DAVIS

However, recent results in the field of machine learning (ML) enable a completely new solution option for meshing: Process steps that were previously performed manually can be learned using data-driven deep learning methods and thus automated. In addition, data-driven approaches enable the incorporation of non-geometric requirements such as load-case-specific, material-dependent and aesthetic conditions, which play a significant role in the assessment of mesh quality but are difficult to formalize mathematically. 

The goal of this project is to develop a novel AI assistance for efficient generation of hexahedral meshes. Such an approach could at least halve the time required to perform simulations in many automotive, medical and mechanical engineering applications. The technology enables the shortening of development cycles and improvement of quality in key sectors of German industry. 

Merkle & Partner brings the experience of 32 years of service and more than 15,000 calculation projects. M&P provides expertise from the practical application of meshing methods in volume/shell separation (the areas where thick-walled and thin-walled components adjoin) and in the detection of critical areas (areas of high gradients, e.g. notch stresses in components). M&P contributes industrial projects relevant for hexahedron meshing from all industries and tests the computational time savings especially for very computationally intensive models. Both strength relevant components like machine parts (structural mechanics) and flow relevant components like flow channels (fluid mechanics) are considered.