摘要
针对结构抗震可靠度分析中存在影响因素复杂、统计资料缺乏、概率模型合理选取困难、计算繁杂等问题,利用遗传算法的全局空间寻优功能和BP网络映射能力强的优点,提出了预测结构抗震可靠度的进化神经网络模型,并利用BP网络进行对比研究.实例分析表明,利用进化神经网络预测结构抗震可靠度,可克服BP网络输入层节点数、隐含层数和隐含层节点数确定的盲目性,适应性更强,精度更高.
There are lots of problems in analyzing the seismic reliability, such as the complex affecting factors, lack of statistic data, difficulty to select the proper probability model and complex computation. Taking advantage of the global spatial optimization function and strong capability of network mapping in BP, this paper put forward an evolutionary neural network model to predict the reliability of seismic structures, and conducted a comparison analysis with BP network.The results show that the model of the GA-ANN not only can avoid the blindness in choosing nodes of input layer, the numbers of concealed layer and nodes of concealed layers in BP, but is better in adaptability and accuracy.
出处
《南昌工程学院学报》
CAS
2007年第6期22-25,40,共5页
Journal of Nanchang Institute of Technology
关键词
抗震设计
可靠度
遗传算法
神经网络
seismic design
reliability
genetic algorithm
neural network
