Selftuning sensorless controlled drives

Controlled induction machine drives at present are considered the working horses of modern industrial applications because of their dynamic properties, robustness, and costs. With the stringent demand to steadily reduce the costs of all industrial components research has focused in the past years on the development of control methods working without speed/position sensor. As turned out from previous research, it is so far only possible to realize speed-sensorless control in the whole operating range including zero flux frequency if the high frequency or transient electrical behavior of the machine is considered. By establishing such a transient excitation in addition to the fundamental wave voltage the transient response of the machine current can be obtained and evaluated. It is modulated due to the inherent saliencies of induction machines caused for example by saturation or slotting. Using appropriate signal processing it is thus possible to estimate the flux and rotor position necessary for the control of the machine. However, all such methods are still limited to laboratory use and are not applied industrially. In the proposed project two of the main problems still existing with zero speed sensorless control will be addressed. One of these main problems is denoted tuning of the control parameters. The design of the machine, the inverter, as well as the kind of excitation has strong influence on the resulting control signal obtained. Thus usually a high number of parameters have to be adjusted in order to set up speed-sensorless control on a new type of induction machine drive. This tuning process up to now is a time consuming task and needs an expert in speed sensorless control to be performed. In the proposed project neural networks together with classical optimization methods and observer approaches will be applied. The final sensorless control is a combination of a fundamental wave method and signal injection method. The goal is to tune the parameters of both methods by an algorithm only supervised by a commissioning engineer using no special test equipment. The second main problem is denoted interference of the different saliencies modulations. Generally there are always flux fixed and rotor fixed saliencies present in standard induction machines. In specific operating ranges the frequencies of those saliencies overlap making it extremely difficult to separate them correctly. This topic will be addressed by applying blind signal separation techniques together with an observer approach. In addition, that one of the two rotor fixed saliencies (slotting, anisotropy) not overlapping in frequency with the saturation saliency will be used to obtain additional information in these critical operating states. In addition, using the developed self tuning and signal separation method the drive should be able to separately track and exploit the dominant stator and rotor fixed saliencies present in standard induction machines and therefore enable speed-sensorless torque as well as speed/position control.

In the past decades the application of inverter-fed drives constantly increased. In comparison to mains-fed electrical machines inverter-fed operation offers the possibility of adjustable speed and high dynamic. Due to its robustness and relatively low price induction machines are the most common type. To control these machines in the whole speed and torque range reliable information of the rotor and/or flux position is needed. In early years of induction machine control this flux information was retrieved from additional sensors attached to the machine. However, these sensors and especially the additional cabling add a further source for possible breakdown of the whole drive system. Nowadays mathematical machine models are used to estimate the flux information needed for induction machine control resulting in increased drive reliability. These mathematical models fail in certain operating points without an additional rotor speed sensor. Though for this approach a mechanical component that increases the drive`s cost and size and decreases reliability has to be added also. Furthermore it is necessary to accurately know the machine`s electrical parameters to ensure proper control of the induction machine. The task of this project was to develop a control strategy that allows operating the inverter-fed machine based on sensors built in the inverter only, without using additional sensors. The application of special test voltages to the machine terminals and the measurement of the current reaction permits to calculate a signal that contains the information on the current rotor and/or flux position. The only sensors needed in this estimation procedure are the current sensors already available in modern drive systems. As the design of the machine and the inverter has strong influence on the control signal obtained, usually a high number of parameters have to be adjusted in order to set up sensorless control on a new type of induction machine drive. This tuning process is a time consuming task and needs an experienced engineer to be performed. In order to make a further step towards a practical application the goal of this project was to develop a method able to determine the necessary parameters autonomously. The tuning can now be done without the need of an experienced engineer being present and even without using special test and measurement equipment. For certain operating conditions the control signal contains disturbing components. Thus the second main task of the project was to separate these components from the information needed for machine control. The project results are very encouraging and a big step towards industrial applicability of sensorless control.