Functional multi-response robust optimization for additive manufacturing process

In the context of the 3D printing process and product quality design, effectively addressing the modeling and parameter optimization problems of functional multi-response, particularly multiple profile responses, remains a significant challenge. Various methods have been developed to tackle these challenges, including functional data analysis and tensor regression. However, existing methods are inadequate for simultaneously handling the modeling and robust optimization problems of functional multi-response, especially when the functional multi-response is densely sampled. In light of the current research status, this study introduces a tensor regression method designed to address the sampling bias problem arising from a limited number of samples in functional multi-response. This method employs a robust optimization technique to determine the optimal process parameter settings. Initially, the high-dimensional parameter tensor in the tensor regression model is decomposed into low-rank components using the CP decomposition method. Subsequently, the parameters are jointly estimated using the Alternating Direction Method of Multipliers (ADMM) and Block Coordinate Descent (BCD) methods. Finally, the Euclidean distance and tensor regression model are utilized to construct a robust optimization model, and a hybrid genetic algorithm is employed to obtain the optimal process parameter settings. Case studies on additive manufacturing validate the reliability an d robustness of the proposed method.