Modeling and control of physical processes using proper orthogonal decomposition (Q5931773)

The authors consider the following modelling problem: snapshots of a function \(U=U(x,y,a)\) are taken for some values of the parameter \(a\). From these snapshots the function \(U\) has to be derived for values of the parameter \(a\) not belonging to the snapshots and without using a mathematical model such as partial differential equations derived from basic laws of physics. The problem solution relies on a suitable basis. Each basis function is a linear combination of the deviations of the snapshots from their mean. This idea results in a (matrix) eigenvalue problem. Further, the truncation problem is investigated, i.e., it is demonstrated that the solution to this eigenvalue problem provides an a priori information on how many basis functions are indeed needed for the construction of \(U\) with a certain relative accuracy. Finally the authors demonstrate the usefulness and accuracy of the proposed method for the temperature of a fluid in a rectangular cavity (Rayleigh-Bénard convection) and demonstrate for this example that this method can also be used to solve the boundary control problem.