LM-BP神经网络在震后建筑损失评估模型中的应用

Application of the LM-BP Neural Network in the EconomicLoss Assessment Model for Post-earthquake Buildings

摘要:

当前震后建筑经济损失评估模型得到的震后建筑经济损失评估精确度、效率低，针对单一神经网络易产生局部极值等问题，对神经网络方法进行了改进，提出LM-BP神经网络在震后建筑损失评估模型中的应用。输入样本要素为影响震后建筑经济损失的5项因素，输出样本是震后建筑经济损失评估结果，在此基础上采用LM-BP神经网络将训练转化成最小二乘问题，结合LM算法重新定义隐含层节点数量，构建基于LM-BP的神经网络震后经济损失评估模型，采用该模型获取最优震后建筑经济损失评估结果。仿真实验结果表明，所设计的评估模型最小评估误差为0.1%，相比同类模型具有高精确度的优势，是一种可靠的震后建筑经济损失评估模型。

Abstract:

The current model for the post-earthquake economic loss assessment of buildings has low evaluation accuracy and efficiency. In this work, the neural network method is improved, and the application of the LM-BP neural network in the post-earthquake loss assessment model of buildings is proposed to resolve the problem of the easy production of the local extremum by the single neural network. The input sample elements are five factors that affect the post-earthquake economic loss of buildings, and the output samples are the evaluation results of the post-earthquake economic loss of buildings. The LM-BP neural network is used to transform the training set into the least-squares problem, and the number of hidden layer nodes is redefined through combination with the LM algorithm. The model based on the LM-BP neural network for post-earthquake economic loss evaluation is constructed and used to obtain the optimal evaluation results for the post-earthquake economic loss of buildings. Simulation results show that the minimum evaluation error of the proposed model is 0.1%. Given its high accuracy, the proposed model is a more reliable model for the assessment of the post-earthquake economic loss of buildings than other models.