Managing Maintenance Error

A Practical Guide

James Reason, Alan Hobbs

• Publisher: CRC Press
• Year: 2017
• ISBN:
  9781351920506
• Edition: 1
• Topic: Technology & Engineering
• Subtopic: Industrial Health & Safety
• Index: Technology & Engineering

Situations and systems are easier to change than the human condition - particularly when people are well-trained and well-motivated, as they usually are in maintenance organisations. This is a down-to-earth practitioner's guide to managing maintenance error, written in Dr. Reason's highly readable style. It deals with human risks generally and the special human performance problems arising in maintenance, as well as providing an engineer's guide for their understanding and the solution. After reviewing the types of error and violation and the conditions that provoke them, the author sets out the broader picture, illustrated by examples of three system failures. Central to the book is a comprehensive review of error management, followed by chapters on: - managing person, the task and the team; - the workplace and the organization; - creating a safe culture; It is then rounded off and brought together, in such a way as to be readily applicable for those who can make it work, to achieve a greater and more consistent level of safety in maintenance activities. The readership will include maintenance engineering staff and safety officers and all those in responsible roles in critical and systems-reliant environments, including transportation, nuclear and conventional power, extractive and other chemical processing and manufacturing industries and medicine.

1	Human Performance Problems in Maintenance

The Bad News

If some evil genius were given the job of creating an activity guaranteed to produce an abundance of errors, he or she would probably come up with something that involved the frequent removal and replacement of large numbers of varied components, often carried out in cramped and poorly lit spaces with less-than-adequate tools, and usually under severe time pressure. There could also be some additional refinements. Thus, it could be arranged that the people who wrote the manuals and procedures rarely if ever carried out the activity under real-life conditions. It could also be decreed that those who started a job need not necessarily be the ones required to finish it. A further twist might be that a number of different groups work on the same item of equipment either simultaneously or sequentially, or both together.

Small wonder, then, that maintenance-related activities attract far more than their fair share of human performance problems. Table 1.1 shows the combined results of four surveys, three relating to US nuclear power plants (NPPs) and one dealing with Japanese NPPs.¹ It can be seen that the proportions of human performance problems associated with maintenance-related activities far exceeded those relating to other kinds of human performance. In three out of four of these studies, maintenance errors accounted for more than half of all the root causes of potentially serious events. Comparable data are not available for other safety-critical industries, but the fact that all maintenance-related tasks have a great deal in common suggests that such numbers would not be wildly dissimilar to those observed in nuclear power generation.

Table 1.1 The relationships between activities and performance problems in nuclear power plant events*

Type of activity	Proportions of human performance problems associated with activity type
Maintenance, calibration and testing	Range = 42−65%
Normal plant operations	Range = 8−30%
Abnormal and emergency operations	Range = 1−8%

* Combined results from three US and one Japanese survey (see note 1)

Maintenance errors have been among the principal causes of several major accidents in a wide range of technologies. These include:

• the Apollo 13 oxygen tank blow out (1970)

• the Flixborough cyclohexane explosion (1974)

• the loss of coolant near-disaster at the Three Mile Island nuclear power plant in Pennsylvania (1979)

• the crash of a DC10 at Chicago O’Hare (1979)

• the calamitous discharge of methyl isocyanate at a pesticide plant near the Indian city of Bhopal (1984)

• the crash of a Japan Air Lines B747 into the side of Mount Osutaka (1985)

• the explosion on the Piper Alpha oil and gas platform in the North Sea (1988)

• the Clapham Junction rail collision (1988)

• the explosion at the Phillips 66 Houston Chemical Complex in Pasadena, Texas (1989)

• the blow out of a flight deck window on a BAC1-11 over Oxfordshire (1990)

• the in-flight structural break of an Embraer 120 at Eagle Lake, Texas (1991)

• a blocked pitot tube contributing to the total loss of a B757 at Puerto Plata in the Dominican Republic (1996)

• the oxygen generator fire in the hold of a DC9 over Florida (1996).

