Interview mit Fazel Ansari, Dr.-Ing, M.Sc., B.Sc.
Assistant Professor, Deputy Head, Research Group of Smart and Knowledge-Based Maintenance, öffnet eine externe URL in einem neuen Fenster (SKBM) at the Institute of Management Science; Senior Researcher in Industrial Data Science, Fraunhofer Austria
[Anmerkung: Das Interview fand auf Englisch statt und wird nicht übersetzt.]
Human data: For us as industrial engineers this is a challenging area!
Can you please describe your current field of research?
I have a background in mechanical engineering but I did my PhD in computer sciences with a specialization in knowledge-based systems, and hold a Postdoc in the industrial engineering and business administration, so I am a kind of interdisciplinary, adopted researcher. Currently, my main focus is on production and logistics management systems, so, industrial management systems in the wider spectrum. I would say we focus less on classical approaches but rather on data, data modeling and analytics methods. For example, we use unstructured data taken from digital shift books and apply text mining as well as AI meta-modeling methods to extract hidden knowledge and support maintenance white- and blue collars in decision-making. Therefore, data is the crucial part of this research.
Since the beginning of 2019, I have had a major role in leading a recently established research group for Smart and Knowledge-Based Maintenance, öffnet eine externe URL in einem neuen Fenster (SKBM). We apply the methods of AI and semantic technology in order to optimize maintenance processes. In particular, we are managing knowledge intelligence in cyber-physical-social production systems. On the one hand, there is the cyber space dealing with computational and data analytics methods; on the other hand, there is the human interaction with the physical artefacts, so called socio-physical space. These complex systems of systems involve several interactions and thus can lead to failures in operation and decision making not only by humans, mechanical and embedded parts of the machines but also from cyber components. Transformation of jobs and emergence of new skills that will be required for maintenance white- and blue collars is another challenge. Robots and AI agents share tasks with human beings and there are new tasks for the humans. For my research, I develop a new topic, together with outstanding scientists and my colleagues at the Institute of Management Science, namely reciprocal learning, meaning that intelligent machines and humans mutually learn from each other through participating in shared tasks and formation of teams.
What are the main challenges you are facing?
We are conducting a number of experiments because we want to prove our hypotheses. This confronts us with a number of legal and ethical issues we need to clarify. Apart from this, the GDPR has to be taken very seriously. For many of the projects and proposals we are writing, there are clear guidelines, but when we collect real human data and our students are involved in this process, there are some additional challenges.
Here is an example from fundamental research: In one of our projects, we try to provide a scenario of what the role of the human is. Therefore, we ask the human as a subject of the test to wear a device like a smart watch. We follow the required processes, meaning the person confirms participation in a written contract, and we store the data during the research process but not permanently. Nevertheless, we have to be very careful that devices do not collect additional body information, like the heart rate. And if this aspect is part of the measurements, we have to keep this sensitive data absolutely secure and protected. We have to be very careful that no harm whatsoever is done to the test persons. The same applies to the analysis of textual data provided by maintenance staff. For us as mechanical and industrial engineers this is an absolutely challenging area.
What would you need in order to resolve some of the challenges you face? Which services would help you most?
Infrastructures for managing the data would help us a lot. We could define spaces for sharing data but also protected areas for sensitive data, which do not allow access without special permission. These systems would also be very useful for the cooperation with industry partners and for fulfilling expectations of funding agencies as well as individuals.
Apart from the infrastructure and questions like data handover, the quality of the data is very important. Without good quality, we cannot get good models, and especially for PhD candidates who depend on these data sources, this is sometimes a very stressful situation. However, there are companies in Austria that see themselves as research companies and which do have a culture of data management and data sharing. In these cases, we usually know the questions we need to ask: how systematically do we collect the data, what can we do with them, whether we need to delete them etc. For us, it would be very welcome if we as institution had concrete, secure services we could offer. The companies and international partners would trust us.
The only problem that I see with systems like repositories: How can we motivate people and organizations to share their data? This is a cultural issue and I feel that we miss a culture of interfacultative data sharing. Researchers should see the benefits they get from sharing data and algorithms. For me, the added value is: I share and I get something!