## Abstract

At the Jinping-I hydropower station the excavation height on the slope of the left bank is about 530 m. The stability of the excavated slope is very important because of the complicated geological conditions. To analyse the process of deformation evolution of the left abutment slope during construction, a comprehensive monitoring system was employed, which combined surface deformation observations, multi-point extensometers and graphite rod extensometers. This paper describes this monitoring system and analyses the deformation development with slope excavation. We also established a 3D numerical model of the slope and simulated the whole process of excavation using the finite-difference method. Results from both monitoring and numerical simulation show that the deformation depth in the left bank slope of the Jingping-I hydropower station is over 150 m owing to large-scale excavation unloading, far greater than that of general engineering slopes. The 3D limit equilibrium method and the strength reduction method are applied to evaluate the stability of the excavated slope. The calculated factors of safety indicate that the left abutment slope is stable and safe as a whole, which is in agreement with the deformation monitoring results.

The Jinping-I hydropower station is located at Pusiluogou in the west of Great River Bend of Yalongjiang River, adjacent to Muli and Yanyuan counties in Liangshan Yi Autonomous Region, Sichuan province, P.R. China (Fig. 1). It is mainly built for power generation, flood control and silt retention. The reservoir is adjusted annually; its normal water level is 1880 m and its storage capacity is 7.76 × 10^{9} m^{3}. The total installed capacity of the power station is 3600 MW, with an annual power generation of 1.66 × 10^{10} kW h. The key structures are a concrete double curve arch dam, spillway tunnels and an underground water diversion and power generation system located in the right bank. The maximum height of the arch dam is 305 m, which makes it the highest dam of its kind in the world (Zheng *et al.* 2003).

The construction site is characteristic of a deeply cut valley environment and poor geological conditions (Chen *et al.* 2015). Geological structures such as bedding planes, fractures, dykes and faults of various scales are widely developed in the left bank slope, which leads to poor slope stability even in the natural state (Qi *et al.* 2004; Jiang *et al.* 2015). To construct the arch dam, a volume of rocks up to 5.5 × 10^{6} m^{3} was excavated in the left abutment between 2110 and 1580 m elevation, forming an excavated slope up to 530 m high and with a slope varying between 1:0.5 and 1:0.3. Such a large-scale excavation inevitably induces a dramatic disturbance to the slope, which further diminishes the stability of the left abutment slope in a difficult geological setting. Stabilization of the left abutment slope, therefore, becomes a critical issue for the safe construction of the dam and the success of the hydropower project.

To effectively predict and control the stability of rock slopes, a variety of methods, both qualitative and quantitative, have been employed to investigate the failure mechanism of slopes and to evaluate the stability of slopes; these include monitoring and measuring techniques, model tests and numerical simulations (Sheng *et al.* 2002; Yeung *et al.* 2003; Corkum & Martin 2004; Yu *et al.* 2005; Xu *et al.* 2011; Marcato *et al.* 2012; Jiang *et al.* 2013; Jiao *et al.* 2013; Wei *et al.* 2016). Because of the complexity and uncertainty of the geological conditions, slope deformation monitoring is by far the most straightforward and effective method to capture pre-failure strains that could potentially lead to landslides. With this consideration, a comprehensive monitoring system was installed in the left bank slope at Jinping-I hydropower station, which combines surface deformation observations, multi-point extensometers and graphite rod extensometers. The deformation of the rock slope was systematically monitored at three depth levels below the surface.

In this study, the monitoring scheme and monitoring instrument arrangements for the left abutment slope of the Jinping-I arch dam are presented. The deformation behaviour of the slope during excavation is analysed and the correlation between slope deformation, geological structures and excavation procedure is discussed. Numerical simulations are conducted using FLAC3D software to obtain an overall view of the slope deformation. Both the monitoring and numerical results provided data underpinning the evaluation of the stability and deformation of the high rock slope in this poor geological setting.

