Text and Images by Pete Walker
Investigation of bulging, bursting and toppling in dry-stone retaining walls
C. Mundell (1), P. F. McCombie (1), A. Heath (1), J. Harkness (2) and P. Walker(1)
(1) BRE Centre for Innovative Construction Materials Dept. of Architecture & Civil Engineering, University of Bath
(2) School of Civil Engineering & the Environment, University of Southampton
As part of an EPSRC (Engineering & Physical Sciences Research Council) funded project at the University of Bath, a series of full scale dry-stone retaining walls have been constructed and tested to destruction. Each wall was built to a different standard in order to facilitate the investigation of various behavioural aspects associated with dry-stone structures. In particular, the phenomenon of bulging was successfully recreated.
Within this article the development and implementation of the instrumentation used to record the tests is described, as well the testing procedure itself. A full version of this article was presented to the 11th Canadian Masonry Symposium, and as such may be found in the conference proceedings.
There are estimated to be 9000 km of dry-stone retaining structures in the UK lining the road and rail network, mostly dating to the 19th and 20th centuries. Though poorly constructed walls presumably collapsed shortly after their construction, the majority of walls have remained perfectly stable over decades of usually steadily increasing loading and weathering of the stone. However, many walls have deformed or bulged and are regarded as potentially unstable. Because little guidance is currently available to assist engineers in the assessment of these structures , they are often replaced, at great cost. They are very rarely rebuilt in dry-stone, as the dimensions required by current design practice make this substantially more expensive than a concrete replacement. It has been estimated that the total replacement cost for the dry-stone walls lining the UK’s highways would be over £10 billion . Indeed, internationally accepted design practice would deem most existing structures to be inadequate.
There is therefore a clear requirement to have means of assessing existing structures that is realistic. Substantial difficulties exist in obtaining information about individual walls, especially their effective thickness and backfill properties, and there is also considerable uncertainty regarding the appropriateness of current design methods for such structures. Research has been carried out at the Universities of Bath and Southampton to address this. The main focus has been on model and full scale testing linked to advanced numerical modelling. However valuable such computational techniques are for research, they are not suitable for routine work by local engineers around the world, who simply do not have the appropriate expertise and resources. A part of the work at Bath has therefore been to develop a simple computer program, which can be distributed freely and used easily to explore the stability of dry-stone retaining structures.
Relevant data regarding dry-stone retaining wall structures is sparse. The largest reported tests to date were conducted in 1834 by Lieutenant-General Sir John Burgoyne, who constructed four full size test walls in Dun Laoghaire, Ireland. Each wall was built using the same overall volume of square cut granite blocks, but arranged in different sections (Fig 1). Testing consisted of backfilling each wall until the full retention height (6 metres) was achieved, unless premature collapse occurred.
From this work Burgoyne proved that wall geometry has a substantial impact on overall stability. However, as the walls described were constructed of carefully shaped and tightly fitted granite blocks, this could have caused different behaviour to that which would be observed in more traditional walls, where the stones may rotate and move more freely. Regardless, his findings and observations remain the basis for the validation of almost all of the numerical studies carried out to date on dry-stone retaining walls, despite consisting solely of dimensional measurements and visual observations reported 19 years after the tests.
Figure 1: Burgoyne’s test wall geometries
Work of such a physical nature was not conducted again until 2005, when a French engineering team led by Jean-Claude Morel and Boris Villemus built and tested five large scale test walls at ENTPE, Lyon. (Ecole Nationale des Travaux Publics de l’État – National school of Public works) The walls were of various sizes ranging from 2m – 4.25m high, up to 1.8m thick and between 2m and 3m long. Each wall was subjected to hydrostatic forces [essentially water pressure -ed] via a large PVC-lined water filled bag, in order to load the wall using purely horizontal pressures that could be precisely known at all times.
The walls were purpose built as short sections with exposed ends. In this way the cross-sections could be observed during testing, and the internal behaviour of the wall recorded. As the walls were increasingly loaded with water pressure via the PVC bags, the movements of the individual stones were measured, and any planes of internal differential movement determined [That is if one part of the wall moves at a different rate to another creating a fault line – ed] . This data was then used to develop more accurate analysis techniques, accounting for the unmortared nature of these walls and the ability of different sections of wall to move relative to each other.
Despite the advantages of being able to view the internal mechanisms of the wall, this method does have drawbacks. As a consequence of using short lengths, the behaviour will be somewhat different to that of a continuous wall, mainly due to the fact that in a longer wall the adjacent sections will provide support to and influence the loaded section. This might in turn influence the location and amount of movement.
