Exploring Purpose-Driven Methods and a Multifaceted Approach in Dam Health Monitoring Data Utilization

Abstract

Dam monitoring tracks environmental variables (water level, temperature) and structural responses (deformation, seepage, and stress) to assess safety and performance. Structural health monitoring (SHM) refers to the systematic observation and analysis of the structural condition over time, and it is essential in maintaining the safety, functionality, and long-term performance of dams. This review examines monitoring data applications, covering structural health assessment methods, historical motivations, and key challenges. It discusses monitoring components, data acquisition processes, and sensor roles, stressing the need to integrate environmental, operational, and structural data for decision making. Key objectives include risk management, operational efficiency, safety evaluation, environmental impact assessment, and maintenance planning. Methodologies such as numerical modeling, statistical analysis, and machine learning are critically analyzed, highlighting their strengths and limitations and the demand for advanced predictive techniques. This paper also explores future trends in dam monitoring, offering insights for engineers and researchers to enhance infrastructure resilience. By synthesizing current practices and emerging innovations, this review aims to guide improvements in dam safety protocols, ensuring reliable and sustainable dam operations. The findings provide a foundation for the advancement of monitoring technologies and optimization of dam management strategies worldwide.

1. Introduction

For dam safety and longevity, SHM is required. Several catastrophic dam failures throughout history highlight the critical need for improved structural safety and real-time monitoring [1]. Several significant dam failure events, caused by a combination of factors such as aging infrastructure, extreme weather conditions, seismic activity, and more, have been discussed to highlight the importance of structural safety and monitoring. The 1975 Banqiao Dam failure in China, one of the deadliest in history, was triggered by extreme rainfall from Typhoon Nina, which exceeded the dam’s capacity and was worsened by design flaws, an inadequate spillway capacity, and communication failures [1]. The 2018 Oroville Dam crisis in the United States resulted from spillway erosion due to design flaws and insufficient maintenance, prompting the evacuation of over 180,000 people. The 2019 Brumadinho Dam collapse in Brazil, which claimed 270 lives, resulted from delayed slip surface growth in weak tailing layers, exacerbated by the construction history, post-closure creep, and insufficient regulatory oversight [2]. The 1976 Teton Dam failure in the United States occurred due to the internal erosion of the earthen embankment during the initial reservoir filling phase [3]. The 2011 Fujinuma Dam failure in Fukushima, triggered by a magnitude 9.0 earthquake—the strongest ever recorded in Japan—caused eight fatalities and highlighted the seismic vulnerability of aging earth fill dams, emphasizing the urgent need for improved design standards and risk assessment strategies [4]. These events stress the need for stronger safety measures and more responsive monitoring systems to help prevent similar incidents.

The multifaceted purpose, advanced technologies, and basic principles of dam monitoring data utilization are examined in this comprehensive work. Dam safety and reliability need strategic sensor data collection [5,6,7]. Safety, early warning, and maintenance optimization are dam monitoring’s core goals [8,9,10,11]. Data provide long-term insights into the dam’s operation, enabling early intervention and targeted maintenance [12]. Engineers and operators can dynamically assess the dam’s structural health thanks to sensors [12,13]. DMD is fundamental despite hurdles due to dam structures’ dynamic nature, environmental impacts, and the need for constant and accurate data gathering [14]. The complexity of protecting these vital assets is shown by the many challenges. Various variables make SHM problematic in the goal of dam safety and reliability [15,16]. Dam structures are dynamic and subject to environmental and operational pressures; therefore, understanding their behavior over time is crucial. The scale and complexity of these structures make continuous and accurate data collection difficult [17,18,19]. Despite this complexity, dam SHM continues to be practiced, ensuring the longevity and reliability of these vital structures. Regularly monitoring water flows and structural stability helps to identify faults and guide maintenance. Water release and electricity generation decisions are safer and more efficient with data monitoring, promoting water resource management and regulatory compliance [20,21]. Detecting abnormalities, analyzing structures, measuring risks, and installing early warning systems improve dams’ resilience, operating efficiency, emergency readiness, and sustainability [13,22,23,24]. Monitoring encourages dam engineering research and improvement, allowing trend analysis to prioritize maintenance and prevent major issues [5]. Data management optimizes dam operation resource allocation and cuts costs. Central systems use data to ensure structural integrity, optimize designs, and achieve operational planning, emergency readiness, and environmental stewardship throughout the dam’s lifecycle, promoting sustainable dam management [25].

In dam management, civil engineers must address structural integrity, operation, maintenance, inspection, and safety [26]. The magnitude, age, and material uncertainty of dams necessitate new solutions [27]. Using localized point sensors to monitor crucial metrics in real time can help to identify concerns early [28]. Sensors record deformations, tension, cracks, seepage, and temperatures to give a complete picture of the dam [13,29,30]. These data are sent to a central system for sophisticated analysis, which informs dam design, building, operation, and maintenance decisions [31]. Beyond safety assurance, monitoring data provide real-world performance data for design optimization [32]. This ensures quality during construction and guides maintenance planning depending on the dam condition throughout operation [33,34]. Innovative methods like smart film crack monitoring [35,36] and the SHM-X Crackmeter [37] measure cracks efficiently. For strain measurement, vibrating wire strain gauges are extensively employed, and displacement, which indicates stability, is important [38]. In addition, FOSs have emerged as indispensable tools in structural health monitoring due to their high sensitivity, immunity to electromagnetic interference, and ability to perform distributed sensing over long distances [39]. These sensors are widely used for temperature and deformation monitoring in dam structures, providing real-time, accurate data that are critical for early damage detection and structural assessment [39,40]. Sensor placement in dam structural health monitoring is crucial, with wireless sensors becoming prevalent. IoT and big data solutions have revolutionized data collection, improving structural health assessments [6,41]. In sensor layout optimization, machine learning (ML) modifications enable informative sensor placements, enhancing system performance and lowering deployment costs [42,43]. This integrated method transforms data into meaningful insights, boosting dam infrastructure over time. Section 3.2 emphasizes data gathering and sensors’ functional roles in dam SHM and shows the responses acquired from dam SHM by typical sensors.

Maintaining dam integrity requires diverse monitoring systems, including numerical, statistical, ML, and hybrid models. Each approach offers unique strengths: numerical models simulate complex systems with precision, statistical models analyze patterns in data, ML models provide predictive adaptability, and hybrid models integrate strategies for comprehensive monitoring [44,45]. These models are influenced by the time coverage and resolution, with differences in application for embankment and concrete gravity dams. HM addresses complex monitoring challenges by combining the strengths of individual techniques. Section 4 includes a table comparing the uses, benefits, and drawbacks of each method, aiding in solution selection. This review emphasizes the importance of combining techniques for advanced predictive analytics, risk assessment, and ensuring dam safety and durability.

Dam safety challenges are increasing due to aging infrastructure, environmental variability, and operational uncertainties, making advanced SHM integration essential. This study highlights the need for a cohesive framework connecting monitoring components, sensor technologies, and analytical models to enhance the predictive accuracy and optimize maintenance. By examining current practices and identifying critical gaps, it offers valuable insights and promotes interdisciplinary collaboration, aiming to drive innovation and improve the safety, resilience, and long-term performance of dam infrastructure.

To address the challenges above, this study explores the objectives and technologies shaping modern SHM systems for dams. It begins with explaining why dam monitoring data are collected, focusing on structural safety, failure prediction, and maintenance optimization, along with criteria for system selection. Next, it examines monitoring technologies, from satellite imaging to embedded sensor networks, highlighting SHM’s evolution. Then, it evaluates conventional and emerging sensors, including fiberoptic systems and drones, to enhance efficiency and real-time response. It subsequently analyzes data utilization through numerical, statistical, and machine learning methods, revealing model effectiveness. By adopting a multidimensional approach, this study advances the SHM discourse and offers practical guidance to enhance dam safety, operational efficiency, and long-term infrastructure resilience.

1.1. History and Motivations for Dam SHM

The origin of SHM in dams stems from the need to ensure safety and longevity, driven by historical dam failures and evolving civil engineering priorities [46]. Early methods included visual inspections in the 1950s–60s [47], progressing to strain gauges in the 1970s–80s [48], automated systems in the 1990s–2000s [49], and satellite/LiDAR in the 2010s [50]. The 2020s saw smart sensors, IoT, and AI improving real-time data and predictive modeling [6]. Future innovations include advanced materials, 5G, and robotic inspections. Figure 1 illustrates this evolution, highlighting the rise of SHM as key to dam safety and reliability.

1.2. Structure of the Review Paper

This review explores the development and objectives of dam SHM systems, highlighting key components, sensor roles, and data types. It examines how monitoring data serve purposes like safety, maintenance, and environmental assessment. Modeling methods, including numerical, statistical, and AI-based, are reviewed with their applications and limitations, as illustrated in Figure 2.

1.3. Research Methodology

In this study, we conducted a systematic review of the literature pertaining to the utilization of DMD. The adopted research methodology is comprehensively illustrated in

Table 1. Summary of research methodology for dam monitoring data literature review.

Stage	Description
Database Search	Primary database: Web of Science, covering academic journals, conference proceedings, and research papers.
Keyword Setting	Keywords grouped into three categories: 1. Dam health monitoring (DHM), dam health surveillance, DSHM, dam structural safety; 2. Dam monitoring sensors; 3. Data enhancement techniques (ML, statistical, numerical, FEM, neural networks). Time-limited to capture recent advances.
Results	Over 400 relevant journal articles identified.
Manual Selection Process	1. Preliminary Screening: Titles and abstracts reviewed to exclude irrelevant articles. 2. Rescreening: Detailed review based on study objectives. 3. In-Depth Analysis: Content, techniques, and relevance examined. 4. Exclusion Criteria: Articles outside the topic removed. 5. Final Selection: 28 review papers chosen for comprehensive analysis from 360 articles.
Quality Assurance	Manual screening ensured inclusion of high-quality publications, strengthening the study’s reliability.
AI Assistance	AI technologies were used to improve the grammar, syntax, and clarity of the manuscript, enhancing readability. Additionally, AI tools helped to organize key sections, ensuring consistency and coherence throughout the text. However, all content, analysis, and insights are the original work of the authors.

