Advanced numerical simulations can boost the design process of new paradigms in space transportation systems, whose necessity is brought by the introduction of small satellites and by the request of space access cost and footprint reduction.
The scope of the project is the numerical characterization of the phenomena involved in space transportation systems. The main pillars of the research are: (i) external flow of space launcher in the lift-off and ascent phases; (ii) hypersonic flows for atmospheric re-entry description; (iii) interaction between turbulent motions and chemical reaction in Liquid
Rocket Engines (LRE); (iv) combustion and propellant grain regression in Solid Rocket Motors (SRM) and Hybrid Rocket Engines (HRE); (v) liquid spray injection, atomization, vaporization and combustion in LRE;
(vi) wall heat fluxes and cooling system for LRE; (vii) wall flows in cooling channels of LRE; (viii) shock-wave oscillations in overexpanded nozzles during the sea-level startup; (ix) thermo-chemical models for propulsive applications.
A variable fidelity approach will be adopted: hi-fidelity techniques will be employed for the characterization of basic physical phenomena and restricted portions of the problem of interest; low-fidelity techniques will be employed to support design phase and system optimization strategies. Big-data analysis and response surface techniques will be employed in conjunction with both hi- and low- fidelity numerical strategies to generate surrogate models to alleviate the numerical and experimental burden of the optimization processes.
For this purpose a mid-size parallel computing server is requested. The server should be equipped with: (i) a number of CPUs sufficient to execute hundreds of threads simultaneously; (ii) an amount of RAM memory of the order of half a terabyte to cope with large physical domains; (iii) an amount of hard memory of the order of the tens of terabytes for the storage of the produced data.
The mid-size server subject of this funding request is a fundamental element to ensure the competitive advances to Sapienza University.
The strength in private and governative funding attraction of the present research group, was made possible by a continuous improvement of the numerical predictive capabilities, which ensures to the present research group a leading position in this research field. To guarantee this leading position in the future, a local mid-size server like the one requested will allow an improvement of the computational power at disposal as well as a better exploitation of external HPC resources.
The evisaged strategy is two-fold: (i) an increase of the group competitiveness in the National and European grants context is ensured by enhancing the physical phenomena characterization capabilities; (ii) the capability to translate expensive computational campaigns into light surrogate models, will make possible to interact with both big private companies and small or medium-sized enterprises.
The requested facility will leverage the actual numerical capabilities in the various ways depicted in the following.
- A mid-size server does not need a load leveler system to manage queues, hence the codes can be run immediately and parametrical testing campaigns can be conducted on the requested server without the bottleneck of the load leveler system, which makes this kind of numerical computations unfeasible on a massively parallel HPC server.
- For the same reason, the requested machine is perfect for surrogate models training campaigns as well as code testing and debugging in preparation for the employment on HPC servers.
- The large amount of RAM memory of the requested server allows the post process of the data generated on the HPC servers, with the advantage of ensuring a real-time, user-guided inspection of the results which is crucial for a successful surrogate model generation. It should be stressed that HPC resources can be exploited only in batch mode, hence real-time inspection of the data is strongly limited if not inhibited.
- The large amount of storage data will ensure a safe and robust data storage. In addition to being a basic requirement of a research work, data storage is the prerequisite for the data-mining activities. It is worth to recall that HPC servers allow to store data for a prescribed period of time, usually of the order of few weeks, elapsed which data are automatically deleted.
More specifically, improved capabilities in external flow of space launcher in the lift-off ascent phases as well as of hypersonic flows for atmospheric re-entry will be achieved. The thermo-chemical model reduction, along with improved turbulence-chemistry interaction closure models, will allow a more detailed description of the reacting flows in the combustions chambers of liquid, solid, and hybrid rocket engines. State-of-the art direct numerical simulations will be employed to shed light on the turbulent mixing and wall interaction of coolants at supercritical conditions. Unsteady three dimensional detached eddy simulations will allow to simulate the high Reynolds number (of the order of one million), typically present in supersonic nozzles. Artificial neural network and Bayesian inference approaches will be employed for the surrogate model generation.
The success of the proposed project is ensured by the fact that all the activities detailed in this proposal are ongoing, although limited by the shortage in computational power at disposal. Moreover we guarantee the full exploitation of the server, this being accessed by all the members of the research group and by approximatively 15 Ph.D. student and research fellows, for a total of a minimum of 25 users.
The server will replace obsolete, more than 10 years old, unserviceable models, hence a proper HPC room (Centro Elaborazione Dati - CED), the power supply chain and the venting systems are already available.