Monitoring of an industrial machining centre based on swarm-intelligence-inspired feature mode decomposition and single-valued neutrosophic cross-entropy

This paper proposes a health-guided, uncertainty-aware fault diagnosis framework for industrial machining centers using acoustic emission (AE) signals. AE-based diagnosis is difficult because signals suffer from strong background noise, non-stationarity, and uncertainty arising from multiple generation mechanisms (shear-zone deformation, chip–tool friction, and localized material deformation). The framework combines two main techniques. First, swarm-intelligence-inspired feature mode decomposition (SFMD): a new fault-aware health index (FAHI), integrating approximate entropy (ApE) and the Gini index (GI), captures both fault-induced impulsiveness and energy sparsity. FAHI serves as the fitness function for the sparrow search algorithm (SSA), which adaptively optimizes SFMD’s filter length and mode number for measurement-consistent decomposition across operating conditions. Second, after decomposition, the root mean square index (RMSI) is computed per mode and transformed into single-valued neuromorphic sets (SVNS) via eigenvalue-based intervals. Faults are identified using the minimum argument principle based on single-valued neuromorphic cross-entropy (SVNCE) between training and testing samples, explicitly modelling measurement uncertainty and feature overlap. Validated on two real-world datasets acquired under milling and idle conditions, the framework achieves average classification accuracies of 98.1% and 99.3% respectively, outperforming state-of-the-art methods and demonstrating its effectiveness and robustness for intelligent machining-center fault diagnosis.