Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - SALAMANDER (Soakback Assessment using LAttice Boltzmann Method and Aerothermal Nodal-network for the Design of the Engine-bay Region)

Teaser

The projects deploys in 3 phases:- First, perform high-accuracy, state-of-the-art LBM simulations, of the full engine.- Second, develop a reduced-cost, Nodal-network model, for the channel region.- Finally,combine LBM (accuracy) & Nodal methods (reduced-cost) to assess...

Summary

The projects deploys in 3 phases:
- First, perform high-accuracy, state-of-the-art LBM simulations, of the full engine.
- Second, develop a reduced-cost, Nodal-network model, for the channel region.
- Finally,combine LBM (accuracy) & Nodal methods (reduced-cost) to assess soak-back at acceptable cost in the TP Demonstrator.

Technical challenges are numerous: Physics (Free Convection/ High Mach numbers); Modelling (LBM/Nodal Network Coupling); Computational capacity (Long-Transient/ Engine Geometry/ 3D Fluid & Solid). The highest levels of numerical expertise are required, especially in the field of CFD.

The present SALAMANDER project offers cooperation between DASSAULT SYSTEMS, world-leader in LBM solutions (amongst other important activities), and ALTRAN, world-leader in Engineering Solutions and outsourced R&D, through its Expertise Center dedicated to Fluids and Thermal Engineering.

Eventually, the outcome model delivery will allow the Topic Leader to improve the design of the TurboProp demonstrator as well as future products. Finally, EU industrials and environment shall greatly benefit from this study, thanks to the dissemination of this work.

Achieving high fidelity and accuracy in modelling enables to:
- Reduce test dependency to validate product configurations, and potentially in mid-term reduce the cost of flight tests campaign and the time to market.
- Develop more efficient and reliable systems, make each trade-off easier, and facilitate decisionmaking about design.
- Optimize the thermal management, gain in mass and in consumption.

Moreover LBM solver is still not considered as the mainstream CFD software among CFD practitioners.
The LBM market represents less than 10% of the complete CFD market, but this proportion is expected to grow strongly in the coming years, notably thanks to the detection of industrial cases where the addedvalue is significant and the growing capacity of fast computation. Besides the turnover associated with the CFD market is expected to triple in the next 10 years. These elements of market also suggest that there will be real opportunities for the new methodologies implemented within the framework of this project.

Work performed

Regarding WP1, agreements and PDER have been delivered as planned. Despite some delays in the first technical WP, papers are in preparation for 2020.
During WP2, the consortium studied the engine compartment and produced a first version of the methodology document.
For the stabilized computation, issues were found as the original nozzle gyration created backflow at the outlet of the core compartment. Another lower gyration was then set to solve that issue. Results were compared with RANS Fluent computation in the aim of understanding this phenomenon. However, no explanation has been clearly identified and it was decided that the real nozzle and test chamber used for the test should be used for the WP3. Comparison between Fluent RANS computation and PowerFlow WLES computation are quite promising.
WP3 consists in studying an engine casing in a detailed test chamber environment and take into account the realistic cut off conditions for the engine. The previous methodologies will then be improved and updated based on these new results.
WP4 is focusing on the internal engine flow. Indeed, internal flow will be modelled and computed with both PowerFlow and PowerTherm and this will be done after end of P1.
A major part of this WP4 is also the 1D modelling of the internal engine flow. A nodal network is currently being built by Altran. All the cavities were identified as were their connections. Correlations for convection and pressure losses should now be set in this 1D model that will be readjusted with computation results.
WP5-6-7 have not started yet, as planned.

Final results

In parallel of traditional CFD approaches (Navier-Stokes based), the Lattice Boltzmann Method (LBM) is emerging for 1 or 2 decades, beginning by automotive industry.
The LBM, thanks to a different approach of resolving fluid dynamics equation allows a better modelling of this phenomenon for a decent simulation time.
Soak-back cases are completely in the range of strength of LBM since the simulations are inherently transient. It has been proved that LBM technology is well adapted to handle natural convection issue including a coupling with a thermal model powering conduction and radiation.

Thermal model will present interesting challenges too. PowerTHERM propose state-of-the-art thermal modelling and will be able to perform the simulation of the whole engine which means:
- Quite huge 3D thermal model
- Radiation modelling by the usage of view factors
- Implementation of 1D fluid network which can be substituted to detailed 3D modelling to save CPU resources
The definition of the 1D nodal network based on high fidelity modelling presents 2 critical aspects:
- Reducing 3D results (several million spatial discretisations) into 1D nodal modelling (some tens or hundreds) implies intrinsically loss of information.
- 1D nodal networks lie on very poor formalism compared with Navier-Stokes or LBM ones. 3D results analysis must take into account this aspect to not only provide macroscopic flow behavior but also the input data which may give adequate behavior of the 1D network.
Thus, the precision of the 1D network results from a compromise: increasing discretisation to improve quality without reducing efficiency (major advantage of this kind of approach is its simplicity) and robustness around the calibration point(s).
Moreover, the disruptive aspect is not so much inherent to the thermal capability as associated to efficient and accurate coupling. Indeed, although both resolutions are unsteady, very different time scales (less than 10-3s for the fluid, up to several minutes for thermal) need adapted coupling process.
Based on EXA’s experience and Best practices, the soak-back simulation will be set with the highest standard and resolution highlighted through numerous past studies [10, 11]. Due to the different size of the gaps, the channel itself and the part to simulate, finest resolution and coupling strategies must be adapted on both cases.