Human reliability assessment during offshore emergency conditions Mashrura Musharraf a , Junaid Hassan a , Faisal Khan a,⇑ , Brian Veitch a , Scott MacKinnon b , Syed Imtiaz a a Faculty of Engineering & Applied Science, Memorial University of Newfoundland, St. John’s, Newfoundland and Labrador, Canada A1B 3X5 b School of Human Kinetics and Recreation, Memorial University of Newfoundland, St. John’s, Newfoundland and Labrador, Canada A1B 3X5 article info Article history: Received 3 October 2012 Received in revised form 14 February 2013 Accepted 4 April 2013 Available online 16 May 2013 Keywords: Human Factor Risk Assessment Offshore Emergency Human Reliability Analysis (HRA) Dependency and Uncertainty in HRA abstract This paper presents a quantitative approach to Human Reliability Analysis (HRA) during emergency con- ditions in an offshore environment. Due to the lack of human error data for emergency conditions most of the available HRA methodologies are based on expert judgment techniques. Expert judgment suffers from uncertainty and incompleteness due to partial ignorance, which is not considered in available techniques. Furthermore, traditional approaches suffer from unrealistic assumptions regarding the independence of the human factors and associated actions. The focus of this paper is to address the issue of handling uncertainty associated with expert judgments with evidence theory and to represent the dependency among the human factors and associated actions using a Bayesian Network (BN) approach. The Human Error Probability (HEP) during different phases of an emergency is then assessed using a Bayesian approach integrated with an evidence theory approach. To understand the applicability of the proposed approach, results are compared with an analytical approach: Success Likelihood Index Methodology (SLIM). The comparative study demonstrates that the proposed approach is effective in assessing human error likelihood. In addition to being simple, it possesses additional capability, such as updating as new information becomes available and representing complex interaction. Use of the proposed method would provide an effective mechanism of human reliability assessment in hazardous operations. Ó 2013 Elsevier Ltd. All rights reserved. 1. Introduction Human reliability, as defined by Swain and Guttmann (1983), is the probability that a person correctly performs system-required activities in a required time period (if time is a limiting factor). Hu- man reliability is related to the field of human factors engineering and involves the study of human Performance Shaping Factors (PSF) (Blackman et al., 2008). PSFs improve or decrease human per- formance. Recognition of the potential contributions of PSFs to accidents leads to the development of different Human Reliability Analysis (HRA) techniques. Swain and Guttmann (1983) proposed THERP (Technique for Human Error Rate Prediction) for qualitative and quantitative analysis of human reliability. Later SLIM (Success Likelihood Index Methodology) (Kirwan, 1994) was proposed to handle the lack of data with expert judgment. With the extension of the human reliability research field from human–machine sys- tems to human inherent factors (psychology, emotion and behav- ior in emergency situations) ATHEANA (A Technique for Human Error Analysis) (Cooper et al., 1996) and CREAM (Cognitive Reli- ability and Error Analysis Method) (Hollnagel, 1998) were pro- posed. Though dozens of HRA techniques are employed today, most suffer from two major limitations. First, they are unable to handle the uncertainty and inconsistency associated with expert judgments. Second, most assume unrealistic independence among human factors, and associated actions. The main focus of the paper is improving HRA method to have better human error probability assessment. The approach has the capabilities of considering the underlying uncertainty and conflict within input data, and repre- sents the dependency among different human factors and associ- ated actions. Specifically the method will be applied to assess HEP to offshore emergency situation. A better estimate of human reliability would help design more effective safety systems and emergency management systems. Due to lack of real or ecologically-valid data, the majority of works in human error prediction consider expert judgment tech- niques such as SLIM and THERP. However, expert judgment from a single expert may be biased and incomplete due to partial igno- rance. Hence, single expert opinion is not sufficient for reliable hu- man error predictions. One potential solution to this problem is to use multiple experts (multi-expert) knowledge and experience. A proper aggregation method is needed to combine this multi-expert knowledge that will minimize the uncertainty and opinion conflict. This paper proposes to use evidence theory to combine multi-ex- pert knowledge and hence increase the reliability of human error prediction. The PSFs that influence human performance depend on the con- ditions or circumstances under which an event occurs and are 0925-7535/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ssci.2013.04.001 ⇑ Corresponding author. E-mail address: fikhan@mun.ca (F. Khan). Safety Science 59 (2013) 19–27 Contents lists available at SciVerse ScienceDirect Safety Science journal homepage: www.elsevier.com/locate/ssci