Low-Cost Vibration Monitoring and Fault Detection for Material Extrusion Process

In the context of Industry 4.0, leveraging new technology to enhance production processes is crucial. This study explores the application of cost-effective dual accelerometers in a Material extrusion 3D printer to monitor vibrations and predict potential issues, aiming to develop a system capable of anomaly detection. Fast Fourier transform and wavelet transform were used for signal processing of the accelerometer for identifying key vibration patterns, analyzed using statistical methods and peak detection. Machine learning was used for training for identifying the faults. Models like Linear Regression, Random Forest, and Support Vector Regression were employed to predict vibration-related anomalies and print quality deviations based on extracted features from the sensor data. These models were optimized through hyperparameter tuning and feature selection, with SVR and Random Forest showing excellent predictive capabilities, achieving Mean Squared Errors as low as 0.0948. The study highlights the importance of low-cost vibration monitoring and analytics in enhancing 3D printing processes, promising significant improvements in reliability and efficiency.