Switched-Mode Power Converter Systems Control and Electromagnetic Compatibility, Ph.D. Thesis by Dejan Kos, December 2007
Sensorless Control of Permanent Magnet Synchronous motor using rotor flux observer, Ph.D. Thesis by Evgen Urlep, December 2006
Matrix based approach to control of discrete event systems: a recursive description of logically controlled systems, Ph.D. Thesis by Ales Polic, December 2005
Low speed sensorless control of induction machine, Ph.D. Thesis by Gregor Edelbaher, 2004
Position Control of Nonlinear mechanisms by Using Fuzzy Logic, Ph.D. Thesis by Andreja Rojko, 2002
Robust Control of Elastic Servodrives, Ph.D. Thesis by Ales Hace, 2001
A Robust Torque Control of an Induction Motor in Servo Drive, Ph.D. Thesis by Alenka Hren, 2001
Efficiency Optimizing of Variable Motor Drive with Induction Motor, Ph.D. Thesis by Dusan Drevensek, 2001
Electromagnetic Compatibility Improvement of the Random Modulated Rectifier With Power Factor Correction, Ph.D. Thesis by Franc Mihalic, July 2000
Ac Motors as Servodrives in Mechatronics, Ph.D. Thesis by Miran Rodic, March 2000
Modelling and Control of Linear Induction Motor, Ph.D. Thesis by Ludvik Kumin, May 1998
Robot in Contact with Environment, Ph.D. Thesis by Suzana Uran, March 1998
Open Structure of the Robot Controller, Ph.D. Thesis by Martin Terbuc, April 1996
Variable Structure System Theory in a Control of Robot Drives , Ph.D. Thesis by Boris Curk, July 1995
Neural Network Control of Robotic Arm, Ph.D. Thesis by Riko Safaric, June 1994
Reliable Control of Electric Drives in Hard Real Time, Ph.D. Thesis by Branko Premzel, May 1994
To achieve spred spectrum of electrical signals and to improve electromagnetic compatibility of switched-mode power converter system, the randomization is introduced to the pulse-width modulation parameters and hysteresis band. Thesis also deals with influence of random signal on low-frequency content of electrical signals. For time-triggered and asynchronous switching functions, power converter output voltage is analyzed with respect to the transistor turn-on and turn-off delay. For basic elements of switch-mode power converter system, models are shown with corresponding parasitics components. Modeling and simulation is suitable method for transients analysis. Switching behavior of transistor is described in details where special attention is put into diode transistor commutation problem. Influence of gate driver parameters on switching delay and other characteristics intervals of transistor turn-on and turn-off sequence are also pointed out. Four-stage gate driver is proposed to achieve minimal switching delay and to reduce influence of diode reverse and forward recovery. Two reconfigurable modules are designed to enable implementation of switched-mode power converter control algorithms on FPGA: reconfigurable differential equation solver where resources are exploited sequentially and reconfigurable finite state machine engine. Described concepts are verified in simulations, experiments and with measurements of electromagnetic interferences.
Sensorless speed control of non-salient PMSM using a non-linear rotor flux observer in stationary coordinates is presented. The behaviour of the rotor flux observer, which uses measured stator voltage and current is analyized and improved esspecially in low speed range. The presented stator resistance adaptation algorithm reduces the rotor flux estimation error, when a load is applied to the machine. The additional current applied in d axis in combination with the estimated rotor flux angle correction enables the limitation of the estimated rotor flux error and improves the behaviour in low speed range including zero speed. The Hall sensors signals, combined with the rotor flux observer reduces the parameter variation sensitivity and enable low and zero speed operation. The additional simplified structure of the rotor flux observer without the transformation angle is featured. It uses the estimated rotor flux directly in the control algorithm. Low speed operation of the observer is improved by a correction signal to reject load disturbance influence. The integration principle of the stator voltage measurement with closed loop linearization of the power inverter is presented. The simplified method of the measurement of the switching elements delays enables the linearization of the power inverter without any additional parameters. The behaviour of the poposed sensorless control and the observer parameter variation sensitivity is shown by several simulations and validated by using the experimental system where presented measurement methods and linearizations are implemented.
This dissertation introduces a new, recursive matrix based approach for the modeling of the systems with inherently event-driven dynamics. Logical engineering systems are considered as discrete event systems. Academically developed approaches are selected as the research starting point and further developed in order to bring the discrete event system theory closer to the traditional control theory viewpoints on one hand and to meet the practically industrial standards on the other hand. Special attention is paid for adaption of new developed academic approach to the practical engineering design and implementation methods. Therefore, the proposed matrix based model of discrete event system enables the formalization of the empirical industrial solutions one hand and the introduction of the academically developed methods to the engineering applications on the other hand.
