Stability of a Borehole during Horizontal Directional Drilling

Introduction

In the densely populated areas in the Netherlands pipeline-crossings beneath important waterways, railways, roads and residential areas are often unavoidable. These crossings are often installed using the horizontal-directional-drilling-method (HDD-method). In order to improve the cost-effectiveness of the technique, a group of Dutch contractors, engineering consultants and knowledge institutes are working together to enhance the understanding of the geotechnical and hydraulic mechanisms that play a role in horizontal directional drilling. One of the issues being examined is the stability of boreholes during and after drilling and installing the pipeline; this is the subject of this present paper.
This study focuses on three issues. Firstly, it is known that arching in the surrounding soil contributes to the stability of the borehole to some extent. However, the influence on the process of arching of different overburden cover-height/borehole diameter (H/D) ratios and the presence of cohesive soil layers is not well understood. Secondly, it is considered that the magnitude of the load from the soil that acts on the pipeline is not well understood. Better knowledge of the soil load acting on the pipeline can result in a reduction of the wall thickness or a reduction of the required ring stiffness of the pipeline. The third issue addressed is the maximum allowable mud pressure during drilling. This is of importance because this pressure determines the maximum length over which a mud flow can be established and thereby controls the maximum length of the boring.
Tests have been carried out in the geotechnical centrifuge of GeoDelft (the GeoCentrifuge) in order to study the above processes and increase our understanding of the mechanisms that play a role in the stability of a borehole filled with bentonite. A programme of four tests has been performed to investigate the phenomena described below:

Arching

During drilling a borehole is protected from collapsing by filling it with bentonite. However, arching in the surrounding soil contributes to the stability of the borehole. As a result, arching also reduces the total amount of soil load acting on the installed pipe.
The mechanism of arching has been described by Terzaghi (1943). Terzaghi described arching in ideal sand due to the local yield of a horizontal support. Arching can be achieved by gradually lowering a strip-shaped section A-B (see Figure 1) of the horizontal support. Before the strip starts to sink, the soil load acting on the strip equals the earth pressures acting at the depth of the horizontal support. However, a lowering of the strip causes the sand located above the strip to follow. This movement is hampered by frictional resistance along the boundaries between the moving and the stationary mass of sand. As a consequence, the total pressure on the sinking strip decreases by an amount equal to the vertical component of the shearing resistance which acts on the boundaries. As a result, the total pressure on the adjoining stationary parts of the support increases by the same amount.

For an ideal sand stratum, Terzaghi’s derivation has up until now been considered to be appropriate for the situation where arching occurs and is accordingly incorporated in Dutch pipeline standards. However, for cohesive soil layers consolidation may occur over a period of time and this will result in a reduction of the arching effect, thereby increasing the vertical load on the installed pipe. The latter process is described by De Kock (1993) and has been added into the Dutch pipeline standards. However, the formulae derived by Terzaghi and further developed by De Kock and Meijers have never been tested and verified for the case of a horizontal borehole. In addition, the mechanism of arching in the surrounding soil of a horizontal borehole filled with bentonite may differ from the arching mechanism triggered by the lowering of a horizontal strip.

Centrifuge Testing

A geotechnical centrifuge presents the possibility of performing tests on a small soil model, wherein the stresses in the soil model are comparable to the stresses in prototype. With an acceleration that is N times the acceleration of gravity in a geotechnical centrifuge, the stresses in a soil model that is 1/N times the prototype will be the same as the stresses in the prototype. Another advantage of centrifuge testing is that the tests can be performed in soils of consistent characteristics with a uniform stratigraphy and under well-established boundary conditions.
In order to enhance our understanding of the arching mechanisms, it is very important to perform tests with stress conditions that are comparable to the prototype situation. Therefore a total number of four tests have been performed in the GeoCentrifuge.

