Standard bearing fault detection features are shown to be ineffective for estimating bearings remaining useful life. Addressing this issue, in this paper we propose an approach for bearing fault prognostics based on features describing the statistical complexity of the envelope of the generated vibrations and a set of Gaussian process models. The proposed features are sufficiently sensitive to the changes in the bearing condition and in the same time are sufficiently robust to variations in the operating conditions. Gaussian process models are non-parametric black-box models which differ from most other frequently used black-box identification approaches as they do not try to approximate the modeled system by fitting the parameters of the selected basis functions, but rather search for the relationships among measured data. In this paper the GP models are used for filtering noisy features and estimating the RUL based on filtered features.
COBISS.SI-ID: 25958695
It is well known that vibration signatures obtained from mechanical drives are generally not independent of the operating speed and torque. In such a case a change in the signature cannot be unambiguously associated with the fault. In this paper we present an approach to the detection and diagnosis of faults in mechanical drives under variable operating conditions. The idea is to use a black-box model to describe the contribution of operating conditions on vibration spectrum. The model then extracts fraction of vibration signature that depends only on the true faults, by compensating effects due to changes in operating conditions. The proposed approach is tested on vibration spectral features from gear pitting experiment in a single-stage gearbox.
COBISS.SI-ID: 26098215
In the paper we propose a novel approach to the diagnosis of gearboxes in presumably non-stationary and unknown operating conditions. The approach makes use of information indices based on Rényi entropy derived from coefficients of the wavelet packet transform of measured vibration records. These indices quantify some statistical properties of instantaneous power of the generated vibration that are largely unaffected by changes in the operating conditions. The analysis is based on probability density of the envelope of a sum of sinusoidal signals with random amplitude and phase. Such an approach requires no a priori information about the operating conditions and no prior data describing physical characteristics of the monitored drive. The fault detection capabilities of the proposed feature set are demonstrated on a two-stage gearbox operating under different rotational speeds and loads with various seeded mechanical faults.
COBISS.SI-ID: 25765159