Abstract
A digital twin is a promising evolving tool for prognostic health monitoring. However, in rotating machinery, the transfer function between the rotating components and the sensor distorts the vibration signal, hence, complicating the ability to apply a digital twin to new systems. This paper demonstrates the importance of estimating the transfer function for a successful transfer across different machines (TDM). Furthermore, there are few algorithms in the literature for transfer function estimation. The current algorithms can estimate the magnitude of the transfer function without its original phase. In this study, a new approach is presented that enables the estimation of the transfer function with its phase for a gear signal. The performance of the new algorithm is demonstrated by measured signals and by a simulated transfer function.
Highlights
Prognostic health monitoring (PHM) by vibration signal analysis is a widespread method for condition-based maintenance (Carden and Fanning, 2004; Randall, 2004a,b, 2011)
Deep learning is a promising approach for PHM by vibration signals
Two main challenges complicate the application of deep learning tools for rotating components: (1) the shortcomings in examples of fault signals from real machinery and (2) generalization across different machines (TDM)
Summary
Prognostic health monitoring (PHM) by vibration signal analysis is a widespread method for condition-based maintenance (Carden and Fanning, 2004; Randall, 2004a,b, 2011). This study demonstrates the importance of transfer function estimation for transfer across different machines (TDM) for enhancing the generalization abilities of digital twins to new systems. The ARMA model estimates the coefficients that best explain the statistical distribution of the noise [minimizing the mean squared error (MSE, Wallach and Goffinet, 1989; Lehmann and Casella, 2006; Shalev-Shwartz and Ben-David, 2013)] between the sample and the predicted value based on the former samples These coefficients correspond to the poles and zeros of the transfer function. The training set was augmented by random circular rotations of the vibration signals to improve generalization abilities (Goodfellow et al, 2016)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
