基于Kriging函数的序贯近似建模方法

多学科设计优化过程中需要多次调用高精度学科分析模型,从而造成计算复杂性问题.为解决上述计算复杂性问题,首先提出基于最优LHD的逆序贯试验设计方法,在此基础上结合模型验证,建立了基于Kriging函数的序贯近似建模方法,提出了序贯近似建模收敛准则,通过权衡训练样本数据数量和近似精度,得到精度与效率折衷的近似模型.数学算例证明了该方法的有效性.

Kriging Function-Based Sequential Approximation Modeling Method

In the process of multidisciplinary design optimization, there exits the problem of calculation complexity due to frequently calling high fideity disciplinary analysis model. For solving this problem, the sequential approximation-modeling method using Kriging function is established. This paper first puts forward anti-sequential experiment design method based on optimal Latin Hypercube Design, then develops Kriging function-based sequential approximation modeling method by taking into account model validation, thus the sequential approximation modeling convergence metric is set up, and the high-fidelity and efficient approximate model can be gained by tradeoff between sampling number and approximation precision. Mathematical example demonstrates the validity of this method.