GEORGE LESLIE LIMITED AGAINST HALCROW GROUP LIMITED
OUTER HOUSE, COURT OF SESSION
[2017] CSOH 106
CA80/16
OPINION OF LORD TYRE
In the cause
GEORGE LESLIE LIMITED
Pursuer
against
HALCROW GROUP LIMITED
Defender
Pursuer: A McKenzie; Pinsent Masons LLP
Defender: Walker; Cameron McKenna LLP, Nabarro and Olswang LLP
2 August 2017
Introduction: Factual Narrative
[1] In 2010 the pursuer, a civil engineering contractor, was engaged by Scottish Water to replace approximately 1.26 kilometres of concrete water main near Westfield, West Lothian, with a new 1,000 mm diameter polyethylene pipe. The pipe was to be gravity fed, and the water was to flow in a generally easterly direction from Gowanbank to Crawhill. On 9 and 10 August 2011, the pursuer entered into a contract with the defender in terms of which the defender agreed to carry out the detailed design for the construction of the works under the main contract.
[2] At a location between chainages 440 and 523, the pipeline required to run up a fairly steep slope with an overall height gain of about 24 metres. Due to the underlying geology, the slope was in two sections. The lower section had an incline of about 15 degrees (or 1 in 3.8) and the upper section had an incline of about 31 degrees (or 1 in 1.7). The ground conditions at the slope were investigated by means of two boreholes at the top and bottom respectively. The defender produced a design for this section of the pipeline specifying various vertical and horizontal bends at angles of 22.5 or 45 degrees. The design also incorporated concrete thrust blocks at pipe bends, the purpose of which was to provide restraint against out of-balance forces arising due to internal pressure and the velocity of the water in the pipe.
[3] Work on site commenced during autumn 2010. The area around the proposed pipeline track was de-vegetated. The pursuer began to excavate the pipeline trench. The defender continued to issue design drawings to the pursuer. On 24 January 2011, the pursuer’s site staff notified the defender’s design manager, Andy Barnett, of the presence of waterlogged ground between about chainages 440 and 480, within the lower section of the slope with the 15 degree incline. On 25 January 2011, Mr Barnett emailed the pursuer’s David Gibson with a proposed detail (“Detail 2”) for crossing the soft area. The detail comprised a basal geogrid layer, overlain by 0.75m of 300-100mm size crushed rock and 0.25m of imported granular Type 1 bedding material. The pipe itself was to be placed on this bedding material and the trench would be backfilled around the pipe with locally as‑dug material. The purpose of the detail was to provide a more solid base to the trench for construction of the pipeline through the soft area. This proposal was accepted and work proceeded accordingly to completion.
[4] On 12 May 2011, the new pipeline was subjected to pressure testing at a maximum pressure of 8.4 bar. The pipeline was commissioned on 19 June and was in operation on 20 June. During 21 June, the area around the pipeline experienced sustained heavy rainfall. Met Office records for Edinburgh (Gogarbank) indicate that 47mm of rain fell on 21 June and a further 5.6mm fell on 22 June.
[5] On 23 June 2011, between 7.30 and 7.45 am, the pipeline suffered a catastrophic failure. As a consequence of the continuing flow of water out of the pipe under gravity pressure, much of the soil and other material above, around and beneath the pipe was washed away, creating a large ravine at the top of the slope. Sections of pipe were left exposed and unsupported. Once the flow of water had been stemmed, the pipeline was found to have failed at two welded joints in the area of the upper section of the slope.
The Issue
[6] The pursuer contends that the pipeline failed as a consequence of breach by the defender of its obligations in terms of the design contract. After proof, the pursuer maintained only one aspect of the case averred on record against the defender, namely that the design of the geogrid solution in Detail 2 was inadequate. Placing and laying the geogrid directly on to soft ground on a slope with a granular material on top had effectively produced a drainage layer which acted as a “slip plane” between the two surfaces. The consequence of the heavy rain was that the geogrid (and the material resting upon it on which the pipeline was bedded) had slipped down the hill, pulling the pipeline with it and causing the pipeline to fracture at the welds. Although the failed welds were weaker than they ought to have been, there was no convincing evidence that they would have failed under normal operating conditions.
[7] The defender contends that the pipeline failed as a consequence of poor welding practices which had produced sub-strength welds. The mechanism of slope failure relied upon by the pursuer was, it was argued, demonstrably wrong. The slope had not failed and, with the exception of an error in the calculation of the thrust blocks (which did not cause or contribute to the pipeline failure), the defender had not been negligent in any of the respects averred by the pursuer. There was, however, evidence of variable quality welds in the pipeline because of the adoption of poor working practices by operatives on site. It was probable that the lower pipe fracture had occurred first, originating as a crack and then spreading to create a brittle weld fracture. After the first weld failure, the escaping pressurised water had washed away the soil, leaving the upper part of the pipe exposed and unsupported.
