A new hazard event classification model via deep learning and multifractal

Hazard and operability analysis (HAZOP) is the paradigm of industrial safety that can reveal hazards process from its node deviations, consequences, causes, measures suggestions, such be considered as hazard events (HaE). The classification research on HaE has much irreplaceable pragmatic values. In this paper, we present a novel deep learning model termed DLF through multifractal to explore where motivation naturally regarded kind time series. Specifically, first vectorized get series by employing BERT. Then, new method HmF-DFA proposed win fractal analyzing Finally, hierarchical gating neural network (HGNN) designed accomplish three aspects: severity, possibility risk. We take HAZOP reports 18 processes cases, launch experiments basis. Results demonstrate compared with other classifiers, classifier performs better under metrics precision, recall F1-score, especially for severity aspect. Also, HGNN effectively promote classification. Our system serve application incentives experts, engineers, employees, enterprises. hope our contribute added support daily practice in safety.