Due to an increasing demand for mineral resources in China, numerous metal mines, such as the Hongtoushan copper mine, the Dongguashan copper mine, the Jiapigou gold mine, the Sanshandao gold mine, the Fankou lead-zinc mine, the Linglong gold mine, and so forth, have operated at depths exceeding 1000 m [1]. Deep mining inevitably creates an increase in and concentration of ground stress, and the maximum principal stress in the deep stope of some metal mines can exceed 50 MPa [2]. Under these conditions, the incidences of dynamic failures such as rockburst have rapidly increased in recent years in metal mines in China.

Factors that induce rockburst include strong blasting disturbance and high stress concentration caused by overlying mining and tectonic structural planes. For example, the Baiyinnuoer leadzinc mine, a large lead-zinc polymetallic deposit in North China, has a goaf volume greater than millions of cubic meters. Because of the existence of such huge-volume goafs in this mine, in situ stress is highly concentrated in the rock mass of some mining zones. Under this condition, dynamic disasters such as rock ejection and roof caving appear during tunnel excavation when the underground mining reaches only a depth of 300 m. These seriously affect the normal production safety of the mine, as shown in Fig. 1. The failure pattern in the tunnels clearly indicates that the direction of the maximum principal stress is influenced by goafs. In the Dongguashan copper mine, more than a tenth of the rockbursts that led to casualties occurred from 1996 to 1999. In April 2006, a rockburst in the Erdaogou gold mine resulted in injury to many workers and a considerable ore loss, and made mining work much more difficult at the lower levels. In January 2013 in the Linglong gold mine of Shandong Province, two workers were injured by a shock wave induced by a rockburst, and a great deal of electrical equipment was destroyed. According to incomplete statistics, in the period from 2001 to 2007, more than 13 000 accidents occurred in Chinese metal mines that became serious threats to the safe production of the mine and that resulted in more than 16 000 workers injured. Moreover, a large amount of valuable resources could not be extracted.

Several strategies have been proposed to reduce the risk of rockburst; these include the development of a new-generation microseismicity monitoring system, increasing current understanding of the rockburst evolution process using laboratory and in situ tests and monitoring, the establishment of a quantitative warning and risk assessment for the location and intensities of different types of rockburst, and the development of measures to dynamically control rockburst. An expert system with neural network modeling has been developed to assess the risk of rockburst in coal mines [3] and in deep gold mines in South Africa [4]. This paper reviews the progress that has been made in reducing rockburst risk, and provides an example to illustrate the applicability of these new strategies.

With the development of computer and communication technologies, modern microseismicity monitoring technology is widely used in the field of rock and mining engineering. However, there are some key technical obstacles that affect the application of microseismic technology. For example, in the absence of global positioning system (GPS) signals, the timing service precision of the collector will greatly affect the location accuracy of microseismic sources. Outside of the sensor array, microseismic source locations are unavailable. In order to solve these problems, further improve the capacity of a microseismicity monitoring system to capture weak signals, and provide more reliable technical support for disaster warnings, a new generation of microseismicity monitoring system was developed (Fig. 2), as described below.

The system hardware consists of sensors, signal fidelity boxes, data-acquisition devices, timing servers, data servers, and related data communications equipment. The software system consists of a microseism acquisition instrument configure (MAC) system based on a web browser for the collectors, microseismic system monitordiagnosis-configure code (MMC) software of configuration and monitoring for the microseismic system, and a real-time waveform dynamic monitoring system for the waveform. Furthermore, realtime geo-disaster microseism signals identification and analysis system (GMS) software and visual geo-disaster microseism signals three-dimensional dynamic display (GMD) software are employed for microseismic information during the development of geological hazards. The MAC system regulates and manages the collectors, while the MMC software monitors and manages the microseismic system. The GMS monitors provide analysis and warning of microseismic activities, and the GMD software displays three-dimensional microseismic information and disaster risks.

