Selection of Design Manners for SVM-based RPD Approach
Robust Parameter Design (RPD) is an important means for quality improvement. Compared with traditional dual response surface methodology (DRSM) which is based on polynomial models, SVM-based RPD approach is more suitable for the process when the relationship between its quality characteristics and influential factors is complex. The paper focuses on the selection among different design manners of the SVM-based approach. Using a leaf spring case study, the paper investigates the influences of four design manners (i.e., full factorial design, fractional factorial design, uniform shell design and face-centered design) on the performances of SVM-based PRD approach. By comparing these four design manners regarding their sample sizes, fitting performances, prediction performances, number of robust points and surface rebuilding abilities, the paper concludes that the full factorial design manner has the best comprehensive performance, and thereby should be the first choice for SVM-based RPD.keywords:Product Design; Quality Assurance; Robustness; Response Surface Methodology; Support Vector Machines
Cui Qingan
School of Management Engineering, Zhengzhou University, Zhengzhou, P. R. China, 450001
国际会议
郑州
英文
2008-09-20（万方平台首次上网日期，不代表论文的发表时间）