Condition Based Maintenance is significantly important for complex mechanical systems in early-stage fault detection and catastrophic failure prevention, in which the key step is the extraction and identification of the operational information. The time-varying condition and complex strong noise will block the vital feature extraction which is the main challenge of the corresponding condition monitoring. Taking the advantages of the superiorities of both the Polish and Chinese side, the project conducts the fundamental research on the mathematical mechanism of the monitoring signal and proposes new methods of processing non-stationary signals (identification, segmentation, extraction, modeling) with non-Gaussian characteristics.
The specific contents include:
- Dynamic modeling for complex mechanical transmission system with extended defects and mathematical expression for fault-induced impulses;
- Impulsive noise modeling and nonstationary operational condition parametrization;
- Hidden Cyclicity/Periodicity detection (non-existing second-order statistics);
- Sparse separation and identification of impulsive features considering multi-source vibration responses and heavy noise interferences;
- Multidimensional data processing algorithms for impulsive sources separation and signal of interest extraction;
- TFR with a high time-frequency resolution for the non-stationary features, multiple time-varying frequency trends, extraction.
The research results have important reference significance for the fault diagnosis of complex mechanical equipment under time-varying nonstationary environments.
The topics for PL team have been bolded.