Identification of compound faults of rolling bearing grounded on 1D-LBP and first-order difference of vibration signal

The fault identification and forecast based on vibration signals is an important approach to guarantee the safety of equipment. Rolling bearing is one of the most important parts of rotary machinery, and its state is crucial to the safety of equipment. When a fault exists in bearings, the vibration signals are highly nonlinear and non-stationary and the fault feature information of signals is weak, complex and compound, which adds up to the difficulty of fault identification. For correct identification of compound faults of bearings based on vibration signals, the paper has brought forward a method of blending first-order difference (FD) of vibration signal and one-dimensional local binary pattern (1D-LBP) algorithm. Impact feature accompanied with the fault of rolling bearings and the first-order difference of vibration signal presents the change rate of signal and can be used to describe this feature. Therefore, the paper has acquired the difference signals from original vibration signals and given a binarized treatment to difference signals with local mean as criterion and moreover, re-transformed the binarized sequence from each moving window into a decimal sequence. A new local texture signal (LTS) is obtained, which can embody the local texture characteristics of signals. To further reduce the noise influence, correlation analysis was combined with obtained LTS. Additionally, features of compound failures were extracted based on the frequency spectrum of autocorrelation function of LTS. Therefore, failure types of bearings were identified. The accuracy and effectiveness of proposed method has been verified through the verification analysis on various situations as well as comparative analysis with conventional methods.