基于CNN的混凝土坝变形预测深度学习模型研究

A deep learning model for concrete dam deformation prediction based on CNN

混凝土坝变形监测与预测分析是一个长期需关注的问题。考虑到混凝土坝变形预测模型是预测结构性态演化、评价安全服役状况的关键措施,充分利用谷歌开源深度学习框架TensorFlow平台,结合深度学习理论中成熟的卷积神经网络技术建立了基于深度学习的混凝土坝变形安全预测模型,并以残差图、均方误差和平均百分比误差等指标评价作为模型的拟合、预测精度评价标准。通过仿真实例表明,相比于浅层神经网络模型和传统的统计模型,基于深度学习的混凝土坝变形预测模型预测精度更高,性能更加稳定,为混凝土坝变形监测提供了一种新方法。研究所得成果为混凝土坝变形预测提供参考依据。

Deformation monitoring and prediction analysis of concrete dam is a long-term concern.Considering that the concrete dam deformation prediction model is the key measure to predict the structural evolution and evaluate the safety service condition,this paper makes full use of Google's open source deep learning framework TensorFlow platform and combines the mature convolution neural network technology in deep learning theory to establish a concrete dam deformation safety prediction model based on deep learning,and takes the residual diagram,mean square error and average percentage error as the evaluation criteria of model fitting and prediction accuracy.The simulation example shows that compared with the shallow neural network model and the traditional statistical model,the deformation prediction model of concrete dam based on deep learning has higher prediction accuracy and more stable performance,which provides a new method for deformation monitoring of concrete dam.The research results provide reference for deformation prediction of concrete dams.