Control feedforward

We have no control over the flow of stream A - indeed changes in its flow are the main disturbance to the process. 

The feedback scheme is limited in that it can only take corrective action once it has detected a deviation from temperature SP. This is particularly important if there is significant deadtime between the PV and the MV, for example if the temperature measurement was a long way downstream of the mixer. No matter how well-tuned the feedback controller, it cannot have any impact on the PV until the deadtime has elapsed. During this time the error ( ) will increase to 

Most of the control systems we have discussed, simulated, and designed thus far in this book have been feedback control devices. A deviation of an output variable from a setpoint is detected. This error signal is fed into a feedback controller that changes the manipulated variable. The controller makes no use of any information about the source, magnitude, or directiop of the disturbance that has caused the output variable to change. 

I The basic notion of feedforward control is to detect disturbances as they enter the process and make adjustments in manipulated variables so that output variables are held constant. We do not wait until the disturbance has worked its way through the process and has disturbed everything to produce an error signal. If a disturbance can be detected as it enters the process, it makes sense to take immediate action in order to compensate for its effect on the process. 

Feedforward control systems have gained wide acceptance in chemical engineering in the past two decades. They have demonstrated their ability to improve control, sometimes quite spectacularly. We will illustrate this improvement in this section by comparing the responses of systems with feedforward control and with conventional feedback control when load disturbances occur. 

Feedforward control is probably used more in chemical engineering systems than in any other field of engineering. Our systems are often slow-moving. 

Feedforward control is a control action in which information concerning one or more conditions that can disturb the controlled variable is converted into corrective action to minimize deviations of the controlled variable. In a true feedforward control system, the controlled variable is not used as one of the Inputs. However, it is usually difficult, as well as not practical, for the controller-computer to include corrections for all variables. Therefore it is customary to include a feedback loop to prevent inaccuracies in the feedforward loop from adversely affecting the process. 

In most evaporator applications the control of product quality is constantly affected by variations in feed rate and composition to the evaporator. In order to counter these load variations, the manipulated variable must attain a new operating level. In pure feedback or cascade arrangements, this new level is achieved by trial-and-error as performed by the feedback controller. 

A control system able to react to these load variations when they occur rather than wait for them to pass through the process before initiating a corrective action would be ideal. The technique is called feedforward control. There are two types of load signals are inputs to the feedforward control system where they compute the set point of the manipulated variable control loop as a function of the measured load variations. The unmeasured load variables pass through the process, undetected by the feedforward system, and cause an upset in the controlled variable. The output of the feedback loop then trims the calculated value of the set point to the correct operating level. In the limit, feedforward control would be capable of perfect control if all load variables could be 

At a practical level, the load variables are classified either as major or minor, and the effort is directed at developing a relationship which incorporates the major load variables, the manipulated variables, and the controlled variable. Such a relationship is termed the steady-state model of the process. Minor load variables are usually very slow to materialize and are hard to measure. Minor load variables are easily handled by a feedback loop. The purpose of the feedback loop Is to trim the forward calculation to compensate for the minor or unmeasured load variations. 

The third Ingredient of a feedforward system Is dynamic compensation. A change In one of the major loads to the process also modifies the operating level of the manipulated variable. If these two Inputs to the process enter at different locations, these usually exists an unbalance or inequality between the effect of the load variable and the effect of the manipulated variable on the controlled variable. This imbalance manifests itself as a transient excursion of the controlled variable from set point. If the forward calculation is accurate, the controlled variable returns to set point once the new steady-state operating level is reached. 

Feedforward control adjusts the manipulative variable as soon as a load disturbance is detected in an attempt to cancel its effect on the controlled variable. Sampled-data feedforward algorithms are designed readily from knowledge of the dynamics of the process. Feedforward control is best employed in conjunction with feedback control to correct for any offset due to inaccuracies in the process model. A typical feedforward/feedback sampled-data loop is 

C z) is set equal to zero, which says that there will be zero error at each samplmg point. Di z) is solved from equation (21.12). 

