Tunnel seismic advanced prediction method is essential for detecting abnormal bodies ahead of the tunnel face and minimizing risks during tunnel construction. Selecting the seismic source plays a crucial role in influencing the precision and effectiveness of data acquisition. At present, tunnel seismic data is usually collected using explosive and sledgehammer sources. Nevertheless, the various sources are located in different positions within the tunnel excavation zone, resulting in distinct characteristics observed on the surface and reflected waves in the acquired tunnel seismic data. The explosive source has minimal surface wave interference, but it is expensive. However, the sledgehammer source is economical yet plagued by inadequate energy and substantial surface wave disruption. Regrettably, there is a lack of research on seismic sources in tunnels, which impairs the precise interpretation of forecasting conclusions from these sources. This paper seeks to investigate how various sources impact tunnel seismic prediction and suggests a new method that integrates data acquisition and processing from these sources. The explosive source is used once, while the sledgehammer source is used 24 times. Cross-correlation calculations are conducted to enhance the resolution of sledgehammer source data, reducing surface wave interference, based on seismic data obtained from the explosive source. Extensive numerical simulations and tunnel experiments support the validity of this method, highlighting its potential to lower data acquisition expenses and enhance tunnel seismic prediction accuracy.

Tunnel seismic advanced prediction is now a crucial job during tunnel construction, as it has been shown to successfully lower risks linked with the process [1]. Different techniques used for tunnel seismic prediction (TSP) [2, 3], integrated seismic imaging system [4], horizontal seismic profiling [5], sonic soft ground probing [6], tunnel seismic tomography [7], tunnel seismic-while drilling method [8], and 3D true reflection tomography (TRT) [9]. Selecting the right source is important for obtaining data efficiently and accurately predicting outcomes. These sources can be grouped into two primary classifications. The initial source type necessitates drilling and stimulating in the hole, like electric spark sources [10] and explosive sources. On the other hand, the second source type does not need drilling and directly impacts the tunnel wall. Instances of this kind comprise sources like sledgehammer [9], automated [11], electromagnetic shock [12], mechanical shock [13], and hydraulic shock [14] sources. Explosives and sledgehammers are frequently utilized as tools for predicting tunnel advancement in these references. The sledgehammer is frequently chosen over explosives in high-risk gas tunnels. The 3D TRT also utilizes a sledgehammer as its source [9].

The excavation of tunnels leads to the development of an excavation damage zone (EDZ). Nevertheless, there are difficulties in measuring the thickness of the EDZ [15, 16]. The calculation formula for predicting the EDZ was summarized in the literature [17], indicating its correlation with field stress, material compressive strength, and passive soil pressure. The stress distribution in the EDZ varies significantly compared to the surrounding rock mass [18]. Research on the mechanical characteristics of rocks shows a strong connection between the stress and the velocity of the nearby rock [19]. In particular, a linear correlation is evident at low-stress levels (60 MPa or less) [20]. The velocity in the EDZ is not the same as in the surrounding rock. The explosive source is set off in a borehole that is 1.5 m deep, reducing the effects of the EDZ. Alternatively, the sledgehammer source is used directly on the tunnel wall and is impacted by the EDZ. Utilizing explosives necessitates 24 boreholes, with each one being drilled to a depth of 1.5 m. This method is expensive, takes a lot of time, and harms the tunnel walls. On the other hand, opting for a sledgehammer removes the necessity of drilling, leading to reduced expenses and increased productivity. Even though sledgehammers are cost-effective and time-saving, they have inherent limitations when it comes to generating excitation energy. Only a small number of researchers have studied the differences in tunnel seismic data when using explosives and sledgehammers as sources, and the specific influence of the EDZ on the seismic signal is still unclear.

The resolution of the reflected wave has a significant impact on the accuracy of TSP. Different elements make up the tunnel seismic wavefields, including direct waves, surface waves, reflected waves, acoustic waves, and interference waves. In order to enhance tunnel seismic data processing quality, various methods are used to boost reflected waves and reduce surface waves [21]. Nevertheless, depending solely on seismic data processing techniques used for surface data to improve the resolution of reflected waves is not feasible due to the scarcity of tunnel seismic data. Hence, it is necessary to investigate other methods to improve the resolution of reflected waves in tunnel seismic data, as it directly affects the accuracy of data processing outcomes and forecasts.

