AI-powered ferroalloy production optimization system

Modeling and optimizing ferroalloy furnace operation based on raw material quality data


  • Manufacturing


  • Artificial intelligence


  • Keras
  • Numpy
  • TensorFlow
  • pandas
  • СatBoost
  • GLFW
  • Pillow
  • Plotly
  • Jupyter
  • PyOpenGL
  • TQDM
  • Scikit-Image
  • CatBoost

The metal melting time directly depends on the raw material properties. With that, materials are supplied from different deposits, undergo different treatment, and differ in composition, dimensions, and humidity. Down the road, it may be hard for the companies to plan their processes. It's barely possible to predict production time per batch (3 hours or 8 hours), calculate costs, and envisage the logistic journey. 

We were approached by a ferroalloy manufacturer that had been seeking ways to curb and calibrate the melting time. We analyzed 2-year historic data including information about raw material composition, melting time and temperature, current, and output metal quality. Leveraging those records, we designed an AI-powered system that helps produce a batch of ferroalloy products within the shortest time possible, 2 hours. The program analyzes the chemical and mineralogical composition of raw materials, size and humidity of particles. Based on that, it figures a perfect recipe: how much material and how many reagents to load, and in what order.

After integration, the system has shortened the ferroalloy production time by 35%, helped revamp process quality and performance 2.5 times. 

Let’s discuss your project

Tell us about your requirements and we’ll get back with a possible technical solution

By clicking «Submit», you consent to the processing of your personal data

Call us

Write to us