An Evolutionary Rule Based Framework for Real-Time Risk Governance of Noise and Fumigant Hazards in Marine- Manufacturing Systems

Hosein  Esmaeili; Mohammad Ali  Afsharkazemi; Reza  Radfar; Nazanin  Pilevari

doi:10.18502/jhsw.v15i3.21409

Hosein Esmaeili Department of Industrial Management, Science and Research Branch, Islamic Azad University, Tehran, Iran
Mohammad Ali Afsharkazemi Department of Industrial Management, Science and Research Branch, Islamic Azad University, Tehran, Iran
Reza Radfar Department of Industrial Management, Science and Research Branch, Islamic Azad University, Tehran, Iran
Nazanin Pilevari Department of Industrial Management, Science and Research Branch, Islamic Azad University, Tehran, Iran

DOI: https://doi.org/10.18502/jhsw.v15i3.21409

Keywords: Threshold-based scenario generation, Occupational noise-induced illness, Occupational exposure limits, Human factors in accidents

Abstract

Introduction: Fumigant gases in maritime and container chains, along with occupational noise in marineand manufacturing industries, are among the most significant chronic risk factors. They are usuallyassessed separately, despite their simultaneous impact on workers’ health. The importance of this studylies in presenting an integrated approach for real-time monitoring of combined risk and aligning it withoccupational exposure limits (OELs). The aim is to develop and validate an interpretable, regulation-oriented framework for predicting combined risk.

Material and Methods: This research integrated and normalized data from the Global Burden of Disease(GBD) 2021 study including age-standardized disability rates (ASDR) and average annual percentagechange (AAPC) for 204 countries with occupational exposure limit tables for fumigants. A Sugeno-typefuzzy inference system with three inputs and four rules was designed. Weights and membership functionboundaries were optimized using the Prairie Dog Optimization algorithm, and a threshold-based scenariogeneration module was applied to produce high-risk synthetic data. Model performance was evaluatedthrough an OEL compliance test.

Results: Findings revealed that the proposed optimization reduced the loss function by 42% comparedto random search. The mean absolute error (0.028 ± 0.006) and root mean square error (0.041) wereobtained. Threshold-based scenario generation improved data coverage in high-risk regions from 0.62to 0.90 and increased the accuracy of critical condition detection from 0.71 to 0.89. The OEL complianceindex reached 0.93, confirming input weighting as the most influential factor.

Conclusion: The proposed framework simultaneously ensures numerical accuracy, interpretability, andregulatory compliance with occupational exposure limits. It can be deployed within real-time monitoringdashboards for ports and factories. Future research should integrate IoT sensors and multi-objectiveoptimization to enable dynamic updates in response to evolving regulations and operational conditions.