The characteristics and advantages of this system are summarized as follows:

(1)This system realizes 32-bit analog-to-digital (AD) conversion and improves the ability of the microseismic system to capture weak microfracture signals in weak rocks.

(2)The system employs a precision time protocol to update time-synchronization strategies. Moreover, in the absence of GPS signals, the time-synchronization precision of the stations reaches the sub-microsecond level, making it possible to reach high-precision orientation of the fracture sources.

(3)The system can collect all signals within 24 h in order to avoid signal loss due to improper thresholds.

(4)Based on an artificial neural network (ANN) and particle swarm optimization (PSO) algorithms, signals of microseismic sources outside of the sensor array can be effectively and accurately identified, allowing operators to locate fracture sources rapidly and accurately.

(5)Using Internet technology, system faults are diagnosed in real time in order to issue early warnings and send information to designated mobile phones, e-mail addresses, or service centers.

(6)Using public platforms, triggering and continuous data are original and open to access; therefore, users can carry out first-hand scientific research and secondary development according to their specific needs.

The technical parameters of the main hardware of the system are described below.

2.1.1. Technical parameters of the sensors

Velocity sensors. GU(T)10 sensors (Institute of Rock and Soil Mechanics, Chinese Academy of Sciences) are used. Their sensitivity, coil resistance, frequency range, and measurable range are 100 (±5%) V·(m·s⁻¹)⁻¹, 4000 (±5%) Ω, 10–1000 (±10%) Hz, and 10–2000 Hz, respectively. The specification sizes of unidirectional and three-directional sensors are, respectively, ϕ33 mm × 120 mm and ϕ58.5 mm × 180 mm.

Constant current acceleration sensors. AU(T)2000 sensors (Institute of Rock and Soil Mechanics, Chinese Academy of Sciences) are used. Their sensitivity, resolution, frequency range, and measurable range are 2 V·g⁻¹ (g = 9.8 m·s⁻²), 0.04 milli-g, 0.2–3000 (±10%) Hz, and ±2.5g, respectively. Their supply voltage, constant current source, bias voltage, and proper temperature are 18–30 V, 2–10 mA, 9–14 V_DC, and –40–120 °C, respectively. The specification sizes of unidirectional and three-directional sensors are ϕ33 mm × 120 mm and ϕ58.5 mm × 180 mm, respectively.

Constant voltage acceleration sensors. AU(T)30000 sensors (Institute of Rock and Soil Mechanics, Chinese Academy of Sciences) are used. Their sensitivity, resolution, frequency range, and frequency response are 30 (±5%) V·g⁻¹, 0.00005g, ±0.16g, and 50– 5000 Hz ± 3 dB, respectively. Their supply voltage, bias voltage, and proper temperature are 24–28 V_DC, 19 V_DC ± 1.5 V_DC, and –20–55 °C, respectively. The specification sizes of the unidirectional sensors are ϕ24.8 mm × 122 mm, and they weigh 156 g.

2.1.2. Parameters of collectors

As they are able to realize 32-bit AD conversion, the signal collectors can sample at frequencies of 4 kHz, 2 kHz, 1 kHz, 500 Hz, and 250 Hz. The dynamic range is not less than 120 dB, and the timing is realized using GPS or the timing server. In addition, the collector has eight channels and a power consumption of less than 5 W. Collectors with geometrical dimensions of 245 mm × 192 mm × 92 mm and a weight of 3 kg are able to work at temperatures ranging from −20 °C to 80 °C.

2.1.3. Synchronization parameters of the timing server

(1)Through GPS timing and network transmission, the timesynchronization precision of the monitoring nodes reaches the sub-microsecond level.

(2)Through GPS timing and long-distance digital subscriber line (DSL) transmission, the time-synchronization precision of the monitoring nodes reaches the sub-microsecond level.

(3)Utilizing the timing server for timing and network transmission, the time-synchronization precision of the monitoring nodes reaches the sub-microsecond level.

(4)Utilizing the timing server for timing and long-distance DSL transmission, the time-synchronization precision of the monitoring nodes reaches the microsecond level.

