基于遗传算法和神经网络的梁板结构可靠性优化

基于结构系统静强度可靠性分析、神经网络和遗传算法，对空间梁板结构系统进行了可靠性分析和基于可靠性的优化设计。结构可靠性分析中，给出了安全余量以及安全余量对各变量敏度的显性表达式，便于各安全余量间相关性计算和可靠性计算精度提高。结构优化中，用神经网络和遗传算法，每代遗传操作中只需用传统方法计算1次结构系统可靠性指标，将该代最优解对应的数据加入神经网络的训练样本，从训练样本中删除最次样本，使训练样本不断处于更新状态。数值算例表明：该法收敛平稳、用时较少，具较好的收敛性和较高的计算效率。

Space Beam and Plate Structural System Optimization Based on Reliability Using Neural Network and Genetic Algorithm

Based on the structural static strength reliability analysis, the neural network and genetic algorithm, the space beam and plate structural system reliability and optimization design were analyzed in this paper. The expressions of the safe margin and sensitivity of safe margin to variable were given in the analysis of the structure reliability, which was easy to calculate the relativity of safe margin and improve the calculation precision of reliability. The neural network and genetic algorithm at the structural optimization program were used in the optimization, and only one time calculation of the structural system reliability index by conventional method at every genetic operation was needed, and the best optimal solution was added to the training samples and the worst sample to update the training samples of neural network was deleted. The numerical example showed that the method had strong optimization capability and high optimization efficiency.