Dataset for Evaluating Accuracy and Runtime Trade-Offs of Simulation-Based Virtual Sensors

This dataset provides benchmark data for the evaluation of simulation-based virtual sensors in mechanical engineering. It supports the analysis of accuracy, runtime behavior, and deployment trade-offs for physics-based virtual sensing workflows. The dataset includes processed and derived results from two representative case studies: (A) a nonlinear tension-bar benchmark and (B) a steady-state thermal model of a multi-core CPU. The data cover live finite element virtual sensors (SVS-LFE) as well as lookup-database virtual sensors (SVS-LDB) constructed using different design-of-experiments strategies and database sizes. For both case studies, the dataset contains accuracy metrics, runtime measurements, mesh convergence results, lookup-database design data, and aggregated comparison tables. The data enable verification of the quantitative results reported in the accompanying publication and provide a reference for future method development and benchmarking. The repository contains processed numerical data only. Full finite element models, meshes, solver input files, and proprietary software outputs are not included.