基于似无关模型的改进多元损失函数方法

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

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。

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.