考虑温变效应的热物性参数RBF辨识方法

在稳态温度环境下,针对温变系统的参数识别问题进行了研究,提出了两种基于代理模型的热物性参数辨识方法:多目标全域法和分段等效法。前者在整个分析域内构建多种温度分布形式,将参数识别问题转化为多目标优化问题,引入代理模型与快速非支配排序遗传算法(NSGA-II)进行求解,在提高识别效率的基础上有效地保证了该方法的鲁棒性。后者通过将材料热物性参数按温度区间分段,建立了温度均方根残差和在区间段内代理模型,识别出每个子区间内的等效物性参数,并采用回归分析获得热物性参数随温度变化规律。通过典型算例对两种方法进行了验证,研究结果表明:两种识别方法都具有较好的识别精度,相对而言,全域识别方法具有更优的抗噪性,而分段识别则更便于实际操作。

IDENTIFICATION OF TEMPERATURE-DEPENDING THERMOPHYSICAL PARAMETERSBASED ON RBF METHOD

The problem of a temperature-dependent-parameters identification has been researched in a steady-state temperature environment, and two methods, the whole domain method of a multi-objective and the equivalent piecewise method based on metamodeling, are developed for parameters-identification problems. The former method is designed to construct different temperature-distribution forms for an optimization objective. Then, the improved method of non dominated sorting genetic algorithm (NSGA-II), as a multi-objective optimization method, is employed to estimate the thermophysical parameters based on metamodels. This methodology could improve the search efficiency and circumvent difficulties of an ill-posed problem. In the latter method, the metamodeling of residuals between calculated and experimental results was constructed to identify equivalent parameters of each temperature section, and then regression analysis was used to identify the law of parameters varying with temperature. Finally, examples were given to demonstrate the effectiveness of these two methods. The results show that the whole domain method performs better in robusticity, while the equivalent piecewise method has a better maneuverability.