Study on VDT fatigue detection for complex industrial control room operators based on multi-modal fusion.

Abstract

In complex industrial control rooms such as nuclear power plants and coal mines, Visual Display Terminal (VDT) fatigue caused by sustained monitoring is a key factor contributing to human error. Therefore, VDT fatigue detection technology is of critical importance. This study evaluated fatigue states by analysing subjective and objective physiological data collected from 16 operators before and after shifts, including electroencephalogram and eye-tracking signals. Paired-sample statistical tests confirmed significant changes in fatigue-related indicators. Based on small-sample data, a TabNet-based model was employed to achieve multimodal fusion, yielding an Accuracy of 0.881 and F1-score of 0.919. Compared with the traditional comparison models, the TabNet-based model showed comparatively favourable classification performance in the present small-sample field dataset. The findings provide a feasible algorithmic pathway for intelligent monitoring of operator fatigue in complex industrial control rooms and offer methodological reference for fatigue management and safety protection in high-risk monitoring environments.