A novel threshold segmentation instantaneous frequency calculation approach for fault diagnosis.

Instantaneous frequency can well track and reflect the transient information of signal, so it plays an important role in the analysis and processing of the non-stationary signal. In this paper, the single component signal is compared with the Second Order Differential Equation in polar coordinates. Based on this, a threshold segmentation instantaneous frequency calculation method is proposed. This method is mainly for characteristics of the non-stationary signal, use the change of the area around the signal and the x axis to determine the amplitude mutation point of each single component signal, and perform segmentation. Simulations, mathematical derivations and experimental tests are used to highlight the performance of the proposed method. It is not only simple in calculation, but also can reduce the unnecessary influence of non-stationary signal amplitude mutation on instantaneous frequency, and can effectively judge the fault of rolling bearing in fault diagnosis.

A rolling bearing fault detection method based on compressed sensing and a neural network.

The high sampling frequency of traditional Nyquist sampling theory not only puts greater requirements on the sampling equipment, but also generates a large amount of data, which increases the difficulty of information transmission and storage. To this end, this paper proposes a rolling bearing fault signal detection method based on compressed sensing combined with a neural network. Based on the theory of compressed sensing, the observations obtained from compression sampling are divided into two sets of data. Given the one set of data, the predictive ability of the nonlinear time series through the neural network can predict the second set of observed values. The predicted observations are used to reconstruct the signal, thereby reducing the amount of data to be stored and transmitted and realizing secondary compression of the signal.