Integrated Analysis of the Formation Mechanism of Cracks in a Concrete Dam Using Microseismic Monitoring and Numerical Simulation

The dam of Guanyinyan hydropower station is composed of a concrete gravity dam in the left bank and a rockfill dam in the right bank.During the operation of the hydropower station, several surface cracks occurred in the concrete gravity dam,which threatened the stability of the dam. To evaluate the evolution trend of the cracks and forecast the potential risk of the dam, the microseismic (MS) monitoring technique and finite-element method were used. First, the concrete three-point bending field test was performed to prove the reliability of the MS technique in monitoring the concrete cracks. TheMS monitoring results were consistent with the simulation results. Then, the MS monitoring system was installed in the dam body. By analysing the MS activities before and after the impoundment, the evolution trend of the cracks and potential risk of the dam were evaluated and forecasted. The simulation results were also consistent with the monitoring results. These results can provide significant references for the operation safety of the dam and also present a new thought for the risk evaluation of similar dam engineering.


Introduction
Concrete dams are commonly large with complex structure design and construction process.Thus, the safety and stability of a dam are likely affected if any mistakes occur during the design and construction processes.Cracks are the major reflections associated with the aging and deterioration of hydraulic concrete structures, which can significantly endanger the concrete structures.Serious cracks accelerate the carbonation and erosion of concrete and destroy the integrity and impermeability of the structures.As a result, the strength and stability of the concrete structures are reduced.Slight cracks also affect the durability and appearance of the structure and may develop into serious cracks.As shown in Table 1 [1,2], concrete cracks have occurred in many hydropower dams.
Presently, stress, strain, displacement, and other conventional monitoring measures are often used to evaluate the stability of dam bodies, side slopes, super-large tunnels, and underground cavities.Conventional monitoring can efficiently detect major deformation or macroscopic instability in engineering structures.However, the precursors prior to the major deformation or macroscopic instability can hardly be monitored.As the main material of a dam body, concrete has an apparent brittle feature.When the microfractures initially occur, the surface deformation is not prominent, and the energy release gradually increases when the microfractures accumulate and propagate in concrete.Generally, before the formation of a large fault zone, a large number of microfractures will form around potential cracks.Using the microseismic (MS) monitoring technique, the microfracture signals can be received.By analysing these signals, the properties of the microfractures (i.e., time, location, energy, and magnitude) in the concrete dam can be deduced using data inversion.Finally, according to the Table 1: Examples of dams with cracks [1,2].

Status of cracks Consequences
Single-buttress massive head dam in Xinfeng River in China A horizontal penetration crack 82 m in length was caused by an earthquake of 6.1 magnitude on March 19, 1962.
The reservoir had to operate at a low water level because of water seepage.

Meishan multiarch dam in China
A 70 L/s sudden seepage occurred at the foundation of the right shore, and dozens of cracks appeared in the dam body, resulting in the dam being in danger.
The reservoir had to be emptied for reinforcement.

Chencun gravity arch dam in China
The dam operated at a low water level for a long time, and large horizontal cracks significantly expanded.
The dam was seriously damaged.

Revelstoke Gravity dam in Canada
Abutment cracks occurred on the upstream face of most sections of the dam.The depth of the largest crack was up to 30 m, and the water percolating capacity was 174 L/s.

LegGage arch dam in France
After the initial impoundment, cracks occurred in a wide area on the upstream and downstream faces of the dam and continuously developed afterwards.
The dam had to operate at and maintain a low water level.

Tona arch dam in France
During the initial impoundment, cracks occurred in a large area on the downstream face close to the skewback and continuously developed in the subsequent eight years.
Arch rings and buttresses were added to the downstream side to reinforce the original arch.

Koehibrein arch dam in Austria
After two years since the dam construction was completed, cracks with serious seepage occurred inside the foundation passage, and the uplift pressure on the whole basic plane in the dam section of the river bed reached the total head.
The operation of the dam had to be stopped for reinforcement.

