Filtern
Dokumenttyp
- Zeitschriftenartikel (5)
- Buchkapitel (2)
- Beitrag zu einem Tagungsband (1)
- Corrigendum (1)
Sprache
- Englisch (9)
Schlagworte
- Machine Learning (3)
- Thermography (3)
- Computer-aided design (2)
- Porosity prediction (2)
- 3D image processing (1)
- 3D image registration (1)
- Additive Manufacturing (1)
- Additive manufacturing (1)
- Computed tomography (1)
- Data Compression (1)
Organisationseinheit der BAM
Paper des Monats
- ja (1)
The formation of flaws such as internal porosity in parts produced by Metal-based Powder Bed Fusion with Laser Beam (PBF-LB/M) significantly hinders its broader industrial application, as porosity can potentially lead to part failure. Addressing this issue, this study explores the efficacy of in-situ thermography, particularly short-wave infrared thermography, for detecting and predicting porosity during manufacturing. This technique is capable of monitoring the part’s thermal history which is closely connected to the flaw formation process. Recent advancements in Machine Learning (ML) have been increasingly leveraged for porosity prediction in PBF-LB/M. However, previous research primarily focused on global rather than localized porosity prediction which simplified the complex prediction task. Thereby, the opportunity to correlate the predicted flaw position with expected part strain to judge the severity of the flaw for part performance is neglected. This study aims to bridge this gap by studying the potential of SWIR thermography for predicting local porosity levels using regression models. The models are trained on data from two identical HAYNES®282® specimens. We compare the effectiveness of feature-based and raw data-based models in predicting different porosity types and examine the importance of input data in porosity prediction. We show that models trained on SWIR thermogram data can identify systematic trends in local flaw formation. This is demonstrated for forced flaw formation using process parameter shifts and, moreover, for randomly formed flaws in the specimen bulk. Furthermore, we identify features of high importance for the prediction of lack-of-fusion and keyhole porosity from SWIR monitoring data.
In this study, we present an enhanced deep learning framework for the prediction of porosity based on thermographic in-situ monitoring data of laser powder bed fusion processes. The manufacturing of two cuboid specimens from Haynes 282 (Ni-based alloy) powder was monitored by a short-wave infrared camera. We use thermogram feature data and x-ray computed tomography data to train a convolutional neural network classifier. The classifier is used to perform a multi-class prediction of the spatially resolved porosity level in small sub-volumes of the specimen bulk.
Laser powder bed fusion is one of the most promising additive manufacturing techniques for printing complex-shaped metal components. However, the formation of subsurface porosity poses a significant risk to the service lifetime of the printed parts. In-situ monitoring offers the possibility to detect porosity already during manufacturing. Thereby, process feedback control or a manual process interruption to cut financial losses is enabled.
Short-wave infrared thermography can monitor the thermal history of manufactured parts which is closely connected to the probability of porosity formation. Artificial intelligence methods are increasingly used for porosity prediction from the obtained large amounts of complex monitoring data. In this study, we aim to identify the potential and the challenges of deep-learning-assisted porosity prediction based on thermographic in-situ monitoring.
Therefore, the porosity prediction task is studied in detail using an exemplary dataset from the manufacturing of two Haynes282 cuboid components. Our trained 1D convolutional neural network model shows high performance (R² score of 0.90) for the prediction of local porosity in discrete sub-volumes with dimensions of (700 x 700 x 40) μm³.
It could be demonstrated that the regressor correctly predicts layer-wise porosity changes but presumably has limited capability to predict differences in local porosity. Furthermore, there is a need to study the significance of the used thermogram feature inputs to streamline the model and to adjust the monitoring hardware. Moreover, we identified multiple sources of data uncertainty resulting from the in-situ monitoring setup, the registration with the ground truth X-ray-computed tomography data and the used pre-processing workflow that might influence the model’s performance detrimentally.
