemical systems is between 10-4 to 1 Hz(one complete cycle per second). Expressed in other measurement units the range can become 0.006-60 min-1 or 0.00063 -6.28 rad/s or 0.037-377 rad/min. The sinusoidal input passing through the system is both amplified and delayed (Figure 24). ( ) ( ) where K is the gain factor and is the phase angle. (38)

(39)

Figure 25 The phase angle is calculated from the time shift .P is the period of oscillation. (40) 20

The frequency response can be calculated from the transfer function, replacing s with j. Thus, the transfer function becomes a frequency function Y(j) characterized by the real part Re() and the integer part Im() becomes ( ) ( ) ( ) (41) ( ) with the module ( ) ( ) ( ) and the phase angle ( ) The representation of these characteristics is done through the Bode, Nyquist or Nichols plots (Figure 26).

Figure 26 21

Types of dynamic behavior All systems include one of five types of dynamic behavior: proportional, integral, derivative, capacitive and with dead time. Any dynamic behavior of a more complex system is a combination of the five types above mentioned. Elements with Proportional behavior (P) The proportional behavior is described by the function ( ) ( ) (42)

where ( ) and ( ) are the output and input variables and KP the gain factor. The response of the element with proportional behavior at a step input is given in Figure 27.

Figure 27 The P element has no delay in transmitting the signal from the input to the output. Very rarely in the practice of chemical processes one can say there are systems responding instantly, unless we consider negligible one delay (in this case of the P element) in comparison with another one. The transfer function and frequency function are ( ) and ( ) (43)

The module and the phase angle are ( ) ( ) (44)

The Bode diagrams are presented in Figure 28

22

Figure 28 Example of proportional behavior (Figure 29). The electric resistive circuit responds apparently at the same time with the turn on of the switch. In reality, there is a time constant of several ms but in comparison with the time delays from the chemical systems, this is negligible. A similar response has a small recipient, with a volume of several cm3 when it is filled with gas from a compressor.

Figure 29 Elements with Integral behavior (I) The integral comportment is described by the equation (45) ( ) ( ) (45)

Where Ti is named the integral time constant, integral time or reset time. The response of the element with integral behavior at a step input is given in Figure 30. 23

Figure 30 The I element gives a ramp at the output, whose slope depends on the integral time constant. Elements with I behavior are for example the tanks or reservoirs with constant outflow when the inflow is changed. The transfer function and frequency function are ( ) and ( ) (46)

The module and the phase angle are ( ) ( ) ( ) (47)

The Bode diagrams are presented in Figure 31

Figure 31 24

Example of integral behavior (Figure 32). A tank with the constant outflow maintained by a metering pump behaves integrally when the input flow is increased abruptly. Of course, the signal increases only until the tank is filled, this being the limit.

Figure 32 From the total mass conservation law applied to the tank ( ) (48) with where is the cross area of the tank and h is the height of the liquid inside, the equation becomes and through integration ( ) (49)

Because the flow Fvi changed abruptly and F is constant, the quantity under integration is a constant, and the result of the integration is , a linear function of t. (Figure 32) The integral behavior is typical for the control function of the regulator, function destined to eliminate the steady state error. Element with Derivative behavior (D) The derivative behavior is described by the equation (50) ( ) (50)

Where Td is named the derivative time constant orderivative time. The response of the element with derivative behavior at a step input is given in Figure 33.

25

Figure 33 The D element gives an impulse at the output, when the input is a step signal. Naturally the derivative elements who practically anticipate the input action do not exist. They are created artificially for the so called derivative kick which can compensate some delays in the systems. The transfer function and frequency function are ( ) and ( ) (51)

The module and the phase angle are ( ) ( ) ( ) (52)

The Bode diagrams are presented in Figure 34

Figure 34 26

Example of derivative behavior. Derivative behavior is characterizing an electronic circuit containing an operational amplifier with positive reaction. (Figure 35)

Figure 35 Writing the second Kirchoffs rule for the ABCDEFG electric network, = 0, from the law of electric charge conservation around the capacity, with the second Kirchoffs rule on the network ABFG (55) and keeping in mind Ii is very small (input current to an operational amplifier), ui , equation (46) becomes = (56) The equation (50) indicates a derivative behavior with Td = RC. The derivative behavior is typical for the control function of the regulator, function destined to compensate the time lags in the controlled process. The derivative kick from Figure 7a, given in the manual operation by the operator, is given in an automatic control action, by the controller. Element with capacitive behavior (C) The capacitive behavior is described by the ordinary linear differential equation (57)
( )

(53) (54)

( )

( )

(57)

where T is named the time constant and K gain factor or simple, gain. The solution of the differential equation (57) is ( ) ( ) (58)

The response of the element with capacitive behavior at a step input is given in Figure 36.

27

Figure 36 The C element is characterized by a delayed response at the output, when the input is a step signal. Most of the processes are capacitive (accumulators of mass, heat, movement) of first or superior order. It is noted that the time constant is the value of time at which the response is at 62.3% of its final value. It can be determined from the construction of the tangent to the response curve, in the 0 time point. T is the segment at the intercept of the tangent with the final steady state value. K, is the ratio between the change of the output (xe) and the change of the input (xi). The transfer function and frequency function are ( ) and ( ) (59)

The module and the phase angle are ( )

( )

)(60)

The Bode diagrams are presented in Figure 37

Figure 37 28

Example of capacitive behavior. Capacitive behavior is characterizing the accumulation of specie A in a CSTR (Figure 14). From equation (14), it can be determined the differential equations in the canonic form (57) where xe is CA and xi isCAi. The equation (14) can be brought to the form
( )

( ) and

( )

(61)

where =

When the input concentration of specie A, CAi, increases suddenly to a superior value (due to a change of the raw material batch), the concentration of the same species is increasing accordingly to the Figure 38.

