Natural Language Processing and Bayesian Networks for the Analysis of Process Safety Events

Adequate Process Safety Management System (PSMS) is fundamental to ensure the safety of the operation of process industry assets and prevent the occurrence of Process Safety Events (PSEs), such as unplanned or uncontrolled releases of product. In this work, we consider as source of information a repository of reports of PSEs and we aim at the identification of the factors that influence PSE occurrence and severity. A methodology based on the combination of Term Frequency Inverse Document Frequency (TFIDF) and Normalized Pointwise Mutual Information (NPMI) is developed for the automatic extraction of keywords from PSE reports, and a Bayesian Network (BN) is developed for modeling PSEs consequences.