The particle rebound behaviors of particle-wall collisions have significant impacts on the particle motion and the separation efficiency in the gas-solid separation process. Previous studies have focused on the collision behavior of the spherical particles. However, inactual industrial processes, particles such as coal powder, biomass, and ore are all non-spherical particles. There are significant differences in the rebound behavior between the non-spherical particles colliding with the wall and the spherical particles. To explore the rebound behavior of the non-spherical particles colliding with the wall, an experimental device for particle-wall collisions was established.High-speed photography and image processing methods were used to obtain basic data of particle-wall collisions of the non-spherical particles. The influence of the key parameters such as particle material, sphericity, wall roughness, impact angle, and impact speed on particle-wall rebound behavior was analyzed. Based on the established four - parameter model of particle - wall collisions and neural network models, the rebound behavior between non-spherical particles and the wall was predicted. The results indicate that there is consistency in the rebound behavior of non-spherical particles colliding with the wall. Sphericity plays an important role in particle-wall collisions. The four-parameter model can predict collision results and random distribution characteristics well, while neural network modelstrained based on experimental data can achieve better prediction results.