HordaGroup´s AI-journey

Important results from the project

The purpose is to use ML-models and production data. The following goals have been met: - in real time visualize the process parameters in graphical process windows - using ML models predict the future trend 10 minutes ahead - develop a dynamic and easily usable HMI - create a cloud-based digital twin of the production cell - competence development of staff and customers We have also created a simulation tool to understand the process and which parameters affect each other. This has been the basis for building "explainable machine learning".

Expected long term effects

All goals in the project are met. In addition, we have developed a simulation tool for increased and common understanding of the process and visualization of which parameters affect each other and based on that we have created "explainable machine learning". It has taken more time and resources than expected to validate data from different machine manufacturers and to integrate the data into the cloud-based digital twin. The project has created competitive advantages for us in the short and long term with better customer relations, better quality and higher utilization of machines.

Approach and implementation

In order to be able to use the correct ML models, a validation of the collected data was first required. Then we started with the integration of the data into the digital twin, which is a cloud-based solution on a virtual server. To understand the process and obtain documentation of domain knowledge, we created a simulation tool (see appendix). In it you can simulate change and then see how it affects other parameters and the process as a whole. The simulator has been necessary to get an overall picture of the process and to be able to use "explainable machine learning".

External links

Publicerad på Hordagruppens hemsida, interntidning/nyhetsbrev och Facebook.