Multi-Scale Enveloping Spectrogram (MuSEnS) for Machine Condition Monitoring and Health Diagnosis
Robert X. Gao, Ph.D.
MuSEnS is a new multi-domain signal processing technique that integrates the wavelet transform, Fourier-based spectral analysis, and color mapping into one entity, to more effectively detect hidden features within a time-varying, dynamic waveform, such as vibration signals from a manufacturing machines, thus enabling better defect detection and diagnosis than using one of these techniques alone.
MuSEnS has overcome the limitations of a predetermined, fixed-time signal decomposition interval as required by other techniques. By means of the wavelet scaling operation, which is an integral part of the MuSEnS algorithm, complex, dynamically-changing features can be flexibly extracted from the signal.
The MuSEnS algorithm, through a user-friendly software interface, produces a three-dimensional scale-frequency map indicating both the intensity and location of the defect-induced frequency lines over the entire spectrum, thus helping identifying the existence of defect at an early stage, to diagnose or predict potential failures.
The MuSEnS technique can benefit various commercial and industrial applications, such as civil and mechanical structure health monitoring, transportation, aerospace, manufacturing, and defense. The algorithm can be integrated into popular scientific computing or data acquisition software packages (such as MatLab or LabView), or it can be coded as a stand-alone package for use on the factory floor to monitor machine tools, on-line, and in real time.
Specifically, when applied to monitoring operation conditions of factory equipment, such as bearings, spindles and gear boxes, MuSEnS can help improve the ability in early defect detection and prediction, to ensure continuous, safe operation; it can aid in planning condition-based instead of fixed-interval maintenance schedules.
For operation monitoring of commercial transportation vehicles, this technique can contribute to enhancing reliability and safety of passengers and cargo.
The technique can be adapted for human health monitoring by enhancing the capability of biomedical/medical signal decomposition, such as electrocardiogram (ECG), electromyography (EMG), and in portable and remote monitoring units.
- Overcomes deficiencies and limitations of Short Time or Repetitive Fourier Transform based analysis that require a predetermined, fixed time interval, by introducing adaptive time-frequency resolution that improves the signal-to-noise ratio in signal decomposition, making the technique less sensitive to noise contaminations.
- User-friendly software interface.
US Patent 7,602,985 issued
Office of Commercial Ventures and Intellectual Property