2025 Spring Meeting and 21st Global Congress on Process Safety

(32j) Comparison of Technology-Driven Approaches for Pre-Causing in Process Hazard Analysis (PHA): Pre-Causing Parser Vs. NLP and Chatbot Integration

Authors

Kaylee Miranda - Presenter, Saltegra Consulting LLC
Michael Saura, Risk Management Professionals
Neo Piolo Manansala, Saltegra Consulting LLC
James David Abiog, Saltegra Consulting LLC
Ann Pearl Triana, Saltegra Consulting LLC
Process Hazard Analysis (PHA) plays a critical role in identifying and assessing risks in industrial settings. However, the preparation phase of PHA, particularly the pre-causing step, where probable causes for potential hazards are identified, can be labor-intensive and requires specialized expertise. An engineer with in-depth knowledge of operational hazards performs this step, thoroughly assessing causes that could trigger top events or lead to severe consequences. This study examines two distinct technology-driven approaches designed to streamline the pre-causing step: the Pre-Causing Parser and an NLP-based system with chatbot integration. Both methods aim to reduce preparation time, lower project costs, and enhance efficiency without compromising the accuracy or thoroughness of hazard identification.

The Pre-Causing Parser is a streamlined automation tool that generates probable causes by analyzing structured data of the nodes, equipment, and system information. Engineers provide input files to the parser, which produces an extensive list of potential causes. Engineers then refine this output, enabling faster and more comprehensive preparation than traditional manual processes.

The NLP and Chatbot Automation approach leverages Natural Language Processing (NLP) and an interactive chatbot to support engineers during pre-causing. Piping and Instrumentation Diagrams (P&IDs) are converted into structured formats, allowing NLP models to efficiently identify potential causes and suggest them interactively. The chatbot further assists by answering queries, highlighting relevant data, and helping refine scenarios in real time. This approach improves consistency in cause identification and reduces manual input.

The study finds that the Pre-Causing Parser is effective for rapid, high-level cause generation, making it suitable for projects requiring quick preparation. Conversely, the NLP and chatbot approach provides advanced predictive insights and interactive support, which is beneficial for complex analyses with high data density. This comparison underscores how these approaches can be tailored to meet specific project needs, offering insights for PHA practitioners seeking to optimize preparation processes with the right technological support.