Forensic Failure Analysis

This example shows how Zencrack was combined with Abaqus to predict 3D crack growth as part of an investigation to explain failure of single cylinder bridge roller bearings. The rollers in question, part of a key bridge structure on the UK motorway network, began to fail several years after a major refurbishment programme. The decision was taken to replace all 148 bearings with considerable implications:

• the bridge is over a mile long, more than 100 feet high, weighing over 58,000 tons
• it carried 150,000 vehicles per day
• the bridge has 35 piers
• repairs required 320km of scaffolding
• £52 million cost for all repair work

Finite element analyses were conducted to gain an understanding of the stresses caused during operation and to explain the possible cause of failure modes and cracking mechanisms. The fatigue load cycle is generated by repeated rolling contact between the roller and plates due to ambient temperature changes.

A number of analyses were performed to investigate growth of initial cracks at one end of the roller. Of those analyses, model plots are shown here for analysis of a perfectly aligned roller (i.e. no contact with end guides) having diameter 120mm, length 480mm and initial corner cracks of radius 10mm.

Key features of the analyses included:

• The bearings allow for expansion and contraction that occurs due to changes in temperature and also to allow for movement of the viaduct as it carries traffic. Analysis therefore includes weight of the bridge deck and a full roll cycle representing the expansion and contraction of the deck.
• Contact is used between the roller and the surrounding plates. Rigid surfaces are used for the plates with the bottom plate fixed and the top plate given a vertical load followed by horizontal translation to represent the load cycle.
• Contact is used between the crack faces where there is potential for mode II and mode III effects.
• A mode II crack growth law is used due to dominance of shear cracking.

The analysis was able to produce crack growth profiles that matched those seen in failed rollers. The predicted failure time was consistent with the actual failure time of the rollers.