基于集成Dropout-DNN模型的盾构掘进速度预测方法

Prediction Method of Shield Tunneling Speed Based on Integrated Dropout-DNN Model

为了提升土压平衡盾构机的掘进速度预测精度,提出一种集成Dropout技术和深度神经网络(deep neural network,DNN)模型的盾构掘进速度预测方法。依据济南地铁R1线盾构隧道段工况数据,将数据集划分为5份,并选取刀盘转速、刀盘扭矩、总推进力、螺机转速、土仓压力这5个参数为输入参数,分别建立了5个Dropout-DNN模型并进行集成实现了盾构掘进速度的预测,进一步对不同的预测方法进行了对比分析。研究结果表明:各Dropout-DNN模型预测精度具有一定的差异性,但基本良好,其决定系数均大于0.6,平均绝对百分误差均小于10%,而集成的Dropout-DNN模型决定系数为0.695,平均绝对百分误差小于5%,可见集成模型预测精度较高;基于反向传播(back propagation,BP)神经网络、DNN模型实现的盾构掘进速度预测模型其决定系数分别为0.502、0.566,可见提出的集成Dropout-DNN模型预测精度提升明显。

To improve the prediction accuracy of tunneling speed of earth pressure balance(EPB)shield machine,a shield tunneling speed prediction method by integrating dropout and deep neural network(DNN)model was proposed.The data set was divided into five parts according to the working condition data of shield tunnel section of Jinan metro line R1.Five parameters of cutter head speed,cutter head torque,total propulsion force,screw machine speed and chamber pressure were selected as input parameters.Five Dropout-DNN models were established and integrated to predict the shield tunneling speed.The different prediction methods were further compared and analyzed.The research results show that the prediction accuracy of each Dropout-DNN model has some differences,but it is basically good with coefficient of determination more than 0.6 and average absolute percentage error less than 10%.The coefficient of determination of integrated Dropout-DNN model is 0.695 and the average absolute percentage error is less than 5%.It can be seen that the prediction accuracy of the integrated model is very high.The coefficient of determination of the shield tunneling speed prediction model based on BP neural network and DNN model are 0.502 and 0.566 respectively.It can be seen that the prediction accuracy of the proposed integrated Dropout-DNN model is improved significantly.