Optical Sensor and Neural Networks for Real-Time Monitoring and Estimation of Hazardous Gas Release Rate

Abstract

A real-time monitoring and estimation method is proposed for managing facilities that store hazardous materials. It relies on optical sensor networks and neural networks trained for Gaussian dispersion model for the detection and analysis of hazardous (gas) releases. This method generates estimated values of the release rate, which provide the basis for the corresponding planning and response actions. While previous monitoring methodologies take less accurate assumed values for release rates, resulting in overestimation of hazards, the proposed method provides calculated values that are within acceptable differences from those estimated after intensive, time-consuming trials with commercial software like PHAST or TRACE. The proposed method does the efficient calculation of the extent of damage, in a shorter period of time, because it directly estimates the release rate with the given inputs, while the commercially available software takes a trial-and-error approach because of the absence of information on the leak size, which is difficult to get in real cases of accidents. It also produces relatively accurate estimation results even for releases of untrained materials. The results indicate that the proposed method can improve the accuracy and availability of information that is crucial to the success of emergency information systems.