## Geological setting of the left abutment slope

The left abutment slope at the dam site is characterized by a high and steep geometry and difficult geological conditions. Figure 2 shows a geological plan view of this slope, together with the layout of excavation. A typical geological profile (II1–II1) of the dam site is shown in Figure 3. The bedrock of the slope is composed of metamorphic rocks belonging to the Zagunao group of the middle and upper Triassic (T_{2-3Z}). It can be divided into three members: a greenschist member (T_{2-3Z}^{1}), a marble member (T_{2-3Z}^{2}) and a slate and low-grade metamorphic sandstone member (T_{2-3Z}^{3}). The left bank is a typical antidip slope and the bedding typically dips at an angle of 30 – 45° (see Fig. 3). There are marble outcrops below elevations of 1850 – 1900 m and the slope gradient is about 50 – 70°. The slate and low-grade metamorphic sandstone outcrops appear above elevations of 1850 – 1900 m and the slope gradient is 40 – 50°. The integrity of the rocks is fairly poor.

As shown in Figure 3, the horizontal distance of unloading in marble is generally 10 – 20 m, and about 50 – 90 m for slate and low-grade metamorphic sandstone.

A large number of faults and fractures are developed in the left slope, as shown in Figure 4. The NE–SW- to NNE–SSW-striking faults form the most developed and extensive faults. The orientation of this group of faults is N30 – 50°E, SE∠60 – 80°, and these fault zones are generally 1 – 3 m in width (e.g. faults f5, f8, f2 and lamprophyre dyke X). Other faults striking NE–SE to east–west, such as fault f42-9 (with an orientation of east–west, S∠40 – 60°), are less well developed. There are three main sets of joints in the rocks of this slope. The first set includes bedding joints with an orientation of N15 – 35°E, NW∠30 – 50°. The second set is oriented at S–N30°E, SE∠60 – 80° and extends over 10 m. The third set has an orientation of N50 – 70°E, SE∠50 – 70° and a common extent of 5 – 10 m. Most of these joints are fresh and free of fillings. In addition, deep-seated tension cracks are widely developed in the subsurface of the slope, to 100 – 200 m depth, as shown in Figure 3. It can be observed that the geological conditions at the dam site are extremely complex, and lead to poor stability conditions of the left abutment slope.

From the spatial distribution and intersections of the weak discontinuities in the left abutment slope, it can be concluded that the overall stability of the slope is mainly controlled by a large potential failure block bounded by fault f42-9, lamprophyre dyke X and deep-seated tension crack SL44-1. Figure 5 illustrates a stereonet showing the orientations of f42-9, SL44-1, X and the slope surface. It can be observed from Figure 5 that wedge failure from the intersection of f42-9 and SL44-1 is kinematically feasible. The total volume of the large block is about 1.08 × 10^{6} m^{3} after the completion of excavation. To increase the stability of this large block, three shear-resistance tunnels at elevations of 1883, 1860 and 1834 m were constructed along the trend of fault f42-9. The cross-sectional size of the shear-resistance tunnels is 9 m × 10 m, and they were filled with steel-reinforced micro-expansion concrete.

## Monitoring scheme of slope deformation

A monitoring system was designed and installed to examine the deformation of the large potential failure block and the discontinuities that form this block. Depending on the depth of the monitoring instruments and monitoring purposes, the slope monitoring scheme can be divided into three main components: surface deformation monitoring, shallow slope deformation monitoring and deep-seated tension crack deformation monitoring. With this scheme, the deformation and stability of the high slope were monitored at three depth levels from the surface to the interior of the slope.

### Deformation monitoring of slope surface

For deformation monitoring of the slope surface, the geodetic surveying method is used. Eighty surface observation piers were installed on different levels of berms and the cutting edge of the slope (see Fig. 6). Of these, 10 piers are on the natural slope above the excavated zone, 27 on the excavated slope above 1885 m elevation, and 43 on the excavated slope between 1720 and 1885 m, forming eight longitudinal monitoring sections. The arrangement of the surface monitoring points aimed to monitor the deformation of both walls of fault f42-9, lamprophyre X and the deep-seated tension crack SL44-1 exposed on the excavation surface. The instruments for the slope surface monitoring are TCA2003 total stations, and the following three components of deformation are observed: along the river (referred to as the *x*-direction), across the river (referred to as the *y*-direction) and in the vertical direction (referred to as the *H*-direction).

