Introduction

The distribution pattern of energy resources in China dictates that coal, as one of the major fossil fuels on Earth, plays a critical role in the nation’s energy structure. Moreover, for a considerable period in the future, coal resources will continue to serve as a core strategic energy source for China’s development1,2. With the gradual depletion of shallow coal reserves under conventional underground mining technologies, mining depths have been continually extended into deeper strata to ensure a stable coal supply, leading to a growing number of kilometer-deep coal mines each year3,4,5. However, under deep mining conditions, disasters triggered by the degradation and instability of coal mass and rock—due to complex stress environments and intense mining disturbances—have become increasingly prominent6. It has been demonstrated that the fundamental cause of coal and rock deterioration and instability lies in energy transfer and conversion7,8, a process inevitably accompanied by the evolution of physical information such as acoustic and electrical signals. Therefore, investigating the evolutionary behavior of acoustic emission, ultrasonic waves, resistivity, and other parameters during the deformation and failure of coal mass and rock under external loading is of great practical significance for understanding the mechanisms of damage and deterioration and for early warning of instability precursors.

In response, researchers have employed various monitoring techniques and, guided by engineering practice and laboratory experiments, conducted extensive studies on the physical information response characteristics during coal and rock damage and degradation9,10. In the field of non-destructive testing, acoustic emission has emerged as one of the most representative monitoring methods11. Key parameters include directly monitored indices such as acoustic emission ring-down counts12 and energy13,14, as well as derived indicators like the acoustic emission b-value15 and three-dimensional localization of acoustic emission damage16. For instance, Guo et al.17 investigated acoustic emission characteristics of coal–rock composites with different height ratios by analyzing stage variations in acoustic emission events and energy counts that correlate with the stress–time curves of the specimens. Liu et al.18 classified acoustic emission signals based on the proportion of low-frequency (< 200 kHz) to high-frequency (>200 kHz) components to distinguish between tensile and shear failure, providing a scientific basis for classifying rock fracture signals in field monitoring. Rodríguez and Celestino19 qualitatively and quantitatively characterized the fracture process in two rock types during compression tests using acoustic emission parameters, petrographic analysis, time–frequency signal processing, and three-dimensional source localization. Akdag et al.20 quantified damage evolution and strain burst mechanisms in brittle rocks under true-triaxial loading using b-values, cumulative AE energy, and event rates. Wang et al.21 studied the damage evolution of sandstone under different strain rates via acoustic emission three-dimensional localization. Hasan Ali Abbas et al.33 innovatively identified crack initiation and damage thresholds in composite rock masses by integrating acoustic emission and ultrasonic monitoring, utilizing ultrasonic amplitude and energy. Luís Reis et al.34 analyzed fracture behavior under various biaxial loading conditions in ultrasonic fatigue tests, examining crack propagation paths, critical initiation angles, and fracture surfaces.

As heterogeneous porous media, coal and rock contain internal pores, fissures, and mineral interfaces, which directly influence the propagation paths, energy attenuation, and vibration characteristics of ultrasonic waves and electric currents22,23. Furthermore, during the dynamic degradation of coal and rock under load—such as compaction of initial defects, elastic deformation, crack initiation, and propagation—internal structural changes are reflected in real time in key ultrasonic and electrical parameters24,25. In this context, Zhao et al.26 observed that P-wave velocity in composite specimens shows an increasing-steady trend with rising rock strength and further explored the relationship between P-wave velocity and mechanical parameters of coal–rock composites. Wang et al.27, based on fractal theory and accounting for ultrasonic attenuation in coal, established a fractal wave velocity evolution model for water-bearing coal and investigated the correlation between structural parameters of complex pore–fracture systems and ultrasonic velocities. Jia et al.28 conducted laboratory tests to study precursory wave velocity changes in coal and sandstone under cyclic loading. Tan et al.29 performed uniaxial equal-amplitude cyclic loading and unloading tests on water-saturated coal to investigate the resistivity response during cyclic loading and established a relationship between resistivity and damage under such conditions.

Previous studies confirm that clear precursor signals emerge before the instability of coal and rock. Identifying and interpreting these precursors can provide critical references for early warning of coal and rock failure30,31,32. However, significant differences exist in the response patterns of different monitoring signals during the deterioration and instability process. Further research is needed on how to achieve effective fusion of multi-source information to accurately represent the laws governing coal and rock damage and instability.

Therefore, this study employs an independently developed integrated acoustic-electrical-wave testing system to monitor acoustic emission, ultrasonic wave velocity, resistivity, and other multi-factor signals during the uniaxial rupture process of coal rock specimens. Focusing on parameters such as stress, strain, acoustic emission characteristics, ultrasonic wave velocity, and apparent resistivity, we conduct a coupled analysis of multi-parameter precursor information throughout the instability and fracture process of coal rock samples. Furthermore, an early warning model for precursor identification of coal-rock damage and instability, based on the fusion of multi-source information parameters, is established.

Methodology

Sample Preparation

The coal rock samples were obtained from the #3 coal seam and the associated roof and floor strata of an outburst-prone mine in Shandong. The samples comprised three lithological types: coal, sandstone, and mudstone. For experimental purposes, the specimens were machined into rectangular prisms measuring 50 mm × 50 mm × 100 mm (length × width × height) to ensure close contact with the acoustic emission sensors, copper electrode plates, and ultrasonic probes. The prepared coal rock samples are shown in Fig. 1.

