Datei im Netzwerk der BAM verfügbar ("Closed Access")
Filtern
Dokumenttyp
- Vortrag (9)
- Beitrag zu einem Tagungsband (4)
- Zeitschriftenartikel (1)
- Corrigendum (1)
- Posterpräsentation (1)
Schlagworte
Organisationseinheit der BAM
Paper des Monats
- ja (1)
Eingeladener Vortrag
- nein (9)
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.
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.
For the wide acceptance of the use of additive manufacturing (AM), it is required to provide reliable testing methods to ensure the safety of the additively manufactured parts. A possible solution could be the deployment of in-situ monitoring during the build process. However, for laser powder bed fusion using metal powders (PBF-LB/M) only a few in-situ monitoring techniques are commercially available (optical tomography, melt pool monitoring), which have not been researched to an extent that allows to guarantee the adherence to strict quality and safety standards.
In this contribution, we present results of a study of PBF-LB/M printed parts made of the nickel-based superalloy Haynes 282. The formation of defects was provoked by local variations of the process parameters and monitored by thermography, optical tomography and melt pool monitoring. Afterwards, the defects were characterized by computed tomography (CT) to identify the detection limits of the used in-situ techniques.
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.
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.
The appearance of irregularities such as keyhole porosity is a major challenge for the production of metal parts by laser powder bed fusion (PBF-LB/M). The utilization of thermographic in-situ monitoring is a promising approach to extract the thermal history which is closely related to the formation of irregularities. In this study, we investigate the utilization of convolutional neural networks to predict keyhole porosity based on thermographic features. Here, the porosity information calculated from an x-ray micro computed tomography scan is used as reference. Feature engineering is performed to enable the model to learn the complex physical characteristics of the porosity formation. The model is examined with regard to the choice of hyperparameters, the significance of thermal features and characteristics of the data acquisition. Based on the results, future demands on irregularity prediction in PBF-LB/M are derived.
The formation of irregularities such as keyhole porosity pose a major challenge to the manufacturing of metal parts by laser powder bed fusion (PBF-LB/M). In-situ thermography as a process monitoring technique shows promising potential in this manner since it is able to extract the thermal history of the part which is closely related to the formation of irregularities. In this study, we investigate the utilization of machine learning algorithms to detect keyhole porosity on the base of thermographic features. Here, as a referential technique, x-ray micro computed tomography is utilized to determine the part's porosity. An enhanced preprocessing workflow inspired by the physics of the keyhole irregularity formation is presented in combination with a customized model architecture. Furthermore, experiments were performed to clarify the role of important parameters of the preprocessing workflow for the task of defect detection . Based on the results, future demands on irregularity prediction in PBF-LB/M are derived.
Quality assessment of components produced by metal based additive manufacturing (AM) technologies such as laser powder bed fusion is rising in importance due to the increased use of AM in industrial production. Here, the presence of internal porosity was identified as a limiting factor for the final component quality. The utilization of thermography as an in-situ monitoring technique allows the determination of the part’s thermal history which was found to be connected to the porosity formation [1]. Combining the local thermal information derived from thermography with the porosity information obtained by x-ray micro computed tomography, machine learning algorithms can be utilized to predict the porosity distribution in the part. In this study, a first approach for the prediction of keyhole porosity in a cylindric specimen from AISI 316L stainless steel is presented. It is based on data augmentation using the “SmoteR” algorithm [2] to cure the dataset imbalance and a 1-dimensional convolutional neural network.
[1] C.S. Lough et al., Local prediction of Laser Powder Bed Fusion porosity by short-wave infrared thermal feature porosity probability maps. Journal of Materials Processing Technology, 302, p. 117473 (2022)
https://dx.doi.org/10.1016/j.imatprotec.2021.117473
[2] L. Torgo et al., SMOTE for Regression. Progress in Artificial Intelligence, Chapter 33, p. 378-289 (2013)
https://dx.doi.org/10.1007/978-3-642-40669-0_33
Avoiding the formation of defects such as keyhole pores is a major challenge for the production of metal parts by Laser Powder Bed Fusion (LPBF). The use of in-situ monitoring by thermographic cameras is a promising approach to detect defects, however the data is hard to analyze by conventional algorithms. Therefore, we investigate the use of Machine Learning (ML) in this study, as it is a suitable tool to model complex processes with many influencing factors. A ML model for defect prediction is created based on features extracted from process thermograms. The porosity information calculated from an x-ray Micro Computed Tomography (µCT) scan is used as reference. Physical characteristics of the keyhole pore formation are incorporated into the model to increase the prediction accuracy. Based on the prediction result, the quality of the input data is inferred and future demands on in-situ monitoring of LPBF processes are derived.