Induction motor speed sensorless control employing new rotor flux and speed observer, allowing operation at low and zero speed, is presented in this work. The rotor flux magnitude and the orientation angle are determined by the output of the closed-loop rotor flux observer based on the current loop and calculation of the extended electromotive force, which is derived from continuous sliding mode controller. The proposed rotor flux observer is robust to parameter variations so that also low stator frequency operation is permitted. Additionally is the observation problem at zero stator frequency solved with the use of frequency dependent correction factor. The performance of the proposed observer is investigated by simulations and verified experimentally on the digital signal processor.
In this work a control scheme, which belongs to the class of the control schemes known as a sliding mode control with disturbance estimation, is proposed. The scheme is derived by using the theory of the variable structure systems. Good disturbance estimation eliminates the need for the switching control signals. Adaptive fuzzy disturbance estimator identifies the variable part of the system dynamics. Adaptation algorithm is derived by using the Lyapunov stability theory and provides the global asymptotic stability of the state errors and therefore a quasi-sliding mode regime. With the proposed physically based fuzzy logic subsystems for the implementation on the robot, estimator’s transparency is enhanced and complexity reduced. It also enables a systematic inclusion of the linguistic knowledge, although it does not require it. The control scheme is tested by simulation and experiments on a laboratory direct drive robot. In both cases the results show good tracking and robustness.
The thesis deals with the topic of elastic-drive control. The main issue that has to be addressed in the field of elastic servodrives is a frequency dependent characteristic of transmission drive's elements. Additionally, undesired resonance phenomenon also can occur. Modern machine tools have to perform higher productivity and quality. Consequently, these require servodrives' operation at higher speed and higher accuracy. Thus, the control algorithm has to be more robust and has to ensure wider frequency bandwidth of position control closed-loop. However, a conventional control approach fails to achieve these demands. Moreover, efforts to achieve rapid response and accurate position tracking cause vibrations in the mechanical drive. This is also a case in servodrives with timing belts. Therefore, a novel robust position control algorithm is proposed in the thesis. The algorithm has been developed for a linear drive with a timing belt. It has been derived by the use of variable structure system control theory, namely, it has been proven, that in this case so called sliding mode occurs that is invariant of system perturbation, parameter variation and external disturbances. However, several obstacles appear in real applications. The discontinuous control of infinite frequency switching which ideally guarantees invariant sliding mode cannot be applied in the practice. Thus, the undesired chattering unavoidable occurs. The proposed algorithm copes with the problem in such a way, that it replaces the switching control scheme with dynamical continuous control. Consequently, the control scheme can be used in real applications. It has been applied to controller of an industrial laser-cutting machine. Experiments showed, that the scheme suppresses vibrations efficiently and significantly reduces position tracking error. Moreover, steady state error was eliminated.
A novel approach to the control of IM drive that assures its high dynamic
performance and robustness against the motor parameters
variation during the motor operation is presented in this
work. The proposed control approach can be classified as a direct
vector control
scheme, since the calculation of the rotor field orientation
is performed in the feedback control loop by the rotor flux observer
that
employes the measured stator voltage and stator current.
The calculation of the reference quantities is performed similarly
as in an indirect
vector control scheme. The structure of the proposed
rotor flux observer belongs to the class of the systems with the variable
structure
and its robustness analysis gives the guidelines for the
gain matrix selection which assures the prescribed dynamic performance
of the
observer in wide speed range inspite of the rotor time constant
parameter perturbation.
The main contribution presents the robustness analysis of the current
control loop with the continuous variable structure current controller.
The analysis based on the discrete complex transfer function
of the control loop and the root locus method shows that the implementation
of the current controller in a rotor field coordinate system minimizes
the effects of the discretization. An algorithm for the rotor resistance
parameter is also presented. The proposed control
scheme was verified through simulation and experiment. The comparison
of the achieved results confirms its superiority over the classical
indirect vector control scheme.
Induction Motor, Motor Drive, Efficiency, Efficiency optimizing, Core Losses, Control, Indirect Rotor Field Orientation Abstract: This dissertation thesis deals with the operation of induction motor fed by an inverter. Induction motor operates with best efficiency in the region near to the rated torque and rated speed. When induction motor controlled by classical control algorithms doesn't operate in vicinity of rated operating point the efficiency deteriorates. Efficiency is especially poor in the low torque region. The objective of this thesis was analysis of losses and synthesis of energy optimal control that would assure maximal efficiency at all operating points. This thesis presents analysis of induction motor with consideration of core losses. Control algorithm with efficiency optimization is proposed. This algorithm is based on optimal rotor flux calculation. Variable motor parameters that are influenced by the temperature changes and magnetic saturation are taken into account. Moreover, operating limits determined by available DC bus voltage and maximal stator current are considered in addition. Sensitivity of efficiency to parameter perturbations is discussed as well. Results were proven by calculations, simulations and experiments on laboratory set-up with IGBT inverter and digital signal processor. Results from this thesis can be applied to electric vehicle end electric boat propulsion systems, and variable speed drives with moderate dynamics requirements.
