空间反射镜的支撑结构决定了反射镜的性能。通过有限元分析来优化支撑结构参数需要很大计算量,而且无法找到最优解。为了缩短设计周期,提出用人工神经网络来模拟反射镜组件输出特性和支撑结构参数间的非线性关系,并用MATLAB编制程序实现神经网络的建立和泛化,最终找到使反射镜输出特性最优的支撑结构参数。该方法不但能找到多变量优化的最优解,而且计算表明用神经网络泛化得到结果和有限元分析计算得到的结果相差在5%以内,精确度足以满足工程应用要求。

The quality of space mirror depends on its supporting structure. To optimize the supporting structure parameters with finite element analysis requires huge amount of calculation, moreover, the optimal solution could not be found. In order to shorten the design cycle, artificial neural network was proposed for imitating the nonlinearity between the space mirror characteristics and its supporting structure parameters. MATLAB program was adopted to realize the foundation and generalization of artificial neural network and the optimal supporting structure parameters for the optimal space mirror characteristics was found in the end. In addition to the finding of the optimal solution for multi-variable optimal, the results of the method also suggested that they were very close to the ones obtained with finite element analysis, with the difference less than 5%. Such an accuracy was enough for engineering.