Abstract
In order to prevent tool breakage and resultant decrease in productivity in unmanned turning operations, many researchers have attempted to developed tool wear estimation and classification models. These include neural network models, fuzzy logic models and working scenario for quantitative models. The worn tools need to be replaced before their wear exceeds the allowed limits. Normally, cutting forces, AErms and cutting conditions are employed as inputs in these models. In the recent past, however, many researchers have focused on flank wear prediction and off-line tool wear prediction systems. Additionally, the accuracy of tool wear prediction for these models needs to be increased. Therefore, a new on-line tool wear estimation system having higher accuracy for estimating the length of flank wear and the maximum depth of crater wear in CNC turning operations needs to be developed.
In order to archive the aim mentioned above, a new on-line tool wear estimation system was developed. This system employed a fuzzy neural network model to predict flank and crater wear. Cutting force, AErms, the derivatives of cutting forces, the derivatives of AErms and cutting conditions were used as inputs of this neural network model. Additionally, the fuzzy neural network model can also eliminate a tool wear estimation error due to a variation in actual cutting tool geometry. Due to the fact that tip fracture, chipping at the major cutting edge, or both on tool inserts cause greater forces and AErms signals, tool inserts having these defects could be detected from the significant increase in force signals. The detection of chipping and fracturing at tool cutting edges was also incorporated in the tool wear estimation system developed by the author.
Experimental results indicated that the new on-line tool-wear estimation system can estimate flank and crater wear accurately and eliminate tool wear estimation error due to a variation in actual cutting geometry. The computational time for this tool wear estimation was about 16 seconds. However, it decreased to 8 seconds for the subsequent flank and crater wear estimation during turning operation.