Theoretical study of the Thuong Kon Tum tunnel, by: PABLO PANTEGHINI
From the very moment I was described the complicated situation of the Thuong Kon Tum project, it caught my attention and I was passionate from the beginning. Perhaps because I had fresh memories of the many hours spent inside a tunnel boring machine, on how suddenly used to arise inside me the need of “failure without resolution” every time that the excavation stopped for different reasons. Once again being called for duty, and once again applied myself fully to solve it. A technical fault was always within my reach; I was accustomed, but too many times the nature of the fault was not such, and I used to feel it as a reflection of inability. Not being able to understand the geology or the behavior of the ground used to disappoint me but fascinated me. It was a gap between engineerings absolutely insurmountable.
The ingredients of a great challenge were served in Thuong Kon Tum.
Tunnel predominantly in granite massif with virtually no aggressive curves, and with steep slopes. The optimal and most reasonable choice of the TBM type Main Beam for the hardest rock excavation.
However, something unexpected happens. The cards are shuffled again and the elements that had previously been forgotten become part of a more complex equation. Probes, rain, runoff, materials, essays, geologic-cores, overburden, support, lining, convergences, temperature, boreholes, everything. Most of the longest Vietnam’s tunnel of 17500m had been excavated with some difficulties, a few important water inflows but without major geological accidents, and with impressive production standards. Only 250m were missing to the target, but the ground did not agree. Simply exciting.
LOCATION AND SCHEME
The Thuong Kon Tum hydroelectric power tunnel had its entrance in the Central Vietnam area near the city of the same name. It has 4.53m in diameter and a total length of 17480m of which 13,000 were excavated with a TBM and the rest with conventional methods.
The stratigraphic profile of the tunnel from pk3000 to pk 6600 was: 65% rock class II; 15% class III rock, 10% class I rock and 5% class IV and V rocks respectively.
- Seismic intensity: I_max ＜ 6 (MSK-64)
- Average annual flow: 16.2 m3 / s
- Annual water runoff: 510.88 × 10 ^ 6m3
- Average annual rainfall: 2800 mm
- Average annual runoff depth: 1366mm
- Annual sediment runoff of 56.9 x 10 ^ 3m3
GEOLOGICAL STRUCTURE OF THE Thuong Kon Tum TUNNEL
The analysis is made specifically in regard of the 9 km upstream slope tunnel headrace’ section
• The rocks were mainly biotitic granite in medium grain particles, massive, dense and hard structure.
• The existence of 18 class IV failures and between 24 and 29 class V failures are highlighted.
• Most faults were almost orthogonal to the tunnel axis and have a steep angle of inclination (75-85 °).
• The faults were plotted based on surface recognition and physical examination, without any boreholes verification
• The width of fractured areas decreases with the increasing depth.
• Important to note that none of the failures detected during the design phase did not match minimally with those that were found during the excavation phase, fact that spoke of the inappropriate geological survey campaign.
• The surface rocks along the tunnel layout were quite eroded, and the characteristics of weathered and thickness vary with the composition of the rock, the shape of the terrain and the location.
• In general, the groundwater level along the supply tunnel was quite low (6 ~ 28.1m aprox.).
The TBM had to be retracted to improve the support regime in which a collapse occurred and a cavity formed in front of the machine (F6).
• The collapse did not extend below the TBM.
• The presence of sand and clay, of pulverized material (gap) and mica was recorded.
• The natural surface in the area was totally inaccessible and unexplored.
• There was no significant presence of water.
• A 20 m horizontal core survey was carried out from the TBM and it was discovered that the fault extension was up to 8 meters.
• This area was exposed for approximately 6 months and there was a continuous collapse of up to 22m above the key (F7).
TBM MB Robbins 1612-285-2: action and reaction against the ground
Drilling and boreholes
Steel ribs’ Ring Beam
Erector for wire mesh mounting
Mobile arm to apply shotcrete
PRE-SIMULATION CALCULATION TOOLS
• Protodyakonov parable calculation approach
The phenomenon of instability in the excavation front that happened in Thuong Kon Tum was progressing and affecting an increasing volume of rock. Finally it reached a state of equilibrium, but at the cost of the formation of a “bell” above the key and a breakage of the nucleus by the plane of maximum shear.
The principle of Protodyakonov roughly what it does is:
- To consider a tri-articulated parabolic arch working with compression.
