Dynamic early-warning model of dam deformation based on deep learning and fusion of spatiotemporal features

作者：

Highlights：

• Dynamic early-warning model under spatiotemporal feature deep learning is proposed.

• DKELM is improved by correntropy to restrain the interference of input outliers.

• Combining POD and the improved DKELM to establish the robust deep nonlinear mapping.

• Cloud model is embedded to fusion time-domain feature with randomness and fuzziness.

• The proposed model shares a good prospect in dam deformation safety early warning.

摘要

•Dynamic early-warning model under spatiotemporal feature deep learning is proposed.•DKELM is improved by correntropy to restrain the interference of input outliers.•Combining POD and the improved DKELM to establish the robust deep nonlinear mapping.•Cloud model is embedded to fusion time-domain feature with randomness and fuzziness.•The proposed model shares a good prospect in dam deformation safety early warning.

论文关键词：Deformation dynamic early warning,Spatiotemporal features,Dynamic updating of early-warning indicator,Robust deep learning,Time-domain feature fusion,Randomness and fuzziness

论文评审过程：Received 13 July 2021, Revised 21 September 2021, Accepted 23 September 2021, Available online 25 September 2021, Version of Record 2 October 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.107537