Fig. 1
figure 1

Prepared coal and rock samples.

Full size image

A total of 16 standard specimens were prepared for the experiment, comprising four samples each of coal, mudstone, and sandstone. Under identical loading conditions and monitoring methodologies, three sets of specimens were tested, while one set was retained as a reserve. The experimental grouping and the fundamental physical and mechanical parameters of the coal rock samples are summarized in Table 1.

Table 1 The basic physical and mechanical parameters.
Full size table

Testing loading system and monitoring system

The experimental setup primarily consists of a uniaxial loading system and an integrated information monitoring system, the latter of which includes acoustic emission monitoring, ultrasonic testing, and real-time parallel electrical measurement acquisition, as illustrated in Fig. 2.

The loading device employed is an MTS816 electro-hydraulic servo-controlled testing machine, capable of applying a maximum axial load of 1459 kN with an actuator stroke of 100 mm. This system is characterized by high precision, excellent reliability, and fast response. Loading of the coal rock samples is performed under displacement control until complete failure occurs, with a loading rate of 0.10 mm/min and a sampling frequency of 10 Hz.

Fig. 2
figure 2

The integrated test system consisted of AE, resistivity and P-wave.

Full size image

Using an acoustic emission monitoring system combined with three-dimensional positioning technology, this study monitors in real time the acoustic emission characteristics of different coal rock masses during damage and fracture under uniaxial loading, with the aim of determining the extent of structural damage in coal rock. By tracking variations in wave velocity throughout the uniaxial compression process, the relationship between ultrasonic wave velocity response characteristics and coal rock damage is analyzed, thereby identifying precursor signals related to wave velocity changes during the fracture instability of different coal rock types. A parallel electrical method real-time acquisition system is employed to collect various electrical parameters during coal and rock failure, with a focus on analyzing transient resistivity changes during loading and fracture, so as to reveal the apparent resistivity response characteristics of coal mass at different stages of loading.

Monitoring methods and parameters

The information monitoring system mainly includes an acoustic emission monitoring system, an ultrasonic testing system, and a parallel electrical method real-time acquisition system. The monitoring layout is shown in Fig. 3, and the parameter settings are as follows.

Fig. 3
figure 3

The 3D layout diagram of the sensors.

Full size image

(1) Acoustic emission monitoring system

In this experiment, we employed the PCI-8 acoustic emission testing system produced by Physical Acoustics Corporation (PAC), which includes a data acquisition system, preamplifier, acoustic emission sensors, signal cables, and data analysis software. Four PAC-R6α acoustic emission sensors were used to collect AE signals, arranged symmetrically and orthogonally on the upper and lower end faces, each positioned 15 mm from the respective end faces. The acoustic emission threshold and preamplifier gain were set to 40 dB, with the sensor resonance frequency ranging from 20 to 400 kHz and a sampling frequency of 1 MHz. Vaseline was applied as a coupling agent between the rock samples and the sensors to enhance coupling effectiveness. During the experiment, key parameters such as the ringing count, energy, amplitude, event count, and threshold values of the acoustic emissions generated during the development of internal cracks under uniaxial compression were synchronously collected.

(2) Ultrasonic testing system

The ultrasonic testing system utilizes a built-in high-pressure pulse emission card to generate voltage signals, which are converted into vibrational signals by the transmitting transducer and received by the receiving transducer. The waveform is recorded to study the propagation characteristics of the waves. During the testing process, the acoustic pressure heads were placed on either side of the square samples to measure the longitudinal and transverse wave velocities of the rock under uniaxial loading. Both the excitation and reception frequencies of the ultrasonic waves were set to 500 kHz, with a signal amplification factor of 40 dB.

(3) Real-time data acquisition system using parallel electric methods

The parallel electric method real-time data acquisition system primarily consists of a host computer, data acquisition devices, copper sheet electrodes, and real-time electrical parameter acquisition software. To prevent damage to the rock samples from drilling, this experiment employs a patch-type electrode, adhering 3 mm × 3 mm (length × width) copper sheet electrodes to the surface of the specimens using conductive adhesive. To comprehensively understand the electrical variation characteristics during the rock failure process, the electrodes are arranged in a three-dimensional configuration. Two rows of electrodes are symmetrically positioned on the front and back, as well as the left and right sides of the sample, with each row containing six electrodes, totaling 48 electrode channels. During the experiment, to prevent direct current transmission through the press machine, an insulating membrane is placed between the upper and lower pressure heads of the testing machine and the rock sample.

Experimental results and analysis

Characteristics of acoustic emission precursors

Changes in ringing count characteristics

Figures 4, 5 and 6 illustrate the relationships among ringing count, cumulative ringing count, stress, and time during the uniaxial loading failure process of coal, mudstone, and sandstone. A comparison of the stress variation curves for different rock types reveals that the compaction stage in coal samples is not pronounced, while the post-peak stage in sandstone and coal exhibits significant brittle characteristics. In contrast, the post-peak behavior of mudstone shows certain plastic failure characteristics. The temporal evolution of stress indicates that the loading-induced instability process is generally similar across different coal rock samples, consisting of four main stages: the compaction stage (OA segment), the elastic deformation stage (AB segment), the plastic deformation stage (BC segment), and the post-peak failure stage (CD segment). The variations in acoustic emission ringing count and cumulative ringing count exhibit distinct characteristics at different loading stages.

