Case Study: a metal processing and recycling company in Bulgaria achieves more accurate copper scrap evaluation with InterConsult Bulgaria's machine learning solution

a metal processing and recycling company in Bulgaria achieves 0.6–0.8% copper estimation deviation with InterConsult Bulgaria

The customer, a metal processing and recycling company in Bulgaria, faced the challenge of accurately estimating the copper content in incoming scrap metal piles. This process was traditionally done by expert employees through visual inspection, but it was prone to human error. The company partnered with vendor InterConsult Bulgaria to automate and digitalize this evaluation using a machine learning solution.

InterConsult Bulgaria developed a computer vision model trained on thousands of scrap pile images to classify copper content percentages. They utilized pre-trained Keras Resnet and MobileNetV2 neural networks, which delivered the best results. The solution achieved an extremely good average deviation of just 0.6 to 0.8 percent from the actual copper composition, significantly improving the accuracy of the customer's invoicing and raw material valuation.