, which delineates the sequential process, ranging from database selection and keyword formulation to manual screening and final inclusion criteria, ensuring a rigorous and structured approach to literature selection.

1.4. Comparison of Relevant Literature Reviews on SHM

While SHM has been widely reviewed, dam-specific SHM remains underexplored. As shown in

Table 2. The pioneering reviews works in the literature on SHM. The data collected included studies published up to February 2024.

Ref.	Title	Remarks/Objectives	Implications for Future Work
Liu et al. [5]	A critical review of statistical models of dam monitoring data	Statistical models address abnormalities, structural responses, and non-linear parameter interactions.	Explore advanced statistical methods for enhanced dam safety monitoring.
Hariri-Ardebili et al. [51]	The role of artificial intelligence and digital technologies in dam engineering	Reviews AI and ML in dam engineering, focusing on forecasting dam responses.	Advocate for AI integration with traditional methods for predictive modeling.
Hassani et al. [28]	A Systematic Review of Advanced Sensor Technologies for Non-Destructive Testing and Structural Health Monitoring.	Explores modern sensors in SHM, merging traditional and advanced technologies.	Enhance sensor systems to address SHM accuracy and reliability challenges.
Chen et al. [52]	Marine Structural Health Monitoring with Optical Fiber Sensors: A Review	Discusses optical fiber sensors for marine SHM, with a focus on AI/ML applications.	Develop AI-enhanced optical sensor systems for maritime SHM precision.
Payawal et al. [53]	Image-Based Structural Health Monitoring: A Systematic Review	Reviews image-based SHM applications and suggests ML and drone integration.	Design ML-driven, image-based SHM systems for efficient structural evaluation.
Cheng et al. [54]	A Literature Review and Result Interpretation of the System Identification of Arch Dams Using Seismic Monitoring Data.	Uses seismic data for modal identification and structural monitoring of arch dams.	Improve dynamic models for vibration analysis and time-variant characteristics.
Prakash et al. [55]	Recent advancement of concrete dam health monitoring technology: A systematic literature review	Examines numerical and hybrid models and monitoring technologies for dams.	Investigate excitation methods, seismic dynamics, and calibration techniques.
Azimi et al. [56]	Data-driven structural health monitoring and damage detection through deep learning: State-of-the-art review	Reviews DL in SHM, covering methodologies like vision-based monitoring and data science.	Advance SHM automation, data quality, and hierarchical damage assessment frameworks.
Clarkson et al. [57]	Critical review of tailings dam monitoring best practice.	Establishes global safety practices for tailing dam monitoring through collaboration.	Foster international cooperation to refine tailing dam safety protocols.
Tan et al. [58]	Computational methodologies for optimal sensor placement in structural health monitoring: A review	Reviews ML-based SHM systems and optimal sensor placement methods.	Optimize algorithms for precise sensor placement and improved damage detection.
Huđek et al. [59]	A review of hydropower dams in Southeast Europe—distribution, trends, and availability of monitoring data using the example of a multinational Danube catchment subarea	Analyzes environmental monitoring and ecological impacts of hydropower dams.	Balance energy production with biodiversity through mitigation strategies.
Sanuade et al. [60]	Prediction of transmissivity of aquifer from geoelectric data using artificial neural network	Uses geophysical methods to monitor dam seepage and stability.	Enhance ANN models with diverse geological inputs for improved transmissivity prediction.
Salazar et al. [61]	Data-Based Models for the Prediction of Dam Behavior: A Review and Some Methodological Considerations	Discusses ML-based dam safety analysis and predictive modeling.	Refine model validation, data preprocessing, and generalization for better safety analysis.
Bukenya et al. [19]	Health monitoring of concrete dams: a literature review	Reviews long-term SHM methods for static and dynamic concrete dam monitoring.	Integrate *GNSS and remote sensing data for comprehensive SHM approaches.

*GNSS: Global Navigation Satellite System.

, existing studies cover the methodology and objectives, but gaps remain in integrating monitoring data.

Table 3. Gaps in the pioneering review works in the literature on dam SHM.

Gap in Existing Research	Objectives of Current Research
Dam monitoring components and processes	Conduct a comprehensive analysis of dam monitoring components and processes, highlighting opportunities for technological advancement and process optimization.
Dam SHM response through various common sensors	Investigate and compare the performance of commonly used sensors in dam SHM, focusing on improving sensor integration and enhancing data accuracy.
Purpose of monitoring	Clarify and expand on the multifaceted purposes of dam monitoring, aiming to align these purposes with current safety, operational, and environmental standards. Also, emphasize the importance of fostering interdisciplinary collaboration in SHM systems for dams to improve overall monitoring effectiveness.
Strengths and drawbacks of various models	Systematically evaluate and document the effectiveness, limitations, and applicability of various SHM models in real-world dam monitoring scenarios.

shows future trends and emerging technologies.

2. SHM System for Dams

The surveillance and monitoring of critical infrastructure, like dams, are essential for safe operation and in preventing catastrophic failures [5]. Dams generate power, supply water, and protect against floods, but their failure risks financial losses and lives downstream [5,23,62,63,64,65,66]. Despite safety factors in design, uncertainties in material properties and working conditions pose risks [67,68,69,70,71,72]. Structural deterioration over time necessitates continuous monitoring [73,74,75]. Instruments and visual inspections assess construction, foundations, slopes, and the environment, with data analyzed to verify design quality and ensure safety [20,57,76,77,78,79,80,81]. Figure 3 illustrates the importance of monitoring.

Behavior models predict dam responses to various loads, linking external causes to dam reactions [61,82,83,84]. Mechanistic models use physical laws but face material uncertainty, while statistical models, such as regression, analyze monitoring data efficiently [5,84]. Automated dam equipment has increased data availability, enabling ML models to gain insights [85]. However, hydraulic complexity presents challenges in capturing nonlinear, time-varying behaviors under fluctuating conditions, necessitating hybrid models combining physical laws and machine learning to improve prediction and real-time decision-making, as mentioned in Section 4.

SHM systems enhance resilience by verifying design assumptions, detecting abnormalities, and informing repairs [15,55,86,87]. These systems facilitate maintenance, reduce risks, and extend dam longevity despite their costs and complexity [20]. Real-time technologies enable proactive management, ensuring critical dam safety and functionality [78].

2.1. Components of Dam SHM Systems

The development of SHM systems for dams has progressed from manual, mechanical instruments to advanced digital platforms [88]. Early methods were labor-intensive and lacked real-time capabilities. Over time, electronic sensors, wireless communication, and cloud computing have enabled automated, accurate, and real-time monitoring [89]. Recent advances in machine learning and data analytics have transformed SHM into an intelligent framework supporting timely maintenance and improved operational reliability [90]. SHM systems consist of key components, as shown in Figure 4. Wireless communication, although costlier, suits large structures. Data are collected via sensors and processed, stored, analyzed, and used to guide inspection and maintenance based on damage assessment [91,92].

2.2. Data Acquisition and the Purposes of Sensors in SHM

Monitoring data are essential in evaluating dam behavior through strategically placed sensors that capture structural and environmental parameters [13,15,93,94]. SHM systems depend on real-time data acquisition related to deformation, stress, temperature, and pore water pressure, enabling informed decisions for maintenance and safety [29,94,95,96,97,98,99,100]. Effective data storage and analysis frameworks are critical in interpreting sensor outputs [62]. Table 4 outlines the sensor types, locations, relevance, and trade-offs for SHM. Sensor selection is based on their sensitivity, stability, precision, and response time, aligned with monitoring objectives [101,102,103,104,105]. Wired sensors are widely used for long-term applications [106,107], whereas WSS technologies offer portable, cost-effective alternatives with decentralized operation [91,92,108,109,110,111]. WSS integrates sensing, processing, and transmission, supporting high-density deployments [112]. Techniques like cosmic-ray myography offer high-resolution imaging for internal structural assessment [111]. Centralized IT systems enhance data validation and fault detection [43,113].

IoT platforms have transformed SHM, enabling real-time, intelligent monitoring [6,41,107]. Raspberry Pi-based systems with piezoelectric sensors demonstrate scalable solutions for structural assessment [114,115,116]. Big data and ML techniques optimize sensor layouts and damage detection [42,43,117,118,119,120,121,122]. DL models, such as radial basis function networks, along with CV, LiDAR, and satellite imagery, enhance predictive diagnostics and support multidisciplinary applications [23,123,124,125,126,127].

Dam monitoring instrumentation serves several purposes: (i) assessing geological risks to inform design; (ii) monitoring water levels and structural integrity for safe operation; (iii) verifying construction adherence to safety protocols; (iv) comparing actual performance with design assumptions; (v) mitigating risks like seepage; and (vi) ensuring legal compliance and owner interests. Remote sensing technologies also measure precipitation, runoff, groundwater, and reservoir conditions [50].

Table 4. Dam SHM response through various common sensors.

