Close this search box.

CONTINENTAL TIRE – Scrap prediction in the tires manufacturing process using Machine Learning 

The name Dubravko Dolic is well known when it comes to Applied Analytics and AI. He is also the Head of Applied Analytics & AI at Continental Tires. Together with his colleague David Koll, who is a Senior Data Scientist, he presented exciting learnings from his company Continental Tires around the topic: “Scrap prediction in the tires manufacturing process using Machine Learning” at the DATA festival in Munich 2022. “Before I came to Continental, I didn’t even know what my tires were made of.”

Dubravko is mainly involved in the value change process with his team. 
His enthusiasm and excitement for these topics are palpable for every participant throughout his introduction to the presentation. His colleague David took over for showing a use case, which makes the current scrap prediction clear. 

In the first few minutes of their presentation, David talked about the 3 biggest problems they had along the way and explained at the end how they were able to solve them. 

“This was a challenging task” 

“Scrap is when we have material that we produced and which is outside of the tolerances that we need to ensure for the quality of our product. (…) It’s a huge problem when it happens.”, To avoid scrap in any case and thus prevent waste and loss of efficiency, a machine-learning solution was needed. 

“Early scrap prediction is something that we want to detect before it happens.”
His impressive illustration of how many components are involved in a tire makes it clear that this is a complex process. “The goal is to know at the beginning if the material will be okay at the end or knowing if we are running in danger to produce some scrap.” 

During their work, they found 3 major problems. 
1) Input / Target mismatch 
2) Very strict tolerances 
3) Live predictions with 1Hz 

On the third problem, also feel free to check out the video below, which shows an excerpt from his talk. 

How to present these predictions? 

The problems were clear. They found the following solutions for them, which David explained in more detail during the presentation. 

  1. Solve data mismatch
  2. Highly precise model
  3. Live deployment 
  4. Building highly precise models

At the end of the presentation “Scrap prediction in the tires manufacturing process using Machine Learning” the audience could ask their questions about the use case and got detailed answers from the experts. 

Watch an excerpt from the presentation here:


Mit dem Laden des Videos akzeptieren Sie die Datenschutzerklärung von YouTube.
Mehr erfahren

Video laden

Die DATA Party geht weiter!
Das DATA festival #online kommt am 19. und 20. November zurück!
Zwei Tage lang dreht sich wieder alles um die nutzbringende Anwendung von Daten & künstlicher Intelligenz und wie wir damit unsere digitale Zukunft gestalten können. Die Teilnahme ist kostenlos.

Weitere Inhalte entdecken

Data, BI, Analytics oder CPM Software gesucht? Hier finden Sie führende Lösungen im Vergleich!