Solving the complexity of the process industry

FCAI’s AI for Process Industry webinar in October was hosted by associate professor Simo Särkkä from Aalto University, and gathered a total of 120 participants. Video and slides from the webinar can be found on the event page.

VTT’s Research Manager Heli Helaakoski outlined the potential for AI in her opening presentation: “There are three major drivers to find new improved AI-aided solutions: greenhouse gas emissions, waste materials and waste generation. We need technological breakthroughs, and AI can have a supporting role to make them happen.”

“How can we optimize the whole feedstock flow of the industry, for example, or how to decrease production of waste? Could we solve complex supply chains of the industries with AI? Or how to switch rapidly from one raw material to another? In the current world of big changes, the best solutions should be calculated automatically,” says Helaakoski. According to her, AI should be utilized more in Europe. “There is already enthusiasm in Finland, which is shown by the number of participants in our webinar.”

“Process industry produces material for our everyday life: cement, steel, plastics, paper, minerals, chemicals, fuels etc. In Finland, the manufacturing industry’s share of exports is 50 percent, and 30 percent of the Finnish labor force works in the manufacturing industry. So far, the adaption of AI is low in the process industry. We can do better than this. What we need now is collaboration between academia and industry to create practical industrial solutions,” she continues.

“So far, the process industry has not been a typical field for AI researchers, and therefore it is important to get more interaction between AI researchers and industry representatives. Cooperation between different disciplines is also needed to develop practical solutions. When we talk about process industry, it’s good to keep in mind that we are not talking only about optimizing production. You have to focus on the company's core problem in order to reduce emissions, for example,” says associate professor Arto Klami from the University of Helsinki.

“The main goals of the Finnish industry and the research methods of FCAI offer good opportunities for intensifying cooperation,” he continues.

Does AI bring added value for process industry?

“This is a difficult question because there may be numerous changes that are implemented in the development process,” says VTT’s research professor Heikki Ailisto. “For example, the development project presented by Outokumpu in the webinar had about 200 parameters that changed during the process. On the other hand, an improvement of even one percent in production efficiency would be significant in a large production process.”

Virtual laboratory

Arto Klami presented the concept of a virtual laboratory to combine process automation of the industry with AI models. Virtual laboratories are a way to combine the state-of-the art AI methods with state-of-the art knowledge from process industry. Its operating idea is to connect the company's own modelling systems with those from researchers.

Companies are competing for AI talent

“The FCAI hub with its AI experts from the universities and VTT is an opportunity for companies to fix their shortage of AI talents. Companies are welcome to join the FCAI network. FCAI's projects solve long-term challenges in joint innovation projects of research organizations and companies,” says associate professor Simo Särkkä from Aalto University.

“There is a lot of AI expertise in Finland, but let’s not lose it. We should offer interesting opportunities for talented people that, for instance, FCAI has attracted to Finland. I wish process industry companies were more involved in these projects. Business projects can also be implemented as masters or doctoral theses,” says corporate relations manager Terhi Kajaste from Aalto University.

Sufficient knowledge resources can also be ensured in companies by setting ambitious goals and long-term funding for development projects.

Rethinking wastewater

Associate professor Francesco Corona from Aalto University demonstrated how AI can be used to utilize the potential of wastewater. He encourages a rethink of the management of water as a resource-rich material. This would result in reducing costs, minimizing chemical input, improving energy neutrality and producing marketable products.

AI in chemical industry

“In chemical research there are plenty of opportunities for AI due to the complexity of science and large amount of data”, said digitalization manager Rupesh More from Neste. “AI has a huge role in creating high-quality products from low-quality materials. We are working with complex renewable raw materials, and we aim to solve the toughest combinatorial and computational problems,” More said. He presented two success stories, one related to improving a product design process with Aalto University, and the other to creating a prediction model for equipment fouling behavior.

“When working with new materials, discovering molecules, and trying to find new methods to convert and refine interesting feedstocks into sustainable fuel products, an AI-driven assisted experiment design can be seen as an opportunity. Laboratory experimentation is expensive, time-consuming and currently uses human judgement-based decision-making in selection and prioritization of experiments.”

The potential solutions would involve developing machine learning models using historical data to assist researchers in experiment design with predictions for optimization and prioritization of experiments.

“Such an AI model would gradually learn using real feedback on applicability of predictions from researchers, and mature in accuracy over a period of time, thus enabling reduction of physical experimentation performed in the laboratories,” says More.

Optimizing furnace melting time

Digital manufacturing in the stainless steel industry at Outokumpu was the topic of research manager Esa Puukko’s presentation: “Optimization of electric arc furnace melting time was chosen for this machine learning project because it is an energy-demanding process and therefore there is a lot of potential for savings. Full melting of the batch is a complex system, and it has been more or less controlled by an operator’s subjective decisions. The target was to make a machine-based model to prevent unmolten material being left in the furnace, avoid overheating, to save time and energy and to improve quality and yield of the output.”

Puukko was pleased with the results. “Four percent less time was used for the processing and in this case saving time means saving energy. Also, carbon dioxide savings were recorded, for example. The experience can be copy-pasted to our other furnaces.”

Another use case comes from Outokumpu’s automatic surface inspection, a key process in quality management in stainless steel making. It detects marks that occur during the processing of products. All those defects must be picked up. “Now we are using machine-learning algorithms for better quality control. We have taught our system to pick up certain defects from the camera image and this helps us to react faster to the defects. Now we are moving towards a semi-automated system. Our final target is to have a more or less automated system,” said Puukko.