论文部分内容阅读
针对目前的地铁隧道沉降变形预测方法忽略了对沉降变形影响因素的综合协调考虑这一问题,该文将遗传算法(GA)结合极限学习机(ELM)的方法引入地铁隧道沉降变形预测。该方法借助最大信息熵理论,充分挖掘地铁隧道沉降主要影响因素与沉降量间的信息特征,并将遗传算法与极限学习机相耦合,利用遗传算法的全局搜索能力获取ELM神经网络优化的初始权值和阈值,形成熵权遗传算法-极限学习机模型,并编制相应计算程序。采用该模型对西安某地铁隧道沉降变形进行预测,并与遗传算法-极限学习机、极限学习机、传统的BP神经网络预测结果进行比较,结果表明熵权遗传算法-极限学习机模型与实测值吻合更好,预测结果更稳定。
In view of the fact that the subsidence deformation forecasting method for metro tunnels ignores the comprehensive and coordinated consideration of the factors influencing subsidence deformation, this paper introduces genetic algorithm (GA) combined with extreme learning machine (ELM) method to predict subsidence deformation of Subway Tunnels. This method exploits the theory of maximum information entropy to fully exploit the information characteristics between subsidence factors and subsidence of subway tunnels. Coupled with the limit learning machine, genetic algorithm is used to obtain the initial weights of ELM neural network optimization Value and threshold, the formation of genetic algorithm entropy weight - limit learning machine model, and the preparation of the corresponding calculation program. This model is used to predict the subsidence and deformation of a subway tunnel in Xi’an. The results are compared with the results of genetic algorithm - extreme learning machine, extreme learning machine and traditional BP neural network. The results show that the entropy weight genetic algorithm - extreme learning machine model and measured value Good agreement, the forecast results more stable.