Bias feedforward

Substituting these into Equation (6.17) shows that K should be set to 1. This is always the case for ratio feedforward. For bias feedforward... 

The tuning method for the decoupler is exactly that described for bias feedforward in Chapter 6, i.e. 

We apply dynamic compensation in the form of a deadtime/lead-lag algorithm. This is tuned in exactly the same way as described in Chapter 6 covering bias feedforward. By performing open loop steps on the MV we obtain the dynamics of both the inferential and... 

Dual fuel firing was cited in Chapter 6 as an example of bias feedforward control. Here we will expand on this technique. The general control problem is illustrated in Figure 10.8. 

In this example fuel A is a gas over which we have no control. Fuel B is a liquid and its flow may be manipulated to control the heater outlet temperature. The scheme includes a bias feedforward scheme so that changes in fuel A are immediately compensated for by adjusting the flow of fuel B. For this to succeed we have to convert the units of measure of fuel A to be consistent with those of fuel B. By including the heating value of fuel B (NHVi,) in Equation (10.5) we get... 

From Equation (6.19) we know that the gain in the bias feedforward is given by... 

As an alternative other algorithms, such as bias feedforward and deadtime compensation, can be implemented in the MVC - depending on which approach is better for operator understanding and what back-up scheme is necessary if the MVC is out of service. It is also possible to move averaging level control from the DCS to the MVC. This should only be considered if it is desirable to let the MVC select which flow to manipulate (i.e. vessel inlet or outlet) depending on where the process is constrained. The DCS controller will still be required as back-up. 

In many eases the feedforward eontrol is usually eombined with a feedbaek system to eliminate any offset resulting from inaeeurate measurements and ealeulations. The feedbaek eontroller ean either bias or multiply the feedforward ealeulation. 

Feedforward control is usually combined with feedback control to eliminate any offset resulting from inaccurate measurements and calculations and unmeasured load components. The feedback controller can be used as a bias on the feedforward controller or in a multiplicative form. 

The basic feedforward neural network performs a non-linear transformation of the input data in order to approximate the output data. This net is composed of many simple, locally interacting, computational elements (nodes/neurons), where each node works as a simple processor. The schematic diagram of a single neuron is shown in Fig 1. The input to each i-th neuron consists of a A-dimensional vector X and a single bias (threshold) bj. Each of the input signals Xj is weighted by the appropiate weight Wij, where] = 1- N. 

As with ratio feedforward, some DCS include the bias as an option within the PID algorithm. Some also permit configuration of the operator display to make the algorithm appear like a true controller with both the actual bias and the bias SP. 

Another difficulty may be in the determination of the feedforward gain (K) that should be used. Unlike feed rate feedforward, feed composition feedforward requires a bias not a ratio algorithm and so K is not 1 (see Chapter 6 for explanation). Figure 12.114 shows how each of the possible MVs should be adjusted as feed composition changes. The shape of these lines was explained earlier in the chapter (see Figure 12.68). K is the gradient of the line. 

Alternative ways of incorporating feedback trim into a feedforward control system include having the feedback controller output signal adjust either the feedforward controller gain or an additive bias term. The gain... 

---