Monday 08.12.25

at 11:30 in the 1rd floor gallery in Amado building

Wood-based Bio-Composite Liquid Deposition Modelling (LDM) enables sustainable, low-energy, customizable fabrication using abundant, biodegradable, locally sourced materials. Existing research lacks a clear understanding of how these materials deform after drying. Current analytical methods capture volumetric deformation but fail to resolve localized deformation patterns and offer no mechanism for linking these effects back to the toolpath for deformation compensation.

This research develops a Machine Learning (ML) approach to predict deformation in wood-based LDM using Daika, a commercially available bio-composite. Digital Image Correlation (DIC) was used to capture more than 50,000 localized displacement points per printed sample. These data were mapped to the 3D-printing toolpath and used to train and evaluate multiple ML models. Classic regression models, Multi-Layer Perceptrons (MLPs), and Convolutional Neural Networks (CNNs) were tested for their ability to forecast final deformed geometries.

The best-performing model achieved 97–98% prediction accuracy across axes and geometries, enabling precise dimensional control and improved design fidelity. These results enhance the reliability and scalability of sustainable additive manufacturing by predicting post-curing deformation, transforming wood waste into high-performance, customizable construction components.

Contact information: Kushagra Gautam  |    +972 58-559-3303 |   e-mail: kushagragautam09@gmail.com