Figure 38 If the involved values are V= 1m3, F=1m/h3, k=0.1h-1, then the calculated values are: T=0.9h, K=0.9 V= 10m3, F=5m/h3, k=0.1h-1, then the calculated values are: T=1.66h, K=0.83 ; if .

Obviously, the systems with larger dimensions have larger time constants. The increase of the flow through the reactor has a consequence the increase of the gain, because of the decrease of the residence time which involves a stronger effect on concentration at the same conditions of reaction. Capacity of the 2nd and nth order (C2andCn) The behavior of a system with two capacities is described by the equation (62)
( ) ( )

( )

( )(63)

Where n is the natural frequency of oscillation of the second order element, K is the gain and is the damping factor. 29

The solutions of the differential equation are ( ) ( ) ( ) ( ) (Figure 39) ( )[ ( )[ ( )( ( )[ ( )for ( ( )] , for )] , for showing overdamped response

, showing critically damped response (64) and , showing underdamped response

showing oscillatory response )]for 0

Figure 39

The transfer function is thus ( ) (65)

But because the second order element can come from two first order elements in series, the transfer function can be expressed ( ) (66) 30

Let us consider the case with two evaporators in a cascade (evaporator with two effects) (Figure 40).

Figure 40 The frequency function is thus ( ) With the module and the phase angle ( ) and ( ) (68)

(67)

The Bode diagrams for this system is depicted in Figure 41

Figure 41

31

The nth order element is a sequence of first order elements (Figure 42)

Figure 42 and is defined by a system of n differential equations of the first order with the output variable of the element jas input for the element j+1. ( ) ( ) + ( )
( )

+ ( )

( )

. .(69) . .

( )

+ ( )

( )

( )

( )

( )

The response of the nth order element is given in figure 43

32

Figure 43

33

The first capacity gives an immediate capacitive response, the second capacity in the row is a little bit slower with a small induction times the induction time increases with the position of the capacity in the row the behavior being characterized in the end by an equivalent dead time ( ) and an equivalent time constant (Te). Imagine a queue of 10 vehicles at a stoplight. When the light goes green, the first vehicle increases immediately its speed according to the inertia of the body of the car; the second vehicle is waiting a little bit, not knowing the intentions and fastness of the first one, having a little delay of around 3 seconds and accelerating normally after that; the last vehicle in the row doesnt start for 30 seconds around and only after that develops its capacity response. A similar response is given by a cascade of reactors or evaporators. The transfer and frequency functions are, according to equation (34) ( ) and ( ) (70)

with module and phase lag ( )

and

( )

(71)

Because the sequence of first order elements produce an equivalent dead time and time constant, the equivalent transfer and frequency functions can be expressed in the equation (72) ( ) and ( ) (72)

The Bode diagrams are given in Figure 44

Figure 44

Example of a cascade of two CSTRs. (Figure 45)

34

Figure 45 The CSTR behaves as first order element from the point of view of the transfer path . In the case of a cascade of two reactors, the behavior on the path is that of a 2nd order system. The mass transfer equations describing the variation of the molar concentration CA, with the simplifying assumptions V, F and k are constant for both reactors: ( ( ) ) and (73)

The global equation describing the whole system is

) (74)

Thus, the reactor cascade system behaves as a critically damped second order system with the correspondent response (Figure 46)

35

Figure 46 Element with dead time () behavior. The comportment of the dead time element is described by the equation (75) ( ) ( ) (75)

where is named the dead time. The dead time we are discussing at present is not the former mentioned equivalent dead time but the so called pure dead time provoked by the mass transportation along a line of important length (pipeline, conveyor etc.). The response of the element with dead time behavior at a step input is given in Figure 47.

36

Figure 47 The element gives an output of the same form and magnitude with the input, but only shifted with . Elements with behavior are for example the conveying belts at quarries, or the long methane gas transportation pipelines or any pipeline of important length. ) in McLaurin series around point t, Expanding ( ( ) ( )
( ) ( ) ( )

+..

(76)

and applying Laplace transform to (76) ( ) 4 5 ( ) ( )

The transfer function and frequency function are ( ) ( ) and ( ) The module and the phase angle are ( ) ( ) ( ) The Bode diagrams are presented in Figure 48

) (77)
( ( )

( )

(78)

37

Figure 48 Example of behavior for the dead time element (Figure 49).Let it be a system of mixing hot with cold water; the consumer is placed at the end of the common pipeline after the mixing point. The sudden increase of the temperature at the junction of the two pipelines due to the increase of the hot water at the tap is felt at the end of the pipeline only after the dead time where lc is the length of the pipeline and is the velocity of the liquid with Fc and Fh are the volumetric flows of cold and hot water and Ac is the cross area of the pipeline.

Figure 49 The dead time behavior is typical for long transportation pipelines, conveyors, cracking furnaces, rivers (pure dead time) but also for the distributed parameter systems, constituted from a sequence of elements (distillation columns - sequence of trays; electrolyzers - sequence of anodic frames etc.). The dead time is a difficult element to be handled by the control systems.

38

As a conclusion of this paragraph, one may say that: a. The dynamic state is almost always present because of the frequent disturbances; the steady state is seldom the reality. b. All systems can be examined from the dynamic point of view as a combination of the standard elements: P, I, D, C, . And this is valid not only for chemical or mechanical systems, but also for social, administrative or behavioral systems. c. Mathematical models of the processes are constructed either for steady state and then are used to calculate the constructional dimensions of the parameters or to predict the technological behavior or the economic circumstances, or to elaborate the structure and calculate the components of a control system; or for dynamic state being then used for training the personal of operation, or for predicting the fluctuations in processing capacities, or for tuning the parameters of the controllers. d. Transfer functions and frequency functions are tools for establishing the dynamic behavior and the capabilities of the control systems to control the processes.

