Dam Health Research Articles

Self-Supervised Dam Deformation Anomaly Detection Based on Temporal-Spatial Contrast Learning.

The detection of anomalies in dam deformation is paramount for evaluating structural integrity and facilitating early warnings, representing a critical aspect of dam health monitoring (DHM). Conventional data-driven methods for dam anomaly detection depend extensively on historical data; however, obtaining annotated data is both expensive and labor-intensive. Consequently, methodologies that leverage unlabeled or semi-labeled data are increasingly gaining popularity. This paper introduces a spatiotemporal contrastive learning pretraining (STCLP) strategy designed to extract discriminative features from unlabeled datasets of dam deformation. STCLP innovatively combines spatial contrastive learning based on temporal contrastive learning to capture representations embodying both spatial and temporal characteristics. Building upon this, a novel anomaly detection method for dam deformation utilizing STCLP is proposed. This method transfers pretrained parameters to targeted downstream classification tasks and leverages prior knowledge for enhanced fine-tuning. For validation, an arch dam serves as the case study. The results reveal that the proposed method demonstrates excellent performance, surpassing other benchmark models.

Missing data imputation model for dam health monitoring based on mode decomposition and deep learning

Missing data imputation model for dam health monitoring based on mode decomposition and deep learning

New Method for Measuring Diagnosis Indexes Based on Concept Lattices in Dam Health Diagnosis

New Method for Measuring Diagnosis Indexes Based on Concept Lattices in Dam Health Diagnosis

An Automated Framework for the Health Monitoring of Dams Using Deep Learning Algorithms and Numerical Methods

Aiming to investigate the problem that dam-monitoring data are difficult to analyze in a timely and accurate automated manner, in this paper, we propose an automated framework for dam health monitoring based on data microservices. The framework consists of structural components, monitoring sensors, and a digital virtual model, which is a hybrid of a finite element (FE) model, a geometric model, a mathematical model, and a deep learning algorithm. Long short-term memory (LSTM) was employed to accurately fit and predict the monitoring data, while dynamic inversion and simulation were used to calibrate and update the data in the hybrid model. The automated tool enables systematic maintenance and management, minimizing errors that are commonly associated with manual visual inspections of structures. The effectiveness of the framework was successfully validated in the safety monitoring and management of a practical dam project, in which the hybrid model improved the prediction accuracy of monitored data, with a maximum absolute error of 0.35 mm. The proposed method can be considered user-friendly and cost-effective, which improves the operational and maintenance efficiency of the project with practical significance.

Multi-Point Deformation Prediction Model for Concrete Dams Based on Spatial Feature Vector

Deformation can effectively reflect the structural state of concrete dams and, thus, establishing na accurate concrete dam deformation prediction model is important for dam health monitoring and early warning strategies. To address the problem that the spatial coordinates introduced in the traditional multi-point deformation prediction model of dams not being able to accurately and efficiently reflect the spatial correlation of multiple-measuring points, a 2D-1D-CNN model is proposed which expresses the spatial correlation between each measuring point through spatial feature vectors, replacing the spatial coordinates in the traditional multi-point model. First, the spatial feature vector is extracted from the historical spatio-temporal panel series of deformation values of measuring points via a Two-Dimensional Convolutional Neural Network (2D-CNN); second, the vector is combined with the environmental impact factor of dam deformation to form the final input factor of fused spatial features; and, thirdly, this vector is combined with the environmental impact factors of dam deformation to form the final input factor of fused spatial features, and the non-linear linkage between the factors and the measured displacement values is constructed by the efficient feature processing capability of a One-Dimensional Convolutional Neural Network (1D-CNN) to obtain the prediction results. Finally, the actual monitoring data of a concrete dam in China are used as an example to verify the validity of the model. The results show that the proposed model outperforms the other models in most cases, respectively, which verifies the effectiveness of the proposed model in this paper.