### Deformation monitoring in the shallow rocks of the slope

Multi-point extensometers were employed to monitor the deformation of rocks and sliding of local blocks induced by blasting and stress relaxation in the slope at shallow depth (within 80 m of the excavated surface). A total of 54 sets of multi-point extensometers were arranged in the left bank slope, of which 12 sets are located above 1960 m, numbered –, 15 between 1885 and 1960 m, numbered –, and 27 below 1885 m, numbered –. The layout of the multi-point extensometers is illustrated in Figure 7.

### Deformation monitoring in the deep rocks of the slope

To monitor the deformation of rocks and cracks in the deep parts of the slope, graphite rod extensometers were installed along the axial direction in some geological exploration adits and drainage tunnels. Four sets of graphite rod extensometers were installed in PD42 and PD44 at an elevation of 1930 m, as illustrated in Figure 8. Another set of graphite rod extensometers was installed in the drainage tunnel at an elevation of 1985 m.

## Analysis of monitoring data

### Analysis of surface deformation observations

The excavation of the left bank slope started in September 2005. In December 2006, the slope was excavated to the platform of the cable crane at an elevation of 1960 m. In June 2007, the excavation reached the crest of the dam at an elevation of 1885 m. The slope was excavated to the concrete seat cushion of the dam at an elevation of 1730 m in August 2008. The whole excavation was completed in August 2009 when the slope was excavated to an elevation of 1580 m.

To examine the correlation between the deformation of slope surface monitoring points and the excavation process, a statistical analysis was conducted on the surface deformation monitoring results. Figure 9 shows the variation of deformation (in the *y*-direction) at some typical surface monitoring points with time. We found that the deformation on the slope surface can be divided into three stages. In the initial excavation stage before May 2006, the slope deformed at a low rate, because at that time the excavation involved only cutting of the rock mass above the cable crane platform, with a small volume of excavation. After May 2006, the slope was excavated rapidly and the main excavation stage began. Therefore, the deformation rate at the surface monitoring points increased dramatically. After August 2008, the excavation of the dam foundation rock mass below the concrete seat cushion started and by then the excavation of the main body of the slope had been basically completed. Hence the slope deformation rate slowed down and showed a tendency towards asymptotic values. Overall, the deformation of the left bank slope shows a good correlation with the excavation process, indicating that the excavation is the main reason for inducing the slope deformation.

To analyse the global deformation tendency of the left bank slope, the distribution of displacement vectors at some typical monitoring points after the excavation was completed is illustrated in Figure 6. It can be seen that the *y*-component (across-river direction) of the deformation on the slope surface is the largest, the *x*-component (along-river direction) the smallest, and the *H*-component (vertical direction) shows intermediate deformation. The excavation-induced deformation of the left bank slope in the *y*-direction mainly points towards the free surface of the slope with a maximum accumulated displacement of 73.7 mm. In the vertical direction, the deformation mainly shows subsidence with a maximum accumulated displacement of 56.8 mm, whereas in the direction along the river, the deformation mainly points upstream with a maximum accumulated displacement of 47.7 mm.

At the surface monitoring points located in the toppling deformation zone above an elevation of 1990 m, the accumulated total displacement was up to 80 – 93 mm according to the observation results between December 2005 and October 2009. The main reason for this is that the toppling deformation zone is strongly weathered and unloaded. Thus fractures are well developed and the quality of rock mass is very poor, which leads to a relatively large deformation induced by the excavation.

To find out whether the large block has discontinuous deformation along the lamprophyre dyke X, we compared the displacements at monitoring points located in the hanging wall and footwall of this dyke. It can be observed from Figure 6 that points TPL-9, TPL-10 and TPL-14 are located in the hanging wall of the lamprophyre dyke X, whereas point TPL-15 is in its footwall. The readings of the four monitoring points started at the same time. We can observe from Table 1 that the horizontal displacements and the total displacements at the three points located in the hanging wall of the lamprophyre dyke X (i.e. points TPL-9, TPL-10 and TPL-14) are slightly larger than the displacement at point TPL-15 located in the footwall. The displacement difference varies from 4 to 10 mm. The monitoring results in Table 1 demonstrate that the slope excavation implies an unloading and stress-relief process, leading to relaxation deformation along the lamprophyre dyke X.