2. Analysis and synthesis of the feedback control systems In Figure 9 is described the classical feedback control system. The analysis of the control system is destined to evaluate the construction and behavior of the components to be able to synthesize the control loop as a whole. It is not sufficient to put together all elements, but also to choose and to tune them to work properly together. 2.1. The controlled process In the elaboration of the control scheme everything starts from the understanding the controlled process. There are cases in industry of inappropriately understood processes which are equipped with the state of the art controls without any success. The steady state equations are obtained from the dynamic model annulling the right side of the equation and making the left side equal to 0. Thus, for the heat exchanger in Figure 5, from the heat transfer equations for the jacket and heated fluid ( ) ( )(79) ( ) ( )(80) The steady state equations are obtained making 0 the time derivatives. Thus, (81) giving a dependence To=f(Fag) illustrated in Figure 50.

39

Figure 50 This function expresses the behavior of the heat exchanger on the transfer path xm . It can be noticed that the nonlinearity of the process induces imprecision in the control of the temperature. P1, P2 are the operating curves for the diurnal perturbations (let say Toi=17oC during the day and 10oC during the night). The controller receives the error at the set point ToN= 52oC and in both cases elaborates a control signal xc to correct the deviation. Supposing a P controller with the gain factor KP, it increases or decreases the steam flow in order to keep constant the temperature. Because of the different curvature of the characteristics (higher slope in the region of small agent flow and lower slope in that one of larger agent flow), the control action of the same magnitude in both ways, xc =KP , does not reach the same ToN in both cases. That is why the nonlinearities of the characteristics are not welcome. There are many attempts to linearize the process characteristics using compensatory characteristics of the control valves (Figure 51).

Figure 51 40

At the same time, the slope to the curve N (nominal regime) in the point (FagN, ToN), represents the gain of the process at the set point Kpr=To/Fag. This is used when tuning the controller parameters. If P1 and P2 are the extreme disturbances (let say Toi =22oC during the summer and Toi=4oC during the winter), the values of the Fag corresponding to these operational curves for ToN are the minimum and maximum steam flows for which the control valve should be designed. The steady state characteristic is used also to determine the control solution for the process, this depending on the aspect of the characteristic and the operating point on the characteristic. The dynamics of a process first determine the controllability of the process. Shinskey defines this notion in the most practical way: the controllability is the capacity of the process to be controlled with a classical feedback control system. Considering a process as being a series of capacities, the controllability depends on the ratio Te/e where both are the equivalent time constant and dead time respectively. The intervals for the state of the controllability of a process are given in Table 1. Table 1 Te/e Feedback control system response to a step input Characteristic features of the response damped oscillations Type of the control system needed for a good quality response* feedback control

Underdamped response with closer to 0

cascade control

oscillatory response

feed forward or MPC

*good quality response is understood according to the paragraph 1.2.3, the one respecting the quarter decay rule

Then, the dynamics of the process, together with the dynamics of the equipment in the control loop are used to tune the controller parameters. But more than that, the dynamics help the design of the technological scheme too and also for constructing the simulators for training. 41

In the following section there are given three examples of simulation. The first (Figure 52) describes the time change of tray molar concentration when xF is changed stepwise (the simulation is done using Matlab environment); the second (Figure 53) describes the behavior of a non isothermal CSTR when inputs are changed (the simulation was done in LabView). Both applications were done in the laboratory of Process Control of the Department of Chemical Engineering of the University BabeBolyai, Romania. The third (Figure 54) is a demo produced by COMSOL and describes the heating of a room from a heater positioned under the window. All files are exe. files and can be run under Microsoft Office.

Figure 52

Figure 53 42

Figure 54 2.2 Transducers and measuring systems The transducer is the device in the control loop which measures the controlled variable and transforms it in a signal compatible with the other devices of the system. Usually, although not mentioned, the transducer is linked to a measurement device which indicates the value of the measured variable. A measuring system (Figure 55) is made of two parts: the transducer, the measuring device which can be a recorder or an indicator and the transmission line.

Figure 55

43

The transducer is formed of the sensing element (sensor) and the adaptor which transforms usually the non electric parameter (flow, temperature etc.) in an electric one. The transmission line is either a cable, an optic fiber, but also the signals can be transmitted by radio. For the compatibility of the devices inside the measuring system or the control loop, the adaptor conditions the signal in the so called unified signals. The international standards imposed the ranges of variation2 10 mA, 0 20 mA, 4 20 mA, or 0 5V, 0 10V, . Supposing a flow transducer with its domain of measuring 0 5m3/h, when measuring 2.5m3/h, the signal produced by the adaptor is either 6 mA, 10 mA, 12 mA if the signal is a current, or 2.5mA, 5mA, 0 V if the signal is voltage. (Figure 56). For the pneumatic signals used in actuating the final control elements, the unified signal domain is 0.2 1 bar.

Figure 56

From figure 56, one can observe the usual characteristic of a transducer is linear. There are five main parameters measured and controlled in the process industry: temperature, pressure, flow, level, concentration. In the instrumentation diagrams, these parameters have the symbols T, P, F, L, A (analysis) respectively. One pressure control loop for example, having the functions of recording (R), control (C) and alarm (A - when the pressure exceeds one limit) is presented in figure 57.The circuit has also a tag/ inventory number to make it distinct among the pressure control or measuring systems of the plant.

44

Figure 57 A more detailed block scheme of a flow control and recording system is given in figure 58.

Figure 58 45

Sensors There are many ways of measuring the parameters, each way being adapted to the conditions and range of measurement. All sensors are based on physical or chemical principles of measurement, and all of them are adapted to the measured environment and measured values. Details about the principles of functioning and construction are given in Mc.Millans Process /Industrial Instruments and Control Handbook. Temperature Measuring temperature is important for the reaction conditions and energy balances determining the functioning or economics of a process. The temperature sensors are usually the resistance temperature detector or resistance thermometers (RTD), the thermocouple, pyrometer. The construction of such a RTD is presented in Figure 59.

