Application of Acoustic Emission Technique in Landslide Monitoring and Early Warning: A Review

Abstract

Landslides present a significant global hazard, resulting in substantial socioeconomic losses and casualties each year. Traditional monitoring approaches, such as geodetic, geotechnical, and geophysical methods, have limitations in providing early warning capabilities due to their inability to detect precursory subsurface deformations. In contrast, the acoustic emission (AE) technique emerges as a promising alternative, capable of capturing the elastic wave signals generated by stress-induced deformation and micro-damage within soil and rock masses during the early stages of slope instability. This paper provides a comprehensive review of the fundamental principles, instrumentation, and field applications of the AE method for landslide monitoring and early warning. Comparative analyses demonstrate that AE outperforms conventional techniques, with laboratory studies establishing clear linear relationships between cumulative AE event rates and slope displacement velocities. These relationships have enabled the classification of stability conditions into “essentially stable”, “marginally stable”, “unstable”, and “rapidly deforming” categories with high accuracy. Field implementations using embedded waveguides have successfully monitored active landslides, with AE event rates linearly correlating with real-time displacement measurements. Furthermore, the integration of AE with other techniques, such as synthetic aperture radar (SAR) and pore pressure monitoring, has enhanced the comprehensive characterization of subsurface failure mechanisms. Despite the challenges posed by high attenuation in geological materials, ongoing advancements in sensor technologies, data acquisition systems, and signal processing techniques are addressing these limitations, paving the way for the widespread adoption of AE-based early warning systems. This review highlights the significant potential of the AE technique in revolutionizing landslide monitoring and forecasting capabilities to mitigate the devastating impacts of these natural disasters.

1. Introduction

Landslides are a major global hazard, causing significant socioeconomic losses and over 50,000 fatalities worldwide in the past decade [1]. These destructive events occur in nearly every country, triggered by factors such as heavy rainfall, earthquakes, erosion, deforestation, and human activities like urbanization. Mountainous regions, characterized by steep terrain and heavy precipitation during the wet season, face a notably higher frequency and risk of landslides [2]. For instance, the southwestern region of China, with its proximity to active seismic zones, experiences a higher prevalence of slope failures (Figure 1). These events endanger human lives and the environment by posing threats to engineering structures, disrupting critical infrastructure, causing structural damage, and contaminating water sources.

Given the complex interplay of factors that contribute to landslides, effective monitoring and early warning systems are essential for mitigating their devastating impacts. The causes of landslides are multifaceted, and involve a combination of short-term triggering factors and underlying predisposing conditions [3,4]. Research has identified a combination of short-term triggering factors and underlying predisposing conditions that contribute to the occurrence of landslides [5]. Major triggering factors include heavy or prolonged rainfall [6], rapid snowmelt [7], changes in groundwater level [8], and earthquakes [9]. Intense rainfall plays a critical role in triggering landslides by weakening the shear strength of soil and rocks through two primary mechanisms. Firstly, it increases the water content in soils and pore water pressure, leading to a decrease in effective stress [10]. This reduction in effective stress makes the soil and rocks more susceptible to failure. Secondly, heavy rainfall adds mass to the already unstable slope material, creating an additional driving force for landslides [11]. The combined effect of increased water content and added mass significantly increases the likelihood of slope failure [12]. In addition to rainfall, rapid snowmelt can also contribute to landslides by saturating the soil and increasing pore water pressure [13]. Changes in groundwater levels can further exacerbate landslide risk by altering the stability of slopes. Earthquakes, with their ground-shaking effects, can directly trigger landslides by destabilizing already vulnerable slopes. To address these challenges, the implementation of advanced monitoring techniques, such as acoustic emission (AE) monitoring, becomes increasingly necessary.

Landslides often develop gradually, characterized by incremental deformations that culminate in sudden failures [14]. This gradual progression emphasizes the importance of continuous monitoring technologies to detect early signs of destabilization and provide timely warnings [15]. However, the geological heterogeneity of terrain complicates the identification of potential failure zones and the accurate quantification of slope movements. Various surface and subsurface monitoring techniques have been employed to investigate landslide characteristics [16]. Geodetic methods, such as synthetic aperture radar (SAR) [17], aerial/satellite imagery [18], and terrestrial laser scanning (TLS) [19], provide valuable spatial data but are limited in capturing subsurface information. In contrast, geotechnical instrumentation, including inclinometers [20], wire extensometers [21], and time domain reflectometry (TDR) [22], allow for the direct measurement of slope displacements but are restricted to specific monitored areas. Geophysical techniques, including electrical resistivity tomography (ERT) and ground penetrating radar (GPR), offer insights into subsurface structures but require meticulous data processing [23,24]. Conventional measurement methods often detect landslides only after failure has occurred, limiting timely evacuation efforts [25,26]. Therefore, there is a pressing need for monitoring approaches that can continuously capture early warning signals preceding slope collapses.

In this context, acoustic emission (AE) technology emerges as a promising monitoring method. AE captures elastic wave signals generated by stress-induced deformation within materials, offering a novel perspective for landslide monitoring. It can detect micro-damage within rock and soil masses during the early stages of landslide incubation, addressing the limitations of traditional monitoring methods. AE involves the generation of stress wave emissions resulting from irreversible changes in material microstructures under stress, such as friction, cracking, fracturing and melting/crystallization of rocks, resulting in the emission of stress waves [27,28]. Unlike active ultrasonic techniques, AE employs a passive monitoring approach by detecting naturally emitted elastic waves from deforming structures [29]. This enables the faster monitoring of landslides compared to traditional methods. Figure 2 illustrates the schematic diagram of the landslide process and AE monitoring. When integrated with transient signal recording and analysis capabilities, AE monitoring shows potential as an early warning indicator for impending failures in geo-materials [30]. Recent field deployments of AE systems have demonstrated a clear linear relationship between cumulative displacement and AE data [31]. However, challenges remain in capturing signals in highly attenuative soil conditions and establishing precise correlations between AE parameters and deformation states [32].

This paper aims to offer a thorough review of the latest advancements in AE-based landslide monitoring techniques. After introducing the topic, the paper compares different methods used for monitoring landslides. It then delves into the basic principles and equipment used in the AE technique. The paper also examines how AE testing is applied in studying the stability of granular and rock slopes in both laboratory and field settings. Lastly, it discusses ongoing challenges and recent progress, and provides insights into potential future developments in the field.