Sensor		References and Source of Image	Function and SHM Measurement	Purpose in SHM	Sensor Location *	Advantages and Limitations	Image of Instrument
Structural Monitoring	Vibrating Wire Crackmeter	[37,128,129]	Long-term monitoring; Measures changes in vibrating wire length to detect cracks or structural deformation.	Monitors the development and progression of cracks, providing early warnings for maintenance and repair.	Installed at critical points in the dam structure prone to cracking.	Easy installation, robust, accurate, waterproof, long-term stability, remote reading, over-voltage protection. Sensitive to environment, range restrictions, regular maintenance.
	Tiltmeter	[130,131]	Long-term monitoring; Records angular displacement or inclination data to identify structural tilting or movement.	Detects excessive movement or deformation, aiding in assessing dam stability and identifying potential risks.	Placed at various locations to measure tilt or movement of dam structure.	Precise, real-time monitoring, versatile installation, calibrated accuracy. Depth constraints, cost, maintenance needs, installation complexity, data interpretation challenges.
	SHM-X Crackmeter	[37,132,133]	Long-term monitoring; Measures changes in crack and expansion joint widths using simple mechanical tools.	Monitoring changes in crack and joint widths.	Typically installed across cracks or expansion joints in various structural elements	Simple, low cost, quick installation, weather-resistant, flexible modes. Nature dependency, expert measurements required.
	Smart film crack sensor	[35,36,134]	Long-term and short-term monitoring; Detects cracks, including their initiation, propagation, shape, and location.	Critical in assessing crack initiation, length, propagation, shape, and location, preventing structural risks.	Typically installed across cracks or expansion joints in various structural elements.	Detailed crack data, advanced processing, reliable communication. Signal instability, interference, environmental sensitivity.
	Strain gauge	[135,136,137]	Short-term monitoring; Measures strain in concrete and steel structures, providing insights into stress distribution and potential structural weaknesses.	Provides insights into stress distribution, structural behavior, and potential weaknesses.	Mounted on steel, iron structures; embedded in reinforced concrete.	Operates in wide temperatures, cost-effective, dynamic load measurement. Installation complexity, electromagnetic interference, expensive data equipment.
	Gecko Tremor	[138,139]	Short-period seismograph; Detects and records local earthquake activity near the dam site.	Designed for local earthquake monitoring, filling gaps between broadband stations, and aftershock monitoring.	Placed at any point on the dam or near the dam.	Affordable, records up to 254 mm/s, quick setup, solar-powered. Limited channels, vague data processing routines.
	Multipoint extensometer	[140,141,142]	Long-term monitoring; Provides deformation data at multiple points within the dam structure for structural assessment.	Monitors deformation, assesses structure, identifies issues, ensures safety.	Installed at various locations to monitor deformations.	Precision monitoring, collapse prediction, slope tracking. Installation complexity, sagging in deep boreholes, high initial costs.
	FOS	[39,40]	Long-term and short-term monitoring; Measures strain, temperature, pressure, and deformation using light-based signals.	Monitors internal stress, thermal effects, crack development, and displacement.	Within dam body, galleries, tunnels, foundations, abutments.	Immune to EM interference, high sensitivity, long-distance capabilities, durable in harsh environments. Higher cost, complex installation and calibration requirements.
Environmental Monitoring	Meteorological station	[143,144]	Long-term and short-term monitoring; Records environmental conditions including temperature, humidity, wind speed, and precipitation.	Assesses dam safety by understanding environmental conditions, aiding in informed decision making.	Placed near the dam to measure weather conditions.	Accurate data, hourly updates, comprehensive weather insights. Site influence, transmission reliance, limited accessibility.
	Temperature sensor	[132]	Long-term and short-term monitoring; Measures temperature variations within dam materials to detect stress and thermal effects.	Monitors temperature variations that affect material properties and structural behavior, helping to identify potential issues.	Distributed throughout the dam structure	Durable, versatile, advanced analysis. High initial cost, expertise required, sensitivity to conditions.
Geotechnical Monitoring	Earth pressure cell	[145,146,147]	Long-term and short-term monitoring; Records lateral earth pressure data to assess soil–structure interaction and foundation behavior.	Provides information on soil behavior and potential movement and aids in the assessment of dam stability and soil–structure interaction.	Installed in the dam foundation or surrounding soil.	Precise pressure monitoring, remote data access, versatile. Skilled installation, regular maintenance, high project costs.
	Piezometer	[132]	Long-term and short-term monitoring; Measures water pressure, seepage, and groundwater levels within and around the dam.	Monitors water pressure, seepage, groundwater levels, and potential leakage, crucial for dam safety and maintenance.	Placed within the dam or surrounding areas.	Reliable, cost-effective, easy automation. Freezing issues, clogging, unsuitable for artesian conditions.
Safety Monitoring	Vibration monitor	[132]	Long-term and short-term monitoring; Detects vibration levels and frequencies to assess structural integrity and identify potential issues.	Detects excessive vibrations indicating structural issues or risks, enabling proactive measures for stability and safety.	Distributed throughout the dam structure.	Comprehensive monitoring, durable, standard-compliant. Costly, technical expertise required, battery dependence.
	Geophone, also called jug or tortuga	[148,149]	Long-term and short-term monitoring; Monitors ground movement at the foundation or dam body to detect settlement or instability.	Foundation for settling, dam body for deformation, slide-prone areas for warning, seismic fault lines for earthquakes.	Could be installed on the dam foundation or dam body.	Sensitive, portable, cost-effective, efficient. Surface noise, depth limits, data complexity, environmental sensitivity.
	Strain gauge load cell	[132,150]	Long-term and short-term monitoring; Measures load or force acting on structural components for performance evaluation.	Measures actual loads on structural elements, aiding in the assessment of structural integrity and performance.	Installed at critical load-bearing components of the dam structure.	Accurate, dynamic, remote monitoring. Calibration needs, damage susceptibility, costly maintenance.
	** InSAR	[132]	Long-term monitoring; InSAR provides high-precision ground movement data for detection of subsidence, landslides, and structural shifts.	Provides detailed measurements of ground movement, subsidence, and landslides, helping to identify potential risks and their impacts on the dam structure.	Focusing on infrastructure in targeted, specific project areas. It focus on the dam structure and surrounding areas for stability analysis.	High precision, remote sensing, no ground setup. Costly, atmospheric disturbances, radar reflectivity dependence.

* The locations of sensors in dam structures or in geological surveying depend largely on the objectives of the monitoring or exploration project. The column outlines optimal locations or gives general guidelines for sensor placement. ** InSAR: Interferometric Synthetic Aperture Radar.

Table 4. Dam SHM response through various common sensors.

Sensor		References and Source of Image	Function and SHM Measurement	Purpose in SHM	Sensor Location *	Advantages and Limitations	Image of Instrument
Structural Monitoring	Vibrating Wire Crackmeter	[37,128,129]	Long-term monitoring; Measures changes in vibrating wire length to detect cracks or structural deformation.	Monitors the development and progression of cracks, providing early warnings for maintenance and repair.	Installed at critical points in the dam structure prone to cracking.	Easy installation, robust, accurate, waterproof, long-term stability, remote reading, over-voltage protection. Sensitive to environment, range restrictions, regular maintenance.
	Tiltmeter	[130,131]	Long-term monitoring; Records angular displacement or inclination data to identify structural tilting or movement.	Detects excessive movement or deformation, aiding in assessing dam stability and identifying potential risks.	Placed at various locations to measure tilt or movement of dam structure.	Precise, real-time monitoring, versatile installation, calibrated accuracy. Depth constraints, cost, maintenance needs, installation complexity, data interpretation challenges.
	SHM-X Crackmeter	[37,132,133]	Long-term monitoring; Measures changes in crack and expansion joint widths using simple mechanical tools.	Monitoring changes in crack and joint widths.	Typically installed across cracks or expansion joints in various structural elements	Simple, low cost, quick installation, weather-resistant, flexible modes. Nature dependency, expert measurements required.
	Smart film crack sensor	[35,36,134]	Long-term and short-term monitoring; Detects cracks, including their initiation, propagation, shape, and location.	Critical in assessing crack initiation, length, propagation, shape, and location, preventing structural risks.	Typically installed across cracks or expansion joints in various structural elements.	Detailed crack data, advanced processing, reliable communication. Signal instability, interference, environmental sensitivity.
	Strain gauge	[135,136,137]	Short-term monitoring; Measures strain in concrete and steel structures, providing insights into stress distribution and potential structural weaknesses.	Provides insights into stress distribution, structural behavior, and potential weaknesses.	Mounted on steel, iron structures; embedded in reinforced concrete.	Operates in wide temperatures, cost-effective, dynamic load measurement. Installation complexity, electromagnetic interference, expensive data equipment.
	Gecko Tremor	[138,139]	Short-period seismograph; Detects and records local earthquake activity near the dam site.	Designed for local earthquake monitoring, filling gaps between broadband stations, and aftershock monitoring.	Placed at any point on the dam or near the dam.	Affordable, records up to 254 mm/s, quick setup, solar-powered. Limited channels, vague data processing routines.
	Multipoint extensometer	[140,141,142]	Long-term monitoring; Provides deformation data at multiple points within the dam structure for structural assessment.	Monitors deformation, assesses structure, identifies issues, ensures safety.	Installed at various locations to monitor deformations.	Precision monitoring, collapse prediction, slope tracking. Installation complexity, sagging in deep boreholes, high initial costs.
	FOS	[39,40]	Long-term and short-term monitoring; Measures strain, temperature, pressure, and deformation using light-based signals.	Monitors internal stress, thermal effects, crack development, and displacement.	Within dam body, galleries, tunnels, foundations, abutments.	Immune to EM interference, high sensitivity, long-distance capabilities, durable in harsh environments. Higher cost, complex installation and calibration requirements.
Environmental Monitoring	Meteorological station	[143,144]	Long-term and short-term monitoring; Records environmental conditions including temperature, humidity, wind speed, and precipitation.	Assesses dam safety by understanding environmental conditions, aiding in informed decision making.	Placed near the dam to measure weather conditions.	Accurate data, hourly updates, comprehensive weather insights. Site influence, transmission reliance, limited accessibility.
	Temperature sensor	[132]	Long-term and short-term monitoring; Measures temperature variations within dam materials to detect stress and thermal effects.	Monitors temperature variations that affect material properties and structural behavior, helping to identify potential issues.	Distributed throughout the dam structure	Durable, versatile, advanced analysis. High initial cost, expertise required, sensitivity to conditions.
Geotechnical Monitoring	Earth pressure cell	[145,146,147]	Long-term and short-term monitoring; Records lateral earth pressure data to assess soil–structure interaction and foundation behavior.	Provides information on soil behavior and potential movement and aids in the assessment of dam stability and soil–structure interaction.	Installed in the dam foundation or surrounding soil.	Precise pressure monitoring, remote data access, versatile. Skilled installation, regular maintenance, high project costs.
	Piezometer	[132]	Long-term and short-term monitoring; Measures water pressure, seepage, and groundwater levels within and around the dam.	Monitors water pressure, seepage, groundwater levels, and potential leakage, crucial for dam safety and maintenance.	Placed within the dam or surrounding areas.	Reliable, cost-effective, easy automation. Freezing issues, clogging, unsuitable for artesian conditions.
Safety Monitoring	Vibration monitor	[132]	Long-term and short-term monitoring; Detects vibration levels and frequencies to assess structural integrity and identify potential issues.	Detects excessive vibrations indicating structural issues or risks, enabling proactive measures for stability and safety.	Distributed throughout the dam structure.	Comprehensive monitoring, durable, standard-compliant. Costly, technical expertise required, battery dependence.
	Geophone, also called jug or tortuga	[148,149]	Long-term and short-term monitoring; Monitors ground movement at the foundation or dam body to detect settlement or instability.	Foundation for settling, dam body for deformation, slide-prone areas for warning, seismic fault lines for earthquakes.	Could be installed on the dam foundation or dam body.	Sensitive, portable, cost-effective, efficient. Surface noise, depth limits, data complexity, environmental sensitivity.
	Strain gauge load cell	[132,150]	Long-term and short-term monitoring; Measures load or force acting on structural components for performance evaluation.	Measures actual loads on structural elements, aiding in the assessment of structural integrity and performance.	Installed at critical load-bearing components of the dam structure.	Accurate, dynamic, remote monitoring. Calibration needs, damage susceptibility, costly maintenance.
	** InSAR	[132]	Long-term monitoring; InSAR provides high-precision ground movement data for detection of subsidence, landslides, and structural shifts.	Provides detailed measurements of ground movement, subsidence, and landslides, helping to identify potential risks and their impacts on the dam structure.	Focusing on infrastructure in targeted, specific project areas. It focus on the dam structure and surrounding areas for stability analysis.	High precision, remote sensing, no ground setup. Costly, atmospheric disturbances, radar reflectivity dependence.

