An in-depth analysis of the flow patterns and mixing dynamics in a twin paddle blender for bi-disperse non-spherical particles were implemented through the use of the discrete element method (DEM) and experiments. This study aimed to explore the mixing efficiency of a double paddle blender containing two different shapes of non-spherical particles. The study focussed on the demonstration of the applicability of the GPU-based DEM model. To achieve this, calibration tests were performed using a classical rotary drum to validate the accuracy of the model. The next step was to examine the impact of various parameters on the mixing performance, including the paddle rotational speed, particle number ratio and vessel fill level. The relative standard deviation (RSD) was employed as a measure of mixing performance. Results revealed that both the rotational speed of the impellers and the particle number ratio had a significant impact on the mixing performance. In addition, it was found that an increase in fill level and a decrease in impeller speed can lead to an increase in total particle contacts, indicating greater mixture compactness. The Peclet number and diffusivity coefficient were calculated in order to gain insight into the underlying mixing mechanism. The results indicated that diffusion was the dominant mixing mechanism, and the best mixing results were seen when the diffusion rates of both cube-shaped and cylinder-shaped particles were nearly equal.