Use predictive models to identify chemical contamination and reduce downtime

Background

Our client is an Austrian chemical company and one of the largest producers of polyethylene (PE) and polypropylene (PP), with manufacturing operations in Belgium, Central Europe (Austria and Germany), Finland, and Sweden.

Challenge

Our client wanted to implement a solution that collects data from various production lines to predict the origins of contamination in order to avoid production waste. The overall goal was to leverage previously unused data to its full potential. Online equipment generates images when contamination occurs on the production line, but lab technicians needed to manually investigate these pictures with a microscope to determine the origin of the contamination.

Approach

- Ensure images are recognized and decoded in real time to extract data to feed models. - Build predictive models, with self-learning capabilities, to predict the origin of the contamination. - Deliver advanced defect reports with a web application, allowing interventions on the production line in real time when a contamination occurs.

Key results

Benefits

Our client gained real time insights on contamination origin so that corrective actions are possible even during the production process. Additionally, they can avoid having to dispose of entire batches of contaminated products by identifying only the compromised ones. All information about defects is centralized in a single advanced report, allowing automatic matching of defects with machines to avoid production process downtimes.

Share this key play