Development of new fault detection methods for rotating machines (roller bearings)

Abstract Early fault diagnosis of roller bearings is extremely important for rotating machines, especially for high speed, automatic and precise machines. Many research efforts have been focused on fault diagnosis and detection of roller bearings, since they constitute one the most important elements of rotating machinery. In this study a combination method is proposed for early damage detection of roller bearing. Wavelet packet transform (WPT) is applied to the collected data for denoising and the resulting clean data are break-down into some elementary components called Intrinsic mode functions (IMFs) using Ensemble empirical mode decomposition (EEMD) method. The normalized energy of three first IMFs are used as input for Support vector machine (SVM) to recognize whether signals are sorting out from healthy or faulty bearings. Then, since there is no robust guide to determine amplitude of added noise in EEMD technique, a new Performance improved EEMD (PIEEMD) is proposed to determine the appropriate value of added noise. A novel feature extraction method is also proposed for detecting small size defect using Teager-Kaiser energy operator (TKEO). TKEO is applied to IMFs obtained to create new feature vectors as input data for one-class SVM. The results of applying the method to acceleration signals collected from an experimental bearing test rig demonstrated that the method can be successfully used for early damage detection of roller bearings. Most of the diagnostic methods that have been developed up to now can be applied for the case stationary working conditions only (constant speed and load). However, bearings often work at time-varying conditions such as wind turbine supporting bearings, mining excavator bearings, vehicles, robots and all processes with run-up and run-down transients. Damage identification for bearings working under non-stationary operating conditions, especially for early/small defects, requires the use of appropriate techniques, which are generally different from those used for the case of stationary conditions, in order to extract fault-sensitive features which are at the same time insensitive to operational condition variations. Some methods have been proposed for damage detection of bearings working under time-varying speed conditions. However, their application might increase the instrumentation cost because of providing a phase reference signal. Furthermore, some methods such as order tracking methods still can be applied when the speed variation is limited. In this study, a novel combined method based on cointegration is proposed for the development of fault features which are sensitive to the presence of defects while in the same time they are insensitive to changes in the operational conditions. It does not require any additional measurements and can identify defects even for considerable speed variations. The signals acquired during run-up condition are decomposed into IMFs using the performance improved EEMD method. Then, the cointegration method is applied to the intrinsic mode functions to extract stationary residuals. The feature vectors are created by applying the Teager-Kaiser energy operator to the obtained stationary residuals. Finally, the feature vectors of the healthy bearing signals are utilized to construct a separating hyperplane using one-class support vector machine. Eventually the proposed method was applied to vibration signals measured on an experimental bearing test rig. The results verified that the method can successfully distinguish between healthy and faulty bearings even if the shaft speed changes dramatically.