PORTAL: Prediction of mechanical properties of additively manufactured parts by means of Machine Learning

The PORTAL project is researching AI-based optimization strategies for the laser additive manufacturing of endoprostheses. A forward model is to be developed in order to make statements about properties such as fatigue strength of the manufactured component in accordance with the intended use. A further aim is to develop a backward model in order to parameterize the manufacturing process precisely to the intended use.

The PORTAL project is researching laser additive manufacturing processes (LPBF) that are to be modelled and optimized bidirectionally with the help of machine learning (ML). The aim is to meet the high material requirements in the field of prosthetics and thus enable a wide range of variants of prostheses that can be customized for individual patients if required.

Prostheses such as hip or knee joints are manufactured and optimized by combining experimental manufacturing processes and theoretical modelling. Advanced AI methods such as neural networks and regression trees are used to improve the manufacturing processes and precisely predict the properties of the workpieces.

A key aspect of the project is the transition from statistical evidence to model-based predictions of component properties in LPBF. The bidirectional models optimize both the manufacturing parameters and the component properties, which represents a paradigm shift in the manufacture of prostheses.

For patients, this means improved and customized prosthetics that meet the highest quality and regulatory requirements. This could lead to more comfortable, more durable and more accurately fitting prostheses that significantly improve quality of life.