Model certainty in optimising control

An investigation is made into the use of certain and uncertain models in optimising control when employing techniques for integrated system optimisation and parameter estimation. It is seen how a standard two-step technique, which possesses mathematical deficiencies, can still be usefully employed to give time saving benefits in the optimisation procedure when a model is used which possesses certainty. The technique is also seen to be able to be used in the initial phase of the optimisation procedure before switching to another algorithm, such as the modified two-step method, to home in to the process optimum. The point at which this switch is made is the point at which the model becomes uncertain as far as the optimisation procedure is concerned. These situations are illustrated by conducting the on-line optimisation of a mixing process which is controlled from within a distributed control system