Before
A number of defects are generated during the Electron-Beam manufacturing process, but the mechanisms are not clear.
After
Able to spot and understand inner structural defects in terms of microcracks, porosities and bubbles effectively. Reduced defect rates by up to 30%.
The How
- High-speed infrared camera installed to capture real-time data during the manufacturing process
- Use unsupervised learning to find patterns in the image data to understand the mechanisms causing the defects
- Combine with supervised learning to develop a data driven ML model
- Use this model to predict, monitor and optimise the manufacturing process
