Using Bayesian belief networks to improve distributed situation awareness in shift changeovers: A case study

Abstract

This paper presents a case study about communication between workers during shift changes based on empirical research at a continuously operating industrial operation.The premises of Situation Awareness (SA) and Distributed Situation Awareness (DSA) are the theoretical lens for conducting the present study. The objective of the study is the probabilistic modeling of the SA of process operators by means of complementarity among the premises of SA and DSA during shift change.The methodology used semi-structured interviews that followed a questionnaire created to categorize verbalizations at the three SA levels. Verbalizations were analyzed with the “digital humanities” approach that allowed the integrated modeling of SA using Bayesian Belief Networks (BBN) theory. The research was performed in an industrial plant that carried out a polymer injection process, which was classified as a physical arrangement for the continuous manufacture of products. In this process, production was conducted in three continuous shifts with one operator per shift.As a result, the BBN modeling showed the dynamic relationships among the core variables monitored, and also demonstrated the complementarity between SA and DSA. In this sense, BBN was used as a communication protocol between shifts and served as a mechanism that influences the operator’s cognitive load. The developed algorithm showed an excellent performance, because it represents the conditional probability between the variables according to the operators' indications regarding the criticality and level of monitoring of the process. The modeling allowed representing the operators’ complete SA, thus promoting DSA.