When comparing the acoustic emission signal variations of coal, mudstone, and sandstone under identical loading conditions, the overall trend in ringing count and cumulative ringing count during the instability process is largely similar. The ringing count remains relatively low during the compaction and elastic deformation stages, becomes more active in the plastic deformation stage, and declines again after peak stress. A sharp increase in the acoustic emission ringing count occurs as the peak failure strength is approached. Meanwhile, the cumulative ringing count shows no significant rise during the compaction stage, increases slightly in the elastic stage, and then rises in a step-like manner during the plastic deformation stage. These experimental results demonstrate that the variations in ringing count and cumulative ringing count during loading correlate well with the initiation and propagation of internal cracks in the rock samples, effectively reflecting the stages of internal damage and failure in coal rock.

Fig. 4
figure 4

Acoustic emission ringing count, cumulative count and stress of coal sample M2 with loading time.

Full size image
Fig. 5
figure 5

Acoustic emission ringing count, cumulative count and stress of mudstone sample N1 with loading time.

Full size image
Fig. 6
figure 6

Acoustic emission ringing count, cumulative count and stress of sandstone sample S2 with loading time.

Full size image

Furthermore, due to the inherent properties of the rock materials, different lithologies exhibit varying sensitivities in acoustic emission signals. For instance, coal samples demonstrate relatively active acoustic emission signals during the compaction phase, whereas sandstone and mudstone show nearly zero ringing counts. In coal, the development and propagation of internal cracks are notably vigorous before peak strength is reached, with multiple high ringing count peaks occurring during the plastic deformation stage. The maximum ringing count observed throughout the loading process is approximately 1.45 × 103, recorded at the peak strength. In contrast, sandstone samples maintain low ringing counts during both the compaction and elastic deformation stages. However, due to their dense internal structure and high strength, they accumulate substantial elastic energy during the early loading phase. When local failure occurs, the ringing count increases abruptly, reaching a maximum of about 2.78 × 103 just before the peak strength—significantly higher than that of coal samples. Meanwhile, mudstone samples remain relatively stable in ringing counts throughout the compaction, elastic, and plastic deformation stages. The sensitivity of internal crack development and propagation to acoustic emission signals is low in mudstone, which is attributed primarily to its high clay mineral content. The development of internal cracks occurs mainly through particle sliding, resulting in relatively weak acoustic emission signals during fracture, with a maximum ringing count of approximately 1.28 × 103 at failure.

Spatial evolution characteristics of acoustic emission

The acoustic emission localization technique is used to spatially locate the acoustic emission events during the coal-rock rupture and instability process, to reproduce the process of crack initiation, development and expansion of the specimen from a microscopic point of view, to effectively predict the macroscopic rupture location and damage degree of the coal mass, and to reveal the spatial and temporal evolution law of acoustic emission of different coal mass. The spatial distribution of acoustic emission events in different loading stages of coal, mudstone and sandstone specimens are shown in Figs. 7, 8 and 9, respectively.

Using coal sample M2 as an example (Fig. 7), the specimen exhibited a considerable number of acoustic emission events during the compaction stage due to its relatively developed internal primary cracks and pores. The acoustic emission localization points first appeared in the upper right end of the specimen, with the number of events reaching 72. As the load continued to increase, cracks at the upper end of the specimen continued to develop, and the localization points reached 203, which is 181% higher than that in the compaction stage. After entering the plastic deformation stage, the expansion of internal microcracks increased, and a large number of localization points also appeared at the lower left end of the specimen, which gradually penetrated to form a “main damage zone”. The number of acoustic emission localization points rapidly increased to 835, representing a growth rate of 312%. Upon reaching the peak strength, the specimen became destabilized as a whole, and localized cracking near the main damage zone triggered additional acoustic emission events, bringing the total number of localization points to 962. When the specimen ultimately failed, the total number of monitored acoustic emission events reached 1,075, and the spatial distribution of acoustic emission localization closely corresponded to the actual damage pattern of the specimen.

Fig. 7
figure 7

Spatial evolution of acoustic emission of coal sample M2.

Full size image

Compared with coal samples, the mudstone specimens generated fewer acoustic emission events throughout the entire loading process (Fig. 8). Only 8 acoustic emission events were detected during the compaction stage, primarily concentrated in the middle section of the specimen. After entering the elastic deformation stage, the number of acoustic emission localization points increased slightly, with new points beginning to appear at the lower left end of the specimen. In the plastic deformation stage, internal cracks developed extensively, and a large number of acoustic emission localization points emerged, expanding from the middle part of the specimen toward both ends. The number of acoustic emission localization points rose to 124, representing a growth rate of 463%. After reaching the peak strength, unlike the coal samples, the mudstone did not undergo immediate destabilization and failure; instead, cracks continued to propagate upward along the main fracture paths, indicating a gradual failure process. Newly occurring acoustic emission localization points were mainly distributed along the central region of the specimen, which is consistent with the final failure mode observed, where a slight crack developed on the specimen surface. By the final stage of destabilization, the total number of acoustic emission localization events monitored throughout the process was only 406, approximately 37% of the total acoustic emission events recorded in the coal samples.

