Predict component failure of 3D-printed metal parts with artificial intelligence

Quickly and flexibly manufactured: 3D-printed metal parts offer many advantages to industry. Due to their porosity, their behavior, both under static and dynamic loads, is very difficult to predict. Using artificial intelligence (AI) and the combined know-how of materials science and computed tomography, a project team from maXerial, OST, inspire and RhySearch has succeeded in simulatively determining component failure.

3D-printed parts made of metal can be manufactured without tools and are flexible in geometry. That is why they are being used more and more, both in mechanical engineering and in the automotive industry or in vacuum technology. However, pores, cracks and other material defects that occur during manufacturing can lead to component failure or breakdown - and possibly the entire system.

Normally, a new component is tested during development with elaborate tensile tests - whereby a new, slightly modified part must be manufactured, or in this case printed, for each test. This is enormously costly, prolongs development time and consumes a lot of resources. Although X-ray computed tomography (CT) makes it possible to analyze external and internal material defects in a non-destructive manner, it is hardly possible to evaluate the huge amounts of data generated in the process without high-performance computers and use them to predict component behavior.

Combination of materials science and computed tomography

In an innovation check financed by the Office of National Economy of the Principality of Liechtenstein, maXerial, RhySearch, the Fachhochschule OST as well as inspire bundling their know-how from additive manufacturing, machining, materials science and computed tomography to form a project team. "The idea was that if you could simulate porosity with sufficient accuracy, you could calculate at the design stage under which loads and at which point the component would break," explains Dr. Thomas Liebrich of RhySearch.

The project team set itself two goals:

  • Develop a data-driven material model that can be used to predict failure under static loads in 3D-printed components;
  • Build a cloud-based database solution with a user-friendly, browser-based interface that enables industry to evaluate CT scans without powerful computers and specialized knowledge.


Predictions thanks to artificial intelligence

For the model, the inspire competence center first printed samples with standardized geometries but different manufacturing parameters and thus different porosity. Their material defects and inhomogeneities were then characterized by means of CT at maXerial in Vaduz.

"Subsequently, we combined the data obtained in the CT with modern data analysis methods, such as artificial intelligence (AI), and various failure models. This enabled us to predict the failure or failure location of the samples," explains Dr. Patrick Bleiziffer, CIO and co-founder of maXerial.

To verify the results of the tests, the samples were analyzed in a quasi-static tensile test at the Institute of Microtechnology and Photonics, Materials Engineering Group, at the OST University of Applied Sciences.


Shorter development time for 3D printed components

Thanks to this joint project of maXerial, inspire, Fachhochschule OST and RhySearch, the time between the design phase and commissioning is shortened enormously. It also makes CT services accessible to a wider range of customers, especially SMEs.

Are you also facing a challenging problem in the field of (ultra) precision manufacturing? Contact us, we will be happy to advise you!


Our website uses cookies so that we can continually improve the page and provide you with an optimized visitor experience. If you continue reading this website, you agree to the use of cookies. Further information regarding cookies can be found in the data protection note.
If you want to prevent the setting of cookies (for example, Google Analytics), you can set this up by using this browser add-on.