Research on intelligent fault diagnosis of equipment parts based on improved wavelet domain Perona-Malik filtering algorithm

An improved Perona-Malik filtering algorithm in wavelet domain was put forward for rotating mechanical vibration signals in multiplicative noise by integrating wavelet transform and Perona-Malik nonlinear anisotropy diffusion filtering model. At first, partial differential equation theories were introduced and Perona-Malik model, as one of important nonlinear anisotropy diffusion filtering algorithms, was brought in. From their applications in image denoising, it can be analyzed and inferred that both wavelet threshold denoising algorithm and Perona-Malik filtering model have their respective disadvantages. However there were a few interrelations and complementarities between them. Thus by integrating wavelet transform and Perona-Malik filtering model, an improved Perona-Malik filtering algorithm in wavelet domain was put forward for multiplicative signal denoising. Furthermore its algorithm principle was elaborated with flow chart. At last, by comparison with wavelet threshold denoising algorithm and Perona-Malik filtering model on bearing vibration signals, it can be indicated that this new improved Perona-Malik filtering algorithm is very appropriate for mechanical vibration signals in multiplicative noise. Above all, this new improved Perona-Malik filtering algorithm can not only realize signal denoising but also reserve signal features without signal distortions. It is a perfect synthesis of wavelet transform and Perona-Malik filtering model with better denoising performance and better anti-noise robustness.