In many industrial applications, to cut down either the cost of natural experiments or the computational cost of complex, high fidelity scientific and engineering simulations, they are often substituted with regression models (in this context also referred to as surrogate models) that mimic the behaviour of the original system as closely as possible while being much cheaper to evaluate. Primary objectives of surrogate modelling are to obtain a model that is as accurate as possible and to minimize the required computational and experimental effort, including minimizing the necessary number of sample points and utilizing an efficient modelling method. In this paper, a surrogate modelling procedure is proposed which incorporate: 1) a space-filling method for designing of experiments based on an analogy of uniform distribution of charged particles in alongside placed multidimensional phantom spaces; 2) an adaptive regression modelling method based on a heuristic search through an infinite model space.