Peripheric sensors-based leaking source tracking in a chemical industrial park with complex obstacles

Hazardous gas leakage can cause irreversible damage to the environment and human health. When it happens, it's necessary to find the accurate position of the leaking source efficiently and take effective measures to reduce or prevent more irreversible losses. However, source tracking in the scenario with complex obstacles faces the challenge caused by turbulent wind flow. In this paper, ethane leak scenarios with different leaking sources and environmental conditions are simulated using the Flame acceleration simulator (FLACS). Considering that sensors are often deployed at the boundaries of industrial parks for the detection of hazardous gas leakage, the concentration information of these peripheric sensors is mapped to images, which serve as inputs to a convolutional neural network (CNN) to determine the location of the leaking source and wind direction in a chemical industrial park with complex obstacles. The results show the effectiveness of the proposed method. In addition, fixed failure rates of the sensor along with additional meteorological conditions are considered to evaluate the performance of generalization.

Automated summary

This paper proposes a method to determine the location of leaking sources in a chemical industrial park with complex obstacles using a convolutional neural network. The effectiveness of the method is demonstrated by simulating ethane leak scenarios with different sources and environmental conditions.