2. Current Landslide Monitoring Methods

2.1. Geodetic Methods

Geodetic methods are surface movement monitoring techniques that can be applied over extensive areas, making them particularly useful for landslide investigations. These methods offer the advantage of being able to map large zones and have the potential for satellite-based applications [33]. Three prevalent geodetic techniques include TLS, global positioning systems (GPS), and SAR.

TLS is an active remote sensing method that captures surface topographic data across diverse field conditions [34]. By generating point clouds that represent the geometry of the entire scene, TLS enables the differentiation of terrain changes over time. However, the lack of standardized metrics for precision and accuracy can limit the reliability of the results [35]. Additionally, factors such as incidence angles, surface reflectivity, and ambient light variations can significantly influence scanning quality [36]. Despite these limitations, TLS has proven effective for mapping landslide-induced terrain modifications across various scales, from local to regional levels [37,38].

GPS provides spatial point positioning by triangulating signals from orbiting satellites. This technique measures displacements to assess slope stability by periodically determining the coordinates of reference markers placed within the hazard zone [39,40]. Key strengths of GPS include accuracy levels down to the sub-centimeter range and the capability for autonomous data logging [41]. However, the requirement for an unobstructed view of at least four satellites poses challenges in areas with dense vegetation or deep valleys, which can limit its applicability [42].

SAR functions as an active microwave remote sensing tool that captures terrain echoes in terms of sensor-target distances and azimuth directions during platform flight [43]. This non-contact imaging radar operates effectively under all weather conditions and at any time of day, making it ideal for both short- and long-term slope monitoring without the need for on-site instrumentation [44]. To address atmospheric phase screen effects encountered during monitoring in steep mountainous regions, Karunathilake and Sato [45] proposed a semi-empirical model. However, environmental contaminants, such as atmospheric heterogeneities, can introduce additional noise into the signal [46]. This complicates data processing and interpretation, necessitating advanced techniques to enhance the reliability of SAR data for accurate slope monitoring [47].

In summary, while geodetic methods are effective for mapping landslide shapes and identifying surface movements, they are not suitable for the continuous, real-time monitoring required for effective early warning systems. In complex landslides where failures extend deep underground, these methods provide incomplete information. Furthermore, the high costs and operational requirements associated with establishing geodetic networks in remote or mountainous areas present additional challenges. These methods are also limited in their ability to detect small deformations that may indicate potential slope issues before they escalate, as shown in

Table 1. Summary of geodetic methods.

Geodetic Methods	Advantages	Disadvantages
TLS	Active remote sensing; adaptable to diverse field conditions; capable of generating point clouds to identify terrain changes; applicable to mapping landslides at various scales.	Lack of standardized precision metrics; scanning quality affected by incidence angles, reflectivity, and ambient light.
GPS	Provides triangulation positioning via satellite signals; measures displacements to assess stability; offers sub-centimeter-level precision; capable of autonomous data logging.	Requires unobstructed observation of at least four satellites; limited applicability in areas with dense vegetation and deep valleys.
SAR	Active microwave remote sensing; operational in all weather conditions and at all times; does not require on-site installation; suitable for both short-term and long-term monitoring; includes models to mitigate extreme interference.	Affected by environmental pollutants; signal noise complicates data interpretation; data processing is complex.

.

2.2. Geotechnical Methods

Geotechnical methods utilize in situ instrumentation to conduct targeted monitoring in potential failure zones identified through prior investigations. Key techniques include TDR [48], fiber optic sensing (FOS) [49], inclinometers [50], and settlement gauges [51].

TDR is employed for monitoring landslides by sending electromagnetic pulses through coaxial cables and analyzing reflected signals to detect changes caused by underground movements [52]. This technology tracks landslide activity by measuring moisture fluctuations that indicate unstable conditions. The effectiveness of TDR lies in its ability to transmit and reflect electromagnetic pulses. Over the years, TDR has evolved, with improved installation methods, data interpretation techniques, and simpler data processing approaches [53]. Additionally, it has been integrated with other monitoring technologies to enhance the accuracy and early warning capabilities of landslide monitoring. However, TDR’s spatial coverage is limited by the depths and lengths of the installed cables [54].

FOS technology utilizes light signals transmitted through optical fibers to measure varying axial strains resulting from soil deformation [55]. This advanced sensing method allows for continuous strain profiling along the entire length of the sensor, effectively overcoming the coverage limitations of traditional TDR technology. By strategically placing FOS in landslide-prone areas, the real-time monitoring of soil movement and strain can be achieved [56]. These sensors detect physical changes along the optical fiber line through phenomena such as Rayleigh, Brillouin, and Raman scattering, which alter the intensity, phase, and frequency of the scattered light, providing crucial data on temperature and strain [57]. Recent studies have demonstrated that distributed FOS technology is highly effective for landslide monitoring due to its exceptional spatial and temporal resolution, offering nearly instantaneous measurement results. Moreover, FOS is increasingly utilized in long-term monitoring and early warning systems due to its lightweight design, environmental sustainability, resistance to electromagnetic interference, cost-effectiveness, and rapid data transmission capabilities [58]. Nevertheless, operational challenges related to moisture ingress and exposure to sunlight remain significant concerns that need to be addressed [59].

Inclinometers are installed within vertical casing pipes and are used to measure lateral displacements by tracking changes in the inclination of the casing [60]. However, their effectiveness is limited by the difficulty of installation in pre-existing boreholes, restricting their application to known sites. Conversely, settlement gauges are utilized to quantify vertical settlement levels, providing insights into areas experiencing excessive compression [61]. While both inclinometers and settlement gauges are useful for detecting vertical movements, they are unable to monitor internal shifts occurring beneath the ground surface without causing visible displacements.

The complementarity of these techniques is vital, as no single instrument is adequate for comprehensive monitoring [62]. For example, an integrated network combining GNSS, extensometers, drone photogrammetry, and hydro-meteorological sensors was employed to synergistically investigate the Urbas landslide in northwestern Slovenia [63]. Joint interpretation of FOS and TDR data also enabled differentiation between soil suction and consolidation effects [64].

A significant limitation of geotechnical methods is their inability to capture pre-failure behaviors, as these instruments typically detect displacements only after a triggering event has occurred [65]. Furthermore, the requirement for direct installation in the subsurface may not always be feasible in hazardous zones where failures are incipient or developing. High costs associated with geotechnical networks also render large-scale implementation impractical, particularly when continuous monitoring is essential, such as during heavy rainfall events.