The prior research on TSP [18] did not thoroughly take into account the impact of the EDZ and was deficient in its comprehensive analysis of seismic data gathered from explosive and sledgehammer sources. This research specifically examines the surface waves and reflected waves found in the seismic data. Even though explosives are renowned for their precise forecasts in TSP, they come with a high price tag. However, the reliability of forecasts made with inexpensive sledgehammers remains uncertain. Thus, this study seeks to explore how various source types impact TSP and suggests a combined approach to obtaining and analyzing data from these sources. This approach overcomes the restrictions of explosives and sledgehammers in terms of data collection and accuracy of predictions. This paper introduces an escalating velocity model to thoroughly study the effects of the EDZ, examining the properties of surface waves and reflected waves at various borehole depths of the geophone and source. The proposed joint processing method’s effectiveness is further shown by using both simulated and real data.

Prior research [22, 23] has thoroughly analyzed the simulation of tunnel seismic full-wave fields. Nevertheless, previous studies failed to consider the impact of the EDZ. A linear relationship has been noticed in the mechanical properties of tunnel rock between velocity and stress at low-stress levels [20]. In order to study the wavefield features of tunnel seismic data with a focus on the EDZ, this research integrates a stress model around the EDZ [18] and develops a velocity model that rises incrementally as shown in Figure 1(a). The velocity of the P-wave of the surrounding rock ranges from 600–3500 m/s, as indicated in Figure 1(b). The model has dimensions of 60 by 180 m and a constant grid spacing of 0.4 m. The Ricker wavelet utilized in this research has a peak frequency of 500 Hz and functions as a source of stress concentration. Geophone points G1 and G2 are 10 m apart, with source points S1 and S2 being between 30 and 76 m in Y-direction. The distance between shots is 2.0 m. The geophone locations G1 and G2 are 10 m apart, while the source locations S1 and S2 are positioned at varying depths. G1 and S1 are positioned at −5 m in the X-direction, which corresponds to a drilling depth of 2 m. G2 and S2 are positioned at −7 m in the X-direction, which corresponds to a drilling depth of 2 m. The finite-difference time-domain elastic wave equations are used for forward modeling [23]. The perfectly matched layer is the boundary condition that is used. A time increment of 35 μs is utilized, with a recording duration of 140 ms.

2.1 0. m Geophone Borehole Depth

The geophone is located at -5.0 m on the tunnel wall, with sources placed between −5.0 m and −7.8 m, representing depths of 0 m to 2.8 m in the borehole. The findings in Figure 2 indicate that the red box S represents the surface wave, direct S-wave, and sonic wave, but is unable to differentiate between them. It is noted that the surface wave and direct S-wave maintain consistency irrespective of alterations in the borehole depth. Yet, the powerful energy of the surface waves obscures the reflected waves, complicating the identification of the reflected waves from interfaces I and II in Figure 2.

2.2 2.0 m Geophone Borehole Depth

The geophone is placed at −7.0 m in the tunnel wall, with the geophone borehole reaching 2.0 m deep. The sources vary from −7.8 m to −5.0 m. The results of the simulation, displayed in Figure 3, demonstrate the existence of a surface wave and direct S-wave highlighted by the red box labeled S. The duration of the time recording is over 60 ms, and the red box S encompasses sonic waves as well. The reflections from interfaces I and II are identified as R1 and R2, respectively. When the sources are located between −5.0 m and −6.2 m, the surface wave (B) interference is quite robust, effectively covering up the reflected wave (C). On the other hand, if the source is located between −6.6 m and −7.8 m, the surface wave interference (B) is not as strong, making it possible to identify the reflected waves (R1 and R2).

Simulation results show that increasing the depths of the geophone and source boreholes decreases surface wave interference and enhances identification of reflected waves. In the TSP geometry being studied, there are a total of two geophones and twenty-four source points. In order to find a middle ground between drilling expenses and signal accuracy, it is recommended to drill the geophone and modify the amount of source boreholes as needed. It is crucial to place the geophone in the borehole during data collection to minimize its impact on the EDZ. Next, we will analyze the migration imaging precision of the wavefields acquired from the −5.0 and −7.8 m source locations, which represent the sledgehammer and explosive sources, respectively.

