Motorway safety fences - welded barriers in the spotlight
TWI Bulletin, September/October 1988
Martin Ogle, BSc(Eng), ACGI, PhD, CEng, MICE, MWeldI, is Principal Design Consultant in the Engineering and Materials Group at The Welding Institute.
The Welding Institute has been conducting a research programme for the Department of Transport to assess the quality of butt welds used in corrugated beam safety fences. These fences have been installed on the central reserves of most motorways and many dual carriageway trunk roads in the United Kingdom over a period of almost 20 years. They are now considered to be an essential requirement for all such roads, where traffic flows are high, to avoid the horrific consequences of head on collisions.
The corrugated beam safety fence must be more familiar to the travelling public at large than any other structure. It is continuously seen in the corner of our right eye, whenever we are travelling on main roads. In fact, it is the very continuity of these structures that is essential to their proper performance.
The fence is in fact a very long tension member supported on light steel posts. The horizontal member or corrugated beam with its familiar 'W' shaped profile is made from 3.5m lengths of rolled steel strip 3mm thick. These are lap jointed together with bolts located within the hollows of the 'W' to maintain a smooth rubbing face to the traffic. Bolting ensures that damaged sections of fence can be removed and new beam sections inserted quickly in the event of a serious impact. After bolting together the sections the beam is then tensioned, to take out any slack in the joints, using long threaded rods.
These fences are designed to contain a 1.5t vehicle travelling at 70 miles per hour (112 km/hr) hitting the fence at an angle of 20°. The mechanism of containment is that the beam is deflected backwards by a metre or more. In the process, the posts are bent over by the vehicle and the small bolts attaching them to the beam break off. The beam thus remains at its original height and the vehicle slides along, the twin grooves made by the 'W' of the beam preventing the rail from riding over or under the vehicle. At the peak deflection an axial tension of up to 33 tonnes may be mobilised in the rail. Provided this force is resisted by the beam's lap joints the vehicle will eventually be redirected back on to the carriageway at a small angle to the fence without severe damage to itself or its occupants.
The corrugated safety fences are ground anchored using an angled beam section
The safety fence is designed to contain and redirect a 112 km/hr 1.5t vehicle approaching at 20°.
The problem
In the last two to three years a particular problem has been highlighted by reports of very unusual behaviour of the safety fence in the region of terminal anchorages. Anchorages occur where there has to be an interruption in the fence at say crossing points for emergency services. At such locations the beam is ramped down at an angle of 6° to a buried anchorage block which is designed to carry the tension from the beam into the ground.
The change of direction is achieved by using a special beam section called an angled beam. In the past these have been made by cutting a standard beam in half using a mitred cut and re-welding the halves with a full strength butt weld at the required 6° angle. This operation demands careful jigging and electrode manipulation to ensure proper penetration of the joint and freedom from adverse defects. On earlier beams the weld was made from one side only, but later on this was changed to both sides.
The unusual behaviour of the safety fence in this location which came to the public's attention involved failure of this butt weld. Although these reported failures have been isolated incidents in recent years, they have in some cases involved very severe casualties and fatalities.
The reported mechanism of failure has generally involved impact by a vehicle, usually a car, at a position some metres away from the angled beam. The forces generated by the impact have been transmitted down the fence ahead of the vehicle. In the particular cases of interest the weld appeared to have fractured instantaneously. This resulted in the downstream half of the angle beam springing out like a barb in front of the upstream half. As the car slid along the fence this barbed section pierced the vehicle, sometimes with appalling consequences. On subsequent examination of the fence, the welds were found to be inadequately penetrated.
This was clearly a serious malfunction of the fence and although the incidence was low, it was important to take some corrective action. In 1986 the Department of Transport resolved to replace angled beams with welds made from one side only, as these tended to be inferior to double sided welds and could be identified on site by eye.