### Analysis of shallow deformation observations

Fifty-four sets of multi-point extensometers were placed in the excavation slope of the left bank dam abutment. The deepest measurement points of the extensometers are located at 48.7 – 89.5 m depth from the excavation slope face. When the excavation of the slope was completed in August 2009, the displacement at the top points of the multi-point extensometers varied within the range of −2.37 to 20.11 mm. A few of the extensometers showed negative values of the top point displacement (deformation direction was toward the interior of the slope). Figure 7 shows the distribution of displacements at the top points of the multi-point extensometers after the excavation was completed. It can be observed that the shallow deformation of the slope section between elevations 1885 and 1960 m is generally greater than that of the section above elevation 1960 m, although the observation time of the extensometers installed in the slope above elevation 1960 m is longer. The main reason is that the installation depths of the extensometers at elevations 1885 – 1960 m range from 68.7 to 89.5 m. These extensometers can survey the deformation of deeper rocks than the extensometers above elevation 1960 m, which have an average depth of 48.7 m.

There are 11 sets of multi-point extensometers drilling across the discontinuities forming the large block, as shown in Table 2. To analyse the deformation of the discontinuities during excavation, we list the deformation differences of the extensometer measurement points located in the hanging wall and footwall of the discontinuities in Table 2. It is shown that the displacements at the top point of the six extensometers are relatively large, and there is an obvious deformation difference between the measurement points across the discontinuities. This deformation is mainly induced by the opening or dislocation of the discontinuities. We can conclude that the main discontinuities show unloading relaxation deformation during the excavation, but no obvious sliding deformation occurred. After the excavation was completed, the deformation of the slope tended to be convergent.

Figure 10 shows the deformation at the measurement points of extensometers and located at the 2–2 monitoring section above elevation 1960 m. We can see that after was installed crossing the lamprophyre dyke X, its top point displacement increased to 9.21 mm by August 2009 when the excavation was completed. This deformation mainly resulted from opening between the hanging wall and footwall of X. The displacements of the other two extensometers and are relatively small (less than 1.1 mm). Figure 11 shows the displacement variations at each measurement point of during the excavation process. The deformation process at these points can be divided into three stages, which is similar to the observations at the surface monitoring points of the slope. After the excavation of the main body of the slope was basically completed in October 2008, the displacement variation of tended to a convergence state.

### Analysis of deep deformation observations

To survey the deformation of deep-seated tension cracks in the left bank slope, two sets of graphite rod extensometers were installed successively in adit PD44 at elevation 1930 m. The first installed set is named G44 and the later installed set is named G44X. Observation of G44 started on 6 September 2007. The total deformation at each measurement point is shown in Figure 12, and the variation of deformation with time is shown in Figure 13. It can be seen from Figures 12 and 13 that inside adit PD44, the deformations at segment 0 + 76.5 – 0 + 122.5 (measurement points 3–4) and segment 0 + 122.5 – 0 + 151.1(measurement points 4–5) are relatively large. It can be concluded that the deformation of the rocks in adit PD44 mainly occurred within segment 0 + 76.5 – 0 + 151.1. The accumulated total displacement at the top point of G44 increased to 39.53 mm when the excavation finished in August 2009.

It can be observed from Figure 12 that G44 is installed across the lamprophyre dyke X, the deep-seated tension crack zone and fault f42-9 successively. The lamprophyre dyke X is located in segment 0 + 76.5 m–0 + 122.5 m, whereas fault f42-9 is in segment 0 + 122.5 m–0 + 151.1 m. The deep-seated tension crack zone, which is classified as class IV2 rocks and has poor integrity, has been well developed between these two discontinuities. During the excavation of the slope, relaxation or tensile deformation developed along the pre-existing deep-seated tension cracks. Because the excavation height of the slope is up to 530 m and the scale of excavation is very large, the unloading effect induced by excavation is intense and the unloading zone is large. According to the monitoring data, the depth of deformation in the slope rocks reaches over 150 m. The large block formed by fault f42-9, the lamprophyre dyke X and the deep-seated tension crack SL44-1 is mainly located in segment 0 + 12.7 m–0 + 76.5 m, where the deformation of the large block is relatively small. The deformation measured by the graphite rod extensometers mainly reflects the relaxation or tensile deformation along the deep-seated tension crack zone owing to excavation unloading.

