Regarding the manufacture of new lightweight, yet sturdy components for modern passenger jets, researchers are handling the process like attempting to brew the tastiest cup of coffee.
Argonne scientists have developed a new machine-learning algorithm that will help enhance the manufacturing of jet components. ((Image Credit: Shutterstock/peterschreiber.media.)
By employing artificial intelligence (AI) and machine learning, scientists at the U.S. Department of Energy’s (DOE) Argonne National Laboratory are shrewdly and automatically picking the flawless settings for a diverse kind of hot brew — the process of friction stir welding, a common constituent required to manufacture airplane parts.
In a new partnership with Edison Welding Institute, GE Research and GKN Aerospace, Argonne computer scientists are harnessing the power of the laboratory’s automated machine learning expertise and supercomputers.
By decreasing the number of expensive experiments and laborious simulations with a new machine learning method, they can produce exact models that offer crucial information regarding the welding process in a lot less time and at a relatively cheaper cost.
This method, referred to as DeepHyper, is a scalable automated machine learning package put together by Argonne computational scientist Prasanna Balaprakash and his colleagues at Argonne. Machine learning is a process wherein a computer can teach itself to locate the best answers to a specific question.
If you’re trying to brew the best cup of coffee, you can spend several hours fiddling with the many settings on the best machines. In trying to make airplane parts, we can avoid this by using machine learning, which gives us the ability to learn from a handful of example settings and identify the best one from a set of a billion possible configurations.
Prasanna Balaprakash, Computational Scientist, Argonne National Laboratory
Balaprakash states that the machine learning algorithm uses a training dataset of diverse welding conditions and parameters from which airplane component properties can be established. Based on this dataset, a huge number of possible inputs are immediately examined and ranked to establish which would give the best possible parts.
Manufacturing airplane parts involves highly complex, sophisticated and expensive machines, and automating their manufacturing can save money and time, and improve safety and efficiency.
Prasanna Balaprakash, Computational Scientist, Argonne National Laboratory
Just as a person may prefer their coffee light and mellow or strong and bitter, researchers who use machine learning have to create different models that explore a number of different properties of the welding process, offering various answers to which is ideal for different properties.
The design and development of machine-learning-based predictive models are automated by DeepHyper, which frequently involve expert-managed, trial-and-error processes.
As, in Balaprakash’s words,“no model is an absolute reflection of the truth,” he and his colleagues are not mainly attempting to locate the single best predictive model and the related welding condition. Rather, they are producing hundreds of extremely accurate models, merging them to evaluate ambiguities in the predictions, and then attempting to employ these more tested predictions in the manufacturing process.
The team’s computationally rigorous work is being facilitated by supercomputing resources at the Argonne Leadership Computing Facility, a DOE Office of Science user facility.
The collaboration between Argonne, Edison Welding Institute, GE Research and GKN Aerospace is supported by a grant from DOE’s Advanced Manufacturing Office. The project is entitled ‘Probabilistic Machine Learning for Rapid Large-Scale and High-Rate Aerostructure Manufacturing’.