Structural noise models have several applications in ultrasonic NDE: forward models can help predicting the performances of inspection setups, and inverse models can be used to characterize materials (evaluation of the structure from attenuation and noise simulation/experiment fitting). For polycrystalline materials, an approach based on the Born approximation is widely used in the literature. It has been adapted to various materials and has proven to be particularly accurate for metals such as nickel and titanium alloys. However, its results in the case of strongly scattering metals such as stainless steel used in the nuclear industry have not been fully satisfactory. In this communication, the limits of the Born approximation applied to polycrystalline metals are discussed. A second order Born approximation is proposed as a more accurate alternative. Results are presented and compared to those obtained with other models, in order to highlight the limitations of the first order approximation and the possibilities for improvement. This study has been conducted within the framework of collaboration between IRSN and CEA.