Repair of a steel viaduct

Martin Ogle, BSc(Eng), ACGI, PhD, CEng, MICE, MWeldI, is Principal Design Consultant in the Engineering and Materials Group.

When fatigue cracks were found in the studs, bolts and fillet welds on a viaduct carrying the M4 motorway in London, The Welding Institute was appointed technical adviser on the repair. The task undertaken involved monitoring structural performance, modifying the design accordingly, and specifying the appropriate repair programme.

When The Welding Institute was involved in design and validation of procedures for repairing extensive fatigue damage in a large steel viaduct, the complexity of the structure's behaviour was such that theoretical calculation could not be relied upon to predict the effects of the repair procedure. By integrating a programme of instrumentation within the contractor's repair programme, the scope of repair work was kept to a minimum needed to ensure satisfactory long term performance of the structure.

This article gives a short account of the structural monitoring and repair work which was carried out between 1979-80 on the viaduct which carries the M4 motorway over Boston Manor Park in London (Fig.1). ^[1] This section of the M4 carries over 80,000 vehicles per day - one of the highest traffic flows in the country. It was therefore essential to devise a programme of repair which kept disruption of traffic to a minimum.

The viaduct consists of seven independent structural systems, one of truss construction and the remainder plate girder. The 13 separate spans are carried on common piers. The reinforced concrete deck is supported by steel cross girders, which in turn are supported by the main girders ( Fig.2).

Discovery of damage

Repair work became necessary following the discovery of fatigue damage to many of the steel cross girders and their connections where they rested on the main longitudinal girders ( Fig.3). ^[2]

The first signs of trouble were noticed in 1972, seven years after the viaduct was opened, when approximately 370 of the 10,000 holding down studs and bolts were found to be fractured through fatigue. Early the following year the nuts on the majority of holding down studs were removed and the broken bolts replaced.

At the same time cracking had been observed in the fillet welds between the bearing stiffeners and the top and bottom flanges of the cross girders. Surveys of the complete structure in 1975/76 showed that 20% of the 6000 bearing stiffeners had fatigue cracks in one or both of the end fillets.

Cause of damage

Preliminary monitoring of stresses and deflections of the fatigue damaged areas of selected cross girders in one of the plate girder spans in 1976 by the Transport and Road Research Laboratory confirmed that the prime cause of fatigue damage was racking of the cross girder under traffic loading. The racking was a result of the relative longitudinal movement between the deck slab and the main girders as the top flanges of the latter shortened and lengthened under bending action. The mechanism, which is illustrated in Fig.4, tries to impose rotations between the flanges of the cross girder and the deck slab above and the main girder below.

Preliminary monitoring confirmed that the predominant source of fatigue damage was individual heavy goods vehicles loaded to their maximum 300-350 kN. Although they only caused stress fluctuations of about 15 - 25 N/mm ² in the main girders, the fluctuations in the bearing stiffeners because of racking could be as high as 100 - 200 N/mm ² . These stress levels could be expected to give lives of two to five years.

The resulting rotations are resisted by the shear connectors in the deck slab and any holding down bolts or studs to the main girders. The presence of the bearing stiffeners above the main girders stiffens the cross section of the cross girders so that there is a natural tendency to pull one side of each flange away from both the deck and main girder as rotation occurs. This results in a number of highly stressed points, which explains the observed fatigue damage in the bearing stiffener welds and the holding down studs and bolts ( Fig.5). Figure 5 indicates that the shear connectors and the top flange of the cross girders were also at risk, although no fatigue damage had been reported in these components.

Choice of repair method

Whilst the fatigue damage was extensive in terms of the number of components showing visible cracking, none of the cross girders was considered to be significantly weakened as far as their primary load carrying function was concerned. For this reason the viaduct had been allowed to carry normal traffic without, as far as is known, any special restriction on loading within its original design limit of HA loading only. Nevertheless, the possibility that fatigue cracks might eventually propagate into more critical components, such as the web-to-flange welds or cross girder flanges themselves, could not be ruled out. As the top welds were more widely damaged it was urgent that a solution be found, firstly for rectifying the existing damage, and secondly for minimising further damage.

