Data preprocessing algorithm based on gearbox configuration

For variable speed rotating machinery such as fan, in the process of operation, various signals reflecting its operation information are repeated periodically with the mechanical operation, and the fault of components will also show periodic impact response. Therefore, the periodic domain can reflect the fault characteristics better than the time domain.

The periodic domain signal can be obtained by synchronous resampling the vibration response signal in time domain and the Keyphase pulse signal of shaft speed. The periodic domain signal contains the operation information and noise signal of the components with the same speed as the specified shaft.

The key to obtain the periodic domain signal is to obtain the synchronous pulse of shaft rotation corresponding to the vibration response of each position. The rotation speed of each gear in the gearbox is the same as that of the corresponding shaft. According to the physical parameters of the gearbox, the amplification ratio of each stage can be obtained. According to the speed Keyphasor pulse signal of the gearbox output shaft collected, the speed pulse signal of the physical shaft related to the vibration response of each position can be further calculated, which is used as the trigger signal, and the periodic domain signal can be obtained by software re sampling at the same time.

In the process of obtaining periodic domain signal by synchronous resampling, two important parameters, sampling points per cycle and number of cycles, are involved. The number of sampling points per cycle will affect the accuracy of the periodic signal itself, the number of cycles will affect the accuracy of the synchronous average analysis method used to reduce the signal-to-noise ratio, and jointly affect the frequency resolution of spectrum analysis.

In addition, there is multi component coupling vibration during the operation of wind turbine, and the background noise is generally strong, which makes the collected vibration signals show non-stationary and non-linear. However, with the increase of vibration signal transmission path, the vibration signal of other parts is weak. Moreover, the magnification ratio of common fan gearbox is about 1 ∶ 100, and the rotational speed of different parts varies greatly. Time domain data discretization is to collect the vibration signal of each position with the same sampling rate for an equal length of time. For high-speed parts, the rotation cycle is more in unit time, while for low-speed parts, the rotation cycle is less in unit time. Therefore, in order to achieve the purpose of data reduction, and considering the frequency resolution and the average number of synchronization, the vibration signal collected by each sensor is defined to be analyzed only for the adjacent components. In order to reduce the error of velocity derivation, when dividing the test range, for the vibration response of a position, when it is necessary to detect the multi-level velocity components, the axis with high velocity should be taken as the reference.