Each of the tests described in this paper were carried out consecutively in a unique outdoor test laboratory situated on the University of Bath campus. To avoid the problems associated with using short sections of wall (as described for the tests at ENTPE), each wall was required to have a significant width/height ratio. Wall spans of 12 metres were chosen, having a height of 2.5 metres through the central test area (including a 300mm cope). The central 4 metres of each wall rests upon a mechanically jacked platform, which has the ability to move vertically as well as tilt forwards or backwards. This allows both foundation and backfill settlement to be imitated, with movements being directed from a remote control station at a rate of up to 10mm/min. In addition, a steel frame is erected over the central portion of each wall, from which a 20 tonne capacity hydraulic jack is suspended, allowing an additional load to be applied through a steel plate onto the backfill (fig. 2). The purpose of this load is to replicate the heavy loads (such as lorries) to which existing walls might be subjected.
Figure 2: Diagram of test setup
The material used to construct the walls was an undressed Cotswold limestone provided by the Natural Stone Market Ltd., requiring approximately 30 tonnes for each test. Limestone quarried from this region generally comes in two varieties, which can be identified by their colour – either grey or a lighter, creamier colour. The grey limestone is generally considered to be much more durable and so has been used throughout this project.
The retained material for each test is a 14mm single sized aggregate, requiring 100 Tonnes to completely backfill the wall and ensure that any failure planes [the surfaces of a fault line – ed] which develop are not impeded by the wall extending either side of the immediate test area. This particular backfill has been chosen to ensure that the retained material is completely free draining, allowing no build up of water pressures which would cause complications when attempting to analyse wall behaviour, because to some extent the actual pressure distribution would inevitably be unknown. Elevated water pressures are certainly a factor in deformation and failure of dry-stone walls, but this phenomenon is better addressed by mathematical calculation/modelling than by practical testing.
TEST WALLS 1 & 2
The first wall was built in June 2007, constructed over 5 days to a high standard, having tightly packed faces with a well finished appearance (fig. 3a). The overall thickness ranged from 600mm at the base, with a battered front face tapering the wall to 400mm at the coping. Through-stones were incorporated at several levels following standard walling practice.
The backfill was placed in layers of 300mm and compacted using a 1kN vibration-plate compactor. Through plate loading tests, the angle of repose of the backfill material [ie the angle at which the backfill remains stable, at higher angles it begins to cascade- ed] was found to be 50.1o, which is much higher than most in-situ retained fills. The fill was placed until a height of 2.2m was attained – 300mm below the crest of the wall, level with the base of the coping stones.
The wall was then monitored for one month and after it was determined that there had been no settlement or movement, the wall was tested over the course of five days. The first day consisted solely of a 40mm raise of the mechanical platform. This was to ensure that the maximum amount of friction possible was generated between the wall and the backfill. In practice, this is a very difficult parameter to measure, and is dependent on both the relative movement of the wall and backfill and also on the properties of the materials themselves. For in-situ walls, it is common for the backfill to settle over time, whereas the wall itself generally will not. Even a slight relative movement is enough to generate the full amount of friction possible, however as this is a long, ongoing, process it is not suitable for these tests. Instead, the wall is raised, causing the same relative movement found behind in-situ walls, hence ensuring the maximum amount of friction.
Following this initial movement, a combination of surcharging [adding additional loads/overloading – as would be the effect of heavy transport – ed.] and forward rotation of the platform was carried out. A rotation of 3.75o was achieved by lowering the front jacks a total of 75mm, imitating the effects of settlement under the toe [outermost edge – ed]. The surcharging was initially at a distance of 500mm from the back face of the wall, applied via a plate. This was later moved to 1m from the back face as it was found that the initial position was too close to the wall, and only affected its upper portion. To compensate for the additional distance, a larger plate of 500mm x 600m was used, allowing larger loads to be applied. The loading mimicked heavily loaded vehicles passing over the backfill.
The first wall eventually toppled. Although some slight bulging occurred, the wall generally moved monolithically (as a single mass), rotating about the toe (fig. 4a). The toppling failure was encouraged by both the initial surcharging near to the coping, and also by the rotation of the platform. The monolithic behaviour was due in large to the tight-knit construction, as the meticulous construction process ensured that each block was placed securely, with no rocking or movement possible.
In an attempt to instil flexibility and encourage bulging during testing, the second wall was built with a sectional profile 100mm thinner throughout. In addition, the backfill was placed uncompacted, giving an angle of repose of 41o, which is much closer to that of typical retained fills. The build time was shorter, reflecting a slightly less precise build, with a looser face and poorer internal fit, thus allowing more individual stone movement.
To maintain continuity between each wall, the second test was in general identical to the first test procedure, but the platform rotation was not applied as it was seen to be mainly promoting a toppling failure. The surcharge load was applied after the initial platform raise, however only the larger plate was applied, at a distance of 1m from the back face of the wall.