Eye Witness Evidence
[8] In a witness statement agreed by the parties to constitute his evidence, Mr Robert Waddell, a semi-retired farmer who lived locally and who had also worked in the construction industry, described what he had observed at about 6.30 am on 23 June 2011 as he drove along a road with a view of the upper part of the slope. A V-shape seemed to have developed towards the top of the steep slope at the location where the pipeline failure subsequently occurred. There appeared to be a damp patch in the ground directly above the V-shape. Mr Waddell could also see about three or four metres of pipeline exposed near the brow of the slope, with a bend in the pipe visible.
Expert Evidence
[9] Expert evidence was given on behalf of the pursuer by (i) Mr Adrian Dolecki, a chartered engineering geologist who is technical director and head of sector for ground engineering with Parsons Brinckerhoff Ltd, Cardiff; and (ii) Mr Philip Clisham, a chartered civil engineer who is technical director of pipelines with GHD, an engineering consultancy, and also deputy chairman of the Pipeline Industries Guild. Expert evidence was given on behalf of the defender by (i) Dr Jeremy Love, a fellow of the Institution of Civil Engineers who is senior partner of Geotechnical Consulting Group LLP, London; and (ii) Dr Jeremy Bowman, a director of Plasticpipes Ltd, a company providing technical consultancy and support to gas and water utility companies in the area of plastics pipes. Mr Dolecki and Dr Love provided expert opinion on whether the slope had failed due to faulty design by the defender and, if so, whether the work of the defender had fallen below the requisite professional standard. Mr Clisham and Dr Bowman provided expert opinion on whether it was likely that the pipe had failed due to weak welds. All four witnesses had considerable expertise and many years of experience in their respective fields and were amply qualified to give evidence on the matters covered by their written reports and oral evidence.
[10] Each of the expert witnesses had produced a report and a supplementary report commenting on the report of his counterpart. By agreement between the parties and with the leave of the court, each pair of expert witnesses (ie Mr Dolecki and Dr Love, and Mr Clisham and Dr Bowman) gave their oral evidence concurrently. Each witness was invited to give a brief initial presentation of his evidence, focusing on matters where there appeared to be disagreement. Counsel were then invited in turn to put questions, it being understood that either expert witness could, if he wished, respond to any question put or to any answer given by his counterpart. In a case dependent upon technical evidence such as this one, I found the procedure extremely helpful as a means of identifying the true areas of dispute, and of assisting my understanding of the reasons for the witnesses’ respective points of view as well as the arguments that could be advanced against those points of view.
[11] The pursuer’s claim that the failure of the pipeline was caused by negligent design by the defender is predicated on establishing that the failure occurred due to the pipeline having slipped down the slope, rather than for any other reason. It is logical, therefore, to begin my assessment of the evidence by examining the issue of causation, because success for the pursuer depends upon my accepting the evidence of Mr Dolecki as to the mechanism of causation of the pipeline failure.
Soil and Groundwater Conditions
[12] Before narrating Mr Dolecki’s evidence, it is necessary to describe the soil and groundwater conditions of the slope prior to commencement of any construction work. The following description, which I do not understand to be controversial, is derived from Dr Love’s first report.
[13] The local geology comprises glaciofluvial deposits overlying Devensian till (boulder clay). Groundwater is likely to sit above the relatively impermeable boulder clay at around 118-119m OD (approximately 8 metres below the crest of the slope), forming a natural spring line on the surface of the hillside at around this level. In comparison to the boulder clay, the glaciofluvial deposits (which consist primarily of sands and gravels) may be considered to be effectively free-draining. The change in soil type together with the emergence of the spring line explains the shallower slope angle which exists over the lower half of the slope. There is also likely to be a thin covering of weaker, superficial head material and soft alluvial soils overlying the boulder clay, over the lower part of the slope.
Mr Dolecki’s Opinion
[14] Mr Dolecki’s evidence, in his written reports and oral evidence, may be summarised as follows. De-vegetation of the slope prior to commencement of construction would have increased the moisture content and pore water pressures, which was likely to have had a detrimental effect on its stability. The crushed rock placed on the geogrid had a particle size, specified by the defender, of 300 to 100mm, whereas the geogrid manufacturer’s guidelines required a maximum particle size of 75mm. This type of pipe support system depended on the tensile load gained by the geogrid over its entire length, and on the interlocking of the stone/granular material with the apertures in the grid. The performance of the grid depended upon using well compacted, free draining, well graded granular soils both above and below the geogrid: ie the geogrid should be positioned in the middle of the layer, not at the base. The defender’s design solution of placing and laying the geogrid directly on to boulder clay with low shear strength on a slope with a granular material on top, effectively produced a drainage layer which would act as a slip plane between these two surfaces. The size of the specified crushed rock was too big to lock into the 33mm apertures within the geogrid. Additionally, no anchor detail, which should have been included, was shown for the geogrid. It would also have been sensible to extend the geogrid reinforcement into non-waterlogged ground upslope and downslope of the soft ground.
