TWI Ltd
Chapter from:
Pressure Systems Casebook: Causes and Avoidance of Failures and Defects John Wintle (Editor)
London, Professional Engineering Publishing, 2004.
ISBN: 1860584217
Synopsis
A reliable technical failure investigation is an essential starting point for any subsequent analyses of the broader aspects of the cause of a failure, efforts to prevent recurrence and/or legal proceedings.
Such an investigation can be broken down into five principal stages: initial data gathering; determination of failure mechanism; determination of the sequence of failure; determination of the primary cause of failure; and reporting. At each stage, in all but the simplest cases, a multidisciplinary input is required, and wherever possible, hypotheses should be tested by experiment.
1. Introduction
A failure investigation which is to produce reliable technical conclusions requires many things - expertise, experience, good facilities, and many other skills and qualities. Two aspects which are of particular importance, however, are a multidisciplinary approach, and the testing of hypotheses. In very few cases is it possible for even the most experienced consultant to 'go it alone'.
TWI has a procedure for conducting failure investigations, and before any of the more obvious activities occur, there is a requirement for data gathering. Design and fabrication data may contain the cause of the failure; for example inadequate strength, no allowance for fatigue loading, or incorrect materials selection. Such data will in any case, be needed for comparison with forensic evidence. Was the intended material actually used? Were weld sizes according to design? Again, information on the history of operation may contain the explanation of the failure. Information is required on service loading and environments, including whether there was any cyclic stressing, and the durations and frequencies of different service conditions.
Finally, the context of the failure must be determined. What was happening at the time? The value of witness statements should not be underestimated; even when an eyewitness account may not be sufficiently robust for a court of law, it may provide valuable pointers to investigators.
It will be evident from the above, that a reliable investigation requires the assembly of a multi-disciplinary team, comprising at least a materials specialist and a structural integrity engineer, but also with input from process engineers and others.
From the occasion of the initial site visit and data gathering, a failure investigation will cover the following four main activities:
- Determination of failure mechanism(s).
- Determination of sequence of failure.
- Determination of cause of failure.
- Reporting.
It is then desirable to go one step further, and to apply what has been learned to the avoidance of similar events in the future. In practice, the execution of these stages will overlap, but it is helpful to consider them independently.
2. Determination of failure mechanism(s)
The first stage of the examination requires close attention to detail, and meticulous record keeping. Initial examination will include photography, dimensional checks, and possibly preparation of fracture replicas. There will be no opportunity to go back later, once destructive work has started.
As soon as it is certain that destructive work can start, detailed laboratory examination will be needed to help determine fracture modes. Typically examination at higher magnification, using both light and scanning electronmicroscopes (SEMs), metallographic sectioning, and chemical analysis will be involved. Micro-analysis in the SEM is often particularly revealing. At this stage, however, there is generally no need for experimentation, unless the fracture mode is unusual, and has to be reproduced to allow the mechanism to be identified with confidence. One occasion when this was the case involved explanation of cleavage fracture in super duplex stainless steel ( Fig.1).
A sub-sea super duplex stainless steel manifold hub had leaked, prior to first service, and the fracture mechanism was cleavage. In this material, this can be due to inherent low toughness or hydrogen cracking. Charpy tests indicated that the former was unlikely, but it was believed that stress levels approximately equivalent to the UTS would be required to induce hydrogen cracking from cathodic polarisation (the only feasible source of hydrogen). The unusual coarseness of the microstructure ( Fig.2) was clearly a factor, but laboratory experiments were necessary to confirm that the cracking was due to hydrogen induced embrittlement leading to cracking at stresses of above yield.
It must not be assumed, however, that sophisticated equipment is the key to successful determination of failure mechanisms. A well equipped laboratory is desirable, but the correct interpretation of micrographs, microanalysis results and the like requires expertise and experience. Also, many quite simple techniques can be as important as the sophisticated equipment. Fig.3 shows, for example, how just the correct choice of etch was able to reveal that a defect in an electric resistance weld was a forging lap rather than a bond line crack.
