Staged feature extraction method based on sawing vibration noise characteristics

Signal feature extraction is the initial stage of original data processing in tool damage monitoring. The quality of the extracted feature values will have a significant impact on the recognition accuracy. In this paper, aiming at the shortcomings of insufficient and inaccurate feature extraction of traditional feature extraction methods, a staged feature extraction method based on sawing vibration noise characteristics is proposed. According to the sawing vibration noise characteristics, the whole sawing process is divided into six sawing stages, and the feature values of each sawing stage and the entire signal stage are extracted. A sensitive feature screening method based on feature value change rate is proposed to screen out more sensitive feature values to circular saw blade damage. Finally, establish a circular saw blade damage classification recognition model based on CNN and conduct training and testing. The test results show that compared with the traditional feature extraction method, the accuracy of model recognition is significantly improved after using the feature extraction and screening method proposed in this paper. The results prove the superiority of the feature extraction and screening methods proposed in this paper in improving the accuracy of model recognition.