We envision global optimization as finding a suitable initial guess of a considered optimization algorithm. One can imagine that this possibility improves the existing optimization methods. To do so, we present a methodology that consists in the resolution of a sub-optimization problem by using a multi-layers line search algorithm. We check the efficiency of this approach by considering benchmark and industrial problems. In particular, we focus on the optimization of the shape and hydrodynamic properties of a bioreactor in which a wastewater substrate is degraded by a microbial ecosystem