摘要
针对结构可靠度计算,以BP神经网络代替一般响应面的多项式函数、结合MCS蒙特卡洛算法组合的模型较为常见。该方法充分发挥了BP神经网络与MCS的优点,利用神经网络泛化能力解决了隐式功能函数的结构失效概率的求解难题。但目前大多研究忽视了BP-MCS组合模型的样本容量选取问题,对一个已有显式功能函数的岩质边坡数值算例进行研究,发现BP神经网络训练样本与MCS算法抽样的选取对BP-MCS模型计算精度有重要的影响,分析结果可为实际工程结构可靠度计算提供建议。
Numerous studies focus on the calculation of structure reliability.It is common to use the BP neural network instead of the polynomial function of the general response surface combined with the Monte Carlo algorithm,to form a combined model.Advantages of BP neural network and MCS are highly brought into this model for using the generalization ability of neural network solving the problem of answering the structure failure probability under implicit function.However,the sample size selection of the BP-MCS combined model has been ignored in most researches currently.In this paper,it is found that the sample selection of the BP neural network training and the MCS algorithm has a significant impact in calculation accuracy of the BP-MCS model through a rock slope numerical example with an explicit function.The analysis can provide some directional guidance for the reliability calculation of actual engineering structures.
作者
章浩龙
刘洪君
Zhang Haolong;Liu Hongjun(Construction Institute of Guangdong Technology College,Zhaoqing 526100,China)
出处
《黑龙江科学》
2023年第4期22-25,33,共5页
Heilongjiang Science
关键词
结构可靠度
失效概率
样本选取
神经网络
蒙特卡洛算法
Structure reliability
Failure probability
Sample selection
Neural network
Monte carlo algorithm
