基于深度学习的盾构竖向姿态组合预测

Combined Prediction Model for Shield Vertical Posture Based on Deep Learning

为解决盾构竖向姿态的精确预测问题,提出一种基于长短期记忆(long short term memory,LSTM)神经网络-支持向量回归(support vector regression,SVR)的深度学习组合预测模型。在对采集到的竖向姿态数据进行相应的数据预处理的基础上,分别构建LSTM、SVR竖向姿态预测模型,并基于最优组合赋权的方式对二者的预测结果进行赋权,以得到LSTM-SVR盾构竖向姿态组合预测模型。为验证所构建的LSTM-SVR组合深度学习预测模型的可靠性,依托昆明地铁项目,将预测结果与LSTM、SVR、BP(back propagation)模型的预测结果进行对比。结果表明:所构建的LSTM-SVR组合深度学习预测模型具有较高的预测精度。

To accurately predict the vertical posture of shields,a deep-learning combination prediction model based on long short-term memory(LSTM)neural network and support vector regression(SVR)is proposed.The LSTM and SVR vertical-posture prediction models are developed by performing the corresponding data preprocessing operations on the collected vertical-posture data.Further,the prediction results of the two models are weighted by the optimal combination of weights to obtain a LSTM-SVR combined prediction model.Finally,to verify the reliability of the developed LSTM-SVR combined deep-learning prediction model,the prediction results are compared with those of the LSTM,SVR,and BP models based on the Kunming metro project.The results show that the LSTM-SVR combined deep-learning prediction model has high prediction accuracy.