In this chapter, the idea of ‘observer models’ will be developed. The use of digital computers in industry, and even computerised measurement instruments, has led to the notions of ‘intelligent sensors’, ‘soft measurements’ and ‘data mining’. An example might be the varying temperature in the centre of cooling cast steel ingots, estimated from the surface temperature. Such calculated variables were considered briefly in Section 4.6, but the more complex issues of state and parameter estimation during dynamic variation, for multivariable systems, must be considered in more detail here owing to the growing importance of these techniques industrially.
Figure 6.1 shows the general case of control on the basis of an observer model output – a form of internal model control (IMC). Typical examples include state estimation and parameter estimation for adaptive control. However, such an estimator need not be part of a feedback control loop – other important applications include fault detection and management information such as process efficiency. An observer like this might also be used to reconcile process data, for example to provide a ‘best-fit’ mass balance as input to an optimiser, which might, or might not, operate in closed loop.