Cutting Tools Operation Reliability Assessment Based on Support Vector Space

For a single or a small batch CNC turning cutting tool reliability estimation, the traditional reliability estimation methods large sample statistics-based is inefficient. Which are limited for some reasons, such as the difficulty for time-dynamic process description, inaccurate model and non-individual characteristics. In order to improve the precision and credibility of reliability assessment to the cutting tools under the condition of a single or a small batch sample, a new operation reliability estimation method based on singular value decomposition (SVD) transform and support vector space is proposed. Firstly, the vibration signals of the tool holder are acquired during the cutting process. The salient features closely related to the tool wear are extracted by wavelet packet decomposition, energy distribution and time-frequency statistic analysis. The SVD method is employed for the dimensionality reduction of high dimension feature data so as to reduce the computational complexity and the redundant component. Secondly, dimension reduced data are substituted into support vector space model to establish a hyper sphere. Then the relative distance between the sample points and the hyper sphere is calculated and used to describe the degradation of the tool. The semi normal function is further introduced to reflect the mapping relationship of the relative distance and the operation reliability of the tool. Two tools whose wear states are failure and normal are used as evaluation cases respectively. The data before and after dimension reduction are used to established the support vector hyper sphere space for reliability evaluation. The results show that dimensional data reduction can effectively reduce the deformation of the hyper sphere coming from the data dispersion. Finally, the unified data of multiple tools under the uniform failure threshold are used for reliability evaluation of the 1st tool. The results show that the changing trend of the tool's relative distance and running reliability are more obvious and whose volatility is reduced when the training data conditions are sufficient. Especially, at the end of tool wear, it is conducive to the accurate evaluation of the tool service performance. The proposed operational reliability assessment method is free from the dependence of the traditional reliability assessment method on large-sample statistical data, which enriches and develops the reliability evaluation theory. © 2018, Editorial Department of Advanced Engineering Sciences. All right reserved.