Soutenance de thèse : Xiaowen ZHU

A reference-free mill monitoring method based on the inter-insert periodic correlation in angular domain

Doctorante : Xiaowen ZHU 

Laboratoire INSA : LVA
É​cole doctorale : ED162 Mécanique, Energétique, Génie Civil, Acoustique de Lyon

Based on the tendency of Industry 4.0, this thesis targets the tool condition monitoring (TCM), which is the terminal process of flexible production. The aim is to detect abnormal tool behavior as early as possible to improve the surface quality of workpieces and to prevent subsequent losses from serious tool failures.
In this context, a methodology for monitoring the wear of end mills in real-time production based on the inter-insert correlation is presented. The approach takes advantage of the angular domain characteristics to segment the signal into periodic cycles of the same angular duration, which are then amenable to correlation analysis. Under high rotational speeds, the external working environment experienced by the individual teeth can be considered quasi-equivalent. Through the correlation analysis, the impact of the non-stationary operation on the monitored signal is effectively reduced. From a wide range of correlation algorithms, singular value decomposition (SVD) is selected to proceed with the analysis, and an ordered separability index with latent correlation characteristics is extracted to assess the current condition of the tool. The feasibility of the proposed indicator was valid and evaluated via the simulated signal and a series of experimental data collected by designed milling patterns.
The results demonstrate the promising development of this method in forming an efficient TCM system. The proposed approach is more independent of the cutting conditions (changes in speed or direction) than the traditional teach-in method and does not require a trial run. It partially fills the gap of tool monitoring demand in flexible manufacturing for customized small batch production. At the same time, inter-insert correlation is also seen as part of a broader framework for monitoring and maintaining rotating machinery. It has great potential to be applied to the analysis of different signals generated by other mechanical components with a rotating nature.