In today’s safety-critical systems such as process and manufacturing plants, operators are often moved to a control room far away from the physical environment, and increasing amounts of information are passed to them via automated systems, they therefore need a greater level of support to control and maintain the facilities in a safe condition. This is especially important when operators confront abnormal situations in which the information flow is quite high and poor decisions may lead to serious consequences. Therefore, they need to be supported from a cognitive perspective to reduce their workload, stress, and consequent error rate. Of the various cognitive activities, a correct understanding of the situation, that is situation awareness (SA), has been found to be a crucial factor in improving performance and reducing error. However, existing system safety researches focus mainly on technical issues and often neglect SA.
This research reviews the role of SA in accidents of safety-critical environments and introduces a clear definition for abnormal situations based on risk indicators. It then relies on mental models that embody stored long-term knowledge about the systems, and develops an abnormal situations modelling (ASM) method, that exploits the specific capabilities of Bayesian networks (BNs). In this sense, it is assumed that the operator’s mental model can be modelled using BNs as a representation of static cause–effect relationships between objects in the situation. Following this, the research presents an innovative cognition-driven decision support system called the situation awareness support system (SASS) to manage abnormal situations in safety-critical environments in which the effect of situational complexity on human decision-makers is a concern. The SASS consists of five major components: (1) a knowledge–base that contains the abnormal situation models of the intended environment developed by the ASM method, (2) a situation data collection component that provides the current state of the observable variables based on online conditions and monitoring systems, (3) a situation assessment component that uses risk indicators and a fuzzy logic system to generate the assessment result, (4) a situation recovery component that provides a basis for decision-making to reduce the risk level of situations to an acceptable level, and (5) a human-computer interface. The performance of the SASS is demonstrated by three cases investigated by the US Chemical Safety Board in which poor operators’ SA has created industrial disasters in recent US history. The results of performance demonstrate that the SASS provides a useful graphical, mathematically consistent system for dealing with incomplete and uncertain information to help operators maintain the risk of dynamic situations at an acceptable level.
The SASS is partially evaluated by a sensitivity analysis, which is carried out to validate the BN-based situation models, and a multi-perspective evaluation approach is proposed based on SA measures to determine the degree to which the SASS improves not degrades the operator’s SA. The approach consists of three SA metrics: the Situation Awareness Global Assessment Technique, the Situation Awareness Rating Technique, and the NASA Task Load Index. The first two metrics are used for direct objective and subjective measurement of SA, while the third is used to estimate the workload of operators. The approach is applied in a safety-critical environment, and ten operators participate in two 40-minute simulation trials using a virtual plant user interface, both with and without the support of the SASS. The results indicate that the SASS improves operators’ SA, and specifically has benefits for SA levels 2 and 3. No significant correlations between the participants’ SA scores have been found. In addition, it is concluded that the SASS reduces the workload of operators, although further investigations in different environments with a larger number of participants have been suggested.