Periodic Reporting for period 1 - InVIGO (INtake Vortex Ingestion on Ground Operations)
Reporting period: 2019-10-01 to 2020-12-31
During early design phases, only reduced instrumentation can be used, whereas in-depth measurements would be required for ground vortex characterization. Therefore, InVIGO project's main objective is to build a predictive model to be used during early design phases with such reduced instrumentation. This model will use machine learning algorithms that will be trained with database coming from wind tunnel testings and numerical simulation performed within Invigo. Both reduced and detailed measurements will be carried out to build the training database. This main objective is therefore related to two essential steps. First is carrying out two wind tunnel campaigns in CSTB facility with a engine nacelle model and many instruments (pressure probes and rakes, Stereo-PIV, PTV, ...). Second is performing a hundred of numerical simulations (CFD) to complement wind tunnel measurements and addressing topics that cannot be dealt with experimentally (scale effect for example).
Then all three main objectives were tackled :
- the methodology for numerical simulation was built by ALTRAN. Two software (ANSYS Fluent and LEMMA ANASTAR) were tested. The methodology was eventually implemented with ANSYS Fluent. Steady and unsteady (Hydrib RANS-LES) simulations were implemented. Mesh convergence and turbulence model comparison was carried out. Vortex post-processing was also a huge challenge due to interaction between ground vortex and flow detachment in the engine inlet.
- the instrumentation and test mock-up was designed and built in CSTB facility. Specific numerical simulations allowed refining the characteristics of the large fan to buy to generate the engine suction flow. A first wind tunnel campaign was successfully carried out by CSTB during two weeks in early December 2020.
- first machine learning algorithms have been implemented by ALTRAN and tested with preliminary database. Both direct (direct vortex characteristics prediction) and indirect (velocity field prediction, to be post-processed in a later time) were tested and showed the need for a large number of inputs.
Regarding communication and dissemination, several communication channels were used and ALTRAN presented a paper at NAFEMS France 2020 congress.
Numerical runs also simulate vortex with very refined mesh, in comparison with existing literature. Post-processing methods are also more precise than what is generally performed with simple manually positionned circles.
Until the end of the project, the first and second campaign results will be analysed and integrated with numerous CFD calculations to build a wide ground vortex database. Machine learning algorithms will be implemented to link pressure probe measurements (so called reduced instrumentation) with ground vortex characteristics.
This model will be then used by the Topic Leader during early design stages to improve fan design and provide more performance and safety on next engine generations.