Research on fault diagnosis of gearbox

As a device that can effectively change the speed and transfer power, gearbox is widely used in various types of mechanical equipment. Because the gearbox often operates in the harsh environment of high speed and heavy load, the probability of failure is greatly increased, and it is also easy to cause mechanical equipment failure. In all mechanical equipment failures, gearbox failure accounts for about 15%. Therefore, in order to ensure the safe operation of mechanical equipment, people pay more and more attention to the fault diagnosis of gearbox, and the research is of great significance.

Zhao Yanli et al. Applied the fuzzy correlation method to the fault diagnosis of the gearbox. By establishing the membership function and calculating the membership degree of each parameter, four kinds of faults (pitting, wear, gluing and broken teeth) of the gearbox were identified; Liu et al. Used particle swarm optimization algorithm to select signal features and used them in fault diagnosis; Du Shiliang et al. [4] constructed BP neural network with 10-5-4 structure by extracting 10 fault features to diagnose four common faults of gearbox; Wang Kai et al. Used wavelet neural network to diagnose gear fault; Chen P, et al. Used rough set theory combined with neural network for fault diagnosis of rotating machinery; Ai Li et al. Put forward a method of combining GA and BP neural network, which takes the weight and threshold as the independent variables of ant colony algorithm and finds the optimal solution through iteration, which is used for the training of BP neural network and the fault diagnosis of gearbox.

In order to further improve the accuracy of the algorithm and accelerate the convergence speed of the algorithm. The BP neural network algorithm is improved, and a ga-aco algorithm is proposed to optimize BP neural network. Firstly, the vibration signal of planetary gearbox is de-noising and feature extraction by wavelet transform; Then, the pheromone distribution in ant colony is optimized by genetic algorithm; Then, the ant colony algorithm is used to find the optimal solution of the weights and thresholds of the neural network parameters, and the optimization results are taken as the optimal parameters of BP neural network. The BP neural network is established to diagnose the four typical faults (pitting, wear, pitting wear, broken teeth) and normal working conditions of planetary gearbox. Compared with the neural network optimized by ant colony, ga-aco-bp has better fault recognition rate and diagnosis speed.

Scroll to Top