Various options for repairing and containing the fatigue damage were considered by the Department of Transport. These ranged from straightforward repair of known damage, to major changes in the articulation of the deck and girders.

The governing criterion had to be the best economic return on money spent over a reasonable working life of the structure. This had to take into account not only the cost of the immediate repair programmes and any consequential disruption to traffic, but also the likely costs of future inspection and repair at the appropriate discounted rate.

The simplest option of repairing the cracked welds and replacing the broken bolts was ruled out as this would not cure the problem of the high stresses and the repair cost would recur regularly. The option of changing the articulation of the structure - so as to eliminate unwanted stresses in the joints - was considered to be very expensive, not only in terms of the work needed, but also in terms of disruption to traffic. Articulation could be changed in various ways, including:

1. Introducing intermediate expansion joints in the deck;
2. Mounting the deck and cross girders on sliding bearings;
3. Fixing the deck compositely to the main girders;
4. Reducing the deflections of the main girders by intermediate pier supports.

Following discussions between The Welding Institute and the Department of Transport it was decided that there was a possibility that the stresses could be reduced and, at the same time, the strength of the joints increased, by replacement of the existing fillets by butt welds. The shrinkage caused by welding would, in effect, make the deck to cross girder joint more freely articulating. The only problem was that the shear connectors would still be attached and some of these would need cutting.

Bracing between the deck and main girders would also be needed to stabilise the more flexible deck sections and the bolts holding down the cross girders at the ends of each deck section would have to be made more flexible to accommodate the articulation.

This scheme contained a degree of uncertainty, because of the difficulty of predicting the complex behaviour of all the affected joints throughout the structure. Preliminary monitoring of stresses in three of the cross girders had shown considerable variations in stress response in joints with similar racking. This difference was attributed to natural variations in fit-up between main girders, cross girders and deck slab at the time of construction.

It was therefore decided to conduct a pilot repair programme in 1979 in three plate girder spans. This was closely monitored by The Welding Institute using strain and deflection gauges. The results showed that the principles were sound, so that the programme was extended over the next two years to cover the whole viaduct. The Welding Institute was responsible for design of all modifications to the structure, monitoring its performance and specifying the scope of repair.

Repair programme

Scope

The overall repair contract carried out in 1979 and 1980 extended over the full length of the viaduct between piers 105 and 122 ( Fig.1). All 429 cross girders were repaired. The main items of repair were:

Replacement of all 6032 bearing stiffener to top flange fillet welds by full penetration butt welds. This was expected to have two effects. Firstly, it would increase the fatigue strength of the joint by a factor of 2.4, which is a theoretical improvement in life of fourteen times. Secondly, the shrinkage of the weld was expected to pull the flange away from the deck slab by at least 0.5mm, and hence lower the stress transmitted ( Fig.6). Vodex electrodes, 3.25mm diameter, were used without preheat for this repair.

ii. Replacement of all 408 bearing stiffener to bottom flange fillet welds on end cross girders by full penetration butt welds and on other specified cross girders by partial penetration butts and fillets. This was necessary because of the tension forces applied to the joint by the holding down bolts on these girders.

iii. Replacement of 1784 selected bearing stiffener to bottom flange fillet welds on intermediate cross girders by 8mm fillets. These were either required where existing welds were cracked or where new bracing was to be attached.

iv. Replacement of all 768 holding down bolts and studs at the heavy cross girder at the end of each by new bolts with Neoprene washers. This was necessary to increase the axial flexibility of the new holding down bolts so that they could still provide vertical pre-stress and carry horizontal loading, see Fig.7.

v. Installation of 142 sets of angle bracing diagonally between top and bottom of adjacent cross girders to resist longitudinal forces, such as braking and out of balance racking forces. These were installed, where possible, at the positions of minimum racking movement ( Fig.8 and 9).

