Due to the non-stationary characteristics of vibration signals acquired from rolling element bearing fault, the time-frequency analysis is often applied to describe the local information of these unstable signals smartly. However, it is difficult to classify the high dimensional feature matrix directly because of too large dimensions for many classifiers. This paper combines the concepts of time-frequency distribution(TFD) with non-negative matrix factorization(NMF), and proposes a novel TFD matrix factorization method to enhance representation and identification of bearing fault. Throughout this method, the TFD of a vibration signal is firstly accomplished to describe the localized faults with short-time Fourier transform(STFT). Then, the supervised NMF mapping is adopted to extract the fault features from TFD. Meanwhile, the fault samples can be clustered and recognized automatically by using the clustering property of NMF. The proposed method takes advantages of the NMF in the parts-based representation and the adaptive clustering. The localized fault features of interest can be extracted as well. To evaluate the performance of the proposed method, the 9 kinds of the bearing fault on a test bench is performed. The proposed method can effectively identify the fault severity and different fault types. Moreover, in comparison with the artificial neural network(ANN), NMF yields 99.3% mean accuracy which is much superior to ANN. This research presents a simple and practical resolution for the fault diagnosis problem of rolling element bearing in high dimensional feature space.

For the spindle of machine tool, rotation error is a main factor affecting the machining quality, so many investigations have been focused on the measurement of the rotation error. In this paper, a measurement and separation method of the rotation error is proposed to extract the orbit of spindle section centre and to quantize the rotation error with a special measurement system. Thereinto, systematic measurement strategy has been especially designed to achieve the measurement and analysis of the rotation error. When the measurement does not conform to the small error conditions, the method can reduce the noise level from sensors and improve measurement accuracy in measuring the orbit of spindle section centre. When it is necessary to accurately quantify the rotation error and the roundness error, the method also can effectively separate them with independent coordinate system and sensor bracket. Additionally, considering the effect of high temperature on spindle, the method can reduce the form error caused by thermal expansion in the measurement process, which is very important in industrial online measurement. Finally, the theoretical analyses are shown good agreement with experimental results, and these also reveal that it is an effective measurement method for the rotation error.

Wheel is an important component of bogie, and its geometry affects the dynamic behaviors of rail vehicles. To accurately obtain the geometric feature information of wheel tread profile, an on-machine measurement method with the 1-D laser sensor is investigated in this article. Combined with the specific morphology of tread profile and measurement model of laser sensor, the corresponding on-machine measurement model of wheel tread profile is established according to the characteristics of inclined and curved profiles. To improve the measurement accuracy, an adaptive error correction is studied to strengthen the measurement of complex profiles and suppress the measurement error of laser sensors. And the method achieves error suppression by more than 30% in the test and calibration experiments. Finally, the measurement method based on the laser sensor is put into the practical measurement of the wheel tread profile on an underfloor wheelset lathe, and the effectiveness of the proposed method is verified with experiments. These works can not only improve the measurement accuracy of wheel geometry but also enhance the machining capability of machine tool, which is important for the safe operation of rail vehicles.

For decades, bearing factory quality evaluation has been a key problem and the methods used are always static tests. This paper investigates the use of piezoelectric ultrasonic transducers (PUT) as dynamic diagnostic tools and a relevant signal classification technique, wavelet packet entropy (WPEntropy) flow manifold learning, for the evaluation of bearing factory quality. The data were analyzed using wavelet packet entropy (WPEntropy) flow manifold learning. The results showed that the ultrasonic technique with WPEntropy flow manifold learning was able to detect different types of defects on the bearing components. The test method and the proposed technique are described and the different signals are analyzed and discussed.

In this paper a stochastic resonance (SR)-based method for recovering weak impulsive signals is developed for quantitative diagnosis of faults in rotating machinery. It was shown in theory that weak impulsive signals follow the mechanism of SR, but the SR produces a nonlinear distortion of the shape of the impulsive signal. To eliminate the distortion a moving least squares fitting method is introduced to reconstruct the signal from the output of the SR process. This proposed method is verified by comparing its detection results with that of a morphological filter based on both simulated and experimental signals. The experimental results show that the background noise is suppressed effectively and the key features of impulsive signals are reconstructed with a good degree of accuracy, which leads to an accurate diagnosis of faults in roller bearings in a run-to failure test.