- To propose the balance of forces, compensating the vertical and horizontal loads by means of the “f” factor.
- To find the highest stable height “h” that the ground can develop. Assuming that the coefficient of strength in the granite of the Thuong Kon Tum tunnel is f = 17, performing the calculations you get:
This is the previous step to understand at what moment a deformation of 14cm led to a loss of ground of more than 1350 m3 of rock and why.
•RMR system used on site (F9)
By: RMR_89 = R1 + R2 + R3 + R4 + R5 + R6
•Q system used on site (F10)
• Calculation of Hoek-Brown breakage constants
The current edition of the Hoek-Brown criterion is expressed through the following equation:
where m_b is the reduced value of the characteristic parameter of the intact rock, defined by the relation:
GSI is the index of geological structures; it is the value of the monoaxial compressive strength of the intact rock while s and a are constants of the rock mass described by the equations:
With D = 0 since the tunnel was being excavated with a TBM and for a GSI = 90 The calculations are solved as follows:
That is the data that I will use later for the analysis of granite material with the specialized software.
The resistance of a fractured rock mass depends on the properties of the pieces or blocks of intact rock and, also, on their freedom to slide and rotate under different stress conditions.
THE TBM, THE GROUND AND ITS INTERACTION
On how TBM affects the ground at the fault crossing
Situation 1: Brittle Failure
The effect of slab detachment that can vary in thickness with large openings is explained. The progression of the fissure results in the formation of a “V” shaped notch that deepens as the tunnel progresses until steady state conditions are reached (F11).
The effects of this condition are:
- Reduction of grain size.
- Dramatic drop in resistance compared to the source rock.
The concept of the trilinear failure criterion to capture the non-constant dependence on the strength of the rock mass in the confinement of materials that are prone to disintegration is explained (F12). The contours of the radial or confinement strain are almost parallel to the excavation geometry. As the rock deforms, the cohesive bonds fail and the friction resistance component develops at a different speed. The cohesive force is lost to the extent that the fragile rock “disintegrates.”
A bi or trilinear failure criterion is required to capture this non-constant dependence on the strength of the rock mass in confinement in materials that are prone to disintegration. Tensile shedding occurs when the tension path moves above the damage threshold and to the left of the shedding limit. The damage thresholds and the shedding limit depend on the scale, the type of rock and the characteristics of the rock mass.
The trilinear shell explains the dominant processes of damage initiation and fracture propagation when limiting tensions are relaxed (in the inner layer to the left of the shedding limit). The transition limit between the extensional failure and shear cutting processes is represented by the shedding limit, defined by a constant stress ratio of σ_1 / σ_3
Situation 2: CHD Detention
The detention explains the triggering of an extrusive movement of the ground, across the front, and the consequent pre-convergence that in turn would have unleashed the exceptional convergence of the cavity becoming uncontrollable by mere radial consolidation.
Depending on the various possible stress-strain states, the behavior of the ground in the front can be traced to approximately three different situations: stable core (elastic field), stable short-term core-front (elastoplastic field), unstable core front (breaking field).
Situation 3: CHD Withdrawn
When a cutting head is removed from a fault zone for post-treatment, a loosening effect can occur.
There is therefore the probability of a problem in the grippers area due to the delayed treatment of the rock that really required a previous treatment.
Ultimately the extremely adverse conditions could have been caused by a combination of erosion of the defective material and on the other hand the greater formation of voids could have occurred due to sub-parallel fault orientations as recorded in past works with similar situations.
Situation 4: Main Beam Performance Factors
The parameters of the Main Beam, thrust and power, together with the properties and characteristics of the rock mass, are the main parameters used to estimate the performance of TBM.
The orientation of the joints and the faults, together with the direction of the advance of the machine, must be quantified for the estimation of the performance.
On how the ground affects the TBM in the fault zone
Situation 1: Granite-MB Interaction
In order to understand the ground-tunneling relationship and the effects that the rock had on its normal excavation process in Thuong Kon Tum, I based the investigation on previous studies of the analysis in a three-dimensional model of dynamic finite elements of interaction between the cutter and rock; through which real thrust and torque are calculated in the tunneling process with the Main Beam.
Through this study results are obtained such as the displacement distribution throughout the main Beam structural system (F18);
Or the strain distribution on the surface of the Cuttehead (F19).
