Fourier spectrum related properties of vibration signals in accelerated motor aging applicable for age determination

Question of condition monitoring and predictive maintenance is an important one in modern industry. In case of electrical motors, the question of their bearings’ health and remaining useful life of the whole device can be asked. In order to diagnose abnormal states and faults, a wide range of indicators have been used, including stator current, vibration and sound [7]. In this study, we will focus on the properties of vibration signals and ability to do the motor diagnostics based on vibration content. Such approach is widely used in practice [10] with various parameters extracted from the vibration signal. While some authors choose to base their methods of vibration diagnostics on the features of signal in time domain [12], the others use the features in frequency domain [17], which is an attractive area of research and being often considered invariant opposed to amplitude variations in time domain analysis, influenced by the motor type or the sensory equipment employed in the experiment. In case of random vibrations, the classical power spectral density (PSD) is serving just as a foundation for more complex methods of frequency based analysis [16]. Often, in case of motor vibrations, decomposition is applied, for instance the Hilbert Huang Transform or Wavelet Transform, with a possibility of applying frequency analysis to components obtained that way [2], or simply taking time domain characteristics of such components [1]. In this work, we have chosen to work with the whole signal, without any decomposition, avoiding analysis of components’ significance and noise level within each of them. While these works often focus on known bearing faults, showing methods of detecting and identifying them, the practical situation often resembles the one seen in case of artificial motor aging: motor bearings, shaft and windings suffering the unpredicted failures and faults [4, 5]. In that case, it is important to be able to determine what the level of aging the motor is achieved, so the remaining life can be estimated [12]. An important question that arises there is what is the difference between ordinary isolated, controlled bearing faults and those emerging in (artificial) motor aging processes. Hence, this study aims to fill the gap in knowledge about the nature of bearing faults occurring in artificial aging processes of induction motors, as well as to investigate on the frequency spectrum properties of vibration produced in motor aging applicable for determination of motor age. Conclusions are given to support the claim that the aging process is not possible to approximate with a single type of bearing fault, but that characteristics common for certain types of faults appear within the frequency signature of motor vibration in characteristic aging signals. Furthermore, simple statistical measures that are selected for demonstration of emerging patterns in motor aging can readily be used for detection of motor age, without the need for introduction of more complex machine learning and intelligence components. Simplicity of patterns also indicates an existence of a logical theoretical basis for the results obtained. The paper is structured as follows: after an introduction, frequency-based methods and experimental data from motors used in this study are presented in the second part. Third part presents the results Harun SiljAk Abdulhamit SubASi