Sinusoidal synthesis based adaptive tracking for rotating machinery fault detection. (15th January 2017)

• Record Type: Journal Article
• Authors: Li, Gang; McDonald, Geoff L.; Zhao, Qing
• Is Part Of: Mechanical systems and signal processing. Volume 83(2017)
• Page Start: 356
• Page End: 370
• Publication Date: 2017-01-15
• Subjects: Sinusoidal synthesis -- Time series modeling -- Adaptive -- Vibration signal -- Fault detection; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2016.06.019
• Languages: English
• ISSNs: 0888-3270

Abstract

Abstract: This paper presents a novel Sinusoidal Synthesis Based Adaptive Tracking (SSBAT) technique for vibration-based rotating machinery fault detection. The proposed SSBAT algorithm is an adaptive time series technique that makes use of both frequency and time domain information of vibration signals. Such information is incorporated in a time varying dynamic model. Signal tracking is then realized by applying adaptive sinusoidal synthesis to the vibration signal. A modified Least-Squares (LS) method is adopted to estimate the model parameters. In addition to tracking, the proposed vibration synthesis model is mainly used as a linear time-varying predictor. The health condition of the rotating machine is monitored by checking the residual between the predicted and measured signal. The SSBAT method takes advantage of the sinusoidal nature of vibration signals and transfers the nonlinear problem into a linear adaptive problem in the time domain based on a state-space realization. It has low computation burden and does not need a priori knowledge of the machine under the no-fault condition which makes the algorithm ideal for on-line fault detection. The method is validated using both numerical simulation and practical application data. Meanwhile, the fault detection results are compared with the commonly adopted autoregressive (AR) and autoregressive Minimum Entropy Deconvolution (ARMED) method to verify the feasibility and performance of the SSBAT method. Abstract : Highlights: This proposed SSBAT model can predict and trend vibration signals. It is a dynamic adaptive model incorporating sinusoidal constraints. The rotating speed is assumed to be unknown, leading to unknown model parameters. The unknown model parameters are estimated using a modified least-squares algorithm. It is suitable for on-line rotating machinery impact fault detection.