JOURNAL OF FACULTY OF ENGINEERING (UNIVERSITY OF TABRIZ)
Year:2008 | Volume:36 | Issue:1 (53) ELECTRICAL ENGINEERING|Papers Count:6

This paper investigates the complete calculations of the different losses in matrix converters such as switching and conduction losses. A small R-C snubber circuit is usually connected across two terminals of a bidirectional switch to limit the device voltage at an appropriate level. Also, the complete analysis of the snubber circuit losses is given. The results show that this simple snubber arrangement causes the total converter losses to increase. In other words, it has been found that the snubber circuit results in a high power loss compared to the total semiconductor losses. Another result of this research is that the snubber losses in the matrix converter are independent of the input and output frequencies and the power factors. The snubber losses mainly depend on the input voltage, the snubber capacitance and the switching frequency. Also, in this paper it is shown that the total energy losses is independent of the switching pattern and the load power factor and the majority of the turn-on energy losses comes from the snubber capacitor. Very good correlation between the analysis and the measured results were obtained.

In this paper, a new algorithm is proposed through the clustering-based modeling and genetic algorithms for color-defects detection of ceramic tiles. The algorithm consists of two stages: feature extraction and inspection. In the former phase, the parameters of the algorithms are regulated by using one (some) reference defect-free ceramic - tile image(s) in order to produce the contextual model of colors. Then, during the inspection phase, the colors of the examined ceramic tiles images are compared to the color model; and each difference indicates a color defect. The proposed algorithm integrates a new color clustering algorithm based on min-max criterion and genetic algorithms such that the former is used by the later as the evaluation function. The experimental results on a database including 120 ceramic tiles images categorized to six groups demonstrated acceptable performance in terms of solution quality, computational burden, and adjustable sensitivity.

The performance of static VAr compensator (SVC) for flicker mitigation can be improved with prediction of reactive power for half cycle ahead. For this purpose it is necessary to obtain the coefficients of prediction equation to minimize prediction error. The proper performance had been achieved for most cases with offline calculating of these coefficients. But for example the frequency of arc furnace reactive power variation is not constant due to the nature of load and system condition. In this paper an adaptive genetic algorithm for online calculation of prediction coefficients is suggested. The performance of new algorithm has been proved with using it in prediction of different built up signals and also applying it in the SVC control system of Khuzestan steel plant. The real arc furnace system has been simulated with PSCAD/EMTDC software and the online genetic algorithm is written as MATLAB function that is linked to main system.

In this paper a novel fractal shape structure is introduced. Application of the proposed fractal shape to a square patch antenna represents a very good reduction size in comparison to the previously reported works. To compensate for the reduction in the antenna bandwidth, which occurs during the miniaturization process, the T-probe feeding method is used. By using this feeding method, the impedance bandwidth of the antenna is increased, which makes the antenna appropriate for PCS applications in the frequency range of GSM 1900. The effects of various parameters of the feed structure on the input impedance of the antenna are also presented in this paper. In order to validate the simulation results, the antenna was fabricated. The results of testing and measurements of the fabricated antenna in the antenna Laboratory of Iran Telecommunication Research Center (ITRC) verified the simulation results.

This paper presents a new fault location method based on synchronized phasor measurement for two–terminal transmission lines. Due to some common problems of current transformers (CT) in power system protection such as low accuracy, this method wipes off the CT requirement. In other word, the method is only based on voltage measurement. Samples of voltage at both ends of line are taken synchronously and used to calculate the location of fault. The algorithm uses thevenin equivalent model of faulted system and symmetrical components theory for estimation of fault distance with reasonable error (0.7%). The proposed method is also independent of fault type, fault resistance, fault inception angle and line angle. In addition to these advantages, the proposed technique is improved by using modified algorithm in order to reduce the estimation error to less than one meter. Simulations with EMTP/ATP were performed to verify the accuracy of the proposed method.

Multicarrier methods have long been identified for optimizing data transceivers on band limited channels. Discrete Multitone is a multicarrier transmission method which can be implemented using DSP techniques. Transmission rate in a simple DMT system can be increased by adding a full duplex echo canceller. So far, there have been several problems with the proposed echo cancellers regarding their high computational complexity and the relatively low stability. In this paper, a new high speed technique based on neural network with a good stability is represented. By using of this method, the echo of DMT modulation in ADSL system is cancelled. This technique optimizes the learning factor and consequently increases the echo cancellation rate in comparison with LMS method in about 2.37dB. In this method, only the operational phase of neural network is applied to DMT system. After training network, the weights are frozen and the computational cost is reduced to an acceptable level.