Zeuzier arch dam in Switzerland
When the reservoir was about to be filled, significant displacement to the upstream occurred on the dam, and multiple cracks 15 mm in widths occurred on the downstream face.
The water level of the reservoir had to be reduced, and large-scale repair work was carried out.
size, density, and concentration of the microfractures, the development trend of the macrofractures may be deduced and the potential risk of the dam can be forecasted [3].
In the past two decades, as a three-dimensional, realtime monitoring technique, the MS monitoring technique has been widely applied to assess engineering hazards that involve rock slopes [3][4][5], deep mining [6][7][8][9][10][11][12][13], tunnels [14][15][16], underground powerhouse caverns [17,18], oil and gas storage [19,20], and nuclear waste disposal [21,22].For example, Xu et al. [3] established a dynamic relationship between geological structures and the spatial distribution of MS events and assessed the stability of a high rock slope based on MS activities.Hudyma and Potvin [10] proposed an engineering approach to manage seismic risks in hardrock mines.Based on the evolution laws of MS events, Feng et al. [15] summarized and explored the methods to forecast rockbursts in deep tunnels.Dai et al. [17] introduced the MS monitoring technique to the underground powerhouse and analysed the relationships among excavation, geological structure, and MS clustering.Two in situ experiments (Atomic Energy of Canada Limited's (AECL's) Mine by Experiment at the Underground Research Laboratory (URL) [21] and SKB's Zedex Experiment [22]), which were conducted by the nuclear waste industry, applied MS monitoring to quantify excavation-induced rock fractures.
In this paper, the MS monitoring technique is applied to a concrete dam project.The three-point bending destruction test of concrete and the evolution trend of dam cracks are studied using MS monitoring technique and numerical simulation.The potential fracturing region of the dam body is evaluated and forecasted based on the monitoring and simulation results, which provides a significant technological method to forecast massive volume concrete engineering in hydropower construction.

MS Monitoring Principle
Under internal and external forces or temperature changes, the elastoplastic energy concentrates in the local elastic and brittle materials.After the energy accumulates to a certain threshold, it will cause the generation and expansion of microfractures with the release of elastic waves or stress waves [14,23].The elastic waves with low energy are defined as microseisms in geology.Each MS signal contains abundant information such as the location, source radius, apparent stress, moment magnitude, and energy release, which can reveal the internal changes of the rock mass.According to the sensor sensitivity and concerning MS magnitudes in engineering, an MS signal can be received by sensors that are installed within approximately 100 m from the MS location, as shown in Figure 1.By analysing the MS seismograms, the seismic source information can be obtained.Consequently, the potential failure trend of the macrofractures of the rock can be deduced based on the size, concentration, and density of the microfractures [3].
The main characteristics of the MS monitoring technique are as follows: (1) this technique can capture MS signals before macrodestruction occurs; (2) this technique can directly determine the time, location, and energy release of the internal fractures of the engineering; (3) the sensors can be installed far away from the destruction to ensure that the monitoring system is in operation for a long time without being destroyed; and (4) this technique can cover a large monitoring area.The seismic magnitudes that are triggered by the destruction of the rock mass, which often occurs in engineering, are commonly below 3 (as shown in Figure 2).Because the discussed range of magnitudes in engineering is commonly smaller than the range of natural earthquakes (higher than 3), seismic phenomena can be referred to as "microseisms" in rock mechanics [24].

MS Monitoring System of the Three-Point Bending Destruction Test.
Concrete is a type of composite material comprised of cement paste and aggregate particles.There are many protogenetic microholes and microcracks before any load bearing occurs.When the load is added, new microcracks gradually propagate and form macrocracks.Then, the macrocracks continuously expand and cause the failure of the concrete.To study the MS features of concrete fractures, a concrete sample (50 × 50 × 300 cm) was manufactured on site, and an antibending destruction test with the simply supported beam three-point slow loading method was performed.The uniaxial compressive strength of the concrete sample was 25 MPa.The left side of the sample bottom was restrained in the , , and  directions, whereas the right side of the sample bottom was restrained in the  and  directions.The loading increments were 5∼20 kN in the middle of the concrete sample.Six accelerometers, which have frequency responses of 50 to 5000 Hz (tolerance band ±3 dB), were placed in an array at the end of the concrete sample.The installation areas were rubbed with a polisher, and the sensors were fixed with plaster.The spatial layout of the sensors is shown in Figure 3.The sampling frequency of the acquisition system was set at 20 kHz with real-time data acquisition.The loading and monitoring time was approximately 1.5 h.

