Conductive specimens such as aluminum plates are tested in order to extract information about possible cracks, flaws and other mechanical damages. Nowadays, eddy current testing (ECT) despite its major benefits (e.g. low cost, high checking speed, robustness and high sensitivity to large classes of defects) implies the utilization of fully coil based architecture probes or hybrid coil-magnetoresistive probes. This work presents an eddy-current testing system based on a giant magnetoresistive sensing device. The application detects and estimates the size of cracks in an aluminum plate specimen. A neural network processing architecture is used to find out the correspondence between the cracks and the signal characteristics measured on the eddy current probe. The crack detection and the estimation of its size using different eddy-current frequencies are described in the paper.

The Indian power system is a large network with an installed capacity of 140 GW. It is unique in that it is the only network which allows the operating frequency to vary over a band from 0.5 Hz to -1.0 Hz from the nominal 50 Hz. The pricing for power purchased is a function of the operating frequency following a commercial mechanism called availability based tariff. Since the network consists of several states which are each commercially independent, an elaborate metering system is required at the interconnecting 400 kV grid. Since 2004, 461 special energy meters with 0.2 class accuracy have been erected across the network. They provide online information about key network parameters in addition providing the data for commercial transactions. We show several examples of the utility of this measuring system with regard to scheduling plants in the network, planning and control studies. We propose a grid transaction index that would help assess the state of the network and the performance of various state’s dispatchers. All discussions are based on the data of Tamil Nadu Electricity Board (TNEB), a utility in the Southern region.

The characteristics of actual converters, including sensors, are – to smaller or larger extent – nonlinearity. With respect to the mathematical model describing the characteristics, they can be characteristics with smooth nonlinearity or nonlinearity with discontinuities. The author presented the problem of linearization of the characteristic of converter with nonlinearity with discontinuity that appears in the small value area, i.e. with insensitivity area. In general, this is not an easy task because of the type of nonlinearity and the local character of linearization. A method utilizing the technique of added noise (dither) is applicable in this purpose. It effectively removes nonlinearities in output signal, caused by the area, i.e. removes nonlinear distortions of the signal processed. Using dither makes it also possible to process signals with values wholly contained in the area, which would be impossible in a converter without applying the method. Side effects of applying the method are: amplitude reduction and phase shift of the processed signal, because of amplification error and phase error of the AV/LPF system, respectively. If we apply averaging it extends the time of processing.

The performance of current devices is mostly limited by the analogue front-end and analogue-to-digital converter’s (ADC) imperfections. ADC performance is not, as commonly known, ideal. An important parameter is the nonlinearity, which (if it is known) can be corrected in the output data. The performance of three different approximations of ADC nonlinearity (common polynomials, Chebyshev polynomials and Fourier series), the approximations accuracy, and their coefficients’ sensitivity to noise is analyzed in the paper. Also an application of the approximated ADC nonlinearity for its correction is mentioned. The accuracy evaluation showed comparable results for small orders of all examined approximations. For higher number of estimated coefficients (few hundreds) the method using common polynomials fails while results using Chebyshev polynomials and Fourier series are similar. The noise sensitivity evaluation showed that an increase of maximum error is similar for all approximations in the case of the low number of coefficients. A rise of the number of coefficients escalates noise sensitivity in all examined cases. The nonlinearity correction was tested using the proposed approximations as well. This possibility was verified and in the presented case the improvement in the output signal spectrum achieved about 30 dB.

Some damages of mechanical parts such as bearings, rotor, shaft or the damages in the elements of electromagnetic system, for example in the wiring of stator, rotor or of the magnetic circuit can appear in induction motors. Statistics say that bearings are this part of induction motors which is most susceptible to damage. That is why the issues connected with diagnostics of bearings are so important. This paper shows diagnostic methods which employ the acceleration of vibration measurements. The appearance of a specific kind of damage in the motor is a source of additional forces whose character is oscillatory and which can be noticed in the damaged area. These forces result in vibrations which form the basis for diagnostic research. When the course of these vibrations is subjected to spectral analysis, it can be obtain a number of components related to specific types of damages. A scanning laser vibrometer PSV-400 Polytec has been employed for vibration measurements. The research has been conducted on a number of artificially damaged bearings. This paper presents exemplary results of testing damaged bearings with the usage of laser vibrometer PSV-400. The research of technical condition of the same bearings was simultaneously carried out with DREAM – a vibration diagnostics system.