Approximate Model Identification via Set-Theoretic Estimation

The performance achievable with a model-based process control strategy increases with the fidelity of the model used, but so do the cost and complexity of these models. This trade-off has led Prater to propose the "principle of optimum sloppiness"--make the model only as detailed as necessary to minimize cost and complexity. But model accuracy is difficult to quantify a priori, especially for simplified models of complex processes. This paper considers the "parameter uncertainty intervals" described by Milanese and Belforte as a method of determining model uncertainty and develops a simple recursive algorithm for evaluating these parameter bounds. A numerical example is included to illustrate the utility of the method, and its application to the problem of approximate process modeling is discussed.