DI-UMONS : Dépôt institutionnel de l’université de Mons

Recherche transversale
(titres de publication, de périodique et noms de colloque inclus)
2020-04-06 - Travail avec promoteur/Doctorat - Anglais - page(s)

Equeter Lucas , "Optimal Replacement of Cutting Tools through Degradation Monitoring Techniques and Residual Lifetime Prediction Models", 2015-10-15, soutenue le 2020-04-06

  • Codes CREF : Sciences de l'ingénieur (DI2000), Probabilités (DI1132), Mécanique (DI1240), Mécanique appliquée générale (DI2100), Technologie de la construction mécanique (DI2130), Façonnage des métaux (DI2733), Usinage (DI2131), Processus stochastiques (DI1131), Statistique appliquée (DI1133)
  • Unités de recherche UMONS : Génie Mécanique (F707)
  • Instituts UMONS : Institut de Recherche en Science et Ingénierie des Matériaux (Matériaux), Institut des Sciences et du Management des Risques (Risques)

Abstract(s) :

(Anglais) In spite of advances in cutting tool technology, the timely replacement of cutting inserts remains a challenge to the machining industry. Their improper management is a major center of costs for machined workpieces, because the tool wear has a direct impact on the production quality. Condition monitoring approaches have been suggested for estimating the tool wear evolution, hence its Remaining Useful Life (RUL). The condition monitoring methodology is only seldom applied in the industrial practice, as the measures of interest (tool wear, cutting forces, vibratory or acoustic signals, temperature…) are impractical to acquire in real time given the conditions of machining (workshop environment, cutting fluids…). The present thesis therefore aims at using the least intrusive measurements, which shall be acquired in real time, to estimate the RUL of the cutting inserts, through statistical methods, coupled with information obtained by numerical models and experimental investigations. After a review of tool technology and degradation processes, the results of a Finite Element (FE) model of orthogonal cutting with worn tools are analyzed as a first approach to the cutting tool monitoring. In order to study the statistical aspects of the tool degradation, a stochastic model is fitted on experimental data in turning and used to generate degradation paths. It is shown to provide a RUL estimate based on few direct wear observations. Further, the Cox Proportional Hazards Model (PHM) is used to link condition monitoring data with the tool estimated lifetime. First, its use in the case of time-independent variables (i.e., the cutting parameters) is demonstrated. In this context, a data transformation is proposed to improve the model predictions, on the basis of analytical relationships between the monitored variables and the tool useful life, which are obtained from the FE model. Second, the use of the Cox PHM with time-dependent variables is demonstrated on condition monitoring variables, thus allowing an updated prediction of the tool RUL. Finally, experimental investigations allow investigating several condition monitoring variables (workpiece surface roughness, cutting forces, electrical currents, chip morphology) with respect to the evolution of tool wear, and assessing the Cox PHM approaches. The developed methodology shows the use of the Cox PHM to estimate the RUL of cutting tools in turning of ferrous alloys. Industrial practice may aim at incorporating a similar methodology for on-line estimate of tool RUL, through an inevitable first phase of data collecting necessary for the model fitting. The developed approach also contributes to answering the recurrent question of data transformation in using the Cox PHM. The condition monitoring variables are statistically assessed as tool wear indicators, and as covariates to these models, and show the importance of the cutting forces ratio as an indicator of tool end-of-life.