AI/ML for underground loaders

Important results from the project

The primary goal of the project was for Agio to learn more about AI / ML and how to apply the knowledge in a production environment. Agio has fulfilled the goal as they know much more about AI / ML today than when their journey started at the end of 2019. Agio has understood the importance of analysing the data you have, you need to understand and find out which data is relevant to the issue you’re working with. Agio has come to the conclusion that machine learning can be useful for improving the predicted rate of the loading production by training in historical data.

Expected long term effects

Through this project, Agio has started its AI / ML journey towards becoming experts in the field. Various ML methods have been tested to evaluate how well machine learning is able to predict the loading rate of planned production and we have built up a contact network with experts in the field. The result will be reported to LKAB to show the possibilities with AI and show new concrete applications for their production. The result that the ML model calculated is 1.5 - 2 times better than the existing model.

Approach and implementation

Step 1 was to get to know existing data, to understand and find out which data is relevant for this specific issue. The database was analyzed and meetings were held with LKAB to find out which parameters affect the loading rate (ton/h). Step 2 was dataextraction to reduce the amount and exclude protruding data. Data that was missing had to be reconstructed. We studied different forums and discussed with LTU what type of data an AI / ML needs. ML.NET and the ML models for regression problems were tested and compared with a reference model.