Need for Adaptive Signal Processing Technique for Tool Condition Monitoring in Turning Machines

This study deals with a comparative study of the processing of tool-emitted sound signal using conventional signal processing technique, FFT and an adoptive signal processing technique, HHT for Tool Condition Monitoring (TCM) in a turning machine. The tool-emitted sound signal obtained for the purpose of TCM is used to classify the condition of the cutting tool insert into one of the three states: Fresh, slightly worn and severely worn. Signal processing techniques are used in this study for extracting features from the tool-emitted sound to train a Competitive Neural Network (CNN) for tool-wear classification. Results of the study show that the CNN trained by the features extracted using HHT performs more accurate classification than the same CNN trained by the features extracted using FFT. Hence, this study leads to the conclusion that adaptive signal processing technique, HHT is more suitable than FFT for designing accurate machine tool condition monitoring systems