Health monitoring and performance degradation prediction of large-size and wide-tooth milling machine blades

The disc milling cutter is the key moving part in the milling process of large-size and wide-tooth milling machine, which determines the machining accuracy and quality of the gear surface. Contact and collision between the blade and the gear surface leads to gradual dulling of the blades, affecting the accuracy of the tooth surface. The monitoring of blades during continuous machining has been a hot and difficult research topic. In this paper, a method for blade health state assessment and performance degradation trend prediction based on semi-supervised convolutional ladder network (SSCLN) with maximum mean discrepancy (MMD) is proposed. Wavelet packets are used to decompose the original vibration signals, from which the reconstructed vibration signals are computed to obtain the RMS thresholds for blade failure, and based on the results, the training and testing sample set labels are constructed. This approach solves the problem of indirect monitoring of blades, where health degradation labels cannot be obtained under continuous non-stop operation. The large, wide-toothed, disc-shaped gear milling machine tool manufactured by our team was used to practical process large inner-ring gear workpieces to obtain a sample dataset, which verifies the superiority of the methodology proposed in this paper. The obtained health assessment indicator has a better trend relative to the common indicators, and the predicted health degradation labels and remaining useful life (RUL) results are more accurate, which suggests that the method has practical application.