A general optimization framework for complex PDE models based on data interactive mechanism

For many PDE-related optimization problems, the cost function is usually computationally expensive and its derivatives are not available or may not even exist. In this paper, we establish a kind of general optimization framework, which can couple Matlab-based optimization algorithms into any PDE-related simulation program that reads/writes its input/output from text files. The key part of this optimization framework is a kind of data interactive mechanism, which passes parameters between simulation programs and optimization function. A numerical optimization of ventilation system operation is conducted according to the established framework. Performance analysis will demonstrate the effectiveness of the proposed optimization approach.