Takagi-Sugeno Model Based Simulation Data Mining for Efficient Product Design

Simulation-driven design has become the mainstream design scheme. Generally, the designer have to design from scratch and repeat the design and analysis cycle until the design satisfies its expected objectives whenever faced new design requirements as the product development is an iterative process. Therefore, it is not an easy task for the designer to efficiently reuse the huge simulation data to improve the design process since what can be reused is the knowledge rather than the simulation data themselves. In this study, Takagi-Sugeno (TS) model based simulation data mining approach is proposed to uncover the intrinsic relationship between the key design parameters and global performance parameters in order to guide the design optimization with the aim to improve product design process. Firstly, an intermediate model which supports global prediction of the simulation results is proposed and its corresponding simulation results are computed based on the mapping and interpolation algorithm. Then the TS model is trained based on the identification algorithm and is further optimized to improve the prediction accuracy. Finally, the approach of TS model based simulation data mining is applied to the actual simulation data sets to demonstrate its feasibility and effectiveness. The experimental results show that the proposed approach can achieve great prediction accuracy on the global performance evaluation as well as local performance evaluation.