Optimal simulation of gas flows_
Physical simulation and optimization algorithms at the service of engineering!
The context_
As part of the development and certification of its liquefied gas containment and transport systems, our client GTT must anticipate and control the behavior of its technology during leak tests or in the event of a leak. The size of containment systems (generally of the order of tens of thousands of cubic meters) and their cost make the production of full-scale prototypes and experiments difficult. It is therefore crucial to be able to digitally simulate the behavior of the technology in different situations.
These simulations of fluid flow in very complex environments represent a real scientific challenge due to the presence of characteristic lengths of very different orders of magnitude: the size of the system is several tens of meters and the size of the flow ducts a few centimeters. The 3D simulation methods by finite elements are thus too greedy in calculation time.
GTT has therefore developed a reduced physical model, called PGaz, to simulate these flows in an acceptable calculation time. The challenge is to control the precision of this reduced model and to readjust its predictions in the event of drift over time.
Coupling, quesako ?_
The coupling is carried out as follows: the flow simulation is carried out by the reduced model in the complete system at each instant. This simulation consists of a local resolution of the flow in elementary patterns (the complete system is spatially broken down into a juxtaposition of these patterns, called VER: Elementary Reference Volume) and this at each time step. The 3D solver, which consumes much more computing time, is called upon at certain time steps and in certain relevant VERs: the result of these simulations is used both to monitor the quality of the predictions of the reduced model and to readjust it for the step next time.
Setting up this coupling is also a computer challenge: the reduced model and the 3D solver are not written in the same programming language and it is crucial to ensure effective interaction between the two so as not to deteriorate the calculation.
The quality of this coupling was assessed through several test cases defined by the customer. On each of these cases a complete 3D simulation has been carried out and is considered as a reference. The results of the simulation of the coupling between the reduced model and the 3D solver were compared with this reference in order to validate the performance of the method.
The objective of our mission: to implement a coupling between the reduced model and a 3D solver by finite volumes and to propose a coupling strategy delivering an optimal compromise between computation time and precision.
Optimized coupling, guaranteed satisfaction_
The optimization of the coupling is at the heart of the success and the efficiency of the method.
It is based on several key elements:
🟠The construction of a cost function, representing the differences (in flow rate, concentration, etc.) between the coupling simulation results and reference simulations.
🟠The estimation of the calculation time generated by a given coupling strategy.
🟠Identification of optimization levers: temporal frequency of calls, number and spatial location of VERs used by the 3D solver.
🟠The optimization of these parameters to guarantee the best simulation precision while respecting the calculation time constraints.
Â
2 OSE developers mobilized
For 5 months
In collaboration with our client GTT
