【摘要】 为了解决多因素影响下岩溶塌陷预测评价问题,采用改进型BP-神经网络方法,推导带参数的神经网络公式,分析学习因子、动量因子和陡度因子的变化对系统收敛性的影响,确定最优的参数组合,并以唐山市区为例,建立了岩溶塌陷安全性评价模型.研究结果表明:基于改进型BP-神经网络的岩溶塌陷预测与实际情况吻合很好.该成果对中国地震多发区和地下水严重开采区的岩溶塌陷预测具有借鉴作用.更多还原

【Abstract】 In order to solve the prediction and assessment problem of karst collapse under multi-factor conditions,this paper adopts the improved BP-neural network method,deduces the formula of ANN with parameters,and analyzes the impact of system convergence caused by variation of three factors of BP-neural network including learning factors,momentum factors and steepness factors,determines an optimal parameter combination,and constructs a safety assessment model of karst collapse in Tangshan City.The result of calculation shows very good consistency with actual conditions.The production holds a good reference to the karst collapse prediction of earthquake-prone areas and serious groundwater mining area in China.更多还原