Case Study: Severstal achieves reduced plant downtime with Confluent Platform

Severstal Reduces Plant Downtime with Real-Time Streaming Data

Severstal, one of Russia’s largest steel and mining companies, faced the challenge of making more than 9 TB of weekly time‑series sensor data available in near real time so machine‑learning models could predict equipment failures and reduce costly downtime. To solve this, Severstal adopted Confluent’s solution—Confluent Platform (including Schema Registry, Kafka Streams and Kafka Connect)—to stream data from manufacturing sites, integrate microservices and feed analytics and ML pipelines.

Confluent implemented a Kafka‑based streaming platform—standardizing event schemas with Schema Registry, using Kafka Connect to move data in and out, and building real‑time processing with Kafka Streams—while providing deployment and security support. The Confluent solution handles about 30,000 messages/sec and ~3 TB streamed per broker per week, delivers ~1‑second latencies from raw data to prediction, enabled a fast initial rollout, and in the first six months resulted in several‑times fewer mission‑critical incidents, reduced downtime, improved product quality and lower energy consumption.

Severstal

Donat Fetisov

Principal Architect