Deformation prediction model of concrete face rockfill dams based on an improved random forest model

The unique structure and complex deformation characteristics of concrete face rockfill dams (CFRDs) create safety monitoring challenges. This study developed an improved random forest (IRF) model for dam health monitoring modeling by replacing the decision tree in the random forest (RF) model with a novel M5' model tree algorithm. The factors affecting dam deformation were preliminarily selected using the statistical model, and the grey relational degree theory was utilized to reduce the dimensions of model input variables. Finally, a deformation prediction model of CFRDs was established using the IRF model. The ten-fold cross-validation method was used to quantitatively analyze the parameters affecting the IRF algorithm. The performance of the established model was verified using data from three specific measurement points on the Jishixia dam and compared with other dam deformation prediction models. At point ES-10, the performance evaluation indices of the IRF model were superior to those of the M5' model tree and RF models and the classical support vector regression (SVR) and back propagation (BP) neural network models, indicating the satisfactory performance of the IRF model. The IRF model also outperformed the SVR and BP models in settlement prediction at points ES2-8 and ES4-10, demonstrating its strong anti-interference and generalization capabilities. This study has developed a novel method for forecasting and analyzing dam settlements with practical significance. Moreover, the established IRF model can also provide guidance for modeling health monitoring of other structures.

Combined prediction model of concrete arch dam displacement based on cluster analysis considering signal residual correction

Displacement is a crucial indicator reflecting the safety operational state of concrete arch dams. Consequently, the application of displacement monitoring models for arch dam structural health assessment has become widely prevalent. However, conventional singular displacement monitoring methods exhibit limited information extraction capabilities and lack precision, resulting in underexplored residual sequences from the monitoring model and unexcavated valuable information. To address this, this paper integrated cluster analysis with Long Short-Term Memory (LSTM), Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Least Squares Support Vector Machine (LSSVM), and Particle Swarm Optimization (PSO) methods to establish a combined prediction model of concrete arch dam displacement based on cluster analysis considering signal residual correction. Firstly, a model combining cluster analysis k-means++ with LSTM (k-LSTM), was proposed for displacement prediction of arch dams. Secondly, CEEMDAN was employed to adaptively decompose the residual sequences, where Pearson correlation coefficient (PCC) was introduced to filter and reconstruct intrinsic mode components. Subsequently, PSO-LSSVM was utilized to predict the reconstructed sequences and obtain residual correction values. Finally, based on the summation of residual correction values and k-LSTM model predictions, the arch dam displacement combination prediction model was constructed. Engineering case demonstrates the efficiency of this model in arch dam displacement prediction. By effectively mining and utilizing valuable information from residual sequences, this combined model exhibits superior generalization capabilities over traditional singular models, thereby providing technical support and novel research perspectives for concrete arch dam health monitoring and operational management.

An Explainable Probabilistic Model for Health Monitoring of Concrete Dam via Optimized Sparse Bayesian Learning and Sensitivity Analysis

Machine learning has become increasingly popular for modeling dam behavior due to its ability to capture complex relationships between input parameters and dam behavior responses. However, the use of sophisticated machine learning methods for monitoring dam behaviors and making decisions is often hindered by model uncertainty and a lack of interpretability. This paper introduces a novel model for dam health monitoring, focused on monitoring radial displacement and seepage, using optimized sparse Bayesian learning and sensitivity analysis. The model hyperparameters are optimized using an intelligent optimization method integrating the multi-population Rao algorithm and blocked cross-validation, while sensitivity analysis is employed to calculate the relative importance of input variables for a better understanding of the dam’s state. The effectiveness of the proposed model is verified by using long-term monitoring data of a prototype concrete arch dam. The results confirm that the proposed model provides satisfactory performance on both the point predictions and the interval predictions for dam structural behaviors while obtaining effective explainability.

Dam Health Diagnosis Model Based on Cumulative Distribution Function

Dam Health Diagnosis Model Based on Cumulative Distribution Function

Multi-expert attention network for long-term dam displacement prediction

Monitoring and predicting the dam displacement of concrete dams has attracted increasing attention for ensuring the long-term health conditions. Most existing models focus on just temporal features and ignore the spatial features relations of monitoring data. To address these problems, a dam displacement prediction model based on multi-expert network is developed. In the proposed model, the long short-term memory network is employed to extract the temporal features of each monitoring sensor. Then the multi-head attention network is employed to obtain the spatial features and the adjacency values. The multi-expert graph is employed to describe the spatial relations between different monitoring sensors. The graph convolutional network is employed to integrate the spatio-temporal features and predict the long-term dam displacement. Through a real-world comparative study against eight prediction models, the proposed model performs better than other models in both cyclical and non-cyclical time series. Therefore, the proposed model is suitable for evaluating the dam health condition.

