Know the how and why - learning from failure

TWI Bulletin, January - February 2004

TWI investigates new failures all the time, not just in metallic components, but also in plastics, adhesively bonded parts, microelectronics, printed circuit boards and a host of consumer products.

John Wintle is a Consultant Engineer for structural integrity and is a leader in the development of reliability engineering at TWI. He takes a keen interest in failure investigations, particularly of pressure equipment, and is a strong advocate of TWI's multi-disciplinary approach.

As well as being a world centre for materials joining, TWI is expert at knowing how and why materials come apart. Failure investigation is an important and increasing part of TWI's business. John Wintle describes the kind of failures that TWI is called on to investigate, the reasons why the knowledge gained may be useful to our clients, and the approach that TWI takes. Failures in which TWI played a part illustrate the process.

Failure investigation can involve equipment at any stage of the product lifecycle, from faults resulting from failings in the manufacturing process, early service failures due to poor design, to later failures where equipment may be older and damaged through service. TWI has investigated failures of many different types of equipment from all kinds of industries. From small microelectronic components through to large offshore structural members: railway wagons, pipelines, bridges, turbines, plastic valves and piping, printed circuit boards, and ships - TWI has seen it. In the most serious of cases, the failures have led to catastrophic consequences in terms of damage to other equipment, loss of production, and risks to workers' health and safety.

Failure can also be less than catastrophic if detected early enough. It may mean a state of defectiveness, a lapse in manufacturing quality control or inspection effectiveness, the use of incorrect or below specification material, or unexpected in-service deterioration or damage through poor operation or maintenance. Even in these cases the financial consequences can be high in terms of rework, repair or remediation.

Motives for investigation

There are many reasons why an organisation may want to know the cause of a failure. In most cases, investigations are made for commercial reasons with a financial interest in establishing the cause. When this is established, someone or some organisation may be asked to explain or compensate those who have lost out financially as a result of the incident.

As many organisations are insured against financial loss from engineering failures, insurers often pursue the task of seeking compensation. They will have to fund necessary repairs, pay for lost production and even make good lost profits subject to the relevant insurance policies in force. Insurers are interested in detailed and rigorous investigations of the causes when they need to know whether or not the failure was the result of a peril covered by the insurance policy. Loss adjusters are frequently engaged to determine the extent of liability of the claimant and of insured losses. They often rely on TWI to carry out and communicate the results of the more technical aspects of the investigation.

In some cases there are many parties implicated in a failure. Each will be just as keen as the party that suffered the loss to know the cause and assess the merits of any possible claims or defences to potential claims for compensation. TWI's procedures for handling these complex investigations are designed to add technical insight but avoid TWI becoming embroiled in potential conflicts between Industrial Members. As a result of its independent status, parties have often accepted TWI as the sole investigator in contentious disputes.

Failures always have the potential for disputes: between insurers and claimants, between suppliers of goods and services and their purchasers, and between the UK's HSE and the parties implicated when health and safety has been put at risk. On some occasions these disputes boil over into litigation, and in the case of health and safety, a criminal prosecution. The potential for litigation is always in mind when TWI undertakes failure investigations, and many of its experts are trained to provide expert evidence and attend court.

On other occasions, the only party may be a single organisation that has suffered loss, sometimes a company with sole responsibility for what has happened. Here, the motives for investigation may be to highlight failings in company procedures and practices, and more importantly to make recommendations for changes that will prevent reoccurrence. TWI may work alongside company investigators, providing specialist services and technical knowledge.

Many incidents have more than one cause in the sense that there are often several factors contributing to the failure, and possibly several parties in error. It may not be clear where responsibility should fall in such circumstances. These are matters to be debated by lawyers, insurers and loss adjusters. TWI's role is to undertake technical failure investigation and to be good team players with other professionals.

The TWI approach

A reliable technical failure investigation is an essential starting point for any subsequent analysis of the root causes of the failure. Technical failure investigation aims to establish the technical facts surrounding the failure, starting from a scientific description of the damaged parts, and then to give an analysis from the standpoint of the materials and the engineering of the ways by which the damage could have occurred. Producing reliable technical conclusions requires many things: expertise, experience and diagnostic facilities are needed, but above all, a multi-disciplinary approach of materials specialists and engineers with a breadth of vision.

The TWI approach to failure investigation takes place in five stages:

• Gathering the initial data
• Determining of the failure mechanisms
• Determining of the sequence of failure
• Determining of the causes of failure
• Reporting conclusions

At the start of any investigation, there is a need for data gathering. Information on the technical design and fabrication specification of the equipment, the service history and its mode of operation will be needed for comparison with forensic evidence from the failed parts. A site visit to examine and photograph the damaged equipment in-situ is often helpful and sometimes essential. Such information may indicate the cause of failure. Was the intended material actually used? Were weld sizes according to design? Was cyclic loading allowed for? What was happening at the time of failure?

In the second stage, the failed equipment or damaged parts are removed from site and sent to TWI for closer examination. This will include photography, dimensional measurement and possibly the preparation of replicas of fracture surfaces. There will be no opportunity to go back once destructive work has started. Once complete, samples of material can be extracted for detailed laboratory examination, chemical analysis and mechanical testing.

