General Electric (GE) reports that researchers at its Additive Research Lab, GE Global Research, Niskayuna, New York, USA, are working to combine computer vision and machine learning to develop a type of metal Additive Manufacturing system with the ability to self-inspect its manufacturing process in real-time.
According to GE, the eventual goal of the project is to build a system which can achieve ‘100% yield’ – wherein a machine only produces perfect parts, eliminating the need for post-processing and inspection.
Some large and complex parts, in particular those produced by GE Aviation, can take days or weeks to produce by metal Additive Manufacturing. Where post-process inspection is the only validation method available, errors and faults in parts may only be discovered after this extensive build-time.
A number of factors can contribute to sub-optimal build results in metal Additive Manufacturing. Among them are variations in powder particle size, the way new powder layers are added and potential property changes arising from powder reuse.
GE has previously applied its computer vision technology to areas such as the study of diseased tissues and the detection of microscopic cracks in machine parts, and stated that it will now use it to detect potential errors during manufacture.
“We do a tremendous amount of work on additive powders to understand what characteristics lead to a good build,” explained Kate Gurnon, a member of GE’s research team. “We want to apply this automatically to machines and, in real time, observe the dynamic behaviour of the powder delivery to the build plate. In this way, we will have a better chance of getting to the 100% yield, faster.”
The team will then attempt to ‘teach’ machines which types of powder behaviour correlate with part defects. “Using artificial intelligence and machine learning, we will turn 3D printers into essentially their own inspectors,” stated Joseph Vinciquerra, who runs the Additive Research Lab.
Read more: GE sees potential in ‘self-inspecting’ metal Additive Manufacturing systems