Outer race defect width measurement in taper roller bearing using discrete wavelet transform of vibration signal

Abstract

Rolling element bearing is one of the important components in rotary machines. Although a significant quantum of work has been done on bearing defect monitoring, estimation of defect size in bearing elements is still a challenge. In this paper, a technique based on decomposition using Symlet wavelet is proposed for measuring outer race defect width of taper roller bearing. Experiments were carried out on a customized test setup. Seeded defects of different size were introduced separately in the form of an axial groove on the outer race of taper roller bearings using laser engraving technique. It is not only difficult but ambiguous to detect the entry point in the groove defect by making use of the signal. The ambiguity gets reduced by using Symlet5 wavelet due to its linear phase nature which maintains sharpness in the signal even when there is a sudden change in signal. The proposed technique has been successfully implemented for measuring defect width over a range of 0.5776–1.9614 mm. The resulting measure of defect width has been also verified using image analysis. Maximum deviation in result has been found to be 4.06% for the defect width of 1.1820 mm at no load which was found further reduced to 1.02% with increase in load.

Introduction

Bearing defect is one of the causes of breakdown in rotating equipments. Different techniques based on the principle for measuring vibration, acoustic, thermal condition and wear analysis are used for detection of the bearing defects. Techniques based on vibration and acoustic emission are developing at a faster pace for monitoring rolling element bearings due to their non-invasive nature and their high reactivity to incipient faults [1], [2]. In general, the diagnosis capability of the techniques is proportional to their complexity [3]. When bearings are installed as part of a complex mechanical system, the measured signal is often heavily clouded with various noises resulting from the compound effect of interferences caused by other machine elements and the background noise present in the measuring device [4]. The defect position can be identified by time–frequency analysis of the vibration or acoustic emission signal. Junsheng et al. [5] constructed an impulse response wavelet by using continuous wavelet transform to extract the feature of vibration signals. They proposed two methods, namely, the scale-wavelet power spectrum comparison and the auto-correlation analysis of time-wavelet power spectrum for detecting the faults of roller bearing and identifying the fault patterns. Yan and Gao [6] presented a signal processing algorithm based on multi-scale enveloping spectrogram for vibration signal analysis. This technique was evaluated for localized structural defects and experimentally found successful. Patil et al. [7] developed an analytical model to predict the effect of a localized defect on the ball bearing vibrations by considering the contact between the ball and the races as non linear springs. For deep groove ball bearing they obtained numerical results by using the model which yielded both the frequency and the acceleration of vibration components of the bearing for simulating the defect.

The analytical mode has its own drawback in meeting real-time scenario. To overcome this, the envelop analysis – the most popular signal processing method for bearing is proposed. In this method the vibration signal is first passed through a band-pass filter to obtain a high signal-to-noise ratio (SNR) signal, and then Hilbert transform is used to obtain the envelope [8]. Patel et al. [9] proposed a method for detection of local defects on races of deep groove ball bearing in presence of external vibration using envelope analysis and Duffing oscillators but could not attempt for measurement of defect size. Kumar et al. [10] investigated Analytical Wavelet Transform (AWT) based acoustic emission technique for identifying not only the presence but also the severity of the defect in the inner race of radial ball bearing. A comparison of wavelet decomposition-based de-noising method and wavelet filter-based de-noising method was made by Qiu et al. [11] for signals from mechanical defects. The finding of the comparison reveals that the wavelet filter based de-noising method is more suitable to detect a weak signature. The wavelet decomposition de-noising method achieved satisfactory results on smooth signal detection.

Al-Ghamd and Mba [12] made an attempt to determine the bearing outer race defect width directly from the raw signal. The relationship between the defect size and acoustic emission of burst size was a significant finding but had not dealt with vibration signal in detail. Sawalhi and Randall [13] attempted to measure the seeded defect width by using two different approaches. In first approach pre-whitening of signal is done and then octave band wavelet analysis is carried out to allow selection of the best band (or scale) for balancing the two pulses with similar frequency content. In the second approach, separate treatment is applied to the step and the impulse responses, so that they may be equally represented in the signal. They emphasized the theory that the rolling element could strike at the end of a spall, as the ball moved half way through it and that impact could get reflected in the signal. The impact in the signal might be due to contact of the ball with the spall base. The peak of impulse in signal would have emerged when the ball might have come out from the spall after touching the exit corner of the spall. So there is possibility of error in measurement of defect width using the above method implemented by Sawalhi and Randall [13]. Clearly there appears a room for improvement in the methods suggested for identifying defect size in ball bearing system as locating exact position of commencement and end of the defect by vibration signal is still a challenge.

The current work is carried out to estimate defect width in outer race of taper roller bearing using Symlet based wavelet decomposition. Extracting desired information from the signal is difficult especially when there is a mismatch between direction of defect and the roller-outer race line contact while passing over the defect. The acceleration signals resulting from the entry of the rolling element into the defect and that from its exit are of different nature. The Symlet wavelet is a near symmetrical/linear phase filter, which makes it easier to deal with the small discontinuity present in the signal without any major loss of information which helps in properly locating the point of commencement and exit of roller from the groove defect.