Fig. 8
figure 8

The spatial evolution of acoustic emission of mudstone sample N1.

Full size image

During the compaction and elastic deformation stages, the sandstone specimen exhibited relatively calm acoustic emission activity (Fig. 9), with only about 126 localization points detected, primarily in the central region. Upon entering the plastic deformation stage, however, the acoustic emission events increased abruptly, and a substantial number of localization points accumulated in the middle part of the specimen, surging to 1968—an increase of 1461% compared to the previous stage. After reaching the peak strength, the central rock mass rapidly fractured outward, accompanied by a loud explosive sound. The specimen experienced extensive failure, and a total of 2214 acoustic emission events were recorded up to the final destabilization.

Fig. 9
figure 9

Spatial evolution of acoustic emission of sandstone sample S2.

Full size image

The spatial distribution of acoustic emission events and the final failure patterns of the specimens demonstrate that acoustic emission localization accurately reconstructs the internal process of crack development, propagation, and coalescence, thereby providing a reliable indicator of the damage extent and rupture location. A comparison of the spatiotemporal evolution of acoustic emission across different lithologies reveals that—due to differences in material properties such as porosity, density, mineral composition, grain size, and strength—the number of acoustic emission events triggered during each loading stage varies, as does the rate of increase in acoustic emission activity. The intensity of final damage in the specimen is proportional to the number of acoustic emission events: the more acoustic emission events recorded, the more severe the degree of specimen failure.

Acoustic emission b-values

The b-value was first proposed by Gutenberg and Richter to characterize the degree of seismic activity by establishing the relationship between earthquake magnitude and frequency. Studies have found that acoustic emission events generated during rock damage exhibit certain similarities with seismic activities, and the amplitude of acoustic emission is positively correlated with the scale of rock damage. Therefore, in recent years, many researchers have replaced the magnitude in seismic analysis with the acoustic emission amplitude divided by 20, establishing the following relationship between acoustic emission events and amplitude:

(1)

where AdB is the peak amplitude of the acoustic emission event, N is the cumulative number of events with amplitude greater than AdB, and a and b are constants. The acoustic emission b-value is defined as the slope of the log10N and AdB fitting curve, representing the relative proportion of large and small-amplitude acoustic emission events. A lower b-value indicates a higher proportion of large-amplitude acoustic emission events and a more severe degree of coal-rock fracture damage.

In this study, the acoustic emission data were divided into 11 segments according to the stress levels at 10%, 20%, 30%, …, 100% of the peak stress, as well as the post-peak stage. The acoustic emission amplitudes AdB were graded in 5 dB intervals to analyze the variation of the b-value at different damage stages for coal rock, mudstone, and sandstone.

Taking sandstone S2 as an example, Fig. 10 shows the distribution of acoustic emission events across different amplitudes at various stress levels. It can be observed that at the initial loading stage, the number of acoustic emission events with amplitudes greater than 55 dB is relatively small. As the loading stress increases, the number of high-amplitude acoustic emission events rises significantly, and their overall proportion increases, indicating an intensification of rock fracture damage.

Fig. 10
figure 10

Amplitude distribution of sandstone in different stress stages.

Full size image

Based on Fig. 10, the relationship between log10N and AdB/20 at different loading stages can be fitted, as shown in Fig. 11. A strong linear correlation is observed between these parameters, and the acoustic emission b-values obtained from the four acoustic emission sensors are relatively consistent. Following this approach, the acoustic emission b-values at various loading stages for coal rock, mudstone, and sandstone were calculated, revealing the evolution of the b-value throughout the entire loading process, as illustrated in Fig. 12.

Fig. 11
figure 11

Calculation of b-value of sandstone in different stress stages.

Full size image

As shown in Fig. 12, the acoustic emission b-values of different coal rock bodies exhibited distinct magnitudes: sandstone showed the highest values, maintained in the range of 1.5–2.2; coal rock displayed intermediate values, ranging from 0.9 to 1.4; while mudstone presented the lowest values, between 0.5 and 1.0. Throughout the loading process, the acoustic emission b-values of different coal rock bodies showed similar trends with increasing stress. In the early loading stage, the b-value increased, which can be attributed to the compaction and elastic deformation stages where crack development was limited, large rupture events were infrequent, and acoustic emission events were predominantly of small amplitude, resulting in a lower proportion of large-amplitude events. As loading stress increased, the coal rock entered a stage of continuous crack expansion, internal damage intensified, and the number and proportion of large-amplitude acoustic emission events rose significantly, leading to a decreasing trend in the b-value. With further stress increase, cracks continued to propagate and coalesce, releasing substantial energy, and the b-value continued to decline. The initial sudden drop in the acoustic emission b-value can thus be regarded as a precursor indicator for coal rock damage destabilization. Specifically, the b-value drop for coal rock and sandstone occurs at 70%–80% of σc, whereas for mudstone, it occurs later, at 80%–90% of σc.

