Fused filament fabrication manufactured biological scaffolds: An investigation of mechanical properties by using the Taguchi method and machine learning techniques.

Abstract

Tissue engineering scaffolds are three-dimensional, biocompatible, biodegradable, and porous structures designed to support cell attachment, proliferation, and differentiation, leading to new tissue formation. Designing optimal scaffolds is complex, requiring the optimisation of various physical, chemical, and biological properties, which are cell- or tissue-dependent. For hard tissue applications such as bone, compressive strength is a critical property and can be adjusted by modifying printing conditions. The mechanical properties of scaffolds produced using different microstructural polymers (semi-crystalline and amorphous) depend on parameters that significantly impact filament extrusion and the crystallisation process. This study investigates the effect of key process parameters (printing temperature, printing speed, and flow) on scaffold mechanical properties using the Taguchi method. Three biocompatible polymers with different microstructures-polycaprolactone, polylactic acid, and polyethylene terephthalate glycol-were examined. Results show a high correlation between process parameters and compressive strength using the Taguchi method, but prediction accuracy remained low. Therefore, four machine learning algorithms-Random Forest (RF), Support Vector Regression (SVR), K-Nearest Neighbor (K-NN), and Gradient Boosting Regression (GBR)-were applied to enhance predictive performance. Notably, the RF and GBR algorithms achieved approximately 99 % prediction accuracy when evaluated on the test dataset.