Application of vibration analysis using time-frequency analysis to detect and predict mechanical failure during the nut manufacturing process

Abstract

The nut manufacturing machine operates almost continuously throughout the day and, consequently, requires frequent maintenance, the duration of which depends on the type of fault. Unpredicted downtime can cause loss of production volume, inability to meet critical timelines, and product quality degradation. Vibration research was conducted on a six-die nut manufacturing machine, which experienced frequent faults on the fourth and sixth dies that exerted the greatest impact on the nut material. The application of several analytical techniques on vibration data – the fast Fourier Transform, Empirical Mode Decomposition, Multiscale Entropy, Shock Response Spectrum, and correlation – enabled the identification of fault indicators, which would be useful for predictive maintenance. These features were detectable up to 8 h 28 min and up to 1 h 9 min prior to failure or deterioration resulting from severe and regular damage, respectively. Faults included quality deterioration and die and non-die breakage, fusion, bending and cracking. This research can optimise machine efficiency through timely upkeep facilitated by predictive maintenance.