Major accidents still occur in the process industry. They tend to occur either when there is a breakdown in the safety management systems that goes undetected, allowing adverse events to occur and escalate, or when the system faces abnormal situations that the safety management system cannot handle.
Breakdowns in safety management systems, and the failure to detect such breakdowns, are often attributed to “human error”. However recent research has shown that “human error” is not a cause of accidents but is rather a symptom of the fact that complex socio-technical systems are seldom adequately designed to support the work of operators. To reduce accidents, work systems are needed that support operators’ ability to monitor, diagnose, and respond in ways that deliver safe system control, and that do so across normal, expected abnormal, unexpected, and unanticipated situations.
Most human factors methods focus on identifying how humans handle normal operating conditions and abnormal but expected operating conditions. The methods identify ways in which human behaviour might deviate from a predetermined norm and they suggest design changes that might prevent unacceptable variations in human behaviour--or that might at least protect the system from such variations. Further risk assessment methods such as Hazard and Operability (HAZOP) studies and Job Safety Analysis (JSA) that are used in process industries also identify ways that human performance might deviate from a “preferred way” in normal and expected abnormal situations. The above methods have improved safety to some extent, but more probing methods are needed to address the undetected, unexpected, and unanticipated situations that still lead to major accidents.
In this thesis a new method is suggested that may help analysts develop work systems that help operators do the following: (1) detect and diagnose current system states and anticipate future system states, (2) exercise safe control during normal and expected abnormal situations, and (3) safely control situations that the system has not been designed to handle with the degree of adaptability needed. The new method is inspired by principles from functional systems framework, cognitive work analysis, and resilience engineering, which are outlined in the thesis.
The new approach is accompanied by an analytic method dubbed “SAfER” (Strategies Analysis for Enhancing Resilience) that helps analysts identify ways to make it easier for operators to handle normal operations as well as expected, unexpected and unanticipated abnormal situations. The method has been developed in conjunction with industry practitioners and it is supported with a software application, also called SAfER. The objective was that industry practitioners should be able to use the SAfER method and software application, without expert help, to perform the above-mentioned analyses.
A key focus of the new method is to identify the range of strategies that human operators might use to complete a task within a system. The method for analysing strategies draws from the literature on decision making, cognitive work analysis, and resilience engineering. The result is a new way to perform the so-called strategies analysis phase of cognitive work analysis, as well as a new way to analyse strategies that is independent of cognitive work analysis.
This thesis describes the SAfER method and the development process used to arrive at the current version of the SAfER software application. It also shows the SAfER method at work, with a retrospective application to an industrial accident, and with a prospective analysis of a proposed shut-down job. To assess the effectiveness of the SAfER method, testing was conducted with industry practitioners and with novices who used the SAfER software application.
Results revealed that the SAfER method could be applied both retrospectively and prospectively and the method could be used by industry practitioners and by novices without expert assistance or training. However, test participants noted that more should be done to improve the usability of the SAfER software and that more training would be helpful. Test results also revealed that most participants thought that the SAfER method (1) was effective in identifying the range of human related hazards and control strategies, and (2) would be a useful in helping improve safety in industry. Specifically, they reported that the SAfER method helped them to consider possible human responses and to identify design requirements that would not have been raised with existing methods.
Overall, the research in this thesis suggests that the SAfER method provides practitioners with a more probing way to identify and analyse the range of strategies that operators might use to monitor, diagnose, and control processes. In the future, further work will be done to improve the usability of the software. It is hoped that SAfER will help practitioners create work systems that better support the operator adaptability needed to successfully control systems especially during situations that are not addressed with existing risk assessment techniques.