A Knowledge-Based Artificial Intelligence Approach to Risk Management

Abstract

This paper details experiences gained while developing a novel technology-driven approach to Risk Assessment methodologies, e.g., Process Hazard Analysis (PHA), Hazard Identification (HAZID) and Hazard Operability (HAZOP), in oil & gas. Emphasis has been placed on combining encoded human knowledge with Artificial Intelligence techniques, in a way which fosters safer designs and operations, while maintaining Subject Matter Experts (SMEs) at the center of decision making. Encoding of human knowledge (e.g., Subject Matter Expertise, Industry best practices) in digital applications has traditionally been associated with creating static pieces of information, such as lessons learned documentation and validation activities for hazard analysis. New digital technologies, however, make it possible to create truly dynamic knowledge representations, which capture key concepts and their relationships, creating a new type of "source of truth." As a result, corporate and external knowledge can be made more readily accessible to engineers and operations personnel participating in decision making. Digital corporate knowledge can also be supplemented with Artificial Intelligence (AI) techniques which can help uncover latent threats and better guide optimal decision making. This is particularly relevant in Workforce, Health & Safety (WH&S) and Process Safety contexts, where the impact of flawed or suboptimal decisions can lead to catastrophic consequences. Practical examples from an oil & gas major show how the risk assessment domain can be represented in a computational knowledge graph, in a format which is comprehensible not only to software developers, but more importantly, to oil & gas SMEs. A presentation of different AI techniques overlaid on top of this computational knowledge graph, can also offer a glimpse of the possibilities of marrying SME expertise with emerging digital technologies.