Table 2. Summary of geotechnical methods.

Geotechnical Methods	Advantages	Disadvantages
TDR	Monitors via electromagnetic pulses; tracks moisture fluctuations; mature technology; integrable with other systems; reliable for long-term monitoring.	Limited spatial coverage based on cable length and depth; installation can be challenging in rugged terrains.
FOS	Measures strain using optical signals; allows continuous strain profile analysis; enables real-time monitoring; offers high spatial and temporal resolution; widely applied due to multiple advantages.	Vulnerable to moisture ingress and damage from UV exposure; initial setup costs can be high.
Inclinometers	Effective for measuring lateral displacements.	Challenging installation; limited application; unable to monitor internal changes underground.
Settlement Gauges	Effectively quantifies vertical settlement; straightforward installation and operation.	Limited to measuring vertical movements; unable to monitor lateral or internal changes.

summarizes the advantages and disadvantages of various geotechnical methods for landslide monitoring.

2.3. Geophysical Methods

Geophysical techniques leverage physical field properties to characterize subsurface conditions in a non-invasive manner, offering significant advantages for landslide investigations [66]. Common applications of these methods include ERT, self-potential (SP), GPR, and downhole seismic techniques.

ERT utilizes variations in electrical resistivity among geological materials to map subsurface structures and properties related to slope instability processes [67]. Resistivity distributions are inferred from surface electrode measurements, facilitated by multichannel systems and inversion modeling. Low resistivity anomalies often indicate developing shear zones due to increased fracturing and pore fluid saturation [68]. However, resistivity values can be influenced by various factors, including temperature, pressure, and the saturation levels of pore fluids. Additionally, the presence of metallic minerals or conductive fluids can adversely affect resistivity measurements.

SP techniques map natural subsurface potential fields generated by fluid and mineral interactions or thermal and electrokinetic coupling associated with slope deformation [69]. This method has been employed to identify preferred directions of landslide movement and areas of seepage. However, temperature variations and electromagnetic field influences can introduce contaminants, complicating the interpretation of SP responses related solely to hydrological factors.

Downhole seismic measurements involve profiling compression and shear wave velocities through boreholes, allowing for a direct assessment of elastic modulus variations within unstable rock and soil masses at different depths [70]. Nonetheless, the effectiveness of this method diminishes with increasing depth due to amplitude attenuation in fractured or weathered zones, where fractures can facilitate weakening. Furthermore, limited access to potential slip zones can restrict the applicability of this methodology.

GPR transmits short-pulse electromagnetic radar signals into the ground to image subsurface features through reflected wave amplitude maps [71]. GPR offers rapid, high-resolution stratigraphic profiling; however, signal penetration can be significantly impaired by water and clay content in the subsurface. The accurate determination of velocity is essential for converting signal data to depth scales, but this is not always feasible under all conditions.

Integrating geophysical techniques with geotechnical and geodetic investigations enhances the characterization of landslides through a unified interpretation approach. AE monitoring can provide insights into the accumulation of preparatory damage before surface displacements become detectable, particularly after surpassing specific thresholds [72]. This method effectively captures internal micro-event emissions, reducing signal attenuation. By combining datasets from various scales and domains, a comprehensive understanding of landslide evolution can be achieved, ultimately leading to improved early warning systems.

Table 3. Summary of geophysical methods.

Geophysical Methods	Advantages	Disadvantages
ERT	Maps subsurface using resistivity differences; infers distribution with multichannel systems and modeling; identifies shear zone anomalies.	Influenced by temperature, pressure, and saturation levels; measurements can be interfered with by metallic minerals and conductive fluids.
SP	Maps natural subsurface potential fields; identifies directions of landslide movement and seepage areas.	Temperature and electromagnetic fields can introduce interference; limits clear interpretation of hydrological factors.
Downhole Seismic Techniques	Measures wave velocity in boreholes to determine elastic modulus; directly assesses subsurface stability.	Effectiveness decreases with increasing depth; limited access to potential slip zones can restrict application.
GPR	Transmits signals for imaging, providing rapid and high-resolution profile measurements; effective for identifying subsurface features.	Signal penetration is affected by water and clay content; velocity determination may not always be achievable.

summaries the strengths and limitations of geophysical methods.

2.4. Remote Sensing Methods

Remote sensing utilizes aerial or satellite-based photogrammetric imaging platforms to delineate areas at risk of landslides [73]. Key techniques include SAR interferometry and the interpretation of optical satellite imagery.

SAR is an active microwave remote sensing technology mounted on satellite platforms and capable of capturing images day and night, regardless of weather conditions [74]. It generates surface backscatter by transmitting microwave signals and measuring the distances of echoes and antenna flight bearing angles. Differential SAR interferometry (DInSAR) analyzes phase differences between interferometric SAR image pairs to map ground deformation with millimeter accuracy by comparing reference and current acquisitions [75]. However, DInSAR requires the precise co-registration of images and correction of atmospheric artifacts, which can be challenging for operational applications. Additionally, frequent imaging repeats are necessary to dynamically monitor continuously evolving slopes.

Passive optical sensors acquiring multispectral data enable morphological terrain characterization through supervised classification routines assigning each pixel into distinct zones linked to terrain/slope features [76]. High-resolution commercial satellites such as IKONOS, QuickBird, and WorldView constellation facilitate detailed feature identification in stereoscopic compositions [77]. Change detection for identifying areas of significant slope gradient/stability variations through pre-/post-event comparisons is another popular application.

Despite its successful applications, remote sensing faces several challenges in its role within landslide early warning systems [78]. Real-time monitoring is hindered by the intermittent revisit intervals of satellites and the availability of data acquisition, which are influenced by unpredictable factors such as weather conditions, thereby limiting the time available for preparedness [79]. Although multi-temporal data can aid in identifying past landslide occurrences, directly measuring preparatory slope deformations is often impractical. Furthermore, surface features may not accurately reflect the subsurface failure mechanisms controlled by hidden geological characteristics.

In general, remote sensing enhances traditional field mapping by providing a comprehensive overview that is conducive to advanced territorial zoning, retrospective analysis of detailed landslide histories, and identification of changes following major events. However, the limitations in dynamically monitoring evolving slope movements in real-time necessitate the integration of subsurface characterization methods. This integration can supplement remote sensing data and observations with on-site geophysical datasets capable of detecting precursory deformations.

