Reducing Downtime with Real-Time Streaming Data
Severstal is among Russia’s largest integrated steel and mining companies, producing more than 11 million metric tons of steel and close to $8B in revenue annually. The company’s key performance indicators—including revenue, profitability and dividend payouts—all improved last year, in part due to Severstal’s strategy of defensive growth, which is focused on increasing earnings via enhanced efficiency and product quality rather than increased scale of production. To further solidify its position as a global leader in value creation, the company has embarked on the next phase of this strategy, a digital transformation in which resources are invested in big data, the Internet of Things, predictive maintenance and machine learning initiatives.
Make use of the multiple terabytes of time series data generated weekly by industrial equipment to reduce downtime and increase efficiency.
Use Confluent Platform to feed machine learning models and data analytics algorithms with near real-time data streams of plant data.
- Reduced plant downtime
- Completed initial deployment quickly
- Received support for securing in-transit data
- Achieved one-second latencies
Predicting Problems Before They Occur
As part of this digital transformation, Severstal is using Confluent Platform to stream data from manufacturing sites, integrate microservices and feed machine learning models for predicting problems before they occur.
“With Confluent Platform we have achieved low-latency transportation of time series data from manufacturing equipment and reliable distribution of that data throughout the company,” says Donat Fetisov, Principal Architect at Severstal. “The solution we developed is being used to alert engineers to potential equipment downtime and failures, to optimize equipment operating modes and to suggest changes to key manufacturing parameters. Further, it is scalable to support not only the data we are generating today, but also the data we will generate in the future as the number of sensors used in manufacturing increases.”Lire l'étude de cas