AnomalieKI | AI-based quality assurance of cast components

Quality assurance in casting technology through AI-based anomaly detection in computed tomography images.

The quality assurance of primary-formed semi-finished products and components is associated with high costs and time expenditure for manufacturing companies due to the complex, three-dimensional geometries and shapes. The quality assurance methods used range from visual inspections and leak tests for cavities to three-dimensional imaging methods using X-rays and computer tomography. Computer tomography, in particular, is an inspection method with a high information density. However, making this usable in operational reality is associated with high costs and time expenditure: Time-consuming inspection of component images by experts for the identification of defects, high number of images of sample components for the generation of digital positive images as well as their necessary updating in case of component changes.

The goal of the Anomaly AI project is to develop an approach using Artificial Intelligence (AI) to identify anomalies in cast components to save both cost and time. Anomaly detection based on AI offers the potential to increase the reliability of identifying defects as well as reduce the cost of using CT equipment in quality assurance by bypassing the need for a positive image.

Transfer of the project results of AnomalieKI to further application fields

Symbol image for structure recognition in the AnomalieKI project
© Fraunhofer IGCV
Symbol image for structure recognition in the AnomalieKI project

Within the framework of the research project AnomalieKI, an efficient procedure for anomaly detection in computer tomography images is to be developed. Furthermore, the potential of different algorithms and methods of artificial intelligence and machine learning for quality assurance in casting technology will be analyzed. In particular, methods from the field of Unsupervised Learning are to be used, which can reduce the effort required to create labels of the data.

Another project goal is the scientific comparison of defect clusters based on three-dimensional data and the historical defect categories of foundry engineering based on two-dimensional sections. More than 1000 three-dimensional CT images of different cast aluminum components from the series production of a foundry will be used.

The project is building a showcase that can have an impact far beyond the foundry industry for all manufacturing technologies and image-based inspection methods. A transfer of the project results to two-dimensional camera images and thus a very wide range of applications is also possible.

Three-dimensional similarity metrics enable quality assurance of cast components and a new way of looking at casting defects

The AnomalieKI project was launched in June 2020. First, the collected CT data is processed and clustered on the basis of various similarity metrics. The resulting clusters are evaluated in terms of their geometric separation acuity with respect to standard geometries and anomalies and assigned a label by statistical analysis. In the last step, these new labels are transferred back to the data set and a classification model is trained for fast anomaly detection in CT data.

In the course of the project, a variety of similarity metrics for component geometries, algorithms and methods of artificial intelligence and machine learning will be tested and analyzed for their suitability in real production engineering operations for application in image-based quality assurance. The result of the project is an applicable pipeline for processing computed tomography data for quality assurance of cast components as well as a new way of looking at casting defects based on three-dimensional similarity metrics.

Cooperation with Fraunhofer IGCV

Contact us for an individual solution that suits your business needs.

Industry solutions

The key sectors of Fraunhofer IGCV:

  • Mechanical and plant engineering
  • Aerospace
  • Automotive and commercial vehicles


We are shaping the way into the future of efficient engineering, networked production and intelligent multi-material solutions.