Monitoring the condition of a component is typically based on an empirical model that estimates the values of some measurable variables (signals) in normal conditions and triggers the fault alarm when the reconstruction deviates from the measured signal. When condition monitoring is performed during plant operational transients the intrinsically dynamic behavior of the signals should be taken into account. To this purpose, two approaches are proposed in this work. The former is based on the development of several reconstruction models, each one dedicated to a different operational zone of the component. The latter is based on the preprocessing of the signals by means of Haar wavelet transforms. The performance of the two proposed approaches are compared with that of the traditional reconstruction approach used for stationary conditions, with respect to a case study concerning the condition monitoring of a gas turbine during start-up transients.
ReifmanJ. Survey of artificial intelligence methods for detection and identification of component faults in nuclear power plants. Nucl Technol1997; 119: 76–97.
2.
VaidyaPRausandM. Remaining useful life, technical health, and life extension. Proc IMechE, Part O: J Risk and Reliability2011; 225(2): 219–231.
3.
BaraldiPCompareMDespujolsA. Modeling of the effect of maintenance actions on the degradation of an electric power plant component. Proc IMechE, Part O: J Risk and Reliability2011; 225(2): 169–184.
4.
BaraldiPCanesiRZioE. Signal grouping for condition monitoring of nuclear power plant components. In: 7th American nuclear society international topical meeting on nuclear plant instrumentation, control and human-machine interface technologies (NPIC&HMIT 2010), Las Vegas, Nevada, 7–11 November2010.
5.
HinesJWGarveyDR. Development and application of fault detectability performance metrics for instrument calibration verification and anomaly detection. J Pattern Recogn Res2006; 1: 2–15.
6.
StrangGNguyenT. Wavelets and filter banks. Wellesley-Cambridge Press, 1996.
7.
OgdenRT. Essential wavelets for statistical applications and data analysis. Boston: Birkhausen, 1997.
8.
VetterliMHerleyC. Wavelets and filter banks: theory and design. IEEE Trans Signal Processing1992; 40(9): 2207–2231.
9.
RoversoD. Fault diagnosis with the aladdin transient classifier. In: Proceedings of system diagnosis and prognosis: security and condition monitoring issues III, AeroSense2003, aerospace and defense sensing and control technologies symposium, Orlando, FL, 21–25 April2003.
10.
RoversoD. Soft computing tools for transient classification. Info Sci2000; 127: 137–156.
11.
BaraldiPPedroniNZioE. Application of a niched pareto genetic algorithm for selecting features for nuclear transients classification. Int J Computational Intell Sys2009; 24(2): 118–151.
12.
BaraldiPCanesiRZioE. Generic algorithm-based wrapper approach for grouping condition monitoring signal of nuclear power plant components. Integrated Computer-Aided Engng2011; 18(3): 221–234.
13.
GrossKCKumenikKE. Sequential probability ratio test for nuclear power plant component surveillance. Nuclear Technol1991; 93: 131–137.
14.
GrossKCLuW. Early detection of signal and process anomalies in enterprise computing systems. In: International conference on machine learning and applications (ICMLA), pp.204–210, June 24–27, 2002, Las Vegas, Nevada, US.
15.
ZioEBaraldiPPedroniN. Selecting features for nuclear transients classification by means of genetic algorithms. IEEE Trans Nuclear Sci2006; 53: 1479–1493.
16.
EfronB. Estimating the error rate of a prediction rule: Improvement on cross-validation. J Am Stat Ass1983; 78: 316–331.
17.
EfronBTibshiraniRJ. Improvements on cross-validation: the .632+ bootstrap method. J Am Stat Ass1995; 92: 548–560.
18.
KohaviR. A study of cross-validation and bootstrap for accuracy estimation and model selection. San Mateo, CA: Morgan Kaufman, 1995, pp.1137–1143.
19.
PolikarR. Bootstrap-inspired techniques in computational intelligence. IEEE Signal Processing Magazine2007; 59: 59–72.
20.
RoversoD. Multivariate temporal classification by windowed wavelet decomposition and recurrent neural networks. In: International topical meeting on nuclear power plant instrumentation, controls and human-machine interface technologies (NPIC&HMIT 2000), Washington, DC, November2000.
21.
BaraldiPZioEDi MaioF. Differential evolution for optimal grouping of condition monitoring signals of nuclear components. In: Proceedings of ESREL 2011 conference, Troyes, France, 18–22 September2011, pp.410–418.
22.
BaraldiPZioEGolaG. Genetic algorithms for signal grouping in sensor validation: a comparison of the filter and wrapper approaches. Proc IMechE, Part O: J Risk and Reliability2008; 222(2): 189–206.
23.
ZioEBaraldiPGolaG. Ensemble feature selection for diagnosing multiple faults in rotating machinery. Proc IMechE, Part O: J Risk and Reliability2007; 221(O1): 29–41.