Examination using optical and electron microscopes can reveal features of the fracture surfaces that can indicate fracture modes such as ductile growth or cleavage and mechanisms preceding failure such as fatigue, stress corrosion cracking and embrittlement. It can reveal defects in welds and the quality of welding and weld microstructure. Chemical analysis can confirm whether the materials were as specified and detect corrosion products. Mechanical testing of tensile and fracture properties can provide quantitative measures of strength and susceptibility to defects.

Few major failures are the result of one single event, although there is generally a point in time or place at which an irreversible train of subsequent damage ensues - the breaking of a member or the penetration of a pipe wall. The sequence of failure will usually involve several stages: a poorly designed weld may lead to a fabrication defect that grows by fatigue to fail by ductile fracture and causes other failures from overload. It is essential to separate the primary failure from consequential damage, but where there are many failed components this may not be obvious at first sight.

An analysis of the operating history may provide clues to the sequence of events. Evidence can also be obtained from microstructural examination of the fracture surface, which can identify prior damage such as fatigue or stress corrosion. The appearance and thickness of oxidation, paint or marine organisms on a fracture surface can approximately date its formation. Establishing the sequence is often like a detective story where several possible scenarios need to be logically tested to determine the one that fits all the facts. Sometimes there will be more than one.

Once a credible sequence of events has been suggested, the hypothesis is more closely analysed and quantified, and all the influencing factors are determined. Confirmatory testing may be required at this stage, commonly corrosion testing of material susceptibility, fatigue or fracture mechanics tests. Sometimes new data changes the theory or introduces a new element. In one instance, what at first sight appeared to be failure of a defective casting turned out to have been caused by a considerable overload.

The investigation in the explosion of the amine absorber tower at Union Oil's refinery in Chicago in 1984, which killed 17 people and cost an estimated $100M, found a hard heat affected zone microstructure as a consequence of a non-optimised weld repair procedure. This had led to hydrogen induced cracking (HIC) in the sour H ₂S environment and the growth of large cracks in the vessel. When fracture tests were made on material retrieved from the failed vessel, the results were not consistent with the low toughness required for the defects and stresses known to have been present in the vessel to have caused failure. It was pointed out that hydrogen could have been an embrittling factor at the time but had since diffused from the metal. Tests on hydrogen charged specimens did indeed prove this to be the case.

The possibility of human error and fallibility must always be in the failure investigator's mind. An awareness of industrial conditions and human factors is essential to be able to identify situations where failures could have initiated. Sometimes otherwise conscientious people take unnecessary risks, such as forcing a part and damaging it in the process or neglecting to inspect it properly.

Presenting the evidence and drawing the right conclusions is probably the most important part of an investigation. Failure investigation reports need to be logical, precise, concise and technically complete, and understandable by people who are not technical specialists. A good report takes no hostages in presenting the facts as they appear, yet avoids being judgmental or conciliatory. As reports may be used as evidence in court, every statement needs to be defendable and weighted appropriately.

Well known failures

Most failures that TWI investigates are confidential to the client and the results never reach the public domain. For understandable reasons, organisations experiencing failures do not want them publicised. However, a few failures in which TWI has been involved have been of such importance that the results of the investigations have been published.

The failure of the Alexander L Kielland accommodation platform in the North Sea in 1980 cost 123 lives. One of the five columns, which were the principal buoyancy aids, broke off completely, the platform heeled over and eventually capsized. The separated column and the fracture surfaces from the broken elements on the platform were recovered for investigation. Although the main investigation was carried out in Norway, TWI was called to act as an independent technical adviser for one of the parties and took part in the ensuing legal action.

The results of the enquiry concluded that structural failure had originated in one of the braces from a pre-existing fabrication crack in the fillet welds between the brace and a support to a hydrophone. Poor weld penetration and unsatisfactory weld bead shape combined with low through thickness ductility of the hydrophone material led to cracking of the fillet weld. The fabrication crack then grew by fatigue into the brace, extending partly by ductile tearing until it was two thirds of the circumference when final failure took place by brittle fracture. A welded attachment that may have seemed not to have any structural significance turned out to be the critical element.

The Merchant Vessel Kurdistan was an all welded oil tanker that in 1979 literally cracked around the hull, hinged about the deck, and then separated into two parts. It was designed as an ice class vessel, and at the time of the failure was carrying a cargo of heated oil at a temperature of around 60°C in the icy seas off Nova Scotia where the air temperature was around zero. A considerable amount of oil was lost into the sea with consequential environmental damage.

TWI's examination of the hull revealed that the fracture had originated at a defective butt weld in the port bilge keel. There was lack of penetration in the butt weld and in some places there was no weld at all!

Although specified to construction category Ice Class 1, the Kurdistan was built almost entirely in Grade A steel with no specified Charpy requirements for ductility. Subsequent testing after the failure showed that the ship's shell plates had 27J Charpy transition temperatures of between five and 20°C. The steel in contact with the cold sea water would therefore have been below its ductile to brittle transition temperature.