[15] The design calculations for the thrust blocks included some errors, particularly for the design of a block located at a 22.5 degree bend near the crest of the slope. Because the indicated concrete density of the block was a thousand times too high (24,000kN/m3 instead of 24kN/m3), the block size was insufficiently large and heavy to perform its intended function. The pipe was liable to movement creating additional forces acting on it, especially at changes of direction.
[16] When the pipe was commissioned, the thrust blocks were inadequate to provide sufficient restraining pressure. The heavy rainfall on 21 June 2011 saturated the slope and lowered its factor of safety by reducing the soil shear strength and the friction properties of the slope materials. The saturated ground conditions were exacerbated by the presence of the free-draining pipe bedding materials on top of the geogrid at the base of the trench, which provided the slip plane causing temporary drawdown of the water table at or near the foot of the slope. This led to additional pipe stress and, with the trench acting as a drain, resulted in the ultimate and catastrophic failure of the pipe. The landslip occurred before the pipeline failed, as evidenced by Mr Waddell’s observation of a V-shaped slip with the main pipe failure event shortly thereafter. An adequately designed slope section should have been able to cope with high groundwater and to withstand expected changes to the groundwater regime as well as any seasonal rainfall events.
[17] In his supplementary report, Mr Dolecki provided the results of a finite element analysis (FEA) carried out using a Plaxis 3D computer model. For the modelling exercise, Mr Dolecki used ground condition information obtained in the course of remedial work after the pipeline failure as well as information available at the time of the original works. Geometric features of the slope were based upon information in Dr Love’s first report. The angle of soil shearing resistance (or angle of friction), to which the symbol φ is attributed, was assumed to be 29° for loose glaciofluvial deposits and 31° for medium dense glaciofluvial deposits. Because of the severe groundwater flow after the heavy rainfall, running down the slope along the unprotected faces of the trench, Mr Dolecki considered that an interface coefficient (or strength reduction factor) of no more than 0.3 was justified. In order to simulate the events of 21-23 June 2011 in the model, (a) the density weight of ground materials was increased to a 90% degree of saturation to take account of the heavy rainfall, and then (b) the reduced strength interfaces were activated to take account of the severe groundwater conditions. The combined result of (a) and (b), which was presented in graphic form in an appendix to Mr Dolecki’s supplementary report, was the formation of a triangular wedge on the upper slope and the sliding of trench material on the lower slope. This was consistent with observations made after the actual slope failure, and thus supported the hypothesis that the pipeline fracture was preceded by sliding of the trench materials on the lower section of the hillside and by rotational slope stability failure of the upper slope, all of which induced considerable additional deformation and forces on the pipeline and caused the welds to fracture.
[18] In his oral evidence in response to Dr Love’s criticisms (see below), Mr Dolecki emphasised that actual soil conditions within the slope would have been extremely variable and would not necessarily have conformed to soil science theory. In particular, conditions at the interface between the geogrid and the underlying material would change as the boulder clay became wetter and softer. A small area of clay with no or reduced shear strength would be sufficient to cause a slip of the geogrid and the material upon it to occur. The changing parameters in this slope were too complex for the application of a simple formula. It was therefore necessary to be extremely conservative when selecting the appropriate angle of friction and interface coefficient.
Dr Love’s Opinion
[19] Dr Love’s opinion, also provided in two written reports and oral evidence, was that the pipe failed because of sub-standard welds. As regards the quality of the welding, Dr Love relied upon the reports by Dr Bowman and by Exova, to which I refer later in this opinion. Saturation of the overlying backfill material after the heavy rainfall was likely to have reduced its restraining effect on the pipeline, placing more reliance on the pipe itself to resist the out-of-balance forces with which, due to the poor welds, it could not cope. It was not clear whether the first joint failed catastrophically in one go, or whether it failed only partially to begin with. If it was the latter, then water under pressure would have started to leak uncontrollably from the pipeline, progressively washing out trench fill material from around the pipeline immediately down slope of the partial rupture. The washing out of trench fill material would have placed progressively more stress on the damaged pipeline, which was likely to have led to the pipe rupturing entirely. As more and more material was washed downhill under the force of the water flow, this would have led to the second rupture and the formation of the deep ravine within the hillside. The fact that the thrust block was under-designed was not the cause of the failure, since the pipeline should not have required any thrust blocks. If the welds had not been poor, the pipeline would not have failed, despite the thrust block being under-designed. Similarly if the thrust block had not been under-designed, the pipeline would still have failed due to the poor welds.
[20] Dr Love disagreed with Mr Dolecki’s conclusion that slope failure had occurred. He sought to demonstrate, by means of slope stability calculations, that the pipeline was stable despite the presence of the geogrid at the base of the trench. The principal reason for this was the effect of friction acting on the sides of the trench, which did not appear to have been considered in Mr Dolecki’s calculation. In addition, the angle of friction used by Mr Dolecki for the interface between the geogrid and the underlying soil was too low. Adequate friction would develop without interlock between the backfill material and the geogrid apertures. Even making the pessimistic assumption of an infinite slope with a water table at the ground surface, the slope with the geogrid was stable. In any event, forward motion of the lower section of the pipeline would not have caused it to fail where it did, at the top.