Dam Safety Evaluation Method after Extreme Load Condition Based on Health Monitoring and Deep Learning.

The safety operation of dams after extreme load is an important frontier research topic in the field of dam engineering. The dam health monitoring provides a reliable data basis for a safety evaluation after extreme loads. This study proposes a novel data-driven fusion model for a dam safety evaluation after extreme load based on monitoring data derived by sensors. First, the relationship between dam environmental quantity and effect quantity is deeply excavated based on bidirectional long short-term memory (BiLSTM) network, which is a deeply improved LSTM model. Aiming at the parameter optimization problem of BiLSTM model, sparrow search algorithm (SSA), which is an advanced optimization algorithm, is integrated. Second, conducting the constructed SSA-BiLSTM model to estimate the change law of dam effect quantity after the extreme load. Finally, the Mann-Whitney U-test theory is introduced to establish the evaluation criterion of the dam safety state. Project case shows that the multiple quantitative prediction accuracy evaluation indicators of the proposed method are significantly superior to the comparison method, with mean absolute percentage error (MAPE) and mean absolute error (MAE) values decreasing by 30.5% and 27.8%, respectively, on average. The proposed model can accurately diagnose the dam safety state after the extreme load compared with on-site inspection results of the engineering department, which provides a new method for dam safety evaluation.

Multi-source monitoring information fusion method for dam health diagnosis based on Wasserstein distance

The fusion of multi-source monitoring information has become the main trend in the field of dam health diagnosis because of the increasing amount of monitoring data that can be obtained from different sensors. However, the Dempster–Shafer (D–S) evidence theory, an important method in multi-source information fusion, may produce counter-intuitive results when fusing conflicting pieces of evidence. To some extent, existing distance measures can deal with highly conflicting evidence, however, the fusion of completely conflicting evidence (indicated by the conflict coefficient K, K = 1) is typically ignored. This study mainly focuses on the fusion of pieces of completely conflicting evidence considering K = 1. The Wasserstein distance in the field of deep learning is introduced to the D–S evidence theory. Based on the foregoing, the belief Wasserstein–1 distance (BWD) method combined with basic probability assignment is proposed to measure evidence distance. The application of the BWD method in dam health diagnosis is presented to demonstrate the validity and effectiveness of this method in multi-source fusion with completely conflicting evidence.

Displacement prediction model for high arch dams using long short-term memory based encoder-decoder with dual-stage attention considering measured dam temperature

Structural health monitoring method can provide important information to evaluate operational status of concrete dams, by establishing accurate models to predict concrete dam behavior with monitored data. This study proposed a model using encoder-decoder based on long short-term memory network with dual-stage attention mechanism (DALSTM) to predict the displacement of concrete arch dams. Encoder-decoder based on long short-term memory network is a deep learning technique that can perform time series prediction, and dual-stage attention mechanism focuses on the key information in the dam displacement series to improve the performance. The effectiveness and accuracy of the proposed prediction model are analyzed on a high arch dam using measured temperature in the dam body instead of the seasonal functions to represent the thermal effect. Compared with traditional stepwise regression, multiple linear regression models, radial basis function networks, and other deep learning models, results show that the proposed approach performance is more accurate and robust for dam health monitoring.

Towards online monitoring of concrete dam displacement subject to time-varying environments: An improved sequential learning approach