Df(z) can be determined from equation (21.13) once the nature of the load and the process model are known. Let us calculate the feedforward algorithm for a feed-composition disturbance from the infexmation in Table 21.1 for a step-load change. 

The time output Dpit) from equation (21.14) is the required feedforward change in the manipulative variable after a change in the feed ccxnposition is deterted. 

If the load form is not known exactly, L s) may be approximated by a staircase frinction L (s)H s). The asterisk dmotes that L s) is a sampled variable. This simply says that the load is fictitiously sampled and the value is held constant for the sampling period. The numerator of equation (21.13) be x mes  

However, since control action can only occur if a deviation occurs between the set point and the measured variable, perfect control is not possible. Therefore feedback control fails to provide predictive control action to compensate for the effects of known disturbances. A more serious limitation, which is particularly important for polymer reactor control, is that the controlled variable cannot always be measured on-line. 

Feedforward control was developed to counter some of these limitations. Its basic premise is to measure the important disturbance variables and then take corrective compensatory action based on a process model. The quality of control is directly related to the fidelity and accuracy of the process model. Two implementations of feedforward control which are widely used in polymer reactor control will be discussed, namely feedforward control design based on steady-state models, and 

A feedforward controller has the potential of perfect control, however, in reality it has several disadvantages  

1 Feed-Forward - Feedback Control. In practical applications, feedforward control is normally used in combination with feedback control. The feedforward part is used to reduce the effects of measurable disturbances, while the feedback part compensates for inaccuracies in the process model, measurement errors and unmeasured disturbances. The feedforward and feedback controllers can be combined in several different ways, as discussed in most standard control text books. 

when tuning the primary controller there should be no interaction between the primary and secondary loops. If there is, it means that the primary loop is not slow enough in comparison with the secondary. 

One of the most common forms of cascade is the output of a primary conttoller acting as a set point to a valve positioner. 

In its simplest form, a feedforward controller merely proportions the corrective action to the size of the disturbance. In other words, the control equation is merely a gain based on steady state, i.e. mass or energy balance at steady state. This does not take into account any of the process dynamics of the system. If there is a difference, or lag, in the speed of the process response to the control action when compared with that of the disturbance, then it may be necessary to introduce some dynamic compensation into the control equation. The dynamic compensation correctly times the control action and response, thus giving increased accuracy in the feedforward control. 

In general, the feedforward dynamic elements will not be physically realizable. In other words, they cannot be implemented exactly. For instance, if the process disturbance measurement contains dead time, or lag, then the feedforward dynamic compensation would have to be a predictor, which of course is impossible unless an exact 

When feedforward control is used, equations are needed to calculate the amount of the manipulated variable needed in order to compensate for the disturbance. This sounds simple enough however, the equations must incorporate an understanding of the exact effect of the disturbances on the process variable. Therefore, one disadvantage of feedforward control is that the controllers often require sophisticated calculations, as even steady models can be nonlinear and thus need more technical and engineering expertise in their implementation. 

Feedback control can never be perfect as it reacts only to disturbances in the process outlet. Feedforward control can theoretically be perfect, because the inlet disturbances are measured, and their effects on the process are anticipated via the use of a model. If the model is perfect then the calculated action to be taken will be exact. 

The example simulation THERMFF illustrates this method of using a dynamic process model to develop a feedforward control strategy. At the desired setpoint the process will be at steady-state. Therefore the steady-state form of the model is used to make the feedforward calculations. This example involves a continuous tank reactor with exothermic reaction and jacket cooling. It is assumed here that variations of inlet concentration and inlet temperature will disturb the reactor operation. As shown in the example description, the steady state material balance is used to calculate the required response of flowrate and the steady state energy balance is used to calculate the required variation in jacket temperature. This feedforward strategy results in perfect control of the simulated process, but limitations required on the jacket temperature lead to imperfections in the control. 

The success of this control strategy depends largely on the accuracy of the model prediction, which is often imperfect as models can rarely exactly predict the effects of process disturbances. For this reason, an additional feedback loop is often used as a backup or to trim the main feedforward action, as shown in Fig. 2.25. Many of the continuous process simulation examples in this book may be altered to simulate feedforward control situations. 

