结构可靠度分析的全局响应面法研究

对功能函数不能明确表达的可靠度分析问题,常采用响应面法。本文提出把响应面区分为局部响应面和全局响应面。常用的二次多项式响应面及常规的神经网络响应面均属于局部响应面,其仅在验算点附近(±σ)处与真实极限状态曲面符合较好。本文提出的全局神经网络及模糊神经网络响应面、改进全局神经网络及模糊神经网络响应面则属于全局响应面,其在全局范围均与真实极限状态曲面符合良好,故其对全局有较好预测效果。文中对以上各种响应面法计算效果及其对全局预测效果通过算例进行了对比分析,其中改进全局神经网络及模糊神经网络响应面法计算精度较好,进行有限元分析次数最少,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。

The study of the whole response surface method for structural reliability analysis

The response surface method for structural reliability analysis is often used when the performance function is unknown. The paper points out that the response surface methods should be divided into the local response surface method and the whole response surface method. The quadratic polynomial method and the artificial neural network (ANN) response surface method are in common use and belong to the local response surface methods, which only match the real limit state surface well in the area near the confirmatory calculation point. The whole and the improved whole response surface methods of ANN and fuzzy neural network (FNN) put forward in this paper belong to the whole response surface methods which match the real limit state surface well in the whole area. Through illustrations and contrast analysis, it can be known that the improved whole response surface methods of ANN and FNN are the best among all above methods, by which the amount of calculation of finite element analysis can be reduced as well as the accuracy of calculation can be increased.