KFD-AEEScan: a Katz fractal dimension acoustic emission event scanner for single-sensor multi-fault diagnosis in milling machines

Abstract

Reliable condition monitoring of milling machines (MMs) is essential for maintaining quality of the output product and avoiding downtime, but traditional acoustic emission (AE)-based approaches using features such as root mean square (RMS), kurtosis, or band-limited energies either blur burst events with background activity or rely on ad hoc thresholding of AE hits. As a result, rotation-synchronous AE bursts generated by tool wear, bearing damage, and gear defects are incompletely captured and difficult to exploit for robust multi-fault diagnosis. This article presents Katz fractal dimension AE event scanner (KFD-AEEScan), a single-sensor diagnostic framework that processes AE signals into time-scale maps of local fractal complexity, replacing purely amplitude- or energy-based traces with a multiscale representation more sensitive to transient burst activity. A sliding-window Katz fractal-dimension estimator is applied at multiple window lengths, producing KFD traces in which burst-dominated regions gravitate toward the lower bound of KFD, while quasi-stationary background occupies higher-complexity levels. These traces are synchronized and stacked into compact two-dimensional time-scale maps, in which multiple KFD scales jointly encode the prominence and timing of AE bursts along each cutting pass. The framework is validated on real milling-machine AE data under multiple spindle speeds and four health states: normal, cutting tool wear, spindle bearing fault, and gearbox fault. KFD-AEEScan achieves high, well-balanced macro-F1 across all classes and outperforms representative wavelet-fractal, short-time Fourier transform-based, and Variational Mode Decomposition (VMD)-fractal baselines. The results indicate that localized KFD mapping of AE bursts provides an effective representation for multi-fault diagnosis and predictive maintenance in MMs.