TOOL WEAR PREDICTION IN MACHINING A HYBRID FINITE ELEMENT AND EMPIRICAL METHODOLOGY

Abstract

Tool wear rate prediction using either empirical or finite element (FE) method has its respective advantages and drawbacks. Major drawbacks with empirical approaches are extensive experimental data are required to calibrate the empirical model constants and are bounded by the unique combination of the tool/workpiece material. Whereas with FE approach, long computational time are required even with current central processing unit speeds and peripheral component interconnection frequencies. The ratio between computation time and real cutting time is about 9600. The objective of this research is to develop and validate a unique hybrid empirical and finite element approach to predict tool wear rate that is computationally efficient and limited number of tests required to calibrate the empirical model. Four different empirical wear rate models were evaluated based on least number of quantitative assumptions involved and independency from the proportion of dominant wear mechanisms. The empirical wear rate models selected for evaluation are dependent on stresses and temperatures rather than process parameters. The stresses and temperature predicted by the FE model served as input for the empirical wear rate models. FE computational time was substantially reduced as this approach only needs to simulate the cutting process up to steady state with different flank wear length. Each unique flank wear length simulations are not dependent on the other. Therefore, the simulation can be performed in parallel, which increased computation efficiency in an exponential rate. An initial wear rate model was also developed, which is dependent on average stresses acting on the cutting edge. Experimental orthogonal cutting tests of AISI 1045 (165-190 BHN) with uncoated tungsten carbide were carried out to calibrate and validate the wear rate models. Usui wear rate model was the most robust when compared to the other models investigated. This was because there is unbounded restriction in selecting process parameters during model calibration. Good agreement between predicted and experimental tool wear rate with only two tests required to calibrate the empirical model. The computation time with such an approach is independent on tool wear rate. The initial wear rate model substantially improved the accuracy of the prediction only at the more aggressive cutting conditions as the wear mechanism were based on mechanical effect rather than thermal.