In mining projects, consideration must be made in microseismicity monitoring to elaborate the evolution of rock mass fractures, both in the entire mining area and at a particular working face. Fig. 3 [5] shows a typical sensor network in a metal mine. A sensor arrangement can be applied through the excavated caverns at the sublevel in order to monitor the stability of the entire mining region (Fig. 3(a)). The radius of this kind of monitoring may extend for several hundred meters or even more. Goafs are a continual issue in the development of ore extraction, and severe monitoring performance loss of the microseismic system will occur around these goafs. Unfortunately, there is a large chance of rockburst in these regions. In order to solve this problem, sensor density should be increased around goafs to improve sensitivity and reduce position error. At the same time, the anisotropic wave velocity of the rock mass (i.e., the different speeds of Pand S-waves propagating from the fracture source to each microseismic sensor) should be adopted for location rock mass fracture events near these goafs. On the whole, the sensor arrangement in the partial working face approximates that of cavern projects, as shown in Fig. 3(b) and Fig. 3(c). Partial sensors are installed through the excavated tunnels near the monitored working face. Other extra sensors are supplied as needed to focus on the regions with rockburst risk. In addition, a dynamic sensor network based on monitored microseismicity and on the risk of rockburst assessment is suggested, using the recycled installation of sensors method.

Rather than an accelerometer, a geophone is suggested for monitoring rockburst in deep mines. The monitored distance of an accelerometer is insufficient for the whole mining area. For the specific stope, the complete radiated wave from a rockburst source sometimes cannot be recorded by an accelerometer. Information on the communication of the microseismic system and the positioning of monitoring parameters can be found in Ref. [5].

Adequate noise-filtering and microseismic source locating are the bases of microseismic data analysis. Conventional data analysis methods based on the far-field microseismic source are suitable for regional monitoring. For the local stope, the signal-to-noise ratio of the rock mass fracture signals is low due to the strong noise; meanwhile, the types of noise vary and some of their related characteristics are similar to rock mass fracture signals. As a result, conventional human-made and signal-index noise-filtering methods become unreliable. A multi-index noise-filtering method such as neural networks is thus essential [6]. Due to the limits of the sensor layout, rock mass fracture sources are sometimes outside of the sensor array. The reliability of conventional linear event-locating methods, such as the Geiger, conjugate gradient, and steepest descent methods, is questionable. For this situation, a perfect non-linear method is a better choice, such as the Monte Carlo, ANN, or downhill simplex and PSO methods [6]. Calibration shots should be executed frequently to ensure location accuracy, and the constant velocity model for event location should be used carefully. For a local stope with complex geological conditions, the anisotropy or layered velocity model is essential.

In situ microseismicity monitoring has revealed that most monitored rockbursts have a microseismic precursor [5,6]. This microseismicity can be seen as a rock mass fracturing process underpinning the rockburst. The mechanism of the rockburst can be interpreted if the types of related rock mass fractures (tensile, mixed, and shear) can be identified. Energy ratio, moment tensor analysis, and P-wave development methods (Table 1) [6‒9] based on real microseismic information are extensively used to identify the types of rock mass fracturing. It should be noted that the reliability of these methods must first be checked against changes in the monitoring situation. A comprehensive method that combines these methods can effectively improve the judgement accuracy [9]. The macro failure characteristics of a rockburst, as observed by means of a field survey, along with the fracture modes (tensile or shear) of the fracture surfaces at explosive rock blocks, as examined by scanning electron microscope, serve as auxiliary information to determine the mechanism of rockburst evolution.

Fig. 4 shows the evolutionary characteristics in the development process of a typical rockburst. It can be seen that:

(1)For a strain burst, the events are mostly documented through a tension mechanism. Only a few events of mixed and shear types can be distinguished during the development process of this kind of rockburst.

(2)In the beginning of a strain-structure slip rockburst, no matter whether there is a single, a set, or two sets of stiff structures, several events with low radiated seismic energy occur in the tensile mechanism. Next, the three types of mechanism event appear alternately.

