A Two-loop Optimization Strategy for Multi-objective Optimal Experimental Design

Abstract

A new strategy of optimal experimental design (OED) is proposed for a kinetically controlled synthesis system by considering both the observation design and the input design. The observation design that combines sampling scheduling and measurement set selection is treated as a single optimization problem arranged in the inner loop, while the optimization of the input intensities is conducted in the outer loop. This multi-objective dynamic optimization problem is solved via the integration of the particle swarm algorithm (for the outer loop) and the interior-point method (for the inner loop). Numerical studies demonstrate the efficiency of this optimization strategy and show the effectiveness of this integrated OED in reducing parameter estimation uncertainties. In addition, process optimization of the case study enzyme reaction system is investigated with the aim to obtain the maximum production rate by taking into account of the experimental cost.