Figure 59 The measurement principle of the RTD is its variation with the temperature (82) ( ) (82) where is the value of the resistor at and is the coefficient of variation of resistivity. Thus, the steady state characteristic is a linear one (Figure 60a). The RTDs and thermocouples are protected in the measurement environment by sheaths manufactured from a material resistant to mechanical and chemical corrosion. The sheaths induce delays in the temperature transmission expressed generally by a first order capacitive function with time constants between 30 s (rapid thermo resistors) and 5 min (case of ceramic thermowells) (Figure 60b).

Figure 60 46

Pressure Measuring pressure is important especially in the gaseous systems, the pressure indicating the mass of gas in the system. Thus, the mass balance can be determined. Otherwise, the differential pressure is important to be measured in the packed absorption or distillation columns indicating the state of the packing or for flow measurement with variable area flowmeters. The pressure sensors are for absolute, relative and differential pressure. The general traditional principle of measuring the pressure is the shift produced by a membrane, usually, subjected to the pressure of the measured environment. But pressure instrumentation ranges from membrane Bourdon- and bellows actuated devices to modern electro-optic transducers(e.g. hot-filament ionization, piezoresistive/ piezoelectric or capacitive ceramic) without moving parts (Figure 61).

Figure 61 47

The membrane exposed to the pressure shifts its position, modifies the position of a diaphragm and obstructing partially a light beam, proportionally with the pressure change (Figure 61.a). Then, the electric signal of the measuring diode is proportional to the pressure. The static and dynamic characteristics of the pressure transducers are illustrated in Figure 62a and 62b. Because of the negligible mass of the moving elements (membranes, bellows) or the negligible delay of the electronic systems (hot-filament, cold cathode etc.) the dynamic behavior is of a capacitive element with T of the order of seconds.

Figure 62 Flow Measuring the flow is important for knowing the mass input, output or accumulation. It is important for controlling the mass balance of a process. One can categorize the flow measurement in three categories: very small flowrates (mg/s or cm3/s), usual flowrates( < 10m3/h) and large flowrates (> 10m3/h). For these categories, there are different ways of measurement: turbine, thermal mass or magnetic for the first category; rotametric, differential pressure, magnetic for the second category. For large flows (e.g. oil pipelines) turbine, Vortex principle, Doppler or Coand principle flow meters are used. All these flowmeters based on the mentioned principles measure the volumetric flow. There are flowmeters measuring the mass flow, based on Coriolis force principle. The advantage of the mass flowmeters is that the measurement is not affected by fluid properties like temperature, pressure, density, viscosity. Figures 63 and 64 illustrate the principles of functioning of some flowmeters.

48

Figure 63 49

Figure 64 The flow transducers have similar static and dynamic characteristics with pressure transmitters. Level The level indicates the mass accumulated at a certain moment in a system (e.g. bottom of a column, tank, storage bin or chemical reactor). For the chemical reactors the level indicates the residence time which influences the kinetics of the reaction. The principles of level measurement are depending on the operating conditions in the process: use of floats or displacers (Figure 65) for liquid in tanks, columns or reactors; ultrasonic measuring principle or time-domain reflectometry for solid storage bins; use of pressure transmitters placed at the bottom of the recipient containing liquid (Figure 66).

Figure 65 50

Figure 66 In many cases, especially when there is storage of solids, the level measurement can be replaced with weighing systems. The level transmitters have similar static and dynamic characteristics as those presented in Figure 62. Concentration The concentration or the composition of a separation product or mass of reaction is important defining the quality of the product or the conditions in which the process evolving. There are actually two types of concentration transducers: non selective (based on the measurement of the physical properties: density, viscosity, conductibility, pH, humidity etc.) and selective (based on the separation of the components of the flux and the measurement of the physical property of each component - liquid or gas chromatographs). Density is the parameter measured for detecting the composition of a binary liquid mixture (e.g. alcohol-water, acid-water) (Figure 67). Viscosity is the measure of the degree of completion of a polymer reaction and hence it is used in the polymer manufacturing processes. The

51

Figure 67 electric conductivity of solutions is a measure for the concentration of the electrolytes (acids, bases, salts); it is based on the correlation conductibility-concentration (Figure 68, 69).

Figure 68 52

Figure 69 The scheme of a conductibility transducer is presented in Figure 70. Since the electrical conductivity depends on temperature through the solution resistivity ( ( )), in the Wheatstone bridge it is introduced a compensation with temperature resistance.

Figure 70 The gas analyzers measure the concentration of a certain gas in a mixture (e.g. CO2 in a mixture of CO2, N2, CO or H2 in a mixture H2 with air) based on some physical principles as thermal conductivity or infrared absorption of gases. An analyzer of H2 based on thermal conductivity is presented in Figure 71. The thermal conductivity is defined in equation (83) 53

(83)

where Q is the heat quantity transferred through the area A in the time t at the temperature gradient . In the Table 2 are presented the values of the thermal conductivity for different gases. The dependence Table 2

of on the concentration of the gases in a mixture is given by an additive relationship (84), + (84)

where are the mass fractions of the components of a binary gas mixture (e.g. air - H2 or air CH4). The gas analyzer based on thermal conductivity is presented in Figure 71. To slightly overheated resistors are placed in two cells: one is sample cell filled with a reference gas and the other one measuring cell (filled with the gas mixture). The gas whose concentration is interesting in the mixture should have a high (e.g. H2 or He) and different from the others one (e.g. N2). In this way, the resistor from the MC is cooled much stronger than that in the SC and thus, the bridge is unbalanced, producing a voltage e proportional with the concentration of H2 in the mixture.

Figure 71 54

The static and dynamic behavior of an analyzer is correspondent with the behavior in Figure 62. The gas chromatography is actually the only selective method for analysis. The working principle is that a solid stationary phase, placed in the so called chromatographic column (Figure 72) retains differently the different components of the analyzed gas, function of their molecular mass. The successive separation is done by another carrier gas which is not retained by the packing of the chromatographic column.

