16. Aug. 2023
Finding a fault in an electric motor is not always easy. However, its effective diagnosis can be facilitated by utilizing new mathematical algorithms with parametric estimation. Lukáš Zezula, a Ph.D. student at the Faculty of Electrical Engineering and Communication at Brno University of Technology and a researcher at CEITEC Brno University of Technology, has been dedicated to this research for several years. He received the Brno Ph.D. Talent scholarship for his work. His algorithms can not only detect the presence of a fault but also determine its location and severity. They are also useful in subsequent fault compensation, ensuring that an electric vehicle remains operational even with a damaged motor.
Among experts focused on this topic, Lukáš Zezula's research stands out as rather unique
In the event of a motor fault in an electric vehicle, several scenarios can occur. In the worst-case scenario, a short circuit can cause a high current that may lead to the ignition of the electric motor. To prevent this, fault detection algorithms are implemented in motor control units to ensure the capture of faults and switch the motor to a safe state. However, this could result in the vehicle coming to a halt, which is unsafe when driving on a highway or through a tunnel.
Therefore, electric motors should remain operational even during a fault. This can be achieved through redundant drive systems utilizing diagnostic algorithms with subsequent fault compensation. Thanks to these algorithms, the motor's control system can handle the fault, allowing the driver to reach a service station or a safe location. Lukáš Zezula, the CEITEC BUT researcher and Ph.D. student at FEEC BUT, is specifically working on this scenario. He focuses on diagnostics using parameter estimation models with discrete time, describing both a healthy motor and a motor experiencing a fault.
"The vast majority of parametric estimation algorithms rely on discrete-time models. There is a plethora of publications on modeling faults in electric drives. Essentially, at every conference on electric motors, someone deals with it. However, discrete models are related to solving a system of nonlinear differential equations describing the motor in the continuous domain, which is more of interest to mathematicians than engineers specializing in electric drives," explains Zezula, providing a rationale for his approach.
During parametric estimation, certain otherwise unknown properties are determined based on known system characteristics and its mathematical model
"It's like guessing the weight of an apple from its size. The exact weight of the apple is not known, but we know that the larger the apple, the heavier it will be - which is, in fact, a mathematical model. If we then have two apples of different sizes in front of us, we can at least say with certainty which one will be heavier. Similarly, during a motor fault, it is possible to estimate, based on the model and measured voltages and currents, the number of turns in the stator winding that are short-circuited. However, for this to work, a sufficiently accurate model must be created, along with a suitable algorithm for parametric estimation," explains Zezula the principles of his calculations. This complexity of the entire issue is what attracted the Ph.D. student to this research.
The proposed algorithms offer the advantage of relatively low computational demands and high reliability. However, the downside is the lengthy development process
"It doesn't always proceed smoothly. I spend many hours in the office, calculating something that is usually impossible to calculate because there is no analytical solution for the given problem. Then I start simplifying the equations - once, twice, three times. It's an iterative process, and I fail constantly until I reach a functional solution. When it's eventually proven that some of the algorithms really work, even on an experimental motor in the laboratory, I am almost surprised," evaluates Zezula.
Zezula has currently designed diagnostic algorithms for short circuits in one of the phase windings of the electric motor and tested them on a workplace with an experimental motor and dynamometer, which is part of the RICAIP Testbed Brno. However, he aims to further focus his research on developing algorithms for other types of faults and their integration with artificial intelligence methods, which he explores in the AI4CSM project. "Currently, I am examining short circuits between turns in one winding, but this can be expanded to short circuits between phases as well. I am also considering how to combine this narrowly specialized expertise with other approaches to increase the reliability and timeliness of diagnostics," concludes Zezula.