Data-based models are widely applied in concrete dam health monitoring. However, most existing models are restricted to offline modeling, which cannot continuously track the displacement behavior with dynamic evolution patterns, especially in time-varying environments. In this paper, sequential learning is introduced to establish an online monitoring model for dam displacement behavior. This approach starts by considering the timeliness difference between old and new data using the forgetting mechanism, and a novel adaptive forgetting extreme learning machine (AF-ELM) is presented. A primary predictor based on AF-ELM is then formulated, which aims to sequentially learn the complex nonlinear relationship between dam displacement and main environmental factors. Considering the chaotic characteristics contained in the residual sequence of the primary predictor, a multi-scale residual-error correction (REC) strategy is devised based on divide-and-conquer scheme. Specifically, time-varying filter-based empirical mode decomposition is adopted to decompose the raw chaotic residual-error series into a set of subseries with more stationarity, which are further aggregated and reconstructed by fuzzy entropy theory and suitable approximation criterion. Finally, the corrected residual sequence is superimposed with the preliminary predictions from AF-ELM to generate the required modeling results. The effectiveness of the proposed model is verified and assessed by taking a real concrete dam as an example and comparing prediction performance with state-of-the-art models. The results show that AF-ELM performs better in displacement prediction compared with benchmark models, and the multi-scale REC can effectively identify the valuable information within the residual sequence. The proposed online monitoring model can more closely track the dynamic variations of displacement data, which provides a fire-new solution for dam behavior prediction and analysis.

A methodology for dam parameter identification combining machine learning, multi-objective optimization and multiple decision criteria

Physical and mechanical parameter identification could provide useful information for dam health monitoring. Traditional methods for parameter identification commonly use subjective weights to convert multi-objective problems into single-objective problems, which may result in unreasonable identification results. In recent years, multi-objective optimization algorithms have been applied to avoid introduction of subjective weights in parameter identification. Since multi-objective optimization algorithms produce a massive number of Pareto solutions, how to select acceptable parameters from the Pareto solutions is an intractable problem, and effective decision criteria for selecting acceptable solutions is required. In this paper, a methodology combining machine learning, multi-objective optimization, and multiple decision criteria is proposed to effectively improve efficiency and credibility of parameter identification for dams. A surrogate model constructed using the Support Vector Machine (SVM) based on grid search and cross-validation is employed to replace time-consuming finite element calculations in the process of parameter identification. Multiple decision criteria incorporating influencing factors of parameter identification (such as the membership degree of identified parameters with respect to designed ones, fitting error of monitoring data, and the degree of overfitting) are devised to compare Pareto solutions derived from multi-objective optimization and determine the optimal one. Parameter identification of an actual gravity dam is implemented to illustrate and verify the proposed methodology. It is shown that the results yielded with the proposed methodology are superior to those produced by traditional methods in terms of robustness of reasonable parameter identification, prediction accuracy of monitoring data, and reasonable physical meaning of material parameters.

Associations between putative risk factors and poor colostrum yield in Holstein Friesian cattle

When dairy cows produce little or no colostrum, calves are likely to suffer from failure of passive transfer (FPT). Volume of colostrum produced by the dam may be affected by: nutrition; environmental conditions; time from calving to milking; parity; dry period length; calving difficulty; calf weight; calf sex; calf viability; cow BCS / body weight; milk production in previous lactation and dam health. While risk factors for poor IgG concentration in colostrum have been extensively studied, there is little published literature on low colostrum yields and associated risk factors. The aim of this observational study was to identify risk factors for poor colostrum yield in dairy cows and to quantify the effect of variation in those variables that the farmer can directly control (length of dry period, and close-up transition period nutrition) using inverse probability modelling. 439 animals from a single Scottish dairy farm were used as a convenience sample. First milking colostrum yield was measured by farmers using a graduated bucket. For each animal, risk factor data and meteorological data from the nearest weather were obtained. Decreased colostrum yield was associated with increasing daily temperature range, while increasing UV index, lactation number and number of days dry increased colostrum yield. Marginal effect models revealed that for each additional day of dry period, colostrum yield increased. Higher UV indices the month before calving were positively associated with colostrum yield, particularly in primiparous cows, and higher temperatures in the months before calving were positively associated with colostrum yield in all cows.

SSA optimized back propagation neural network model for dam displacement monitoring based on long-term temperature data

Featured with the harmonic sinusoidal function to reflect temperature effects, the hydrostatic-season-time (HST) model is often used to monitor the concrete gravity dam health, but it does not take account of the effects of environment temperatures in real-term and has flaws especially when applied in conditions of significant temperature variations. A model of Sparrow Search Algorithm optimized error Back Propagation neural network (SSA-BP) based on the hydrostatic-temperature-time (HTT) model is proposed in this paper for predicting the concrete gravity dam displacement using the long-term environment temperature variable sets to reflect temperature effects. Successive Projections Algorithm (SPA) is used for the first time for feature selection on long-term temperature variables to further optimize the model (as SPA-SSA-BP). Through a case study with the practical observed data from a reality high concrete gravity dam, the effectiveness of the new model is verified, suggesting that HTT-based SSA-BP models have better performance than HST with the best result obtained when using the 2-year long variable sets. The SSA-BP model has much lower error in predicting the concrete dam displacement than Multiple Linear Regression (MLR). The arithmetic speed and prediction accuracy of the SPA-SSA-BP model is optimized as it can minimize the collinearity among feature variables in the long-term HTT variable sets, bring down the input variable dimension close to the level of HST, and diminish the redundant data information.

