Attempt to predict human error probability in different industry sectors using data from major accidents and Bayesian networks

Morais, C. and Moura, R. and Beer, M. and Patelli, E. (2018) Attempt to predict human error probability in different industry sectors using data from major accidents and Bayesian networks. In: 14th Probabilistic Safety Assessment and Management, PSAM 2018, 2018-09-16 - 2018-09-21. (https://psam14.org/proceedings/index.html)

[thumbnail of Morais-etal-PSAM2018-Attempt-to-predict-human-error-probability-in-different-industry-sectors]
Preview
 Text. Filename: Morais_etal_PSAM2018_Attempt_to_predict_human_error_probability_in_different_industry_sectors.pdf
Final Published Version
Download (901kB)| Preview

Abstract

Looking into aviation, nuclear power generation, oil & gas and chemical industries, one can notice their interaction between organisational factors, technological systems and humans – the so-called complex socio-technical systems. To prevent accidents from occurring, engineers carry out safety analyses, and to calculate the likelihood of some scenarios they have to know the failure rates. It is easy to understand that components’ failure rates are evaluated differently from human’s failure rate. This subject is called Human Reliability Analysis (HRA), and it should be analysed ideally through the cooperation between engineers, psychologists and sociologists. Bayesian network is a probabilistic methodology that allows these three professional groups to better communicate through its intuitive graphical representation of the conditional probabilities. This paper presents a Bayesian model of a dataset of major accidents from different industrial sectors, instead of using scenario simulators and expert elicitation. The steps required to construct a model are presented together with tools for the assessment of the conditional probability and the model validation. The proposed approach allows to calculate the Human Error Probabilities as outputs of the model.

ORCID iDs

Morais, C., Moura, R., Beer, M. and Patelli, E. ORCID logoORCID: https://orcid.org/0000-0002-5007-7247;
CORE (COnnecting REpositories)