Extracting Design Knowledge From Optimization Data: Application to Multi-Split Thermal Management System Configuration

Abstract As engineering systems become more complex, the traditional sources of design knowledge face limits in the rate and complexity of design knowledge generation. Additional sources of knowledge are needed to guide future design efforts, especially for unprecedented engineering systems that have no design heritage. One promising avenue lies in the analysis of design optimization data, which has the potential to offer valuable insights into unprecedented system design, and to break free from incremental improvements over heritage designs. This paper presents a step toward extracting knowledge from design optimization data. The resulting design knowledge can be used as a basis for successfully synthesizing engineering system configurations that are more complex than those the knowledge was derived from. These methods are studied here using the combined system topology and optimal control design of multi-split fluid-based thermal management systems. This knowledge generation approach offers several advantages over traditional strategies, including applicability in cases where there is no design heritage and the ability to provide normative guidance as opposed to descriptive (i.e., how should systems be designed vs. how have they been designed). Four significant case studies with varying levels of complexity are presented that demonstrate the effectiveness of using knowledge extracted from design optimization data in enhancing the design of complex thermal management systems. Our results show that the knowledge extracted in this way provides a good basis for more general design of complex thermal management systems. It is shown that the objective function value of the estimated optimal configuration closely approximates the true optimal configuration with less than 1 percent error, achieved using basic features based on the system heat loads without involving the corresponding optimal open loop control (OLOC) features. This eliminates the need to solve the OLOC problem, leading to reduced computation costs.