The tool wear condition monitoring is key to ensuring product quality. This paper develops a direct technique dealing with cutting tool images to automate the tool wear detection and identification. The constructed U-Net based network can realize an effective and reliable extraction of the tool wear area. The introduction of deep supervision with a Matthews Correlation Coefficient (MCC) based surrogate loss function helps to address the few-shot and data imbalance issues. Experiments on the images with wear on the flank face of cutting tools from a CNC turning machine show the effectiveness, competitiveness, and reliability of the proposed method under different types of loss functions.

It is very difficult to detect weak fault signatures due to the large amount of noise in a wind turbine system. Multiscale noise tuning stochastic resonance (MSTSR) has proved to be an effective way to extract weak signals buried in strong noise. However, the MSTSR method originally based on discrete wavelet transform (DWT) has disadvantages such as shift variance and the aliasing effects in engineering application. In this paper, the dual-tree complex wavelet transform (DTCWT) is introduced into the MSTSR method, which makes it possible to further improve the system output signal-to-noise ratio and the accuracy of fault diagnosis by the merits of DTCWT (nearly shift invariant and reduced aliasing effects). Moreover, this method utilizes the relationship between the two dual-tree wavelet basis functions, instead of matching the single wavelet basis function to the signal being analyzed, which may speed up the signal processing and be employed in on-line engineering monitoring. The proposed method is applied to the analysis of bearing outer ring and shaft coupling vibration signals carrying fault information. The results confirm that the method performs better in extracting the fault features than the original DWT-based MSTSR, the wavelet transform with post spectral analysis, and EMD-based spectral analysis methods.