Gear fault detection and diagnosis under speed-up condition based on order cepstrum and radial basis function neural network

Abstract

Varying speed machinery condition detection and fault diagnosis are more difficult due to non-stationary machine dynamics and vibration. Therefore, most conventional signal processing methods based on time invariant carried out in constant time interval are frequently unable to provide meaningful results. In this paper, a study is presented to apply order cepstrum and radial basis function (RBF) artificial neural network (ANN) for gear fault detection during speedup process. This method combines computed order tracking, cepstrum analysis with ANN. First, the vibration signal during speed-up process of the gearbox is sampled at constant time increments and then is re-sampled at constant angle increments. Second, the re-sampled signals are processed by cepstrum analysis. The order cepstrum with normal, wear and crack fault are processed for feature extracting. In the end, the extracted features are used as inputs to RBF for recognition. The RBF is trained with a subset of the experimental data for known machine conditions. The ANN is tested by using the remaining set of data. The procedure is illustrated with the experimental vibration data of a gearbox. The results show the effectiveness of order cepstrum and RBF in detection and diagnosis of the gear condition.

Fault diagnosis in spur gears based on genetic algorithm and random forest

“…This binary partition is repeated until the best regions are reached, in order to approximate the model output in Eq. (6). The stop criterion could be the number of nodes, or some classification cost.…”

“…In recent years, several analysis techniques for gears faults diagnosis have used Wavelet Packet Transform (WPT), in order to enhance the information that is provided by the classical statistical parameters from the vibration signal, in time and frequency domains [2][3][4]. In case of machine learning-based diagnosis, the most common approaches have been developed by using Neural Networks [5,6], Support Vectors Machines [7], Cluster Analysis [8] and Genetic Algorithms [9,10]. These approaches have been very useful for implementing conditionbased maintenance (CBM), as presented in Jardine et al [11].…”

Fault Tolerant Control for Non-Gaussian Stochastic Distribution Systems

“…Based on the system (6), the nonlinear dynamic model can be transformed into the following equivalent form: (11) where…”

Fault Detection and Diagnosis for Non-Gaussian Singular Stochastic Distribution Systems via Output PDFs

“…For stochastic system,the two kinds of approaches include the system identification techniques [6] and the statistic approaches based on the Likelihood methods, Bayesian theorem,and Hypothesis test techniques ( [1]) can be used to deal with the related FDD problems. Besides, we have known that observers or filters have been extensively applied to generate the residual signal for fault detection and diagnosis( [7], [8]), and many significant approaches of them have been applied to practical processes successfully( [9], [10]). …”

“…However, nonlinearity may lead to non-Gaussian output,where (especially for asymmetric distributions with multiple peaks) mean and variance of the system output are insufficient to characterize their statistical behavior precisely ( [11], [12], [13]). As such, there is need to further develop fault detection and diagnosis algorithms that can be applied to the stochastic system subjected to random parameter.On the other hand,along with the development of advance instruments and data processing technique, the measurements for feedback are the stochastic information which can be described by the probability density functions(PDFs) of the stochastic distribution system output rather than the actual output values.For such non-Gaussian stochastic system, we call stochastic distribution control(SDC) systems ( [3], [7], [10]- [16]). Different from any other previous stochastic control approaches,the stochastic variables are not confined to be Gaussian and the output PDFs of the stochastic system is concerned( [13]- [18]).…”