Induction machine has been widely used in industrial drive, due to its simple structure, high reliability, and rigidity. In recent years, its application expands to electric vehicles and servos with the development of inverters and control technologie...
Induction machine has been widely used in industrial drive, due to its simple structure, high reliability, and rigidity. In recent years, its application expands to electric vehicles and servos with the development of inverters and control technologies. In this regard, the induction machine's output characteristic is closely related to the control algorithm because the inverter and the vector control mostly drive the machine used for this purpose. Therefore, in the motor's design process based on finite element analysis (FEA), it is necessary to consider the control algorithm and its performance. However, it is not easy to test the controller directly with FEA because of the long computational time.
In this thesis, a new simulation model of an induction machine based on the FEA is proposed considering magnetic saturation, saliency, and spatial harmonics. Firstly, magnetostatic FEA was conducted in various current operating points, and flux linkage and torque were obtained. A flux map-based model was then introduced, which is fast and numerically stable without differential operation. Also, it has less burden on memory. To implement the proposed flux-based model, an inverse relation of current-flux is required. Finding the inverse relation of current-flux is equivalent to solve the inverse of a 4-dimensional vector function. Therefore, the existence of the inverse relation was proved, and it is implemented through the whitening transform and the artificial neural network. At learning the neural network, the accuracy was considered with the highest priority, and various learning methods were compared and chosen to reduce the learning time. Through the proposed model, various control algorithms can be tested fastly without conducting the FEA repetitively.
Meanwhile, the rotor bar's skin effect should be considered, which affects the bar's current density distribution. Thus, the magneto-transient FEA was conducted to analyze the variation of resistance and the rotor bar's inductance. The analysis confirms that the rotor bar's skin effect has a different aspect than the general round conductor because of its unique shape for the deep bar effect. Also, the increase of resistance was affected by the frequency and the saturation of the magnetic core. Therefore, an analysis method was proposed, which can consider the induction motor's saturation level in the vector control situation, and the flux-based simulation model was modified with the analysis.
The validity of the proposed simulation model was verified with two different induction machines. The simulation models of the two machines were implemented and compared with FEA and the experimental results. First of all, the no-load test was conducted in the model, FEA, and the experiment. The input voltage varies in the wide range, and it is confirmed that the model shows the magnetic saturation well by comparing the results. Then, the vector control was applied to the model and the experiment, and the results were compared. In the current control, the effect of spatial harmonics appears in heavy load conditions even if the rotor is skewed, and it is shown both in the model and the experiment. Also, the current trajectory in flux weakening control and the model's output torque characteristics are very similar to those of the experiment. At last, it is confirmed that the modified model considering the skin effect represents the saturation induced saliency in dynamic inductances well.