Prediction-based multi-objective optimization method for 3D printing resource consumption

A prediction-based multi-objective optimization (PBMO) method is proposed in this paper to forecast and reduce 3D printing (3DP) resources on demand, including time, energy, and material. In the authors’ previous research work, a hybrid code-based and data-driven modeling (HCDM) scheme was proposed to customize the predictive models based on process parameters, material deposition paths, and machine behaviors. This study further utilizes the models as multi-objectives to be minimized, aiming at the appropriate solution of process parameters that consume the least resources. Non-dominated sorting genetic algorithm II (NSGA-II), one of the commonly used metaheuristic algorithms, is adopted to construct the PBMO framework, where the HCDM process is embedded in the fitness evaluation step. The corresponding computing program is compiled and then validated on two material extrusion (MEX) machines. Based on the optimization results, hypervolume, as a Lebesgue measure, is used to evaluate the superiorities of all near-optimal solutions, thereby recommending the best-performing solutions for real 3DP. Apart from the 3DP process, the proposed optimization method is adaptable to other mainstream computer numerical control (CNC) manufacturing processes and will guide process design to promote resource conservation for cleaner production.