Abstract:Additive Manufacturing (AM) has emerged as a manufacturing process that allows the direct production of samples from digital models. To ensure that quality standards are met in all manufactured samples of a batch, X-ray computed tomography (X-CT) is often used combined with automated anomaly detection. For the latter, deep learning (DL) anomaly detection techniques are increasingly, as they can be trained to be robust to the material being analysed and resilient towards poor image quality. Unfortunately, most recent and popular DL models have been developed for 2D image processing, thereby disregarding valuable volumetric information. This study revisits recent supervised (UNet, UNet++, UNet 3+, MSS-UNet) and unsupervised (VAE, ceVAE, gmVAE, vqVAE) DL models for porosity analysis of AM samples from X-CT images and extends them to accept 3D input data with a 3D-patch pipeline for lower computational requirements, improved efficiency and generalisability. The supervised models were trained using the Focal Tversky loss to address class imbalance that arises from the low porosity in the training datasets. The output of the unsupervised models is post-processed to reduce misclassifications caused by their inability to adequately represent the object surface. The findings were cross-validated in a 5-fold fashion and include: a performance benchmark of the DL models, an evaluation of the post-processing algorithm, an evaluation of the effect of training supervised models with the output of unsupervised models. In a final performance benchmark on a test set with poor image quality, the best performing supervised model was UNet++ with an average precision of 0.751 $\pm$ 0.030, while the best unsupervised model was the post-processed ceVAE with 0.830 $\pm$ 0.003. The VAE/ceVAE models demonstrated superior capabilities, particularly when leveraging post-processing techniques.

Voxel-wise classification for porosity investigation of additive manufactured parts with 3D unsupervised and (deeply) supervised neural networks

Ultrasonic Array Tomography-Oriented Subsurface Crack Recognition and Cross-Section Image Reconstruction of Reinforced Concrete Structure Using Deep Neural Networks

3D total variation denoising in X-CT imaging applied to pore extraction in additively manufactured parts

Attenuation-adjusted deep learning of pore defects in 2D radiographs of additive manufacturing powders

Porosity Evaluation of Additively Manufactured Components Using Deep Learning-based Ultrasonic Nondestructive Testing

Direct voxel classification from X-ray projections for 3D pore detection applied to laser sintered parts

Automatic Volumetric Segmentation of Additive Manufacturing Defects with 3D U-Net

Volumetric defect classification in Nano-resolution X-ray computed tomography images of laser powder bed fusion via deep learning

Deep learning with mixup augmentation for improved pore detection during additive manufacturing

In situ porosity intelligent classification of selective laser melting based on coaxial monitoring and image processing

3-Dimensional residual neural architecture search for ultrasonic defect detection

Detecting and classifying hidden defects in additively manufactured parts using deep learning and X-ray computed tomography

Optimizing porosity detection in wire laser metal deposition processes through data-driven AI classification techniques

Qualify assessment for extrusion-based additive manufacturing with 3D scan and machine learning

ThermoPore: Predicting Part Porosity Based on Thermal Images Using Deep Learning

Three-Dimensional Residual Neural Architecture Search for Ultrasonic Defect Detection

In-Situ Laser-Based Process Monitoring and In-Plane Surface Anomaly Identification for Additive Manufacturing Using Point Cloud and Machine Learning

In-situ porosity prediction in metal powder bed fusion additive manufacturing using spectral emissions: a prior-guided machine learning approach

Automated porosity segmentation in laser powder bed fusion part using computed tomography: a validity study

Machine Learning Driven Instance Segmentation Providing New Porosity Insights into Wire Arc Directed Energy Deposited Ti-22V-4Al

A Hybrid Deep Generative Network for Pore Morphology Prediction in Metal Additive Manufacturing