Modeling dam deformation using independent component regression method

doi:10.1016/S1003-6326(13)62717-X

Transactions of Nonferrous Metals Society of China

Volume 23, Issue 7, July 2013, Pages 2194-2200

https://doi.org/10.1016/S1003-6326(13)62717-X Get rights and content

Abstract

In the application of regression analysis method to model dam deformation, the ill-condition problem occurred in coefficient matrix always prevents an accurate modeling mainly due to the multicollinearity of the variables. Independent component regression (ICR) was proposed to model the dam deformation and identify the physical origins of the deformation. Simulation experiment shows that ICR can successfully resolve the problem of ill-condition and produce a reliable deformation model. After that, the method is applied to model the deformation of the Wuqiangxi Dam in Hunan province, China. The result shows that ICR can not only accurately model the deformation of the dam, but also help to identify the physical factors that affect the deformation through the extracted independent components.

References (16)

R ARDITO et al.
Diagnostic analysis of concrete dams based on seasonal hydrostatic loading [J]
Engineering Structures
(2008)
A SZOSTAK-CHRZANOWSKI et al.
Use of deformation monitoring results in solving geomechanical problems-case studies [J]
Engineering Geology
(2005)
Guang-yong XI et al.
Application of an artificial immune algorithm on a statistical model of dam displacement [J]
Computers and Mathematics with Applications
(2011)
Hong-xiang LI et al.
Dam displacement model based on principal component regression [J]
Hydropower Automation and Dam Monitoring
(2008)
Ying-wei ZHANG et al.
Complex process monitoring using modified partial least squares method of independent component regression [J]
Chemometrics and Intelligent Laboratory Systems
(2009)
A HYVÄRINEN et al.
Independent component analysis: Algorithms and applications [J]
Neural Networks
(2000)
Xue-guang SHAO et al.
A new regression method based on independent component analysis [J]
Talanta
(2006)
Jie XIAO et al.
Analysis and application of automatic deformation monitoring data for buildings and structures of mining area [J]
Transactions of Nonferrous Metals Society of China
(2011)

There are more references available in the full text version of this article.

Cited by (15)

