| 结合半参数方法和贝叶斯抽样技术的多响应优化设计 |
| Multi-response optimization design based on semi-parametric methods and Bayesian sampling techniques |
| 摘要点击 265 全文点击 0 投稿时间：2020-07-06 修订日期：2021-04-07 |
| 查看/发表评论 下载PDF阅读器 |
| 中文关键词 多响应优化设计；半参数方法；质量损失函数；响应曲面方法；贝叶斯方法 |
| 英文关键词 multi-response optimization; semi-parametric methods; quality loss function; response surface methods; Bayesian methods |
| 基金项目 国家自然科学基金项目（面上项目，重点项目，重大项目） |
|
| 中文摘要 |
| 在多响应优化设计中，响应曲面模型的构建对试验结果影响至关重要。传统的响应曲面模型（如双响应曲面方法、广义线性模型等）往往会事先对响应曲面的模型结构（如线性多项式或非线性曲面等）做出一系列的假设。然而，在面向复杂产品的质量设计时往往会出现模型结构错误设定的情况。针对上述情况，本文结合半参数方法和贝叶斯抽样技术提出了一种新的多响应优化设计方法。本文所提方法不仅能够构建更为精确的响应曲面模型，而且还能够充分利用贝叶斯抽样方法评估优化结果的可靠性。最后，案例研究表明：本文所提新方法能够有效地解决模型结构不确定以及小样本数据对研究结果的影响，获得更加稳健可靠的优化结果。 |
| 英文摘要 |
| In a multi-response optimization design, the construction of a response surface model is critical to the experimental results. Traditional response surface models (e.g., dual response surface methods, generalized linear models, etc.) tend to make a series of prior assumptions about the model structure of the response surface (e.g., linear polynomial or nonlinear surfaces, etc.). However, there are frequent cases where the model structure is incorrectly assumed when designing for the quality of complex products. This paper proposes a new multi-response optimization design method combining semi-parametric methods and Bayesian sampling techniques to address the above issues. The proposed method not only enables the construction of more accurate response surface models but also makes full use of Bayesian sampling methods to assess the reliability of optimization results. Finally, the case study shows that the proposed method can effectively address the influence of the model structure uncertainty and small sample data on the research results, so as to obtain more robust and reliable optimization results. |
| 关闭 |
