Understand and evaluate geometries of 3D CAD data for the Digital MockUp


The project was located in the field of digital prototyping in vehicle development. In the so-called Digital MockUp, the components of a vehicle, given as 3D CAD data, are validated with regard to their functionality. This is not only done individually for each component, but also for the interaction of the components, which are constructed by many different engineers. An important aspect is to check that components do not conflict with their neighbours, i. e. do not occupy the same installation space. In the past, the degree of automation here only went so far that large quantities of components could be examined for collisions with their neighbours.

Research question

In some cases, the collisions between components that occur in practice show the engineers relevant errors that require a structural modification of the components. However, the majority is irrelevant to the engineers. A common example is collisions involving small parts such as screws, bolts or clips. Their task is to fasten components to each other and the resulting collisions are wanted. The final evaluation, whether critical or uncritical, had to be made by an expert. In GeoCADUp, methods have been developed for the automated classification and evaluation of collisions, which very often involve fasteners such as screws or clips.


The project focused on the use of deep neural networks based on images or point clouds. Image-based neural networks have been developed that classify 3D objects and in particular different types of fasteners in order to provide the expert with information about the components involved in a collision. The large data base necessary for the training of neural networks was available at our industrial project partner and was prepared for GeoCADUp. In addition, an academic data set of fasteners was created and made available to the scientific community.


For a subsequent evaluation of collisions between a fastener and its hole countergeometry, a neural network using panoramic views is used.

In addition, a neural network based on point clouds was developed for the segmentation of connection points within a component. With the results of GeoCADUp, a higher degree of automation has been achieved in the digital validation of vehicles.

ManagementProf. Dr. Nicola Wolpert
Partner (external)invenio Virtual Technologies GmbH, Johannes Gutenberg University Mainz
Project e-mail addressnicola.wolpert@hft-stuttgart.de
FundingFederal Ministry of Education and Research (BMBF)
Call for proposal"Qualifizierung von Ingenieurnachwuchs an Fachhochschulen" ("Young engineers")
Duration01.05.2017 – 30.04.2021 extend until 30.06.2022


Name and position Field Email and phone Room
Professor Mathematics +49 711 8926 2697 2/368