An adaptive control system can automatically modify its behaviour according to the changes in the system dynamics and disturbances. They are applied especially to systems with non-linear and unsteady characteristics. There are a number of actual adaptive control systems. Programmed or scheduled adaptive control uses an auxiliary measured variable to identify different process phases for which the control parameters can be either programmed or scheduled. The best values of these parameters for each process state must be known a priori. Sometimes adaptive controllers are used to optimise two or more process outputs, by measuring the outputs and fitting the data with empirical functions. 

---

Generic Control Strategies. The two generic strategies for process control are feedback and feedforward control. Most process control strategies are based on one or a combination of these strategies (1 3). 

The feedforward control strategy (Fig. lb) addresses the disadvantages of the feedback control strategy. The feedforward control strategy measures the disturbance before it affects the output of the process. A model of the process determines the adjustment ia the manipulated variables(s) to compensate for the disturbance. The information flow is therefore forward from the disturbances, before the process is affected, to the manipulated variable iaputs. 

The primary advantage of the feedforward over the feedback control strategy is that corrective action is initiated before the controlled variable is upset. Feedforward control, however, has its own drawbacks, ie, variables used to characterize the disturbances must be measurable a model of the response of the controlled variable to the disturbance must be available (when the feedforward strategy is used alone, the control performance depends on the accuracy of the model) and the feedforward control strategy does not compensate for any disturbance not measured or modeled. 

In most process plant situations where feedforward control is appropriate, a combination of the feedforward and feedback control is usually used. The feedforward portion reduces the impact of measured disturbances on the controlled variable while the feedback portion compensates for model inaccuracies and unmeasured disturbances. This control strategy is referred to as feedforward control with feedback trim. 

The decision to implement a feedforward control strategy should be based on the quahty of control requked, the nature and frequency of the... 

Stea.dy-Sta.teFeedforwa.rd, The simplest form of feedforward (FF) control utilizes a steady-state energy or mass balance to determine the appropriate manipulated variable adjustment. This form of feedforward control does not account for the process dynamics of the disturbance or manipulated variables on the controlled variable. Consider the steam heater shown ia Figure 15. If a steady-state feedforward control is designed to compensate for feed rate disturbances, then a steady-state energy balance around the heater yields ... 

Fig. 15. Example of steady-state feedforward controls, where + indicates the summation of signals. Terms are defined in text.

Ratio and Multiplicative Feedforward Control. In many physical and chemical processes and portions thereof, it is important to maintain a desired ratio between certain input (independent) variables in order to control certain output (dependent) variables (1,3,6). For example, it is important to maintain the ratio of reactants in certain chemical reactors to control conversion and selectivity the ratio of energy input to material input in a distillation column to control separation the ratio of energy input to material flow in a process heater to control the outlet temperature the fuel—air ratio to ensure proper combustion in a furnace and the ratio of blending components in a blending process. Indeed, the value of maintaining the ratio of independent variables in order more easily to control an output variable occurs in virtually every class of unit operation. 

Ratio control and multiphcative feedforward control, in general, are subject to the same considerations. Ratio control can be of a steady-state or a dynamic form. It is often implemented using a setpoint as the load variable when the load variable has a controller associated with it and the controller is in auto mode. 

Feedforward Control A reedfoi ward system uses measurements of disturbance vai iables to position the manipulated variable in such a way as to minimize any resulting deviation. The disturbance... 

While the single-loop PID controller is satisfactoiy in many process apphcations, it does not perform well for processes with slow dynamics, time delays, frequent disturbances, or multivariable interactions. We discuss several advanced control methods hereafter that can be implemented via computer control, namely feedforward control, cascade control, time-delay compensation, selective and override control, adaptive control, fuzzy logic control, and statistical process control. 

Feedforward Control If the process exhibits slow dynamic response and disturbances are frequent, then the apphcation of feedforward control may be advantageous. Feedforward (FF) control differs from feedback (FB) control in that the primary disturbance or load (L) is measured via a sensor and the manipulated variable (m) is adjusted so that deviations in the controlled variable from the set point are minimized or eliminated (see Fig. 8-29). By taking control action based on measured disturbances rather than controlled variable error, the controller can reject disturbances before they affec t the controlled variable c. In order to determine the appropriate settings for the manipulated variable, one must develop mathematical models that relate ... 

