The development of a maximum likelihood model for model-based applications

Since model parameter uncertainties affect the accuracy of the model's outputs, this work describes the development of a maximum likelihood model based on robust parameter estimates to improve the model's results. A robust statistical theory framework is used to determine the robust parameter estimates. Next, it is proven that a process model parameterized by robust parameter estimates within their feasible ranges is a maximum likelihood model. A chemical reactor process is presented to demonstrate the development of the maximum likelihood model and its performance properties in a model-based predictive control framework.

► We applied robust statistical theory to determine robust parameter estimates. ► We provide the theoretical foundation to prove the existence of a maximum likelihood model (ML). ► We demonstrate the development of an ML model using a chemical reactor process. ► We further demonstrate the robust properties of the ML model in a model predictive control framework.