An important strain occurs at the junction of the thrust cylinders with the main beam due to the existence of an inspection well. On the other hand, due to the moment provided by the traction motors, the strain in the support of the Cutterhead is distributed in radial direction.
Situation 2: Updated RMR
RMR_14 maintains a structure similar to the RMR_89. Additionally, apart from the correction factor for the tunnel orientation, two new correction factors have been included: one for the case in which the excavation is mechanically performed (TBM) and another that takes into account the effect of the plasticization of the front of the tunnel in which the RMR is determined.
Situation 3: Updated Q
The analysis of the five typical “lines” of performance of an MB is considered. The curves (and crosses) of “unexpected events” are the worst cases associated with low Q value (F23).
It is highlighted how the deceleration gradients (-m) are directly related to the normal Q value when Q <1. The application of the Q value to estimate the slope is shown in TKT. The potential length and penetration rate become critical when the exponent of the equation increases (F24).
Situation 4: Resemblance OSO Tunnel—TKT Tunnel
These graphs (F25-F26) show the comparison between world record performance of the Bear Tunnel exaveted in 1966/1967 and the extraordinary performance of the Thuong Kon Tum Tunnel excavated in 2018/2019 over 50 years later. Two unparalleled feats of engineering carried out by The Robbins Company although for both cases the works stopped for several months before they could be completed using different solutions.
In general, the failure rate will be proportional to the relationship between the effective resistance of the rock and the magnitude of the unbalanced rock strain.
3D SIMULATION WITH MIDAS GTS NX: PRE-PROCESSING
The construction of a tunnel, as a rule, corresponds to the static problem of an elastoplastic solid subject to boundary conditions and initial conditions.
As the practical case tunnel in the jobsite is presented on a spatial domain, after studying and understanding the site geological situation I define a geometry on which I look for the solution to the contour problem. Apart from defining the space occupied by the ground, as the structural elements also occupy space, then such space it must be defined as well.
The next step would be to consider the boundary conditions that the practical case establishes: horizontal ground surface, uniform overload and the TBM circular section tunnel.
The system of equations to solve is formed by equations of balance and constitutive (solid elastoplastic). The balance equations remain defined by the problem to be solved, in this case a static problem in three dimensions under the hypothesis of flat deformation subject to the action of the gravity and without dynamic terms. The constitutive equation governs the strain response and complete the system of equations to solve. With the data provided from the jobsite I consider a Mohr-Coulomb perfect elastoplastic constitutive model for the faulted material and a constitutive model with breaking criteria governed by Hoek-Brown’s laws for the granitic massif.
For this scope it is important to determine the stress state of the ground before simulate the constructive process, that is, the initial conditions.
The finite element method is a procedure to replace the continuous description of a problem, given by a system of equations, by one discrete in which the solution is obtained only in a finite number of points in space. For this it is essential to subdivide the spatial domain through the finite element mesh.
I establish a geometry, including the spatial domain as well as the structures considered, the boundary conditions required, the materials used, the finite element mesh and the initial conditions.
Advancement and simulation procedure of the Main Beam
• In the first step, the initialization is carried out taking into account the voltage field in situ;
• In the second step, the Main Beam enters the model, the first section of the Roof Support is activated and the first cut is excavated
• In the initial three steps, equivalent to approximately 6m in length of the Roof Support, the Main Beam advances towards the model;
• In the fourth step the first cut of granite ground is already excavated without support;
• In the eighth step, the Main Beam enters the fault zone and at the same time the grout filling is activated by means of the respective change of boundary condition;
• In the tenth step the Roof Support element is deactivated;
• In the fourteenth step, the cutting of granite ground without sustaining after 20m of excavation is deactivated;
• In the eighteenth step the grout element is deactivated;
• The stages continue like this until it reaches virtually 4m of failure.
3D SIMULATION WITH MIDAS GTS NX: POST-PROCESSING
3D virtualization of the model
I have elaborated a three-dimensional digitization of some graphics that I have found important and illustrative. I have extrapolated them from the global post-processing of the Thuong Kon Tum tunnel carried out with Midas. At this point I have exported them in pdf format for better appreciation thanks to a multimedia and 3D tool (F33) program.
3D simulation with Midas
Tensions of solid elements tunnel and ground. Initial conditions of tensions of granite ground and fault zone respectively with the excavation front in the granite massif 16m from the fault zone (F35). Verification of ko and state of tensions at rest.
