Matej Emin

CEO, Culmium Advanced Technologies

Below is the video transcript.

Companies and organizations are on different levels of understanding data, of working with data. Having a license for a program is really just the beginning. There are then two aspects of using that software to the full of its capability: one is understanding the procedures behind what needs to be done and when, and, of course, the other aspect is using the program and knowing the program very well: what it can do and where to find specific functions.

​We are currently three people employed at Culmium. We are two physicists and one geologist. We are offering our consultancy services to different customers, ranging from industrial manufacturing companies to academic institutions.

​The background of that procedure was a company who is producing extra white flint glass in Slovenia is required to add selenium dioxide into the batch in order to get the desired quality of glass. This is a normal procedure. This selenium dioxide is very volatile at high temperatures and gets emitted from glass melt. Then it's captured in the filter. This filter dust then becomes a hazardous material. Of course, this is not desired. It's not an ecologically sustainable solution to generate a lot of waste that needs to get disposed. Because this emitted dust is selenium-rich, there was an idea if it would be possible to establish a circular production loop, right? Selenium- rich filter dust would be reintroduced into the furnace in order to be used like a raw material.

We came up with an idea to establish a prediction model based on the manufacturing process parameter values. So, this model would be used to predict the amount of selenium in the filter dust. If that model would work very well, then we would at every time know how much selenium is being emitted into the filter dust and this would then be properly weighed and added to the batch. Industrial data can be very messy. So, we put all those tables into the correct form so that it was suitable for further analysis with JMP. This was done in collaboration with subject matter experts. They helped us with this initial understanding of which parameters might be most influential.

Then we used our prediction model to predict the new values. And when we compared these values with the values we got from chemical analysis of filter dust, the predictions were good. We were satisfied. So then the company did a pilot study where they were returning this filter dust back into the furnace for a certain amount of time. And they were able to return 100% of this filter dust without a negative effect on product quality.

In this context, JMP allowed us to quickly visualize the data that we had at hand. It allowed us to get to this intuitive understanding of data much quicker than we would with probably any other tool. Focus more on the content and less on how to manipulate the program to do something for me. The work we did with this glass company was so important to certain long-term strategies of our country, of Slovenia, that we have won a silver award on a national level for this work.

When companies are moving towards greener production, they need very strong data analytics. We should pursue this goal of minimizing industrial waste as much as possible. The possibilities are endless.