Figure 72 In a chromatogram, a certain gas i is positioned always at the same time ti. The carrier gas transports the i component to the gas analyzer (e.g. for thermo conductibility or infrared absorption) and a peak signal proportional with a certain concentration is emitted. Because the gas chromatographs are smart transducers, they are very expensive, thus being used in a factory for the analysis of n fluxes each with m components. The analyzer analyses the fluxes in the same order, one after the other. In this way, if one component, say Cl2, is the component j appearing at the time tj in the analysis of the flux k<n, next time when Cl2 is measured is n tan,k tj where is the duration of the analysis of the flux k. If the analyzer is used in a control loop, the concentration of Cl2 is interpreted to be the same between two consecutive measurements of the flux k. That is why it is 55

considered that the gas chromatograph is an element with dead time = . In order to increase the accuracy of the analysis, usually the range of analysis is very narrow. Thus, the gain factor of the analyzer is very high (e.g. Kan=1000). 2.3. Controllers The controller C from the block diagram of the control loop (Figure 9), isthe device which calculates the control signal, according to an embedded control algorithm and the sign and magnitude of the error. The controller output is thus calculated to bring the deviation of the controlled variable from the set point, to 0. There are several types of controllers, continuous, numerical and on-off controllers. The last category is used in simple processes as those from the household (refrigerator, washing machine, central heating). The industrial controllers are shown in Figure 73.

Figure 73 The most frequently used controllers are those having P, PI or PID algorithm. The controller P is used to give a proportional kick to the manipulating variable to counteract the error. It has the structure of a P element with its behavior described by the equations (42) - (44). ( ) ( ) ( ) (85)

( )is the controller output The controller gain factor is KP and is the error of the control system. corresponding to the steady state of the process where the process output is equal to the set point of the control system. The larger the gain is, the larger the controller output is and stronger the controller action upon the actuator. This is obvious from the steady state characteristics presented in Figure 74. 56

Figure 74 The industrial controller is standardized and has to cope with different situations: processes with a very high gain (as neutralization of a strong acid with a strong base) or processes with a lower gain the waste water treatment having a pH control loop (Figure 75), for the strong acid strong base system which has a very steep neutralization curve at the set point pH = 7 (Figure 74), the error is large and the dosage of base should be very fine; this implies a very small opening of the control valve which is induced by a small KP. If in the waste treatment station, arrives a weaker acid as effluent from the industrial plant, the dosage should be stronger, because of the lower slope of the neutralization characteristic. Thus, KP should be larger.

Figure 75 An industrial controller has to cope with several situations similar to the above mentioned examples. This is the reason, all controllers have the possibility to change all their tuning parameters in a certain 57

range each. For the P controller, Kc has the possibility to be changed in the range of 0.5 - 20. The industrial controllers sometimes work with the Proportional Band (in%), . The P controller is characterized by the steady state error st. When a disturbance occurs, the P controller shifts abruptly the position of the control valve to compensate the influence of the disturbance but the process never reaches the set point. A detailed explanation is given in Agachis book Automatizarea Proceselor Chimice (Chemical Process Control) in chapter 6. The larger the controller gain is, the smaller the steady state error is, up to the point when the increase of the gain induces instability in the loop (Figure 76). Generally, the rule of thumb for tuning the P controller is that when the gain of the process at the set point is large, the gain of the controller should be small and viceversa.

Figure 76 The PI controller combines the proportional kick which rapidly compensate the dynamic error, with the integral component of the controller output (Figure 77) which cancels the steady state error (the signal increases until the error becomes 0).

Figure 77 58

The equation of definition is (86) ( ) ( ) 0 ( ) ( ) 1 (86)

where is the integral time or reset time. This is the time when a step increase of the controller output ( ) is the controller output corresponding to the steady state of the doubles its value (Figure 77). process where the process output is equal to the set point of the control system. The rule of thumb for tuning Ti is that it has the same order of magnitude with the dominant time constant in the process. If it is too small, induces instability in the system and if it is too large, the I action is not efficient, not eliminating the steady state error in due time. The most used industrial control algorithm is the PID control either in the analogue or numerical form. The PID controller is the combination of the P, I and D elements (paragraph 1.2. 5. Types of dynamic behavior).The function expressing the control algorithm is ( ) ( ) 0 ( ) ( )
( )

(87)

where ( ) is the controller output, ( ) is the error of the control loop and , and are the ( )is the controller controller gain, the integrative/reset time and derivative time respectively. output corresponding to the steady state of the process where the process output is equal to the set point of the control system. The proportional action has the role of immediate response at the error (proportional kick), the integral action has the role of annulling the steady state error and the derivative action to diminish the delays imposed by the process (derivative kick). The response of the P, PI and PID controllers at a step input (e.g. 10% step increase of the set point of the temperature control system from 70 to ), with Kc=1, Ti=60s and Td=5s is illustrated in Figure 78.

Figure 78 59

The transfer and frequency functions are (88) ( ) 0 1 ( with the module and phase lag (89) ( ) ( ) (89) ( ) ( ) ) [ ] (88)

The Bode diagram for the PID controller controlling a heating process is presented in Figure 79

Figure 79 60

When the stability of the control loop is discussed, it is demonstrated that if the sum of the phase lags equals -180o the system is at the oscillation limit, . This means that, if a disturbance occurs and has its fundamental own frequency ( )the same with or close to the of the control system, the system goes unstable. As it was stated before, the in the process industries are at very low frequencies (0.006-60 min-1) meaning that the fastest disturbances perform less than one oscillation per second and the slowest are at less than one oscillation per hour. This means that if, by an appropriate control action, it is succeeded to push the oscillation frequency of the whole control system (process and controller) in a range with higher frequency than the system behaves stable. This is observed in the Figure 79 when an appropriate derivative time Td2 is chosen and the result is . The rule of thumb in tuning the parameters of the PID controller is Ti . Going backwards to the Figure 7, where an intelligent manual operation of the heat exchanger is presented, we may notice that that manual operation is similar to a PID control action with the beneficial effect. P controllers are used for simpler processes, where there is no need of accuracy (e.g. level control in raw material tanks). PI controllers are used generally for fast processes (e.g. flow, pressure control). PID controllers are generally used when slow processes are to be controlled (e.g. temperature or concentration control). 2.4. Control valves and actuators The last element discussed from the control loop(AD), is the final control element, or actuating device. This is the device which, having as input the controller signal, transforms it in a flow of mass or energy which brings the process output back to the set point. The ADs can be of different types: control valves, centrifugal pumps, conveyor belts, screws, all with variable speed etc. The most used AD is the control valve (Figure 80), this being the reason for which it is described in this chapter.