Fig. 12
figure 12

Variation characteristics of b-value of sandstone with loading stress.

Full size image

Characteristics of ultrasonic wave velocity precursor

Ultrasonic wave velocity serves as a comprehensive indicator for characterizing the physical and mechanical properties of coal mass. It is closely associated with the compactness, porosity, moisture content, and stress state of the material, particularly under loading conditions. In such cases, macro-scale wave velocity variations often reflect the development of internal cracks and pores within the coal rock. At the early stage of loading, the original pores and fissures inside the coal rock are closed under the initial load, and the wave velocity tends to increase; after entering the stage of fissure development, the newborn fissures and original fissures inside the coal rock expand, and the wave velocity tends to decrease significantly before reaching the peak load, so the wave velocity drop before the peak can be taken as the precursor information point of coal rock rupture and instability.

The coupling relationship between the longitudinal wave velocity changes and stress during the uniaxial compression process of certain coal, mudstone, and sandstone is illustrated in Figs. 13, 14 and 15. Due to the varying compositions and porosities of different coal rock bodies, there are discrepancies in their wave velocity characteristics; sandstone exhibits the highest wave velocity, followed by coal, with mudstone having the lowest. As uniaxial loading progresses, the longitudinal wave velocity of the coal rock shows significant variations at different loading stages.

Fig. 13
figure 13

Coupling relationship between wave velocity and stress of coal rock M2 with loading time.

Full size image

Using coal sample M2 as an example (Fig. 13), the wave velocity exhibits a clear increasing trend during the compaction stage. As the loading transitions into the elastic deformation stage, the rate of increase in longitudinal wave velocity moderates, rising from 1508 m/s to 1615 m/s—a growth of 6.42%. Upon entering the plastic deformation stage, the initial phase shows only minor fluctuations in wave velocity. As internal fractures progressively develop and interconnect, the wave velocity displays a gradual declining trend. When the loading stress reaches 82.92% of the peak stress, the wave velocity begins to decrease markedly, dropping from 1615 m/s to 736 m/s at the point of peak load—a reduction of 54.42%. This transition point can be identified as a precursor to the instability and failure of the coal sample.

Fig. 14
figure 14

Coupling relationship between wave velocity and stress of mudstone N1 with loading time.

Full size image

The variation in wave velocity of mudstone during the loading process is illustrated in Fig. 14. Throughout the entire loading process, the wave velocity of the mudstone fluctuates within a relatively narrow range of 1080 m/s to 1248 m/s, indicating a weak sensitivity of mudstone wave velocity to the degree of fracture development in the rock mass. During the initial loading phase, the wave velocity of the mudstone increases slightly from 1212 m/s to 1237 m/s, representing a rise of 2.06%. Subsequently, the wave velocity stabilizes around 1233 m/s until the loading stress reaches 94.2% of the peak strength, at which point the wave velocity curve shows an inflection point and begins to decline gradually. Upon reaching peak stress, the wave velocity decreases from 1233 m/s to 1197 m/s, a reduction of 2.91%. In the post-peak failure stage, the wave velocity decreases at an accelerated rate.

Fig. 15
figure 15

Coupling relationship between wave velocity and stress of sandstone S2 with loading time.

Full size image

Similarly, for sandstone illustrated in Fig. 15, the wave velocity demonstrates a clear increasing trend during the loading and compaction stage, rising from 1845 m/s to 1996 m/s—an increase of 8.18%. However, upon entering the elastic deformation stage, the wave velocity remains relatively stable with minimal fluctuation. During the plastic deformation stage, the wave velocity increases gradually before undergoing a sharp decline, which initiates near 82.20% of the peak stress. At the point of peak stress, the wave velocity drops from 2085 m/s to 1736 m/s, representing a reduction of 16.88%.

The above experimental results indicate that wave velocity variations in specimens of different lithologies generally exhibit similar trends with increasing applied stress. During the initial loading stage, wave velocity increases as fractures within the rock specimen are compacted. Upon entering the fracture development stage, the wave velocity begins to decrease.However, due to differences in composition and varying degrees of fracture development during failure, the impact of fracture development on wave velocity varies across different coal and rock bodies. The precursors to wave velocity reduction appear significantly earlier in coal and sandstone than in mudstone. Moreover, coal and sandstone exhibit markedly higher sensitivity in wave velocity reduction compared to mudstone. This may be attributed to the higher clay mineral content and tighter structure in mudstone, which likely restricts fracture development, slows fracture propagation, and results in a less pronounced wave velocity reduction trend. In contrast, the structure of coal and sandstone is relatively loose, facilitating fracture formation and expansion, which leads to more pronounced wave velocity changes. Additionally, coal, with its higher organic matter content and porosity, exhibits heightened sensitivity to fracture development under loading, resulting in more pronounced wave velocity changes in the experiments.

Characteristics of apparent resistivity

In this experiment, the acquisition parameters for apparent resistivity were set to a power supply time of 0.1 s, a sampling frequency of 1 Hz, and a power supply voltage of 96 V. Complete data acquisition was performed for all power supply electrodes every 15 s throughout the entire loading process.At the end of the experiment, the electrical data collected at different times were compiled and analyzed. Using Surfer post-processing software, data from different loading stages are shown in Figs. 16, 17 and 18, which present the apparent resistivity profiles of coal, mudstone, and sandstone specimens at different locations along the measurement line, respectively. It should be noted that the apparent resistivity of the coal sample in the post-peak stage could not be collected due to its pronounced brittle behavior, which typically leads to instantaneous and severe failure.

