An HMM and polynomial regression based approach for remaining useful life and health state estimation of cutting tools

Abstract There has been considerable advances, over the last few decades, in sensing instrumentation, hardware, signal processing algorithms, and internet technology infrastructure that has eventually paved the way for long envisioned concept of smart factories under the concept of Industry 4.0. It has been projected that by 2020, somewhere from 26 to 50 billion “things” will be connected to the Internet. This will lead to astonishing amount of precise data in real time. To leverage on the collected data, factories across the globe is embracing digital manufacturing in the era of fourth industrial revolution, Industry 4.0. Combined with today’s more advanced analytics, these data enable industrial organizations to implement new, more effective maintenance strategies to progress further along on the maturity continuum from reactive, to preventive, to condition-based, to predictive, and – ultimately – to prescriptive maintenance. This instigate development of advanced analytic techniques to large scale deployment of conditioned based maintenance. We present a novel approach for health state estimation to facilitate autonomous diagnostics and thereupon we build our prognostics module based on the diagnostics result using a polynomial regression models. The proposed research presents a model based sequential clustering on time series sensor signals for estimating health states of a cutting tool. Results obtained validates the competitive performance of the method on a CNC machining test-bed outfitted with thrust-force and torque sensors for monitoring drill-bits.