vi. Diamond drill cutting of 443 pairs of selected top flange shear connectors in the neighbourhood of bearing stiffeners to eliminate unacceptable bending stresses in the top flanges (view X-X in Fig.5). This involved locating the studs close to the bearing stiffeners with the help of the original fabrication drawings, and drilling through the soffit concrete ( Fig.10, 11 and 12). This was, as far as is known, the first application of this particular technique, and it worked very satisfactorily.

vii. Provision of bolted shear connections on eight special cross girder joints to provide extra resistance to horizontal forces at the change in direction in the truss spans.

viii. Burr grinding of 64 repair butt weld toes on top flanges of selected end cross girders. Monitoring showed that the fatigue strength of the top flanges of the four most heavily racked cross girders needed increasing by the grinding. The technique is shown in Fig.13. and the contractor's workmen were given instruction by Welding Institute staff.

Phasing of repair and monitoring

It was decided that all repair and monitoring work should be done without closure to traffic on the M4 This meant that no controlled load testing could be carried out, and that the structure was in a constant state of movement during all the repair stages, which put extra demands on the welders in particular. It also meant that special precautions were needed to ensure that unwanted loads were not locked into the new bracing members.

The work was done as a rolling programme one span to the next. The typical sequence operations is shown in Table 1. As each span was structurally different from the next it was important to do the preliminary monitoring work as soon as the scaffolding was erected and before any repair work was started. Monitoring concentrated on measuring the racking movements in the cross girders and strain fluctuations in the tops and bottoms of the bearing stiffeners.

Table 1 Programmed sequence of repair operations

As soon as this was done the top fillet welds of all the bearing stiffeners were air-arc gouged and replaced by butt welds. The results of the initial monitoring enabled the extent of the other items of work to be estimated.

After all welding had been completed the new longitudinal bracing was loosely installed. A time of slack traffic was chosen, usually the early hours of Sunday morning, when the bracing bolts were tightened and the end cross girder holding down bolts loosened and replaced by the new 'soft' bolts. This operation was referred to as 'transfer' as it marked a change in the general articulation pattern between deck and main girders.

The racking movements and strains in critical components, including the new bracings, were again monitored to assess the change in response. Strain measurements were also made on the top flanges of the cross girders and a map of locations for cutting of studs produced for the contractor. The effectiveness of the stud cutting was confirmed by a further check (see below).

Provision of facilities for The Welding Institute's monitoring work was covered in the contract and times allotted in the contractor's programme. Rates for the various repair operations were agreed in advance. As access to each span was only possible shortly before the repair work was due to begin, close liaison between all parties was essential to avoid unnecessary disruption.

Details of monitoring

Principles

The aim of the monitoring was to show that the final fluctuating stress levels were low enough to predict a satisfactory life (nominally 100 years from the time of monitoring). The problem was that the loading at the time of monitoring at any location could not be known. Even if heavy lorries were observed, their loaded weight could not be assessed directly. As a large number of locations needing monitoring (approximately 700), it was not practicable to record a long enough strain history to enable extrapolation for full life. In any case rapid answers were needed if the contractor was not to be delayed.

The principle adopted was to monitor the racking deflection on each span long enough to obtain a representative spectrum of racking deflections. Extrapolation of the linear part of this spectrum on to the y-axis gave the 'datum racking movement' which, by statistical analysis of vehicle weight data, was estimated to be equivalent to that caused by a 365kN vehicle. Thus the racking deflection could, in effect, be used in place of a weighbridge.

The next step was to confirm that the stress range in any critical detail was proportional to the racking deflection in a span ( Fig.14). Whilst the degree of proportionality varied considerably because of local geometry differences, it did mean that, provided that racking movement was always measured together with strain, very few readings of the latter were required to obtain the 'datum stress range' for that vehicle. The principle is shown in Fig.15.

Once the datum stress range for a detail had been obtained, it was possible to predict the probability of failure, knowing the fatigue detail class and the predicted load spectrum. The probability at failure was obtained from the data in Part 10 of the new Bridge Code BS 5400 ^[3] while the predicted load spectrum was estimated from traffic flow data for the viaduct and typical weighbridge data from a site on the M4 near Reading. A typical probability curve is shown in Fig.16.