* The locations of sensors in dam structures or in geological surveying depend largely on the objectives of the monitoring or exploration project. The column outlines optimal locations or gives general guidelines for sensor placement. ** InSAR: Interferometric Synthetic Aperture Radar.

Diverse data types are vital in dam SHM to ensure structural integrity. Data types like guided wave data, vibration signals, and acoustic emissions [151] provide critical insights into dam conditions. Sophisticated sensors and numerical simulations * contribute to identifying and evaluating structural irregularities.

Table 5. Commonly used data types in SHM.

Data Type	Sensors and Tools	Methods and Techniques	Primary Applications	Remarks and Advantages	Challenges
Vibration signals	Sensors such as Acceleration Strain Displacement	Parametric and non-parametric methods Evaluating modal frequency, mass, damping, and stiffness	Monitoring structural changes Evaluating modal parameters	Continuous monitoring of structural integrity Utilizes FEM, public datasets, and SHM system data Real-time data acquisition	Sensitivity to environmental noise Requires extensive sensor network
Acoustic emissions	AE sensors	Sound wave analysis during structural changes	Fatigue crack localization Concrete crack detection	Used for fatigue crack localization and concrete crack detection Incorporates FE simulations	High background noise interference Requires expert interpretation
Images	Cameras Mobile phones Unmanned air vehicles Traffic cameras Online search	Image processing Machine learning (ML) Deep learning (DL)	Crack detection Bolt loosening Structural deformation	Suitable for crack detection, bolt loosening, vehicle recognition Employs public and synthetic datasets	Limited to surface damage Affected by lighting and environmental conditions
Guided waves	Piezoelectric transducer (PZT) sensors	Deployment of ultrasonic waves (Rayleigh, Lamb waves) FE simulations and experimental data	Internal damage identification Stress and strain monitoring	Accurate internal damage localization and assessment	Complexity in signal interpretation Equipment sensitivity
Others	Electromechanical impedance (EMI) sensors Ground-penetrating radar (GPR) sensors	EMI and GPR data acquisition techniques	Damage identification Cable tension prediction Material characterization	Applications in damage identification and cable tension prediction	Limited to specific types of damage Requires specialized equipment

outlines key data types, their sensors, methods, and remarks. These advancements enhance the monitoring efficiency, guiding maintenance and ensuring safety through comprehensive structural assessments.

2.3. Dam Monitoring Process

The structural health monitoring process is divided to four parts, which are presented schematically in Figure 5. It starts with planning and sensor integration, where sensors are placed for optimal data collection; then, data acquisition and real-time monitoring take place, with regular data collection and analysis for risk detection; this is followed by data management, which includes analysis and visualization for predictions; finally, evaluation and maintenance are conducted, where the performance is assessed and recommendations are made for maintenance. As shown in Figure 5 the supervisory control and data acquisition (SCADA) systems are essential for centralized monitoring and control in industrial processes, including dams and power plants, using real-time data from distributed sensors and devices.

2.4. Interdisciplinary Collaboration in SHM Systems for Dams

Improving dam monitoring systems requires the integration of engineering, data science, and environmental sciences [152]. Traditionally, engineering has focused on structural modeling, data science on processing and predictive analytics, and environmental sciences on climate and hydrological factors. An interdisciplinary approach, however, enables more comprehensive and adaptive SHM systems capable of addressing complex structural and environmental challenges. Engineering expertise facilitates structural analysis and vulnerability identification, while data science contributes real-time analytics, anomaly detection, and predictive maintenance using advanced algorithms [153]. Environmental sciences evaluate the effects of climate change, extreme weather, and ecosystem dynamics, offering context for long-term safety strategies. By combining these disciplines, SHM systems can become more resilient, enabling informed decision making that accounts for sedimentation, hydrological variability, and climatic risks [5]. This collaboration transforms monitoring from a reactive process into a proactive, intelligent system that supports dam safety and sustainability from construction to decommissioning, especially amid changing environmental conditions.

3. Fundamental Purposes of Dam Monitoring Data

Dam monitoring data ensure safety, indicate potential issues, and provide early warnings [13,22,154]. By monitoring parameters like water flows, structural stability, and seepage, systems generate data that help engineers to assess dam conditions [23,24,61,64]. One of the fundamental applications of these data is in risk assessment, which involves the systematic identification of potential failure modes, the estimation of their likelihood, and the evaluation of the consequences associated with these failures [155]. This process supports informed decision making and the development of mitigation strategies to reduce risks and enhance dam safety. Moreover, monitoring data aid in predicting aging and failures, facilitating timely maintenance [64,76]. They can also be used to optimize operational efficiency, guide water release, and support hydropower decisions. Dam monitoring ensures adherence to safety standards, regulatory compliance, and environmental assessments. Additionally, it plays a role in research, public awareness, and responsible water resource management, contributing to dam longevity and sustainability. The main purposes of monitoring are outlined below in Figure 6.

3.1. Monitoring Data for Dam Safety and Risk Management

Monitoring data form the backbone of safety evaluations in dam infrastructure. They supports a structured approach to identifying vulnerabilities, quantifying risks, and informing timely interventions. This process involves not only tracking structural and hydraulic parameters but also integrating them into broader risk management frameworks that address both technical performance and public safety [20,21,63].

3.1.1. Risk Assessment and Hazard Identification

Risk assessment in dam safety involves a multi-step process that begins with identifying potential sources of failure. As shown in Figure 7, this includes analyzing data from real-time monitoring systems, such as piezometers, inclinometers, strain gauges, and remote sensing platforms, to detect abnormal trends in seepage, displacement, and loading conditions [10,61,156,157,158]. These anomalies often serve as early indicators of deeper structural or geotechnical issues.

Hazard identification focuses on defining failure modes, including overtopping, internal erosion, foundation instability, and seismic-induced damage. Evaluating these scenarios requires a combination of empirical data, historical case studies, and predictive models [158,159,160]. Probabilistic methods such as Bayesian inference [160] and fault tree analysis allow engineers to estimate the likelihood of each failure mode under varying operational and environmental conditions.

3.1.2. Probability and Consequence Evaluation

The level of risk is typically expressed as a function of the probability of failure and the magnitude of the associated consequences:

$$Risk=ProbabilityofFailure\times ConsequenceofFailure$$

Probability estimates are refined through the continuous monitoring of structural and environmental variables [8,9,10,161,162,163,164,165,166,167,168,169]. Advanced machine learning techniques, including boosted regression trees and time-series analysis, are increasingly used to improve the forecasting accuracy [63,158]. These models offer the flexibility to incorporate both historical and real-time data, allowing for dynamic risk reassessment as conditions evolve. Consequence analysis considers the downstream impacts of potential dam failures. This includes modeling the extent of flooding, estimating the loss of life, evaluating property damage, and assessing environmental degradation [159]. Tools like inundation mapping and vulnerability indices help to quantify these outcomes, thereby informing mitigation priorities and emergency planning.

