Abstract
To find a fast and efficient fault diagnosis method for induction motor using limited data samples, the modelling and application of deep learning in induction motor fault diagnosis was studied. Firstly, the application of traditional convolution neural network (CNN) in fault diagnosis of induction motor was analysed, and a convolutional discriminative feature learning (CDFL) method based on improved CNN training mode used for discriminative learning of induction motor fault characteristics was put forward. The method mainly used BP neural network discriminative learning ability to learn local features of convolutional layer filter, so that the convolution pool model could learn the default features with the fault characteristic invariant of the induction motor vibration signal not adjusting the network parameters. In addition, the correct classification of fault type was achieved by selecting support vector machine SVM. At last, experiment was designed to compare CDFL method and signal processing wavelet packet transform method and it was compared with other deep learning methods. The experimental results showed that the classification results of CDFL fluctuated the minimum; when the filter window size was 200, the classification effect of model was the best and when the pool domain size was 20, it achieved the best effect of classification. To sum up, CDFL has the highest classification accuracy and good robustness, and it can learn the fault characteristics of induction motor more quickly, intelligently and effectively. To study the application of data mining technology in chemical engineering optimization, the data mining technology was adopted in this paper to discretize and analyze the experimental data. Results have shown that after the data discretion by the database technology, the expansion rate was below the optimal value when the deposition rate and recovery rate of the chemical equipment were the best. It was then concluded that the application of data mining technology in chemical engineering optimization, far from unrealistic, can greatly improve economic benefits for the company, for which it can be highly promoted and widely applied.
