Case studies / Cloud data for a steelworks

IoT · Metallurgy

Cloud data for a steelworks: a pilot at a lower cost than in-house infrastructure

ClientCelsaTechnologyAWS IoT Greengrass, AWS IoT Core, AWS IoT SiteWise

The challenge

Celsa, a metallurgical company dedicated to metal rolling, has production plants around the world, and in many of them the processes are the same or very similar. It was looking for the most efficient way to obtain the data of those processes, store it and process it to get metrics and improve the process.

It asked us for a proof of concept (PoC) to test a system based on cloud services (serverless), with these objectives:

  • Optimise the cost of processing the data, from the machine to the database or the cloud.
  • Simplify the data collection infrastructure and have a robust and secure system.
  • Be able to use cloud services for artificial intelligence, machine learning and Business Intelligence afterwards.
  • Distribute the data to the departments according to their needs.
  • Lay the basis for predictive maintenance.

What we did

A data capture system with AWS services:

  • AWS IoT Greengrass bridges the local MQTT topics with AWS IoT Core: it is the encrypted transfer of the data to the cloud.
  • AWS IoT Core takes the data, processes it with rules and passes it to AWS IoT SiteWise.
  • AWS IoT SiteWise generates the data model, stores the data as time series and gives a portal with charts, alarms and warnings.

Services such as SageMaker or QuickSight can be connected to this infrastructure later, for artificial intelligence or Business Intelligence.

The result

The conclusions of the pilot:

  • The system is scalable and flexible: it can be extended or changed according to the needs of each production process.
  • It is robust and secure, with encrypted data and users and roles that are easy to define.
  • There was a significant cost reduction compared with a classic on-premises infrastructure.
  • It is ready for predictive maintenance to be implemented.
  • Scaling it first needs an in-depth study of the data and a clear list of what is connected, with its metrics, model and attributes.

Feeding what is learned from the data back to the machine automatically was also analysed. It is possible, but with Celsa’s type of manufacturing the changes still need a manager’s approval: for now the system would make proposals, which people approve.

I want this on my machine →