2.3. Migration Imaging

Implementing F-K filtering enables the extraction of the wavefield in the advanced direction from the simulation records displayed in Figure 3(a) and Figure 3(h). The representations corresponding to Figure 4(a) and Figure 4(b) are shown. Surface waves (S) are the prevalent wave components in Figure 3(a), whereas residual surface waves (N1, N2, and N3) dominate in Figure 4(a), making it difficult to distinguish reflected waves from the interface. Nevertheless, in Figure 4(b), we can see both surface waves (S) and waves reflected (R1 and R2) from interfaces I and II.

In Figure 4, the wave field is shown with equal-plane migration imaging carried out at a migration velocity of 3500 m/s. The migration outcomes can be seen in Figure 5, with the blue lines I and II showing the real interfaces. Figure 5(a) contains multiple robust impedance interfaces, leading to reduced resolution. The locations of the actual interfaces (blue lines I and II) are indistinguishable in Figure 5(a). Figure 5(b) shows greater resolution than Figure 5(a). The intense impedance interfaces are focused within the lateral positions of 120–140 m, which align with the actual interfaces shown in Figure 5(b).

3.1. Joint Acquisition

The seismic data obtained from the explosive source shows little surface wave interference and offers improved resolution for reflected waves in the advanced direction (Figure 3(h)). Nonetheless, this technique requires the drilling of twenty-four boreholes, each ranging from 1.5–2.0 m deep, leading to expensive drilling and higher costs for building tunnels. However, using sledgehammers can eliminate the necessity of drilling because they are directly applied to the tunnel wall, resulting in decreased expenses for construction. Nonetheless, seismic data acquired from the sledgehammer source displays significant surface wave disruption, making it difficult to distinguish reflected waves in the advanced direction (Figure 3(a)). Taking into account the pros and cons of each source, this paper suggests combining the explosive and sledgehammer sources for acquisition and applying a joint data processing approach. Figure 6 shows the joint geometry of the hammer and explosive source. Both geophones G1 and G2 are utilized, with drilling depths varying between 1.5 m and 2.0 m. The sledgehammer’s source positions are labeled as first to twenty-fourth, each spaced 1.5 m apart. The twenty-fourth point also indicates where the explosive origin is situated, which is also excavated to a depth ranging from 1.5–2.0 m. The offset typically defined as 20 m, is the minimum distance between shot point and geophone. The px measurement represents the distance between the twenty-fourth source and the surface of the tunnel. The sledgehammer collects records 1 through 24, with the 24th record specifically linked to the explosive source. By implementing this joint geometry, the need for data collection can be fulfilled with just three drill holes, leading to reduced expenses and enhanced building productivity. Moreover, this joint geometry also maintains the benefits of conventional TSP linear configuration, allowing for the utilization of F-K filtering and other techniques for precise data processing [21].

3.2. Joint Data Processing

The tunnel seismic data collected by the sledgehammer is heavily affected by surface waves, leading to a decrease in resolution of the reflected wave. The typical method for processing tunnel seismic data cannot extract the reflected wave in the intended direction. Therefore, the seismic data from the tunnel collected by the hammer can be regarded as passive noise data. In this paper, we use the passive source data imaging technique and apply the spatial cross-correlation function (SPCC) [24-26] to capture the reflected waves.

The SPCC function can be expressed as [27, 28]

ψ(r0,t)=1AAu(x,t)v(x+r0,t)dx
(1)

where x is the variable in the spatial region A, xA , r0 is the distance within region A, u(x,t) is the observation record of the at the point x, and v(x+r0,t) is the observation record at the point x+r0 .

Combined with the joint acquisition method in Figure 6, it can be considered that u(x,t) is the 1st to 24th record collected by the sledgehammer, and v(x+r0,t) is the 24th record collected by the explosive. If the tunnel seismic data collected by the explosive is missing, v(x+r0,t) is the 24th record collected by the sledgehammer.

