Due to the advantages of small volume, large transmission ratio, strong bearing capacity and high efficiency, planetary gearbox is widely used in helicopter, wind power generation, ship, heavy truck, mining excavator and other large mechanical equipment. The working environment of planetary gear box is usually very bad. It runs under heavy load, fatigue and strong impact for a long time. It is easy to cause pitting corrosion, crack and tooth breakage of key components such as sun gear, planetary gear and ring gear. Once the planetary gearbox breaks down, the equipment and the whole power transmission system will be destroyed, and the consequences will be very serious. Therefore, it is of great significance to carry out fault diagnosis of planetary gearbox to ensure the stable and safe operation of equipment and avoid personnel and economic losses.

When the gear is damaged, periodic and non-stationary impulse vibration will be generated in the damaged part during meshing. When the vibration is transmitted to the outside, it will be affected by the complex transmission path and will gradually weaken. At the same time, the meshing vibration of many pairs of gears in the planetary gearbox is coupled with each other, resulting in the non-stationary and non-linear vibration response signal picked up from the shell. Time frequency analysis method is a powerful tool to analyze non-stationary and nonlinear signals, such as Wigner Ville distribution, short time Fourier transform (STFT), wavelet transform. However, these methods have their own limitations. Wigner Ville distribution has serious cross interference when it is used in multi-component signals; STFT has the defect of fixed time-frequency resolution; wavelet transform needs to determine the wavelet basis function and decomposition level in advance, which makes it lack of adaptability. The non adaptive signal processing method is difficult to achieve satisfactory results for the analysis of the actual signal.

Empirical mode decomposition (EMD) method adaptively decomposes the signal into the sum of a series of intrinsic mode functions according to its local scale characteristics, so as to reveal the internal nature of the signal. However, there are over envelope, under envelope, end effect and mode aliasing in EMD, which affect the correctness of the analysis results. In order to explore a new and more suitable time-frequency analysis method, Simith et al. Proposed the local mean decomposition (LMD) method, which overcomes the envelope and underenvelope problems in EMD, and has the advantages of unobvious endpoint effect and less iterations. However, like EMD method, LMD still has serious mode aliasing. In order to suppress mode aliasing, Chen et al. Proposed an ensemble local mean decomposition (elmd) method based on ensemble empirical mode decomposition (EEMD) In this method, the signal components of different scales are automatically projected into the uniform reference frame established by white noise, so as to solve the problem of mode aliasing.

Since elmd method was put forward, it has attracted extensive attention in the fields of fault diagnosis and signal processing. However, the selection mechanism of the two key parameters (amplitude and integration times) of white noise in elmd is not clear. At the same time, there are problems of residual noise pollution and large amount of computation in the process of signal reconstruction. In essence, the purpose of adding white noise to elmd is to make the distribution of the extreme points of the original signal uniform and eliminate the mode aliasing. Ideally, the smaller the amplitude of white noise, the better. But if the amplitude of white noise is too small, it will not improve the distribution of extreme points. Therefore, the amplitude of adding white noise should not be too small. In this case, it is necessary to increase the number of integration to eliminate the influence of residual noise, which will eventually lead to a larger amount of calculation. For vibration signal or noise, the extremum density (the average number of extremum points per unit length) represents the frequency of the signal. The higher the frequency of the signal, the greater the extremum density and the denser the extremum points, and vice versa. This shows that by increasing the frequency of the signal, the noise with smaller amplitude can cause enough changes of the extremum, so that the distribution of the extremum is more uniform.

Based on this, an adaptive parameter optimized ensem ble local mean decomposition (apoelmd) method is proposed. Simulation analysis and data processing of planetary gearbox fault experiment show that the method can effectively extract the fault feature information of planetary gearbox, and has strong practicability.

A fast adaptive global local mean decomposition method for noise frequency optimization is proposed. By adding pairs of positive and negative white noise, the amplitude and integration times of white noise are fixed to 0.01sd and 2 respectively, and the relative root mean square error is used to adaptively determine the optimal upper limit frequency of white noise, so as to optimize elmd and solve the problem of elmd It is difficult to select the noise parameters, the residual noise pollution in the process of signal reconstruction and the large amount of computation. The simulation results show that the apoelmd method is better than the original elmd and has strong practicability. The apoelmd method is applied to analyze the vibration signal of planetary gearbox, which can effectively decompose the complex multi-component vibration signal of planetary gearbox into multiple amplitude modulation signals containing fault information Frequency modulation single component, select the carrier frequency as the meshing frequency, double the frequency of the single component signal respectively for amplitude demodulation and frequency demodulation analysis, successfully extract the fault characteristic frequency, realize the accurate diagnosis of the local damage fault of the sun gear and planetary gear, has a certain practical engineering application value.