Andrzej Puchalski

Stable Distributions and Fractal Diagnostic Models of Vibration Signals of Rotating Systems

The main objective of the presented work is improving the knowledge of non-linear effects in vibrodiagnostics of rotating machinery components using multifractal analysis. The modeling is based on the data-driven approach, in which probabilistic models can be constructed to capture the structure of the time series. These models allow the identification of deviations from the training data during the monitoring of the system. The paper proposes the feature vector, being the data-driven empirical diagnostic model, of mechanical vibrations in the vehicle powertrain. Dynamic analysis of wearing and/or defects of complex rotating systems confirms the presence of non-linear, non-stationary and multiscale properties. The recorded time series exhibits the impulse-like nature and fluctuations. Probability distributions often deviate from the Gaussian distribution and exhibit heavy tails. Both broad density functions, typical for stable distributions, and long-term correlations are two important sources of multifractality. Analyses of fluctuations of the recorded vibration time series were carried out utilising the dependence between selected parameters of alpha-stable distributions and measures of singularity spectra.

Online Fault Diagnosis of Automotive Powernets by Kalman Filtering

The paper discusses real-time parameters identification methods of the mechatronics subsystems in the automobile for the on-board diagnostics and adaptive control. In particular, the significant increase in electric power consumption requires predictive management strategy that detects critical loads situations. To accomplish these goals we investigated diverse vehicular powernet circuits and devices. This paper presents the results of the extended Kalman filtering algorithm for the parameters estimation of the wiring harness of a cooling fan system.