Fig. 16
figure 16

Apparent resistivity profile of coal sample M2 at different loading stages.

Full size image

Taking coal rock M2 as an example (Fig. 16), when the specimen is unloaded, the presence of numerous natural defects within the coal rock leads to localized high-resistivity zones in some sections. As the specimen enters the compression and densification stage, the closure of primary pores and fissures increases its density, thereby improving the overall conductivity and resulting in a significant decreasing trend in the apparent resistivity. For instance, in the lower part of profile 1, the apparent resistivity decreased from 3722 Ω·m to 1635 Ω·m, a drop of 56%. During the elastic stage, the apparent resistivity exhibited a further reduction. Upon entering the plastic deformation stage, as the loading stress gradually exceeded the bearing strength of the specimen, a large number of new cracks initiated and propagated within the coal rock sample. This intensification of internal damage was marked by a significant increasing trend in apparent resistivity, with low-resistivity zones gradually transitioning into high-resistivity zones. For example, in the upper part of profile 1, the apparent resistivity increased from 4626 Ω∙m to 9854 Ω∙m, a rise of 113%, indicating that the apparent resistivity of coal samples is highly sensitive to the development of internal cracks.

Fig. 17
figure 17

Apparent resistivity profile of mudstone sample N1 at different loading stages.

Full size image

The apparent resistivity response of mudstone N1 during the whole loading process is shown in Fig. 17. Unlike coal samples, the trend of apparent resistivity decrease in the early loading stage of mudstone is not obvious, which is due to the relatively good densification of mudstone, and fewer primary pores in the natural state. Therefore, the densification of rock samples in the early loading stage does not change much, and there is no obvious change in the electrical conductivity. Moreover, the four profiles do not show significant changes in apparent resistivity during the stages of elastic deformation and fissure development until the peak strength is reached, indicating that crack development inside the specimen has little influence on its electrical conductivity until after damage occurs. The apparent resistivity of the four profiles before reaching the peak strength did not change significantly, indicating that the development of cracks inside the mudstone specimen had less influence on its electrical conductivity, until after the damage occurred. After macroscopic cracks appeared on the surface of the specimen, the range of the high-resistance area in profiles 2 and 3 increased dramatically, and a small range of apparent resistivity increase appeared in the middle of profiles 1 and 4. Compared with profiles 2 and 3, the range of the high-resistance area in profiles 1 and 4 was significantly smaller, indicating that these locations remained basically undamaged, and the damage area was mainly located in the vicinity of profiles 2 and 3. In addition, compared with coal and sandstone, the change in apparent resistivity of mudstone before peak damage is the least obvious, and its apparent resistivity shows the weakest sensitivity to crack development.

Figure 18 shows the apparent resistivity profile during loading of sandstone S2. At the early stage of loading, the extent of the low-apparent-resistivity area increased significantly. For example, in profile 3, the apparent resistivity decreased from 7252 Ω∙m to 3759 Ω∙m, a reduction of 48%. After entering the elastic deformation stage, the apparent resistivity response characteristics did not change much compared with those of the compression-density stage. Upon entering the plastic deformation stage, the range of the high-apparent-resistivity area increased significantly, and the value of the apparent resistivity increased from 9632 Ω∙m to 13,227 Ω∙m, an increase of 37%.

Fig. 18
figure 18

Apparent resistivity profile of sandstone sample N1 at different loading stages.

Full size image

As can be seen from the figure, when the specimen is not subjected to force, significant differences in resistivity among coal masses of different lithology are observed due to variations in their mineral composition, pore structure, and water content. The sandstone exhibits the largest apparent resistivity value, with an average of approximately 5824 Ω·m, followed by the coal rock at about 4552 Ω·m, while the mudstone shows the smallest value, averaging around 282 Ω·m. This indicates that the mudstone has the best electrical conductivity, whereas the coal and sandstone exhibit poorer conductivity. Under uniaxial loading, the apparent resistivity at different profile locations shows a consistent trend of first decreasing and then increasing. However, due to differences in the properties of the coal masses and their sensitivity to conductive characteristics, the apparent resistivity response of different coal and rock bodies during the loading and failure process also varies.

The above test results show that the variation in apparent resistivity of loaded coal mass correlates well with its stress state and can effectively reflect internal damage. Throughout the loading damage process, the mudstone exhibits the best electrical conductivity, but the change in its apparent resistivity is not sensitive; before reaching the peak strength, the expansion of the high-resistivity area is not obvious. In contrast, although the coal and sandstone have poorer conductivity, the change in their internal resistivity during crack development is more pronounced. Therefore, a significant expansion of the high-apparent-resistivity area can serve as precursor information for predicting catastrophic failure of coal rock.

