Model-Based Probabilistic Robust Design With Data-Based Uncertainty Compensation for Partially Unknown System

Model uncertainty often results from incomplete system knowledge or simplification made at the design stage. In this paper, a hybrid model/data-based probabilistic design approach is proposed to design a nonlinear system to be robust under the circumstances of parameter variation and model uncertainty. First, the system is formulated under a linear structure which will serve as a nominal model of the system. All model uncertainties and nonlinearities will be placed under a sensitivity matrix with its bound estimated from process data. On this basis, a model-based robust design method is developed to minimize the influence of parameter variation in relation to performance covariance. Since this proposed design approach possesses both merits from the model-based robust design as well as from the data-based uncertainty compensation, it can effectively achieve robustness for partially unknown nonlinear systems. Finally, two practical examples demonstrate and confirm the effectiveness of the proposed method. [DOI: 10.1115/1.4005589]