[21] Dr Love’s calculation was based upon the text of CIRIA Special Publication 123, entitled “Soil reinforcement with geotextiles”. Section 4.5 of that publication deals with the coefficient of direct sliding of a block of soil across a layer of reinforcement. As noted by Dr Love, SP 123 states that the resistance to direct sliding of a block of soil across a geogrid depends on (a) shear between the soil and the planar surfaces of the grid, and (b) the soil-to-soil shear through the grid apertures. Taking both of these into account, the expression for direct sliding resistance used for design is normally written as
αds = ās (tan δ/tan φ) + (1 – ās)
where:
ads is the coefficient of direct sliding operating at the geogrid interface;
ās is the area ratio (the fraction of the grid surface area that is solid);
tan δ is the skin friction for soil shearing over the planar surfaces of the grid material; and
tan φ is the skin friction for soil shearing over soil (which Dr Love assumed here to be around 30°).
SP123 recommended taking typical values of 0.5 for ās, and 0.6 for (tan δ/tan φ), which would give a typical value of 0.8 for ads, as opposed to Mr Dolecki’s figure of 0.3. This would suggest an angle of interface friction of around 25°, not 15°.
[22] The validity of the FE analysis carried out using the Plaxis computer model was dependent upon the assumptions fed into it. Dr Love was critical of the values used by Mr Dolecki for the coefficient of direct sliding. The friction at the bottom of the trench was not 0 as appeared to have been assumed for the purposes of the model; it should be at least 0.8. As for the sides of the trench, the interface friction should be the full soil friction and not 0.3 of it. The side walls of the trench were not smoothed in any way; they were left naturally rough, and the material used to backfill the trench (which was the same granular material as existed in the hillside) was placed and compacted in intimate contact with the surrounding material. The program ought to have determined the normal effective stresses acting on both of these surfaces (the base and sides of the trench) and then used the correct interface friction angles to calculate the appropriate shear stresses acting to restrain the pipeline trench fill material from moving. If done correctly, the pipeline trench would have been found to be stable. A sensitivity analysis subsequently carried out by Mr Dolecki had shown that once the assumed interface friction acting on the base and sides of the trench exceeded about 60% of the available soil friction (as, in Dr Love’s opinion, it ought to have done), all four of the key variables, namely (a) the axial force in the pipeline, (b) the shear force in the pipeline, (c) the maximum bending moment in the pipeline, and (d) the maximum pipeline displacement, reduced to nominal values, demonstrating that the slope was stable. The reason why the Plaxis analysis presented in Mr Dolecki's supplementary report predicted instability was that unrealistically low interface friction values had been assumed to act on the base and sides of the pipeline trench. The angle of friction of the glaciofluvial deposits had also been underestimated.
Discussion: Slope Failure
[23] These are highly technical matters, but on the basis of the available material, including the discussions between Mr Dolecki and Dr Love during the concurrent presentation of their evidence, I feel able to reach a conclusion. There was no material disagreement between the two experts as to the soil science theory used in their calculations; Mr Dolecki confirmed expressly that he had no difficulty in principle with the methodology utilised by Dr Love or with his mathematical calculations. That is perhaps not surprising because, as I have noted, the formulae used in the calculations are taken directly from CIRIA Special Publication 123. The dispute between Mr Dolecki and Dr Love was a sharp and clearly identifiable one: they disagreed as to (i) the appropriate angle of friction (φ), and (ii) the appropriate coefficient of direct sliding (αds). Of these two parameters, the more substantial disagreement was the latter, where Mr Dolecki advocated a maximum coefficient of 0.3 – 0.4 and Dr Love a minimum of 0.8.
[24] As regards coefficient of direct sliding, the CIRIA publication states (section 4.5) inter alia as follows (with footnote references omitted):
“…The coefficient of direct sliding is in the range 1.00 ≥ αds ≥ 0.60 for a wide variety of woven and non-woven geotextiles and soils. The lower values apply to geotextiles with smooth, even surfaces. The minimum possible direct sliding resistance would be of the order αds = 0.4 which applies for soil shearing over smooth metal. Woven geotextiles with significant surface roughness mobilise greater direct sliding resistance in the range αds ≈ 0.8 to 1.0.”
The reinforcement used in the present case was not a geotextile but a geogrid with 33mm rectangular apertures. I note, however, that in its discussion of geogrids (at the end of section 4.4), the CIRIA publication confirms that “when the fraction of the reinforcement plan area is set to unity, ās = 1, the relations reduce to those relevant to woven and non-woven geotextiles”. On the face of it, therefore, the CIRIA publication supports Dr Love’s methodology.
