BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:MEchanics GAthering -MEGA- Seminar: Robust design of herringbone g
 rooved journal bearings supported turbocompressors using artificial neural
  networks
DTSTART:20220421T161500
DTEND:20220421T173000
DTSTAMP:20260506T045013Z
UID:8ec27924b202ce5bc0edbcecd6971f2d6a1655305621722a0a515bbd
CATEGORIES:Conferences - Seminars
DESCRIPTION:Massoudi Soheyl (LAMD\, EPFL)\nAbstract Although robustness 
 is an important consideration to guarantee the performance of designs unde
 r deviations\, systems are often engineered by evaluating their performanc
 e exclusively at nominal conditions. Robustness is sometimes evaluated a-p
 osteriori through a sensitivity analysis\, which does not guarantee optima
 lity in terms of robustness. Herringbone grooved journal bearings (HGJB) s
 upported turbocompressors need tight manufacturing tolerances on the beari
 ng clearance and the groove depth to run in a stable mode. This presentati
 on introduces the automated design of HGJB supported turbocompressors via 
 multi-objective optimization including robustness as an additional competi
 ng objective. Robustness is computed as 1) the sampled hypervolume of the 
 feasible region under constraints and as 2) the signal-to-noise ratio of t
 he performance metric over the same sampled hypervolume. In order to addre
 ss the high number of additional evaluations needed to compute robustness\
 , ensembles of artificial neural networks are used to generate fast and ac
 curate surrogates of high-fidelity models. The developed methodology was a
 pplied to the design of a small scale turbocompressor. Robustness was incl
 uded as a hypervolume objective and a signal-to-noise ratio of the load ca
 pacity\, alongside losses. An experimentally validated 1D HGJB model and a
 n intersection mode identification model were used to generate the trainin
 g data. The optimization results suggests a clear competition between the 
 two definitions of robustness and the bearings and rotor losses\, while th
 e use of neural networks led to a speed up by 3 orders of magnitude compar
 ed to the 1D code.\n\nBio Soheyl Massoudi currently works as a Ph.D. cand
 idate at the Laboratory of Applied Mechanical Design (LAMD). His research 
 focuses on Design methodology via surrogate-model assisted optimization us
 ing artificial neural networks and evolutionary algorithms. The target app
 lication is gas-bearings supported turbocompressors for heat-pumps and fue
 l cells.
LOCATION:MED 2 2423 https://plan.epfl.ch/?room==MED%202%202423 https://epf
 l.zoom.us/j/64863509346?pwd=RGlkWjRGOVVSaXZxdFdrb3duYy9mZz09
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR