Appointment: this challenging job is based on a fixed-term appointment for a period of three years during which the PhD candidate will be able to gain both academic and industrial experience.
Where: University of Poitiers, SKF Verdon (industrial partner) and Ecole Centrale Lyon, France
Topic: dynamical model learning, optimal input design, mechatronics, magnetic bearings, rotor dynamics
Active Magnetic Bearing (AMB) is a contactless force generator used to levitate rotors and shafts, thus to enable rotation while avoiding friction and thus wear. This type of actuator is frequently used in applications where high speed rotation has to be achieved (e.g., turbomachinery, centrifugal compressors, micro-milling). The rotating device is maintained in a fixed position in the air gap via an advanced control system based on active vibration control technology. A turbomachinery operating with an AMB position control system is known for its high reliability. Because of the complexity of such dynamical systems due to their nonlinear behaviors, it is essential to develop tools for tracking their dynamics, for many purposes such as retuning the controller accordingly when the systems dynamics tend to move away from its nominal values or generating relevant input information for condition monitoring and predictive maintenance. This is where real time model learning play a central role in this project. More precisely, during this 3 year project, it will be asked to the candidate to tackle the following scientific challenges:
- Data collection: a specific attention will be paid to the design and the acquisition of informative data sets for model learning. For this first step, optimal experiment design solutions will be considered. Generating cheap and non-intrusive input signals for online model learning is indeed an essential step for generating reliable dynamical models of the process.
- Real time dynamical model learning: once reliable data sets are available, the problem of real time model learning for controller tuning will be tackled.
- Validation: all along this project, test plans will be designed and performed on actual turbomachinery at SKF Magnetic Mechatronics, Saint Marcel, France, in order to validate the new solutions with real data.
Candidate requirements: applicants should have a MSc degree in engineering from a good-quality engineering school. They should possess a strong background and interest in mathematics and, ideally, in system identification and advanced control. They should have excellent analytical and problem solving skills and, preferably, well-developed programming skills. Applicants should have a good knowledge of Matlab. The candidate should have excellent oral and written communication skills in English.
Application procedure: if you are interested by this challenging project, please contact X. Bombois (xavier.bombois@ec-lyon.fr), G. Mercère (guillaume.mercere@univ-poitiers.fr) and J. Mouterde (joel.mouterde@skf.com ) by email with subject "real time model learning", attaching an academic CV, a cover letter, a pdf of your diplomas and transcript of course work and grades, a recommendation letter from your MSc thesis’ supervisor, a certificate of proficiency in English, as well as any other document which can enrich the application.