3.3. Migration Imaging

The seismic data obtained through the use of a sledgehammer and explosive (as seen in Figure 4) is utilized to determine (1) and choose the 24th channel as the virtual source. Next, cross-correlation calculations are carried out in sequence using records 1 to 24. The findings are displayed in Figure 7(a) , and (b). In order to improve the comparison of cross-correlation results, the source excitation location is simulated at the 24th channel position, as shown in Figure 7(c). In the figures, the surface wave is symbolized as S, the reflected waves from interfaces I and II are labeled as R1 and R2, respectively, and the multiple waves are denoted as M. Multiple waves are classified as multiple waves originating from both interfaces I and II (Figure 1). When compared to Figure 7(a), Figure 7(b) shows less noise energy, particularly in red circles A and B.

The wave field illustrated in Figure 7 is utilized for equal-plane migration imaging with a chosen migration velocity of 3500 m/s. The imaging results can be seen in Figure 8(a) and Figure 8(b). The actual interfaces are depicted by the blue lines I and II. Ⅲ is pseudo-anomalous interfaces. In Figure 8(a), there are many prominent impedance interfaces present, causing a reduction in resolution. Consequently, accurately determining the positions of the actual interfaces (shown by the blue lines I and II) in Figure 8(a) becomes difficult. On the other hand, Figure 8(b) shows greater detail in contrast to Figure 8(a). The main focus of strong impedance interfaces is between lateral positions 120 m to 140 m, mirroring the actual interfaces displayed in Figure 8(b). Figure 8(a) shows numerous fake anomalies interfaces between lateral positions 145 and 175 m (inside black box III), whereas Figure 8(b) depicts a smaller amount of abnormal interfaces and better clarity.

4.1. Data Acquisition

The experimental area is found in a water diversion tunnel in Chongqing, China. The surrounding rock is mainly made up of Tongzi and Honghuayuan formations (O1t+h) from the lower Ordovician period. The rock makeup consists of limestone, dolostone, and biogenic limestone combined with shale. The layers show a posture of 285°∠70°. Categorized by the type of surrounding rock, it is classified as type III. During excavation, the workers came across some fissure water in the vicinity. To collect seismic data, a standard TSP geometry was used, consisting of twenty-four source points and two geophone points arranged as specified [21]. The TETSP-2 tunnel seismic advance detection system was utilized for the collection of three-component seismic data [29]. The time interval for sampling was established at 41.7 μs, with 8,192 samples in each trace. Simulation results show that the wavefield is affected by the depth of the geophone, so a constant geophone depth was utilized during the field test. Figure 6 shows the variables involved in the data collection procedure. The 30 g explosive and the 24 pounds sledgehammer were utilized as resources. The explosive was inserted into a hole drilled 1.5 m deep in the tunnel wall. In contrast, the sledgehammer hit the tunnel wall’s surface head-on, striking five times at every point of impact. It is crucial to mention that individual seismic data is stored for each impact. Figure 9(a) shows the sledgehammer hitting the tunnel wall, while Figure 9(b) displays the waveform data captured by the explosive at the tunnel location.

4.2. Data Processing

In Figure 10, we can see the raw Y-component data that was collected by both the explosive and the sledgehammer. In the data from the explosive source (Figure 10(a)), the direct wave (D) and acoustic wave (A) stand out more, with minimal surface wave (S) interference. In contrast, the direct wave’s energy is low in the sledgehammer source data, with the acoustic wave being obscured by noise and surface wave interference being prominent. The data noise interference from the explosive source (Figure 10(a)) is less than that from the sledgehammer source (Figure 10(b) to (f)). The statistical findings in Table 1 show that the sound-to-noise ratio (SNR) of the explosive source data is three to five times higher than the SNR of the sledgehammer source data across the 30–60 ms range. The results of the spectrum analysis from 0–60 ms can be seen in Figure 11. The explosive source data covers a spectrum bandwidth from 1–1000 Hz, whereas the sledgehammer source data covers a spectrum bandwidth from 1–700 Hz. Therefore, the spectrum analysis results are used as parameters for the band-pass filter in the following data processing phase. Moreover, the explosive source’s energy is seven times more potent than that of the hammer source (Figure 11).