[25] Mr Dolecki’s view was that it was prudent to use a coefficient of direct sliding much lower than the figure in the CIRIA text, in view of the adverse ground conditions created by the period of heavy rainfall. As he put it at paragraph 6.5.3 of Appendix A to his Supplementary Report:
“In the time lapse between this rainfall event and the slope failure on the 23rd June 2011, the severe groundwater flow running down the hillside along the unprotected faces of the trench justifies in my opinion the implementation of a strength reduction factor of no more than Rint = 0.3. It is my understanding that there is a cause and effect relationship between the rainfall event and the slope failure which hindered the initial intimate contact between the trench materials and the surrounding ground.”
His position in oral evidence was that theoretical calculations were of limited value when the actual condition of the slope following the heavy rainfall was unknown. Even if the angle of interface friction between the geogrid (and the material over it) and the underlying clay did not reduce to 0 (which was the figure he had used in the FEA analysis for the angle of interface friction between the oversized backfill and the geogrid), it could reduce to 10 or lower as the clay became softened by waterlogging; one just did not know. The situation was complex; there would be variations of conditions within the slope but it was only necessary for a small area to move for pipeline failure to result.
[26] As Lord Hope of Craighead confirmed in Dingley v Chief Constable, Strathclyde Police 2000 SC (HL) 77 at page 89, the function of the judge in a civil case is to decide whether the case has been made out on a balance of probabilities, and not to apply the standard of proof that a scientist might adopt in forming a view as to whether a particular thesis has been proved or disproved. The onus of proof rests upon the pursuer. Having carefully considered the evidence of Mr Dolecki and Dr Love, I am not satisfied, on balance of probabilities, that it is appropriate to adopt a coefficient of direct sliding of the order of 0.3‑0.4. Mr Dolecki’s adoption of such a low figure appeared to be based upon uncertainty as to the effect of the rainfall on the clay underlying the geogrid. There appear to me to be two difficulties with this. In the first place there is the formal but important objection that uncertainty does not constitute positive proof on the balance of probabilities. In the second place, however, Mr Dolecki’s approach appears to me to betray a confusion – as noted in the course of the hearing by Dr Love – between the two critical parameters. Even if the effect of the heavy rainfall was to radically reduce the angle of friction (φ) of the trench fill material, this does not justify downward adjustment of the direct sliding coefficient (αds) which, as I understand it, refers to the coefficient of direct sliding between the geogrid and the soil. I accept Dr Love’s opinion (in oral evidence) that the fact that the backfill material was larger than the apertures in the geogrid is irrelevant if the slip is said to have occurred beneath the geogrid, and does not afford a justification for a reduced coefficient of 0.3 or 0.4. I note that in the passage quoted above from the CIRIA publication, a coefficient of the order of 0.4 is said to be applicable to soil shearing over smooth metal, which was clearly not the factual situation here. I also accept Dr Love’s opinion (supplementary report, para 2.1.38) that no reduction factor is required in respect of soil friction acting on the sides of the trench, because the interface is soil to soil. It is, as I understand it, common ground following Mr Dolecki’s sensitivity analysis that if the coefficient of 0.3 used by Mr Dolecki for αds in respect of both the trench floor and trench sides were replaced in the Plaxis model by a coefficient of 0.8 or thereby, the slope is demonstrated to be stable.
[27] There are certain further features of Mr Dolecki’s analysis which I have difficulty accepting, and where it seems to me that Dr Love’s criticisms are valid. Mr Dolecki’s suggestion that the angle of interface friction between the geogrid and the clay could reduce to as low as zero was based upon already-waterlogged clay becoming further softened by water flowing down the trench. Dr Love’s calculations assumed that the clay was waterlogged and that the water table was at the ground surface. He did not regard it as credible that the strength of the clay would be further reduced by rainfall; his observation that if this theory were correct then no clay hills would exist seemed to me to have force.
[28] There is also, in my view, a difficulty for Mr Dolecki’s analysis in the amount of time that elapsed between the period of heavy rainfall on 21 June and the failure of the pipeline during the morning of 23 June: a lapse of more than 30 hours. In the course of the hearing Dr Love estimated (without challenge) the speed of the flow of water down the trench containing the pipeline as about 0.001 metres per second, which would equate to around 3.6 metres per hour, or around 108 metres in 30 hours. Dr Love’s described this flow as being “not a torrent”, and not a high enough velocity to cause erosion. For present purposes, however, the significance of the figure is that drainage of the heavy rainfall from the upper part of the slope would have occurred some considerable time before the pipeline failed. Mr Dolecki asserted that there would be a time lag between the rainfall and the pipeline failure but did not, in my view, provide any convincing explanation for such a time lag in the circumstances of the present case.
[29] Finally, it seems to me that Mr Dolecki’s analysis, which is based upon a slip of the geogrid and not of the fill material in the upper part of the slope, does not provide a satisfactory explanation for the fact that both of the pipeline fractures occurred in the upper part of the slope, above the area where the geogrid had been laid. Indeed, the analysis may be inconsistent with the Plaxis model itself, which seems (at paragraph 7.11 of Appendix A to Mr Dolecki’s supplementary report) to show, in figure 3, that following the rainfall event the trench containing the pipeline had become more and not less stable than the rest of the slope.
