Adaptive neuro-fuzzy modeling and vibration analysis of high-speed 3D-printing for prototyping

Abstract

This paper describes a novel integration of the Adaptive Neuro-Fuzzy Inference System (ANFIS) approach with vibration analysis for modeling and predicting the dimensional integrity and surface roughness of Polylactic Acid (PLA) parts produced by high-speed Fused Filament Fabrication (FFF) for rapid prototyping. A full factorial design was implemented considering three layer thicknesses, three printing speeds and three infill patterns. In parallel, vibration measurements were captured during printing using an ESP32 microcontroller and an ADXL345 accelerometer. The acquired three-axis acceleration signals were processed through Principal Component Analysis (PCA) to derive a single representative vibration component for comparative analysis. The developed ANFIS models exhibited strong predictive performance, with coefficients of determination (R²) exceeding 0.9 for both dimensional accuracy and surface roughness. The results reveal that the interaction between printing speed and infill pattern strongly affects dimensional error, while the interaction between the speed and layer height contributes toward surface roughness. Vibration analysis further demonstrated that the choice of infill patterns significantly affects the overall vibration magnitude, and the lines pattern yields the lowest values. The inclusion of vibration characterization via Fast Fourier Transform (FFT) and PCA further enhances the understanding of process dynamics by a data-driven approach. The innovative integration of ANFIS modeling with vibration analysis provides a baseline for modeling and part quality verification for high-speed prototyping applications.