Initial conditions of thrust on the roof support of the Main Beam before starting to bore (F36).
For hard rock granitic massifs of very good geotechnical quality, the analysis of fracturing in parallel faces (spalling) in excavations under high stress suggests that the rock mass behaves in an elastic and fragile manner, as shown in F34 . When the resistance of the rock mass is exceeded, a sudden decrease in its resistance occurs.
The analyzes carried out of the progressive failure of rock massifs of very poor geotechnical quality that in the vicinity of tunnels such as the one found in Thuong Kon Tum, suggests that the post-failure characteristics of this type of rock mass are adequately represented by assuming that the massif behaves perfectly plastic. This means that it continues to deform under a constant level of effort, and that no change in volume is associated with this progressive failure (F37).
Detail of the ground tension XX with a distortion factor of 0.06 in longitudinal section to highlight the deformation effect when the Cutterhead of the Main Beam was 8 m from the Fault (F38). For the detailed analysis of the tensions in the different stages of the machine I have made a double cutting plane to facilitate understanding.
Total flat translation detail TZX. (F39). In this first phase of a sequence, the displacement deformation diagram achieved by making a transverse planar section along the entire block can be seen. As the Main Beam approaches the fault zone, the displacement increases until it abruptly increases as soon as the excavations resume. Here the delta of displacement is still practically null.
Total tunnel displacement
Tunnel condition just before the fault zone; views with deformation and with plane of longitudinal section at 25m (F40). Associated vector simulation. The vector option shows the size and direction of the selected displacement or reaction force component as a vector in each node.
The vector graphic is also affected by the “Cut and Cut Plane”.
Detail of the tension situation of the excavation front (F42).
Diagram of deformation by displacement in phase 2 (F43).
YY bending moment of the roof support (F45). Deformations in longitudinal section with vector simulation detail (F44). Von Mises solid tension; deformation with factor 0.4 and Roof Support detail (F46).
In this image (F47) some ground conditions are extremely detailed: the vectorization, the flow lines, the reaction forces and the plasticization surface before the collapse.
Ground conditions: vectorization, flow lines, reaction forces and plasticizing surface just at the moment when the collapse is triggered (F48). Detail of the situation Von Mises Tension of solid; deformation with factor 0.4 (F49). Detail of the Roof Support and the grout injection further back after entering the fault zone (F50).
YY bending moment of the Roof Support (F51) and phase shift deformation diagram (F52).
I will stop at this point to conduct a deeper analysis of what is happening at this time of excavation. When retaking the excavations and entering the fault zone I have done the deepening on the one hand by means of quantified histograms with the data obtained in the simulation, visualizing the characteristics of the deformation through a virtual “walk” along the simulation frozen at the S8 stage. This flight simulation (F53A), being very visual, allows me to be a little clearer when it comes to explaining the concepts of tense deformative behavior of the ground during which I consider the most critical phase of the collapse in Thuong Kon Tum and the events that were triggered a posteriori as a direct consequence of this situation.
From this graph (F 54) I point the displacement <1mm in the excavation front during the excavated stage before reaching the fault zone. After an abrupt preconvergence in the stop time of the TBM, the travel values increase up to 95mm after resuming the excavation and entering the fault zone.
Although the machine did not actually make more than two advances, the pattern of gradual deformation is recorded but constant even during the time it would have taken the grout to reach the collapsed area. The total breakage process is also noted in some of the nodes under study.
Graph F55 shows the inflection point when entering the fault zone also in terms of stresses. A strong generalized strain situation almost abruptly cancels out as indicated by the values associated with the nodes in the excavation front plane. He had previously talked about the delicate situation that can be generated after stopping the excavation front in a situation of very high tensions due to the large covert. Specifically, he had exposed the behavior of the ground in front of the excavation front in function of the various possible stress-strain states and with the elimination of the minor main tension XX as a result of the advance of the front, there is a risk of generating situation C (F53B): unstable core front field of breakage or collapse. We see how the deformation response in the field and the deformation in code coincide with the previous theoretical study, resulting in the unstable behavior of the excavation core-front.
PLASTIFICATION OF THE GROUND ON THE EXCAVATION FRONT IN FAULTED AREA
Here I explain the sequence of plasticization of the ground in the excavation front with detail of plastic failure, the strain failure and the cyclic strain-stress process of discharge / recharge.
