AGH UST scientists transform industrial plants into smart factories

A commonly accepted consensus states that the world economy has, until now, seen three industrial revolutions: steam, electrical, and digital. The first was related to the invention of mechanised textile production using the power of steam; the second was brought about by the electrification of production lines; and the third has been running its course due to the automation of technological processes with the use of digital systems and robots. Currently, we are experiencing the fourth industrial revolution (Industry 4.0), which is driven by the rapid development of the Internet – its goal is to transform industrial plants into smart enterprises managed with the use of new technologies.

Scientists from many disciplines have their share in this transformation – mostly economists, engineers, but also computer scientists, and statisticians. For many years now, at the AGH UST and within the Department of Automatic Control and Robotics, the Laboratory of Computer Science in Control and Management has been operating and gathering specialists who conduct research in varous areas, such as data analysis and control systems or production management. Applying their extensive knowledge in mathematical modelling and artificial intelligence, the scholars offer support to entrepreneurs in data analysis and production optimisation. All this to improve the productivity and efficiency of their industrial plants, which translates into increasing their competitiveness.

The team of the Laboratory of Computer Science in Control and Management; photo by Marianna Cielecka

Fundamental research

Currently, the team led by AGH UST Professor Jerzy Baranowski carries out a project, funded by a university grant, that focuses on diagnostic methods of technological processes with the use of machine learning and Bayesian statistics. The scientists are trying to develop predictive models that would allow them to predict faults and plan repair work in a given industrial plant. After all, it is crucial to realise that the lack of process optimisation is not the only thing that generates trading loss – it is also malfunctions and repairs that cause downtime. As the risks associated with loss of profits are high, companies need adequate decision support systems, which the AGH UST scientists are providing.

‘To improve the functioning of technological processes, repairs and modernisations must always be taken into account. This, however, creates room for various kinds of decision support systems that will allow us to predict the damage or detect it at a very early stage, if something has already started to happen. Unfortunately, the problem with decision support systems is this: if we want to make certain choices on the basis of process engineering, we have to be able to defend such decisions’, Professor Baranowski explains.

Each high-risk decision – loss of money, for instance – should therefore be interpretable, that is, it should have a reasonable justification that will make it understandable. In the case of managing company infrastructure, damage might be caused by both too rare and too frequent maintenance works, which is why the choice of a specific option must always be supported by arguments. Nevertheless, predicting faults is difficult and looks a bit like hunting for black swans, which are referenced in the title of a bestselling book by Professor Nassim Nicholas Taleb. The researchers claim that machine learning and Bayesian statistics might be of use in making decisions, as the latter is very good at predicting anomalies.

‘Bayesian statistics wasn’t popular for many years for one reason: it requires tremendous computing powers. It’s only been several dozen years that we have obtained enough processing power to solve these difficult problems. It started with the Manhattan Project, when first Markov chain Monte Carlo algorithms were developed. These methods have been developed since then, and so in the last 20 years, there has been another leap, which allowed us to use Bayesian statistics efficiently’, the project leader describes.

Professor Jerzy Baranowski reveals the secrets of the Monte Carlo method; photo by Marianna Cielecka

But pure mathematics will not suffice on its own because – as the scientists claim – the role of a human expert is indispensable to provide scientists with precise knowledge about the whole process. This is because industrial practice usually deviates from the theory of books and manuals. Hence, only the combination of a priori knowledge with expert knowledge and measurements can give the desired results, whereby it is equally important to take all uncertainties into account. Based on expert experience and time-series analyses, it is possible to create an accurate predictive model, which should also be well-adjusted to a specific installation used in a specific factory.

‘We are constantly talking about decision support systems and not decision-making systems. To say that you can get rid of the expert in any system that analyses data is utopian. The expert must be present at the stage of formulating the assumptions and the stage of creating the system that should, after all, be supervised. Especially because these are high-risk decisions and you cannot let them be made automatically. Sometimes, you might need an entire team of experts to make a decision’, Professor Baranowski claims.

Plans for the future

The research conducted by the team as part of the IERU project will soon find its continuation in a project funded by the NSC Opus programme, whose leader will also be Professor Baranowski. It is also worth mentioning that in addition to an active research activity and fruitful cooperation with business, the Laboratory of Computer Science in Control and Management focuses on a broad educational activity, organising numerous classes in numerical methods, introduction to business intelligence, or project management. There is also the Industrial Data Science (IDS) Student Research Club, which focuses on the issue of data analysis in automatics and is closely connected with the Laboratory.

Professor Jerzy Baranowski will also lead the project titled Vacation learning at AGH for Sustainable Industry 4.0 Education, funded by the NAWA Spinaker programme. This interfaculty project offers three virtual schools dedicated to foreign students interested in Industry 4.0. There will be four editions for each school. The participants will have a great opportunity to expand their knowledge on new technologies and learn about Poland and studying at the AGH UST. This is how our university promotes innovative research, Polish culture, and education in English, which definitely brings the AGH UST closer to fully implementing the chief objectives of the IERU project.

****
The project was funded by a university grant within the framework of the Initiative for Excellence – Research University project (the AGH UST 2020-2022, PRA-6).