Various data processing techniques, including track equalization, automatic gain control, band-pass filtering, F-K filtering, polarization filtering, and predictive deconvolution [21] (Figure 12), are applied to the seismic data shown in . Figure 13 shows the P-wave records that were obtained. The reflected waves propagating in the forward direction are recognized as the prominent continuous events with high amplitudes in the wavefields. Nevertheless, the reflected waves in Figure 13 are not clearly distinguishable. In particular, R1 and R2 are regarded as the reflected waves that occur most consistently.

The wavefield record, displayed in Figure 14(a), is derived from processing and is capable of accurately detecting ongoing reflected waves that possess high energy. R1, R2, and R3 are the reflected waves in the forward direction. Through the utilization of cross-correlation calculations on to , waveform records like those depicted in Figure 14(b) are acquired. Performing cross-correlation with to using the 24th record in as the virtual source results in the waveform record displayed in Figure 14(c). In Figure 14(b), the connectivity of R1, R2, and R3 is not strong, and there is a notable variation in energy levels between neighboring channels. Nevertheless, in Figure 14(c), R1, R2, and R3 show improved continuity and consistency in waveform.

4.3. Migration Imaging

The direct wave’s arrival time is found using . The true velocity of the surrounding rock is determined by calculating the slope of the direct wave using the least square method. The approximate velocity of the P-wave is 6006 m/s. Migration imaging is conducted through the use of the equal-plane migration method [30], illustrated in Figure 15. Symbols on the figure represent abnormal mileage ranges I1–I6. Figure 15(a) displays three abnormal ranges (I1–I3) within the migration profiles. Figure 15(b) illustrates four abnormal intervals (I1–I4) and is computed based on Figure 13(a). Figure 15(c) shows four unusual intervals (I2, I3, I5, and I6) and is computed based on Figure 14(b). Figure 15(d) is calculated based on Figure 14(c) and shows three abnormal intervals (I1, I2, and I3). Figure 15(b) and Figure 15(c) stand out with the highest number of abnormal interfaces, which complicates the process of distinguishing true anomalies. The unusual interfaces I1 depicted in Figure 15(a) were obtained by migrating the reflected wave R1. Compared to Figure 15(d), Figure 15(a) shows a higher energy level for I1. The reason for this is the existence of powerful residual surface waves in R1 (Figure 14a). Through cross-correlation calculation, Figure 14(b) shows a reduction in the energetic surface waves, leading to a more even energy distribution in R1. Therefore, the enhancement of interface I5 in Figure 15(c) results in an improved resolution. Out of all the migration profiles, Figure 15(d) shows the least amount of abnormal interfaces and the greatest resolution. It is important to point out that the explosive source contains plenty of energy, whereas the sledgehammer source has restricted energy (Figure 11). Despite surpassing 80 m in predicted distance, Figure 15(a) still displays an abnormal interface (I3) with significant energy, while Figure 15(d) shows a relatively low energy anomalous interface (I3).

Subsequent discoveries during excavation at interface I2 (Figure 15) showed the presence of soft shales, as shown in Figure 16(a)(A). Furthermore, there are small karst caves and fracture zones seen at the interface I3, depicted in Figure 16(b) (B and C). The excavation findings closely align with the predicted results obtained using the method of joint acquisition and joint data processing.

TSP requires taking into account the alterations in stress equilibrium within the surrounding rock caused by excavation, especially in the EDZ, which greatly differs from the actual stress levels in the rock [18]. Changes in stress directly affect the velocity properties of the nearby rock. The stress distribution in the surrounding rock is affected by several factors such as the depth of tunnel burial, rock type, level of fissure development, water content, and distance from the tunnel wall surface. In this paper, we only consider the distance from the tunnel wall surface and do not take into account other factors that may have an impact. Our velocity model predicts a small amount of stress on the tunnel wall surface, which causes the velocity gradient to rise from the surface to the surrounding rock. It should be pointed out that, while the geophone borehole depth is constant, the source depth may change. Numerical simulations and measured data are utilized to analyze the characteristics of reflected waves and surface waves at different source depths. The indirect assessment of the influence of the EDZ is also taken into account.