RFPA 3𝐷 Model of the Three-Point Bending Destruction
Test.RFPA 3D (realistic failure process analysis) is an FEMbased code developed as commercial software by Dalian Mechsoft Company.It is a numerical stress analysis tool used for handling the failure process of materials.It can perform analyses on the stress-strain, acoustic emissions, and potential failure surface of concrete, rock samples, and engineering.The heterogeneity of the mechanical parameters (i.e., Young's modulus, Poisson's ratio, cohesion, and friction angle) is considered using the Weibull distribution in RFPA 3D , as defined by the probability density function: where  is the parameter of the element,  0 is the scale parameter related to the average element parameters, and  is the homogeneity index that defines the shape of the distribution function and represents the degree of material homogeneity [25,26].
The FEM model is assumed to be ideal elastoplastic.Its loading and unloading behaviours are described by elastic damage mechanics.The maximum tensile stress criterion and Mohr-Coulomb criterion are selected as the failure criteria of the elements and are defined as where  1 is the maximal principal stress,  3 is the minimal principal stress,   is the uniaxial compressive strength,   is the tensile stress, and  is the friction angle.The progressive degradation of the materials is induced by the initiation, propagation, and coalescence of cracks.Thus, the numerical results of the fracturing processes can reflect the damage evolution of the materials subject to loading [25,26].
Based on the actual size of the concrete sample in Figure 3, the RFPA 3D model was established with a length of 500 cm, a width of 50 cm, and a height of 50 cm, as shown in Figure 4.This model was completely discretized into 75049 elements.The mechanical parameters of the concrete and reinforcing bar are shown in Table 2 and include Young's modulus (), Poisson's ratio (]), the uniaxial compressive strength (  ), the tensile strength (  ) and the homogeneity index ().The boundary conditions were identical to those of the field test.The progressive failure process of the concrete was achieved under gradual displacement-control loading and each loading step was 0.02 mm.Failed elements occur when the stress gradually grows.Each failed element releases its stored elastic energy, which is considered as an AE source [27].The AE counts are proportional to the number of failed elements in the concrete sample, and the strain energy released by the failed elements are captured.To simulate the three-point bending destruction test more realistically, three reinforcing bars in the numerical simulation were used for loading conveniently.As shown in Figure 3, two sides of the bottom and the middle of the top were arranged of I beams for loading.

MS Monitoring and Numerical Simulation Results.
As shown in Figure 5, the force values, AE counts and Step 9 Step 11 Step 15 Step 23 Step 28 Step 48 (a) Step 9 Step 11 Step 15 Step 23 Step 28 Step 48 Figure 6 shows the spatial distribution of microfractures using numerical simulation and MS monitoring of the field test (a view from the support end face of the concrete beam).According to Figure 6, immediately after the load was added to the sample, the protogenetic cracks in the middle and on both sides of the model were compacted or extended.Thus, the fracturing points at this stage were randomly distributed in the model.At stage 23 of loading on the sample, the fracturing points mainly concentrated in the middle of the concrete beam.When the load increased, the spatial distribution of the fracturing points rarely changed.When the sample finally fractured, the spatial distribution of the fracturing points in the numerical model was consistent with the MS monitoring results.Most fracturing points concentrated in the middle of the sample and caused the sample to fracture.
Figure 7 shows the final fracturing state of the sample in the numerical simulation and field test.The bearing capacity of the concrete beam obtained from the three-point bending field test and numerical simulation were 155 kN and 148 kN, respectively.The relative error between the field test and the numerical simulation was approximately 4.5%.Thus, the results of the concrete three-point bending destruction test, which were obtained from the numerical simulation, were almost consistent with those of the field test.The small difference may result from two main reasons.First, the simulation and field test had similar but not completely consistent loading methods.The field test used I beams to connect with the sample.Although the contact area was small, some difference may remain with the numerical simulation by using the reinforcing bars.Another reason was the heterogeneity.The homogeneity index was used to better reflect the real properties of concrete.In this  simulation, the homogeneity index was set as 4 based on an empirical analysis.However, there remains a difference between the homogeneity index in the simulation and the real heterogeneity in the field test.

