基于似无关回归的多元稳健损失函数方法

发布时间：2018-01-28 10:43

本文关键词： 似无关回归多响应优化损失函数响应相关性稳健性　出处：《数理统计与管理》2017年05期 　论文类型：期刊论文

【摘要】：针对多响应参数优化问题,考虑响应间相关性和可控因子波动的影响,提出了一种基于似无关回归的多元稳健损失函数方法。首先采用似无关回归对模型拟合和过程优化中的相关参数进行估计,更有效地利用响应间相关性信息;然后利用给定点处梯度信息来估计可控因子波动对过程稳健性的影响。算例表明,当响应间存在相关性时,与最小二乘方法相比,采用似无关回归拟合的响应曲面模型精度更高;与传统质量损失函数相比,在采用相同质量成本矩阵时,采用稳健损失函数方法得到的最优解处期望质量损失更小。
[Abstract]:For the problem of multi-response parameter optimization, the effects of response correlation and controllable factor fluctuation are considered. A multivariate robust loss function method based on similar independent regression is proposed. Firstly, the correlation parameters in model fitting and process optimization are estimated by using similar independent regression, which makes more effective use of the correlation information between responses. Then the effects of controllable factor fluctuations on the process robustness are estimated by using the gradient information at the given point. An example shows that when there is correlation between the responses, it is compared with the least square method. The accuracy of the response surface model with similar independent regression fitting is higher. Compared with the traditional mass loss function, when the same mass cost matrix is used, the expected mass loss at the optimal solution obtained by the robust loss function method is smaller.
【作者单位】：军事交通学院装备保障系;天津大学管理与经济学部;
【基金】：国家自然科学基金重点项目(71532008)
【分类号】：O212
【正文快照】： among responses,robust0引言复杂产品质量改进或过程优化中多响应问题非常常见,产品或过程的多个响应共同受一组因子的影响,而且响应之间一般存在着相关性。稳健性优化[11是多响应优化问题的一个重要分支,通过过程参数设计,不但使质量目标得到优化,而且过程对波动、误差等因

【相似文献】