Research On The DOA Estimation Of Compressed Gas Leakage Sources Based On Compressive Sensing

Compressed gas leakage will not only cause serious economic losses and waste of energy,but also bring great harm to People's lives and property.Therefore,the detection and localization of compressed gas leakage is of great importance.At present,people often use hand-held ultrasound detection equipment to obtain the leak orientation,which greatly depends on manual operation.It is not only hard to achieve real-time and comprehensive detection,but also threaten the safety of the testing personnel.In addition,owing to the disadvantages of high sampling rate,large amount of data and heavy computational burden,the existing direction of arrival(DOA)estimation methods of compressed gas leakage based on microphone array are limited in practical application.In order to overcome the drawbacks of current methods,a method combined compressive sensing with microphone array signal processing is applied to solve the DOA estimation of compressed gas leakage source.Firstly,the threshold of the far filed and the near field is set to 10 m,which depends on the characteristics of the compressed gas leakage source.What's more,the diameter of the microphone array is limited to no more than 20 cm,which is based on the threshold formula of the far filed and the near field.Secondly,the mathematical model of the DOA estimation of compressed gas leakage source is constructed based on compressive bearing estimation(COBE)method,which is the extension of COBE from low frequency application to high frequency application.The subspace pursuit(SP)algorithm is used to solve the DOA estimation result.Simulation results show that the method can achieve the accurate DOA estimation of one or more compressed gas leakage sources through a small amount of sampling data.In addition,the influence of microphone array layout,the division density of spatial angular and the sampling frequency of reference microphone(RM)on the DOA estimation performance and the applicable range of source angle are analyzed,which is the supplement analysis of the original COBE method.Simulation results fully verify the correctness of the parameter analysis.However,it is hard for original COBE method to obtain compressed measurements from the analog signal directly,and its' measurement matrix leads to large demand of system storage and heavy computation.An improved COBE method,direct non-uniform random subsampling-compressive bearing estimation(DNRS-COBE),is proposed in this thesis to solve the problems above.The DNRS-COBE mainly provides a direct non-uniform random undersampling scheme,which can obtain the compressed measurements of the analog signal directly without changing the original sampling hardware system and just need to re-design the sampling clock.In addition,an equivalent measurement matrix corresponding to the sampling scheme is constructed,where only a few elements are one and most elements are zero.The Monte Carlo simulation and the correlation simulation are carried out by using the equivalent measurement matrix and the measurement matrix of the original COBE method.It is verified that both matrices satisfy the restricted isometry property(RIP)condition and the mutual incoherence property(MIP)condition.What's more,the demand of system storage and computation of the different measurement matrices are analyzed,the results show that the equivalent measurement matrix requires less system storage and computation.Finally,the simulations of the DOA estimation based on DNRS-COBE method are carried out.Simulation results show that the DNRS-COBE method is effective and accurate.Compared with the original COBE method,the DNRS-COBE method can achieve accurate DOA estimation by using less measurement data.