Gear damage diagnosis and classification based on support vector machines

Abstract

The operation status of gear device can directly affect the working conditions of the whole machine system. Thus, it is crucial to detect the gear damage as early as possible to prevent the system from malfunction. This paper proposes an intelligent diagnosis method for gear damage using multiple classifiers of support vector machines with extracted failure feature vector. The vibration signal of gear box is employed as the analytical data in this paper. In order to illustrate the representative characters of gear conditions, statistical parameters and characteristic amplitude ratios of frequency bands are extracted from the vibration signals in time-domain and frequency-domain respectively, which are served as failure feature vector for the following diagnosis. Moreover, to reduce the dimensions of the failure feature vector, the technology of principal component analysis is adopted to transform the original failure feature vector into a new smaller set of variables as inputs to classifiers of support vector machines. In order to classify different types of gears, multiple classifiers of support vector machines based on the binary tree are designed. The validity of this approach is investigated by the experiment. Three kinds of gears, namely normal gear, spot damaged gear and pitted gear, are tested on the power circulating type gear testing machine. The vibration accelerations of gear box are measured as original data. Most of the samples are correctly classified by the provided method, which demonstrates the effectiveness of the proposed method on the application of gear damage diagnosis. IntroductionSince minor gear damage may cause serious failures of the entire equipment, early detection of gear damage is one of the important measures to prevent the machine system from malfunction. Analyzing the vibration signal adopted from the gear or gear box is one of the effective methods to diagnose gear failures. Researchers have done countless studies in this respect and have developed many methods based on the analysis of vibration signal in time domain, frequency domain and time-frequency domain. In frequency domain, it is well known that the fault condition of gears can be observed at the meshing frequency and its harmonics, together with sidebands by spectra analysis of vibration signal (Dalpiaz, et al., 2000). However, the acquired signals are always inevitably interfered by the vibration of other components in the system, or the environment disturbances. Therefore, in order to strengthen the characteristics of the useful signal, researchers have developed many other signal processing techniques for gear fault detection, such as time synchronous average method in time domain, wavelet transform and Hilber transform techniques in time-frequency domain, advanced statistical approaches and so on. These techniques have been satisfactorily applied to both fault detection and identification of the damaged gear (Houjoh, et al., 2007, Tanaka, et al., 2012, Wang, et al., 2010 Fan, Ikejo, Nagamura, Kawada and Has...

A Review of Early Fault Diagnosis Approaches and Their Applications in Rotating Machinery

“…Shen et al [158] Statistical feature + SVM Liu et al [156] Impact time frequency dictionary + SVM Fernández-Francos et al [157] Band-pass filters and Hilbert Transform + ν-SVM Zhao et al [160] EEMD + multi-scale fuzzy entropy + SVM Tabrizi et al [162] WPD + EEMD + SVM Wu et al [163] Continuous wavelet transform+ SVM Fan et al [155] Statistical parameters + PCA + SVM Kang et al [165] Singular value decomposition+ SVM Konar et al [164] CWT + GA + SVM Saidi et al [159] Spectral kurtosis + SVM…”

“…The SVM-based fault diagnosis is very similar to KNN-based methods. Fan et al [155] used statistical parameters and characteristic amplitude ratios of frequency bands as fault features, then the PCA was used to reduce the fault feature dimensions. The selected features were put into SVM for gear fault type identification.…”

Diagnosis for gear tooth surface damage by empirical mode decomposition in cyclic fatigue test

“…In our previous study (Fan, et al, 2014), a diagnostic method for gear damage using SVMs with extracting amplitude ratios of frequency bands from vibration accelerations as failure feature vectors was provided. To validate the utility of the proposed method, we also diagnosed the conditions of test driving gear using the previous method.…”

“…Since it is difficult to obtain sufficient samples in practice, SVMs are introduced for gear fault diagnosis and classification in this paper. In our previous study (Fan, et al, 2014), a diagnostic method for gear damage using SVMs was proposed, in which amplitude ratios of frequency bands and statistical parameters were extracted as failure feature parameters. In this study, a diagnostic method based on techniques EMD and SVMs is proposed to monitor and diagnose gear conditions.…”

Health monitoring and diagnosis system for heavy roll grinding machine

“…P Parenti et al 10 developed an online waviness identification scheme for cylindrical traverse grinding from acceleration measurements. Q Fan et al 11 proposed an intelligent diagnosis method for gear damage using multiple classifiers of support vector machines with extracted failure feature vector in which the vibration accelerations of gear box are measured as original data. JW Fan et al 12 adopted small sample technology which needs less data while getting higher evaluation accuracy; using the theory of Bayes theorem and failure rate as random variables, the experiment 14 built a semi-supervised learning model for fault detection and classification for a complex process.…”