By using the high power factor correction circuit (HPFCC, i.e. boost rectifier) in single-phase bridge rectifier the unity power factor can be reached. On the other hand, high frequency switching of the main switch generates the electromagnetic interference (EMI) in the radio frequency (RF) range. This thesis is dealing with the wide-band frequency analysis of the boost rectifier with randomized pulse width modulation (RPWM). In the low frequency range, introduction of the RPWM is reflected in smaller increase of the input current's total harmonic distortion factor (THD) for less than 1% and consequently in negligible reduction of power factor (which is still 0.9981). Based on the random process theory, the power spectrum density (PSD) of the input current is estimated and measured for detecting the influence of the randomization in the medium frequency range. In this particular case, the Welch's estimation method has been applied and optimisation method within the Matlab has been done for detecting the crucial parameters of estimations as well. Both, estimations and measurements have confirmed the benefits of the randomization: improvements in the PSD make a good promise for reduced conducted EMI as well. Measuring the conducted EMI assesses behavior of the boost rectifier in the high frequency range. At the end of the thesis it can be concluded, that power spectrum estimation method offers an effective and good prediction for conducted EMI with low cost measuring equipment as well. In this thesis, results of the wide-band frequency analysis in the boost rectifier with PWM as well with RPWM are presented and discussed in detail.
A novel approach to the speed sensorless sliding mode rotor flux observer for induction motor torque and rotor flux tracking control is presented in the work. The main contribution of the work is a new structure of rotor flux observer aimed for the sensorless operation. Proposed approach addresses the problem of sensorless induction machine servo-drive at both, low and high speed, where rapid speed changes can occur. The idea is realized using a non-linear flux observer by introducing a non-linear stator frequency dependant gain in the rotor flux observer, thus enabling the operation at low and high speeds. At low speed observer turns into the closed-loop nonlinear system using stator current, stator voltage and desired rotor flux values as inputs, whereas at high speeds it transfers into well-known voltage model of induction motor with the feedback structure which improves observer's performance. Also an algorithm for stator resistance adaptation is presented. Thus the observer is somewhat more robust to the parameter uncertainties. The performance is investigated and verified by simulations and experiments.
In the thesis the derivation of the dynamical model and control design of a linear induction motor is presented. The model is obtained by the Lagrange method on the basis of the magnetic energy accumulated in the motor airgap and kinetic energy of the moving parts of the motor. The expressions of the primary and secondary currents sheets are formulated for the given winding. Magnetic conditions in the motor airgap are described by the Laplace partial differential equation.It is solved using the expressions of the current sheets as the boundary conditions. as the result the fundamental harmonic of the space distribution of the magnetic field intensity is obtained. The expression for the accumulated magnetic energy in the airgap is derived using the expression of the magnetic field intensity. Lagrange-Euler equations are then used to formulate the lumped parameters three-phase dynamic model of the linear induction motor. This model is not suitable for control design, therefore the three-phase/two-phase and dq transformations of the motor model are performed. The obtained dq model is then used as the basis for the control design. Two different control structures are designed: vector control and the input-output linearising control. It is shown that perfect tracking can be assured if the tracking controllers of the position and the secondary magnetising current are used with the input-output linearising control. The secondary magnetising current estimator, which is used for estimation of the magnitude and position of the secondary magnetising current vector, is derived, because both control structures are based on the dq motor model. The dynamical properties of the estimator and the effects of the secondary resistance perturbations are analysed.
The simulation results clearly show that input-output linearising control guarantees perfect tracking of the position and magnetising current, and better dynamic performances of the drive than vector control. Vector control and input-output linearising control for rotational induction motor were also derived following a similar procedure as described for the linear motor and tested on the laboratory drive. The comparison of simulation results of linear and rotational induction motor control reveals similar dynamical properties of both drives. In addition, very close matching of simulation and experimental results of the rotational motor control is evident.
