Fault causes identification of rotating machinery based on multiphase zoom permutation entropy

Rotating machinery generally operates under adverse working conditions due to its specific function requirement in industrial equipment, the mechanical faults with different root causes contain various oscillation patterns. To extract fault features from vibration signals, multi-dimensional permutation entropy-based methods emerge as promising tools by measuring dynamic information in multiple scales. Unfortunately, these methods suffer from the information loss problem due to extracting features from the single oscillation pattern and ignoring the sharp decrease of the orbit states in the phase space of shortened sub-signals. To tackle this problem, multiphase zoom permutation entropy (MZPE) is proposed to extract more separable dynamic information for different fault causes. First, MZPE employs zoom analysis to extract features from different oscillation patterns. Then, MZPE splices phase spaces of all decimated sub-signals to generate more orbits and promote a more dispersed state probability distribution, which enhances the quantity and separability of the fault characteristics for fault causes identification. At last, the MZPE based fault diagnosis model is constructed to identify mechanical faults with different root causes. The simulation is conducted to verify the superiority of the MZPE. The experiment demonstrates that the proposed method outperforms existing methods in identifying fault types with different root causes.