Conclusion on Causation
[30] For these reasons I hold that the pursuer has failed, on balance of probabilities, to establish that the pipeline fracture was caused by failure of the slope as a consequence of a slip of the geogrid and the fill material over it around the pipeline. It does not necessarily follow from this conclusion that the defender’s design solution as depicted in Detail 2 was not negligent, but the question becomes academic. Mr Dolecki had various criticisms of the manner in which the defender carried out its work in implement of its contractual obligations, including failure to carry out adequate ground investigations, failure to carry out a slope stability analysis (as was done for the remedial works following the pipeline failure), but by the conclusion of his evidence it appeared that these were aspects of an allegedly negligent design in which the geogrid was used without adequate measures, including use of the correct size of stone, anchoring, and extension beyond the waterlogged area, to prevent it from creating a slip plane. Dr Love’s opinion was that the defender was not negligent in any of the respects suggested by Mr Dolecki, with the exception of the under-design of the thrust block to which I have referred but which was not relied upon by the pursuer as causing or contributing to the pipeline failure. In the circumstances of the present case, I consider that my rejection of Mr Dolecki’s analysis that the pipe fracture was caused by a slope failure is sufficient to entitle me to hold also that the pursuer has failed to prove that the defender’s design solution, as depicted in Detail 2, was negligent.
Why Did the Pipeline Fracture?
[31] It is also strictly unnecessary for me to make any positive finding as to why the catastrophic pipeline failure occurred. Even if no plausible reason for the failure could be identified, that would not create any inference that Mr Dolecki’s analysis should, after all, be accepted as correct. Sherlock Holmes’ well-known observation in Conan Doyle’s The Sign of Four that once the impossible has been eliminated, “whatever remains, however improbable, must be the truth” has not been adopted by the courts: see The Popi M [1985] 1 WLR 948, Lord Brandon of Oakbrook at 955, applied in McGlinchey v General Motors UK Ltd [2012] CSIH 91 at paragraphs 33-34. Although it is open to a defender to advance an alternative cause, there is no obligation to do so, or to prove it.
[32] In deference, however, to the time and expense devoted by both parties to investigation of the defender’s alternative cause, namely faulty welds, it is right that I should express my view on this chapter of evidence.
The Butt-Welding Process
[33] In order to facilitate an understanding of the opinions of Dr Bowman and Mr Clisham with regard to weld failure, it is necessary to describe the butt-welding process that was or at least ought to have been followed by the defender’s employees when welding sections of the polyethylene pipeline together. The process, which was carried out on site using one of only two machines in the UK capable of welding a polyethylene pipe of 1000mm diameter, is as follows. The machine brings the two sections of pipe to be welded close together. They are separated by a heater plate against which the pipe ends are pressed. This is known as “bead‑up” because the axial pressure causes beads of molten material to be pushed up outside and inside the pipe circumference. When the beads reach 3 mm in depth, the pressure (at least according to UK practice) is removed and for about ten minutes the pipe ends are allowed to soak up heat from the heater plate, creating a reservoir of heat in the pipe walls. This is known as “heat soak”. Next, the heater plate is removed and the two pipe ends are pushed together to enable them to fuse. The period between removal of the heating plate and the pushing together of the two pipes (known as “dwell time”) must be kept to less than ten seconds because the pipe ends cool rapidly when not in contact with the heating plate. After a short period, the axial pressure on the pipe is reduced to a low level and the hot pipe ends continue to fuse.
[34] In his report, Dr Bowman noted a requirement of Water Industry Specification (WIS) 4-32-08 (2002) and of the Water Authorities Association’s “Pipe materials selection manual” (1988) that two dummy welds be made for 1000mm pipes. Dummy welds are incomplete welds, in that only the bead-up and heat soak phases are undertaken. Their purpose is to clean the active part of the heater plate prior to it making welds that become incorporated into the pipeline. After the heat soak, the weld process is stopped, the pipe ends are cooled, and the dummy welds are discarded. Approximately 90 minutes are required to make two dummy welds, including 30 minutes for heater plate cleaning, re-heating, and temperature measurement.
The Exova Report
[35] After the pipeline failure, the defender instructed Exova Group plc, a laboratory-based materials testing group, to carry out an investigation. Exova was provided inter alia with sections of pipe containing the two failed butt fusion joints; several intact joints; photographs of the weld failures taken on site; and weld records from site. Exova produced an initial report in July 2011 and a final report in October 2011. Among Exova’s conclusions were the following:
- The weld records indicated that a majority of welds were apparently made with much higher heat soak pressure, equivalent to bead up and heat soak pressure, than specified in WIS 4-32-08. This would have had the effect of squeezing out molten material that was required to make the weld and was therefore a possible cause of poor performance.
- The weld records also suggested that in many cases pressure was not reduced to zero during heater plate removal (dwell time). This was not realistic, since there had to be no interface pressure when the heater plate was being removed.
