In many real-world applications, the development of a new product is carried away by making slight modifications to the existing design until the desired properties are achieved. This aspect must be taken into account for the numerical simulations, specially for large models. In these cases, each new simulation of the new geometry of the object can take days, even if the changes in the model were minor.
A promising alternative to tackle these types of problems is the Reduced Order Modelling (ROD). In ROD, the full simulation of the model is carried a way a small number of times. With these results, it is now possible to “train” the algorithm and, and compute results with a minimal cost for new geometries that were never simulated directly. By solving a much smaller problem, of a few degrees of freedoms, accurate results can be obtained as long as the flow behaves similarly as in the simulations used to “train” the algorithm.
Below some examples are presented.