3. Determination of sequence of failure
Few major failures consist of one fracture event. The sequence of failure may involve several stages, for example, a poorly designed weld may lead to an initial fabrication defect that grows by fatigue, fails by ductile or brittle fracture, and causes other failures, from overload. It is essential that the primary failure is separated from consequential damage, and in many instances, the sequence of failure will not be apparent at first sight. Particular care needs to be taken when, as is not uncommon, the entire failed component is not available.
Much will have been achieved in the first stage, but experience of 'reading' fracture faces needs to be applied to help identify the actual initiation site. Many people, seeing the bolt fracture face in Fig.4a will correctly recognise a fatigue failure. Many, however, will also be mislead into identifying the fracture propagation to be from top to bottom, because of the curvature of the beach marks. In fact, multiple initiation around the bottom edge (as evidenced by the ratchet marks in Fig.4b) has produced an initial curved crack front which flattens out as it propagates from bottom to top.
Very often, non metallurgical evidence can be as important as metallurgical evidence. Growth of various organisms on surfaces in maritime environments ( Fig.5) can help date the formation of a fracture surface, as can other contaminants which can be dated, in particular, paint. Paint on the surface of a lamellar tear where the hydrophone port separated from the structural member in the Alexander Kjelland was one of the prime pieces of evidence which confirmed that this fracture was the cause, rather than a consequence, of the failure.
4. Determination of cause of failure
The cause of failure should not be confused with the mechanism of failure. The question which is being addressed is why did this component fail, when others have not failed? Many contributions to ultimate failure (applied stress, imperfections...) are present in most operating equipment. The cause of failure is the condition or combination of conditions which are outside design or normal operation.
One practical example of this is a cleavage fracture in a casting, which had propagated from an approximately 25mm x 12mm shrinkage defect, and this also demonstrates the power of fracture mechanics. When the measured toughness ofthe casting and the known defect size were fed into a fitness for purpose assessment, it became evident that the cause of the failure was not the defect, as might have been assumed, but a significant overload.
It is at this stage of a failure investigation that testing is most often required, to help confirm hypotheses - most commonly corrosion testing, fatigue testing or fracture toughness testing, as in the example above. Nevertheless, it is sometimes necessary to go beyond fracture mechanics, as evidenced by the Union Oil failure in 1984 ( Fig.6).
The geometry of the vessel was simple ( Fig.7), the pressure at the time of failure was known, and the size of the initiating defect was evident ( Fig.8).
Fracture mechanics calculations, however, indicated that fracture toughness of the material should have been 0.064mm, whereas the measured value was 0.17mm. It required a metallurgist with an understanding of the effects of hydrogen on properties of steel to solve the puzzle. When fracture toughness tests were repeated on corrosively hydrogen charged samples, the measured values were 0.064mm - 0.096mm.
In all the stages of a failure investigation, it is important to avoid being side-tracked. The quality of the welding in Fig.9 is poor by any standards, but the fatigue failure from the weld toe has not been influenced by this at all.
5. Reporting
Finally, a very important and frequently underestimated part of the technical investigator's job is to report the findings. It is usually necessary for the report to be understandable by non-specialists, and even people with very limited technical background. The report needs to be logical, precise, concise and yet complete. A very useful touchstone, to apply to every statement in the report, is 'could I defend that in court?' Any reputable organisation will subject reports to an internal peer review procedure prior to release.
6. Conclusions
Reliable technical failure investigation requires above all, experience, and multi-disciplinary expertise. Good practical facilities are clearly necessary, but not sufficient in themselves. Any hypotheses developed in the investigation need to be tested by experiment and/or calculation to allow confidence to be placed in them. Finally, top quality reporting is required to ensure that the results can be put to best use. TWI undertakes technical failure investigations of a range of sizes, to determine the mechanisms, sequence, and causes of failure. The application of metallurgical forensic techniques combined with engineering failure analysis provides a robust basis of fact and insight from which the contributory factors can be identified. Overall, it is found that most components can tolerate fabrication defects, severe environments, or high loads. Failures generally occur when there is a combination of actors, and nothing is going your way.