Final Report Summary - REALTIME (Real Time Computational Mechanics Techniques for Multi-Fluid Problems) Real-time computational engineering systems are computer systems for solving problems that must produce their results within short time intervals similar to the physical time in which the problem occurs. Real-time systems are nowadays applied to flight control, patient monitoring, nuclear power plant control, optimal industrial processing, risk prevention and computer games. Real-time systems are having an ever-increasing impact on the quality of human life. In many cases, human safety depends on their correct performance. The project REALTIME consisted in develop, integrate and validate experimentally new formal approaches and computational methods noted in the principles of the Computational Mechanics that have proved to be useful in building industrial-strength computer codes applicable for solving multi-fluid engineering problems in real time.The main scientific goal of the project was the development of three new generations of real time methods based on innovative mathematical models and efficient computational procedures for fluid dynamics (CFD) problems. These include new CFD methods based on Lagrangian formulations, particle methods, special techniques allowing large time steps in the time integration algorithms, fast continuous re-meshing, reduction methods and, last but not least, the use of state of the art GPUs and parallel processing technology.The main practical outcome of this project was a new code (which will be in a future useful for industries, governments or enterprises) to control in real or quasi-real time different problems that could induce human risk such as industrial processes, fire spread, critical atmospheric situations or patients monitoring in which the security or human life depends of a response in a very short time.We have developed three different methods for different applications and with different performances and accuracy. A first method is quite general. Can solve any problem of incompressible CFD using unstructured meshes, explicitly time integration and large time-steps. With this code one order of magnitude in computer time over previous solutions is obtained. The second method consists in solving the problem in two steps. A first step is called off-line step where basis vectors are calculated. During the second step, which is called on-line step, the problem is solved very fast. The advantage is that once the basis vectors are calculated, different problems can be solved very quickly. With this method the on-line part is two orders of magnitude faster in computing time than the full problem. Finally, the third method is to use structured and regular meshes but very thin. Using finite difference for the approximation and fast Fourier transform (FFT) to solve the equations, the method allows obtaining very fast solutions. The method is limited to simple geometries, but allows about three orders of magnitude faster in computation time than standard methods.
