Abstract:
AISI 4340 steel is very useful in many Engineering applications but difficult to machine due to its mechanical properties of hardness which easily wear out cutting tools. The focus of this research work is to develop some selected devices and model(s) for monitoring cutting tool condition during finish-turning operation on AISI 4340 steel towards achieving adequate surface finish at optimal tool life. A Power-Temperature (PT) Meter was developed to capture machine tool power, current/voltage inputs, as well as tool-tip temperature during machining operation. Also, an Accelerometer-Force (AF) Meter which captures the acceleration/vibration signals during machining was developed. The outputs of the devices were in readable and editable excel format which allows for further analysis. Data were collected from randomized experimental runs conducted on a CBT1640 CNC Lathe machine during finish-turning of AISI 4340 steel with CNMG120408MN KU30T Tungsten Carbide tool. The data obtained for the workpiece surface roughness were processed using Analysis of variance (ANOVA), available in Minitab 19 software, to investigate the effect of cutting condition on surface roughness and as well determine the optimum cutting condition. Also, a model for predicting surface roughness from the cutting conditions was developed and validated. An optimum cutting conditions of cutting speed of 180 m/min, federate of 1.5 mm/rev, and depth of cut of 0.3 mm were obtained and then employed to conduct further turning experiments over the life cycle of the cutting tool during which some process signals were collected using the developed devices. The surface roughness associated with each set of process signal was recorded. Subsequently, the process signals trend with the surface roughness were analysed to investigate their capability to identify tool conditions. Also, the process signal feature that has high correlation with tool condition monitoring in respect of workpiece surface roughness were obtained. Finally,
Multilayer Perceptron (MLP) and Radial Basis Function (RBF) neural networks, using the signal features with high correlation coefficient as inputs, were developed and validated for the determination of tool conditions. Results revealed that the main effect of the feed rate and the interaction effect of the feed rate and cutting speed were found to have statistical significance on the workpiece surface roughness at 95% confidence level. Lastly, the MLP and RBF neural networks developed were both found suitable for determining the tool condition in finish-turning of AISI 4340, although MLP performed better with a performance error of 19.8% as compared with the RBF with a performance training and test error of 83.7% respectively.