Table 4. Summary of remote sensing methods.

Remote Sensing Methods	Advantages	Disadvantages
SAR and DInSAR	Active microwave remote sensing, capable of imaging around the clock under all weather conditions; DInSAR can measure deformation with millimeter-level accuracy and conduct dynamic monitoring of slopes.	Difficulties in image registration and atmospheric correction can complicate analysis; frequent imaging is required, leading to high operational costs and significant demands on data processing.
Passive Optical Sensors	Characterizes terrain morphology using multispectral data; high-resolution satellite imagery allows for detailed feature identification through stereoscopic compositions; detects changes to identify areas with slope variations.	Difficult to achieve real-time monitoring; data acquisition is affected by satellites and weather conditions; unable to directly measure pre-deformation of slopes; surface features hardly reflect the subsurface instability mechanisms.

provides advantages and limitations of remote sensing methods.

2.5. Limitations of Existing Methods

Although current methods for monitoring landslides have demonstrated success in identifying hazardous areas and analyzing past incidents, they possess limitations that hinder their effectiveness as reliable early warning systems.

Table 5. Comparison of slope monitoring methods.

Monitoring Methods	Advantages	Disadvantages
Geodetic Methods	High precision; direct measurement; less affected by weather.	Limited measurement points; low efficiency; high requirements for the environment.
Geotechnical Methods	Strong pertinence; real-time monitoring; combined with engineering practice.	Damage to the rock and soil mass; high maintenance costs; representativeness issues.
Geophysical Methods	Large-area detection; non-contact measurement; multi-parameter measurement.	Damage to the rock and soil mass; high maintenance cost; representativeness issues
Remote Sensing Methods	Macroscopic monitoring; fast data update; not restricted by terrain.	Relatively low precision; dependent on weather and lighting; complex data processing.
AE Technique	Early damage sensitivity; possession of a natural signal source; applicability in multiple fields; comprehensive analysis of signal parameters.	Potential damage to rock and soil masses; high maintenance costs; issues with representativeness.

summarizes the advantages and disadvantages of several traditional slope monitoring methods.

Geodetic techniques are constrained by the intermittent nature of data acquisition, which is not suited for the continuous monitoring required for unstable slopes. Additionally, surface measurement techniques fail to capture subsurface kinematics that are critical for understanding buried structures, infrastructure, and processes such as deep-seated soil creep. Furthermore, the interpretation of the results often involves significant uncertainties due to factors such as vegetation density, atmospheric distortions, and terrain irregularities.

Geotechnical instrumentation faces challenges related to limited subsurface coverage, as the costly deployment of extensive networks along every potential failure plane at sufficient depths is often impractical. Post-failure detection can delay timely responses, even if full slope coverage is achieved. The high costs associated with these systems prohibit large-scale implementations without continuous government subsidies, which is unsustainable for developing countries, which experience the majority of global landslide disasters each year. Additionally, field instrumentation can become ineffective when access to hazardous areas is obstructed during active failure events.

Geophysical methods are limited by resolution constraints and may yield ambiguous results, necessitating integration with additional techniques, as standalone solutions often lack clearly interpretable parameters. Seismic monitoring requires dense station networks for effective source localization, which increases costs and is not feasible over extensive areas.

Remote sensing methods are impaired for continuous monitoring due to infrequent revisit periods and dependence on favorable weather conditions for image acquisition. Detected changes typically correspond to displacements that have already occurred, rather than capturing early-stage precursory deformations. Moreover, discrepancies may exist between the surface morphological expressions delineated from optical imagery and the actual shear planes that govern internal slope failure mechanics.

These limitations render existing methods less effective as early warning tools that can be integrated into community-based disaster preparedness strategies. Such strategies ideally require low-cost sensors that are robust enough for on-site autonomous operations, coupled with reliable analytical models to provide reasonable lead times for response. Additionally, attenuation remains a major challenge, impeding the direct sensing of granular slope deformation by electromagnetic, acoustic, or mechanically coupled waves. An ideal approach must identify strategies to overcome these restrictions, and the AE technique holds promising capabilities in this regard, as will be highlighted subsequently.

3. Fundamentals of Acoustic Emission Technique

3.1. Acoustic Emission Phenomenon

AE denotes the emission of elastic stress waves resulting from sudden stress release due to irreversible internal structural changes within materials [80]. These stress waves propagate through the material as two main types: compressional (primary) waves, characterized by alternating compressions and rarefactions moving in the same direction as the wave propagation, and shear (secondary) waves, which entail particle movement perpendicular to the energy transport direction [81].

AE testing relies on the passive detection of spontaneously emitted transient elastic waves from the surface or subsurface of a structure (body) using transducers, which convert these waves into electrical signals through the piezoelectric effect. These signals are then amplified for recording and analysis [82]. Unlike ultrasonic testing, which transmits mechanical waves artificially, AE sources arise naturally from micro-failures induced by service or testing within the monitored system (structure, component, or ground mass). This constitutes an active non-destructive testing methodology that is sensitive to early damage development, as shown in Figure 3.

Numerous micro- and macro-level processes involving localized stress redistributions are considered potential mechanisms for AE generation in soils and rocks [83]. Inter-particle friction, sliding, rolling, plowing, and collisions occur during shearing or complex loading of granular materials [84]. Inherent heterogeneities facilitate the formation of force chain networks, where load transfers concentrate on narrow contacts experiencing slip-stick fluctuations under changing boundary conditions. These fluctuations are closely associated with localized elastic wave releases. In both laboratory and field measurements, brittle cracking, crack–crack interactions, crack growth acceleration, and coalescence within monolithic rock materials are widely documented as sources of microseismic AE [85]. Furthermore, fracture processes at material interfaces such as joints, fault zones, cementation planes, and clay-rich laminae also emit AE signals [86].

In landslide monitoring, body waves and surface waves both produce AE, depending on their orientation relative to sensor placement [87]. However, body waves typically predominate as the primary component of the failure process due to their higher velocities. Attenuation refers to the gradual reduction in amplitude and energy as waves propagate outward from the AE source. This attenuation is primarily caused by geometric damping resulting from wavefront expansion, material intrinsic absorption that converts mechanical energy into heat, and scattering at internal interfaces that alters wave particle movements. The significant variation in attenuation observed in geological materials has important implications for the detection range of AE, highlighting the need to consider propagation paths when designing AE monitoring systems.

