Extension of Speed Transform for Fault Diagnosis of Rotating Machinery Under Nonlinear Speed Variations

Prompt fault diagnosis of rotating machinery is crucial for minimizing economic losses and preventing safety accidents in industrial applications. Fourier transform (FT) is a fundamental fault characteristic frequency extraction method of rotating machinery under constant speed conditions. However, traditional fault diagnosis methods based on FT struggle with detecting time-varying frequencies of rotating machinery under variable speed conditions. Speed transform (ST), an order analysis approach without angle domain resampling, has been proposed for rotating machinery fault feature extraction with linear speed variations. However, the speed variations in rotating machinery are frequently nonlinear, especially during the start-up and shutdown processes, inhibiting the engineering application of ST. In this article, we extend the ST algorithm to encompass a more general case of nonlinear speed variations for fault diagnosis of rotating machinery. The rationality of this extension is demonstrated through theoretical analysis, numerical simulation, and two experiments focused on rotor rubbing and bearing outer ring fault diagnosis. The results highlight the superiority of ST over FT in extracting fault signatures of rotating machinery from vibration signals under nonlinear speed variations, achieving comparable performance to computed order tracking, showcasing its potential in developing new signal processing methods for variable speed.