To further verify the deformation in the deep-seated tension crack zone, the second set of graphite rod extensometers (G44X) was installed in adit PD44 in July 2008. The displacement at each measurement point of G44X is shown in Figure 14. Table 3 shows a comparison of the monitoring results from these two sets of graphite rod extensometers during the same period. We can observe that the distributions of the rock-mass deformation recorded by these two extensometers are basically identical. Therefore, the deformation of the slope rocks in adit PD44 mainly developed in segment 0 + 76.4 m–0 + 152 m, reflecting the opening of the deep-seated tension cracks owing to excavation unloading.

## Numerical simulation and analysis

The understanding and prediction of the deformation behaviours of a high rock slope largely rely on comprehensive investigations combining geological surveying, site monitoring and numerical modelling. To further explore the deformation mechanism of the left abutment slope during excavation, we employ FLAC3D software for modelling the deformation of the left abutment slope induced by sequential excavation (ITASCA Consulting Group 2002).

### Three-dimensional model setup

For numerical simulations, a 3D numerical model for the left abutment slope was created, as shown in Figure 15. The slope model is 620 m long in the water flow (*x*-) direction, 750 m wide in the transverse (*y*-) direction and 1370 m high in the vertical (*z*-) direction. Anisotropic laminar layer elements were used to simulate the discontinuities, such as the lamprophyre dyke X, faults f2, f5, f8 and f42-9, and the deep-seated tension cracks SL44-1–SL44-9 and SL54-1. The slope model contains 54 345 nodes and 259 168 elements.

It is very important to estimate the mechanical parameters of the rock mass and discontinuities for the stability analysis of the slope. The rock mass of the left bank slope was first classified into four grades according to the RMR system; that is, grades II–V, with grades III, IV and V further divided into III1 and III2, IV1 and IV2, and V1 and V2. Then, by selecting representative positions, 97 points for rigid bearing plate tests were chosen for all types of rock mass. Using the deformation parameters obtained from the tests, a scatter diagram was plotted corresponding to the rock mass of different grades. By excluding outlying data, the average value for a point cluster segment was taken as the recommended deformation parameter. The strength parameters of the rock mass and discontinuities were determined by *in situ* testing using large-scale direct shear. Fifty groups of *in situ* shear tests have been carried out, with 27 groups for rock-mass specimens and 23 groups for discontinuity specimens. On the basis of the results from direct shear testing, the predominant slope method was adopted to determine the strength parameters of the rock mass and discontinuities. From the recommended value of the dam foundation rock-mass classification at Jinping-I hydropower station (Zheng *et al.* 2003), the mechanical parameters of the rocks and discontinuities are listed in Tables 4 and 5, respectively.

According to the measured data and the geological conditions at the construction site (Zheng *et al.* 2003), the initial stresses in the slope before excavation are calibrated as , and , where *h* is the vertical distance from the ground surface in metres (with the stress in MPa). Staged excavations were simulated, with all relevant boundaries being constrained in the normal. An elasto-plastic model with a Mohr–Coulomb failure criterion was adopted to simulate the deformation behaviours of the slope rocks. Bolt, cable and beam models in FLAC3D were used to simulate the supporting structure of the slope, including rock bolts, prestressed anchor cables and concrete grid beams. Because the groundwater level was low in the left bank slope during construction (Fig. 3), the effect of groundwater is not considered in the present study.

### Simulation results and analysis

FLAC3D software was used for the simulation of the stages of sequential excavation of the left bank slope, based on the actual construction procedure. Figure 16 shows comparisons of the calculated and measured displacements (in the *y*-direction) at some typical surface monitoring points after the completion of excavation. It can be observed from Figure 16 that the calculated results agree well with the measured deformation. This demonstrates that the 3D numerical model and the selected values of mechanical parameters of the rocks and discontinuities are reasonable.

