Enhanced condition monitoring of the machining process using wavelet packet transform

Tool wear in machining processes can have a detrimental impact upon the surface finish of a machined part, increase the energy consumption during manufacture and potentially, if the tool fails completely, damage incurred may require the part to he scrapped. Monitoring of the tools condition can therefore lead to preventative steps being taken to avoid excessively worn tools being used during machining, which could cause a part becoming damaged. Several studies have been devoted to condition monitoring of the machining process, including the evaluation of cutting tool condition. However, these methods are either impractical for a production environment due to lengthy monitoring time, or require knowledge of cutting parameters (e.g. spindle speed, feed rate, material, tool) which can he difficult to obtain. In this study, we aim to investigate if tool wear can be directly identified using features extracted from the electrical power signal of the entire Computer Numerical Control (CNC) machine (three phase voltage and current) captured at 50 KHz, for different cutting parameters. Wavelet packet transform is applied to extract the feature from the raw measurement under different conditions. By analyzing the energy and entropy of reconstructed signals at different frequency sub-bands, the tool wear level can be evaluated. Results demonstrate that with the selected features, the effects due to cutting parameter variation and tool wear level change can be discriminated with good quality, which paves the way for using this technique to monitor the machining process in practical applications.