A semi-supervised learning method combining tool wear laws for machining tool wear states monitoring

In tool wear states monitoring, obtaining tool wear labels is an expensive and difficult task, and deep learning models trained under limited labelled samples are prone to overfitting. Moreover, tool wear is a continuous, slow and irreversible process, and this physical law should be taken into account. In response to this issue, this article proposes a semi-supervised learning method combining tool wear laws (SSLTWL) for machining tool wear states monitoring. Specifically, under limited data, establish a deep learning model based on metric learning to learn compact feature representations of labeled samples. The pseudo-label sample selection mechanism is designed based on the confidence level and tool wear law to ensure the diversity and accuracy of pseudo-label samples. In order to effectively update the classifier, the improved generative adversarial networks is used to generate minority samples to adaptively maintain data class balance during the semi-supervised learning process. Adopting an iterative self-training strategy to train the model multiple times. Milling experiments and boring engineering experiments were carried out to verify the effectiveness of the proposed method. Compared with other methods, the proposed method can accurately identify the tool wear state and achieve the optimal performance.