Researchers at Carnegie Mellon University College of Engineering in Pittsburgh, Pennsylvania, USA, have developed a deep learning method that provides a novel method to capture and characterize melt pools in laser powder bed fusion (PBF-LB) additive manufacturing using airborne or thermal techniques. The method, which was recently detailed in Additive Manufacturing Magazine, allows manufacturers to obtain basic melt pool geometry and predict transient melt pool changes almost instantly.
A machine learning model architecture consisting of an autoencoder and a deep CNN model (ResNet).
© Carnegie Mellon
Artificial intelligence is used for process monitoring
© 3D Science Valley White Paper
Monitoring - the key to quality
When additively manufacturing metal, many things can go wrong, and without on-site process monitoring, defects can only be detected and characterized after the product has been manufactured. Most commonly, manufacturers will use high-speed cameras to keep an eye on the melt pool geometry and how it changes over a short period of time during the PBF-LB additive manufacturing process. This requires expensive equipment, a lot of memory storage (i.e., 20,000-30,000 high-resolution photos per second), and laborious manpower to collect and sort the data. These ultimately increase the cost of online visual tracking and process analysis.
The potential thermal interactions between the laser and the raw material in L-PBF selective laser fusion 3D printing and L-DED laser directed energy deposition 3D printing are similar in that they both rely on the laser to provide heat to the powder, which is usually completely melted to form a molten pool. The molten pool then solidifies on top of the previous layer or substrate to form the desired shape. Obviously, there are significant differences in the setting of processing parameters for these processes, so different monitoring methods are required.
Several other methods have been developed for in-situ monitoring of laser additive manufacturing. However, these methods are either suitable for research purposes or not for production-scale monitoring.
Synchrotron X-ray monitoring and schlieren imaging are useful research tools for laser AM additive manufacturing, which can provide valuable insights into the phenomena that occur during laser processing. Synchrotron X-ray monitoring allows for high-resolution imaging of the melt pool area, revealing processing dynamics. Schlieren imaging investigates the fluid dynamics of the laser plume and build chamber, revealing how the AM additive manufacturing process is affected by its environment, both of which require specialized experimental setups.
© 3D Science Valley White Paper
Optical coherence tomography (OCT) and in-line coherence imaging (ICI) monitoring allow to inspect part surfaces and interpret the effects of machining parameters and understand the effects of scanning strategies on surface roughness. Eddy current testing (ECT) is used to detect cracks and sub-surface defects inside metals and has been proposed as a method for in-situ monitoring of additive manufacturing processes. At present, OCT and ICI have been minimally explored in in-situ monitoring, while ECT has only recently been deployed by scientists for in-situ monitoring of L-PBF selective laser molten metal 3D printed AlSi10Mg aluminum alloy materials. As these inspection systems further explore laser metal additive manufacturing, machine learning is likely to be used to assist in the classification and prediction of samples.
© 3D Science Valley White Paper
More economical and agile
Researchers at Carnegie Mellon University College of Engineering in Pittsburgh, Pennsylvania, USA, have developed a deep learning method that provides a novel method to capture and characterize melt pools in laser powder bed fusion (PBF-LB) additive manufacturing using airborne or thermal techniques. This is an innovative development in the field of additive manufacturing. Here are the key takeaways from their research:
- Deep learning monitoring method: This method uses airborne or thermal techniques to capture and characterize the melt pool in laser beam powder bed fusion (PBF-LB) additive manufacturing, enabling rapid acquisition of melt pool geometry and prediction of melt pool changes.
- Advantages of low-cost sensors: The new approach uses low-cost sensors, such as microphones or photodiodes, to reconstruct critical melt pool characteristics, which greatly reduces cost and operational complexity. Defect Detection Capability: This method has the potential to identify spatially dependent lack of fusion (LOF) defects, which are essential for improving the durability and mechanical properties of the final product.
- Experiment-to-data synchronization: Through a series of PBF-LB experiments that simultaneously collected acoustic, thermal, and high-speed imaging data, the researchers were able to reconstruct the melt pool geometry and track the oscillatory behavior of the melt pool.
- Understanding of multimodal process signals: This study contributes to a better understanding of the physical correlation between acoustic signals, thermal emission, and molten pool morphology, which is an important area of exploration for the scientific community.
- Future research directions: The team plans to explore more acoustic and thermal emission data for materials, as well as real-time monitoring applications across different platforms and additive manufacturing processes, with a view to building more advanced surrogate models and digital twins.
© Carnegie Mellon
According to the understanding of 3D Science Valley, the team has conducted a series of PBF-LB experiments to explore various 3D printing parameters of titanium alloy Ti-6Al-4V (Ti-64). Airborne acoustic, thermal, and high-speed imaging data for each respective process condition is collected and synchronized from pre-designed as-built structures to successfully reconstruct accurate melt pool geometry. The team even tracked the oscillatory behavior of the molten pool in just a few milliseconds. The method is also said to exhibit a good ability to effectively detect local defects between two adjacent laser scan lines.
According to Jack Beuth, co-director of the NextManufacturing Center at Carnegie Mellon University, this approach allows for the use of low-cost sensors to monitor the melt pool, which can be installed in any laser powder bed additive manufacturing machine. Artificial video of a high-speed melt pool generated from acoustic and photodiode sensor data is unique in the field of additive manufacturing. Mechanical Engineering Professor also said.
In addition, the team's research is said to be a key step towards a better understanding of the physical correlation between signals from multimodal in-situ processes.
Open up a wider research space
The significance of this research is that it provides a cost-effective, easy-to-implement monitoring method that enables the capture and analysis of critical information in the additive manufacturing process in real time to improve product quality and production efficiency. At the same time, it also provides a new direction for future research, especially in the physical correlation of multimodal process signals.
© 3D Science Valley White Paper
According to the research team's Dr. Liu, by leveraging the underlying physics of multimodal process signals and the benefits of data-driven AI, the AI algorithms developed by Carnegie Mellon enable engineers to reconstruct key melt pool properties using very affordable and easy-to-use sensors.
According to Dr. Liu, by conducting more in-depth research on acoustic waves and thermal radiation, the research team hopes to better understand their relationship to melt pool changes, pore oscillations, and other spatially relevant process features, and one day, may build advanced surrogate models and fully functional digital twins for the entire additive manufacturing process for other process characterization devices such as synchrotron X-ray machines!
Going forward, the team plans to explore more real-time monitoring applications driven by acoustic and thermal emission data for materials other than Ti-64, as well as applications across different platforms and additive manufacturing processes.
If you know deeply, you can go far by doing. Based on a global network of manufacturing experts, 3D Science Valley provides the industry with an in-depth look at additive and intelligent manufacturing from a global perspective. For more analysis in the field of additive manufacturing, follow the white paper series published by 3D Science Valley.
Website Submissions l Send to [email protected]