3.2. Instrumentation for Acoustic Emission Monitoring

Critical AE instrumentation components encompass sensors as transducers, pre-amplifiers, main amplifiers, filters, and an analog-to-digital converter linked to a data acquisition system for parameter extraction and analysis [88]. Figure 4 summarizes the schematic illustration of AE instrumentation.

Highly sensitive broadband sensors are favored for capturing a broad range of frequencies. Piezoelectric wafers made from lead zirconate titanate (PZT) and polarized during manufacturing are widely utilized because they convert mechanical strains induced by elastic waves into voltage signals through crystal deformations [89]. Sensors vary in their frequency ranges, spanning from 0.1 kHz to over 1 MHz, with resonance frequencies adjusted according to specific applications and the properties of the monitored materials.

Sensors must establish effective coupling with monitored structures, either by direct mounting on surfaces or through waveguides, especially in media with high attenuation such as soils [90]. The skin effect, arising from the wavelength of AE exceeding crystal thickness, leads to surface responses that benefit from flush mounting on surfaces to ensure contact with a coupled fluid layer that resonates with the structure for optimal transmission. Coupling materials should possess acoustic impedance similar to that of both sensors and structures to minimize reflections that could weaken signal penetration into the test structures and enhance sensitivity [91].

Signals typically remain weak at nano- and microvolt levels, necessitating pre-amplification to enhance amplitudes before transmission through coaxial cables to main amplifiers positioned at a distance from the sensors to mitigate electromagnetic interferences [92]. Pre-amplifiers with gains ranging between 20 and 40 dB amplify signals to microvolt and millivolt levels, making them suitable for further amplification by main multi-channel amplifiers located far from the sensors. In comparison to traditional cable sensors, optical fiber AE sensors offer several advantages, including high sensitivity, broad frequency response, immunity to electromagnetic interference, capability to operate in extreme temperatures, remote monitoring, and intrinsic safety [93].

Filtering cleans signals from electromagnetic noise, environmental disturbances, and high/low frequency artifacts, respectively, before digitization [94]. High pass filters above 20–30 kHz eliminate seismic/acoustic noises, while low pass filters under 1 MHz attenuate resonance peaks centered at sensor natural frequencies to retrieve genuine signals. Adjustable gain amplifiers regulate final signal amplitudes before analog-to-digital conversion.

Multi-channel data acquisition systems connect amplifiers to computers running customized software for additional processing, event discrimination, and parameter extraction [95]. Fast Fourier transformations are used to analyze frequency domain spectra from time-domain signals, while continuous wavelet transforms extract time–frequency information by converting signals into coded digital bitstream formats [96]. Data analysis provides parameters such as amplitudes, energy levels, rise times, counts, and event rates for empirical correlations with mechanical and failure responses observed through reference techniques [97].

Overall, well-designed AE systems enhance sensitivity by combining components that work together effectively as integrated units to account for the unique attenuation characteristics of geological materials. This ensures signal integrity throughout the entire process, from detection to analysis. Furthermore, the incorporation of automated algorithms for pattern recognition and machine learning modeling improves the interpretation of data from growing datasets collected during continuous monitoring periods.

3.3. Parameters of Acoustic Emission Signals

AE signals consist of transient elastic waves that propagate through materials as a localized response to rapid stress and strain energy releases associated with irreversible damage events occurring internally. As non-stationary stochastic signals, AE signals are primarily analyzed through the extraction of parameters in the time, frequency, and time–frequency domains for empirical quantification [98].

In the time domain, essential features of AE are characterized by the detected waveforms and envelopes, as illustrated in Figure 5 and summarized in

Table 6. Parameters of acoustic emission signals.

Parameter Name	Definition	Function or Significance
Amplitude	The intensity of the peak signal, measured in decibels relative to a reference value or in volts, depending on sensor characteristics.	Represents the energy released by the seismic source and the degree of attenuation during the propagation process.
Rise Time	The time interval from when the signal first exceeds the initial threshold to when it reaches peak amplitude.	Reflects the rate of deformation energy release; a short rise time indicates a sudden release of energy (e.g., brittle intergranular fracture), while a longer rise time suggests a more gradual process, such as ductile yielding.
Count	The number of times the signal amplitude exceeds a preset threshold level within a user-defined time window.	Affected by the amplitude and frequency of the wave; serves as a proxy for the amount of energy released.
Duration	The time interval from the first instance the signal exceeds the threshold to the last instance it does.	Represents the total duration of the signal packet when a discrete event is detected; a longer duration indicates contributions from multiple sources during the event.
Energy	A parameter obtained by integrating the square of the instantaneous voltage signal over the entire duration of the event.	Quantitatively represents the total magnitude of energy released during the event.

.

Moreover, advanced time–frequency domain analyses using wavelet transforms allow for the simultaneous extraction of signal components across a broad frequency range, capturing changes in dominant frequency contributions over time [99]. This technique effectively tracks the dynamic evolution of the dominant frequency in the time dimension, presenting a comprehensive and accurate portrayal of the signal’s frequency characteristics at different moments. This provides crucial data support for studying the developmental patterns and internal mechanisms of the signal. L.M. Spasova et al. have validated the effectiveness and reliability of this method in their research on acoustic emission signals generated during the melting and crystallization of granites [100]. Compared to Fourier analysis alone, this approach is particularly adept at revealing temporal variations in signals with non-stationary characteristics. Parameters such as wavelet energy, cross-correlation coefficients, and standard deviation are valuable for quantitatively comparing signals and serve as references for qualitative interpretations [101].

Empirical observations have demonstrated distinct correlations and sensitivities among various AE parameters in relation to different failure mechanisms in soils and rocks [102]. For example, AE energy, duration, and frequency centroid exhibit greater sensitivity to shear fractures in rock joints compared to amplitude and rise time. Du et al. [103] concluded that energy and peak frequency are effective indicators of rock crack growth processes. Different loading modes also result in varying predominant parameters. Therefore, careful recognition of the significance of these parameters is essential for tailored quantitative analyses of specific source processes in geomaterials.