In the acceleration based robot impedance control approaches external linearization is used in order to linearize nonlinear robot dynamic model. Therefore the implementation of such an approach is very demanding. First accurate robot model has to be built and then considerable large processing power of the computer is needed in order to perform the computation of the robot dynamic model. Previous research results performed in our laboratory have shown that robot dynamic model is not suitable for the decentralization on our robot multiprocessor based controller. Therefore a robust approach to robot impedance control which is based on the PI disturbance estimator is discussed. First robust stability of the robot impedance control with PI estimator by the use of Lyapunov method is discussed. Robust stability of the proposed approach is shown for the case when robot is not in the contact with environment. Then robust stability of the proposed approach is demonstrated for the case when robot is in contact with the environment and next the stability of the robot transition from movement in the free space into the contact with the environment or vice versa is shown. The stability proofs are valid for the whole robot working space and not only for a single working point. In this work the chattering free continuous sliding mode behaviour of proposed approach is shown, where the second order impedance represents the sliding surface and the phase trajectory approaches the sliding surface robustly. Structural analysis has shown that the impedance control with disturbance observer proposed by Ohnishi could be derived from the impedance control with PI estimator. The proposed impedance control with PI estimator was verified by simulations and by the implementation on the direct drive (DD) robot. Both simulations and implementation of the approach have shown that impedance control with PI estimator is an effective approach for robot tasks where the contact between the robot and the environment is needed.
The main subject of the dissertation is the development of the robot controller. It is well known, that new concepts of the automatic control have to be tested not only with the use of simulation techniques, but also with experiments on real mechanisms. Transfer from simulations to experiments is enabled by the use of the open structure of robot controller described in this work. For user, it is also possible to include algorithm in the existing structure, like the assembling of blocks in the simulation programs. Results can be followed by the real--time display of the desired system variables which can also be saved for later processing. Tasks of the robot controller are systematically described in the work. Data exchange is necessary among the different tasks. Since the real--time operation is required, local decisions about the packet contents have to be made. Also, communication connections - paths are derived in such a manner that the fastest possible transfer is enabled. Because of the complexity of the tasks multi--processor, control system with the optimal distribution of tasks is used. Universal descriptions, independent of the robot mechanism, are used. Results are described on the example of the direct drive robot and SCARA mechanism.
A new variable structure control approach is proposed for robust and accurate trajectory tracking of a robotic manipulator with electrical actuators. The approach is based on estimation of joint acceleration signals with introduction of load estimation with the asymptotic observer. Decentralized acceleration controllers are used to generate the local switching function. This novel controller gives a zero steady state error and enables each joint to trace the acceleration command. The parameter variation and disturbance insensitive response provided by this control method is demonstrated on a model of a SCARA robot with AC drives. Simulation results confirm the validity of tracking capability and the robust performance.
Two types of neural network controllers for trajectory tracking: neural network controller in a task space and neural network controller in a joint space were designed.
The neural network controller uses the same approach as computed torque controller, however instead of inverse dynamic models two multilayered neural networks are introduced. Both neural networks controllers were tested as tracking controllers and compared with computed torque controllers in the task and joint spaces. All tests were made and verified trough computer simulations and with experiments on real robot mechanism.
Lyapunov theory for deriving an adaptation law or learning algorithm of neural networks was used to prove robot control system stability. From this prove, the new learning procedure for trajectory tracking robot neural networks controller was developed. The new learning procedure enables position error convergence to global minimum, without danger, that convergence would end in local minima.
Hardware and the software of the robot controller developed on a transputer card MTM-PC with four transputers are described. The hardware and software arhitecture allows a parallel computing of neural networks and their learning algorithms.
An improvement of the BPG-learning procedure for used robot controller, which enables improved position error convergence is also described.
Main topic of this dissertation is the increase of electrical drive reliability. The scope has been restricted to power switching components overload preventation and to redundant software functions for spurious error elimination and permanent error detection.
With an effective current limitation we can influence two main causes of power switching element failures: overtemperature and temperature cycling. Therefore we have integrated a current limiter into the induction motor static control scheme based on DC link current measurement only. The main advantages of this control concept are low cost and low complexity of the drive controller. This control algorithm is suitable for drives, where high dynamics is not required. The modification of this algorithm for electrical vehicle and practical implementation are also shown.
Effective diagnostics is one of the main advantages of digital drive controllers. The presented methods are based on practical experience. One of them is an algorithm for spurious sensor error elimination. Since there is no universal method an algorithm is presented that is as commonly useful as possible. A procedure for permanent fault detection based on error frequency measurement has also been integrated into this algorithm.
Fully digital drive controllers have not only higher capabilities but also higher complexity. Therefore tools for drive testing, installation and maintenance must be adequatly designed. The result of my work are tools which enable a simple and efficient insight into drive controller functions. They are useful for other real-time systems also.
Quality of software tools influences the quality of applications built with their assistance. Problems with real-time software design tools and methods for problem avoidance are also presented.
Most of the methods presented here do not depend on any specific drive type. Some of them are dedicated to induction motor drives for which the experiments presented in this dissertation have been performed.
Last modified Oct. 21, 2000
hajo Joze.Harnik@uni-mb.si