61

Figure 80 The actuator, which is the part of the AD, receives the control action from the controller and transforms it in a shift of its stem (stem travel), proportional with the magnitude of the control action. The pneumatic actuator which is very common, has a membrane inside which shifts its position proportional to the controller signal. A stem is attached to the membrane and at the other end it has a valve plug which clogging a flow orifice (seat ring). The stem travel ranges from 0% (closed valve) to 100% (completely open valve) (Figure 81) Different construction models of the valves are appropriate to different working conditions: for slurry there are used ball valves; for gases at low pressure swing-through butterfly valves; for highly corrosive processes, rotary eccentric plug valves are used.

62

Valve sizing The necessary flow of the process is calculated from its steady state characteristic (Figure 50). The valve is sized in such a way that it allows the maximum flow calculated from the maximally disturbed characteristic. From Bernoullis law, the liquid flow equation is deduced: where (90)

is the volumetric flow passing through the valve, is the valve sizing coefficient, or K Wert, are the upstream and downstream pressure and is the density of the fluid passing through the valve. The valves are produced in series by the valve manufacturers according to the in the catalog. Thus, (91)

and this value is the characteristic one for a valve: is calculated for a certain flow and compared with the manufacturers catalog value . Usually, the next larger is chosen for the sized valve. But not only is defining the control valve. Another important characteristic is the intrinsic characteristic which expresses the relationship between the and the stem travel h, ( ).In Figure 81, there are presented the intrinsic characteristics, meaning own characteristics of the valve. From this point of view, there are four types of control valves: with linear, equal percentage, quick opening and modified parabolic characteristics. is the sizing coefficient at closed position of the valve.

Figure 81

63

The equations defining the four types of behavior are: ( ) linear characteristic

0 (

equal percentage (logarithmic) (92) modified parabolic

and quick opening valve. .

is the maximum valve sizing coefficient for the fully opened valve at

But the control valve is not isolated in the hydraulic/pneumatic system of transportation (Figure 82).

Figure 82 Beside the control valve pressure loss local pressure drop linear pressure drop and = hydrostatic pressure drop. 64 (93) in the system there are:

where and are the local and linear pressure drop coefficient, are the level differences summed in the system, lj and dj are the lengths and the diameters of the pipeline sections, the total system pressure drop is then (94) Including all pressure drops in the characteristic equation of the hydraulic/pneumatic system, the working characteristics of the control valve are obtained (95) and shown in Figure 83:
( ) .
( )

(95)
/ ( )

where

is the maximum flow passing through the valve at

Figure 83 65

It is obvious that the intrinsic characteristic is not respected but in the cases when meaning that the control valve is mounted on a very simple line. The working characteristics are the real operational ones and they can be used in compensating the nonlinearities of the process (Figure 51). 2.5. Synthesis of the control systems In the sections 2.1 - 2.4 the elements of the control system have been presented. This is not enough to make a control system to work properly. There are practical tips of how to design a process to be more stable from construction, how place the operational point in a process to have efficient control, where to place the transducers in the process to avoid lags, what type of transducer or control valve to be chosen in a specific environment etc. Since all these demands are accomplished, one has to tune the controllers parameters, in order to match with the characteristics of the other elements in the control loop. The controller tuning parameters are the only possibilities to assure the matching of all elements. Tuning The oscillations in the control loop are similar to those of any oscillatory system: pendulum, basketball ball and player etc. (Figure 84).

Figure 84 The conditions for a stable oscillatory process (necessary in the basketball game), are that the action on the oscillating object to be applied with the same magnitude and at a constant time lag (one period T). The corresponding module for the control loop is in this case ( ) = 1. ( ) (96)

and the corresponding phase lag is feedback control introduces a ( ) loop remains to be ( )

. Because the summing point in the case of the the total phase lag of the other elements in the control

(97)

Conditions 96 and 97 are the conditions for keeping stable oscillations in the control loop. But the appropriate functioning regime in the loop, according to the quarter decay condition, is that every consecutive oscillation of the controlled variable is damped to one half of the previous one 66

. This imposes the condition (96) to be stricter ( ) (98)

Thus, the conditions for a damped quarter decay regime are (97) and (98). Tuning controllers means choosing the controller parameters (Kc for P controllers, Kc and Ti for PI controllers and Kc, Ti and Td for PID controllers). The tuning can be done experimentally when the process allows to be disturbed, or theoretically, based on a dynamic model, when the process does not allow imposed disturbances which could bring the process to its functioning limits. The experimental methods are simple and do not need sophisticated calculations. Ziegler - Nichols method of quarter decay ratio response by ultimate gain is the oldest and most used because it is simple and does not need special preparations for the experiment. The dynamic characteristics of the process are extracted from process behavior and they are the ultimate gain of the proportional controller and ultimate period of oscillation of the control loop. The steps of the experiment are: Step 1. With the integral and derivative functions of the PID controller switched off (Ti = or 30min maximum value and Td = 0), and the controller in automatic functioning mode, the gain of the controller is continuously increased in small steps until the loop oscillates with constant amplitude. At this point, the ultimate gain (Kcu) and ultimate period of oscillation (Tu) are recorded for afterwards use. For all experiments with a changed controller gain, the same step change (disturbance or set point) is repeated.(Figure 85)

67

Figure 85 68

Step 2. The estimation of the tuning parameters deduced from the dynamic behavior are illustrated by Ziegler and Nichols in Table 3. Table 3

If there is no possibility of bringing the process to instability (dangerous unstable processes), there is another experimental method useful in these cases, process step testing. The step test is carried to a control loop (Figure 86) where the process is understood as all elements of the control loop with the exception of the controller.