Figure 17 shows the distribution of the predicted displacement in the representative profile II1–II1 (marked in Fig. 2) after the excavation. It can be seen that the direction of the deformation induced by the excavation disturbance points towards the free surface of the slope, which is nearly identical to the deformation measured using theodolites and level instruments. Moreover, the deformation depth in the slope is over 150 m owing to a dramatic release of stress induced by the large-scale excavation, far greater than that of general hydropower engineering slopes. To further explore the deep deformation mechanism of the complex rock slope, Figure 18 shows the plastic yield zones in the II1–II1 profile obtained from the 3D simulation. After the slope excavation was completed, plastic yield zones were developed along the lamprophyre dyke X, faults f42-9 and f5, and the deep-seated tension crack zone. Because the intactness of the rocks in the deep-seated tension crack zone is very poor, this zone reached the plastic yield state during the excavation process. From the numerical results, the deep-seated tension crack zone disturbed by excavation unloading was estimated to have a horizontal thickness of 152 m, which is comparable with the field measurements.

### Evaluation of slope stability

The stability of the left abutment slope after the completion of excavation is evaluated by an interactive visualization tool named Slope3D (Jiang *et al.* 2011, 2016). For comparison, Table 6 gives the factors of safety of the left bank slope for various reinforcement measures, which indicates that reinforcement obviously improves the stability of the slope. It can also be seen from Table 6 that the combination of various reinforcement measures including shear resistance tunnels and anchor cables is essential to satisfy the design criterion (safety factor, *F*s = 1.3).

To further verify the safety of the left abutment slope, the strength reduction method (Zheng *et al.* 2005; Griffiths & Marquez 2007) has been incorporated into the FLAC3D program and adopted to evaluate the stability of the slope after the completion of excavation. We can obtain the factor of safety of the slope with an equal proportional reduction of the shear strength parameters of the potential failure surfaces. Because the overall stability of the slope is controlled by the large block, the potential failure surfaces consist of the lamprophyre dyke X, faults f42-9 and the deep-seated tension crack SL44-1.

A completely connected plastic yield zone is used as the criterion for slope failure. When the slope reaches the failure state, the strength reduction factor is taken as the safety factor of the slope. Illustrated in Figure 19 is the evolution of the plastic yield zone with the reduction factor. With increase in the reduction factor, the plastic yield zone along the potential failure surfaces expands gradually. When the reduction factor approaches 1.4, the plastic zone of the failure surfaces is completely connected. Therefore, the safety factor of the excavated slope is regarded as 1.4 after the completion of the excavation and reinforcement measurements, which is basically identical to the results obtained by the 3D limit equilibrium method.

## Concluding remarks

The stability of high rock slopes is one of the most important issues during construction of a hydropower station, as it is critical to the safe construction of the dam and the success of the project. The stability problem of the left bank slope of Jinping-I hydropower station is predominant owing to the poor geological conditions, large-scale excavation and deep unloading features. For safe construction of the slope, the deformation of the complex rock slope was monitored at three depth levels from the surface to the interior of the slope. The deformation mechanism and stability of the slope induced by the large-scale excavation were evaluated by the 3D limit equilibrium and finite-difference methods. The main conclusions are summarized as follows.

The deformation of the high rock slope shows a good correlation with the excavation process, indicating that excavation is the main reason for inducing the slope deformation. Because the toppling deformation zone above an elevation of 1990 m is strongly weathered and unloaded, relatively large deformation induced by the excavation occurred in this zone of the slope.

The results from deformation monitoring in the shallow rocks of the slope show that the discontinuities (f42-9, X and SL44-1) forming a large block experience relaxation or tension deformation during excavation. The magnitude of the discontinuous deformation is less than 10 mm, and the deformation tended to be convergent after the completion of excavation, indicating that the large block is stable as a whole.

By comparison of deformation monitoring of the surface, shallow and deep parts of the slope, it is found that the deformation of the high slope during excavation mainly resulted from deep deformation. The deep deformation mechanism was interpreted by 3D numerical analysis. The deep-seated tension crack zone disturbed by excavation unloading was estimated to have a horizontal thickness of 150 m.

The stability of the high rock slope was evaluated by the 3D limit equilibrium method and the strength reduction method by FLAC3D. The results from the two methods agree well, which indicates that the factor of safety of the excavated slope is over 1.3 (the design criterion) under normal conditions after the completion of the reinforcement measures. Therefore, the left abutment slope is stable and safe.

## Acknowledgements and Funding

The work reported in this paper has received financial support from the National Natural Science Foundation of China (No. 51179137) and the National Basic Research Program of China (No. 2011CB013502). This support is gratefully acknowledged.

- © 2016 The Author(s)