Abstract: Automated fiber placement (AFP) plays an important role in fabricating high-performance carbon fiber reinforced polymer (CFRP) composite products, whose quality is largely influenced by process parameters. The lay-up quality of CFRP composite prepregs manufactured by the self-built AFP machine was discussed, and the influence of process parameters on it was analyzed. The relationship between AFP properties (bonding condition between prepregs, deformation of prepregs, and number of bubbles) and process parameters (tool temperature, placement speed, and compaction pressure) were analyzed by single-factor experimental method. Based on this foundation, the regression model was then established using artificial neural network (ANN), whose optimal architecture was identified by particle swarm optimization (PSO) to improve predictive accuracy. The three-objective optimization problem was solved by multi-objective particle swarm optimization (MOPSO) to get the optimal Pareto set. Complex proportional assessment (COPRAS) combined with analytic hierarchy process (AHP) was applied to determine the best parameter combination from the alternative set.