Gavriil Kupriyanov is a student at the Moscow State University Faculty of Physics and a member of the Design Laboratory. Together with the team, he worked on a case study on optimizing the soda drying process at the alumina plant of a large company, and as a result of the work received a personal grant.
We asked Gavriil to talk about working on a case, his experience leading a project team, and what it means for a student to participate in projects at the intersection of science and industry.
– Tell us more about your project and what kind of problem it helps to solve?
– Our project was dedicated to optimizing the soda drying process at the alumina plant of a large company. The threshold control algorithm implemented in production led to overconsumption of burnt fuel oil. Our team analyzed historical data, estimated the amount of fuel overspending, and developed an improved control algorithm based on Model Predictive Control. According to our calculations, this solution saves 6-10% of fuel oil. The implementation of the developed algorithm at the plant will pay off in about 5 months and will save about 17 million rubles per year.
– What technologies and approaches have you used in your work on the project?
– In the work on the project, we used data analysis, physical modeling and machine learning. To identify the causes of fuel overspending, we conducted a statistical analysis of historical data, as well as assessed the efficiency of the installation in various operating modes. Physical modeling of the drying process was used to enrich the data sample, free from the influence of the current threshold control algorithm. Using machine learning methods, we created models that predict the parameters of the drying plant, and then interpreted them. At the final stage, the obtained models were used in the final control algorithm.
– How were the roles distributed in the team and how did you build the process of working on the case?
– Our team consisted of five ML developers and a business analyst. Tasks were distributed among ML developers based on their preferences, experience, and strengths. This approach allowed each participant to work on the area where he could be most effective.
Weekly team meetings, setting goals, discussing interim results, and interacting with the company's specialists helped keep an eye on the pulse. Thanks to this, by the end of the design laboratory, we were able to present a complete solution that takes into account both the technical features of the task and the practical requirements of production.
– What difficulties did your team face during the project development process and what skills were you able to gain through participation in the Project Laboratory?
– During the work on the project, of course, there were many difficulties: starting with parsing and preprocessing data and ending with analyzing the physical features of evaporation of moisture from soda. The task was at the junction of several areas, so it was important for us not only to build high-quality models, but also to correctly understand the production process itself. However, a gradual immersion in the subject area and interaction with the company's specialists helped overcome these difficulties.
I am sure that each member of our team has acquired something of their own during the project: someone has deepened their data analysis skills, someone has better understood machine learning for industrial tasks, someone has gained experience in communicating with a customer. For me, the most important skill I learned was to lead a team in an uncertain environment. It was not always immediately clear in the project which approach would be the most effective, so it was necessary to make decisions, distribute tasks, synchronize the team and at the same time keep moving towards the final result.
– How did your participation in the Design Laboratory and work with industry representatives affect your professional development and future career plans? Has your idea of what you would like to do in the future changed?
– Participation in the Design Laboratory helped me to see how academic knowledge and technical skills are applied in real industrial tasks. This experience has strengthened my interest in tasks at the intersection of machine learning, mathematical modeling, and industrial analytics. I realized that I was interested not just in building models for the sake of metrics, but in creating solutions that could be applied in a real process and measure their practical benefits. Therefore, participation in the project rather confirmed and strengthened my desire to develop in the direction of applied ML, data science and optimization of production processes.
– What would you improve if you started the project all over again?
– If I were starting a project anew, I would try to dive even deeper into the subject area. And I can say for sure that I would not hesitate for a second when deciding to participate in the Design Laboratory! My participation gave me the opportunity to work on a real task, communicate with experts, and go all the way from analyzing the problem to developing a practical solution.
