SparePartAssist research project by CADENAS, HTW Berlin, GFal, & Keßler optimizes spare parts search
How AR assistance system & geometric similarity search will soon help service technicians in the field find spare parts faster
Since June 2020, the SparePartAssist consortium of CADENAS, the Berlin University of Applied Sciences (HTW), the German Society for the Advancement of Applied Computer Science (GFal) and Keßler Real Estate Solutions GmbH has been on an exciting research and development journey. The aim of the project is to support service technicians in the field. A mobile and AR-based assistance system can help them in the search and assembly of urgently needed spare parts, e.g. for manufacturing plants, escalators or elevators. The search begins by making a 3D scan of the initial components via a smartphone or tablet. The 3D scan is then used to start a geometric similarity search from 3DfindIT.com powered by CADENAS, which generates a list of geometrically similar components. 3DfindIT.com is a technical search engine that searches over 4,000 manufacturer catalogs for matching 3D CAD & BIM models.
Here is an example of how this search process might look like using the app and Geometric Similarity Search for a larger part, like an office chair:
3D scan recognizes components based on visible partial areas
In such cases as above, the previously established object recognition methods from the 2D domain do not always suffice. Components might not get recognized because they are either obstructed and thus not completely visible, heavily worn or partially destroyed. In these cases, search algorithms from the 3D domain are needed. They are able to identify components based on partial areas of the geometry and to reconstruct 3D objects.
Implementation through data recorders & 3D sensors for Android
"When we planned and applied for the project in 2019, integrated 3D sensors for smartphones only existed in the Android world. We therefore initially implemented a simple data recorder app for Android that saves images, depth maps and the respective camera poses. This data is provided by the Camera2 API and the ARCore framework," says Prof. Dr.-Ing. Frank Neumann from the University of Applied Sciences Berlin in the German SparePartAssist Blog.
"SparePartAssist will revolutionize the spare parts market. All industrial customers which we talked to about this topic in the last few months confirmed that such a technology does not exist and is needed on the market. CADENAS will happily make it available to all companies that produce software for the spare parts market", says Jürgen Heimbach, CEO of CADENAS GmbH.
The video shows how spare parts are procured during Christmas season: