基于BP神经网络的普通建筑雷灾风险评估

Risk Assessment Method of Lightning Disaster about Ordinary Buildings Based on the BP Neural Network Process

基于《雷电防护第2部分:风险管理》(GB/T21714.2—2008)及BP人工神经网络理论,利用MATLAB软件建立了一个三级BP网络预警模型,经过网络训练、网络检测得到训练成熟的普通建筑雷灾风险BP网络评价模型。BP神经网络的运用有助于评价模型更加准确客观地反映建筑物雷灾风险各影响因素与最终风险评价结果之间的非线性关系。为了验证该评价模型的评判效果,除对训练成熟的网络评价模型进行检测,还进一步选取《雷电防护第2部分:风险管理》(GB/T21714.2—2008)中提供的评估实例进行验证性评估,二者结果保持一致,表明该方法可以用作普通建筑的雷灾风险评价。

Based on the BP neural network theory and “Protection against lightning-Part 2：Risk management” （GB/T 21714.2-2008）, a three levels BP network prediction model is established by MATLAB and a risk assessment model of lightning disaster about ordinary buildings based on the BP neural network Process which has been trained and tested is obtained. Through the extraction of BP neural network, this model can well reflect the complex nonlinear relations between lightning disaster risk influence factors and the final risk assessment results. In order to verify the effect of evaluation model, In addition to test the well trained network assessment model, the evaluation example which inside the “Protection against lightning-Part 2： Risk management” （GB/T 21714.2-2008） is selected for confirmatory assessment. The results show that the two results keep consistent, so the method can be used as the lightning disaster risk assessment of ordinary buildings.