Intelligent Fault Diagnosis And Location Method Based On Sound Information

With the continuous development of science and technology,mechanical equipment is becoming increasingly sophisticated and complex.While improving productivity and meeting people’s material needs,it also brings urgent problems to be solved: the failure rate of mechanical equipment is becoming higher and higher.At present,the fault diagnosis technology of mechanical equipment often relies on extracting the vibration signal of the equipment for analysis and diagnosis.However,for some compact equipment and high-speed rotating equipment,the installation of sensors is inconvenient,and it is difficult to obtain the vibration signal under some special working conditions and harsh environment,which limits the contact fault diagnosis method based on vibration signal.Therefore,the research on acoustic diagnosis method based on non-contact measurement is imperative.The noise of mechanical equipment is the continuous form of vibration signal propagating to the outside world through the medium,which can characterize its health state as the vibration signal.Acoustic diagnosis technology has the advantages of on-line monitoring,non-contact sampling and fast and convenient operation,which can make up for the shortcomings of contact measurement methods.The specific location of the fault can be located by using sound information combined with sound source location technology,so as to repair the equipment.Based on the above ideas,this paper proposes an intelligent fault diagnosis and location algorithm based on sound signal.The specific contents are as follows:(1)Aiming at the defects of poor real-time performance and low positioning accuracy of traditional sound source location algorithms,a new beamforming deconvolution algorithm-retrospective fast iterative threshold shrinkage algorithm(RFISTA)was proposed.After the principle derivation of the algorithm,based on the indoor sound source location scene,a circular microphone array is built,combined with the dual Bluetooth sound experimental system,and the experimental simulation is carried out,which is compared with the more mature deconvolution approach for the mapping of acoustic sources(DAMAS)and Fourier non negative least squares(FFTNNLS).Experimental results show that this algorithm can greatly improve the realtime imaging and positioning accuracy compared with the traditional deconvolution algorithm.(2)In view of the fact that the traditional fault diagnosis methods rely on a large number of signal processing technology and rich expert diagnosis experience,combined with the deep learning theory,this paper proposes an intelligent fault diagnosis algorithm of "sound signal + deep learning".Taking advantage of the advantages of array signal processing with multi-channel information fusion,the endto-end convolution neural network fault state identification method is adopted to establish the layer by layer nonlinear mapping between multi-layer convolution neural networks,The feature expression is sampled layer by layer to complete the adaptive extraction of fault features,and the identification of equipment fault types under multiple working conditions and large samples is realized.In this paper,a rotor test platform based on digital microphone array is built.The experimental results show that the proposed method not only has better performance than the traditional fault diagnosis method based on vibration signal,but also can overcome the difficulty of obtaining vibration signal during contact measurement and the lack of engineering practical experience of traditional signal processing technology,At the same time,it provides a new solution for fault diagnosis technology.Aiming at the fact that the fault diagnosis technology based on vibration signal can not give consideration to location and diagnosis,and the contact measurement method is easily limited by the environment and working conditions,this paper proposes an intelligent fault diagnosis and location model based on sound signal,introduces deconvolution imaging on the basis of traditional beamforming algorithm,and uses deep learning to train and classify the sound signal while determining the location of noise source,so as to judge the fault type,It takes into account the characteristics of noise source identification and location and fault detection,and further promotes the cross development of fault diagnosis technology in many fields.