Experimental Evaluation and Machine Learning-Based Prediction of Laser Cutting Quality in FFF-Printed ABS Thermoplastics.

Additive manufacturing, particularly Fused Filament Fabrication (FFF), provides notable advantages such as design flexibility and efficient material usage. However, components produced via FFF often exhibit suboptimal surface quality and dimensional inaccuracies. Acrylonitrile Butadiene Styrene (ABS), a widely used thermoplastic in FFF applications, commonly necessitates post-processing to enhance its surface finish and dimensional precision. This study investigates the effects of CO(2) laser cutting on FFF-printed ABS plates, focusing on surface roughness, top and bottom kerf width, and bottom heat-affected zone. Forty-five experimental trials were conducted using different combinations of plate thickness, cutting speed, and laser power. Measurements were analysed statistically, and analysis of variance was applied to determine the significance of each parameter. To enhance prediction capabilities, seven machine learning models-comprising traditional (Linear Regression and Support Vector Regression), ensemble (Extreme Gradient Boosting and Random Forest), and deep learning algorithms (Long Short-Term Memory (LSTM), LSTM-Gated Recurrent Unit (LSTM-GRU), LSTM-Extreme Gradient Boosting (LSTM-XGBoost))-were developed and compared. Among these, the LSTM-GRU model achieved the highest predictive performance across all output metrics. Results show that cutting speed is the dominant factor affecting cutting quality, followed by laser power and thickness. The proposed experimental-computational approach enables accurate prediction of laser cutting outcomes, facilitating optimisation of post-processing strategies for 3D-printed ABS parts and contributing to improved precision and efficiency in polymer-based additive manufacturing.