The numerical optimization of a photoacoustic gas sensor is a challenging problem in terms of computational time. The signal detected by the gas sensor is a non-linear function depending on several geometrical parameters. The optimization requires a very large number of function calls, thus an important simulation time. In the case of photoacoustics, two different numerical models with different fidelity levels are available to simulate the behavior of the component. In order to reduce the

more »

... tational burden of optimizing the gas sensor, a new multifidelity metamodeling framework, based on Radial Basis Function, is proposed. The present method offers an alternative to co-kriging (a widely used multifidelity metamodel). This multifidelity model is then used in an optimization sequence to enrich a training database, via a strategy inspired by Gutmann [1]. The process is applied to the optimization of geometrical parameters in the gas sensor problem.