The major problem remained as to how to assess the quality of the remaining double sided welds of which it was estimated there might be some 20 000 in service up and down the country. At this point the Department of Transport consulted The Welding Institute who recommended that a random sample of about 200 angled beams be collected from safety fences all over the country. These were sent to Abington where they were subjected to detailed non-destructive testing, including visual examination and measurement, magnetic particle inspection and radiography. The aim was to obtain an estimate of the extent of lower quality welding, so that the risk of future malperformance could be assessed.
The welded angle beams were impact loaded to destruction in TWI's test rig
The results of the NDT were compared with the results of subsequent tensile tests on sections of the butt welds. Although there was a general trend for welds with a lower NDT quality rating to have lower tensile properties the correlation was not good enough to produce satisfactory NDT guidelines for use in the field to identify welds with inadequate mechanical performance. The problem was that many of the welds which had a low quality NDT rating appeared to perform acceptably in the mechanical tests.
The test loads were recorded on a 100mm diameter pin load cell
From this it was concluded that a programme of on-site NDT inspection to identify and replace low quality welds was not appropriate. It had to be borne in mind that such a programme would have involved considerable cost, disruption to traffic and danger.
The solution
It therefore became necessary to provide evidence for assessing the risk of inadequate mechanical performance of the angled beams, without conducting a nation-wide NDT survey. Upon the results of this depended the degree of urgency needed for a replacement programme, bearing in mind that the cost of special lane closures at all angled beam sites would be expected to run into some millions of pounds, and in itself be a serious source of disruption and danger. The lower the incidence of inferior welds the greater the case for replacement when convenient as opposed to immediate replacement requiring special closures at all sites.
It was decided that realistic full scale loading tests should be done on a sample of at least 50 angled beams selected at random. This meant simulating the structural assembly of the safety fence with its posts, the loading mode and the rate of load application. There was an urgent need for the test data so that a decision about the in-service angled beams could be made at the earliest opportunity. To this end a purpose made full scale testing rig was successfully designed and built by The Welding Institute in three weeks. Over 50 beams were tested to destruction during the following two weeks.
The test rig consisted of a 4.7m long strong back. A double acting hydraulic jack with a 360mm diameter piston was mounted at one end and a 100mm diameter pin load cell at the other. The test angled beam was bolted to special stub beams attached by pins to the jack and load cell. Thus the test beam did not have to be prepared in any way for fixing into the test rig. The rig was designed so that there was easy access to the beam and the supporting post (which was bolted to the beam at the weld location). The downward thrust on the post was reacted by a transverse plate welded to the underside of the strong back. Thus the geometry and restraints on the test beam were similar to those in the real fence.
The jack was actuated by means of a 34litre accumulator containing nitrogen and oil at a pressure of about 200bar. Both sides of the jack piston were pressurised initially as the accumulator was pumped half full of oil. The test was activated by releasing the oil pressure at the rear of the cylinder instantaneously through a large dump valve. The pressurised gas and oil in the accumulator was then free to propel the piston backwards at high speed. The angled beam was stretched axially until rupture occurred either at the bolted splice or at the weld. The total time to failure was typically about 0.25sec, which was similar to the estimated time taken to reach the peak load in the event of an impact on a real fence.
The weld was strong enough in this test to cause fracture at the bolted splice
In some cases the bolted splice failed in bearing
In other cases the failure was caused by inadequate weld penetration
The impact load was applied using compressed nitrogen gas which forced oil into the jack at high speed
Not only the angled beam but the post-to-beam connections were subjected to realistic loading, as the latter also influenced the overall mode of failure.
The axial load history was recorded on an ultraviolet light oscillograph and the peak loads measured off the traces. These results were reported as testing progressed, so that by the time the tests were completed the Department was in a good position to make a well informed statistically based decision about the degree of urgency of replacement of the 20 000 welded angled beams in service.
Comment
This investigation shows how a sensibly planned programme of testing work can reduce uncertainties to a level where decisions can be well founded and quickly made. Too often important technical decisions are made with inadequate data, in which case the inevitable need for 'safety at any cost' can lead to unnecessary expenditure or delay far in excess of that due to testing.
