Customised optimisation for increased computational efficiency

The control of a large process plant usually requires the generation of a `set-up´, a set of set points which are applied to the plant. Set-up algorithms fall into two types: empirical set-up algorithms; and generally optimised set-up algorithms. Empirical set-up algorithms are usually designed to perform quickly, and are common in on-line systems. They are usually sub-optimal in the set-ups they produce to a greater or lesser extent, but by judicious tuning and application specific experience can be made to perform reasonably well. In the paper the authors consider constrained optimisation, regression analysis with linear constraints, and non-iterative constrained optimisation