结构可靠度分析的改进BP神经网络响应面法

Improved BP ANN Response Surface Method for Structural Reliability Analysis

对功能函数不能明确表达即具有隐式功能函数的问题进行可靠度分析,常采用响应面法。其中二次多项式响应面法应用较为广泛,采用与此方法相同的思路,人们提出了BP神经网络响应面法。在此基础之上,笔者提出了改进的BP神经网络响应面法。对以上方法计算效果,通过算例进行了对比分析,其中改进BP神经网络响应面法计算精度较好,进行有限元分析次数较少。该方法用于大型复杂结构的可靠度分析,可相应提高工作效率和解题质量,具有一定实际应用价值。

The response surface method for structural reliability analysis is usually adopted as the performance function is unknown. An improved method of BP Artificial Neural Network (ANN) is developed. The practical examples with compa rative analysis show that the improved method is suitable one for complex large scale structure reliability analysis with least calculation of finite element analysis and a higher efficiency.