【摘要】 选取房屋破倒率、震时地面加速度峰值、地震区域抗震设防烈度和地震发生时刻作为引发地震次生火灾的4个预测因子,结合中美日76条地震次生火灾统计数据,运用BP神经网络建立城市地震次生火灾起火率预测模型。分别运用BP神经网络模型和国内二项式拟合模型对实际样本起火率进行预测。结果表明:BP神经网络模型的预测结果与样本的实际起火率基本吻合;与国内二项式拟合模型相比,BP模型在地震次生火灾起火率的预测精度上有了较大提高,证实了BP神经网络适用于地震次生火灾起火率预测。以武汉市汉口地区为例,对该地区地震次生火灾起火率进行了预测,为城市抗震救灾、消防设施布置和防灾减灾规划建立提供了依据。更多还原

【Abstract】 Considering four predictors including house collapsed rate,peak ground acceleration,seismic fortification intensity and the time of earthquake,the BP neural network was used to establish the prediction model of frequency of urban post-earthquake fire with 76post-earthquake fire statistics from China,the United States and Japan.The BP neural network model and domestic binomial fitting model were employed to predict the frequency of fire in actual sample respectively.The experimental results indicated that the prediction of the BP neural network model coincided well with the actual frequency of fire in the sample.The prediction accuracy of BP neural network model was improved significantly compared with the domestic binomial fitting model,which meant BP neural network model was suitable for the prediction for frequency of post-earthquake fire.And taking Hankou area in Wuhan as a prediction object,the prediction for frequency of post-earthquake fire in this area was made,which aimed to provide the basis for earthquake relief,arrangement of firefighting devices and establishment of program for disaster prevention and reduction in city.更多还原