Cases Involving Human Error

The following are some cases (Oil Insurance Association, 1971) in which human error is involved as one of several errors in a chain. 

This is the tniditional view of human error. There were several retisons w hy ihe mainlenancc pnx o(lurvs were ni i lollowcd  

1) The highly technical language used by the pump manufacturer, 

2) Obscure format made it difficult to find the information, 

E Hard cover binding unsuitable for plant conditions, 

---

Equipment Cracking Failure Case Studies Equipment fails either alone or in combination with other factors, including substandard materials, improper material selection, poor design, equipment abuse, unexpected stresses or environmental conditions, and poor maintenance practices and/or neglect. Many failures, in one way or another, involve human error to some extent. 

On the basis of Nagel s (1988) information processing stages, 71 per cent of the 284 accidents studied were considered to involve human error, whereby 22 per cent were eoded as information errors, 35 per cent as decision errors, and 43 per cent as action errors (O Hare et al., 1994). However, a different picture emerged when the mishaps were divided according to accident severity. Of the 34 accidents in which there was a fatality or serious injury, 62.5 per cent were attributable to decision errors, while only 25 per eent involved the action stage (O Hare et al., 1994). Of the 169 cases involving minor/non-injury occurrences, only 30.5 per cent were the result of decision errors, while 45.6 per eent were attributed to action errors. These results closely mirror those reported by Jensen and Benel (1977) in their examination of NTSB reeords between 1970 and 1974 in whieh the majority of non-fatal incidents appeared to involve perceptual-motor factors, whereas most fatal accidents were associated with decision-related factors (Jensen and Benel, 1977). 

The other main application area for predictive error analysis is in chemical process quantitative risk assessment (CPQRA) as a means of identifying human errors with significant risk consequences. In most cases, the generation of error modes in CPQRA is a somewhat unsystematic process, since it only considers errors that involve the failure to perform some pre-specified function, usually in an emergency (e.g., responding to an alarm within a time interval). The fact that errors of commission can arise as a result of diagnostic failures, or that poor interface design or procedures can also induce errors is rarely considered as part of CPQRA. However, this may be due to the fact that HEA techniques are not widely known in the chemical industry. The application of error analysis in CPQRA will be discussed further in Chapter 5. 

Until a few years ago, it was widely believed that if a worker was involved in an accident while performing their duties, it was probably the worker s fault. In a review of 75,000 accident cases, an estimated 88 percent of the accidents were due to unsafe acts of people. Similarly, airplane accidents were generally assumed due to human error— the pilot. U.S. Air Force ballistic missiles introduced a new concept of the cause of accidents. When the missiles were launched and then failed, there was no human on board to blame. Thus, the cause of the accident was due to something else, such as design or manufacturing error. This led to the new concept that accidents could be caused by an error due to ... 

The error classification of the test and maintenance human errors involved in unplanned reactor trips showed that planning failure contributed most high to the human-related tmplanned reactor trip events (47%), and contribution of execution failure was also considerably high (33%). In the case of the problem from a work method or work object, the contribution was low against the overall events (19%) but it shows a distinctive feature that the events are strongly related to maintenance categories. Contribution of a rule violation was low as a whole. The results are summarized in Table 1. 

The case studies have illustrated how fundamental assumptions can he proved wrong and how multiple-level redundant systems can fail, especially when applied to human behaviour. Application software is seen to be particularly vulnerable in this regard, as was shown at Milton Keynes. Most if not all of the techniques and measures aimed at systematic failure avoidance involve human behaviour, where error rates ( failures ) are known to be much greater than those for physical devices. In the author s opinion this is where the effort should be directed. 

The model shows that the hazard may be introduced in two w s. It may occur through an error on the part of the subsequent victim, for example, the inadvertent ingestion of a toxin. Alternatively, it may be introduced by an error on the part of another person, remote in time or space - for example, when a design fault causes a weakness in a structnre that remains as an unsuspected hazard for a substantial period before eventual collapse. In both cases the causal sequence remains unchanged. Wigglesworth listed typieal human errors and their causes, involving ... 

Fat contaminated the polyurethane floor at a margarine production factory causing it to break up. It is probable that the floor material manufacturer was not aware of the circumstances of use so that to him/her it would be unexpected service conditions, whereas considering the people who specified the floor it would be a case of incorrect choice of material. Yet again, it could be that the factory owner did not disclose the actual use of the floor area in question. Because the investigators were only required to prove damage from fat the actual human errors involved are not known. 

Thomas McKaig s 1962 book Building Failures is a widely known collection of case studies intended for the use of engineers, architects, and contractors. Although he clearly places the blame on human error, McKaig believes that those involved in building failures are thonselves the victims of accidents, and... 

At the time of the study addressed in the present paper, [Pitblado 88] a total of 17 incidents had been reported in the Netherlands over a 4 year period on computer controlled plants, which related to the computer system or to the human interaction with the computer. These failures led mainly to small and medium scale releases from the flare systems, but in one case led to plant damage and in another to a fireball. Table 1 shows a summary checklist indicating the number of failures in each category (hardware, software, human), from which it can be seen that human errors during operations were associated with 59% of the incidents. Errors were mainly due to inadequate, insufficient or incorrect information supplied to the operators (59%) and a failure to correctly follow procedures (47%). Human errors in design were involved in 29% of incidents. Hardware and software failures were less prominent. 

It is easy to assume that human errors are one amorphous mass but in reality this is not the case as there are distinct types of human error. By knowing the type of error involved in a given situation, it is easier to identify its cause(s) and also the best approaches to remove the error potential or to mitigate the effects of the consequence(s). 

However, systems are usually not made up just of computers -rather they will also involve other entities (e.g., devices and humans) which in many cases will not be able to simply forget some of their recent activity, and so simply go straight back to an exact earlier state when told that an error has been detected. Thus forward error recovery (the typical programming mechanism for which is exception handling), rather than backward recovery will have to be used. Each of these complications individually makes the task of error recovery more difficult, and together they make it much more challenging. This in fact is the topic that I and my colleagues have concentrated on these last few years. 

---