Advanced correlation-based anomaly detection method for predictive maintenance
Authors:
- 1 June 2017
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 78-83
Abstract
Variations in sensor data collected from equipment have been widely analyzed by using anomaly detection methods for predictive maintenance. Our experience shows that correlations between sensors effectively predict failures because the correlations usually reflect the status of equipment with higher sensitivity. In this paper, we present a method that exploits correlations between sensors for pre-processing and enables anomalies to be detected using both sensor data and correlations. The method was evaluated by applying it to compact electric generators, and the results showed it detected anomalies more accurately than when only sensor data were used. This method is expected to predict failures earlier and reduce the cost of downtime and maintenance.Keywords
This publication has 8 references indexed in Scilit:
- The present status and future growth of maintenance in US manufacturing: results from a pilot surveyManufacturing Review, 2016
- An improved correlation-based anomaly detection approach for condition monitoring data of industrial equipment2016
- Machine Learning for Predictive Maintenance: A Multiple Classifier ApproachIEEE Transactions on Industrial Informatics, 2014
- Detecting Novel Associations in Large Data SetsScience, 2011
- State-of-the-Art Predictive Maintenance Techniques*IEEE Transactions on Instrumentation and Measurement, 2011
- Brownian distance covarianceThe Annals of Applied Statistics, 2009
- Predictive maintenance techniquesIEEE Industry Applications Magazine, 2009
- Continuous-time predictive-maintenance scheduling for a deteriorating systemIEEE Transactions on Reliability, 2002