Figure 86 With the controller on manual mode it is applied to the process a step change of the control action xc. The response of the transducer xr is recorded being obtained a diagram of the type represented in Figure 87. On the diagram there are identified the equivalent dead time e, the equivalent time constant Te and the process gain Kpr . The optimal controller parameter values are suggested by Ziegler and Nichols and presented in Table 4. It can be noted that it is a correlation between the controllability ratio Te/e and the changes imposed to the controller parameters values (better the controllability, higher Kc for accuracy and rapidity).

69

Figure 87 Table 4 Controller parameter BP Ti Td Controller structure P PI PID 0.7 K 100 K 100 1.1 K 100 T T T 2 2.5 0.4

If the process does not allow experiments, there are several analytical methods used for controller parameter tuning: minimum error integral tuning, Internal Model Control tuning, controller tuning based on Routh or Nyquist stability criteria and also, the former presented quarter decay ratio tuning. Here it is presented the method based on the quarter decay ratio criterion. As presented in the equations (97) and (98), the criteria for having a quarter decay ratio damping of the closed loop response to either a disturbance or a set point change are: ( ( ) ) ( ( ) ) ( ( ) ) ( ( ) ) ( ( ) ) ( ( ) ( ) ) (99) (100)

From equation (99), which is a transcendent implicit equation, is calculated as being the frequency at which the control system oscillates. With the calculated, the only unknown remains ( ) from which the KP is deduced. If the structure of the controller is PI or PID, the time constants are estimated as follows: Ti is approximated to the dominant time constant of the loop (of the process, final control element or transducer) and Td is calculated from the relationship of phase margin: in Figure 79 70

it is shown that in order to push away in the region of higher frequencies, not common to the chemical processes, and thus to stabilize the loop, the phase margin should be +30o. ( ) . /= (101)

Example for tuning Supposing a flow control loop with the following characteristics: Tpr= 0.5s, Kpr= 1 ; TT =0.2s, KT=2.5 ; TAD=3s, KAD=1.3

and governed by a P controller.

All elements, flowing process, flow meter and control valve have capacitive behavior, defined through equations (59) and (60).The equations (99) and (100) become

and from this equation

and

This controller gain assures the quarter decay ratio damping of the oscillations of the flow control loop output, when a set point change or a disturbance occur. 3. Advanced control systems If the process has a weaker controllability, as from the Table 1 is observed, there are needed more complicated and more expensive control systems. 3.1. Cascade and feed-forward control systems Cascade control The cascade control is useful where a slow process can be divided fictionally in two parts, one faster and the other one slower on the transfer path . (Figure 88). The cascade control system has two control loops: one interior, processing the intermediate process response and one exterior, watching the main variable of the process (Figure 89).

71

Figure 88

Figure 89 The disturbance P2 does not propagate till the end of the process, being corrected in the inner loop. In the case of exothermic CSTR, the disturbance in the temperature of the cooling agent is felt almost immediately by the temperature sensor in the jacket and much later by the temperature sensor placed in the reactor (Figure 90).

72

Figure 90 As it is seen, the structure is more appropriate for rejecting of the D2 type, but also to the D1 type as well. The maximum efficiency of the cascade is when 3 (102)

The structure of the controllers is: C2 has P structure because it is not important to eradicate the inner ; C1 has PI or PID structure depending on the rapidity of the slower part of the process. Tuning the parameters of the controllers Step 1. With C1 in manual mode, the tuning is done for the controller C2 (section 2.5 - Tuning). Step 2. With C2 already tuned and on the automatic mode, the inner control loop is consider an element in the outer control loop. C1 is tuned according to the mentioned methods. The efficiency is illustrated in the following example: a dryer is subjected to two types of temperature control - regular feedback and cascade (Figure 91).

73

Figure 91 Supposing a variation of the quality of the steam heating the dryer (D2 disturbance) with an optimal tuning of the controllers, the response of both control systems is given in Figure 92.

74

Figure 92 It is observed that the overshoot and damping are smaller and much faster respectively for the cascade. The integral of the temperature curve (total area under the curves) gives is the energy consumption. The ratio = 4 showing the energy consumption is four times smaller with a cascade control system. Feed forward control When the controllability of the process is very bad, there are needed more complicated and more expensive control systems as feed forward or model predictive. The feed forward control system (Figure 93) is used when a disturbance is dominant and can be measured. When the disturbance is detected by the transducer of the disturbance, the disturbance controller is so smart that, knowing the disturbance and of the manipulating variable propagation through the process, compensates the deviation of the output due to the disturbance by an output signal opposite as sense and equal as magnitude (Figure 94).

Figure 93 75

Figure 94 Thus, the variation of the controlled variable is zero meaning = 0. The synthesis of the disturbance controller is based on the desired condition
( ) ( )

or

( ) ( )

(103) ( ), ( ) ( ) ( ) ( )- and ( ) ( ) (104)

Thus from the expression of ( ) , and with ( ) ( ) ( ) ( ) ( ), ( ) ( ) ( )( )

(105) (106)

( ) ( )
( ) ( ) ( ) ( ) ( ) ( ) ( )

( )

( )

( ) ( ) ( )

( ) ( )

( ) ( )

( )

(107)

( ) is not influenced by ( ) if the nominator of the From condition (103), one can be deduced that second fraction from (107) is 0. From this condition, results the transfer function of the Disturbance Controller ( )
( ) ( ) ( ) ( )

(108)

