TransAT-SArP module is being developed by Prof. Hatem Ghorbel’s Data Analysis Group at HES-SO, within the SmartFluid Project funded by the Swiss Commission for Technology (CTI).
TransAT-SArP is a genetic algorithm controller (GA) encapsulating TransAT tool and aiming at accelerating convergence of the numerical TransAT computational fluid dynamics (CFD) solver. The GA is based on evolutionary and genetic approaches that optimize the search of best adapted and fitted solutions to the current resolution context. The genetic process goes through real time adjustment of the free numerical parameters of the iterative solver within the constraints of physical convergence and objective achievement of the simulation. In order to enhance the process of selection and evaluation of the fitness of solutions, approximation functions inspired from data mining techniques (decision trees) are constructed from the perspective of historical data.
TransAT-SArP monitoring process is based on the following steps:
- User choice of initial parameter,
- Lunch of CFD solvers analytic resolution,
- Detection of bad resolution performance,
- GA launch to search for fittest solution,
- CFD solvers analytic resolution of the selected solution.
Input of the TransAT-SArP
TransAT-SArP could be launched using the default input parameters. In case of customized launch is desired, the following optional input data is required :
- Input physical parameters needed to be varied,
- Value range of varied parameters,
- Search depth of the GA,
- Complexity of the GA.
Output of the TransAT-SArP
TransAT-SArP generates graphical interfaces to figure out:
- The convergences of the current solution
- The progress of the solution search
- The fitness of the retained solutions