Session: 04-11-01 Posters

Paper Number: 195010

195010 - Reliability-Informed Machine Learning Approach for Predicting Cable and Cabinet Damage in Switchgear Room Fires 

Abstract: 

A fire in the Motor Control Center (MCC) of a switchgear room in a nuclear power plant can significantly threaten safety and operational reliability by damaging critical equipment and disrupting key system operations. This study investigated the effects of fire source locations, cable and cabinet locations, air changes per hour (ACH), and ambient temperatures on cable and cabinet damage time in a switchgear room using Consolidated Model of Fire and Smoke Transport (CFAST) simulations. The findings show that cables positioned close to the ceiling experience damage more quickly, whereas fires originating at lower heights cause cable damage to occur earlier. Furthermore, higher ACH accelerates damage to distant cables. Machine learning–based surrogate models were developed to predict damage times using CFAST simulation datasets. The Shapley additive explanations (SHAP) method was applied to interpret relationships between input features and cable damage. Moreover, uncertainty quantification was performed to assess the prediction reliability. The results indicate that the Light Gradient-Boosting Machine (LGBM) model provides accurate predictions of damage time, and SHAP analysis aligns well with the sensitivity analysis, supporting reliable interpretation of the model results. Uncertainty quantification reveals a strong correlation between predicted uncertainty and errors, providing a reliable measure of prediction confidence. The framework provides a rapid, accurate, and risk-aware tool for damage prediction, supporting practical fire safety measures and emergency preparedness.

Acknowledgment

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2022-00144190).

 

Presenting Author: Van Tuan Ho Chungnam National University

Presenting Author Biography: 

Authors:

Weon Gyu Shin Chungnam National University
Van Tuan Ho Chungnam National University