Indoor 3-D Gas Leakage Monitoring And Localization Methods

At present,the leakage of poisonous and harmful gases is a serious threat to people's lives and property,so it is of great significance to carry out the research of gas leakage monitoring and localization.The current research methods have some critical defects,such as many limiting conditions,simple application environments and only for the two-dimensional(2D)plane.To solve these problems,this thesis focuses on indoor gas leakage monitoring and localization research in three-dimensional(3D)real-world environments.The main work can be concluded as follows.(1)An indoor 3D gas leakage monitoring system and an associated gas source localization strategy are proposed.The proposed monitoring system is inspired by the overhead lifts system in medical domains,which can realize the 3D movement of multiple sensor nodes,and accomplish the task of gas leakage monitoring.The leakage source localization strategy needs to partition the monitoring area into some small tetrahedron areas,and then judges each tetrahedron area to decide whether there is a leakage source or not.This strategy can convert the localization problem into a two-class classification problem.Computer simulation has been carried out to validate the proposed indoor 3D gas leakage monitoring system as well as the biggest gas concentration searching method and the tetrahedron traversal method involved in the leakage source localization strategy.(2)Aiming at the two-class classification problem involved in the source localization strategy,a mass flux criterion(MFC)based method is proposed.The detailed work includes the derivation of the mass flux criterion,the calculation method of the net mass flux on the closed tetrahedron surface as well as the tetrahedron verification experiment in the indoor artificial wind field environment.Experimental results show that the inner source of tetrahedrons can be assessed through the mass flux criterion and the average value of several hundred mass flux values.(3)A machine learning algorithm based two-class classification method is proposed to shorten the time consumed by the MFC based source localization process.The extreme learning machine(ELM)is used to combine with MFC,namely ELM-MFC,in which the mass flux criterion is regarded as a sample filter to acquire reasonable training samples for the ELM.Compared with the MFC method,the ELM-MFC algorithm improves the speed of on-line source localization at the cost of the large off-line training workload and the sacrifice of certain classification accuracy.Finally,the tetrahedron experiment in an indoor natural ventilation environment is designed,and the validity of the ELM-MFC algorithm is verified by comparing with the classification results of the single ELM algorithm.