Abstract: Fault diagnosis method based on vibration signals has limits in certain situations. However, machine noise contains abundant condition information about machines. It possesses the advantage of non-contacting test. It can be used for fault diagnosis as a substitution of vibration signals. Traditional fault diagnosis methods based on noise signals are mainly on the basis of spectrum analysis. They cannot provide the location and the strength of sound sources. So they can only be used in some simple conditions. A fault diagnosis method based on acoustic holography is presented. A microphone array composed of a few microphones is used to acquire sound field. And the wave superposition method is adopted for sound field reconstruction. The sound field around a machine can be easily visualized by using such a method. Once the exterior sound field around the machine is reconstructed, the information extracted from the holography can be used for fault diagnosis. A number of templates of normal condition and fault condition can be made from the holography. A certain condition of the machine can be confirmed by comparison of it with these templates. Then the faults of the machine can be found. Numerical simulations are performed on the basis of a multiple pulse-ball sound source model, and experiments are also done with two sound boxes in an anechoic chamber. Both the simulations and the experiments have accurately identified the changes of the sound field radiated from the radiator, and have found out the faults. It shows that the method is validated and feasible, which lays a foundation for its on-site applications.

Key words: Acoustic holography, Fault diagnosis, Feature extraction, Wave superposition