Some years ago, a number of tools were specially designed for geotechnical centrifuge testing of problems related to micro- and macro tunnelling as part of the CUR/COB L600 project. For this study, the so called ‘reamer module’ could be used. This reamer module simulates the drilling of a pipeline in a soil stratum. For that, the reamer module is installed inside the soil model before the test begins. During the centrifuge test the reamer is pulled through the soil model. While doing so, the reamer creates a borehole in which the pipeline is installed. To prevent the borehole from collapsing, the space between the pipeline and the wall of the borehole is instantly filled with bentonite. The test set-up is given in Figure 2. The mud pressure is controllable during the drilling process. First, the mud pressure is reduced to create arching and possible collapse of the borehole. The mud pressure is increased until failure of the borehole at the end of each test.
The first three tests are executed in a homogeneous, rather dense sand stratum. For these tests, whilst the borehole diameter remained constant, the boreholes were drilled at three different depths. In the fourth test, a thin clay layer is added to the sand stratum at the same level as the borehole.

The differences between the set up of the four tests are summarized in Table 1. In this table, H represents the height of the overburden cover height whilst D is the borehole diameter. The tests were performed at a scale of 1:16. The acceleration level in the centrifuge was therefore 16 times higher than the acceleration of gravity.

In the tests, the reamer was pulled through the sand stratum over a prototype distance of 6.4 meters creating a borehole of the same length. During the drilling process, the reamer moved past six total stress transducers (tst) and five displacement transducers (dt). The total stress transducers were located at different separations from the borehole on both sides, but all at the same height. Three displacement transducers were located in the moving mass of sand above the borehole and the other two were located between the moving mass and the stationary mass of sand. The displacements at surface level were measured as well. Inside the borehole the mud pressures are measured throughout the whole drilling process and the deformations of the installed pipeline. After finishing the tests, visual inspections were performed to establish the shape of the remaining borehole. In Figure 3 the installed reamer module, including six total stress transducers and five displacement transducers are shown before the sand was placed into the container.

Results Geocentrifuge Tests

The boreholes in homogeneous dense sand strata (Tests 1-3) were never found to be unstable, even in the case of the lowest mud pressures (just marginally greater than the water pressures at borehole level). It was also found that the borehole did not collapse where the overburden was minimal (Test 1) or where a thin clay layer was present (Test 4).
In all tests, the measurements of the total stress transducers showed an increase of the total vertical earth pressures in the stationary mass of soil adjacent to the borehole (presumably due to arching). However, when the drilling process was stopped in Test 4, the total vertical earth pressures fell rapidly before reaching an equilibrium. It is presumed that this fall was caused by consolidation of the clay layer. As the borehole remained stable, no soil load acted directly on the pipeline. It was found that only the universal mud pressures acted on the installed pipe. In case of Tests 1-3 these mud pressures could almost be decreased to as low as the water pressures acting at borehole level. However, in Test 4, it was found that the mud pressures in the borehole could not be reduced in the same extent, as was the case in Test 1-3 without a clay layer.

It was found that even small changes in pressure within the borehole caused an immediate change in stress distribution in the adjacent soil mass. For example, it was found that once the reamer was approaching the line of tst 2-5, the stresses measured at tst 2 and 3 (closest to the reamer) started to decrease. The decrease continued until the reamer had passed the line of the total stress transducers by 3.84 m. Once the reamer passed this line and the borehole was created, the vertical stresses immediately increased again. Similar drops of stress occurred when the mud pressures in the borehole were reduced to a lower level. Apparently the total stress transducers closest to the borehole (tst 2, 3 and in some cases tst 4, depending on the H/D ratio) are very sensitive to pressure changes in the borehole, whereas the vertical earth pressures measured at a greater distance from the borehole at tst 4 and 5 remained the same or exhibited a slight increase.

After finishing the drilling process, the mud pressures in the borehole were increased until failure of the borehole. No external blow-out at surface level was found. The failure mechanism in Test 1 appeared a slight lift of a soil wedge between the borehole and the surface. In the sand stratum of Test 2 and 3 hydraulic fractures were found. In Test 4 the bentonite was penetrated in the clay layer.

Finite Element Calculations

The objective of the finite element calculations is to study numerically the mechanisms that were observed in the physical (centrifuge) model, and to gain an understanding of the processes on the basis of a comparison of both the physical and the numerical model observations.
The calculations have been performed using PLAXIS V8.2, a finite element code allowing for the two dimensional analysis of deformation and stability in geotechnical engineering.