The effect of the disturbance on the controlled variable These models can be based on steady-state or dynamic analysis. The performance of the feedforward controller depends on the accuracy of both models. If the models are exac t, then feedforward control offers the potential of perfect control (i.e., holding the controlled variable precisely at the set point at all times because of the abihty to predict the appropriate control ac tion). However, since most mathematical models are only approximate and since not all disturbances are measurable, it is standara prac tice to utilize feedforward control in conjunction with feedback control. Table 8-5 lists the relative advantages and disadvantages of feedforward and feedback control. By combining the two control methods, the strengths of both schemes can be utilized. 

FIG. 8-31 (a) Feedback control of a heat exchanger, (h) Feedforward control... 

The above FF controller can be implemented using analog elements or more commonly by a digital computer. Figure 8-33 compares typical responses for PID FB control, steady-state FF control (.s = 0), dynamic FF control, and combined FF/FB control. In practice, the engineer can tune K, and Tl in the field to improve the performance oTthe FF controller. The feedforward controller can also be simplified to provide steady-state feedforward control. This is done by setting. s = 0 in Gj. s). This might be appropriate if there is uncertainty in the dynamic models for Gl and Gp. 

Other Considerations in Feedforward Control The tuning of feedforward and feedback control systems can be performed independently. In analyzing the block diagram in Fig. 8-32, note that Gy is chosen to cancel out the effects of the disturbance Us) as long as there are no model errors. For the feedback loop, therefore, the effects of L. s) can also be ignored, which for the sei vo case is ... 

These decoupler design equations are very similar to the ones for feedforward control in an earlier section. In fact, decoupling can be interpreted as a type of feedforward control where the input signal is the output of a feedback controller rather than a measured load variable. 

Feedforward control can also be applied by multiplying the liquid flow measurement—after dynamic compensation—by the output of the temperature controller, the result used to set steam flow in cascade. Feedforward is capable of a reduction in integrated error as much as a hundredfold but requires the use of a steam-flow loop and dynamic compensator to approach this. 

FIG. 8 54 Controlling evaporators requires matching steam flow and evaporative load, here using feedforward control. 

Lee, M., and Park, S., A new scheme combining neural feedforward control with model predictive control. AIChE J., 38, 193 (1992). 

Fig. 5.3. Process regulation (a) feedback control, (b) feedforward control.

Apply classical controller analysis to cascade control, feedforward control, feedforward-feedback control, ratio control, and the Smith predictor for time delay compensation. 

To counter probable disturbances, we can take an even more proactive approach than cascade control, and use feedforward control. The idea is that if we can make measurements of disturbance changes, we can use this information and our knowledge of the process model to make proper adjustments in the manipulated variable before the disturbance has a chance to affect the controlled variable. 

We will continue with the gas furnace to illustrate feedforward control. For simplicity, let s make the assumption that changes in the furnace temperature (T) can be effected by changes in the fuel gas flow rate (Ffuei) and the cold process stream flow rate (Fs). Other variables such as the process stream temperature are constant. 

This equation provides us with a model-based rule as to how the manipulated variable should be adjusted when we either change the set point or face with a change in the load variable. Eq. (10-5) is the basis of what we call dynamic feedforward control because (10-4) has to be derived from a time-domain differential equation (a transient model). 3... 

In Eq. (10-5), 1/Gp is the set point tracking controller. This is what we need if we install only a feedforward controller, which in reality, we seldom do.4 Under most circumstances, the change in set point is handled by a feedback control loop, and we only need to implement the second term of (10-5). The transfer function -GL/Gp is the feedforward controller (or the disturbance rejection... 

To see how we implement a feedforward controller, we now turn to a block diagram (Fig. 

In contrast, we could have done the derivation using steady state models. In such a case, we would arrive at the design equation for a steady state feedforward controller. We ll skip this analysis. As will be shown later, we can identify this steady state part from the dynamic approach. 

---