Study of multi-modal learning-empowered effect analysis of process parameters on tensile properties of 3D-printed composites

The fused deposition modeling (FDM) process, as a mainstream 3D printing technique, has attracted extensive attention in the field of additive manufacturing because of its high forming efficiency and lightweight advantages. However, printing parameters significantly influence the mechanical performance of specimens, and the complex interactions among them make it difficult to accurately establish a mapping relationship between parameters and performance. To address this issue, a prediction method based on a multi-modal convolutional neural network (MM-CNN) was developed, which achieves precise modeling by integrating different types of input features. Taking a polylactic acid–carbon fiber composite (CF/PLA composite) as the research object, this study systematically investigated the influence of four key printing parameters—layer thickness, nozzle temperature, material flow, and printing speed—on the tensile properties of printed specimens. A total of 256 parameter combinations were designed, and tensile tests were conducted to construct a multi-modal dataset composed of numerical parameters and image representations, which was then employed as input to the model for feature fusion and performance prediction. The results show that the proposed model effectively captures the complex nonlinear relationship between parameters and mechanical performance. The model was evaluated in terms of accuracy, precision, recall, and F1-score, and all the indicators remained above 0.95, representing an improvement of about 17.3% in prediction accuracy compared with traditional convolutional neural network (CNN) models. These findings demonstrate that the MM-CNN can effectively predict and optimize the mechanical performance of different parameter combinations, providing a reliable and generalizable approach for process optimization in FDM-based additive manufacturing.