additive manufacturing defect detection machine learning

Introduction

Additive manufacturing (AM), also known as 3D printing, is a transformative technology that enables the creation of complex parts layer by layer. However, ensuring the quality of these parts is challenging due to the potential for defects during the manufacturing process. Machine learning (ML) has emerged as a powerful tool for defect detection in AM, offering the potential for real-time monitoring and correction of defects.

Key Insights

In-Situ Defect Detection Using Supervised Machine Learning:
- High-resolution imaging combined with supervised ML techniques, such as support vector machines (SVM), can achieve in-situ defect detection with accuracies greater than 80% during cross-validation experiments.
- Integrating thermal simulations with in-situ sensor data in a supervised ML framework can detect flaws with an F-score approaching 90%, demonstrating the effectiveness of a digital twin approach.
Convolutional Neural Networks (CNN) for Defect Detection:
- CNNs, particularly models like VGG16 and Xception, have shown high accuracy (up to 95.8%) in classifying defects in selective laser sintering (SLS) processes using small datasets and transfer learning techniques.
Point Cloud Processing and Machine Learning:
- In-situ point cloud processing combined with ML methods can achieve rapid surface defect identification with an accuracy of 93.15% in directed energy deposition (DED) processes.
- Synthetic 3D point clouds used for training ML models can effectively detect geometric defects, outperforming traditional methods like the Z-difference method.
Integrated Frameworks and Hybrid Models:
- Combining vision-based and sensor-based data in ML models significantly improves defect classification accuracy compared to using a single data source.
- An integrated framework using computer vision and ML algorithms, enhanced by Generative Adversarial Networks (GAN) for dataset augmentation, can effectively detect defects in real-time.
Advanced ML Algorithms for Enhanced Detection:
- Improved YOLOv3 algorithms, incorporating techniques like weighted k-means clustering and MobileNetv3, enhance detection accuracy and speed for surface defects in metal AM.

Conclusion

Machine learning has proven to be a valuable asset in the field of additive manufacturing for defect detection. Techniques such as supervised learning with high-resolution imaging, convolutional neural networks, and integrated frameworks combining multiple data sources have shown high accuracy and efficiency in identifying defects. The use of advanced algorithms and synthetic data for training further enhances the capability of ML models to detect and classify defects in real-time, paving the way for more reliable and high-quality additive manufacturing processes.

Additive manufacturing defect detection machine learning

Summary

Introduction

Key Insights

Conclusion

Sources and full results

Application of supervised machine learning for defect detection during metallic powder bed fusion additive manufacturing using high resolution imaging.

In situ defect detection using high-resolution imaging in powder bed fusion additive manufacturing can achieve accuracies of over 80%, enabling in-process part qualification and enabling potential for in-process re-melting and correction of detected defects.

Toward the digital twin of additive manufacturing: Integrating thermal simulations, sensing, and analytics to detect process faults

Integrating thermal simulations and in-situ sensor data in a machine learning model significantly improves the detection of flaws in additive manufacturing processes.

A machine learning method for defect detection and visualization in selective laser sintering based on convolutional neural networks

Convolutional neural networks (CNN) effectively detect and visualize powder bed defects in selective laser sintering processes, offering an alternative method for non-destructive quality assurance and manufacturing documentation.

Rapid surface defect identification for additive manufacturing with in-situ point cloud processing and machine learning

This study proposes a rapid surface defect identification method for additive manufacturing using in-situ point cloud processing and machine learning, achieving an accuracy of 93.15%.

Machine Learning in Additive Manufacturing: A Review

Machine learning can improve parameter optimization and anomaly detection in additive manufacturing, with potential future research directions.

In-Situ Defect Detection of Metal Additive Manufacturing: An Integrated Framework

This paper proposes an integrated system for in-situ monitoring and defect detection in metal additive manufacturing, using off-axis cameras and GANs for dataset augmentation.

Geometrical defect detection for additive manufacturing with machine learning models

The proposed machine learning-based scheme effectively detects geometric defects in additively manufactured objects, outperforming existing methods and saving training time and costs.

Towards Machine Learning for Error Compensation in Additive Manufacturing

Machine learning algorithms can enhance additive manufacturing by detecting product defects and improving design complexity, prefabrication, and security in the cyber-physical system.

A Combination of Vision- and Sensor-Based Defect Classifications in Extrusion-Based Additive Manufacturing

Combining vision and sensor-based machine learning models significantly improves defect classification accuracy in 3D printing processes.

Fast and Accurate Defects Detection for Additive Manufactured Parts by Multispectrum and Machine Learning.

The improved YOLOv3 algorithm using multispectrum and machine learning improves surface defect detection in metal additive manufacturing with high accuracy and fast speed.

Try another search