Strain sensing characteristics of 3D-printed carbon nanotubes/polypyrrole/UV-curable composites: experimental validation and machine learning predictions

The integration of wearable electronic devices, particularly strain sensors, with 3D printing technology has gained significant interest due to the versatility and adaptability of additive manufacturing (AM) processes. This article explores the advancements and challenges associated with the integration of wearable electronics and 3D printing, focusing on the potential of carbon-based materials, specifically multi-wall carbon nanotubes (MWCNTs) and polypyrrole (PPy) composites, in enhancing the electrical conductivity and mechanical properties of printed objects. The study employed a comprehensive experimental validation and machine learning predictions to elucidate the strain sensing behaviour of the 3D-printed FPU/MWCNTs/PPy composites. Through a comprehensive experimental analysis, it was demonstrated that the addition of MWCNTs improves mechanical properties, while the incorporation of PPy enhances electrical properties. However, higher concentrations of MWCNTs result in agglomeration and void structures, and the addition of PPy leads to a decline in mechanical performance. Moreover, leveraging machine learning techniques, several machine learning (ML) models are optimised in a regression task to predict the relative resistance change, ∆R/R0 using input features of CNT, PPy, and elongation. The extra trees regressor (ETR) achieved superior performance among the considered models. SHAP analysis revealed a direct relationship between input features and the target property, with feature importance ranking as CNT > PPy > elongation. These findings aligned with experimental results, and the optimised ETR model exhibits accurate predictions, highlighting its ability to capture complex relationships. This demonstrates the potential of ML to expedite advancements in optimising polymer additive formulations.