A critical challenge in Additive Manufacturing is to ensure the safety and dimensional accuracy of produced parts. Since quality control is made by means of different online and offline imaging techniques (e.g. Thermography, X-ray and Optical Computer Tomography), image registration plays an important role in addressing these challenges. This paper introduces the main motivation, challenges, and research gaps of image registration in Additive Manufacturing. Furthermore, it introduces the main transformations, registration methods, similarity matrices and accuracy measurement. The main aim of the paper is to present a comprehensive review on the available methods for image registration in Additive Manufacturing based on the measurement techniques. Various registration methods, including landmark-based, point cloud-based, image intensity-based, and shape-based techniques, are examined in their applications for quality control, defect detection, and distortion quantification.
In Additive Manufacturing (AM), precise rigid three-dimensional (3D) image registration between X-ray Computed Tomography (XCT) scans and Computer-Aided Design (CAD) models is an important step for the quantification of distortions in produced parts. Given the absence of standardized benchmarks for image registration in AM, we introduce a gold standard for 3D image registration, using a reference base plate beneath the build structure. This gold standard is used to quantify the accuracy of rigid registration, with a proof of concept demonstrated in PBF-LB/M. In this study, we conduct a comparative analysis of various rigid 3D registration methods useful for quality assurance of PBF-LB/M parts including feature-based, intensity-based, and point cloud-based approaches. The performance of each registration method is evaluated using measures of alignment accuracy based on the gold standard and computational efficiency. Our results indicate significant differences in the efficacy of these methods, with point cloud based Coherent Point Drift (CPD) showing superior performance in both alignment and computational efficiency. The rigidly registered 3D volumes are used to estimate the deformation field of the printed parts relative to the nominal CAD design using Digital Volume Correlation (DVC). The quality of the estimated deformation field is assessed using the Dice score metric. This study provides insights into methods for enhancing the precision and reliability of AM process.
We present a 3D image registration technique for non-linear deformation estimation in Additive Manufacturing processes. The methodology involves comparing X-ray Computed Tomography (XCT) data with Computer Aided Design (CAD) models for Triply Periodic Minimal Surface (TPMS) lattices and employs the Chamfer distance to refine mesh non-linear deformations.
Quality control in Additive Manufacturing (AM) is vital for industrial applications in areas such as the automotive, medical, and aerospace sectors. Geometric inaccuracies caused by shrinkage and deformations can compromise the life and performance of additively manufactured components. Such deviations can be quantified using Digital Volume Correlation (DVC), which compares the Computer-Aided Design (CAD) model with the X-ray Computed Tomography (XCT) geometry of the components produced. However, accurate registration between the two modalities is challenging due to the absence of a ground truth or reference deformation field. In addition, the extremely large data size of high-resolution XCT volumes makes computation difficult. In this work, we present a deep learning-based approach for estimating voxel-wise deformations between CAD and XCT volumes. Our method uses a dynamic patch-based processing strategy to handle high-resolution volumes. In addition to the Dice score, we introduce a Binary Difference Map (BDM) that quantifies voxel-wise mismatches between binarized CAD and XCT volumes to evaluate the accuracy of the registration. Our approach shows a 9.2% improvement in the Dice score and a 9.9% improvement in the voxel match rate compared to classic DVC methods, while reducing the interaction time from days to minutes. This work sets the foundation for deep learning-based DVC methods to generate compensation meshes that can then be used in closed-loop correlations during the AM production process. Such a system would be of great interest to industry, as it would make the manufacturing process more reliable and efficient, saving time and material.
The ever-growing volume of data in imaging sciences stemming from advancements in imaging technologies, necessitates efficient and reliable storage solutions for such large datasets. This study investigates the compression of industrial X-ray computed tomography (XCT) data using deep learning autoencoders and examines how these compression algorithms affect the quality of the recovered data. Two network architectures with different compression rates were used, a deep convolution neural network (D-CNN) and a vector quantized variational autoencoder (VQ-VAE). The XCT data used was from a sandstone sample with a complex internal pore network as a good test case for the importance of feature preservation. The quality of the decoded images obtained from the two different deep learning architectures with different compression rates were quantified and compared to the original input data. In addition, to improve image decoding quality metrics, we introduced a metric sensitive to edge preservation, which is crucial for three-dimensional data analysis. We showed that different architectures and compression rates are required depending on the specific characteristics needed to be preserved for later analysis. The findings presented here can aid scientists in determining the requirements and strategies needed for appropriate data storage and analysis.