Hierarchical prediction of dam deformation based on hybrid temporal network and load-oriented residual correction
2024, Engineering Structures
Dam deformation prediction methods try to make the predicted values infinitely close to the true values, but no method can capture all the deformation information and residuals are inevitable. For this reason, this paper develops a hierarchical dam deformation prediction method centred on residual correction strategy combined with hybrid temporal network. Firstly, hybrid CNN-LSTM is established based on convolutional neural network (CNN) and long short-term memory network (LSTM), aiming at mining the deformation temporal features and outputting the deformation trend information. Secondly, in the training stage, the mapping relationship between environmental loads and residuals is established by using Extreme Learning Machine (ELM), and the load information embedded in the residuals is mined to output the fluctuation information specific to dam deformation. Then, in the prediction stage, the preliminary deformation prediction results are outputted by CNN-LSTM, and the residual correction values are obtained according to the mapping relationship between environmental loads and residuals, so as to obtain the prediction results. The analysis shows that the outputs of CNN-LSTM can reflect the deformation trend term and provide reliable preliminary prediction results for residual correction, the ELM mapping relation model can capture the loading information embedded in the residuals that cannot be mined by the temporal model, which improves the interpretability of residual correction, and the output results reflect the deformation fluctuations well. The proposed method can realise the adaptive correction of the preliminary deformation prediction results, and the prediction accuracy is satisfactory and adaptable.
A critical review of statistical model of dam monitoring data
2023, Journal of Building Engineering
This paper comprehensively reviews the origin, history, development, applications, and prospects of statistical modeling in the field of dam safety monitoring. The review covers the hydraulic-seasonal-time (HST) model, thermal HST (HSTT) model, hydraulic-temperature-time (HTT) model, hydrostatic-seasonal-state (HSS) model, hydraulic-hysteretic-seasonal-time (HHST) model and time series model, etc. These models were quite commonly used for dam behavior studies. These studies demonstrate the advantage of statistical models in explaining the relationship between structural response and its influence factors and detecting abnormal behavior before any unexpected disasters based on the dam monitoring data. However, the statistical model also has disadvantages, such as difficulty in modeling complex non-linear relationships and a lack of physical meaning for model parameters. Additionally, it will demonstrate the importance of statistical models in controlling data processing and analysis. They can also be used in conjunction with other techniques for data analysis to produce more robust proof for dam monitoring and assessment. We hope that this review will aid in understanding the statistical model of dam monitoring data by researchers and dam engineers, enabling greater practical application of the usefulness and value of statistical models.
A Bayesian approach to model selection and averaging of hydrostatic-season-temperature-time model
2021, Structures
Citation Excerpt :
Using advanced data analytics and various machine learning method, such as neural networks [3], support vector regression [14–16], Gaussian process regression [17], and deep learning method [18] algorithms were developed that can be implemented in real-time. Among other studies [19] using principal component analysis, [20,21] using artificial neural networks, [22] using blind source separation, [23] using artificial immune algorithm, [24] using independent component regression, [25] using deformation separation method, [26] using short term memory model, and [27] using Bayesian dynamic forecasting is worth mentioning here. Most of the previous studies assumed that the what variables to be included in model are known apriori.
This paper presents a probabilistic approach to model the static dam responses, and monitor its performance using structural health monitoring (SHM) data. In general, the main factors that affects the dam responses are hydrostatic pressure, temperature changes, seasonal variation, and age-related deterioration. In this paper, using these variables two models (i.e., Model-1 and Model-2) have been developed to model, predict and monitor the dam performances. The Model-1 is a parsimonious model based on Bayesian model selection principle and can be applied if long-term monitoring data is available. On the other hand, Model-2 is an ensemble model which accounts for the model uncertainty and can be applied during the initial service-life with a few years of monitoring data. In both cases, the model parameters were estimated and updated using Bayesian method where the prior knowledge and SHM data obtained from dam are integrated. Finally, the model residuals i.e., the difference between the measured and predicted dam responses were used to detect any anomaly in the dam performance. The proposed method is illustrated through the crest displacement data obtained from Dongjiang dam located in China.
Structural health monitoring of concrete dams using long-term air temperature for thermal effect simulation
2019, Engineering Structures
Citation Excerpt :
Statistical models based on multiple linear regression and its advanced forms have proven to be more or less successful in dam behavior modeling [5]. The regression models adopted include multiple linear regression (MLR) [3], stepwise regression (SR) [9,10], regression based on principal component analysis [11,12], independent component regression [13] and partial least-squares regression [14], etc. The advantages of MLR models are simplicity of formulation and speed of execution.
Dam health monitoring is usually achieved by using the hydrostatic-season-time model that simulates the temperature effect through harmonic sinusoidal functions. The model is convenient, but does not take into account the effect of temperature variations in different years on dam response. This paper presents a dam health monitoring model using long-term air temperature for thermal effect simulation. In the model, harmonic sinusoidal functions are replaced by long-term air temperature to simulate the temperature effect on the dam response. The machine learning technique RBFN is adopted to mine the temperature effect from long-term air temperature series. The proposed dam health monitoring model was verified using monitoring data of a real concrete gravity dam. Results show that the nonlinear RBFN model utilizing long-term air temperature achieves better results than the model using harmonic sinusoidal functions.
Extracting seasonal deformations of the Nepal Himalaya region from vertical GPS position time series using Independent Component Analysis
2017, Advances in Space Research
Citation Excerpt :
Independent Component Analysis (ICA) is the most widely used BSS method and separates statistically independent signals based on the analysis of the higher order statistical moments of the sources (Comon, 1994). It has been effectively used in ground deformation analysis (Bottiglieri et al., 2007) and dam deformation monitoring (Dai et al., 2013, 2014). Liu et al. (2015) proposed a spatio-temporal filtering method that uses ICA to extract some ‘meaningful’ independent series from GPS networks.
Seasonal variations in vertical GPS time series are mainly caused by the Earth surface mass loading plus bedrock thermal expansion and GPS systematic errors. Relevant surface mass loading phenomena include atmospheric pressure, continental water storage, and non-tidal ocean loading. In regions with significant water storage variations, the hydrological effect makes the greatest contribution to Earth surface seasonal deformation. In this paper, we extract the various seasonal deformations from GPS time series of 17 stations in the Nepal Himalaya region using a blind source separation tool, the Independent Component Analysis (ICA). The vertical displacement components extracted by ICA showed strong correlations with modeled deformation induced by atmospheric and hydrological loading, indicating that independent components may represent different loading effects. Considering the wide application of GRACE in detecting hydrological deformation, we make a comparison between GRACE-derived displacements and the independent displacements contained in GPS. The results show ICA is effective in extracting atmospheric and hydrological deformation.
A multi-variable grey model with a self-memory component and its application on engineering prediction
2015, Engineering Applications of Artificial Intelligence
Citation Excerpt :
Moreover, accurate estimation and effective control of engineering settlement deformation are key to the success of engineering construction. Existing methods for predicting settlement deformation usually adopt the time series analysis model by utilizing monitoring data (Zhou et al., 2010; Xu et al., 2012; Dai et al., 2013), which have effectively promoted the development of the settlement deformation prediction theory and its practice. However, since engineering construction is a complicated nonlinear dynamical system with multiple influencing factors, the issue of settlement deformation involves not only time influence but also spatial effect.
This paper presents a novel multi-variable grey self-memory coupling prediction model (SMGM(1,m)) for use in multi-variable systems with interactional relationship under the condition of small sample size. The proposed model can uniformly describe the relationships among system variables and improve the modeling accuracy. The SMGM(1,m) model combines the advantages of the self-memory principle of dynamic system and traditional MGM(1,m) model through coupling of the above two prediction methods. The weakness of the traditional grey prediction model, i.e., being sensitive to initial value, can be overcome by using multi-time-point initial field instead of only single-time-point initial field in the system׳s self-memorization equation. As shown in the two case studies of engineering settlement deformation prediction, the novel SMGM(1,m) model can take full advantage of the system׳s multi-time historical monitoring data and accurately predict the system׳s evolutionary trend. Three popular accuracy test criteria are adopted to test and verify the reliability and stability of the SMGM(1,m) model, and its superior predictive performance over other traditional grey prediction models. The results show that the proposed SMGM(1,m) model enriches grey prediction theory, and can be applied to other similar multi-variable engineering systems.

View all citing articles on Scopus

: Foundation item: Project (41074004) supported by the National Natural Science Foundation of China; Project (2013CB733303) supported by the National Basic Research Program of China

View full text

Modeling dam deformation using independent component regression method

Abstract

Engineering Structures

Engineering Geology

Computers and Mathematics with Applications

Hydropower Automation and Dam Monitoring

Chemometrics and Intelligent Laboratory Systems

Neural Networks

Talanta

Analysis and application of automatic deformation monitoring data for buildings and structures of mining area [J]

Transactions of Nonferrous Metals Society of China