Blog

How refining data cracks hydrocarbon problems

17th August 2017

Writing as someone who has spent his academic career studying problem-solving techniques, it is always interesting and rewarding to have real-world problems to solve. A recent project with Intertek is a nice example. Intertek provide a range of chemometric and analytic testing services to the hydrocarbon processing industry. This means they frequently look at large datasets from complex processes, like refineries, and so need appropriate tools to analyse them.

Intertek have a software suite called Interpret, which has modules for assessing the quality of crude oils (InBlend) and predicting organic deposition (InFlow). What they wanted us to come up with was a generally-applicable technique, which could quickly generate useful insight into a wide variety of hydrocarbon industry datasets.

We settled on an approach that generates so-called “fuzzy rules”. These are of the type “IF X IS LARGE AND Y IS SMALL THEN Z IS MEDIUM”. They relate dataset variables by combining them in qualitative rules that can be understood by human experts. It is a useful way to refine understanding of what is going on in a complex system without getting bogged down in the numbers. The trick is to extract a small number of useful rules out of a lot of data and a large number of variables.

During a six-month project, funded by Intertek and The Data Lab Innovation Centre, we applied this technique to a range of industrial problems including prediction of: hydrocarbon fractions produced in a refinery; preventative maintenance and advanced process monitoring. The techniques developed on this project now underpin the InProcess module, which provides a general smart analytics capability in Intertek’s Interpret suite.

Key to the success of the project was strong teamwork between experts at Intertek and data scientists at RGU. Close interaction means that focus is maintained on the desired end-product and there is also a lot of very useful information exchange. A pleasing realization was the strong analogy between how Intertek analysts reason about a complex problem and the machine process of generating and testing fuzzy rules. This led to an approach that is natural to the industrial problems that arise and understood by the experts, making it easier to apply the tool to new problems once the project is complete.

The relationship continues with Intertek sponsoring a studentship and continued technical exchanges. Hydrocarbons and water do not mix, but the blend of academic insight and industrial expertise can be a valuable one.

Want to work with academia? If you are interested in finding out more about how Interface can help your organisation or business, please contact 0131 526 3209 or your regional Business Engagement team member.