3.1.3. Emergency Actions and Response Protocols

Monitoring data also underpin the development and activation of emergency protocols. As illustrated in Figure 8, sensor networks continuously feed data into automated systems designed to detect critical anomalies [170,171]. When thresholds are exceeded, the system initiates a series of predefined actions:

• Health monitoring systems are active around the clock, measuring seismic activity, water levels, temperatures, and other relevant indicators [164,172];
• Once data anomalies are detected, they are classified by severity and type to determine whether further action is needed;
• If a significant anomaly is confirmed, the system issues alerts to dam operators, emergency services, and potentially affected communities;
• These alerts trigger established evacuation routes, resource mobilization, and coordination among local authorities [173,174].

This structure ensures that response times are minimized and that decisions are based on objective, data-driven assessments.

3.1.4. Adaptive Risk Management

Dam safety is not a static condition but a continuously evolving challenge influenced by aging infrastructure, changing climate patterns, and operational demands [11]. Accordingly, effective risk management depends on continuous feedback loops where new monitoring data are used to recalibrate models, update hazard scenarios, and refine emergency plans [11,63,79,159,175]. This adaptability supports a transition from reactive maintenance to preventive strategies and long-term resilience planning.

3.2. Operational Efficiency

Monitoring data help to optimize flood control, water release, and power generation [176,177,178], ensuring dam efficiency, adaptability to environmental conditions [14], and enhanced functionality. Key contributions to operational efficiency are detailed as follows. Firstly, the real-time monitoring of water levels, precipitation, and downstream conditions enables effective flood control by adjusting spillway release or retaining water, protecting infrastructure and communities [179]. An innovative early warning model [179] defines a flood alert index to optimize reservoir operations during typhoons. Secondly, accurate monitoring data on reservoir levels, inflows, and environmental factors can be used to optimize water release schedules, balancing downstream user needs and environmental sustainability [180]. Thirdly, continuous temperature, precipitation, and seismic monitoring contributes to adaptive dam stability techniques to withstand climate change and unanticipated events [181].

3.3. Environmental Impact Assessment (EIA)

Addressing the EIA of dam operations is vital for sustainability [182,183]. Monitoring data are used to evaluate ecological impacts, balance water use with ecosystem preservation [5,183], ensure regulatory compliance, and guide research [184,185]. GISs aid in landscape analysis and erosion control [186], as shown in the Three Gorges Project, minimizing environmental harm [187].

3.4. Hydropower Optimization

Hydropower generation, vital for dam operations, relies on detailed monitoring for optimal efficiency [188]. Tracking factors like water flows and turbine performance helps to align energy production with demands while preserving water resources [189]. This data-driven approach boosts efficiency, extends equipment lifespans, and supports sustainability [190]. Monitoring also ensures environmental and economic viability. Studies highlight dam monitoring’s role in hydropower optimization, including meta-heuristic methods like the harmony search algorithm [191], machine learning applications for river flow and reservoir forecasts [188], and high-fidelity models to improve generation while addressing environmental constraints [192]. Effective monitoring enhances hydropower production and sustainability by managing challenges like erosion and cavitation [190].

3.5. Research and Development (R&D)

Continuous monitoring in dam engineering provides important data for R&D [64], enhancing dam designs [193], materials science [172], and structural integrity. This involves evaluating durability, reinforcement, and techniques, fostering sustainable, resilient practices [185]. Insights improve materials and construction, ensuring that new projects benefit from past lessons. Monitoring supports the evolution of engineering methods and sustainability efforts in dam construction.

3.6. Maintenance Planning

Identifying patterns and concerns in concrete structures, spillways, and gates helps to plan dam maintenance and prioritize rehabilitation [33,34]. Proactive maintenance prevents major issues, improving safety, sustainability, and ecology. The “Reservoir App” shows how augmented reality and near-field communication improve operating efficiency [194,195]. Monitoring balances efficiency, environmental protection, and cost-effective hydroelectric maintenance. Repair expenses are reduced through fault identification, enabling real-time operational expense optimization and resource efficiency. Monitoring data help to allocate resources for sustainable dam operations that reduce risks and preserve safety and performance.

4. Methods Employed in Utilizing Dam Monitoring Data

Dam monitoring employs a variety of models to ensure structural safety and operational efficiency [196,197,198]. These include numerical, statistical, ML, and hybrid models, each contributing distinct advantages to a comprehensive DSHM framework. Figure 9 presents the classification of these models. Numerical models, such as those based on FEM, simulate long-term structural responses under varying conditions like reservoir filling and seismic activity [199]. Statistical models extract patterns and correlations from large datasets to support risk identification and informed decision making [61]. ML models have advanced DSHM by enabling real-time anomaly detection, predictive analytics, and adaptive learning from sensor data [122,124]. Hybrid models combine these techniques to improve accuracy and resilience, especially when data complexity limits single-method effectiveness [61,120].

Table 6. Overview of some pioneering works in the literature on models that enhance dam SHM systems, covering their methodologies, purposes, advantages, and limitations.

Model	Ref.	Methods/Techniques and Proposed Approach	Dam Type	Advantages and Limitations
Statistical Models	Wang et al. [200]	FEM, HST, and a new HHST model. Viscoelastic hysteresis and ambient cooling drive deformation, while hydration heat and valley contraction cause long-term structural changes.	Jinping arch dam	HHST model captures temperature drops and hysteretic–hydraulic effects. Offers insights into abnormal dam deformation. Concrete creep and temperature impacts need more study. Isolating temperature-induced consequences in complex fields is a major issue.
	Tatin et al. [201]	A statistical model using water temperature profiles to interpret dam displacements.	Arch dam	Increases physical representativeness for irreversible effects. Requires accurate environmental data and further exploration to balance model freedom and physical soundness.
	Tatin et al. [202]	Introduces HST-Grad model to account for thermal effects in concrete dam displacement measurements.	Gravity dam	HST-Grad refines thermal displacement, reducing dispersion. Incorporates mean temperature and gradient for better accuracy. Homogeneous temperature assumption limits accuracy. Ongoing work addresses non-homogeneous thermal loads.
	Léger et al. [203]	Introduces HTT statistical model; compared with HST model.	Schlegeis arch dam	HTT accommodates diverse thermal loads, enhancing flexibility; frequency-domain methods improve heat transfer handling. Assumes elastic behavior but needs validation against HST displacement.
Machine Learning Methods	Li et al. [23]	Includes 1D-CNNs, Bi-GRU, and transfer learning techniques. BO algorithm optimizes parameters of DL-based paradigm.	High-arch dam	Deep learning excels in accuracy; transfer learning enhances predictions across monitoring points, overcoming statistical model limitations. Relies on accurate environmental data, requiring balance between model freedom and soundness.
	Li et al. [204]	STL method, extra trees, and stacked LSTM models.	Multiple-arch concrete dam	Excels in separate prediction of seasonal, trend, and remainder components, showing superior accuracy. Hyperparameter optimization relies on trial-and-error, handling single monitoring points.
	Liu et al. [205]	LSTM-PCA and LSTM-MA models.	Lijiaxia arch dam	LSTM models outperform traditional models, ensuring precise long-term predictions. Their adaptability to abnormal time series improves reliability. Operational complexity and computational overhead must be considered.
	Ribeiro et al. [206]	RNN with LSTM, SARIMA, SARIMAX, and SARIMAX-NEURAL methods.	Itaipu dam’s buttress block	SARIMAX-NEURAL exhibits a 20–50% accuracy enhancement, capturing linear and non-linear dependencies. Numerical models rely on historical data, limiting applicability. Complex implementation and parameter tuning required.
	Ying Hua et al. [207]	BP-ANN with genetic algorithm (GA).	Arch dam	BP-ANN and GA offer a novel method for processing of monitoring data. May exhibit poor predictive ability with low-frequency monitoring data.
	De Granrut et al. [208]	HST linear model and ANN model.	French arch dam	ANN is efficient for analyzing non-linear phenomena in dam monitoring. Reliable interpretation of couplings and threshold effects. Complexity of ANN models requires careful tuning and quality input data for reliable results.
Numerical models	Oliveira et al. [209]	FEM for key components, automatic management, monitoring data analysis, and simulation software.	Cabril dam (Portugal)	Highlights SSHM benefits, crucial for informed decision making, maintenance, and lifetime management. Software lacks automatic damage detection; engineers manually analyze data for evaluation of dam deterioration. Future improvements planned.

Abbreviations: 1D-CNN: one-dimensional convolutional neural network; Bi-GRU: bidirectional gated recurrent unit; BO: Bayesian optimization; LSTM: long-short term memory; STL: seasonal trend decomposition based on loess; PCA: principal component analysis; MA: moving average; RNN: recurrent neural network; SARIMA: Auto-Regressive Integrated Model of Seasonal Moving Average; SARIMAX: SARIMA with Exogenous Variables; RF: random forest; BRT: boosted regression trees; NN: neural network; SVM: support vector machine; MARS: multivariate adaptive regression splines; BP-ANN: backpropagation algorithm with artificial neural network; GA: genetic algorithm; SSHM: structural health monitoring system.

reviews foundational studies, while Figure 9 provides an overview of the classification of DSHM models.

Table 7. Practical applications and comparative analysis of modeling methods in dam SHM.

