An improved algorithm is derived for obtaining and maintaining the correct optimal steady-stateoperation of an industrial process, inspite of deficiencies in the mathematical model employed within the optimisation procedure. An augmented Lagrangian technique is employed to extend a previous method which iterates between successive solutions of parameter estimation and modified system optimisation. Optimality and convergence conditions are given, and the paper includes a detailed proof of convergence based on Zangwill's approach. The practical utility of the algorithm is demonstrated on a pilot-scale chemical vaporiser plant, where its performance is compared with previous methods.