Ore conveyor belt rollers operate in harsh environments, making them prone to premature failure. Their service lives are highly dependent on the stress field and bearing misalignment angle, for which limit values are defined in a standard. In this work, an optimization methodology using metamodels based on radial basis functions is implemented to reduce the mass of two models of rollers. From a structural point of view, one of the rollers is made completely of metal, while the other also has some components made of polymeric material. The objective of this study is to develop and apply a parametric structural optimization methodology to minimize the mass of the two models of rollers. To represent the mechanical behavior of the rollers, simulations were performed using the finite element method. During the numerical optimization process, the variable parameters were the dimensions of the shaft and external tube. The geometric configuration that corresponded at the same time to the lowest mass and acceptable ranges for the stress and bearing misalignment angle was determined. With the proposed methodology, a 32.3% reduction in mass was obtained for a metal roller design and an 18.9% reduction for a polymer roller. In both cases, the constraints were not violated. For the all-metal roller, the safety factors for the maximum stress and bearing misalignment angle were 1.44 and 1.75, respectively, while for the polymer roller the corresponding figures were 1.50 and 2.23. This work describes a low-computational-cost optimization methodology for roller designs that have been little studied in the literature. Furthermore, the methodology could be adapted for use with other types of rollers and rollers made of different materials.