(3)Aside from strain burst and strain-structure slip rockbursts, most fracturing events are tensile. However, the final occurrence of a rockburst displays the failure mechanism of shear.

(4)The appearance of a major event (lgE > 4.8) depends on the type of rockburst. For most strain-structure slip rockbursts, major events occur at each stage. However, major events can only be seen in the middle and late periods of a strain burst’s evolution.

The strain burst mechanism analysis of a typical deep tunnel indicated that more than 92.5% of fracturing events appear through tension mechanisms and the mixed and shear events are only about 5% on average. As the number of stiff structures increases, the ratio of tensile events decreases quickly. If there is a single stiff structure or one set of stiff structures, the ratio of tensile events is usually less than 86%. This ratio drops to 68% when there are two stiff structures or two sets of stiff structures. In short, stiff structures are a major control factor in the rockburst evolution mechanism. However, more than 70% of events display a tensile mechanism in the rockburst development  process.

In this study, the phrase “rockburst warning” is used to indicate a warning of the location, intensity, and probability of a rockburst, rather than a warning of the time of a rockburst occurrence. The warning result is used for rockburst mitigation.

Studies show that most rockbursts show the following key features during their development processes.

(1)Microseismic events in the spatial distribution approach nucleation as an evolution of the rockburst. Studies show that the spatial distribution of microseismic events during the development process of a rockburst creates spatial fractal behavior [10]. The spatial distribution of microseismic events reveals fractal self-similarity. The nucleation and self-similarity of the spatial distribution of the microseismic events indicate that monitored microseismicity can be utilized to warn of the location of a potential rockburst in the area.

(2)The energy of microseismic events during the development process of the most immediate rockburst creates temporal fractal behavior [10]. The temporal distributions of energy and the number of microseismic events reveal fractal self-similarity. Furthermore, there is a correlation between rockburst intensity and microseismicity. In general, the more active the microseismicity (i.e., higher energy and a larger number of events), the greater the intensity of the rockburst [11]. This indicates that the monitored energy and the number of microseismic events in a temporal distribution can be used to warn of the intensity of a potential rockburst in the area.

Fig. 4. Evolution mechanisms of different typical rockburst types. (a) A moderate strain burst; (b) an intense strain-structure slip rockburst with a set of stiff structures; (c) a moderate strain-structure slip rockburst with two stiff structures.

Based on the in situ microseismicity and rockburst case study, a rockburst warning formula can be established and can be used to provide some warning of rockbursts [11]. In order to improve the accuracy of this rockburst warning, the following characteristics should be considered.

(1)As mentioned above, in terms of the development mechanism and the effect of geological structure, rockbursts can be divided into different types, with a different microseismicity for each type [9,12]. Therefore, different formulas should be developed for each rockburst type.

(2)The microseismicity changes with construction methods, such as drilling and blasting, and tunnel boring machine (TBM) methods [6]. Therefore, different formulas should be developed for different construction methods.

(3)The microseismicity and risk of rockburst change with variations in geological condition, in situ stresses, rock mass properties, excavation, and support. Therefore, there should be a dynamic rockburst warning based on the variation of the microseismicity.

Considering these factors, a warning formula was developed based on the in situ microseismicity and case study that can be expressed as follows [11]:

 

where m is the construction method, r is the rockburst type, i is the rockburst intensity (extremely intense, intense, moderate, slight, or none), j is the microseismic parameter and six parameters in total, w is the weighting coefficient, and P_ji  is the functional relationship between microseismicity and rockburst (one example is shown in Fig. 5). The probability of every rockburst event ranges from 0 to 100%. The bigger the probability, the greater the rockburst risk [11].

Several methods have been developed to reduce the risk of rockburst, such as: optimization of the excavation (or mining) section size and sequences in order to reduce the stress concentration induced by excavation or mining; high stress aided blasting; distressing to release part of the energy; and the support absorbing energy. Here, a method of using high in situ stresses to improve blasting effectiveness in order to reduce the risk of rockburst is introduced.