It has also been estimated that maintenance errors ranked second only to controlled flight into terrain accidents in causing onboard aircraft fatalities between 1982 and 1991.²

Despite these tragedies, the main repercussions of maintenance errors are more likely to be felt in the bottom line than in injuries and fatalities. Maintenance errors cause large and continuing financial losses. Maintenance-induced or maintenance-prolonged outages in US nuclear power plants are priced at around a million dollars a day. At coal-fired power stations, 56 per cent of forced outages occur less than a week after a planned or maintenance shutdown.³ General Electric has estimated that each in-flight engine shutdown—for which maintenance errors are usually the primary cause—costs the airline in the region of $500 000. Boeing rates the costs of each maintenance-related flight cancellation at $50 000, and at $10–20 000 for each hour of maintenance-induced delay.⁴ Litigation costs related to the ValuJet DC9 crash in the Florida Everglades are currently in excess of one billion dollars. And so it goes on. Maintenance errors not only endanger lives and assets, they are also extremely bad for business. Yet they keep on happening in remarkably similar ways—which brings us to the good news.

The Good News

Many people regard errors as random occurrences, events that are so wayward and unpredictable as to be beyond effective control. But this is not the case. While it is true that chance factors play their part and that human fallibility will never be wholly eliminated, the large majority of slips, lapses and mistakes fall into systematic and recurrent patterns. And these patterns are especially evident in maintenance-related activities, as we shall see below.

Far from being entirely unpredictable happenings, maintenance mishaps fall mostly into well-defined clusters shaped largely by situation and task factors that are common to maintenance activities in general. That these errors are not committed by a few careless or incompetent individuals is evident from the way that different people in different kinds of maintenance organisations—often very good people in excellent organisations—keep on making the same blunders. One of the basic principles of error management is that the best people can make the worst mistakes.

In 1997 Alan Hobbs interviewed experienced aircraft maintainers about incidents they had been involved in, or had witnessed, and was told of 86 safety incidents. Human error featured in all but a few of these events. About half of the incidents had implications for worker safety and about half affected the airworthiness of aircraft. When the people who supplied these incident reports were asked to say, in each case, whether or not a similar or identical incident had happened before, they produced the responses summarized in Figure 1.1.⁵

More than half of the incidents (particularly those with bad consequences for the aircraft) were recognized as having happened before. In the majority of cases the maintainers making these judgements were confident that the same or similar errors could happen again. This is an important finding because it shows that there are certain situations and work pressures that lead people into the same kind of error regardless of who is doing the job. These ‘error traps’ clearly imply that we are dealing primarily with error-provoking tasks and error-inducing situations rather than with error-prone people.

Figure 1.1 The frequency of reoccurence of aircraft maintenance incidents (n = 86)

Source: A. Hobbs Human Factors in Airline Maintenance: A Study of Incident Reports (Canberra: Bureau of Air Safety Investigation, 1997).

So the good news boils down to this: the maintenance error problem can be managed in the same way that any well-defined business risk can be managed. And because most maintenance errors occur as recognizable and recurrent types, limited resources can be targeted to achieve maximum remedial effect. It should be stressed, however, that there is no one best way of limiting and containing human error. As discussed in Chapter 2, effective error management requires a wide variety of counter-measures directed at different levels of the system: the individual, the team, the task, the workplace and the organization as a whole. First, however, we will look at the patterns into which maintenance errors fall.

Removal Versus Replacement

Regardless of the industry or sphere of operations, many maintenance activities involve two repeated activities: (a) the removal of fastenings together with the disassembly of components, and (b) their reassembly and installation followed by the replacement of the fastenings. Anyone who has ever taken anything apart and then tried to put the bits back together again knows that the former is far easier to accomplish than the latter. There is often something left over when you think you have completed the reassembly.

It does not take a rocket scientist to work out why reassembly is more vulnerable to human error than disassembly. Consider the following example as a micro-model of maintenance in general. Imagine a bolt with eight nuts on it. The nuts are labelled A–H and they need to be removed and then replaced in a predetermined order. There is really only one way of taking the nuts off the bolt, and each step is prompted naturally by the preceding one. All the necessary knowledge is available in the task itself and does not have to be stored in memory or read from procedures. It is ‘knowledge-in-the-world’ rather than ‘knowledge-in-the-head’.

However, when it comes to reassembling the nuts in a particular order, there are over 40 000 ways of getting the sequence wrong (factorial 8 possible combinations: 8 × 7 × 6 × 5 × 4 × 3 × 2 × 1 = 40 320)—and this takes no account of any possible omissions. Moreover, all the necessary knowledge has to be either contained in memory or available on some written procedure. Reassembly imposes a much greater burden upon limited mental resources (that is, memory and attention), thus greatly increasing the probability of error.

In a real-life assembly task (as opposed to the bolt-and-nuts example), the maintainer’s job is made even more difficult by the fact that errors (omissions, improper installations, misorderings and so on) ma...