Your browser doesn't support javascript. This means that the content or functionality of our website will be limited or unavailable. If you need more information about Vinnova, please contact us.

AI enhanced modelling for efficent process modelling

Reference number
Coordinator Calejo Industrial Intelligence AB
Funding from Vinnova SEK 292 400
Project duration July 2019 - February 2020
Status Completed
Venture Strategic innovation programme for process industrial IT and automation – PiiA
Call Digitization of industrial value chains

Purpose and goal

The main objective has been to investigate whether AI technology can be used in combination with physical modeling to create models of complex processes in a cost-effective way. To verify this, a delimited part of the flotation process in Bolidens Aitik-gruva has been selected. Due to its complexity, the flotation process is difficult to model with traditional physical modeling. Parts of the process are also difficult to measure, limiting the possibilities of creating a pure AI model based on historical data.

Expected results and effects

The model can predict good concentration of copper in both the foam and the remaining flow with good results two hours ahead. As the model handles both the copper concentration and the flows, this means that it is also possible to obtain a value for the amount of copper coming out of the flotation bank. The project shows that it is possible to combine physical and data-based modeling into one model, which can be used to both gain a better understanding of the process and to optimize copper extraction in the future.

Planned approach and implementation

After mapping the process, a so-called gray box model was created, ie a combination of a physical Dymola model and a data-based AI model. The Dymola model was used to determine the outflow from the flotation tanks, this including mass balance calculations based on the concentration levels in the flow. Calculations were then used as a soft sensor value into the AI model. The AI model consists of a neural network, which was trained on approximately nine months of process data. By combining the models, both the copper concentration and the amount of copper can be predicted.

The project description has been provided by the project members themselves and the text has not been looked at by our editors.

Last updated 10 April 2020

Reference number 2019-02511

Page statistics