Discussion

Multi-source information response characteristics

The above test results show that, due to the inherent properties of coal and rock, different coal and rock masses exhibit varying sensitivity to various precursor information at different loading stages. The response signals of acoustic emission, resistivity, and ultrasonic waves released during the loading and failure process show strong regularity and complementarity. Therefore, acoustic emission, ultrasonic, and resistivity signals can be utilized to compare and analyze with each other, so as to accurately obtain precursor information of rock failure. When synchronized signals with large amplitude occur frequently, it indicates that the rock is about to enter the destabilization stage.

The acoustic emission b-value continues to increase; in the plastic deformation stage, the acoustic emission signal becomes more active, the ringer counts show a sudden increase, the acoustic emission localization events increase sharply, and the acoustic emission b-value shows a decreasing trend; when entering the post-peak damage stage, the ringer counts reach the peak, the acoustic emission events continue to increase, and the acoustic emission b-value continues to decrease. In summary, the sudden increase of acoustic emission ringer counts, the accumulation of acoustic emission events and the sudden decrease of b-value in the plastic deformation stage can be regarded as the acoustic emission precursor signals of coal rock instability.

As a comprehensive index characterizing the physical and mechanical properties of coal rock, macroscopic wave velocity variations often reflect the development of internal cracks and pores. At the beginning of loading, the original pores and fissures within the coal rock close under the initial load, leading to an increase in wave velocity; after entering the fissure development stage, newly generated and pre-existing fissures expand, resulting in a significant decrease in wave velocity before the peak load is reached. Thus, the pre-peak drop in wave velocity can be taken as precursor information for coal rock rupture and destabilization. As an important geophysical parameter of coal mass, resistivity is closely related to the degree of fissure and pore development, which substantially influences the electrical conductivity: the more developed the fissures, the poorer the conductivity, and the higher the resistivity value. Therefore, based on the characteristics of resistivity changes during the loading damage process, we can analyze the internal damage state of coal mass and provide early warning for coal rock rupture and destabilization.

Criteria of coal and rock instability

Based on the above analysis, acoustic emission, ultrasonic wave, and resistivity are all closely related to the deformation and damage process of coal rock and can serve as precursor warning signals for coal rock damage and instability. In addition, according to the evolution of acoustic emission, ultrasonic, and resistivity signals during the deformation and damage process of the three lithological samples, it can be seen that the characteristic eigenvalue points of each signal parameter appeared in a sequential order, and the intensity of the signal values also varied. Therefore, relying solely on a single type of monitoring information can easily lead to inaccuracies in precursor warnings of coal rock damage and instability. To improve the reliability of early warning, and given that each type of signal has its own characteristics, the ultrasonic, acoustic emission, and resistivity signals monitored during the coal-rock damage process should be comprehensively analyzed to predict rock burst disasters. This integrated approach can not only enhance the accuracy of rock burst disaster prediction in coal mines but also provide important guidance for safe coal mine production.

The determination of the coal-rock instability hazard index can be based on a guiding principle that progresses from single- to multi-source signals and from decentralized to fused data. Specifically, the respective hazard indices are first determined according to the evolution characteristics of acoustic emission ringing counts, acoustic emission b-value, and ultrasonic wave signals. These single-signal hazard indices are then assigned weighting values and fused to establish a comprehensive multi-source hazard index. The calculation of the single hazard index is shown in Eq. (2).

(2)

Where: W (t) is the single-index hazard index, with a value ranging from 0 to 1 (see Table 2); A (t) is the magnitude of the indicator at time t; A0 is the average magnitude of the indicator, and Amax is the maximum magnitude, both of which are determined from field monitoring data.

The composite hazard index for multi-source information fusion is calculated as shown in Eq. (3).

(3)

where: total W(t) is the multi-source information hazard index at time t; i denotes the type of monitoring information, including acoustic emission ringing counts, acoustic emission b-value, and ultrasonic wave velocity; Wi(t) is the single hazard index of the ith information at time t and fi(t) is the weighting factor of the ith information index at time t.

Table 2 Monitoring information risk indicators.
Full size table

Instability warning model based on multi-source information

In this section, a BP neural network model is employed as the framework to integrate multiple parameters—including acoustic emission, ultrasonic, and resistivity signals—by combining their respective single hazard indices. Based on this integration, a comprehensive hazard index is derived from the individual indices, leading to the proposal of a multi-parameter interactive precursor warning model for coal rock instability and damage. The specific procedure is illustrated in Fig. 19.The instability warning model mainly includes four parts: extraction of single feature signals, fusion of different feature signals, calibration and output of the comprehensive warning hazard index, and judgment of the final stability of the coal mass.

Fig. 19
figure 19

Early warning model of coal and rock instability.

Full size image

(1) Input layer: Monitoring of acoustic emission, ultrasonic wave, and apparent resistivity information from the surrounding rock of the model roadway is conducted to obtain characteristic values of various signals during rock deformation and failure. Four parameters—acoustic emission ringing counts, event number, b-value, and ultrasonic wave velocity—are used as the main indicators for judging precursor signals of surrounding rock failure, while apparent resistivity serves as a testing parameter for verification and calibration.

