From data to decision. AI powered proactive safety of landslide Natech events

The increasing intensity and frequency of extreme weather events underscore the growing risk of natural hazards impacting industrial facilities. Natech (Natural Hazard Triggering Technological) accidents can be triggered by both major and minor environmental phenomena, such as heavy rainfall and landslides. Building on the concept that in the AI era, data can be transformed into valuable knowledge for anticipating risk conditions by feature extraction, this study presents a dynamic risk assessment framework integrating natural hazards with process safety models. The approach dynamically updates failure probabilities of safety barriers depending on landslide probability over time, offering a real-time perspective on risk evolution. A case study on an LPG (Liquefied Petroleum Gas) storage facility in a landslide-prone region in Italy illustrates the application and effectiveness of this methodology. The model leverages Long Short-Term Memory (LSTM) Recurrent Neural Networks (RNNs) to predict landslide occurrences and their cascading effects on industrial safety. Findings of this study provide new perspectives on proactive risk management based on AI applications in Natech scenarios.