ANN-PSO Modelling and Optimization to Reduce Surface Roughness in Additive Manufactured Sports Equipment

Fused Deposition Modelling (FDM) is an additive manufacturing (AM) technology that creates a wide variety of structures and complex 3D model data geometries. Due to the layered manufacturing process, poor surface roughness is the primary issue with FDM, limiting the direct applicability of FDM-printed parts to a significant number of applications such as sports equipment. There were few new meta-heuristic approaches for modelling and optimizing FDM surface roughness. As a result, a newly hybrid method ANN-PSO is developed to effectively assist practitioners in the prediction and optimization of FDM process parameters for improving fabricated part surface roughness. The experimental data from the previous study were used to train, validate, and test the artificial neural network (ANN) model, which finely maps the relationship between input process control factors and output response. The best network has a single hidden layer of 15 neurons. Thus, the 4-15-1 NN model was optimized by using the Particle Swarm Optimization (PSO) algorithm to find the best combination of parameter settings to improve the surface roughness of fabricated parts. The best values for layer height, print speed, print temperature, outer shell speed, and surface roughness were obtained by PSO, which were 0.2674 mm, 19.66 mm/s, 190.51 °C, 22.29 mm/s, and 2.0113 µm, respectively. As a result, the integrated ANN-PSO is a perfect combination for improving the surface finish of FDM printed parts especially for sports equipment.