MS Monitoring System in the Dam Body.
Guanyinyan hydropower station is located on the middle reaches of Jinsha River between Sichuan and Yunnan provinces, China, as shown in Figure 8.The total installed capacity of this station is 3000 MW.The dam is composed of a concrete gravity dam in the left bank and a rockfill dam in the right bank.The concrete gravity dam is 159.00 m in height and 816.65 m in length, and the rockfill dam is 71.00 m in height and 341.35 m in length.However, during the construction of the concrete gravity dam, dozens of cracks in the lower dam appeared with lengths of several to tens of meters.The crack development will undoubtedly endanger the safety of the dam, particularly in the process of impounding and with variations in the water level.Thus, an MS system was applied to monitor the evolution trend of the cracks in the dam.The MS monitoring system is manufactured by Engineering Seismology Group (ESG), Canada, and is mainly composed of 6 sensors, the Paladin digital signal acquisition system, and the Hyperion digital signal processing system.According to the crack shape in the field, 6 sensors, including 5 uniaxial acceleration sensors and 1 triaxial acceleration sensor, were installed in the relative dam body.The sensors formed a spatial grid structure for better positioning accuracy [28], as shown in Figure 9.The installation holes of the sensors were symmetrically distributed on both sides of the crack.The distances between the installation holes and the crack were approximately 3.5 m.The depth of the installation holes for the sensors was consistent with that of the crack, ranging from 1.5 to 6.0 m.The MS monitoring system operation from March 2014 to December 2014 and the normal operating ratio of the MS monitoring system were approximately 85%.

RFPA 3𝐷 Model of the Dam Body.
Because of the great depth and length of the crack in dam section #12, this section was typically simulated.The cross section of dam #12 is shown in Figure 10(a).The dam foundation scope in this model was extended by 30 m in the upstream and downstream directions, whereas the vertical direction was 40 m (see Figure 10(a)).In addition, a 3.5 m long and 10 mm wide crack was prefabricated near foundation grouting gallery #1 (see Figure 10(b)).This model was discretized into a mesh that contained 296990 elements (see Figure 10(c)).Table 3 shows the mechanical parameters of the concrete of the dam body and rock of the dam foundation.Except for the upstream and downstream dams, all boundary surfaces were normally restrained.In addition, the upstream water pressure  and uplift pressure were considered after the impoundment.

MS Monitoring and Numerical Simulation Results
. The actual working conditions of the dam body were as follows.
(1) From March to June 2014, the upstream water level stayed at an elevation of approximately 1030 m, which was approximately 15 m higher than that in the monitored area.(2) From July to September 2014, the upstream water level stayed at an elevation of approximately 1030 m-1080 m, which was 15-65 m higher than that in the monitored area, and the water level occasionally increased or decreased during this period.
(3) In late October 2014, the impoundment of the reservoir officially started and, till the end of December 2014, the water level was at an elevation of approximately 1130 m, which was 115 m higher than that in the monitored area.Till December 31, 2014, several original events were recorded, most of which were interference signals.By comprehensively analysing the automatic waveform identification and scanning and through comparison with the monitoring results of the concrete model test, frequency of the waveform, energy and magnitude characteristics of the events, positioning results, and other characteristic parameters, 104 concrete MS events were identified and screened.According to the MS monitoring results and field working conditions, a comparative analysis was performed for the MS monitoring and numerical simulation before and after the impoundment.

4.3.1.
Before the Impoundment.Figure 11 shows the comparative analysis between the MS monitoring and numerical simulation results before the impoundment.According to the MS monitoring results of March-May 2014, the MS events of the concrete mainly occurred near and on the right side of the existing crack on platform 1015, dam section #12 (few MS events occurred on the upstream side of the crack), and the depth was within 6 m.Most MS events with small magnitudes occurred near the surface, whereas the MS events with larger magnitudes and high energy occurred deeper than the existing crack (i.e., at an elevation of 1009-1012 m).According to the projection drawing of the passage base plate, the MS events were distributed in the form of a stripe at an elevation of 1015 m, and the position of the stripe was on the right side of the existing crack.A sphere represents an MS event.The colours of the spheres represent the magnitudes of the MS events, and the sizes of the spheres represent the energy release of the MS events.During the period from June to September 2014, few low-energy MS events occurred in August mainly because of a sudden change of the water level and a significant change of the environmental stress in the monitored object.
As shown in the elevation drawing, in both the upstreamdownstream direction and the vertical direction of the dam body, the range of MS events had been increasing before the impoundment (see Figure 11(a)).The numerical simulation results cannot reflect the temporal characteristics of the microfractures (see Figure 11(b)).Representative steps for loading were extracted from the numerical simulation process before the impoundment, and only the spatial distribution features of both results were compared.According to the numerical simulation results, the density of the fracturing points also increased and vertically extended with the progress of loading.The results are consistent with the MS monitoring.
From the - plane view, we find that the MS events mainly concentrated in two areas before the impoundment (Figure 11(a) within the red dashed boxes): on the left side of the upstream end and around and on the right side of the cracks (the left and the right sides were divided based on the left and right banks of the dam body).According to the scope where the fracture points were located in the field monitoring, a localized scope of the numerical simulation was also selected.The distribution on the - plane is consistent with the monitoring characteristics (Figure 11(b) within the purple dashed boxes).The fracturing points mainly concentrated around the cracks and the upstream end of the existing crack.When the upstream was empty, the dam heel was most seriously damaged in the entire dam body.Thus, more microfractures were distributed on the upstream end near the dam heel, which is consistent with the results of the stress analysis.