- Laboratory mechanical testing was conducted on five intact joints that had been welded on site, and results indicated that three of these welds failed to meet the requirements of the specification. One of these samples gave particularly poor results, with failure stress less than 50% of yield stress for the material in one specimen.
Exova considered that there was a continuing problem that required to be addressed. In the event, all welds in the pipeline were replaced some eight months after the occurrence of the pipeline failure with which this action is concerned.
Dr Bowman’s Opinion
[36] On the basis of the documentation provided to him, which by the time of his supplementary report included the final Exova report, Dr Bowman formed the conclusion that the welds in the pipeline had been of variable quality: some had been of sufficiently low strength to cause them to fail prematurely. This conclusion was, broadly, based upon:
- Examination of one of the fracture surfaces of each of the two welds that failed. This showed that the failures – and in particular the lower of the two pipeline failures – had been brittle and not ductile failures, indicating that adequate fusion had not been achieved during the welding process.
- The fact that the intact welds tested by Exova showed such variable quality, including one weld which was extremely poor and required very low energy to rupture it.
[37] Dr Bowman attributed the defects in some welds, including the two welds that failed, to a combination of the following causes:
(i) The welds were of defective quality because of incorrect heating of the pipe ends by the site operatives. The two elements of the weld heating process, bead-up and heat soak, were wrongly combined to reduce heating time to about 78% of that required. Then, in addition and incorrectly, the higher bead-up pressure was applied for all of the (reduced) heating phase, rather than only the bead-up phase. These errors put too little heat into the pipe ends, and because of the higher pressure applied for all of the heating phase, a significant element of the hot molten material that was created was forced into the beads, which were enlarged. This reduced the heat reservoir in the centre of the wall at the ends of the pipe that were to be joined; this lack of heat in the centre of the pipe wall inhibited the creation of strong ductile welds, and led to brittle welds.
(ii) The weld records contained no information to confirm that the correct procedure for making dummy welds had been followed at the site. The time of the first weld recorded in the data logger records did, however, provide some guidance: on at least some days the times were sufficiently early to suggest that dummy welds had not been made. One of the failed welds (the upper failure, which Exova labelled “weld 6569-1”) was the first of the day; the other (the lower failure, which Exova labelled “weld 6589-2”) was either the second or the third of the day. First and second welds of the day, if made without dummy welds, would probably contain contamination that would further reduce their strength.
[38] Evidence in support of these conclusions was said to be supplied by analysis of the fracture surfaces. Both samples demonstrated a lack of bonding in the centre of the pipe wall where, as a consequence of inappropriate application of pressure during the heat soak phase, the material had been insufficiently hot to fuse. The surface of weld 6589-2 indicated a two stage failure process. First, a process of slow crack growth caused the weld to exhibit a leak during a sufficient time for “acne” type damage to be caused to the surface of the weld by a scouring process at that local point. In the second stage of the failure process the weld failed totally, and failed quickly, by the fast propagation of a crack around the remaining section of the weld. The pipeline then ceased to function. The last segment of the pipe to fail operated as a “hinge” with signs of ductile rather than brittle failure. The fracture surface of weld 6569-1 also showed evidence of the joint leaking, and this was supported by the low quality of the weld (poorly bonded central portions) which would allow a leak path to form easily. This section of the pipeline had, however, been buried and deformed during the landslip at the top of the slope which followed the pipe failure, and its poor condition clouded any conclusion.
[39] Dr Bowman did not consider that doubt was cast upon his opinion by the fact that other welds continued to function for several months after the initial failures. Defects would vary in size, with differing consequences for strength and durability. The rate of slow crack growth would vary from weld to weld depending upon the density of molecules crossing the weld interface. Where parts of some welds were not bonded, it was not unexpected that some would fail early and some would continue to function long after. Nor was his opinion inconsistent with the fact that the pipeline had passed a pressure test; the test itself would have introduced damage into the low quality welds and could have accelerated their failure. Ordinary water pressure was sufficient to cause a crack to propagate if the weld was sufficiently weak.
Mr Clisham’s Opinion
[40] Mr Clisham agreed with many of Dr Bowman’s observations regarding the quality of the failed welds and the welds analysed by Exova. They were of variable quality and there was evidence of contamination. One of the intact welds examined by Exova was especially poor. But for welds to fail after only three days, they would have to be very weak indeed. Dr Bowman’s analysis did not explain why two welds which were operating at the lowest pressures were the only ones to fail, while others at higher pressure continued to function for a further eight months. It was highly improbable that these were the worst two joints because there was nothing to indicate that they were made differently from the others. In Mr Clisham’s opinion, although some of the welds were weak, they were nevertheless fit for purpose.