Method	Practical SHM Applications	Application Scope	Research Orientation	Advantages	Limitations	Suggested Application Contexts
Statistical Models	Early anomaly detection; seepage and deformation trend analysis; validation of historical design assumptions	Applied to monitor dam deformation, seepage behavior, and crack development during construction and operation phases, especially for gravity and arch dams.	Basic data processing, analysis, and application tools	Simple to implement, robust, highly interpretable; well suited for engineers	Assumes linearity; sensitive to data quantity and quality; limited in handling complex non-linear behavior	Best for early-stage analysis, linear behavior trends, or when only limited historical data are available
Numerical Models	Structural stress–strain simulation; thermal analysis; failure mechanism modeling; risk assessment	Used to simulate structural and thermal behavior of dams across design, construction, and operation stages. Applicable to all dam types.	Advanced simulation and prediction	High fidelity in physical simulation, applicable to diverse dam conditions	High computational cost; require domain expertise and detailed structural data	Ideal for safety assessment, structural stress modeling, and failure analysis under known physical laws
ML Models	Anomaly detection; predictive maintenance; deformation forecasting; risk prediction	Suitable for operational monitoring of aging dams. Applicable across various dam types to detect anomalies and predict future behavior using large datasets.	Data-driven analysis and prediction	Capable of learning complex non-linear relationships; handle large datasets	Require large training datasets; potential for overfitting; interpretability and explainability are limited	Effective for real-time monitoring and complex pattern recognition when large data are available
Hybrid Models	Comprehensive risk analysis; multi-factor deformation forecasting; optimal intervention planning	Integrate various monitoring objectives (deformation, seepage, environmental factors) throughout all dam lifecycle stages, especially for complex dams.	Combination of data-driven and simulation methods	Combine physical realism (numerical) with data adaptiveness (ML); improved forecasting	Technically complex; high computational demands; demand expertise in both ML and numerical modeling	Best for complex dam environments with non-linear behavior and multi-source data; suitable for critical decision making

compares these methods, guiding selection based on goals, applicability, strengths, and limitations.

4.1. Practical Applications of Monitoring Methods in SHM Decision Making

In dam SHM, modeling techniques play a critical role in transforming sensor data into actionable insights for anomaly detection, condition assessment, and predictive maintenance. These approaches are broadly categorized into statistical, machine learning (ML), DL, and hybrid methods. Statistical models (e.g., MLR, ARIMA) are valued for their simplicity and interpretability, making them suitable for linear systems and limited datasets [210]. ML techniques, such as support vector machines and random forests, are more effective in capturing non-linear patterns and supporting short-term forecasting [122,124]. DL architectures, including LSTM, CNNs, and Temporal Fusion Transformers, handle multivariate, high-dimensional data, with strong predictive capabilities, but they demand greater computational resources. Hybrid models integrate signal decomposition (e.g., STL, VMD) with ML/DL to enhance robustness under seasonal, hydrostatic, and aging influences [44,45]. Beyond prediction, these models support critical decision making, such as identifying early warning signs, estimating failure risks, optimizing maintenance schedules, and verifying whether dam behavior aligns with design expectations under real-world operating conditions. Table 7 provides a summary of how these methods are applied in practice to support critical SHM decisions.

4.2. Comparison of Models Based on Performance, Input Needs, and Application Domains

4.2.1. Numerical Models in Dam Monitoring

Numerical models are essential in dam monitoring, applying deterministic approaches grounded in physical laws and mathematical formulations, such as FEM and system identification techniques [211,212,213]. These models discretize structures into finite elements governed by compatibility and equilibrium conditions [212,214,215], enabling simulations of irreversible deformations and comparisons between predicted and observed dam behavior [216,217,218]. They are particularly effective in analyzing responses during reservoir filling scenarios [219], as seen in the seismic evaluation of the Tsankov Kamak dam using FEM [220]. Parameter identification methods assist in estimating values like the equivalent elastic modulus for global structural behaviour analysis [78,221]. Various numerical models are employed in dam analysis, including FDM, DEM, FEM, and BEM [5,201]. FDM is widely used for seepage and thermal analysis in concrete dams [222,223,224]. DEM is suitable for rock fill dam evaluation, while FEM addresses complex structural and hydraulic issues. BEM offers efficient modeling in terms of flow and elasticity.

Table 8. Common numerical methods for dam monitoring and their advantages, limitations, and equations.

Method	Ref.	Equation	Advantages	Limitations	Remarks
FDM	[223]	${▽}^{2}h={\displaystyle \frac{{\partial }^{2}h}{\partial x^2}}+{\displaystyle \frac{{\partial }^{2}h}{\partial y^2}}=0$	FDM is relatively straightforward to implement. It is well suited for problems with regular geometries.	Handling irregular geometries can be challenging. It may not be as accurate as other methods for complex problems.	Laplace equation for steady-state seepage
DEM	[225,226]	$F_i{=m}_i{\displaystyle \frac{d^2{r}_i}{{dt}^2}}$ $M_i{=I}_i{\displaystyle \frac{d{\omega }_i}{dt}}$	Effective in modeling complex behaviors of granular materials and can simulate large deformations and failure processes.	Computationally intensive and requires detailed information about the material properties and particle interactions.	Newton’s second law
FEM	[227]	$M_S\ddot{U}+C_S\dot{U}+K_{S}U=F_g+F_p$	Highly versatile and accurate. Suitable for complex geometries and a wide range of materials.	Requires significant computational resources. Accuracy depends on the quality of the mesh.	Dam structure displacement
BEM	[227,228]	$=\underset{\partial \Omega F}{\overset{c\left(x\right)p\left(x\right)}{\int }}[q\left(y\right)G\left(x,y\right)-p\left(y\right)H(x,y){dS}_y$	Reduced computational domain leads to less computational effort. Effective for problems with infinite or semi-infinite domains.	Handling non-linear problems can be complex.	Laplace equation for potential flow problems

Here, $\mathrm{h}$ represents the hydraulic head; $\mathrm{x}$ represents the direction (horizontal) and $\mathrm{y}$ represents the direction (vertical); ${\mathrm{F}}_{\mathrm{i}}$ is the total force acting on particle i; r_i is the centroid position; m_i is the particle mass; M_i is the total moment; ω_i is the angular velocity; I_i is the moment of inertia; $\mathrm{M}\mathrm{S}$, $\mathrm{C}\mathrm{S}$, and $\mathrm{K}\mathrm{S}$ denote the classical mass, damping, and stiffness matrices. The vector $\mathrm{U}$ signifies nodal displacements in the dam’s finite element model. The dot symbol ($·$) indicates differentiation concerning time $\mathrm{t}$; $\mathrm{q}\left(\mathrm{y}\right)$ denotes the normal derivative of pressure $\mathrm{p}\left(\mathrm{y}\right)$ at a point y with coordinates $({\mathrm{y}}_1;{\mathrm{y}}_2)$ on the boundary. Here, $\mathrm{x}$ represents the source point, and $\mathrm{y}$ is the current point. The functions $\mathrm{G}\left(\mathrm{x},\mathrm{y}\right)$ and $\mathrm{H}\left(\mathrm{x},\mathrm{y}\right)$ serve as fundamental solutions for Laplace’s equation in this context.

outlines their advantages, limitations, and applications.

Recent numerical approaches integrate traditional methods with machine learning and real-time data analytics, enhancing dam behavior predictions. FEM and neural networks improve deformation forecasts, while real-time data boost model adaptability, safety, and efficiency.

4.2.2. Statistical Models in Dam Monitoring

Statistical models play a vital role in dam monitoring by providing insights into conditions and performance, which are essential for risk identification and maintenance guidance [5,229,230,231]. Complex, variable datasets, including those on water levels and structural stresses, are analyzed to identify early issues such as leaks or weaknesses [5,78]. The evolution of these models has progressed from fundamental linear methods to sophisticated systems that are adept in managing non-linear interactions and deriving insights from data for precise predictions [61]. Causal models connect environmental factors to dam responses, whereas non-causal models concentrate exclusively on dam behavior [5]. Techniques encompass parametric, nonparametric, and time-series analysis to gain insights into dynamic behavior [78]. Advanced models combine data from sensors, satellite imagery, and historical records to improve the predictive accuracy and guide sustainability initiatives.

Table 9. The most common statistical methods in dam monitoring, elucidating their respective advantages, limitations, and equations.

