Research On Chemical Hazardous Gas Monitoring Technology Based On Multi-Source Data

In recent years,China’s industrialization process has been accelerating,and chemical production has become an important and indispensable part of industry,however,there are potential risks associated with the use of chemical gases.The leakage of hazardous chemicals will pose a serious threat to human health and environmental safety.The thesis combines deep learning theory with gas infrared imaging technology to propose an intelligent monitoring method for chemical gas leaks.At the same time,in order to accurately calculate the location of the leakage source,the thesis proposes a leakage source location method based on a swarm intelligence,which can accurately locate the leakage source in a short time and facilitate enterprises to take timely countermeasures to avoid serious consequences.Finally the thesis combines gas infrared imager,stationary gas sensor array and meteorological parameters to design and implement a chemical hazardous gas monitoring system.The main work of the thesis is:(1)A leaking gas segmentation networks(LGSNet)based on gas infrared imaging technology for leaky gas monitoring is designed.Feature fusion within the network is enhanced by fusing attention mechanisms with jump connections,and the network’s ability to extract features at different scales is enhanced by a pyramid pooling module.The network is trained using a synthetic infrared image dataset,and the performance of the network is tested on a real infrared image dataset.The experimental results show that compared with SegNet,the intersection over union of LGSNet is improved by6.17%,the accuracy rate is improved by 8.1%,and the processing speed meets the actual engineering requirements.(2)A leakage source location(LSL)algorithm is designed.To address the problem that the initialized population generated by the random function is unevenly distributed in the solution space,the population is initialized using Tent chaotic mapping,and then the leakage source information is calculated by MPA(Marine Predators Algorithm)combining the fixed gas sensor array data and meteorological parameters.The results are input to NM(Nelder Mead)algorithm for further optimization.The experimental results show that the average position error of the calculated results of LSL algorithm is reduced by 82.73% compared to MPA,which is feasible in the actual source finding problem.(3)A chemical hazardous gas monitoring system is designed and implemented.The system is based on B/S architecture.The back-end program receives the data collected by sensors and stores them in the database,then inputs them into LGSNet with LSL algorithm for processing,and the system monitors the gas leak and issues an alarm on the front-end page,while calculating the leakage source information,diffusion range and displaying it.After testing,the accuracy and response time of the system meet the actual engineering requirements.