This work deals with the performance improvement study of the direct torque control (DTC) of a Double Star Permanent Magnet Synchronous Machine (DSPMSM) based on Second Order Sliding Mode Control (SOSMC), powered by two voltage source inverters. DTC control using conventional PI regulators has certain disadvantages such as significant flux, torque ripples and sensitivity to parametric variations. To overcome these drawbacks, we apply a new type with more robust regulators such as the second order sliding mode control. Simulation results demonstrate the feasibility and validity of the proposed DTC-SOSMC system by effectively accelerating system response, reducing torque and flux ripple and a very satisfactory performance has been achieved.

This paper presents a study of a surface crack detection in which the volume is filled by conductive substances due to the polluting environment. Hence, this investigation demonstrates by numerical simulation that electric conductivity is a crucial property that has to be added to the other defect geometrical characteristics in order to complete the developed models. Consequently, introducing the tolerance in percent in the measured impedance is necessary in some conditions. So, the obtained results demonstrate that the signal amplitude passes from 0 to 78% of the maximal amplitude when the defect conductivity rises from 0 to 0.5 Ms/m. On the other hand, the relative difference of the resistance partincreases according to defect volume. For example, for a defect of 0.3 MS/m, the relative difference of the resistance varies from 52 to 62% of the maximal amplitude when the defect depth varies from 0.5 to 2.25 mm. These results can be exploited to show the effect of the conductive substances occupying the crack volume. In fact, the controller using EC-NDT technique must take into consideration the presence of conductive polluting elements in the crack volume. So, this condition becomes primordial and necessary according to the degree and nature of pollution.

Nowadays, numerical modelling has become an interesting tool for determining impedance variations due to various conductive flaws in eddy current nondestructive evaluation systems. These kinds of defects, rarely treated in the published works, are taken into consideration in the modelling while introducing them as electrically conductive volumes with a finite electric resistivity. This step is very important since it permits to improve qualitatively several models developed so far by many authors whose consider the defect as loss of material only. However, in several applications, the defect can occur with a finite resistivity such as impurity, small burns and micro-solder. On the other hand, even though the defect appears with a loss of materials, som e polluting materials can fill the affected region. Indeed, the volume of the initial defect will be completely or partially occupied by these conducting pollutant materials. This paper deals with the effect of physical and geometrical characteristics of such kind of defects on the differential sensor response. Furthermore, the necessity of taking the defect electric conductivity (as an important parameter) into account will be explained, in order to develop a reliable and accurate inverse method allowing a full characterization of conductive defects.

This study will focus on the investigation of the effect of electrical discharge on physical, chemical, electrical properties of transformer oil, and on the development of a mathematical model describing the gassing of insulating oil under electrical discharge, using the information contained in the measured values. For predicting the gassing tendency for extensive ageing periods, we use the model developed, for an intelligent system design. The predictor's parameters are chosen based on their influence degree by the electrical field. Various scenarios were considered. The study was carried on two types of fluids, under electrical stress for different ages. The 6802, 6181 and 924 ASTM tests methods were used for the measurements of parameters in degradation. All the results obtained are summarized and compared. The properties which are strongly dependent have been specified, a multiple linear regression model for each fluid as a function of its DDP, DDF, turbidity and aging period is developed. This model is for the estimation of the gas quantity cumulated under electrical discharge. The prediction is made, by implanting the stepwise regression results into a neural network system, which has been tested on experimental results obtained from laboratory samples, and high prediction accuracy has been achieved.