Method		Refs.	Equation	Advantages	Limitations	Remarks
Regression Models	HST	[5,201,232]	$\delta =a_0+{\delta }_1\left(h\right)+{\delta }_2\left(s\right)+{\delta }_3\left(t\right)+\epsilon$	Provides a comprehensive analysis of dam deformation factors. Capable of distinguishing between reversible and irreversible deformations. Useful for long-term monitoring and trend analysis.	May oversimplify complex interactions between different environmental factors. Assumptions in the model might not capture all nuances of dam behavior. Dependent on the accuracy and availability of environmental data.	Used to track reversible and irreversible deformations in concrete dams.
	HSTT	[5,233,234]	$\delta =a_0+{\delta }_1\left(h\right)+{\delta }_2\left(s\right)+{\delta }_3\left(t\right)+{\delta }_4\left(\Delta {\theta }_R\right)+\epsilon$	Accounts for thermal effects, offering a more complete understanding of dam behavior. Enhances the prediction of thermal deformation, particularly in response to climate variability.	Complexity increases with the addition of the thermal component. Requires detailed temperature data, which may not always be available. The lagged response may not be accurately captured in all scenarios.	Assesses impact of temperature variations on dam structure.
	HHST	[200,234,235]	$\delta =a_0+X\left({\delta }_{He}^{’}+{\delta }_{Hv}^{’}\right)+{\delta }_2\left(s\right)+{\delta }_3\left(t\right)+\epsilon$	Incorporates the hysteresis effect of hydrostatic loads, improving model realism. Suitable for analyzing dams with significant viscoelastic and hysteretic characteristics.	More complex to implement due to the additional hydraulic components. Requires extensive data for calibration and validation. May be computationally intensive.	Analyzes abnormal deformation behavior due to hydrostatic loads.
	MPN	[236,237]	$\delta ={\delta }_1\left(h,x,y,z\right)+{\delta }_2\left(s,x,y,z\right)+{\delta }_3\left(t,x,y,z\right)$	Provides a spatially comprehensive analysis, covering the entire dam. Enhances the understanding of deformation distribution across the dam.	Requires a large amount of spatial and deformation data. Increased complexity in model construction and interpretation. Higher computational requirements.	Used for spatial monitoring across the entire dam.
Time-Series Models	ARMA	[238,239]	${\delta }_t={\phi }_1{\delta }_{t-1}+{\phi }_2{\delta }_{t-2}+\cdots +{\phi }_p{\delta }_{t-p}+{\epsilon }_t-{\theta }_1{\epsilon }_{t-1}-{\theta }_2{\epsilon }_{t-2}-\cdots -{\theta }_q{\epsilon }_{t-q}$	Effective in analyzing time-dependent data like dam deformation. Can reveal underlying patterns and trends in dam monitoring data.	Model order determination can be complex and critical for accuracy. Not suitable for non-stationary data without modifications.	Predicts dam deformation over time.
	ARIMA	[238,240,241]	$\left(1-{\phi }_{1}B-{\phi }_2{B}^2-\cdots -{\phi }_p{B}^p\right){\nabla }^d{\delta }_t=\left(1-{\theta }_{1}B-{\theta }_2{B}^2-\cdots -{\theta }_q{B}^q\right){\epsilon }_t$	Capable of handling non-stationary data, making it versatile for various dam monitoring scenarios. Useful in forecasting and predicting future dam behavior.	Determining the appropriate combination of parameters (p, d, q) can be challenging. Requires expertise in time-series analysis for effective implementation.	Utilized in forecasting dam behavior under various conditions.

Here, δ is the recorded deformation; a₀ is a constant value; δ₁(h) is the hydrostatic component; δ₂(s) is the temperature component; δ₃(t) is the aging component; ε is the error of the residual; ∆θ_R is a ∆θ lagged variable; ${\delta }_{Hv}^{\prime }$ is the hysteretic hydraulic component; ${\delta }_{He}^{\prime }$ is the instantaneous hydraulic component calculated and fitted by the FEM model; X is the adjustment coefficient of the total hydraulic component; ($\mathrm{x},\mathrm{y},\mathrm{z}$) are spatial coordinates; δ_t represents measured data on dam deformation at the moment t; φ₁, φ₂,…, φ_p are autoregressive coefficients; θ₁, θ₂,…, θ_q are moving average coefficients; ε_t is a stationary disturbance with mean 0 and variance ${\sigma }_{\epsilon }^2$; B is the delay operator and satisfies Bⁿδ_t = δ_t-i; $\nabla$ is the difference operator; and ${\nabla }^d$ = (1 − B)^d.

summarizes the statistical models, outlining their advantages, limitations, and applications in previous studies, offering essential tools for engineers to ensure dam safety and integrity.

4.2.3. ML Models in Dam Monitoring

Recently, machine learning has significantly improved the structural health monitoring of dams through advanced, data-driven methodologies that enhance safety and performance assessments [42,65,100,117,122,124]. Cheng et al. [242] combined machine learning with traditional analysis to enhance seepage monitoring in rock fill dams, whereas Salazar et al. [243] discussed the strengths and limitations of machine learning in anomaly detection. Machine learning facilitates the advanced analysis of dam behavior under diverse stressors, exceeding traditional methods in both accuracy and reliability [85]. Initially used for sensor data analysis, ML has evolved to include satellite imagery, geotechnical measurements, and environmental data, driven by advancements in sensor technology, computational power, and algorithms [85,242]. Machine learning models utilized in dam monitoring can be classified into three categories [51]: (i) supervised learning employs labeled data for regression and classification to identify deviations in structural behavior, facilitating early interventions; (ii) unsupervised learning analyzes unlabeled data to uncover patterns, enabling clustering and dimensionality reduction, which streamlines analysis and improves data interpretation; (iii) semi-supervised learning integrates both labeled and unlabeled data, optimizing the use of limited labeled samples alongside a more extensive dataset to enhance accuracy.

Table 10. The most common ML methods in dam monitoring, elucidating their respective advantages and limitations.

Method *		Refs.	Advantages	Limitations
Supervised Learning Techniques	DT	[244,245]	Simple to understand and interpret; handles both numerical and categorical data	Can create overly complex trees that do not generalize well (overfitting); sensitive to noisy data
	SVM	[246,247,248,249]	Effective in high-dimensional spaces; works well with a clear margin of separation	Requires a good kernel choice; not suitable for large datasets; difficult to interpret
	EM	[250]	Combines predictions from multiple models to improve accuracy; reduces overfitting	Can be computationally expensive; model interpretation can be challenging
	BM	[124]	Incorporates prior knowledge; handles various types of data	Based on assumptions that may not hold in all scenarios; complex calculations
	NN& BP	[251,252]	Can model complex non-linear relationships; highly flexible	Requires a large amount of data; prone to overfitting; computationally intensive
Unsupervised Learning Techniques	HMM	[253,254]	Models time-series data well; handles variable-length sequences	Assumes independence of features; can be computationally expensive for large states
	k-NN	[255,256]	Simple and intuitive; no training phase	Computationally intensive during testing; sensitive to irrelevant features and scale of data
	GMM	[257,258]	Offers flexible, complex distribution modeling with soft clustering and density estimation, ideal for real-valued data analysis	Sensitive to initialization, assumes Gaussian distributions, can overfit, computationally complex and difficult to interpret in high dimensions
	NN	[124]	Learns feature representations; can handle complex structures in data	Difficult to interpret; requires large amount of data
S-SLM	S-SLM	[124]	Utilizes both labeled and unlabeled data; useful when labeled data are scarce	Dependent on quality of unlabeled data; can propagate errors if unlabeled data are not representative

* These algorithms each have their unique strengths and weaknesses, making them more or less suitable for specific types of data and analysis tasks in dam monitoring. The choice of algorithm depends on the specific requirements of the monitoring task, such as the size and nature of the dataset, the desired outcome (e.g., prediction, classification, clustering), and the computational resources available.

presents an overview of the machine learning methods, outlining their advantages and limitations, thereby providing a comprehensive perspective on their applications in dam safety monitoring.

The application of ML in dam SHM represents a transformative approach to traditional monitoring methods [124]. With ML, SHM systems transition from static, rule-based models to dynamic systems capable of continuous learning and adaptation [259]. This shift allows for the processing and analysis of vast datasets from sensors and monitoring devices, extracting valuable patterns and insights that are essential for dam maintenance and safety. Furthermore, ML algorithms in SHM enhance predictive analytics, enabling the early detection of potential structural issues, and facilitate rapid anomaly detection [122,260,261,262]. These features improve dam monitoring, maintenance, and management, making SHM systems more robust, efficient, and responsive. However, applying ML in this industry is difficult. Overfitting in artificial neural networks and parameter tuning in support vector machines can reduce their effectiveness [263]. ML algorithms can improve SHM systems, exhibiting future potential in civil engineering and infrastructure management despite these challenges.

4.2.4. Hybrid Models

The core principle behind hybrid models is to combine the strengths of diverse methodologies, thereby addressing the limitations inherent in standalone techniques. Traditional monitoring typically relies on physics-based simulations, such as FEM for structural behavior, and statistical analyses, including regression models for time-series data. While these methods provide valuable insights into physical phenomena and long-term trends, they often struggle with the complex, non-linear, and dynamic relationships present in real-time dam monitoring data [61]. Additionally, they can be computationally expensive and demand extensive estimations of material parameters, which can be challenging to achieve [264]. ML-FEM hybrid models improve the accuracy by correcting FEM predictions, estimating parameters, analyzing feature importance, and serving as surrogates. This approach reduces the computational costs and enhances the understanding of dam behavior by bridging the gap between theoretical models and real-world data. This is demonstrated in case studies like the Karun-I Dam [265], where an FEM-DNN-CatBoost hybrid model achieved an R² of 0.9763 and identified the lake level as the most significant factor (74% importance) influencing displacement. Similarly, at the Fei Tsui Dam [120], an adaptive time-dependent evolutionary least squares SVM model, as part of a hybrid AI approach, achieved an R² of 0.993 for dam body displacement prediction, highlighting its efficiency for early and accurate warnings.

Traditional statistical models, such as ARIMA and seasonal trend decomposition using loess (STL), are well established in forecasting short-term trends and capturing linear relationships in data [266]. However, their primary limitation lies in their inability to capture complex non-linear relationships and dynamic processes, which are often influenced by multiple environmental and operational factors. ML-Stat hybrid models overcome these limitations by combining the strengths of both methodologies, often following an “error correction” paradigm where statistical models handle linear components and ML models address non-linear residuals. This strategy, alongside feature extraction, the optimization of statistical models by ML, and parallel ensemble methods, leads to comprehensive modeling capabilities, improved forecasting, and enhanced interpretability [266]. For example, a hybrid model for the Jinping-I Super-High Arch Dam [267], which reconstructed aging and temperature components using a Burgers model and kernel principal component analysis (KPCA), achieved significantly higher accuracy (MSE of 0.3404, R² of 0.9902) compared to the traditional HST model (MSE of 2.2055, R² of 0.9898).

