Analysis Of The Approach To Multivariate Robust Design And Its Application

Robust design is an efficient methodology for process improvement. Most traditional robust design models consider only a single quality characteristic, yet customers' requirement is multidimensional, hence how to optimize multivariate robust design problems becomes one of the issues to be addressed in the area of quality engineering.In this thesis, the research status of multivariate robust design at home and abroad is reviewed, some drawbacks of multivariate robust design models in existence are summarized on the basis of the systemic retrospection. Then the basic principles of robust design are introduced. According to the exist methods and theories, a parametric programming approach to multivariate robust design is suggested. First of all, different ways for standardization are showed for LTB, STB and NTB characteristics; second, the covariance-variance matrix is used to measure the overall variability of the multiple quality characteristic; third, parametric programming is applied as the frame of optimization, in order to balance the variation and bias flexibly; finally, a data-based geometric method to determine the parameter has been suggested for the circumstances that the parameter is uncertain.Multivariate robust design for a process of injection of the shell of cell-phone has been made by the method established above to prove its feasibility. The result suggested that the solution not only makes the overall variation minimized, but also keeps the quality characteristics within the specifications. At the end, sensitivity analysis is conducted on the solution.