3.4. Frequency Characteristics of Acoustic Emission in Soils

The frequency characteristics provide critical insights into underlying AE generation mechanisms in soils as they are directly related to stress/strain waveforms released over source process durations. However, observed AE frequency ranges vary widely depending on the soil types, test conditions, and analysis methods employed, as shown in Figure 6.

Several studies have shown that AE signals from granular soils span a 0.1–1000 kHz range in general [29]. Particle scale interactions including sliding, rolling, and breakage produce higher frequency emissions. For example, single-particle crushing tests using silica/coral sands identified dominant frequencies within 100–600 kHz corresponding to particle failure timescales [104]. Similarly, tropical residual soil fracture occurred at 100 Hz–20 kHz under triaxial conditions [105].

Soil sample shearing experiments indicate that sliding and rolling contacts emit lower frequencies. Results from Dagois-Bohy et al. [106] monitoring avalanching sand flows found dominant 60–160 Hz emissions attributed to grain collisions and friction. Similarly, direct shear tests on glass beads sheared at 60 kPa produced up to 80 kHz AE signals [107]. Lower confined stresses facilitated more grain reorganization, releasing lower energy signals at lower frequencies.

Unsaturated soil testing is less reported. But results indicate an influence of the degree of saturation on detectable frequency ranges. Frequencies up to 10 kHz were recorded during partially saturated sand flow experiments, presumably owing to localized pore structure reconfigurations accompanying fluid redistributions [108].

Regarding field applications, typically lower monitoring frequencies prove practical. Landslide analyses employed seismometers capturing 20–30 Hz MS signals from slipping soil/rock volumes [109,110]. Likewise, excavation slope assessments relied on 5–8 kHz bandwidth detections through embedded waveguides [111]. Lower frequency ranges allow for the monitoring of larger volumes hampered at higher bands by high attenuation in geological media.

Overall, multi-frequency analysis may better characterize failure micro-mechanisms in granular soils by capturing the co-existence of grain-scale and fluid–pore interactions, depending on frequency sensitivities. Additionally, factors like stress-states, saturation conditions, and soil mineralogy introduce complexity, necessitating systematic parametric investigations for establishing quantitative correlations between frequency contents and specific processes driving slope instability, relevant for monitoring and early warning objectives.

3.5. Attenuation Characteristics of Acoustic Emission in Soils

Attenuation refers to the exponential decay in AE signal amplitude as it propagates away from its source location within a material. It poses a major impediment for employing AE techniques for long-range subsurface monitoring applications in geo-materials exhibiting high damping capacities. Understanding attenuation properties specific to soils proves crucial to optimize instrumentation for overcoming restricted detection limits.

Attenuation arises due to factors like geometric spreading, scattering, mode conversion, material intrinsic damping from internal friction, and macro/micro structural heterogeneities [112]. It depends jointly on frequency, confining stresses, and soil mineralogical-textural attributes through their influence on elastic moduli and fabric anisotropy.

Ono [113] investigated the frequency-dependent performance of ultrasonic transducers and AE sensors, demonstrating that the reception sensitivity can be precisely evaluated by utilizing both sinusoidal wave and impulse excitation techniques. At lower test frequencies, the impact of attenuation significantly rises as a result of the escalating sensor impedance. Consequently, adjustments are necessary to determine the intrinsic sensitivity of a sensor, also known as open-circuit sensitivity, which may deviate by over 20 dB from outcomes obtained with typical preamplifiers featuring an input impedance of around 10 kΩ.

Koerner et al. [114] reviewed data on AE attenuation in different geomaterials like sands, silts, gravels, and intact rocks. Sands demonstrated the highest damping, in the order of 10 dB/cm, which decreased geometrically with increasing effective confining stress from 10 dB/cm at 70 kPa to 2 dB/cm at 700 kPa. Attenuation for silts and gravels ranged within 1–10 dB/cm. Figure 7 shows the relationship between AE attenuation and frequency in sandy soil.

Lin et al. [115] discovered that in saturated soils, the interconnected isotropic water phase promotes the transmission of acoustic waves with reduced attenuation, resulting in an increased detection of AE events. In contrast, in dry or unsaturated soils, the propagation of acoustic waves becomes more intricate due to the existence of discontinuities like particle–particle contacts and interfaces between particles and air or water. Consequently, fewer AE signals are identified compared to saturated soils, and the detection sensitivity is influenced by the frequency spectrum of the emitted waves, as varying frequencies experience diverse levels of attenuation during transmission.

Establishing correlations between AE attenuation and geotechnical parameters, García-Ros et al. [116] indicated decreasing trends with increasing consolidation stress and void ratio. While attenuation increased hyperbolically with frequency, moisture content exponentially amplified attenuation.

Attenuation mechanisms remain complex combinations of scattering from void space heterogeneities, grain-to-grain contact impedance mismatches, intrinsic material losses, and moisture influences coupling dissipative fluid motions [117]. Accordingly, reliable attenuation predictions demand calibrated models accounting for predominant physics through multivariable dependencies.

Overall, significant soil attenuation poses a significant challenge for long-range AE monitoring endeavors. Future studies that combine empirical validations with microstructure-informed numerical simulations incorporating accurate grain characteristics, spatial arrangements, and soil moisture conditions offer potential for predictive attenuation assessments. This approach can enhance the optimization of field sensor placement and the design of instruments tailored to particular geological materials and stress conditions.

4. Discussion

4.1. Field Monitoring of Soil Slopes

The primary goal of implementing AE monitoring in soil slopes is supported by foundational laboratory validations [109]. Early implementations encountered significant obstacles due to substantial wave attenuation in porous, unconsolidated soil masses, making remote detections impractical. To address these challenges, effective strategies utilize waveguides to enable localized self-emission generation and transmission, particularly along predefined embedded routes.

Initial attempts monitored three active landslides in Colorado, USA, using steel waveguides in trench beds recording AE counts through portable recorders over months [118]. Despite technical limitations, relationships between increased AE and seasonal slope movements were qualitatively interpreted. Using similar setups, Hardy Jr and Taioli [119] reported thresholds for evaluating stability through establishing activity gradients classified as low, moderate, and high.

Subsequent studies optimized waveguide configurations. Steel tubes backfilled with sands and gravels as “active” waveguides transmitted emissions from deforming host soils towards attached sensors along borehole sections [120]. Cowden and Arseley coastal slopes were monitored quantitatively, establishing empirical proportionalities linking event rates linearly to displacement velocities between 0.001 and 1 mm/min for failure forecasting.