The results of the simulation show how surface wave interference affects seismic data greatly when the geophone and sources are placed on the surface of the tunnel wall, resulting in a reduction in the resolution of the reflected wave. The wavefields are more affected by the depth of the geophone and source. Hence, it is crucial to place the geophone and source inside the borehole in order to prevent the impact of the EDZ in practical TSP. If it’s not possible to use explosives in the tunnel for specific reasons, it’s better to place the geophone in a 2.0 m deep borehole rather than on the tunnel wall.

The precision of forecast is significantly impacted by the resolution of the reflected wave in the forward direction. The resolution of the strong impedance interface in the migration profile is directly tied to the resolution of the reflected wave. The high-resolution explosive source data (Figure 14(a)) covers a frequency spectrum of 1–1000 Hz. In contrast, the sledgehammer data source has lower resolution (Figure 13), typically covering a spectrum from 1–700 Hz, and occasionally only from 1–400 Hz. The tunnel wall surface features various structures, including steel frames, arch-primary support, and empty spaces. Using the sledgehammer as the origin decreases the excitement energy and range of these structures. Moreover, it can be difficult to sustain a steady level of energy for every hit of the hammer. Taking into account these factors, the original data from the sledgehammer typically exhibits poor SNR and resolution. Conventional techniques used to analyze tunnel seismic data have restrictions, resulting in typically poor accuracy when predicting sledgehammer source data. Figures 5(a) ,, 1(b) depict this information.

This paper presents a new approach to handling the problem of expensive acquisition costs and poor resolution in seismic exploration with explosives, along with the challenges of surface wave interference and identifying reflected waves accurately when using a sledgehammer. This approach aims to resolve contradictions and enhance acquisition cost and resolution by leveraging the benefits of both methods. The suggested joint acquisition method decreases drilling expenses by $920 and cuts construction duration by 6 hr. Additionally, it delivers detailed and broad-range information with explosives as the source. Nevertheless, the quantity of data collected is restricted as a result of the high costs of excitation. However, utilizing a sledgehammer for the input yields a high volume of information because of its minimal stimulation expense, although resulting in reduced clarity and a more limited range of frequencies. The joint processing method suggested in this paper combines high-frequency data as a constraint to improve the resolution of low-frequency data and reduce interference from surface waves by using cross-correlation calculations. Both simulated and measured data confirm that the suggested method significantly improves the resolution of sledgehammer source data (Figures 5(a) ,, 1(b)), outperforming the resolution attained by explosive source data (Figures 8(a) ,, 15(a)).

This paper investigates how sources affect tunnel seismic advanced prediction and proposes a method of joint acquisition and data processing. The results, obtained from both numerical simulation and actual measurements, can be described in the following way:

  1. The depth of the geophone and source impacts the resolution of data. Hence, it is essential to implant the geophone within the borehole rather than positioning it on the surface of the tunnel wall when gathering tunnel seismic data.

  2. By increasing the depth of both the geophone and source, it is possible to reduce surface wave interference and enhance the resolution of reflected waves.

  3. The suggested joint collection method provides cost savings during the acquisition process and decreases the time needed for tunnel construction.

  4. The new joint processing method is designed to maximize the use of high-frequency signals in explosive source data and the cost efficiency of the sledgehammer source, leading to a notable rise in data volume. By taking into account the high-frequency explosive source signal as a limitation, it enhances the precision of the low-frequency sledgehammer source signal and successfully diminishes surface wave disruption through cross-correlation computations.

The goal of creating a tunnel seismic source device is to overcome the limitations of using manual sledgehammer sources. Utilizing both electromagnetic and mechanical sources on the tunnel wall allows for reliable and strong energy stimulation. When used with the suggested method of joint processing, the precision of forecasts can be improved even more.

The seismic data used to support the findings of this study were supplied by Xinglin Lu under license and so cannot be made freely available. Requests for access to these data should be made to [Chao Yang; anyufutu@163.com].

The authors declare no conflicts of interest.

This study received funding from the Science and Technology on Near-Surface Detection Laboratory (6142414211607).

Exclusive Licensee GeoScienceWorld. Distributed under a Creative Commons Attribution License (CC BY 4.0).