Search:
Match:
1 results

Analysis

This paper addresses the challenge of predicting multiple properties of additively manufactured fiber-reinforced composites (CFRC-AM) using a data-efficient approach. The authors combine Latin Hypercube Sampling (LHS) for experimental design with a Squeeze-and-Excitation Wide and Deep Neural Network (SE-WDNN). This is significant because CFRC-AM performance is highly sensitive to manufacturing parameters, making exhaustive experimentation costly. The SE-WDNN model outperforms other machine learning models, demonstrating improved accuracy and interpretability. The use of SHAP analysis to identify the influence of reinforcement strategy is also a key contribution.
Reference

The SE-WDNN model achieved the lowest overall test error (MAPE = 12.33%) and showed statistically significant improvements over the baseline wide and deep neural network.