Dam Settlement Prediction Based on Random Error Extraction and Multi-Input LSTM Network
Authors:
- 1 August 2022
- journal article
- research article
- Published by American Society of Civil Engineers (ASCE) in Journal of Surveying Engineering
- Vol. 148 (3) , 04022006
Abstract
The prediction of dam settlement data plays an important role in analyzing whether the dam is in a safe operation state. But in the field of surveying engineering, the original data measured by instruments will inevitably have random and unpredictable random errors, and the settlement of dams usually has a strong correlation with environmental parameters. In this study, the influence of random error and environmental parameters on dam settlement prediction is discussed, and a prediction model based on multi-input long short-term memory (LSTM) network and random error extraction is proposed. Through the settlement data of a concrete face rockfill dam, the analysis shows that removing random errors can significantly improve the short-term prediction performance and considering environmental parameters can significantly improve the long-term prediction performance. In addition, through comparison and generalization experiments, this method not only has higher prediction accuracy, but also can be applied to other surveying and mapping engineering fields.Keywords
This publication has 17 references indexed in Scilit:
- An empirical comparison of machine learning techniques for dam behaviour modellingStructural Safety, 2015
- Development of support vector regression identification model for prediction of dam structural behaviourStructural Safety, 2014
- EMD-Based Signal FilteringIEEE Transactions on Instrumentation and Measurement, 2007
- A Fast Learning Algorithm for Deep Belief NetsNeural Computation, 2006
- Applications of Hilbert–Huang transform to non‐stationary financial time series analysisApplied Stochastic Models in Business and Industry, 2003
- Approximation theory of the MLP model in neural networksActa Numerica, 1999
- The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysisProceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 1998
- Long Short-Term MemoryNeural Computation, 1997
- Learning long-term dependencies with gradient descent is difficultIEEE Transactions on Neural Networks, 1994
- Learning representations by back-propagating errorsNature, 1986