Abstract:
This paper presents a framework for "as-built" computational modeling, which aims to incorporate realistic geometry and material data from fabricated or legacy objects into simulations, moving beyond the idealized "as-designed" models typically used in engineering finite element simulations. The process involves three key steps: inspecting the object using non-destructive characterization (NDC) and metrology techniques, reducing and processing the collected data (including signal and image processing, sensor fusion, and feature extraction), and finally, performing engineering and physics analysis using finite element codes. The authors illustrate this framework with a cylindrical phantom, demonstrating that as-built models can be created with a high degree of confidence and that simulations based on these models can show notable differences from those using as-designed configurations. The example shows a volume-averaged difference in accumulated plastic strain of 3% and local differences of up to 10%, highlighting how the as-built approach can reduce simulation uncertainty and provide a more realistic understanding of an object's performance under a given load. This methodology is particularly valuable for assessing high-value parts, as the extra effort required for as-built modeling is justified by the improved insight it provides into potential performance issues, such as degraded performance or failure from unintended stress localizations. The paper also discusses challenges and future work, such as quantifying accumulated errors and developing more automated methods for geometric parameter extraction to make the process more routine and efficient.
