The primary objective of this study was to demonstrate the effectiveness of machine learning (ML) in predicting the mechanical behavior of sandwich structures. To achieve this, the study focused on assessing the bending response of environmentally friendly sandwich beams consisting of auxetic cores made from polylactic acid (PLA) and flax/epoxy composite face sheets reinforced with halloysite nanotubes (HNTs). Two ML techniques, specifically shallow neural networks (SNNs) and deep neural networks (DNNs), were employed to predict the specific energy absorption (SEA) and load‐deflection curves of these sandwich beams, respectively. The key design parameters under consideration included the HNT content in the face sheets and three geometric characteristics of the auxetic cells. Subsequently, 16 distinct specimens were meticulously designed for manufacturing, following Taguchi's experimental design principles. The cores of the structures were produced using 3D printing techniques, while the face sheets were meticulously fabricated using a hand layup process. These prepared specimens were then subjected to a three‐point bending test, and the collected data were employed to train the aforementioned neural networks. The outcomes of this study revealed that an SNN with a single hidden layer comprising seven neurons effectively predicted the SEA of the structures across various design parameter values. Additionally, the remarkable performance of a DNN, consisting of five hidden layers with 128, 64, 32, 16, and 8 units, respectively, was demonstrated by comparing its predicted results with the experimental results for a randomly designed sandwich beam.Highlights
ML techniques were used for bending behavior of sandwich beams.
The beams were made of flax/epoxy/HNT face sheets and 3D‐printed auxetic core.
SNN was successfully employed to predict the SEA of the structures.
DNN effectively predicted the load‐deflection curves.
HNT content and auxetic cells significantly impacted the bending properties.
