Virtual reality in automotive digital twins : A large language model-assisted content analysis of an expert evaluation

Abstract

Digital twins are gaining significant traction within the automotive engineering industry. Within this environment of data interchange that may reach very high levels of complexity, virtual reality can ease understanding and provide intuitive interaction possibilities. There is, however, a need for more expert-driven investigation into the potential of virtual reality for digital twins. Our research aims to provide such insights into the qualities of virtual reality that could enhance the use of digital twins, as well as suggestions on how this can be integrated into existing processes. Using focus groups with experts in digital twin technology (n=16), we employ a novel large language model-assisted approach for extracting common themes identified. Our research indicates stakeholders can use visualized data in VR to optimize analysis and fault identification during production, thereby optimising the manufacturing process as well. This also allows stakeholders to digitally view components that would otherwise be difficult or even impossible to inspect due to the manufacturing process. Additionally, discussions highlighted how VR can simplify data while ensuring accuracy. Finally, we provide insight into the use of large language models for qualitative analysis, specifically to find themes and use triangulation to determine their prevalence.