Feasibility analysis of black-box processes using an adaptive sampling kriging based method

Accurate knowledge of the effect of parameter uncertainty on process performance is vital for optimal and feasible operation. The objective of this work is to develop a systematic methodology for performing feasibility analysis over a multivariate factor space when the explicit form of a process model is lacking or when its evaluation is expensive. For this purpose a Kriging based surrogate approximation of the process model based on experimental or simulated data is used. In this work, two issues are addressed: feasibility evaluation of black-box processes using Kriging and introduction of an adaptive sampling methodology in order to minimize sampling cost, while maintaining feasibility space accuracy. The adaptive sampling strategy identifies critical regions and directs the search towards regions where feasibility boundaries exist or where the Kriging prediction uncertainty is high. The average error of Kriging prediction as well as cross-validation methods are used to validate the robustness of the produced model of the initial experimental design which is found to highly affect the final prediction.