Thesis supervisor: Attila Magyar
Location of studies (in Hungarian): University of Pannonia, Veszprém, Egyetem str. 10. Abbreviation of location of studies: PE
Description of the research topic:
Increasing environmental and economical expectations towards many industrial sectors like (i) automotive, (ii) aerospace and (iii) precision mechatronic engineering, have led to a surge of complexity increase of engineering systems (e.g. electric and hybrid cars, aircrafts and drones, wafer scanners, etc.) as they are increasingly relying on a fusion of advanced mechanical, electrical and computer technologies. The current cutting-edge system designs exhibit complicated non-stationary, nonlinear and spatial dynamic behavior and they have a dominant Cyber-Physical (CP) nature allowing massive data-logging capabilities. Traditional modeling and control engineering solutions are incapable to cope with this surge of complexity, performance demands, the intense network interaction in CP systems and the flood of data available for performance enhancement. On the other hand, recent advances in Artificial Intelligence (AI) methods have shown high potential in solving data-intensive high-complexity modeling and regulatory problems. A major drawback of these methods is that it is difficult to incorporate prior engineering knowledge in them, to give safety and performance guarantees, to provide interpretability and transparency of the solutions, which are essential expectations for applications in sectors (i)-(iii). This PhD research aims to establish the next generation of modeling solutions by developing a novel synergy of AI methods, which can handle the complex nature of these systems with an automated modeling framework. The proposed methods aim at efficient digital twining of individual applications, i.e., automatic learning of dynamical models using both user-specified input, measured information, and existing engineering knowledge. To provide theoretical guarantees and efficiency, an identification theory for machine learning based modeling (relying on deep auto-encoders [1], GPs [2] and neuroevolution [3]) will be established where automatic complexity optimization of network structures & noise models, experiment design, utilization oriented objectives (e.g., control specifications) and model invalidation can be well understood both in open- and closed-loop setting.
Papers related to the topic:
[1] I. Goodfellow, Y. Bengio, and A. Courville: Deep Learning. MIT Press, 2016.
[2] C.E. Rasmussen, C.K.I. Williams: Gaussian processes for machine learning. The MIT Press, 2006.
[3] K. O. Stanley, J. Clune, J. Lehman and R. Miikkulainen: “Designing neural
Number of students who can be accepted: 1
Deadline for application: 2020-09-30
2024. IV. 17. ODT ülés Az ODT következő ülésére 2024. június 14-én, pénteken 10.00 órakor kerül sor a Semmelweis Egyetem Szenátusi termében (Bp. Üllői út 26. I. emelet).
Login