Calculation of the thermal stresses in the ship resulting from the temperature difference between the cargo of warm oil and the cold sea indicated that a high tensile stress would have been present in the shell and bilge keel. It is thought that the stresses due to the impact of a wave on the bow and bending moments on the hull superimposed on the high thermal stresses were sufficient to trigger brittle fracture from defects in the bilge keel. The toughness of the plate was insufficient to arrest the propagating crack and with no crack arrest hole, complete failure ensued.

The investigation of the failure of the Kurdistan showed two important omissions of the requirements at that time for ships designed as Ice Class vessels. Firstly, that such a ship could be built of steel with no impact test requirements, and secondly that there was no requirement to calculate or to consider the effects of thermal stresses for cargoes at temperatures below 65°C. This failure showed, yet again, how critical the quality of workmanship can be when there is all-welded construction.

Although now quite old, the investigation of the failure of the John Thompson pressure vessel in December 1965 was important because it led to the proposal that fracture mechanics principles should be used when setting fracture avoidance criteria for high strength steels. It also recommended that pressure tests be carried out at temperatures above the ductile to brittle transition in vessels designed to operate in this range in order to reduce the risk of failure during test.

This massive vessel was manufactured for use as an ammonia converter at ICI's Immingham plant. The vessel was designed to be used at 120°C and had a design pressure of 35N/mm ². It was hydraulically tested at a temperature of less than 10°C and suffered catastrophic brittle fracture at a pressure of 34 N/mm ². Four large pieces were thrown from the vessel, one of them weighing approximately two tonnes landed some 46 metres away. A piece of this vessel still stands in the foyer of TWI's new Technology Centre. Although there was only one minor casualty, the financial loss was great.

The investigation revealed two fracture initiation sites. These were two small pre-existing cracks in the heat-affected zone (HAZ) between the submerged arc weld joining the end flange forging to the vessel shell. The cracks were located on the forging side of the weld in regions of segregation where the concentration of carbon and alloying elements was locally increased. This would have increased the susceptibility of the material to hydrogen cracking, the probable cause of the original cracks. Another factor was the welding procedure where pre-heat was discontinued immediately on completion of the weld thus reducing the opportunity for hydrogen escape.

In combination with pre-existing cracks, the weld had poor toughness properties as a result of inadequate heat treatment, and the toughness of the shell and forging material, although meeting the requirements was not sufficient to arrest a running crack through the cross section. Investigation indicated that the required heat treatment temperatures had not been achieved and that unrelieved residual stresses from the weld would have been a contributory factor. The Charpy energy at the test temperature was on the lower shelf for this material, which was sufficient to initiate the cracks and fail the vessel. Had the test been carried out at a temperature closer to the intended operating temperature of the vessel, the Charpy energy would probably have been sufficiently high and the failure would not have occurred.

Recent investigations

More recent investigations cannot be identified but the following gives an idea of the kind of work that TWI has been doing.

In one recent investigation, TWI was called by a client to examine the actuator mechanism from a large valve. The actuator worked by means of a hydraulic piston acting against a large spring which was the fail safe device. The piston turned a yoke arm, which, through a series of keywayed shafts, turned the ball of the valve.

Many parts of this mechanism were broken or had sustained damage, including parts loaded only on either the opening and closing strokes. The puzzle was to fathom out how so many parts came to fail and the failure sequence. An analysis of the stress distribution at different times seemed to suggest that the damage had occurred incrementally on both opening and closing strokes. Linking a complex operating history with the possibility of overload combined with poor material made this a most interesting case.

On another occasion, TWI was called to an oil installation which had been destroyed when a pipe carrying crude oil failed, created a vapour cloud and caused a fire and explosion. It was clear that the pipe was badly corroded, but where had the water come from? A plan showed that a building had existed before the explosion whose corrugated roof had overhung the failed location of the pipe. Did the roof have a gutter and was this the source?

Failure investigations do not necessarily involve large components. TWI's Microelectronics Group was called to investigate a failure on printed circuit boards and packaged devices. Special microscopy equipment was used to examine electrical connections for mechanical damage and poor joining and performance tests carried out using environmental chambers.

The failure of a double walled subsea pipe assembly led to recognition of the problems in designing and fabricating pipe-in-pipe assemblies and a subsequent Group Sponsored Project for TWI. The inner pipe carrying the product had high integrity girth welds whereas less care had been taken with the welding of the outer pipe, which was to retain insulation in the inter-space between the pipes. When the outer weld failed, the loading in the combined assembly was concentrated just on the inner pipe at that point, which subsequently also failed.

Conclusion

This article has illustrated the kinds of failures that TWI investigates and how it goes about the job. TWI is investigating new failures all the time. In addition to mechanical and structural welded components, TWI also investigates failures in plastics, adhesively bonded components, microelectronics and printed circuit boards, and consumer products. TWI hopes you do not experience failures in your business... but should you have a failure, you know where to come.

* Crown Copyright material reproduced with permission of the Health and Safety Laboratory