Pultrusion is a continuous process for manufacturing polymer composite with uniform cross-sectional profiles. In this process the die temperature process variable can be used to improve the chemical and mechanical properties of the polymer composite. Several works have been done regarding simulation tools for pultrusion analysis. Usually, the development of manufacturing processes is mostly a trial-and-error exercise. However, few studies focus on optimizing this manufacturing process. In this paper, the pultrusion process of a glass fibre-epoxy set with different geometries are simulated by using a three-dimensional finite element-nodal control volume (FE/NCV) approach, a quadratic programming algorithm and a particle swarm heuristic was used to optimize the manufacturing process. The aim is to show that to optimize pultrusion process, quadratic programming algorithm can also be very efficient in finding the optimal distribution of the temperature in the pultruded material. The second proposal of this paper is to present a simulation for pultrusion process based on a total number of 16 heaters embedded into die block, instead of external planar heaters. The results show that the mean of the cure degree is satisfactory when used with many internal heaters.