The stability of coaxial powder feeding laser cladding process is affected by many factors, which makes it difficult to estimate the optimal process parameters. This study designs a central composite experiment, which considers the process parameters (laser power, powder feeding speed, and scanning speed) as input and outputs the characteristic parameters that reflect the cladding morphology and quality. The regression model and neural network in the response surface method are applied to the prediction of the single-pass cladding results, and their effects are compared. Based on this, a multi-objective optimization algorithm, i.e., the non-dominated sorting genetic algorithm II (NSGA-II), is used to optimize the three aforementioned process parameters. The results denote that the optimized process parameters can improve the surface hardness of the repaired parts by 17.11%, reduce the depth of the base heat-affected zone by 13.90%, and improve the cladding efficiency by 6.10%.