Product Development Via Computer Experiment and Statistical Optimization

• Main Authors: Fran-Cois Liang, 梁志鴻
• Other Authors: Angus Jeang
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/fp6wwa

碩士 === 逢甲大學 === 工業工程學所 === 91 === This paper presents a systematic approach to establish a methodology and procedure for product development that associates with the domain knowledge to disclose relations between the design variables space and quality / specification characteristic space including the effect of noise for the sake of understanding of product and making decision in early design stage. This methodology and procedure includes the Design of experiment method which design matrix is built of Box-Behnken Design for three level design variables’ requirements, Monte Carlo simulation which to simulate the variations’ effects induced by services condition, material and manufacturing process (tolerances) in each experimental treatment, finite element method which to execute the experiments by computer and then using the probability theory to transfer the quality characteristic to probability of specification dependability and reliability ,and then to establish the response surfaces via stepwise regression in order to find the optimum design variables set by mathematical programming based on the statistical consideration. The approach disclosed in this paper was successfully to solve the complicated bike frame product development, which contains 16 design variables with 3 levels and 2 noise factors. The objective is to minimum the expected bike frame weight which is subject to the reliability, cost and dependability.