A new method of gear fault diagnosis in strong noise based on multi-sensor information fusion

Abstract

In the work process of mining machinery gears, vibration signals are not only influenced by friction, nonlinear stiffness and non-stationary load, but also influenced by strong noise. How to extract the fault feature information effectively, identify the fault status accurately and eliminate the uncertainty in the identification process is the key to evaluate the fault status in strong noise. A new gear fault diagnosis method in strong noise is proposed based on multi-sensor information fusion, which combines wavelet correlation feature scale entropy (WCFSE), self-organizing feature map (SOM) neural network and Dempster-Shafer (D-S) evidence theory algorithm. The noise is reduced by the way of a wavelet transform correlation filter, to calculate the Shannon entropy of denoised wavelet coefficients, which can reflect the vibration signal complexity. The WCFSE of standard training samples defined as the input vectors of an SOM neural network is used to train the neural network, and the gear status clusters in a competitive layer. In order to improve the accuracy and completeness in the identification process, multi-sensor fusion technology is introduced to establish the recognition framework of D-S evidence theory and the basic belief function allocation method based on the recognition rate of statistical SOM neural network. Each sensor can provide sub-evidence, and the gear fault diagnosis is analyzed according to the combination rules and basic belief function. The experimental results show that the proposed gear diagnosis method in strong noise can identify the gear fault accurately, eliminate the identified uncertainty and imperfection with a single sensor. The recognition rates of two-sensor fusion beyond 80%, and the recognition rates of three-sensor fusion beyond 88%, especially the status with tooth loss beyond 95%. The fault recognition rates have been greatly improved compared with a single sensor, so this is an effective method of gear fault diagnosis in strong noise.

Multi-sensor data fusion based on the belief divergence measure of evidences and the belief entropy

“…With the fast development of electronic technologies, a variety of sensors were developed and applied in many engineering fields, like the fault diagnosis [1][2][3][4], wireless sensor networks [5][6][7], risk analysis [8], and so on [9][10][11]. Because multisensor-based applications can provide more reliable and accurate information than that of a single sensor alone, multi-sensor data fusion technologies have attracted considerable attentions in many fields of practical applications for the past few years [12][13][14].…”

Study on a novel fault diagnosis method based on information fusion method

“…Xu et al [26] proposed a sensory information fusion-based diagnostic methodology and framework. Cheng et al [27] proposed a new gear fault diagnosis method by using multi-sensor information fusion based on combining wavelet correlation feature scale entropy, self-organizing feature map neural network and D-S in strong noise. Xiong et al [28] proposed an information fusion fault diagnosis method based on a static discounting factor, K-nearest neighbors and dimensionless indicators.…”

A New Method of Two-stage Planetary Gearbox Fault Detection Based on Multi-Sensor Information Fusion

“…This level of data fusion contains the most direct and reliable information of the measured object and can provide the most accurate troubleshooting results [26]. However, in the research on fault monitoring and detection of rotating machinery based on multiple sensors, data-level fusion methods mostly rely on artificial intelligence methods such as neural networks and support vector machines [27][28][29], and there are few research works on multi-sensor information fusion in the field of data layers [30,31].…”