Example of feed forward control of the bottom molar composition ( ) of a binary distillation column (Figure 95). The main and measured disturbance is the feed flow F. The ( ) function is calculated ( ) function is calculated on the transfer path Fag thus on the transfer path F . The 76

Figure 95 On the F path there are n trays in series, and taking into consideration that each tray of the column has a capacitive behavior described by a transfer function (59), the equivalent transfer function similar to that in equation (72), is ( ) On Fag function ( ) (109) path there is only one main element, the bottom of the column and thus the transfer ( ) is (110)

77

where KB and TB are the gain and the time constant of the heating process in the bottom. With the other transfer functions needed for the synthesis of the feed forward controller, ( ) ( ) , the function of the actuating device and , the function of the disturbance transducer, (111) (112)

The disturbance controller transfer function is ( ) (113)

Niesenfeld et al. show the advantages of the feed forward control in an experiment made on a column (Table 5). Table 5 Disturbance , Control structure Settling time of the concentration oscillations [min] 10 No oscillations Maintained oscillations 3

60 - 70 70 - 80

PID Feed forward PID Feed forward

The authors mention an energy saving of 15%. Marioniu and collaborators mention an experiment made on a propane - propylene column with 90 trays and a steam consumption of Fs = 4 t/h. With feed forward control there was reported a saving of S=20% steam consumption. Calculating simply the value of the energy saved in one year one may estimate the value of such a control system. Annual Energy Saving = Fs S lv functioning time/year = . If the average price is around 40 - 50 USD/Gcal, the saving is around 120,000 - 150,000 USD/year. If there are calculated the initial investments in equipment and labor, they sum up at around 60,000 - 80.000 USD/year, meaning the investment is recuperated in less than six months. 3.2. Advanced Process Control (APC) APC is the state of the art control technique applied to process engineering. It involves the computer techniques (hardware and software) applied to control, the good understanding of the process (process modelling and design), the good understanding of control techniques (Model Predictive Control, Inferential, Adaptive, Optimal and Fuzzy Control, Distributed Control Systems etc.). In the processes with multiple variables (hundreds or thousands) an efficient operation of the process, of the plant or of the industrial platform, can be done only through APC. Moreover, the concept of heat integration (recovery of the heat from the hot effluents for heating the cold ones) which is economically very effective, brings an increased instability of the controlled processes: the effect of all disturbances in the process is brought by heat integration to the entrance. APC is the appropriate tool for coping with this 78

problem. Central Control Rooms looking like cosmic flight control centres are supervising and operating the whole plant. New skills of the operators are requested, because the implications of their actions are on a large scale. With APC one can save energy, raw materials, time, reduce costs and make the processes more competitive in terms of quality and costs. A comprehensive picture of APC is given in Figure 96.

Figure 96

79

Model Predictive Control The most prominent control technique of APC, is the Model Predictive Control. The MPC approach is simply presented in Figure 97.

Figure 97 The present moment is considered to be at t = kT. The future process behaviour is considered for a time span ranging from t = kT to t = pT, interval usually denoted as the output horizon. On the basis of the step input model and the current measurements, the process outputs are predicted over the output horizon. The predicted output is function of the known past, but also of not yet known future manipulated variables. The future manipulated variables will be computed on a time span ranging from t = kT to t = mT, interval usually denoted as the input horizon. The future manipulated variables will be computed such as the predicted outputs conform to a desired (setpoint) output trajectory. The first computed manipulated variable is sent to the process and maintained constant up to the new sampling moment t = k+1, when the all the computation is performed again with the output and input horizons are shifted one time step forward. This procedure produces the so called moving or receding horizon control approach which is one of MPCs characteristics. In Agachi et.al., a broad range of applications is given: PVC reactor, NaCl electrolyzer, FCC plant, bioreactor, distillation column.

4. Automation of complex processes An industrial process like that of synthesis or separation is more complicated from the control point of view. There are many interdependencies between input, state and output variables which produce the character of uncontrollability. These interdependencies can be treated either through APC approach for MIMO systems, or in the classical way, with regular feedback control systems, with controllers tuned wisely. The processes are of synthesis (in reactors of different types as CSTR, plug flow, mass transfer reactors etc.) and separations (distillation, absorption/ desorption, extraction, evaporation, drying etc.). When a control structure is established, it is used, of course, the industrial experience of previously operated processes and the steady state modeling techniques. 80

To illustrate this approach, we take the example of the evaporation with multiple effect. In Figure 40, it is shown an evaporator with double effect. To establish the control solution for the evaporator, one has to construct the steady state mathematical model, expressing the final concentration function of the process parameters (inputs and states). The following simplifying assumptions are done: The feed of the two evaporators is done at the boiling temperature of the solution, manipulating the pressures in the evaporators, p1 and p2; thus, the heat consumed for evaporation is only the latent heat of vaporization, lv; - The vapors do not entrain the solid fraction; in means that in the component mass balance, there is no solid fraction in the vapor phase; - The steam used for heating condenses totally delivering to the evaporator only the latent heat lv; this means that the whole heat delivered in condensation is given fully for evaporation; - The heat losses to the exterior are considered null; thus no term of heat transfer to the exterior is in the energy balance. The model, in these conditions is: (114) mass balances for both evaporators (115) energy balances for both evaporators (116) -

where F, F1, F2 are the liquid flowrates, V1, V2 are the solvent vapor flowrates, xF, x1, x2 are the mass concentration of the solid solved and lv0, lv1, lv2 are the latent heat of vaporization of the steam and solvent in the two evaporators respectively. From equations (114) - (116) result the expression of the output x2 function of inputs and states ( ); are functions of evaporators pressures.
( )

, or generalizing for n effects,

(117)

(118)

From relations (117) and (118), the control solution can be deduced: a ratio control for , level controls for keeping the mass balances from (114), pressure control for keeping constant lv. Thus, the control scheme is given in Figure 98.

81

Figure 98 Such an approach for establishing the control solution is used for all processes.

82

Você também pode gostar