Date of Publication :7th October 2025

Abstract: Arc flash events pose significant hazards to personnel and equipment in electrical systems, particularly in industrial and utility operations. Accurate estimation of incident energy and effective risk mitigation remain critical for compliance with established electrical safety standards. This paper presents a comprehensive evaluation of three principal arc flash protection frameworks: NFPA 70E, IEEE 1584-2018, and OSHA 1910.269. The study integrates empirical modelling, historical incident data, and machine learning techniques to assess the predictive accuracy and compliance challenges associated with each standard. Incident data from OSHA and NFPA sources (2010–2024) were analysed to identify patterns in fault current, voltage class, arc duration, and PPE usage. Incident energy was computed using IEEE 1584-2018 equations and compared with reported injury severities. The findings indicate that while IEEE 1584 predictions align with observed outcomes in most configurations, notable underestimations occur in horizontal conductor and open-air systems. NFPA 70E, although widely adopted, provides qualitative guidelines and relies on external methods such as IEEE 1584 for energy calculation. A logistic regression model trained on the incident dataset achieved 87% accuracy in predicting severe injury outcomes based on system parameters. This model was extended with a neural network architecture to support real-time classification of arc flash risk. The integration of sensor data through IoT enabled monitoring and predictive analytics enables dynamic hazard assessment and supports pre-emptive mitigation. A comparative analysis highlights the strengths and limitations of each standard. IEEE 1584-2018 offers robust empirical modelling but depends on configuration-specific inputs. NFPA 70E provides structured procedural guidance but lacks inherent computational capabilities. OSHA 1910.269 enforces general safety compliance but does not prescribe detailed modelling techniques. This study proposes a data-driven framework that enhances arc flash hazard prediction through validated equations, statistical analysis, and AI-based risk models. Recommendations for standard refinement and predictive system integration are presented to support proactive electrical safety management.

Reference :