# Additively Manufactured Carbon-Reinforced ABS Honeycomb Composite Structures and Property Prediction by Machine Learning

## TL;DR

This paper explores 3D-printed carbon-reinforced ABS honeycomb structures and uses machine learning to predict their properties.

## Contribution

The study introduces a machine learning model to predict properties of carbon-reinforced polymer composites based on print parameters.

## Key findings

- CNT-ABS achieved higher ultimate compressive strength due to reduced porosity.
- A machine learning model with 0.92 accuracy and 0.96 coefficient of determination was developed for property prediction.
- Compressive strength decreased with increasing hex size across all materials.

## Abstract

The expansive utility of polymeric 3D-printing technologies and demand for high- performance lightweight structures has prompted the emergence of various carbon-reinforced polymer composite filaments. However, detailed characterization of the processing–microstructure–property relationships of these materials is still required to realize their full potential. In this study, acrylonitrile butadiene styrene (ABS) and two carbon-reinforced ABS variants, with either carbon nanotubes (CNT) or 5 wt.% chopped carbon fiber (CF), were designed in a bio-inspired honeycomb geometry. These structures were manufactured by fused filament fabrication (FFF) and investigated across a range of layer thicknesses and hexagonal (hex) sizes. Microscopy of material cross-sections was conducted to evaluate the relationship between print parameters and porosity. Analyses determined a trend of reduced porosity with lower print-layer heights and hex sizes compared to larger print-layer heights and hex sizes. Mechanical properties were evaluated through compression testing, with ABS specimens achieving higher compressive yield strength, while CNT-ABS achieved higher ultimate compressive strength due to the reduction in porosity and subsequent strengthening. A trend of decreasing strength with increasing hex size across all materials was supported by the negative correlation between porosity and increasing print-layer height and hex size. We elucidated the potential of honeycomb ABS, CNT-ABS, and ABS-5wt.% CF polymer composites for novel 3D-printed structures. These studies were supported by the development of a predictive classification and regression supervised machine learning model with 0.92 accuracy and a 0.96 coefficient of determination to help inform and guide design for targeted performance.