A weighted multi-scale morphological gradient filter for rolling element bearing fault detection

Abstract

This paper presents a novel signal processing scheme, named the weighted multi-scale morphological gradient filter (WMMG), for rolling element bearing fault detection. The WMMG can depress the noise at large scale and preserve the impulsive shape details at small scale. Both a simulated signal and vibration signals from a bearing test rig are employed to evaluate the performance of the proposed technique. The traditional envelope analysis and a multi-scale enveloping spectrogram algorithm combining continuous wavelet transform and envelope analysis (WT-EA) are also studied and compared with the presented WMMG. Experimental results have demonstrated the effectiveness of the WMMG to extract the impulsive components from the raw vibration signal with strong background noise. We also investigated the classification performance on identifying bearing faults based on the WMMG and statistical parameters with varied noise levels. Application results reveal that the WMMG achieves the same or better performance as EA and WT-EA. Meanwhile, the WMMG requires low computation cost and is very suitable for on-line condition monitoring of bearing operating states.

Introduction

Rolling element bearing failure is one of the foremost causes of failures in rotating machinery, and such failure may result in costly production loss and catastrophic accidents. Early detection and diagnosis of bearing defects while the machine is still in operation can help to avoid abnormal event progression and to reduce productivity loss. One of the most popular tools for diagnosing bearing problems is vibration signal analysis, due to its effectiveness and ease of measurement. By employing appropriate data analysis algorithms, it is feasible to detect changes in vibration signals caused by faulty components and to make decisions about the bearing health status [1], [2], [3], [4], [5].

If a localized defect occurs in the surface of bearing, an impulse of short duration will be generated, and this excites resonance of the bearing and other components in the machine. The impulses will be periodically produced with the rotation of the bearings, and the frequency is determined by the defect location, i.e., the inner race, outer race, or the rolling element. Hence, it is possible to identify the defect type by extracting the impulsive features in the vibration signals [6], [7], [8]. The envelope analysis (EA) technique is the most commonly used approach for detecting bearing defects over the last few decades [9], [10], [11], [12], [13]. However, a critical limitation of this technique is that it requires prior knowledge of the filtering band. Due to this limitation, detecting machine defects at the incipient stage when defect-characteristic components are weak in amplitude and without a distinctive spectral pattern poses a challenge to the conventional enveloping spectral analysis technique [8]. In order to overcome this limitation, the wavelet transform (WT) was incorporated with the conventional envelope analysis in recent years. The wavelet-based envelope analysis technique enables simultaneous multi-scale decomposition to extract and separate envelopes of the repetitively excited mechanical vibrations with different frequency coverage, thus improving the robustness in signal analysis [8], [10], [11], [14].

In this investigation, we present an alternative signal processing scheme based on mathematical morphology (MM) theory for rolling element bearing fault detection. MM theory is a relatively new nonlinear signal processing technique which is based on set theory [15], [16], [17], [18]. It has become an efficient tool for various aspects of signal processing [17], [19], [20], [21], [22] in recent years. Increasing attention has been focused on applying MM theory for detecting machinery faults in recent years because it has provided an alternative means of extracting impulsive signals purely based on time-domain analysis [18], [23], [24].

In this work, a novel morphological signal processing tool, named the weighted multi-scale morphological gradient filter (WMMG), is proposed to analyze the bearing vibration signals. It has been reported that the morphological gradient filter (MG) is capable of enhancing the impulsive features in the signal [25], [26]. An MG with large-scale structure element (SE) can depress the noise effectively but blur the impulse details, while one with a small-scale SE can preserve the impulse details but fail to suppress the noise [27], [28]. The motivation under the WMMG is to take advantage of MG with both a small-scale SE and a large-scale SE to obtain satisfactory impulsive feature extraction results. We evaluate and compare our new method on a simulated signal and on vibration signals measured from bearing test rig with the traditional EA and WT-EA approaches. The computation cost of the WMMG is also investigated. Experimental results have demonstrated the presented WMMG to be an effective approach for detecting rolling element bearing defects. We also explore the classification performance of three bearing states using a WMMG as a preprocessor. The effects of noise level on the classification accuracy of the WMMG and WT-EA are also investigated. Fig. 1 gives the flow chart of this work.

The remainder of this paper is organized as follows. In Section 2, the basic theory of mathematical morphology (MM) is simply introduced. Then the presented WMMG scheme is detailed. Section 3 presents an experiment applying the presented WMMG scheme to analyze a simulated signal. In Section 4, the vibration signals acquired from rolling element bearings are employed to evaluate the WMMG technique. The conclusions of this investigation are summarized in Section 5.