A team from Notre Dame has demonstrated a machine learning-assisted 3D printing process for manufacturing thermoelectric materials with unprecedented performance. By combining extrusion-based printing with Bayesian optimization, they produced a Bismuth Antimony Telluride (Bisbte) structure showing the highest value of room temperature ZT reported in printed thermoelectrics.
Published in the Journal of Materials Chemistry A, this study introduces a data-driven strategy to quickly identify the optimal ink formulation and printing parameters. The team adopted Gaussian Process Regression (GPR) to predict thermal power factors, implement printability constraints to support vector machines (SVMs), ensuring high geometric fidelity. Optimized inks have been important for applications such as waste heat recovery of irregular surfaces, allowing for the creation of complex 3D structures such as curved tubes and lattices, with 83 wt% visceral particles, 0.5 wt% Xanthan gum in water, 1.4 mm filament spacing and 1.0 mm standoff distance, allowing for the creation of complex 3D structures such as curved tubes and lattices.
Machine learning accelerates thermoelectric development
The team's Bayesian Optimization Framework addressed the critical challenges of thermoelectric printing. Competing demand for particle loading (enhancing conductivity and reducing printability) and rheological modifications (improving shape fidelity but improving porosity). By modeling the four-dimensional parameter space, the algorithm identified the optimal conditions that human intuition may have missed.
The Gaussian Process Regression (GPR) model showed strong prediction accuracy for power factors. In this case, the predicted experimental values and experimental values are closely matched over multiple optimization rounds. Bayesian optimization reduced the experimental iterations required to identify the optimal parameters, but this paper does not quantify accurate time savings. ”
Scalable and shape forced device
The researchers demonstrated printing of a complex heat exchange-inspired architecture, including a 60° tilted tube that can fit curved surfaces, a hexagonal honeycomb grating for improved mechanical stability, and a multi-layer spiral structure that optimizes surface area to volume ratios.
Post-treatment via a high temperature etiometric press (hip joint) at 480°C below 200 MPa pressure maintained the printed geometry while increasing material density. SEM/EDS analysis revealed the separation of Tellurium at grain boundaries, a microstructural feature that contributes to high ZT. The paper notes that although it does not specify an exact dimensional deviation or density percentage, the hip joint achieved comparable thermoelectric performance to cold pressed samples while maintaining better structural integration.
Water-based ink systems offer potential cost advantages over organic solvent-based alternatives by eliminating expensive solvents and simplifying the processing. The use of aqueous chemistry and Xanthan gum suggests material savings compared to traditional approaches.
Machine learning and 3D printing optimization
Machine learning is increasingly being used to optimize 3D printing processes across materials and applications. Researchers at the University of Toronto have recently introduced a machine learning framework to optimize laser parameters for metal additive manufacturing, improve quality and reduce material waste. In the energy sector, Bambu Lab collaborated with researchers to develop an AI model that predicts the production of wind turbine blades based on 3D printing performance metrics.
Elsewhere, another study showed how machine learning can track the origins of 3D printed parts with 98% accuracy, providing new features of quality assurance and traceability. These advances highlight how AI tools are becoming essential for predictive control and smart manufacturing in the additional processes.
what 3D printing trends Do you need to be careful in 2025?
What do you think? The future of 3D printing Shape it up?
Don't forget to sign up to stay up to date with the latest 3D printing news 3D Printing Industry Newsletter Or follow us Twitteror like our page Facebook.
Why not subscribe to us while you're here YouTube channel? It features discussions, debriefs, video shorts and webinar replays.
The distinctive image shows a printed 3D composite structure. Song et al./Journal of Materials Chemistry A.