(2) Fusion layer: Based on the characteristic values of different signal types and the criteria for determining single-information hazard indices, the individual hazard indices are first determined, and a comprehensive hazard index is derived according to their weighting relationships. The neural network unit in the fusion layer determines the weights across different time points and signal types by learning from extensive early-stage monitoring data. Once the fusion layer possesses an adequate memory function, it can automatically integrate the input data and output a reasonable comprehensive hazard index.

(3) Output layer: The comprehensive hazard index is verified against test data such as apparent resistivity. When the error between the output and measured data is large, the result is fed back to the fusion layer for reprocessing; when the error is small, the output is forwarded to the decision-making layer.

(4) Decision-making layer: Based on the multi-source fused hazard index and its verification with test data, the stability of the roadway surrounding rock is evaluated. The hazard index is further classified into risk levels, ultimately providing an assessment of the likelihood of surrounding rock failure under varying coal-rock layer conditions and mining technological settings, thereby enabling precursor warning for coal-rock damage instability.

Table 3 Multi source information and stability prediction.
Full size table

Table 3 provides 30 sets of sample data and decision-making results. The correlation and comparison between the predicted and actual values from the sample BP neural network model are shown in Figs. 20 and 21, respectively. The linear relationship between the model-predicted values and the actual calculated values is strong, with a high degree of agreement, indicating that the BP neural network model can accurately predict the stability of the coal mass. By integrating multi-source information, the neural network system effectively fuses these inputs and evaluates rock mass stability based on the output results, thereby providing an effective precursor warning of damage instability.

Fig. 20
figure 20

Correlation comparison between predicted value of BP neural network model and actual calculated value.

Full size image
Fig. 21
figure 21

Comparison between the predicted value of the neural network model and the actual value.

Full size image

Conclusion

In this study, utilizing a self-developed acoustic-electrical-ultrasonic integrated test system, we synchronized multi-parameter data—including acoustic emission, resistivity, and wave velocity—during the deformation and failure process of different coal-rock masses under uniaxial loading. The intrinsic relationships among precursor signals such as acoustic emission ringing counts, b-value, ultrasonic wave velocity, and apparent resistivity during coal mass rupture and instability are thoroughly analyzed, leading to the following conclusions:

(1) The acoustic emission signal response reflects the initiation and evolution process of internal damage deterioration in coal and rock samples. The acoustic emission ringing count corresponds to the macroscopic stress–strain process, primarily comprising four stages: the signal initiation stage during compression and densification, the signal quiet period during elastic deformation, the signal surge stage during plastic deformation, and the signal decline stage during post-peak failure. Three-dimensional localization of internal damage cracks in the specimen, inverted from the acoustic emission event count, visually reproduces the process of crack initiation, growth, propagation, and coalescence from a microscopic perspective. This enables effective prediction of macroscopic fracture locations and damage extent in coal and rock masses, revealing the spatiotemporal evolution patterns of acoustic emission in different coal and rock types. In contrast, the acoustic emission b-value, determined based on signal intensity, exhibits significant variations among different lithologies: sandstone has the highest b-value, maintained between 1.5 and 2.2; coal samples are intermediate, ranging from 0.9 to 1.4; and mudstone has the lowest b-value, maintained between 0.5 and 1.0. Therefore, the sudden increase in acoustic emission ringing counts and the abrupt drop in the b-value can be regarded as precursor indicators of instability failure in coal and rock samples.

(2) Macroscopic wave velocity variations can effectively reflect the development of fissures and pores within coal and rock. The influence of fracture development on wave velocity changes varies among different coal and rock types: the precursor of wave velocity reduction appears significantly earlier in coal and sandstone than in mudstone, and the sensitivity of wave velocity reduction is also markedly higher in coal and sandstone. However, the wave velocity variations in specimens of different lithologies follow generally similar trends with macroscopic stress–strain. In the initial loading stage, primary pores and fissures within the coal and rock close under load, leading to an increase in wave velocity. After entering the fissure development stage, newly generated and pre-existing fissures expand, resulting in a significant decrease in wave velocity before the peak load is reached. Therefore, the point of abrupt wave velocity drop before the peak can be taken as precursor information for coal and rock rupture and instability.

(3) The variation in apparent resistivity closely corresponds to the stress state of coal and rock, making the significant expansion of high-apparent-resistivity zones a reliable precursor for predicting catastrophic failure. Throughout the loading process, mudstone exhibits the best electrical conductivity; however, its apparent resistivity changes only slightly during damage development, and the expansion of high-resistivity zones before peak strength remains limited. In contrast, although coal and sandstone show poorer conductivity, their resistivity changes more noticeably during internal crack development, supporting the use of apparent resistivity as an effective precursor indicator.

(4) A precursor warning model for coal rock damage and instability is proposed, which employs a BP neural network to integrate multi-source information including acoustic emission, ultrasonic wave, and apparent resistivity. The instability warning model mainly consists of four components: extraction of individual feature signals, fusion of multi-type feature signals, calibration and output of a comprehensive hazard index, and stability assessment of roadway surrounding rock. This model enables real-time inversion of the coal rock damage deterioration process, thereby providing accurate precursor warnings of instability. By using acoustic emission and ultrasonic wave information as inputs and apparent resistivity as a verification parameter, the reliability of multi-source information fusion is evaluated through a feedback mechanism, offering a novel reference framework for early warning of dynamic disasters in coal mines.