After the Impoundment.
Figure 12 shows the spatial distribution of the MS monitoring and numerical simulation results after the impoundment.Generally, between November and December 2014 after the impoundment, few concrete MS events occurred, and the MS aggregation was weak in the monitoring area (see Figure 12(a)).The MS events were mainly distributed around and on the left side of the existing cracks (at an elevation of 1012-1015 m) and the depth was 0-3 m.In general, few MS events were detected, but most of them were clearly concentrated with small magnitudes and low energy.
In Figure 12(b), when the burial depth of the water upstream of the dam body reached 120 m, the fracturing points were mainly concentrated in the dam heel and the floors of two foundation grouting galleries near the dam heel.Note that a small number of microfractures occurred in the downstream end of the floor of foundation grouting gallery #1 where the existing cracks were located.These results are consistent with the field monitoring data.Step 16 Step 18 Step 20 In normal operating conditions, typical microfracturing points in the dam section have significantly decreased because, during the construction of the dam body, the weight of the dam body continuously increased, and the dam body was in a stress adjustment stage.As a result, the microfractures in the concrete gradually compacted and even disappeared.When the impoundment began, the number of cracks in the concrete material decreased, and the noncontinuity was slightly eased.Thus, the number of microfracturing points decreased.

Concluding Remarks
Using the numerical simulation and MS monitoring methods, the three-point bending test of a concrete model and the microfracture evolution characteristics of a dam body during impoundment were studied.The overall destruction trend of the dam body was summarized and forecasted.Additionally, the stress status, change in displacement of the concrete sample, and "dam body-dam foundation" system were analysed.The main conclusions are as follows: (1) According to the comparative analysis between the field three-point bending test and the numerical simulation, the fracture mode of the concrete in the numerical simulation is consistent with the result of the field test, which proves the feasibility of the MS monitoring system in monitoring massive concrete microfractures.
(2) An MS monitoring system was established in a gravity dam.The MS events mainly concentrated on the left side of the upstream end and around and on the right side of the cracks before the impoundment.After the impoundment, the number of MS events decreased.The MS events were mainly distributed around and on the left side of the existing cracks (at an elevation of 1012-1015 m), and the depth was 0-3 m.Step 10 Step 10-1 1# foundation grouting gallery (3) The finite-element method was used to study the evolution of microfractures in the dam body, particularly near foundation grouting gallery #1.The effect of different working conditions on the cracks was explored, and the crack-generation mechanism was analysed.
A comparative analysis was performed between field MS monitoring and numerical simulation results.The results were consistent with each other.Thus, a new idea can be provided for subsequent mass concrete monitoring of hydraulic structures.

Figure 5 :
Figure 5: The variations of force, AE counts, and accumulated AE counts with displacement during the failure process of three-point bending destruction test.

Figure 6 :
Figure 6: Spatial distribution of microfractures of three-point bending destruction test.(a) Numerical simulation and (b) MS monitoring.

Figure 7 :
Figure 7: The final fracturing state of the concrete beam (a) Numerical simulation, and (b) Field test.

Figure 9 :
Figure 9: Spatial layout of sensors in the dam.

Figure 11 :
Figure 11: Spatial distribution of microfractures in the dam body before the impoundment (a) MS monitoring data, and (b) Numerical simulation data.

Figure 12 :
Figure 12: Spatial distribution of microfrctures in the dam body after impoundment (a) MS monitoring data, and (b) Numerical simulation data.

Table 2 :
Mechanical parameters of concrete and reinforcing bar.MaterialsYoung's modulus/GPa Poisson's ratio Uniaxial compressive strength/MPa Tensile strength/MPa Homogeneity index

Table 3 :
Mechanical parameters of concrete and rock.