Switching Kalman filter for failure prognostic. (February 2015)

• Record Type: Journal Article
• Authors: Lim, Chi Keong Reuben; Mba, David
• Is Part Of: Mechanical systems and signal processing. Volume 52/53(2015)
• Page Start: 426
• Page End: 435
• Publication Date: 2015-02
• Subjects: Switching Kalman filter -- Rolling element bearing -- Bayesian estimation -- Prognostics -- Remaining useful life; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2014.08.006
• Languages: English
• ISSNs: 0888-3270

Abstract

Abstract: The use of condition monitoring (CM) data to predict remaining useful life have been growing with increasing use of health and usage monitoring systems on aircraft. There are many data-driven methodologies available for the prediction and popular ones include artificial intelligence and statistical based approach. The drawback of such approaches is that they require a lot of failure data for training which can be scarce in practice. In lieu of this, methods using state-space and regression-based models that extract information from the data history itself have been explored. However, such methods have their own limitations as they utilize a single time-invariant model which does not represent changing degradation path well. This causes most degradation modeling studies to focus only on segments of their CM data that behaves close to the assumed model. In this paper, a state-space based method; the Switching Kalman Filter (SKF), is adopted for model estimation and life prediction. The SKF approach however, uses multiple models from which the most probable model is inferred from the CM data using Bayesian estimation before it is applied for prediction. At the same time, the inference of the degradation model itself can provide maintainers with more information for their planning. This SKF approach is demonstrated with a case study on gearbox bearings that were found defective from the Republic of Singapore Air Force AH64D helicopter. The use of in-service CM data allows the approach to be applied in a practical scenario and results showed that the developed SKF approach is a promising tool to support maintenance decision-making. Highlights: Use of Switching Kalman filter for prognosis of bearings. Different degradation process can be modeled under one framework. Most probable degradation process is inferred and used for predictions. Method shown with measurement data of defective bearings from the AH64D helicopter.