Video analytics automated the control of agglomerate sintering at an alumina plant (Northwestern region of Russia)
At a large alumina plant, the technological process of preparing agglomerate (charge) for subsequent metallurgical processing requires precise maintenance of the temperature regime during the sintering stage.
Project background
At a large alumina plant, the technological process of preparing agglomerate (charge) for subsequent metallurgical processing requires precise maintenance of the temperature regime during the sintering stage. Depending on the temperature, three states of the agglomerate are distinguished: weak, normal, and strong sintering. Operators visually assessed the state of the material, which was subjective, dependent on human factors, and did not allow for prompt adjustment of furnace heating. This led to deviations from the regulations, a decrease in the quality of the final product, and additional processing costs.
An automated solution was required for classifying the state of the agglomerate, instantly notifying operators of deviations, and integrating with the corporate data collection system.
Implementation process
Specialists from OSMOVIEW LLC developed and implemented a solution for automatic sintering control based on the OSMOCV platform. The platform was deployed on the customer's server infrastructure within a local network.
A key feature of the project is the interaction model in which OSMOVIEW LLC acted not as a turnkey solution provider, but as an integration hub. The platform combined three components into a single stream: video data from 6 IP cameras, a custom neural service BENTOML (developed by the customer), and the corporate data collection system Kafka.
Cameras were installed in the agglomerate sintering control zone, and shooting parameters were configured for operation in complex industrial conditions. The neural service classifies the state of the agglomerate into three categories — weak, normal, and strong sintering — and is integrated with the platform via REST API.
Additionally implemented: a data converter for exchanging images and analysis results, event generation upon deviations with operator notifications (screen, Telegram, email), and a producer for transmitting structured data to Kafka. Online training for the customer's administrators was conducted (6 people, 5 hours).
Achieved results
- Response time reduced — notifications of violations are delivered within seconds, eliminating prolonged downtime and overconsumption of energy resources.
- Centralized data collection ensured — transmitting results to Kafka provides a single storage for production analytics and deviation statistics.
- Management costs reduced — automation of sintering control reduced the need for constant manual monitoring, freeing up operator resources for other tasks.
- 24/7 monitoring ensured — the system operates without interruptions, providing control throughout all shifts, including night and weekend shifts.
Thus, the OSMOCV platform performed the role of a link between the customer's own neural services, video data, and the corporate data collection system. This allowed the solution to be integrated into the existing IT infrastructure without replacing components or restructuring processes. This approach demonstrates the key advantage of the "integration hub" model: the customer retains control over their developments and systems, while the platform ensures their seamless interaction, acting as a universal data conduit between sources and consumers.