Earch On The Methodology Of Machinery Fault Diagnosis Based On Time-Frequency Manifold Analysis

Rotating machinery is the key equipment in modern metallurgy, chemical engineering, power engineering and other industrial fields, so it is necessary for its operation status monitoring and fault diagnosis. Vibration data from a mechanical system often associate with important measurement information for machinery fault diagnosis. For rotating machines, the localized faults of key components generally represent as periodic transient impulses in vibration signals. However, the existence of background noise will corrupt transient impulses in practice, and will thus increase the difficulty to identify specific faults.Manifold learning is a nonlinear data dimension reduction method developed in recent years. It can extract the main nonlinear inherent characteristics of high-dimensional data components. This paper proposes to obtain the time-frequency manifold (TFM) by addressing manifold learning on the time-frequency distribution (TFD), which has the merits of noise suppression and resolution enhancement. In view of these merits, the TFM is applied to machinery fault diagnosis, and the effect is satisfactory. In this paper, the researches of TFM-based machinery fault diagnosis are focused on two aspects including TFM correlation matching for periodic fault identification and vibration signal denoising using TFM.TFM correlation matching for periodic fault identification combines the concept of time-frequency manifold (TFM) and image template matching, and proposes a novel TFM correlation matching method to enhance identification of the periodic faults. This method is to conduct correlation matching of a vibration signal in the time-frequency domain by using the TFM with a short duration as a template.The proposed vibration signal denoising method using TFM intends to realize data denoising by synthesizing the TFM using time-frequency synthesis and phase space reconstruction (PSR) synthesis. The denoised signal would have satisfactory denoising effect, as well as good effect of inherent time-frequency structure keeping. Moreover, this paper presents a new diagnostic approach, called frequency probability time series (FPTS) spectral analysis, to show its effectiveness in fault diagnosis.The proposed TFM-based machinery fault diagnosis method has been employed to deal with a set of bearing defect data and gearbox fault data, and the results verify that the method is rather superior to traditional methods for machinery fault diagnosis.