The geometries investigated are as previously stated in Table 1. In the framework of the current investigations it was decided to verify the existence of mechanisms that were observed in the physical model tests, rather than to investigate the suitability of specific numerical material models to predict prototype behavior. It was therefore decided to model the prototype behavior with the well known standard Mohr-Coulomb model. As many of the soil parameters in the physical model are known within small boundaries or have been measured directly, it was decided to adopt the following parameters as given fact:

• the mud pressure at the top level of the borehole is equal to the hydrostatic water pressure increased by 2 kN/m²;
• the bulk unit weight (g) of the drilling mud is 10.28 kN/m³;
• the effective angle of internal friction (φ’) of the sand is 40° ;
• the Poisson ratio (n) of the sand and clay equals 0.3 and 0.35 respectively.

A parametric study was carried out for the soil parameters below (the ranges between which parameters were varied are shown between brackets):

• the Young’s modulus (E50) of the sand (17500 to 35000 kN/m^²)
• effective cohesion (c’) of the sand (0.2 to 1.0 kN/m²);
• the dilation angle (y) of the sand (0 to 10°);
• the undrained shear strength (S_u) of the clay (13 kN/m^² to very high, the latter to prevent failure of the clay). Undrained behavior was modeled assuming undrained material behavior in a so-called effective stress analysis with c’ = S_u and φ’ = 0° (reference is made to the hand-outs of the Experienced Plaxis Users Course 2005: method B).

The calculations are first performed with a minimum amount of cohesion in the sand. In order to avoid numerical complications the minimum value was set to 0.2 kN/m^². As borehole stability was not achieved in most of the modeled geometries, the value was raised to 1.0 kN/m^², simulating some shear strength at low stresses due to the mud cake along the inside of the borehole. This resulted in a stable borehole in almost all situations and thus also in situations in which the arching effects could be analyzed.
As the geometry is symmetric, half the geometry has been modeled in PLAXIS. At the locations of the stress transducers and the displacement recorders in the geocentrifuge tests, stress- and displacement points were selected in the PLAXIS model. The modeled geometry of Test 1, including the displacement points is shown in Figure 4. Figure 5 shows the locations of the stress points in the model.

Results Finite Element Calculations

Distinction is made between the runs carried out for the model in sand (Tests 1- 3) and the runs carried out for the model with a clay layer (Test 4). Failure mechanisms and the influence of parameter variations are discussed for the models in sand and clay separately.

Model with sand

The results of the numerical calculations for the model with sand are summarized below:

• Failure mechanism: In case of instability of the borehole, the upper part of the borehole collapsed (see Figure 6).
• Stability: The borehole was found to be stable in all tests when applying a dilatation angle of 5 to 10°, and a cohesion of 1 kN/m² in the sand.
• Dilatation angle: An increase of dilatation angle led in all tests to an increase in borehole stability.
• Cohesion: A slight increase of cohesion (0.2 to 1.0 kN/m²) led to a significant increase of borehole stability. Due to a decrease of plasticity, the calculated displacements decrease as well.
• Stiffness: Variation of stiffness only led to a variation of calculated deformations. Borehole stability and arching effects were not influenced by a variation of stiffness.
• Influence of H/D: No correlation was found between borehole stability and H/D. With a decrease of H/D, the width of the increase of soil stress due to arching effects decreased approximately linearly with H/D in the calculations.

Model with clay

Parameter variations were firstly carried out with a constant undrained shear strength of Su = 100 kN/m² for the soil model with a clay layer (Test 4), see Figure 7 & 8. This was done to prevent failure of the clay in order to be able to study the effect of its presence on the development of the arching in the overlying sand layer.
Other parameter variations were carried out with a constant undrained shear strength of Su = 13 kN/m², which represents the estimated initial in-situ undrained shear strength.

Both calculations are referred to respectively as Calculation 4 (Su =100 kN/m²) and Calculation 4b (Su = 13 kN/m²).

• Failure mechanism / stability: In some cases, local instability was calculated due to differences in stiffness between the sand and clay layers. In general calculations resulted in a stable borehole when applying a dilatation angle of 0 to 10 degrees and a cohesion of 1.0 kN/m² in the sand.
• Dilatation angle of the sand: An increase of dilatation of the sand in all cases led to an increase of borehole stability and a decrease of deformations.
• Cohesion of the sand: A slight increase of cohesion (0.2 to 1.0 kN/m²) led to a significant increase of borehole stability.
• Stiffness of the sand: The borehole stability and the arching effects are slightly influenced by the applied variation of stiffness of the sand. In the calculations with low strength of the clay, the stability increases with lower stiffness of the sand as the decrease of difference in stiffness between the sand layer and the clay layer results in less local instability near the layer boundaries.
• Strength of the clay: Arching effects and modes of deformation are very much influenced by the strength of the clay layer. Two typical deformation modes are shown in figures 7 and 8, figure 7 shows local failure (high strength clay) whereas figure 8 shows squeezing (low strength clay).

