Celsa Group

Customer Reference


iScrap, Scrap Classification and Quantification with AI Vision

CELSA Group™ is one of the largest manufacturers of steel long products in Europe, the most diversified and vertically integrated.

The Group focus on offering the highest quality and attention in the services that their clients demand. Through their products they provide services to sectors such as construction, automotive, energy and equipment goods, among many others.

The Group is present industrially in Denmark, Spain, Finland, France, Norway, Poland, Sweden and the United Kingdom. Each location allows them to operate with great strategic and competitive advantages to supply customers from anywhere in the world.

CELSA Group™ and the companies that operate under its brand represent the largest producer of long products in Spain and one of the main European producers.

Celsa Group Pervasive client References
Advanced Factories
Advanced Factories 2019
Innovation Prize Winner Project

The Challenge: Classify and Quantify automatically Scrap, basic material for the steelworks

CELSA Group™, with eight million tons of ferrous scrap melted annually in its steelworks, is the largest ferrous scrap recycler in Europe. The Group, in its firm commitment to the circular economy, provides an opportunity for waste to rejoin the circular economy, thus forming an integral part of the new products that are created in the production system.

Through the recycling of steel, which by its intrinsic properties can be recycled many times without losing its essential properties, we use 74% less energy, 90% less primary materials and 40% less water.

CELSA Group™ is integrated throughout the steel recycling value chain, with companies dedicated to the recovery and treatment of ferrousscrap and other materials in Spain and Poland for subsequent recycling.

Today, Scrap classification and quantification is performed in place, by classifiers, the challenge is to reduce the time spent to classify, be more precise and do it remotely, improving security Classify and Quantify automatically Scrap, basic material for the steelworks.

Technical Challenges in the Barcelona pilot site:

  • Process 380 Scrap trucks per day
  • More than 9800 Tn/day of scrap to classify
  • Predict from handheld devices images
  • Select between 50 types of scrap

Expected output from the pipeline:

  • 3 main classes composing the truck load and percentage per class

The Solution: Deep Learning Convolutional networks for characterization of individual images and recurrent neural network to combine the results of multiple images

The project was developed in two phases with Google and Pervasive Technologies working together with a Celsa Team to develop a iterative models to obtain better output precision in each iteration.The pipeline is based in an AI Vision solution to execute the online and remote classification of the type and amount of scrap for each truck load by means of the automated processing of images with technology based on Artificial Intelligence - Deep Learning from Google Cloud and Pervasive Technologies We use Convolutional networks for characterization of individual images and recurrent neural network to add the results of multiple images

Pervasive Technologies has implemented a solution based on Google Vision products, Gooogle Cloud Platform and our own algorithms.


  • Improved Security
  • Remote Classification
  • Improved Classification precision


The automation of the Scrap Classification and Quantification process, with AI Vision, has disruptive potential for the steel industry.  Results suggest that operational costs can be drastically reduced, precision augmented and security improved, by using the proposed highly automated system.


Listo para comenzar?

Contacte con nosotros para agendar una demostración o para conocer más sobre nuestros servicios y soluciones, o simplemente para saber cómo podemos ayudarle.