Customers: Khimprom of PJSC Novocheboksarsk
Contractors: Digit Product: Artificial intelligence (AI, Artificial intelligence, AI)Second product: Individual development of electronic state services Project date: 2018/04
|
Khimprom and Tsifra company announced on April 5, 2018 the beginning of development of an analytical system using methods of deep machine learning for workshop of production of chlormethanes.
It is planned that a system will issue recommendations to the operator of workshop at the choice of optimal parameters and the technology modes for achievement of maximum capacity. A system will allow to increase release of a high-marginal product of methylene dichloride to 46 tons per day, thus, 5 percentage growth on the current production capacities are expected. Economic effect of implementation is predicted at the level of 11 million rubles annually, at the same time investments in Khimprom are expected to be paid back less than in a year.
It is expected that on project implementation about 4 months will be required. At the first stage experts in data (data scientists) from Digit will analyze arrays of managing parameters of Khimprom and will develop a mathematical model. At the second stage the software product (software) which task — to analyze and issue recommendations to the operator of workshop will be developed. At the third stage will pass tests of the developed product, then a system will be brought into trial operation.
We have an opportunity to pay attention to issues of optimization of production processes, updating of business and other assets within the implementable program of renovation of the enterprise — the CEO of Khimprom Sergey Nauman considers. — In case of positive result it is supposed to implement the similar systems on other productions of the Novocheboksarsk chemical plant. |
The developers of artificial intelligence technologies who are torn off from real production do not achieve desirable results. Joint work with experts of PJSC Khimprom will allow to test as fast as possible the developed algorithms and also to train a system at real production data. As a result we should receive not simply "digital model", and a real and unique case of application with the confirmed effects — Andrey Kondratyev, the director of strategy in the process industries Digit told. |