Building Digital Twins to Simulate Manufacturing Variation

In order to improve the quality of a manufactured part in industry, a variety of techniques are used to scan a built geometry to bring it back to the physics based simulation world to assess its true performance. There are various laser measurement techniques (GOM), CTScan as well as touch-point probes in the form of the CMM cloud of data.  However, there are many challenges on how to construct the digital geometry scanned not to lose any deviations and defects and yet being able to mesh a solid manifold. In this paper, a novel method based on multi-layered AI (Artificial Intelligence) method is developed to produce a meningful engineering design space to perturb the design intend geometry to match the manufactured data cloud. The inverse mapped geometry has been applied to a range of real turbomachinery components to demonstrate its flexibility and robustness, even when the original GOM is not perfect. 

Since the perturbation design space could be very large of the order of hundreds for a typical blade, a sophisticated multi-level optimisation strategy is devised to inverse map a large number of blades automatically. The resultant hot-running digital twin is then automatically meshed using a templated multi-block structured mesh generation [1], followed by conducting a high-fidelity RANS-based CFD to assess the blade performance.

Multiple strategies can then be followed to improve the performance of the manufactured part by altering and devising new manufacturing techniques whilst making sure the cost is not prohibitively increased [2]. Alternatively, a new, more robust blade design can be produced where the resultant blade performance would be more stable in the presence of the same (or similar) manufacturing variations.

Manufacturing variation in some critical components such as the LP System (Low pressure) can lead to significant performance variations and deterioration, for example it has been discovered that a few millimetre variations in some BOGVs (By-pass Outlet Guide Vanes) of a modern high-by-pass-ratio jet engine can lead to 0.1% SFC (specific Fuel Consumption) increase which is worth millions of pounds to an aircraft operator.  

In this paper, detailed physics-based analysis of the digital twin of the manufactured BOGVs have been carried out to gain understanding which key characteristic parameters is responsible for the significant performance deterioration. Recommendation is then made to the factory to apply minimal changes needed to avoid paying performance penalty. It is shown that the proposed change indeed reduces the impact of manufacturing variations on performance.

References:

 [1] S. Shahpar and L. Lapworth, “PADRAM: Parametric Design and Rapid Meshing System for Turbomachinery Optimisation":  ASME Turbo Expo IGTI, GT2003-38698, Denmark, June 2003.

[2] Wen Yao Lee, W.N. Dawes and J.D., Coull, “Physics-Based Part Orientation and Sentencing: A Solution to Manufacturing Variability”, ASME GT2019-91591, Phoenix, Arizona, USA, June 2019.