The choice of waveguides for AE monitoring is crucial, as it involves balancing coupling efficiency with flexibility. The selection process is also influenced by the monitoring system’s overall effectiveness. Aluminum waveguides outperformed steel ones in transmission efficiency, primarily due to their better compatibility with soil conditions owing to lower seismic impedance mismatch [110].

Recent advancements have consolidated quantification schemes. Linear relationships describe normalized AE event rates directly scaling with real-time displacement rates reproducibly for triaxial modeling, landslide flume, and field slope experiments [121]. Soil-type-dependent proportionality coefficients were proposed to enable early warnings based on gradient criteria of “essentially stable”, “marginally stable”, “unstable”, and “rapidly deforming”, according to linearly extrapolated velocities from AE trends.

Multi-location arrays also located AE epicenters, validating that sources corresponded to slope shear planes and not noise. Slope deformation fronts and zones were successfully mapped using portable arrays of passive detectors linked to centralized recorders for spatially distributed sensing along potential failure courses ahead of morphological imprints left along displaced paths [122]. Data analytics established kinematic fingerprints like b-value variations indicative of preparatory damage build-ups facilitating slope behavior forecasting.

Overall, AE provided subsurface non-contact access addressing major constraints of traditional geotechnical inclinometer readings strictly limited to post-failure stages. Well-designed active waveguides capable of generating localized measurable self-emissions from incipient deforming soils underpinned a fully fledged early warning system validated through decade-long developmental research and numerous successful landslide premonitory detections.

4.2. Monitoring of Rock Slopes

Rock slopes experience failures differently from soil slopes due to anisotropic lithological variations determining strength and disturbance zones depending on discontinuity orientations. Brittle cracks nucleate under stress perturbations, permitting sudden collapses once thresholds are surpassed. Successful AE applications provide insights into failure precursors, potentially forecasting impending breakdowns.

Investigations were performed to examine cliff instabilities [123]. Five seismic stations monitored chalk cliffs in Normandy, France, observing rising activity hours preceding collapse interpreted as progressive damaging. Three months of monitoring of unstable columns in Vercors massif limestone cliffs distinguished fracture process regimes from microseismicity [124].

Large-scale mining-induced slope failures motivated the monitoring of collapsed opencast coal mines in Indonesia using seismic sensors [125]. Back-analyses established empirical relationships between precursor bursts and unstable mass dimensions, permitting predictive modeling simulating failure propagation sequences and outcomes.

Laboratory studies established correlations between AE parameters emitted during true-triaxial tests and failure processes within Aztec sandstone [126]. Results suggest the feasibility of employing AE techniques for seismic hazard evaluation from earthquakes, mining-induced seismicity, dynamite explosions, etc.

Waveguides offered the means of embedding multiple sensors inside boreholes for enhanced detection. Wireless earthquake detection devices incorporated fiber-optic waveguides transmitting Raman backscatter-modulated strain signals from fractured rock masses [127]. Simulations showed the potential for structural health monitoring as damage signals were resolved.

Recent studies have utilized a convolutional neural network (CNN) to automatically classify seismic signals collected over a 15-year period on the Åknes rock slope in Western Norway [128]. The CNN effectively distinguished and categorized eight types of events, achieving an accuracy rate of approximately 80%. This analysis yielded important findings regarding the microseismic activity at Åknes, which exhibited pronounced seasonality correlated with annual temperature fluctuations. By automating the classification process, this research makes a notable advancement in landslide monitoring, enabling the real-time analysis of seismic events associated with slope movements.

Overall, rock slope monitoring through borehole-embedded passive/active AE waveguides furnished direct access into internal damage evolution when sub-surface interrogation was not possible through surface techniques. Emerging machine learning capabilities promise augmented quantitative understandings linking the onset and growth of precursor emissions to slope kinematics, improving forecast lead-times in urgent demand for advanced warning applications.

4.3. Quantification of Acoustic Emission Signals

Quantifying AE parameters into physical unit representations provides a crucial rationalization permitting the interpretation of monitored responses as real deformation indicators. Early investigations qualitatively classified activity levels as shown below:

• Low: below 1000 AE counts with a low probability of failure [129].
• Moderate: 1000–10,000 counts indicating marginal stability requiring monitoring [118].
• High: 10,000–100,000 counts, representing an unstable state demanding remediation [119].
• Very high: above 100,000 counts, signaling failure onset necessitating evacuation [118].

Nevertheless, establishing AE parameters as objective indicators correlated with externally measured deformation rates has been essential for the development of failure prediction models that incorporate specific warning thresholds, with the following capabilities.

• Establishing linear scaling between the natural logarithm of the AE event rate (normalized by duration) and monitored landslide flume displacement rates [121].
• Rating slope stability into four categories of “essentially stable”, “marginally stable”, “unstable”, and “rapidly deforming” through gradient classification of scaled AE–displacement rate relation [121].
• Demonstrating proportional correlation between normalized cumulative AE energy and measured daily displacements of slow-sliding landslide monitored using optical leveling for validation [109].

In addition to quantification, damage classification and the identification of deforming zones have utilized pattern recognition tools, in the following ways:

• Through identifying slope instabilities from stable regions through cluster analyses of AE dataset attributes like b-value, partition coefficient, and skewness [130].
• Through the spatio-temporal tracking of slope structural failure processes through Hidden Markov Models, decomposing AE sequences into discrete states signifying damage evolution stages [122].

Overall, quantitative indices constitute the main focus of contemporary AE monitoring towards enabling dependable early warnings through establishing objective success/failure criteria scientifically validated against reference measurements. In the long run, advanced analytical techniques like machine learning may revolutionize data-driven quantitative evaluations of monitored slope health.

5. Challenges and Recent Advances

5.1. Signal Detection Under High Attenuation Conditions

Attenuation-induced amplitude decay impedes distant subsurface signal propagation, presenting a formidable challenge for long-range in situ monitoring applications employing the AE technique. Overcoming high damping capacities exhibited by geological media constitutes an active area of ongoing research and development.

