Model for process control - for prediction of future values of inaccessible control magnitudes in e.g. distillation

摘要

Model comprises in series a generator of a function (G), receiving the output vector (X) of the model and/or of the process, with an integrator. A parameter vector (P) defines the parameters of the model and if required, an input vector (U) for the process and the model. The model produces a vector G (X, U, P) equal to the derivative with respect to time of the vector, (X) while the output of the integrator is connected to the output of the model. The parameter vector (P) is generated in a computer block comprising at least (a) a multiplier which multiplies a generalised error vector (E) equal to the difference between the output vectors of the model and the process, by the matrix of the partial derivs. of the components of the vector (G) with respect to components of the parameter vector (P). It further comprises (b) an integral chain following the multiplier and formed by putting in series, a matrix of adaptation and an integrator the output of which is connected to the output from the computer block which gives (P). The model can be adapted at every instant so that it reflects the actual characteristics of the process. The unmeasured fundamental magnitudes can therefore be evaluated very much more precisely and an estimate-free from background noise - of the variation of these magnitudes can be integrated t provide their predicted future values.