Multi-sensor real-time monitoring of dam behavior using self-adaptive online sequential learning

The behavior monitoring model is the most widely used method in dam health monitoring, but existing methods still concentrate mainly on offline modeling or batch learning, neglecting the timeliness requirement. This paper describes an online model based on sequential learning for real-time monitoring of dam displacement behavior. The proposed method involves two major modeling stages. First, kernel principal component analysis (KPCA) is used for multi-sensor data fusion to extract the more essential contextual components and remove the multicollinearity from the raw sensory data. Second, a novel self-adaptive online sequential extreme learning machine (SOS-ELM) is presented to efficiently capture the complex nonlinear mapping from environmental variables to displacements by coupling the classic OS-ELM with regularization technique and forgetting mechanism. The proposed model is verified using monitoring data of a real-world concrete dam. The results show that the proposed sequential model can obtain satisfactory prediction accuracy with a low computational cost, and hence, is a competitive modeling tool for dam behavior monitoring.

A combined finite element and deep learning network for structural dynamic response estimation on concrete gravity dam subjected to blast loads

Social infrastructures such as dams are likely to be exposed to high risk of terrorist and military attacks, leading to increasing attentions on their vulnerability and catastrophic consequences under such events. This paper tries to develop advanced deep learning approaches for structural dynamic response prediction and dam health diagnosis. At first, the improved long short-term memory (LSTM) networks are proposed for data-driven structural dynamic response analysis with the data generated by a single degree of freedom (SDOF) and the finite numerical simulation, due to the unavailability of abundant practical structural response data of concrete gravity dam under blast events. Three kinds of LSTM-based models are discussed with the various cases of noise-contaminated signals, and the results prove that LSTM-based models have the potential for quick structural response estimation under blast loads. Furthermore, the damage indicators (i.e., peak vibration velocity and domain frequency) are extracted from the predicted velocity histories, and their relationship with the dam damage status from the numerical simulation is established. This study provides a deep-learning based structural health monitoring (SHM) framework for quick assessment of dam experienced underwater explosions through blast-induced monitoring data.

Data-based model with EMD and a new model selection criterion for dam health monitoring

Data-based models are extensively applied in concrete dam health monitoring. In this paper, the Empirical Mode Decomposition (EMD) is adopted to extract and visualize the irreversible component of deformation monitoring sequence. Consequently proper mathematical formulation of the irreversible component of deformation monitoring model for concrete dams can be established, and the problem of improperly assuming the mathematical form of irreversible component when constructing data-based models is well addressed. The periodic component of the deformation time series is modelled by regression and three machine learning algorithms (i.e., Extreme learning machine, Support Vector Machine based on Grid search and Cross-Verification, and Back Propagation Neural Networks optimized by Genetic Algorithm and Simulated Annealing) respectively. The data-based models with the EMD are able to effectively improve the performance of formulating irreversible component of concrete dam deformation and decrease the over-fitting levels. A new model selection criterion, namely Over-fitting coefficient, is proposed for evaluating and selecting the optimum model among candidate monitoring models. The Over-fitting coefficient can better reflect over-fitting levels of individual monitoring models compared to traditional model selection criteria. As a case study, the displacement and strain monitoring data from the YL gravity dam are modeled by data-based models with the EMD and machine learning algorithms. It is shown that data-based models with the EMD are of higher prediction accuracy and lower rate of false alarm in deformation monitoring than traditional models. The Over-fitting coefficient can better reflect over-fitting levels of monitoring models compared to the error measurement criteria and information criteria. The case study highlights the merits of the proposed data-based models and the model selection criterion for health monitoring of concrete dams.