Comparison

The soil loads presented in Table 2 should be accounted for with regard to the pipeline design when applying the Dutch pipeline standards (NEN 3650 and 3651). As mentioned before, the Dutch pipeline standards are based on Terzaghi's theory and further developed by De Kock and Meijers. According to the standards, the soil load can only be reduced by taking arching into account if the pipeline is situated in a homogeneous soil stratum at a depth of at least 8 times B₁, where B₁ stands for half of the width of the subsiding soil column above the borehole. Accordingly soil loads aren’t reduced by arching for H/D ratios of less than 5.0.

As was mentioned earlier the borehole remained stable in all GeoCentrifuge tests. Consequently no soil load acted directly on the pipeline. The mud pressures could almost be decreased to the point where they equaled the water pressures at borehole level during drilling for Tests 1-3. However, in Test 4, some soil load did act indirectly onto the pipe. This conclusion is based on the fact that it was not possible to decrease the mud pressures below a magnitude of 2 kPa above the actual water pressures caused by pulling the reamer without injecting bentonite in the borehole. This value should be compared to the difference in mud pressure and actual water pressure for Test 2, which appeared negative for a short period of time. After equilibrium it was 0.8 kPa higher than the actual water pressure. Despite of the reduction of the arching effect in cohesive or layered soils, no collapse of the borehole at Test 4 was found. A logical explanation resides in the fact that the drilling mud was confined and had no way out.

The numerical calculations showed that by adding a realistic angle of dilation to the sand (5–10 º) and a little cohesion of 1 kPa, a stable borehole could be obtained as was found in the geocentrifuge tests. In the calculations, the mud pressure at the top level of the borehole is equal to the hydrostatic water pressure increased with 2 kPa.

From the test and numerical results it can be concluded that the design loads calculated in accordance with NEN 3650 and 3651 are conservative for the cases studied. A similar conclusion can be drawn from the increase of vertical earth stresses in the adjacent soil masses next to the borehole. The increase of vertical earth stresses is due to the fact that the soil column above the pipeline will settle while the soil next to the pipeline won’t (at least for a homogeneous sand stratum). Due to this effect shear forces are developed and vertical load is transferred from the settling soil to the sides. According to Terzaghi’s theory the integral of the total stress increase times the area at pipe level is equal to the integral of the shear stress developed on the sliding planes times the area of the planes.

Figures 9 to 11 show the increase of total vertical earth stresses for Tests 1-3 at various distances from the borehole calculated using the approach of Terzaghi, (and subsequently modified by De Kock and Meijers). The stress increases measured in each of the four tests have been added to these figures as have the results obtained from the numerical calculations. The tst’s are located at the positions shown in Figure 3, being at the mid-height level of the borehole and at a distance of 1.06 D to the left (tst 2) and 0.66 D, 1.46 D and 2.26 D to the right (tsts 3,4 and 5 respectively). Despite the limited amount of total stress transducers installed during the tests a clear trend is visible with regard to arching in a homogeneous sand stratum. Apart from the measurements at tst 2 and 3 for test 3 (H/D = 7.0) all total stress transducers and numerical results show a higher increase of total vertical stresses due to arching than proposed by Terzaghi and Meijers. This indicates that more load is transferred to the sides than calculated by their approaches and less load is therefore to be expected on the pipeline.
The relatively low increase of total vertical stresses at tst 2 and tst 3 for test 3 can be explained by a redistribution of stresses due to very low pressures in the borehole. During test 3, borehole instability was triggered by pulling the reamer over a relatively long distance without injecting bentonite. This did not result in borehole instability but instead a decrease of build up stresses at tst 2 and 3 was visible and at the same time a stress increase at tst 4 and 5 occurred. Apparently, the existence of even very low mud pressures in the borehole cause the arching zone at borehole level to expand.

