Multispectral Automated Roller and Ball Bearing Inspection System Using Machine Learning

To address the Air Force’s need for a non-destructive inspection (NDI) system for silicon nitride ceramic bearings that is significantly faster and more cost-effective than current systems, Intellisense Systems, Inc. (Intellisense) proposes to develop a new Multispectral Automated Roller and Ball Bearing Inspection System Using Machine Learning (MARBIL) that uses both multispectral imaging to detect all potential flaws in the bearing, including surface and interior flaws, without requiring any liquid penetrant and machine learning (ML) algorithms to rapidly distinguish faulty bearings from acceptable ones. MARBIL’s ML-based imaging approach can accommodate multiple sizes of rolling elements, both rollers and ball designs, as well as bearings of different materials, including different ceramics and metal. This allows a single MARBIL unit to inspect all possible bearings at a production facility, significantly reducing costs compared to those at facilities that utilize specialized NDI systems. Its ML algorithms allow for training at a fraction of the time and cost required for conventional artificial intelligence training, further reducing the cost of the system. Furthermore, MARBIL’s parallelization approach for imaging and image processing minimizes the overall inspection time and supports a high throughput of bearings. In Phase I, Intellisense will demonstrate the feasibility of MARBIL by constructing a proof-of-concept system and documenting its ability to meet the minimum flaw size and probability of inspection requirements established by the Air Force. Intellisense will also develop an initial transition and business plan for further development. During Phase II, Intellisense will validate the methodology developed in Phase I for conditions replicating those found in practice with physical testing and show feasible build processes and stable quality assurance processes. Intellisense will also work with interested original equipment manufacturers and bearing suppliers to begin transitioning the technology to production facilities. By the conclusion of Phase II Intellisense expects to have a demonstrator model complete and validated in one or more production facilities in preparation for full MARBIL production.