Abstract
Rapid source locating of high-risk gas leakage is the key to identification and early pre-control of risks in chemical industrial parks. In this work, a multi-scale dispersion model is established by coupling the traditional atmospheric diffusion model with the diffusion parameters accounting for the turbulent vortex structure, to improve the accuracy of gas diffusion prediction. The locating of the leakage source is performed based on fixed-points monitoring of the gas concentration produced by simulations with multi-scale dispersion model. The relationship between the monitored dataset and the source location is then established through artificial neural network. Finally, a rapid locating method for gas leakage is built. The results show that the simulation accuracy of the diffusion area predicted by the current multi-scale dispersion model with considering the influence of turbulent vortex structure can be significantly improved in comparison with the traditional method. And the leakage source location can be accurately located within 3 s by using the rapid locating method. This source locating method can provide support for the identification, early pre-control of risks, emergency response and rescue of accidents in chemical industrial parks, thereby enhancing security resilience.
