Feature dimension reduction method of rolling bearing based on quantum genetic algorithm

In order to improve the correct recognition rate of rolling bearing failure, a feature dimension reduction method based on quantum genetic algorithm (QGA) is proposed. A kind of rolling bearing test bench with casing measurement points is introduced. Two one-way accelerometers are installed on casing at a 90-degree angle to monitor the acceleration signals of the casing. Firstly, with the method of wavelet analysis, the acceleration signals are decomposed into 5 levels, while distinct time-frequency features based on wavelet packet energy are obtained and treated as high dimension features. Secondly, these high dimension features are mapped to the quantum bits of the chromosomes in the quantum bit coding system of QGA. Then, from the rolling bearing experimental data samples, training samples are randomly selected to train the rolling bearing fault diagnosis model established based on QGA and artificial neural network. Meanwhile the prediction accuracy of the rolling bearing fault diagnosis model is used to construct the fitness function of QGA. Finally, by constant renewal calculation of the rolling bearing fault diagnosis model, the sensitive features can be selected from the high dimension features. The experiment results show that the signal feature dimension of rolling bearing can be effectively reduced by QGA, which contributes to rolling bearing fault diagnosis.