Teaching New Motors Old Tricks: My PhD Journey in VFD Behavior Replication
- Apr 9
- 3 min read
Updated: 3 hours ago
Text: Leyla Shojaeifard
Doctoral Researcher, DSII
Bridging theory and practice, my doctoral research aims to solve a significant industrial challenge: automating the complex process of configuring replacement Variable-Frequency Drives (VFDs) to match the behavior of their predecessors precisely. Working directly with industry engineers has confirmed both the practical need and potential impact of this work as I build the necessary domain expertise to develop ML solutions for this real-world problem.
My research focuses on solving a practical problem in the electric motors industry. More specifically, my goal is to develop a solution for automatically configuring a given VFD to mimic the behavior of another VFD model. In practical terms, the use case is that an industrial VFD controlling a heavy-duty electric motor needs to be replaced, and the replacement option is of another make and model.
The language of drives
Configuring a VFD to operate a motor effectively requires managing a long list of parameters, making it a substantial task to do manually. Additionally, there is a desired behavior that must be replicated as closely as possible. Industrial VFDs must be configured differently depending on their application; a motor controlling a conveyor belt requires precise speed regulation and gradual acceleration, while a quarry pump motor needs high starting torque and protection against sudden pressure changes.
It is the practicality and the potential for solving real-world problems within the industry that got me interested in this research topic and the DSII doctoral position. I am not belittling the importance of theory, but I find it more meaningful to work toward solving engineering problems present in day-to-day work. The industrial collaboration aspect of DSII is what got me inspired to apply for the position and get started with the topic at hand.
My first year has been slow in terms of progress, for I have been deeply invested in learning more about the application domain so that I am able to speak the same language as the electrical, control, and grid engineers working at the collaborating company.
The latest status update on my research is that I have been to the company’s facilities both in Vaasa, Finland and in Gråsten, Denmark, interviewing the people working with the VFDs. I have heard their perspectives on the topic, the motivation for why this VFD behavior replication is important, the challenges they face, as well as the concerns they have with development.
I have yet to analyze all the interview data, but I can already share that based on the feedback I have received from the engineers, the general consensus seems to be that they are happy to hear that people are working on solving this problem. I think hearing this directly from the people working in the field does well to concretize both the practicality and the significance of the research topic for industrial day-to-day work.
Machine learning meets motor control
Before I started my work, I was expecting to be working with data analysis, statistics, and machine learning during my first year of studies. I soon learned that I may need to get a bit more involved with the practical work, and since then I have been refreshing my memory on the related areas of physics, and I am working on learning more about electrical and control engineering. While this was a bit surprising, I am looking forward to what the future will bring. There are still a few years of studies ahead of me, and I will try my best to make the most of what may be thrown my way. By building the groundwork for the domain knowledge I prepare myself for the real task that is looming around the corner. When the time comes to start training the machine learning models for automating VFD parameter configuration, I will be ready.
Bình luận