The deep rock mass is in a high-stress (high-energy) environment. During deep mining by the drilling-and-blasting method, rock fragmentation is completed under the combined effect of high static stress and dynamic stress induced by explosion. The blasting vibration is one of the main causes of dynamic disasters such as rockburst in surrounding rocks. In the blasting process of deep high-energy rocks, the dynamic stress from blasting can induce the disordered release of high energy in the rock mass. As a result, over-breakage and lack-breakage occur to a slight degree, and affect the formation of excavation peripheries. In severe cases, rockburst occurs. Since it is known that the dynamic stress from blasting can induce high energy release in the rock mass, the orderly release of high energy can be achieved by optimizing the engineering layout and the blasting program. Thus, the energy is used in the rock fracture process to improve the blasting efficiency and effect, and the risk of rockburst may be reduced. A new method involves high-stress-induced rock cracking to promote rock blasting and to control rockburst risk. The core idea of the method is as follows: The energy in the rock blasting zone is used to promote rock fragmentation, reduce the explosive consumption, and improve the blasting effect; the energy in the non-explosive rock mass is absorbed and controlled, while the energy level of the rock and the risk of rockburst are reduced; and finally, the blasting vibration is reduced and the regulation of rockburst is realized.

The first step of this method involves an investigation of the mechanism of rock fragmentation under coupled deep high stress and dynamic stress by blasting, and of the mechanism of dynamic catastrophe induced by blasting vibration. Here, numerical simulations and experiments are used to study the blasting failure laws of the energetic rock mass under high stress; the activation, transfer, and release characteristics of the static energy in the rock mass during the blasting process; and the disaster mechanism and the influence factors of the high-stress rock mass induced by the blasting vibration. The experimental results show that static stress has a significant effect on the shape and volume of the blasting zone in the presence of a free surface, as shown in Fig. 6 [13]. The design of the drilling-blasting scheme in deep rock engineering must consider the influence of static stress.

The second step involves real-time monitoring of the stress field in the working area and in the surrounding rock mass during drillingblasting excavation. Through a combination of microseismicity monitoring, stress measurement, and numerical simulation, the stress and energy distribution characteristics of the working area and surrounding rock mass are obtained and used as the basis for the design of the drilling-blasting scheme.

The final step uses the drilling-and-blasting design theory and method, which consider the difference of the stress field. According to the stress transfer law, the stress field and the energy field are regulated by the cutting position, the caving sequence, and the zone of each caving. Thus, the stress field is more favorable to rock fragmentation by blasting and rockburst control.

According to the stress distribution characteristics, the blasting parameters, charge structure, and delay time are optimized and regulated. The shape of the blasting area and the rock fragmentation effect are controlled, and the blasting vibration is reduced.

The Hongtoushan copper mine, one of the deepest nonferrous metal mines in China, has a mining depth of over 1300 m. Ground hazard events such as collapse and rockburst have rapidly increased in frequency, threatening life and causing financial losses. For example, a strong rockburst occurred in the ramp near the stope at the −647 m level on 18 May 1999, resulting in 10 m of long rock mass being damaged and about 60 m3 of rocks being thrown to the opposite side. On 8 January 2005, a rockburst caused dozens of rocks to be thrown from the wall and roof of the stope, as shown in Fig. 7. In addition, due to the extremely complicated geological conditions, impact caving was easily induced when tectonic structural planes were encountered in the mining process. For example, at the No. 10 stope at the −767 m level, an upward slice cutting-and-filling stoping method was applied, with an exposed area of more than 2400 m2. While mining the topmost layer, a structural plane was revealed between the wall and roof of the stope, resulting in a large area of orebody caving from the roof. The caving area was about 400 m2 with a height of more than 3 m and an ore quantity of about 3500 t. To address this situation, microseismicity monitoring technology was employed in 2007 below the −647 m level, and a few rockburst preventative measures were adopted to strengthen the risk management.

