A gradient-improved particle swarm optimizer using surrogate modeling

Marc Lambert (1) , Charles Boulitrop (2)

1 GeePs - Laboratoire Génie électrique et électronique de Paris
2 GeePs - Laboratoire Génie électrique et électronique de Paris

An optimization scheme combining a PCE metamodel and a PSO optimizer has been implemented and tested on both a well-posed and an ill-posed ECT configurations. The algorithm has been improved by integrating the information of the gradient of the PCE, improving the quality of the parameter retrieval.

Mots clés: inverse problems; particle swarm optimization; surrogate modeling; polynomial chaos expansion

Domaines: Electromagnétisme

Marc Lambert, Charles Boulitrop. A gradient-improved particle swarm optimizer using surrogate modeling. 25th International Workshop on Electromagnetic Nondestructive Evaluation (ENDE'22), Jun 2022, Budapest, Hungary. ⟨hal-03847594⟩