Research On Mechanical Noise Sources Identification And Feature Extraction

As the result of vibration emission in air, machine sound signal carries affluent information about the working condition of machine and it can be used to make mechanical fault diagnosis. The fundamental problems for sound diagnosis are to estimate the number of mechanical noise sources and localize them. Sound source identification is a technique that applies the partial information of the acoustic holography to reconstruct the acoustic parameters including sound pressure, velocity and sound intensity at every point in 3-D sound field. In this paper, the research history and applicaiton of sound source identification are studied detailedly in overseas and home. After the investigations and research about implement procedures, characters and existing problems of three sound source identificaiton technologies: Near-field Acoustic Holography (NAH), Helmholtz Equation Least Square (HELS) and Superposition Wave, a combined wave superposition method is developed to overcome time consuming and high cost for sound source identification. It allows for reconstruction of acoustic field radiated from an arbitrary object with few relatively few measurements, and the efficiency of reconstruction can be significantly enhanced. The first step in the combined wave superposition method is to establish the Helmholzt equation least squares formulation based on a finite number of acoustic pressure measurements taken on or beyond a hypothetical spherical surface that enclose the object under consideration. Next enough field acoustic pressures are generated using the Helmholtz equation least squares formulations and taken as the input to calculate the source strength. The acoustic pressures, velocities and sound intensities at the discretized nodes on the reconstructed surface are determined by solving the matrix equation based on the wave superposition. In order to avoid spatial Fourier transform-related truncation error and windowing effects based on NAH, the statistically optimal NAH (SONAH) method is introduced which performs the plane-to-plane calculations directly in the spatial domain. The SONAH algorithm is described and some numerical simulations are presented. In addition, a new technology based on SONAH method is also develop for the seperation of machinery's acoustic signal.After the estimation of sound source numbers and positions, each source's spectum is obtained from the mixed signal. The main content of this paper can be summarized as follows:1. Firstly, the research significance of sound source identification is discussed. The research history is reviewed and the research application of is also investigated. The advantage and disadvantage of every sound source identification methods are analyzed and compared detaily,and the concrete research points are decided, then the research content of this paper are defined.2. The basic theory of sound radiation from vibrating structure is introduced, and the near-filed acoustical holography (NAH) algorithm is deduced. Some filters are also introduced and discussed for sound reconstruction accuracy. Two pulsing spheres with the same phase are investigated to reconstruct sound field on the reconstruction plane. The results show that the use of spatial FFT and multiplication with a transfer function in the spatial frequency domain is computationally very efficient, but it causes"wrap-round errors"and windowing effects in the calculations.3. A new sound field reconstruction method of combined wave superposition is proposed for the first time. It not only can avoid singularity present in the integral equations and the non-uniqueness of the solution at critical wave numbers,but also reconstruct of acoustic field radiated from an arbitrary object with few relatively few measurements. It is disscussed about the non-singularity solution problems by using mono-layer potential form for Dirichlet inner region problem and using double potential form for Neumann inner region problem at the Eigen-frequency.By applying pulsing spherical source as the example, the influences of number of virtual source, virtual source location and distribution shape of virtual source on the accurateness of reconstruction result are analyzed, and the instability of the combined wave superposition is investigated.4. In order to avoid spatial Fourier transform-related truncation effects, the measurement aperture (i.e., the hologram surface) must typically extend well beyond the sources, a statistically optimal NAH (SONAH) method is introduced which performs the plane-to-plane calculations directly in the spatial domain. In oder to avoid the estimation errors of the sound sources number and sources positon, a new feature extraction method based on SONAH is developed to get the separated sources from the mixed signal. Firstly, the sound preasure field is reconstructed based on SONAH and the number and position of sound sources are estimted efficiently.Then, the tranfer matrix between microphones and sound sources is established. Finally, by inverse solution, the waveform in time domain and the exatrcted featureg in the frequency domain are obtained, which can be used as a machine diagnostic tool.5. In a semi-anechoic chamber, the sound sources are set up as one high fidelity loudspeaker, two loudspeakers and the motor. The location, the sound pressure and the properties in frequency domain of the sound sources can be found through this method precisely. The experimental results demonstrate that the SONAH is very effective in the low-to-mid regime, and can potentially become a powerful noise diagnostic tool.