综合时序信息的试验优化设计及稀疏求解

考虑不同试验点的试验成本及试验时序信息,在性能指标与影响因素间的响应模型未知情况下,利用装备系统的先验信息构造模型驱动的基函数,构造相应的回归模型来刻画系统的内在关系,分析不同试验点成本及试验时序对试验方案的影响。定义了一种考虑空间权重及批序贯时序的非均匀最优设计模型,将稀疏理论引入到回归试验设计中,把不同批次试验下的回归试验设计模型参数估计问题转换为结构化稀疏重构问题,讨论了非均匀最优设计的构造定理,提出结构化稀疏参数求解算法。算例表明:利用本方法生成的试验点考虑了试验点空间权重和试验时序信息,更符合实际试验规程;采用本方法获得的响应函数得到的期望预测误差更小。

Optimization of complex sequential experiment design and sparse analysis

Considering the experiment cost and experiment sequence information of different experiment points, where the response model between the performance index and influence factors is unknown, this paper utilizes a priori information of the complex system to structure model oriented basis functions and construct the corresponding regression model to depict the internal relationship of the system. The effects of experimental cost and batch sequence of different experiment points are analyzed. A non uniform optimization design model is defined taking into consideration spatial weight and batch sequence. Sparse analysis theory is introduced to transform the parameter estimation problem of the regression model from different stages into a sparse reconstruction problem. As for the non uniform optimum designs, this paper proposes a structured sparse parameter solving method. According to analysis and computational validation, it can be found that experimental points generated from the above mentioned model are more consistent with practical experimental regulations, and the expected prediction error of the response function solved by proposed method is smaller.