Side view of the model in the excavation front with visualization of the plasticizing zone just at the moment the TBM enters the fault zone (F56).
Diagram of horizontal displacements at different distances from the excavation front (F57).
The plasticization of the nucleus-front and the channeling of the flow lines of the horizontal stresses give rise to a spatially evolutionary phenomenon that originates an inwardly extrusive mechanism evidenced by the vectors.
In the case of the natural core in Thuong Kon Tum, in which it is a core without reinforcement or pre-support, the analysis of the numerical results has clearly shown that the progressive cancellation of the containment pressure in the front of the Main Beam, which simulates the excavation, leads, under given conditions, to the plasticization of the core face (F58) and the consequent channeling of the horizontal strain flow lines (F59) that corresponds to a progressive increase in the normal pressures exerted by the ground on the covering for an area of length equal to approximately an excavation diameter.
In particular, the evolution of an extrusion mechanism is highlighted, in which the ground in a band around
The key flows to the digging front. This is a spatially evolutionary phenomenon in which the plasticization of the ground around the tunnel key continues to extend in the longitudinal direction behind the front.
Actually, when moving forward, the plasticization of the ground around the cavity, which is an irreversible phenomenon, will affect the entire length of the tunnel. The above is also evident in terms of the characteristic curve. In fact, if the characteristic curve of the front is normalized, it shows how a vertical asymptote never reaches: its rigidity is progressively reduced without ever becoming null, showing ductile behavior with a resurgence that does not depend on the constitutive bond, but consistent with the resurgence due to the spatial propagation of the plasticized area.
Standardization of the characteristic curve of the excavation front (F60).
Plastification values of the excavation front inside the fault (F61) intuitively create an envelope that matches the normalized curve.
3D PLAXIS ANALYSIS OF THE SAFETY FACTOR IN THUONG KON TUM
In this section, I perform the calculation of the global safety coefficient against ground resistance (c, ⱷ) before the fault crossing in this section of Thuong Kon Tum Tunnel simulated with the PLAXIS 3D software and the analysis of the stability of the front excavation For these purposes, I will consider in a first phase that the roof support has reached its position within the block of the faulty ground described above. In this way, in a first stage (after the calculation of the initial tensions) I simulate the exact moment in which the excavation of the tunnel penetrates the fault, the behavior of the “Roof Support” of the Main Beam is detailed, the water pressures inside the tunnel are eliminated and the initial pressure will be applied to the front. In order to obtain the “characteristic curve” in the front. In a second phase I will consider that the pressure in the front is canceled. In this way, in this calculation phase, when the pressure decreases gradually from its initial value, I will reach a value for which the collapse of the front into the tunnel takes place, thus obtaining the “characteristic curve” in the front and the front breaking load limit as an asymptotic value of this curve.
Introduction, description and methodology
• Phase 0: Self weight. Initial phase.
• Phase 1: Excavation of the tunnel, the installation of the roof support including friction interface, water pressures inside the tunnel will be eliminated and the pressure on the initial front will be applied.
• Phase 2: Gradual reduction of pressure in the front until it collapses.
• Phase 3: Calculation of the global safety coefficient, from Phase 1, through the reduction of the shear resistance parameters.
The Phi-c reduction procedure consists in progressively reducing the resistance parameters tanϕ and c of the soil until the structure breaks. The resistance of the interfaces, if active, is reduced in the same way.
Simulation and results
• Shear stress γ_s phase 2: where the most important distortions accumulate (F67);
• Phase 3 Displacements (F64);
• Phase 3 distortion (F66);
• Arc Effect (F65);
• Phase 3 global collapse (F68);
• Mstage development and Safety Factor curve (F69);
As the phase develops, the pressure in the front is decreasing and at the same time the displacements of the node A that is in the axis of the tunnel in the front are increasing. The asymptotic value that we find is 5000 KN / m2 which would be the breaking load on the front.
This is the way to obtain the characteristic curve of the front (F70).
Along this document I have briefly explained the main incidents that occurred during the construction of the last section of the Thuong Kon Tum Project tunnel in Vietnam, the possible causes of the collapse and subsequent formation of the void, the state of the art in terms of the excavation technology associated to a hard rock Main Beam, the terrain-machine interaction in a cause-effect system and the subsequent 3D simulation to complete the analytical research model.