Monitoring equipment

The requirement for the monitoring equipment was that it should be easily transported and set up, robust and reliable, and give instantaneous results. A typical monitoring station is shown in Fig.17.

Strain measurements were originally made using bonded electric resistance gauges. However, the large number of locations needing monitoring precluded their use on the majority of work. A frictional strain gauge was found to be most convenient. This is held on to the cleaned steel surface with a spring loaded magnetic clamping device (see Fig. 18a). Details of this gauge are given in ref. ^[4]

Racking movement was measured with a linear displacement transducer which was mounted on a movable bracket (Fig.18a and b).

Recording was made on a 10 channel light beam oscillograph. This enabled rapid screening of locations to be made when frictional gauges were used. It also enabled malfunctions to be spotted early. Figure 19 shows a typical record of a strain and deflection channel which has been correlated with traffic observations.

Scope of measurement

The extent of the work throughout the viaduct is shown in Table 2. Most strain monitoring was concentrated in the immediate neighbourhood of the bearing stiffeners ( Fig.14b). However, the opportunity was taken to monitor other sundry details in passing to ensure that they could be given a 'clean bill of health.' The loadings in the new longitudinal bracing were monitored carefully to ensure that the end connections were not at risk in the long term.

Table 2 Coverage of monitoring programme

Results

Detailed discussion of all the measurements made during the two year programme is beyond the scope of this article. However, typical examples are illustrated in Fig.20-22.

Figure 20 shows how the pattern of cross girder racking rotations θ _D in an anchor span matches the incidence of observed fatigue damage (shaded area). The measured rotations were less than predicted by simple theory, showing that the effects of friction and limited composite action between deck and main girders were not insignificant. Note also that the end cross girders tend to be the most heavily racked.

Figure 21 shows the effects of repair welding on the stresses in the tops of the bearing stiffeners on a suspended span. On three of the four cross girders, including both end cross girders, the weld shrinkage produced sufficient drops in stress range. Together with the improvement in fatigue class resulting from changing the original fillet welds to butts, the life expectancy of the majority of these joints has been raised from less than five years to over 100 years.

Figure 22 shows the marked drop in top flange bending stresses when the stud shear connectors adjacent to the bearing stiffeners were cut. The points, which are from more than one span, have been plotted for 0.1° rotation of the cross girders in question. The results were used to prepare a cutting schedule for the whole viaduct, as it would have not been possible to monitor all locations. Reliance on purely theoretical analysis of such a joint would have been out of the question because of the uncertainties of fit-up and fixity of the stud in the concrete.

The opportunity was also taken to monitor strains in trusses. In this instance secondary and local stresses in the region of the gusset were of interest as these are not simple to predict when the joint region is large. At all locations, the stress levels were small enough to predict satisfactory performance for the design life of the structure.

The full results of the monitoring programme have been reported to the highway authority, with recommendations for future periodic inspection of selected areas of the structure. These have been designed to suit the Department of Transport's recommendations for inspection of highway structures. ^[5]

Summary

1. In the repair programme for the Boston Manor Viaduct, instrumentation of highway response under operating conditions by The Welding Institute played an important role in identifying and remedying problem areas.
2. With careful planning at the tender stage, monitoring was used to control the extent and method of repair to keep costs to a minimum, but without disrupting the contractor's programme.
3. Monitoring of structural response is important in effective planning of inspection programmes, particularly in complex structures where calculation methods may be insufficient for prediction of fatigue performance.

Acknowledgements

The author thanks the Department of Transport for agreeing to the preparation of this article, and acknowledges the assistance of B R Piggott, B A Martin and D Jarvis with the instrument programme and of G Slater with the instrumentation, traffic recording and analysis.

References

Footnote: The highway authority involved in this work was the Department of Transport (GLRT and BET divisions); the maintenance agent was the London Borough of Hounslow, main contractor was Linkweld Engineering and Construction; and inspecting engineers were Messrs Sandberg.