Key strategies focus on optimizing hardware designs and embedded sensor configurations tailored for specific material conditions to enhance sensitivity and detection efficacy under severe signal weakening scenarios. Notable developments include the following:

• Employing wideband piezoelectric transducers preferentially tuned to lower frequency ranges resulted in reduced damping compared to higher bands [113].
• Deploying localized denser multi-sensor arrays configured in advanced geometries like concentric circles and nested quadrants to maximize detection through coincidence logic combining signals [131].
• Coupling sensors and waveguides firmly using optimized viscoelastic materials with acoustic impedances closely matching both for minimal reflection losses and maximized transmission [132].
• Implementing advanced low-noise pre-amplifier designs integrated very close to sensors to boost initial gains, countering losses through early-stage amplification [133].

Overall, hardware customized according to specific propagation environments and application depths coupled with expertise in strategic sensor configurations holds immense potential for realizing long-range subsurface AE monitoring aspirations hitherto restricted by attenuation limitations.

5.2. Improvements in Sensing and Data Acquisition

Advancements in sensor technologies and data recording systems increasingly aid in higher-efficacy subsurface AE monitoring under adverse conditions. Notable progresses encompass the following:

• Micro-electro-mechanical system-based ultra-compact sensors with broad 20 kHz to 1 MHz bandwidths, low self-noise, resonance frequencies up to 1 GHz, and diameters less than 1 mm for dense embedding [112].
• Fiber-optic-distributed acoustic sensing utilizing Rayleigh backscatter in optical fibers for continuous stroke monitoring over extended lengths, circumventing signal attenuation limitations of traditional point sensors [134].
• Wireless battery-powered low-power sensor nodes streamlining sensor arrays through autonomy, avoiding cabling and overcoming field installation complexities [135].
• Cloud-hosted data repositories augmenting computation and storage capacities facilitating advanced pattern recognition, machine learning, and visualization tools, empowering data-driven interpretation [136].

In general, continuous progress in sensing and data acquisition systems holds the potential to enhance the capabilities of AE techniques for the more accurate monitoring of subsurface defects across large areas, even in challenging conditions. This evolution includes the development of advanced quantification and autonomous methodologies.

5.3. Developments in Signal Analysis Techniques

Contemporary analytics frameworks increasingly exploit expanding datasets from optimized acquisition platforms. Prominent developments include the following:

• Advanced statistical tools analyzing spatio-temporal attributes like b-value distributions, inter-event times, covariance, and clustering, identifying damage progression signatures;
• Machine learning algorithms automating pattern recognition through supervised classifiers discerning stable/unstable slope conditions and source characteristics;
• Deep neural networks trained on massive labeled datasets, achieving source localization and forecasting slope behavior through predictive modeling, often outcompeting conventional linear regression techniques;
• Numerical wavefield simulations incorporating virtual complex structures and sensor layouts, optimizing hardware designs through design validation prior to deployments under the guidance of full-waveform analyses.

Current advancements are moving towards the integration of augmented datasets with physics-based comprehensive numerical–statistical frameworks, leveraging increasing computational capabilities for precise quantitative data-driven damage assessment and prognostics. When coupled with optimized data acquisition systems, these developments hold great promise for enhancing long-range subsurface monitoring with high accuracy.

5.4. Integrating Acoustic Emission with Other Technologies

Leveraging complementary attributes of multiple techniques presents a pragmatic approach overcoming limitations of individual methods. Hybrid monitoring schemes integrating AE hold immense potential for comprehensive evaluations.

Combining AE with traditional geodetic methods addresses uncertainties from isolating surface displacements. Cross-correlating underground fracture growth from AE activity with ground-based interferometric SAR deformation patterns augmented slope stability assessments through validated subsurface–surface couplings [15].

The fusion of AE parameters with continuous hydrological monitoring through in situ pore-pressure sensors quantitatively enabled the discrimination of dilatant failure regions from stable compressive zones, as directly evidenced through hydraulic head changes [137]. Establishing damage–hydraulic couplings optimized predictive modeling by enabling physics-based extrapolations beyond calibration sites.

Exploiting mutual complementary attributes between techniques like wider spatial/temporal coverage of geodetic/hydrological methods against the subsurface sensitivity of AE presents an effective paradigm for advanced long-term predictive monitoring schemes deployed in operational warning systems. Hybrid schemes promise the overcoming of limitations inherent to the isolation of single modalities optimizing monitoring scopes. Combined with optimized signal processing and machine learning frameworks, future applications utilizing such integrated multi-technology platforms appear most promising.

6. Conclusions and Future Perspectives

This study reviewed the fundamentals and state-of-the-art applications of the AE technique for landslide monitoring and early warning. Key conclusions and future outlooks are as follows.

• The AE method exploits the high-resolution, non-contact sensing of elastic emissions generated by evolving subsurface damage as soils and rocks deform. While laboratory investigations have established qualitative and quantitative understandings of the failure mechanisms, signal attenuation remains the foremost challenge for long-range field implementations. To address this, ongoing innovations are optimizing embedded sensor configurations, waveguides, and hardware designs tailored to the specific propagation conditions.
• Advancements in MEMS/optical sensing, data acquisition systems, and wireless automated networks are transforming monitoring capabilities towards higher fidelities over extensive volumes with continuous coverage. However, validating the evolving quantification schemes under diverse geological settings remains necessary to establish their reliability and widespread applicability.
• The integration of AE monitoring with multimodal sensing techniques, such as surface deformation measurements and hydrological monitoring, has enhanced the comprehensive characterization of subsurface failure mechanisms. By leveraging the complementary attributes of individual methods, these hybrid monitoring frameworks overcome the limitations of isolating single modalities, thereby optimizing the monitoring scope and improving the accuracy of predictive models.

In light of these findings, this review recommends future research to explore machine learning-based analysis techniques for AE signals, which may significantly enhance predictive capabilities and improve data interpretation. Considering the complexity of AE data and their correlations with various landslide-related factors, data-driven approaches can help uncover intricate signal features and patterns that are difficult to detect through traditional analysis methods, enabling the more accurate prediction of landslide occurrences. Additionally, the development of coordinated multi-scale monitoring frameworks, optimized through virtual simulations, can further advance the integration of AE with other sensing technologies for enhanced landslide forecasting capabilities.

Overall, the AE technique holds great promise for achieving the long-standing goal of comprehensively characterizing subsurface slope responses through quantitative early precursory detections. This capability can substantially augment disaster mitigation efforts by providing reliable early warning systems. Continued research and development, particularly in the context of climate change and the increasing frequency of extreme weather events, is essential for strengthening the effectiveness of AE as a robust solution for landslide prediction.