基于机器学习技术的盾构荷载预测研究

Study on Shield Load Prediction Based on Machine Learning

盾构荷载直接反映了盾构施工对周边环境的扰动情况,准确的盾构荷载预测可以保证周边环境的稳定和工程施工安全。鉴于传统的预测方法的精度较差的局限性,本文以北京某盾构工程为研究背景,提出了一种结合最大信息系数(MIC)、卷积神经网络(CNN)和时间卷积神经网络(TCN)的新型盾构荷载预测模型(MCT)。首先用MIC选取合适的输入参数,然后通过CNN模型挖掘荷载数据的空间特征,接着通过TCN模型提取荷载数据的时序特征,最后得出预测的结果。以实际工程的监测数据作为数据集,通过与现有的4种算法进行对比实验,结果表明本文提出的模型具有更好的预测能力,可以为以后类似的工程施工提供指导。

Shield load directly reflects the disturbance of shield construction to the surrounding environment,and accurate shield load prediction can ensure the stability of the surrounding environment and the safety of engineering construction.In view of the limitations of traditional prediction methods with poor accuracy,this paper proposes a novel method that combines Maximal Information Coefficient(MIC),Convolutional Neural Networks(CNN)and Temporal Convolutional Network(TCN)to predict the shield load.Firstly,MIC is used to select suitable input parameters,then the CNN model is used to mine the spatial features of the load data,followed by the TCN model to extract the temporal features of the load data.Finally,the prediction results are obtained.The monitoring data of the actual project used as the dataset,the experiment results show that the proposed model in this paper has better prediction performance by comparing with the existing four algorithms.The proposed model can provide guidance for the construction of similar projects in the future.