Statistical techniques for automating the detection of anomalous performance in rotating machinery
The level of technology utilized in automated systems that monitor industrial rotating equipment and the potential of alternative surveillance methods are assessed. It is concluded that changes in surveillance methodology would upgrade ongoing programs and yet still be practical for implementation. An improved anomaly recognition methodology is formulated and implemented on a minicomputer system. The effectiveness of the monitoring system was evaluated in laboratory tests on a small rotor assembly, using vibrational signals from both displacement probes and accelerometers. Time and frequency domain descriptors are selected to compose an overall signature that characterizes the monitored equipment. Limits for normal operation of the rotor assembly are established automatically during an initial learning period. Thereafter, anomaly detection is accomplished by applying an approximate statistical test to each signature descriptor. As demonstrated over months of testing, this monitoring system is capable of detecting anomalous conditions while exhibiting a false alarm rate below 0.5%.
- Research Organization:
- Oak Ridge National Lab., TN (USA)
- Sponsoring Organization:
- USDOE Environment and Safety
- DOE Contract Number:
- W-7405-ENG-26
- OSTI ID:
- 6258858
- Report Number(s):
- CONF-790435-2; TRN: 79-017531
- Resource Relation:
- Conference: CAM-I international spring seminar, New Orleans, LA, USA, 9 Apr 1979
- Country of Publication:
- United States
- Language:
- English
Similar Records
Statistical algorithm for automated signature analysis of power spectral density data
Day, night and all-weather security surveillance automation synergy from combining two powerful technologies