DL, particularly LSTM networks, has emerged as a highly effective component in hybrid models for dam monitoring, especially due to its inherent suitability for sequential and time-series data [268,269]. Dam behavior, such as deformation, seepage, and stress patterns, is fundamentally time-dependent, exhibiting long-term dependencies influenced by factors like the water level, temperature, and aging. LSTM networks, as a specialized type of RNN, are designed to overcome the vanishing and exploding gradient problems common in traditional RNNs, allowing them to effectively model long-term dependencies and capture dynamic trends in sequential data. LSTM networks are integrated into hybrid frameworks in various ways, including coupling with physics-based models like FEM, statistical models, dual-stage deep learning approaches, and advanced DL architectures like KANs [269].

4.3. Technical Challenges and Research Gaps

The integration of advanced modeling techniques into SHM systems for dams holds immense potential to enhance the safety, efficiency, and resilience of these critical infrastructures [270]. In this section, we identify and discuss the key technical challenges and research gaps that need to be addressed for SHM systems to reach their full potential, while providing insights into areas where future research could contribute to overcoming these limitations.

4.3.1. Sensitivity Analysis and Calibration Gaps

The accurate calibration of models is essential in ensuring that SHM systems provide reliable and realistic predictions of dam behavior [210]. Sensitivity analysis plays a crucial role in determining how changes in various parameters, such as material properties, environmental conditions, and operational factors, influence model outputs [271]. However, many existing models lack robust sensitivity analysis techniques, particularly when dealing with complex, high-dimensional datasets from large-scale sensor networks [272]. In traditional dam monitoring, methods such as FEM are widely used, but these methods often require accurate estimates of material properties and boundary conditions, which can be difficult to obtain in real time. Furthermore, there is a lack of standardized procedures for sensitivity analysis that can be applied across different types of dams, sensor configurations, and environmental conditions [273]. Addressing these calibration gaps will require the development of more sophisticated sensitivity analysis tools that integrate with real-time data streams. These tools must be capable of identifying the most influential parameters for each specific monitoring scenario, thus improving the reliability of SHM predictions in dynamic operational conditions [272].

4.3.2. Uncertainty Quantification (UQ)

UQ is another significant challenge in SHM systems, as it impacts the reliability and accuracy of predictions. Dam monitoring involves dealing with uncertainties in input data, including sensor readings, environmental conditions, and operational parameters [271]. Existing methods of UQ, such as Monte Carlo simulations or Bayesian inference, can be computationally expensive and impractical for real-time applications [272]. Moreover, the integration of real-time sensor data into models introduces additional uncertainties, particularly when sensors experience errors, fail, or provide incomplete data [271]. This uncertainty can have a significant impact on the performance of SHM models, leading to inaccurate predictions and, ultimately, safety risks. Therefore, there is a growing need for the development of more efficient UQ methods that can quantify the uncertainties inherent in real-time SHM systems without imposing excessive computational demands. New techniques, such as hybrid models that combine ML algorithms with traditional physical models, may offer a promising solution for efficient uncertainty quantification in SHM applications [272].

4.3.3. Model Convergence, Fairness, and Bias Evaluation

Combining traditional methods like FEM with data-driven techniques such as ML presents challenges in terms of model convergence [273]. In particular, when using ML algorithms to refine the predictions of traditional models, ensuring that these models converge to a solution becomes crucial. Without proper convergence, predictions can be unreliable, and safety assessments can be compromised. Additionally, the use of AI-based models introduces concerns regarding fairness and bias [274]. Model fairness ensures that predictions are not disproportionately affected by certain variables, and model bias evaluation ensures that decisions made by the system do not favor certain dam types, geographical regions, or sensor technologies [275]. For example, if a model is trained on data that are predominantly from concrete dams, its predictions may not be as reliable for embankment dams. This type of bias could lead to unfair safety assessments, which is particularly problematic in safety-critical applications like dam monitoring [274]. Future research must address model convergence, fairness, and bias to improve the reliability and accountability of SHM systems, ensuring that predictions are both accurate and equitable across different dam types and monitoring configurations.

4.3.4. Dynamic Database Updating and Real-Time Integration

Real-time integration allows for the timely detection of structural issues, enabling prompt interventions. However, the sheer volume of data generated by modern sensor networks poses significant challenges for real-time data integration and dynamic database management [272]. Existing SHM systems often struggle to process high-frequency sensor data without causing delays in decision making or overwhelming computational resources. Furthermore, sensor failures and data inconsistencies can lead to inaccurate predictions if not addressed quickly [272]. There is a pressing need for dynamic databases that can efficiently handle real-time data streams, perform necessary data validation, and update the models accordingly. Moreover, the models should be able to adjust to changes in environmental conditions or operational parameters without requiring manual recalibration. Developing such systems will require advancements in both data management and model integration techniques to ensure that SHM systems remain responsive and accurate in real-world operational scenarios [276].

4.3.5. Future Challenges in AI + SHM Integration

The integration of AI with SHM systems holds great promise in improving the accuracy, adaptability, and efficiency of dam monitoring [270]. However, the adoption of AI in this domain faces several challenges, particularly in terms of model interpretability, data requirements, and computational demands. One of the most significant hurdles is the “black box” nature of many AI models, especially deep learning models like LSTM networks [275]. While these models often provide highly accurate predictions, their lack of transparency makes it difficult for engineers to understand the reasoning behind the model’s outputs [275]. This is a critical issue in safety-critical infrastructure, where engineers must be able to trust and understand the decisions made by the system. Additionally, AI models require large datasets for training, which may not always be available, especially in the case of newly constructed dams or those located in remote areas [275]. Therefore, developing AI systems that are both interpretable and capable of functioning with limited data is essential for their successful integration into SHM systems [277]. Future research should focus on improving the transparency and explainability of AI models, as well as enhancing their ability to work with smaller or incomplete datasets, to make them more applicable and reliable in real-world dam safety applications.

5. Summary and Prospects

The history, technologies, and trends of DSHM are critically examined in this research. Examining SHM system components, methods, and applications exposes research gaps, notably in merging monitoring strategies with developing technologies. This review advances dam safety theory and practice, guiding future innovations in monitoring systems, data utilization, and system optimization for improved dam safety and reliability.

The principal novel insights and innovative contributions of this review to the field are as follows.

(a)

This paper introduces a structured framework connecting monitoring methodologies (numerical, statistical, and ML) to operational and safety goals. While it does not include direct forecasting, it identifies how time-dependent models such as ARIMA and LSTM can be applied to dam monitoring datasets to improve trend analysis and anomaly prediction. The review emphasizes the importance of incorporating such time-based methods in future DSHM systems to transition from reactive to predictive monitoring.

(b)

The paper highlights the revolutionary potential of integrating smart materials with IoT technology in DSHM systems. This integration facilitates the creation of real-time, autonomous monitoring systems that provide improved self-diagnostic functionalities. These solutions facilitate expedited decision making and proactive maintenance, presenting novel opportunities to enhance the efficiency, dependability, and safety of dam infrastructure. This amalgamation of technology signifies a notable divergence from traditional monitoring methodologies.

(c)

This study emphasizes the need for collaboration between engineering, data science, and environmental sciences to address dam monitoring challenges, combining diverse abilities to create more resilient and versatile SHM systems that evaluate structural integrity and environmental considerations for a holistic dam safety approach.

(d)

This research provides a thorough evaluation of dam monitoring models, including numerical, statistical, and ML methods. This work categorizes such models and evaluates their strengths and shortcomings.

(e)

This analysis emphasizes the need to adapt SHM systems to address climate change dangers, such as extreme weather and chronic environmental changes. The combination of predictive climate models and adaptive sensor technologies can improve dams’ resistance to shifting climates.

Future Research Directions

Building upon the gaps identified in the literature, the following specific research directions are proposed to advance DSHM.

(a)

The current literature lacks a framework linking dam monitoring methodologies, purposes, and analyses. Future research should create a framework that integrates these factors and provides clear instructions for the selection of monitoring systems depending on safety, operational, or environmental goals. This methodology will optimize monitoring procedures for their intended outcomes, serving to improve dam safety management and DSHM system decision making.

(b)

Future studies must examine how SHM systems might adapt to climate change’s growing effects. Integrating predictive climate models with monitoring systems helps researchers to analyze dams’ long-term resistance to harsh weather and environmental changes. This research will assist SHM systems in anticipating and minimizing climate variability hazards, enhancing dams’ environmental stress tolerance.

(c)

Given the complexity and scale of modern dam projects, existing sensor technologies often face limitations in monitoring all aspects of dam performance, especially in remote or hazardous areas. In order to address these problems, future research should integrate autonomous monitoring equipment like drones and robotics. These devices can collect data in high-risk regions for human inspectors. Furthermore, a comprehensive DSHM model should use real-time, high-resolution data to increase the monitoring accuracy and decision making. These improvements would improve inspections and increase the data available for predictive maintenance and risk assessment, improving dam infrastructure safety and longevity.

(d)

Future research should implement time-series forecasting models, such as LSTM and ARIMA, to analyze temporal patterns in deformation, vibration, and seepage. Integrating such techniques would allow DSHM systems to detect trends and anticipate structural anomalies before they pose serious risks.

(e)

Due to the structural intricacy of dams, DHM has garnered significantly less attention than the more extensively researched domains of bridge and building monitoring. Consequently, a substantial research gap exists, underscoring the necessity of the creation of advanced algorithms that can evaluate damage and forecast the remaining useful life of a dam. A heightened emphasis on this domain will be necessary to improve dam safety and ensure their sustained reliability.