International Journal of Science, Engineering and TechnologyAuthors: Kush Joshi, Aryan Yardi, Anshika Goyal, S Ahmed Hamdan, Saravanakumar R
Abstract: – Unplanned tool failure in CNC machining leads to poor surface quality, scrapped parts and machine downtime. This paper addresses the design and implementation of a low-cost, real-time system for monitoring tool wear and estimating the remaining useful life (RUL) of cutting tools. The hardware architecture comprises an ADXL345 3-axis accelerometer mounted on the spindle housing, an electret microphone for acoustic emission sensing and an ACS712 Hall-effect sensor for spindle motor current. The signals are conditioned by a custom low-pass filter board and digitised by an Arduino Mega or an NI myDAQ. The software is developed in LabVIEW and Python and applies DC offset removal, Daubechies db4 wavelet denoising, windowed FFT and statistical feature extraction (RMS, kurtosis, crest factor, tooth-passing amplitude and sideband energy). Tool condition is classified as Healthy, Early-Wear or Critical using calibrated thresholds, and RUL is estimated by extrapolating the polynomial trend of the RMS feature, with machine learning models identified as the next stage of development. A real-time dashboard with alert thresholds presents the results to the operator. The system is being validated on a benchtop CNC router against four known tool conditions, with targets of at least 90 % classification accuracy, a false alarm rate below 5 % and an RUL error below 10 %. By enabling the early detection of tool wear, this framework is intended to support predictive maintenance, reduce unplanned downtime and extend tool life in automated manufacturing.