[41] In many years of practice, having examined thousands of welds, Mr Clisham had encountered only two butt fusion weld failures in polyethylene pipes in the UK. Welding failures were exceptionally rare in the United Kingdom, despite there being evidence of much substandard practice. It was hard to believe that the level of stress to which this pipeline would have been subjected could cause slow crack growth. Even the pressure test was not at a high pressure; the pipe should have been well capable of surviving it and probably did. It was accepted that the welding operatives had not followed the guidance in WIS 4-32-08, although this was understandable as it appeared that they might have been misled by the manufacturer’s instructions, and did not imply poor quality control on site. It was likely that some of the welds treated by Exova as failures would now pass. In other countries, notably the United States, different methodologies were employed which included continuing to apply pressure to the pipe ends during the heat soak phase, which was what the operatives here were criticised for having done. One consequence of the adoption of the incorrect method would have been to push any contamination out into the beads. If contamination due to failure to carry out dummy welds had been the cause of failure, one would have expected to see more failures occurring during the next eight months.
[42] Mr Dolecki’s analysis was, in Mr Clisham’s opinion, a much more credible explanation for the pipeline failure than substandard welding. The stresses generated by a failure of the slope would have been sufficient to result in the failure of the pipe even if the joints had been welded in strict accordance with specification. Even if Mr Dolecki was wrong, it would still have been extraordinary for the pipe to fracture because the welds failed. If, however, there was no slope failure, it was the only credible explanation.
Discussion: Butt Weld Failure
[43] Mr Clisham’s evidence, based upon many years’ experience, that butt weld failures in polyethylene pipes are extremely rare cannot be lightly disregarded. Bearing this in mind, and also the fact that other welds in this pipeline, including at least one weld subsequently found to have been of very poor quality, did not fail during the eight months before they were taken out of service, it seems that if the pipeline fracture is to be attributed to a weld failure, the circumstances must have been exceptional. On balance of probabilities, I am persuaded by Dr Bowman’s evidence that such exceptional circumstances did exist. In the first place, it was common ground that 1000mm pipelines are themselves very rare, although Mr Clisham considered that the pipe width (which was not unusually large) was more important than the pipeline diameter. Dr Bowman noted that the vast bulk of UK butt welding is on small diameter pipes, and is carried out by automatic welding machines which eliminate operator error. It is not disputed that the procedure adopted on site here was disconform to industry guidance, and that it was capable of producing welds with a cold area in the centre of the pipe where adequate fusion would not occur. It does not appear to me to assist that other countries adopt methods in which axial pressure is maintained during the heat soak phase, because what was done here did not conform to those methods either.
[44] If contamination were founded upon as the major cause of the two weld failures that occurred, I would have some difficulty with the analysis because, as Mr Clisham pointed out, the postulated systematic failure to carry out dummy welds would have been likely to affect a great deal more of the welds in the pipeline, yet there is no evidence of such widespread weakness. I understood Dr Bowman’s conclusion to be that contamination was a contributory factor in the failure of welds which were already weak due to inadequate fusion. In the end, the defender’s hypothesis, based upon Dr Bowman’s evidence, depends upon the presence of at least one weld which was so weak as to be capable of failing under normal water pressure. For my part, I find the hypothesis to be plausible. Support for it is, in my view, provided by Dr Bowman’s analysis of the surface of weld 6589-2. I accept his opinion that the “acne” effect in one segment of the pipe is consistent with scouring damage that occurred during the first phase of the crack growth, followed by (i) rapid brittle failure of most of the rest of the weld, and (ii) a hinge effect at the opposite segment which was last to fail. It may be that the same occurred in weld 6569-1, and that both failed, as Dr Bowman believed, at a similar time. Either way, having regard to the undisputed variability in quality of the welds due to use of incorrect methodology, I find that crack propagation due to at least one exceptionally weak weld was the most likely cause of the pipeline failure. I bear in mind also Mr Clisham’s very fair acknowledgment that if Mr Dolecki’s analysis was not accepted, a weld failure was the only credible alternative.
[45] Nor, in my opinion, is this conclusion inconsistent with the eye-witness evidence of Mr Waddell. It was suggested on behalf of the pursuer that if the failure of the pipeline had occurred by means of an initial leak, the water pressure in the pipe would have created a feature akin to a fountain which would have been observable by Mr Waddell. There is some support (having regard to the apparent orientation of the pipe in the photographs of the slope following the failure) for a finding that the initial leak postulated by Dr Bowman must have taken place in the top segment of the pipeline, although it is debatable whether this is sufficient to support the notion of an observable fountain of water at some stage of the process. In any event there is no reason to conclude that the process was at that particular stage at the time when Mr Waddell drove past. His description of a V-shape in the slope, a wet patch above, and a section of visible pipeline suggests that the pipe rupture had reached a fairly advanced stage, albeit not yet to the point where the flow was interrupted altogether, and that if indeed there had been a stage at which a fountain of water from the initial leak could have been observed from the road, that was no longer the case.
Disposal
[46] For all of the foregoing reasons, the pursuer’s case fails. I shall repel the pursuer’s pleas in law, sustain the defender’s second plea in law, and grant decree of absolvitor. Questions of expenses are reserved.