Fire risk of electrical installations: a fuzzy petri net approach applied to the National Museum of Brazil

Abstract

Over time, electrical fires have been recurring disasters around the world. As a common preventive strategy, the use of a standard checklist has become a traditional technique widely adopted today. However, this procedure might not capture the full spectrum of risk levels or address interconnected issues, thus potentially compromising effective risk management. Conversely, allowing for prioritization and immediate remediation of fire hazards, the incorporation of expert knowledge and data-driven insights to gauge risk factors is noteworthy. In this context, the approach proposed, rooted in fuzzy Petri nets, optimizes resource allocation by focusing on the highest-risk elements, leading to a more substantial risk reduction. An enhanced methodology was implemented, considering the pre-fire conditions at the Brazilian National Museum. Simulation results indicate that simple corrections in the electrical installation could have significantly reduced the fire risk, possibly preventing the tragedy. These findings underscore the method's effectiveness, emphasizing the importance of thorough electrical risk management. It suggests that the catastrophe might have been avoided had the risks been appropriately addressed. The model emerges as an essential instrument for enhancing risk assessment and strategic resource allocation, especially vital in resource-constrained environments characteristic of developing countries like Brazil. However, it is important to clarify that the proposed methodology is based on expert systems, offering an alternative approach to risk quantification when statistical data and deterministic methods are unavailable. This methodology integrates expert judgment and fuzzy logic for qualitative risk assessments, enabling the identification and prioritization of risk factors despite the lack of quantitative data. While sensitivity analysis is not applicable in this context, validation can be achieved through consensus among expert groups who evaluate the model's assumptions and outcomes.