Engineering assistance for hybrid decision support based on machine learning surrogate models

• Applicants: Professor Dr.-Ing. Detlef Gerhard; Professor Dr.-Ing. Bernd Kuhlenkötter
• Subject Area: Engineering Design, Machine Elements, Product Development ; Production Automation and Assembly Technology
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 543079297

The advancements of digital engineering tools have continuously been changing the engineering design process. As computing capabilities have increased, more complicated and comprehensive digital models have been introduced to simulate the complex properties and behavior of products. Techniques such as generative design can be used to explore a design space with many solutions optimized for different goals that may even exceed the capabilities of experienced human designers. However, these approaches to digital engineering design support are not without shortcomings. Due to the complexity of simulations, only selected parameters can be optimized. Consequently, the solution space explored is often limited and progress incremental. Decisions predominantly rely on human expertise to select parameters and optimize, which does not always lead to the best solutions given limited time and budgets and the possibility of human bias. In recent research towards performance driven engineering design, models known as Machine Learning based Surrogates (MLS) have been introduced to enhance or partially substitute conventional simulations in product analysis and validation tasks. MLS techniques significantly reduce computing effort compared to traditional simulations. Thus, a huge design and solution space can be explored with a comparatively low effort. The accuracy of MLS is, however, adequate to find the best possible solution for given constraints and desired performance characteristics. Only the final solution has to be validated by conventional simulation with high fidelity. The aim of this project is to transfer MLS approaches from analysis tasks related to physical phenomena to the analysis of later product creation phases like manufacturing and assembly. Design phase decisions are not only driven by the performance of a product or component, but also by the impact on downstream processes, in particular on effort and associated costs, on quality, and also on sustainability metrics such as energy demand or carbon footprint. If the analysis of an envisioned solution is carried out instantaneously within the design process and provides insights from subsequent processes, product developers receive direct feedback on the consequences of their decisions. This enables proactive effort reduction in later stages of product creation by allowing them to influence production criteria. For example, time, costs and resources can be saved by rationalizing production steps, adapting assembly orders or designing for automation. In addition, products that are easy to manufacture are generally also easy to dismantle or repair, which leads to additional savings during the product use, maintenance and recycling phases. Thereby, the intended results of this project provide a hybrid decision support for engineering design and form the methodological basis for the implementation of corresponding assistance functionalities in CAD and other digital engineering tools.

• DFG Programme: Priority Programmes
• Subproject of: SPP 2443: Hybrid decision support in product creation