高层建筑火灾风险的神经网络评价

High rising building fire risk assessment based on the artificial neural network

通过对模糊综合评价和人工神经网络评价进行分析比较,表明神经网络采用非线性转换函数的处理方法更符合火灾风险评价的非线性特性。针对神经网络易陷入局部极小而引起评价指标权值分布不合理的缺陷,提出运用遗传算法克服神经网络的这种缺陷,在此基础上建立了基于遗传算法和神经网络的建筑火灾风险评价模型。研究实例证明了此模型的可行性。图1,表3,参11。

High building fire is one of the most important aspect in city safety. Fire prection based on pre-formance is becoming a primary way in building fire prection design. The difference between new design method and the traditional is performing fire risk assessment to decide the fire prelection equipment, so that design can meet the needs of building safety by the lowest economic cost. Fuzzy Comprehensive Assessment is a method widely used in safety assessmnet, but it can not non-linear issues very well. Artificial Neural Network is compared with Fuzzy Comprehensive Assessment approach in this paper. The result of comparison shows the former is more suitable. A new risk assessment model for building fire based on the Artificial Neural Network and Genetic Algorithm is established after analyzing the limitation of the Artificial Neural Network such as its searching and optimizing method of weight. An example is given and the results shows this model is reliable. 1fig. ,3tabs. ,11refs.