基于神经网络的建筑物火险评价

为了快速检测建筑物当前火险等级，应用神经网络技术，建立了火险评价系统。首先构建三大类15项评价指标体系，然后请经验丰富的消防专家判定建筑物火险等级，生成60条专家样本。前50条用于神经网络的训练，后10条用于神经网络检验。通过训练，神经网络获得了较高的评价精度，训练样本的总相对误差绝对值为7258%，检验样本总相对误差绝对值为0%。实践表明，采用神经网络实现建筑物火险评价，无需推导数学模型，操作效率高，使用成本低

Fire risk assessment for buildings based on neural network

In this study, a fire risk assessment system for buildings was built based on neural network in order to quickly identify the current fire risk level Of buildings. Firstly, an assessment index system containing 3 categories, 15 items was established. Then, some experienced fire protection experts were engaged to determine the fire risk level of buildings and 60 expert samples were generated. The first 50 samples were used for neural network train- ing, while the last 10 samples were used for testing. After the training, the neural network obtained relatively high assessment accuracy. The absolute value of the total relative error of the training samples was 7. 258%, and the ab- solute value of the total relative error of the testing samples was 0%. The practice showed that it does not require deriving mathematical models by adopting neural network to achieve fire risk assessment. Meanwhile, this system a- chieves a high operating efficiency at a low cost.