Interview with Pavel Pupyrev, Head of the Data Monitoring and Analysis Center of VTI
For the third year, VTI OJSC has a division created to develop predictive maintenance systems - the Data Monitoring and Analysis Center. About the Center and the results achieved during this time: an interview with the head of the Center - Pavel Pupyrev
Where did the idea of creating a predictive diagnostic system come from?
This idea is far from new. The first attempts to create such a system were made in the 70s. Only in our institute over the past 30 years, the development of such a system has twice begun, but, unfortunately, due to a number of objective reasons, the work was suspended. However, these developments allowed us to accumulate knowledge and start not from scratch, but having a serious background in this area. Now there is a more favorable situation for the successful implementation of this project. Firstly, the creation of domestic software is supported at the state level, in particular, it is impossible not to mention the departmental project "Digital Energy" developed by the Ministry of Energy of Russia, as part of the implementation of Presidential Decree No. 204 of 07.05.2018, the purpose of which is to transform the energy infrastructure of the Russian Federation through the introduction of digital technologies and platform solutions. Second, information technology has evolved significantly not only in terms of the availability of convenient software for creating software and analyzing data, but also in terms of hardware, which have become smaller in size, work more efficiently and cost much less.
Tell us a little about your Center and what goals and tasks are facing it?
The center was created in early 2018 with the main goal - the development of a predictive maintenance system, which would allow energy companies to switch from a system of planned repairs of equipment to a repair system according to technical condition by analyzing data obtained from sensors installed at the power unit. Using such a system, power companies can significantly improve the reliability and quality of energy supply to consumers, reduce the cost of repairing electric generating equipment and reduce the number of unplanned equipment shutdowns, which often lead to large fines. In general, creating such a system is not an easy task, since unlike many other data analysis tasks, it is necessary to involve high-class technologists in order to create a system together with IT specialists. Our institute is well known for its technologists, so the creation of the Center was necessary, including the creation of a strong IT team.
What results can you boast of for the work done during the two years of the Center's existence?
A lot of work has been done. By the end of 2019, we created our own system for diagnosing and predicting equipment and launched it into pilot operation at one of the energy facilities in Russia. In addition to data analysts and programmers of our Center, more than 30 technologists participated in the creation of this system. Diagnostic and prediction algorithms were developed not only for the main equipment (boiler, turbine, generator and transformer), but also for the auxiliary (circulation, condensation and feed pumps, HPH/IPA, traction mechanisms). The created system was developed completely "from scratch" based on open source projects (Apache Kafka, Docker, PostgreSQL, etc.), which make it independent of Western sanctions and do not require additional monetary investment when installed on other facilities.
What do you see next steps for the next two years to develop the system?
Algorithms, algorithms, and once again algorithms. The main difficulty in creating a predictive service system is the development of algorithms that can correctly determine the birth of a defect and predict its further development. That is why our first priority will be both the improvement of existing algorithms and the development of new ones. One of the interesting new directions for the institute is the creation of a digital twin of the power unit, which we will begin to develop this year.
In addition to working on algorithms, we will also devote enough time to unify the developed platform, which is the foundation for a predictive service system. In our system, more than 100 unique algorithms are programmed to detect a defect. It is very time consuming to program, test and support these algorithms at such a number, so in the system architecture we use unified approaches to their implementation, which help us quickly implement and test a new or modified algorithm for correctness of operation. I have a great desire at a certain stage in the development of the platform to publish the source code of the system so that other organizations and inventors can use it. I think that this should allow everyone to focus on developing algorithms rather than spending their resources on creating a system that we have already developed.
Will the institute be able to independently create such a complex system?
The answer is definitely yes. Moreover, I believe that in our country only our institute is able to master this task. Our technologists used decades of experience and knowledge to develop algorithms. We used this experience to build the system and, of course, we will use it in the future.
At the same time, for the larger development of the project, it would be very useful to attract high-class specialists from around the world to develop an exhaustive list of algorithms, so we are ready to work with other enterprises, educational institutions and private owners to jointly develop diagnostic algorithms. It will also be possible to introduce already developed algorithms into our system in order to offer them as part of the entire system to our existing and new customers.