Many factors influence the accuracy of microseismic data, including the number of sensors, microseismic event location precision, and precise processing of data. Therefore, rockbursts are predicted using microseismic multi-parameters including the apparent volume, spatial correlation length, energy index, fractal dimension, and b value [14]. Before a large-scale fracture or rockburst occurs, these microseismic multi-parameters display distinct precursory characteristics. This method can significantly improve the accuracy of rockburst prediction.

In deep mining practice, high in situ stress is conducive to breaking rocks in a hard rock mass [15,16]. If methods of using stored strain energy are used to break the rock mass when blasting, less explosive means can be used for efficient and safe mining. The equivalent stress, which is calculated according to microseismic data, is used for mining sequence and blasting parameters optimization. The energy index, EI, is the ratio of the seismic energy, E, to the average energy released by the events of the same seismic moment, E(M 0 ) ,expressed as follows [17]:

The higher the energy index, EI, the higher the driving stress at the source of the event at the time of its occurrence. The equivalent stress, defined as the average stress per cubic meter of rock mass, was used in this study. When at least two regions are prepared for stoping, their equivalent stresses are compared, and the region with a higher equivalent stress is selected for mining, as shown in Fig. 8. In this case, the blasting parameters were changed to reduce the use of explosives by using the stored strain energy to break the rock mass when blasting. The row spacing changed from 1.3 m to 1.2 m, while the hole-bottom spacing increased from 1.7 m to 2.2 m. The optimization of the mining sequence and blasting parameters, according to the equivalent stress distribution, not only reduced the rock mass over-breakage and lack-breakage, that is, the ore dilution and loss (shown in Fig. 9), but also reduced the explosive quantity and blast vibration in order to maintain rock mass stability.

For in situ microseismicity monitoring, it should be noted that there is no linear correlation between the distribution of deformation and the stress in a rock mass. According to this view, dynamic supporting should be used for a rock mass with better integrity and high stress, which is prone to rockburst, while ordinary anchor net support can be adopted for a rock mass with serious degradation, which is prone to hazards in the forms of spalling and stripping. Using control measures for ground pressure hazards that correspond with the occurrence mechanisms can reduce the risks for workers and equipment.

By carrying out the abovementioned prediction and prevention measures for rockburst in the Hongtoushan copper mine, explosive consumption during the mining process was decreased from 0.308 kg·t⁻¹ to 0.255 kg·t⁻¹, and the volumes of over-breakage and lack-breakage, that is, ore dilution and loss, were notably decreased. More importantly, the occurrences of rockburst decreased from more than 20 in 2007 to less than five in recent years, as shown in Fig. 10. Rockburst prediction and prevention measures have provided and will continue to provide a great deal of useful technical support for risk management in the Hongtoushan copper mine, especially for deeper mining in future.

Reducing the risk of rockburst is a major challenge in deep mining. It can be concluded that, for most rockbursts in deep mining, microseismicity can be monitored during the rockburst evolution process and can then be used to warn of the intensities and locations of rockburst due to the self-similarity of the evolution process of rockburst. In situ tests and monitoring can be used to understand the evolution mechanism of rockburst. Big data, including histories of rockburst cases, can be extracted using deep learning techniques. Dynamic and positive measures including the dynamic adjustment of mining parameters and sequences, distressing techniques such as drilling blasting boreholes and microwave fracturing, techniques to absorb energy and vibration waves, and so on, can be used to reduce the risk of rockburst. These can be conducted for different types of rockburst. The effectiveness of such measures is verified by the case of the Hongtoushan copper mine. Nonetheless, a great deal of work remains, for example, in the prediction of the occurrence time of rockbursts and in the mechanism and monitoring of fault-slip rockbursts.

The authors gratefully acknowledge financial support from the National Natural Science Foundation of China (51621006, 413200104005, and 11232014).

Xia-Ting Feng, Jianpo Liu, Bingrui Chen, Yaxun Xiao, Guangliang Feng, and Fengpeng Zhang declare that they have no conflict of interest or financial conflicts to disclose.