Figure 4: Surveyed movements: a) test wall 1; b) test wall 2
Failure was again via toppling, but prior to collapse the wall profile was far from linear. The lower half of the wall had bulged out, with the upper half retaining its integrity and form above this. From figure 4, the two walls can be compared with the differences being apparent. Although the main failure mechanism was toppling, it should be noted that for this wall there was another factor which instigated failure. As the loading progressed, and the deformations occurred, the bulging in the lower courses caused several blocks to move much further than those directly below. Another consequence of these deformations was to shift the wall’s centre of mass further and further forwards. At the time of failure, some of the lower face stones were overhanging those below significantly, with almost the entire weight of the wall passing through them. These loads combined with the overhang were sufficient to force these key stones to rotate and fall out, consequently causing the failure of the entire structure.
During the course of the investigation, tests were conducted to determine the density of the walls. The density of the stone itself was measured at 26kN/m3 [about 2.5 tonnes per cubic metre – ed] and by weighing set volumes of wall the overall density could be determined. To ensure the volumes measured were as accurate as possible, small ‘test cubes’ were constructed, consisting of open-sided timber frames of specific dimensions (generally 600mm x 500mm in plan, 500mm tall). The masons then constructed sections of wall inside the confines of these boxes to a variety of standards (fig 5). In addition to recreating the general standard of the walls used in the tests, examples of high quality and poorer quality walls were constructed.
Figure 5: Sample wall blocks: a) best practice; b) poor construction
It was found from these tests that the lowest percent of voids that a traditional dry-stone wall can have is roughly 20%, although if each block was carefully squared and tightly fitted this value could potentially be lower. Conversely, the poorly constructed block had 40% voids, and even higher values have been reported by Walker et. al.. From further testing, the walls constructed for tests 1 and 2 were found to have approximately 28% voids.
TEST WALLS 3 & 4
It was determined that the third wall should examine further the effect of block rotation, and the ability of individual stones within the wall to roll and rotate more freely. To investigate this process, the wall was built identical in profile to the first wall (600mm wide at the base, tapering to 400mm). This wide cross-section was intended to provide as much stability as possible against overturning failures [essentially toppling – ed], allowing other mechanisms [forms of failure] to develop. The internal make-up of the wall was much different to the first wall, consisting of a much rougher construction, utilising much smaller unfaced blocks ensuring limited use of slabs. The standard practice of back wedging the stones to eliminate movement was largely disregarded, giving each block the ability to rock slightly in place. In short, the wall more closely resembled a much older construction, subjected to several years of weathering and erosion, with a much higher percent of voids than the first two walls. Indeed, once testing began, a small section of the wall failed, leaving a large hole whilst the rest of the wall remained stable (fig. 6a).
Figure 6: Test wall 3: a) local failure; b) prior to total failure
The testing procedure remained identical, with an initial platform raise followed by a surcharge at a distance of 1m until the collapse mechanism occurred. A stable bulge quickly developed, forming the distinct ‘belly bulge’ which is commonly found in many in-situ walls (fig. 6b, fig 7a). This bulge continued to develop as the loading progressed, until collapse occurred mainly as a bursting failure.
Figure 7: Surveyed movements: a) test wall 3; b) test wall 4
As the third wall displayed the bulging and eventual failure mechanisms described within the project goals, the fourth wall was used as a control, repeating the third test wall. The wall construction technique was the same, utilising similarly sized blocks in the same manner as before. This wall was also tested in an identical manner to the third wall, and as figure 7 shows, the repeatability of this test was demonstrated.
The main goals of the physical tests described were to recreate and understand the mechanisms which occur in dry-stone structures in the field, in particular the phenomenon of bulging. Through the four wall tests described in this paper, various aspects of dry-stone behaviour have been investigated, culminating with the repeatable recreation of a stable bulge. This has been linked to block rotation, build quality and overall wall geometry.
The eventual goal of the project discussed in this paper is to develop guidelines and codes of practice to use in the field. Although still requiring further testing, particularly in the area of stabilising distressed walls, this work represents a large step forwards, giving an understanding of the forces at work and the important behavioural aspects of dry-stone walls.
1. Powrie, W., Harkness, R.M., Zhang, X., and Bush, D.I. (2002) “Deformation and failure modes of dry-stone retaining walls” Geotechnique
2. O’Reilly, M.P. and Perry, J. (2009) “Dry-stone retaining walls and their modifications -condition appraisal and remedial treatment”
3. O’Reilly, M.P., Brady, K.C., and Bush, D.I. (1999) “Research on masonry-faced retaining walls” 2nd European Road Research Conference
4. Burgoyne, J. (1853) “Revetments or retaining walls” Corps of royal engineers
5. Villemus, B., Morel, J.C., and Boutin, C. (2007) “Experimental assessment of dry-stone retaining wall stability on a rigid foundation” Engineering structures
6. Walker, P.J. and Gupta, V.P. (1995) “Stability of medieval dry-stone walls in Zimbabwe”
Our thanks to Pete Walker of the University of Bath for given us permission to republish the article Investigation of Bulging, Bursting and Toppling in Dry-Stone Retaining Walls which was published in Issue No.18 of Stonechat (Summer 2009).
This article was first published in Stonechat 18, Summer 2009