The Selection Of The Characteristic Parameters In The Car The Main Gear Assembly Quality Fault Diagnosis Method Of Research

The driving axle locates at the end of automobile transmission system, which is used to enhance the torsion, decrease the speed of running and change way. The quality of axle is closely related to the gears in the retarder. The driving axle is one of main sources to the truck in the vibration and noise. The impact coming from the gear transmission system would excite the vibration of the rear axle to emit noise. A method to evaluate the final drive with mixed eigenvector extracted by the statistical characteristics and wavelet theory is presented, and then pattern recognition is applied to identify quality of the final drive assembly.Based on introduction about the structure and functions of the final drive, the vibration model of the gear in the final drive is built, the mechanism, and the origin of the noise and the relation between noise and vibration is analyzed. Then the fault types and causes of the final drive are discussed. The paper is the core of the composite measurement machine and confirms series of statistical characteristics in time and frequency domain and wavelet analysis according to the character of the vibration from the final drive. Several characteristics are calculated based on the date from the composite measurement machine, such as Kurtosis Value, Shape Factor, and Crest Factor. These statistical parameters lie on probability density function, and reflect directly the condition of the final drive without the effect of the equipment. The vibration signal is processed with the wavelet theory to decompose into different frequency bands to identify continuous noise. And the energy distribution at every frequency band is distinguished in energy spectrum as part of the fault eigenvector. The quality interpolator of the final drive condition is designed with neural network. The mix eigenvector made up of the statistical characteristics of time and frequency domain and the energy spectrum in frequency bands is the input of neural network identification. The type and size of the neural network based on the known inputs and outputs is designed, and this neural network is trained with the samples. The final drive assembly is diagnosed with the trained network.Root Mean Square, Clearance Factor, Kurtosis Value, Six-order Moment and wavelet function db6 are the distinct element of the vibration signals related to the typical states. The interpolator on BP neural network identifies quality of the final drive based on the eigenvector. The method has been proved of validity with the experiment.