With regard to test 4 (H/D = 4.5, with clay layer) it can be noted that the increase of total vertical earth stresses, after equilibrium, is evidently less than was the case for test 2, without a clay layer. It is interesting to note that the numerical results confirm this phenomenon. The differences in total stress increase between both tests reduce with distance away from the borehole. Also, in line with De Kock’s theory, the arching effect reduces due to consolidation of cohesive layers and correspondingly the load acting on the installed pipeline increases.

The relative increase in effective stress and the stress distribution in the arching zone observed in the numerical model differ from the physical model. No clear reasons were found to explain the differences. However, the drilling process in the centrifuge tests may be of influence. As was said earlier, the total amount of load transferred to the sides and the stress distribution in the arching zone is highly sensitive to pressure changes in the borehole. Also vibrations caused by the drilling process could have influenced arching in the physical models. 

Discussion

With some reservations it can be concluded from Figures 9-11 that the total amount of increase in total vertical earth stresses measured at one side of the borehole roughly equals half of the neutral weight of the subsiding soil column above the borehole. Considering no soil load acting directly on the installed pipeline (at least for test 1-3) it must be concluded that almost all neutral weight was transferred to the sides. In other words, either higher shear stresses (positive friction) seem to develop along the sliding planes above the borehole or a different mechanism to that proposed by Terzaghi exists. No visible shear planes were found in the centrifuge tests except from one accidental borehole instability due to a borehole filled with water instead of bentonite. The shear planes developed because of this accident at the beginning of Test 2 are identical to the shear planes predicted by both the numerical model (see Figure 6) and Terzaghi’s formulae. Therefore it is to be expected that higher shear stresses are being developed.

  

As hardly any cohesion (c) is present in sand and the internal friction (f) of sand is well known higher horizontal effective stresses (s’_h) are to be expected in the soil column above the borehole than assumed in the calculations. From equation (2), it is apparent that the coefficient of horizontal earth pressure K is an important parameter in the arching process.  A value for K of 1 to 1.5 is suggested by Terzaghi but Meijers and De Kock consider a value of K = 1 – sin f (i.e. 0.36 for f = 40º) to be more likely. From the results of this study, indeed higher K values are to be expected. The numerical calculations confirm this assumption.

Not all of the questions have been answered yet.  However, from the parametric numerical study it was concluded that an increase of the dilation angle leads in all cases to an increase in arching and thus in an increase in borehole stability. The tests were performed in a homogeneous rather dense sand stratum (relative density of the sand was approximately 65%).  The extents to which the relative density of the sand and the dilation angle influence the arching effect are still to be investigated.

The effects of vibrations (e.g. vibrations caused by sheet piling, pipe vibrations induced by pumps, earthquakes) were not part of this study. Neither was the interaction between the drilling mud and the soil grains and the effects of elastic deformations and long term relaxation (creep) studied. All these phenomena can result in a stress redistribution but they did not form a part of this study.  

Conclusion

The GeoDelft centrifuge tests on borehole stability during horizontal directional drilling showed that a borehole drilled in a homogeneous dense sand stratum can be very stable due to arching.  It was also found that only the universal mud pressure acts as a load on the installed pipeline.  From the results of Test 4 it can be concluded that due to consolidation of cohesive soil layers the arching effect reduces and likewise the load acting on the installed pipeline increases. The centrifuge test results, supported by PLAXIS calculations, demonstrate that borehole stability can be achieved with only very small mud pressures in the borehole due to the development of arching. No or only minor load develops on the pipeline, other than induced by the mud pressure, which suggests that optimization of the design is still possible (compared to designs following the Dutch Pipeline Design codes). The results show that the borehole remains stable for all H/D ratios investigated. No correlation has been found between borehole stability and H/D ratio. 

References

1. De Kock, R.A.J. (1993), Personal communication.
2. Meijers, P. (1993), Review of a calculation method for earth pressure on pipelines installed by directional drilling, Report CO-341850/4, march 1993 by Grondmechanica Delft.
3. Meijers, P. and De Kock, R.A.J. (1993), A calculation method for earth pressures on directionally drilled pipelines, conference, Belgium.
4. NEN (1994), Nederlands Normalisatie Instituut, Additional requirements for steel pipelines in crossings of important public works, code NEN 3651, Delft.
5. Terzaghi, K. (1943), Theoretical soil mechanics, John Wiley and sons, New York, 66-76.