工业工程

工业工程

• 专题论述 • 上一篇    下一篇

基于似无关模型的改进多元损失函数方法

  

  1. 1. 军事交通学院 装备保障系，天津 300161; 2. 天津大学 管理与经济学部，天津 300072
  • 出版日期:2016-12-31 发布日期:2017-02-22
  • 作者简介:张旭涛（1981-），男，河南省人， 讲师， 博士，主要研究方向为质量管理、试验设计等.
  • 基金资助:

    国家自然科学基金委杰出青年基金资助项目（71225006）

An Improved Loss Function Approach for Multi-response Optimization Using Seemingly Unrelated Regressions

  1. 1. Department of Equipment Support, Military Transportation University, Tianjin 300161, China; 2. College of Management and Economics, Tianjin University, Tianjin 300072, China
  • Online:2016-12-31 Published:2017-02-22

摘要：

对复杂产品或过程中存在相关性的多响应优化问题进行了研究，通过引入似无关回归模型对质量损失函数方法进行改进，将响应间的相关性综合考虑到模型拟合和参数优化两个阶段。算例表明，与传统质量损失函数方法相比，改进方法得到的最优解处期望质量损失更小。该方法克服了传统方法忽视响应间的相关性或只在优化阶段考虑相关性的不足，是对多元损失函数的重要改进。

关键词: 似无关回归, 多响应优化, 损失函数, 响应曲面方法, 响应相关性

Abstract:

Multi-response optimization with correlated responses in complex product or process is studied. The seemingly unrelated regression method is employed to improve traditional loss function approach. Correlations among responses are considered both in the model fitting phase and in the optimization phase. Illustrated example shows that the proposed approach can lead to an optimal solution with smaller expected quality loss. The approach overcomes the drawback of traditional method which ignores the correlation or only considers the correlation in the optimization phase. It is an important improvement on the conventional loss function method.

Key words: seemingly unrelated regressions, multiresponse optimization, loss function, response surface methodology, correlations among responses