On the basis of this case study the main conclusions are summarized below.
About the tunnel
- Excessive confidence in the tunnel experience
- Importance of anticipating stress states in weakened areas
- Identify the factors that determine a failure process
- Effective modeling of a possible fragile failure
- Practical implications of the transition from continuous to discontinuous behavior
About the TBM
- Take into account the strain routes around the TBM
- Of the plasticization in the front of excavation
- Main advantages of the Main Beam before this type of geological difficulties
About the Results Obtained
- A work guideline has been established that could be used in the future as a complementary study to the database in terms of tunnels in granite terrain with large coverts. Both at the level of correct use of the updated classification systems and the three-dimensional representation of a tunnel made specifically with hard rock TBM crossed by a fault zone.
- An appropriate construction phase simulation has been developed and in its essence I have been able to satisfactorily illustrate the problem in Thuong Kon Tum
- It has been possible to use the concepts referred to a real work that always increases the degree of difficulty.
As Richard J. Robbins once observed, it may be more feasible, and in fact it may become more realistic, to expect adaptable tunnel boring machines that can cope with the variable and unknown conditions of deep rock tunnels. to wait, in the same period of time, for the development of information prepared by the geotechnical fraternity that can be accurately interpreted by tunnel builders.
BARTON, N.R. TBM performance overview, and why the QTBM prognosis model. Oslo (Norway).
BARTON, N.R. (2016). TBM Tunnelling in sheared and fractured rock and the application of QTBM model concepts. Oslo (Norway).
BARTON, N.R. (2018). TBM Tunnelling under Difficult Conditions: Too Massive, Too Faulted, Too Wet, Too Deep. Oslo (Norway).
THE ROBBINS COMPANY. (2016). Service Manual MB1612-285-2 Rev2. USA.
CLARK, J.; CHORLEY, S; The Robbins Company. (2014). The Greatest Challenges in TBM Tunneling: Experiences from the Field. North American Tunneling Conference. 101-108. USA.
ROBBINS, R.J; The Robbins Company. The application of tunnel boring machines to bad rock conditions. Washington (USA).
THE ROBBINS COMPANY. (2017). Geological basis for the Claim on Kon Tum HEPP 27. Vietnam.
THE ROBBINS COMPANY. (2018). Kon Tum Project Cavity Traversing Methodology. Vietnam
CARTER, T.G. (2011). Himalayan Ground Conditions challenge innovation for successful TBM Tunnelling. Toronto (Canadá).
ZHAO, K; JANUTOLO, M y BARLA, G. (2012). A Completely 3D Model for the Simulation of Mechanized Tunnel Excavation. Department of Structural and Geotechnical Engineering, Politecnico di Torino. 475–497. Turin (Italy).
PLAXIS. (2019). PLAXIS 2D Reference Manual.
LINER PLATES. American Commercial Incorporated.
KAISER, P. (2010). Practical implications of brittle failure on hard rock tunnelling. Jornada Técnica “Túneles en granito”. 45-56. Barcelona (España).
KAISER, P.; AMANN, F.; STEINER, W. (2010). How highly stressed brittle rock failure impacts tunnel design. Canada.
MEIDONG Han; ZONGXI Cai; ChuanyoNG Qu; KUI Chen. (2015). Tunneling Simulation and Strength Analysis of Cutterhead System of TBM. ICIRA 2015, Part III, LNAI 9246, 445–455, 2015. Switzerland.
GAROGLIO, M.M. (2013). Modellazione agli elementi finiti in 3D (Galleria ferroviaria del brennero). Corso di laurea in Ingegneria Civile. Milano (Italy).
GEOCONTROL; CELADA; BENIAWSKY (2014). Actualización_RMR; publicado en Ingeopress n°234
LUNARDI, P. (2015). Il controllo dell’estrusione del nucleo di terreno al fronte d’avanzamento di una galleria come strumento di stabilizzazione per la cavità. Supplemento al Numero 114 della Rivista “Gallerie e Grandi Opere Sotterranee”. Milano (Italy).
OTEO, Carlos. Caracterización geotécnica en túneles (de la roca al suelo y viceversa)
HA THANH TÚNG (Vietnam Field Engineering). (2017). Engineering Geological Expansion Drawin