This paper investigates how Gaussian mixture models (GMM) may be used to detect and trend fault induced vibration signal irregularities, such as those which might be indicative of the onset of gear damage. The negative log likelihood (NLL) of signal segments are computed and used as measure of the extent to which a signal segment deviates from a reference density distribution which represents the healthy gearbox. The NLL discrepancy signal is subsequently synchronous averaged so that an intuitive, yet sensitive and robust, representation may be obtained which offers insight into the nature and extent to which a gear is damaged. The methodology is applicable to non-linear, non-stationary machine response signals.
Condition-based maintenance is concerned with the collection and interpretation of data to support maintenance decisions. The non-intrusive nature of vibration data enables the monitoring of enclosed systems such as gearboxes. It remains a significant challenge to analyze vibration data that are generated under fluctuating operating conditions. This is especially true for situations where relatively little prior knowledge regarding the specific gearbox is available. It is therefore investigated how an adaptive time series model, which is based on Bayesian model selection, may be used to remove the non fault related components in the structural response of a gear assembly to obtain a residual signal which is robust to fluctuating operating conditions. A statistical framework is subsequently proposed which may be used to interpret the structure of the residual signal in order to facilitate an intuitive understanding of the condition of the gear system. The proposed methodology is investigated on both simulated and experimental data from a single stage gearbox.
This paper proposes a novel framework for monitoring the condition of a rotating machine (for example a gearbox or a bearing) that may be subject to load and speed fluctuations. The methodology is especially relevant in situations where no (or only noisy) shaft angular position measurements are available. Shaft angular position reference measurements are often not available due to physical constraints that render it difficult to install tachometers or encoders on the shaft of interest. The proposed methodology aims to simplify the task of monitoring a time-varying vibration signal by using a neural network to filter out the normal vibration components that generally tend to dominate the signal. The neural network may be optimised without the need for extensive datasets that are representative of different machine fault conditions. The envelope of the filtered signal is referred to as a discrepancy transform, since the discrepancy signal indicates the presence of fault-induced signal distortions. The discrepancy signal tends to be significantly simpler (smoother) than the original vibration waveform and may thus be resampled using a less accurate reference signal than would be required to resample the original waveform. A numerical gear model is used to illustrate the diagnostic potential of the proposed methodology.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.