The Detection Method Of Cnc Machine Tools Cutlery Based On Pattern Recognition

The detection of computer numerical control (CNC) machine tools cutlery can be affected by the quality of training samples for BP neural network. On the other hand, the recognition rate of one kind of samples can not match the average recognition rate of the whole samples. For this two reason, the recognition result is not so satisfactory as we hope. By the mothed of Online Feedback, which includes the Genetic Algorithms used to improve the quality of training samples, and Adaboost Algorithms used to improve the recognition rate of one kind of samples by the way of paralleling multiple BP neural network, the average recognition rate of the whole samples and the recognition rate of one kind of samples are both improve to a satisfactory level.This thesis aims to develop the accuracy of the detection of CNC machine tools cutlery by building Online Feedback.A way of feature extraction methods of the detection of CNC machine tools cutlery is presented. By analyzing the time domain and frequency domain of force signal and vibration signal,13parameters are presented as the eigenvalue.And the main means of the detection of CNC machine tools cutlery and the main means of identification methods are introduced in this paper.The method that based on the Online Feedback of BP neural network, which including the Genetic Algorithms and the Adaboost Algorithm is presented in detail. In addition, the defects of BP neural network are listed in this paper.And in response to these shortcomings, the combination algorithm of Genetic Algorithms and BP neural network is presented for improving the average recognition rate of the whole samples, which is proved by experiments. On the other hand, the way of paralleling multiple BP neural networks based on the Adaboost Algorithms is presented and the effect of improving the recognition rate of one kind of samples is proved by the comparison of the recognition rate.In the end, the actual results will be analyzed by comparing he recognition rate between Online-Feedback and Non-Online-Feedback to prove the significant impact of Online Feedback.