基于灰关联和神经网络的大型商场火灾风险评价方法

A renovated method of fire risk evaluation of large emporium based on ANN and greycorrelation

根据大型商场火灾的特点,建立了大型商场火灾风险评价的指标体系。通过比较几种安全评价模型,提出了基于遗传算法的神经网络评价模型,并利用灰关联度筛选评价指标,改善BP神经网络的泛化能力。在灰关联的计算中使用Spearman等级相关系数组合主、客观权重,解决单一赋权的缺陷。通过实例验证了该模型具有较好的应用价值。

Large emporiums are usually faced with higher fire risk due to their flow of large crowds of people, large space for commodity and facilities and high combustible materials. Based on the study of the said characteristic features of large emporiums, this paper is aimed to present a renovated method of fire risk evaluation of large emporium based on the grey correlation. The renovated method we have proposed is established on the theoretical basis of grey correla- tion, genetic algorithm and BP neural network for evaluating the safe- ty conditions in large emporiums. More specifically, such a multiplex system involves a lot of factors that are likely to affect the construction safety, which makes it necessary to use the grey correlation to select assessment indexes. In this paper we have adopted the Spearman cor- relation rating coefficient to combine the subjective and objective weights to exclude the deficiency of the single weight case. The com- bination weight makes the comprehensive safety assessment more sci- entific and reasonable. And the indexes that are selected by grey cor- relation can be chosen as input parameters of neural network. At the same time, we have established a safety evaluation model of large emporiums via the BP neural network based on genetic algorithm. The proposed model intends to avoid being too complicated for getting into a local value and a too-long training period. Such a set has to be trained and evaluated by using this model so that an integrated sys- tematic evaluation of the safety conditions of large emporiums can be confirmed via grey correlation and BP neural network. Last of all, we have chosen a Xi ＇ an building as our case study for the evaluation. The application of grey and ANN proves that the renovated method is a valuable means for its purpose to a great extent.