Deep Neural Networks for Source Tracking of Chemical Leaks and Improved Chemical Process Safety

Abstract Leak accidents of chemical plant belong to a major industrial accident that can cause huge damage to human and facilities. Accurate detection and prompt response to leaks, earlier in the event of accidents, is crucial to alleviate damage. This research presents an approach to predict the location of leak source in real-time, by using the measured data from a sensor network as an input data to a pre-trained deep neural network. Computational Fluid Dynamics (CFD) simulations for various leak scenarios are conducted for a real plant geometry, and the massive data derived from the simulations are used for training and verification of the neural networks. Proposed deep neural network models perform well as a solution to the problem of leak-source tracking, one of the challenging inverse problems in chemical process safety.