Accelerating High-Fidelity Fixed Point Schemes with On-the-fly Reduced Order Modeling

A general method for accelerating fixed point schemes for problems related to partial differential equations is presented in this article. The speedup is obtained by training a reduced-order model on-the-fly, removing the need to do an offline training phase and any dependence to a precomputed reduced basis (e.g. a fixed geometry or mesh). The surrogate model can adapt itself along the iterations because of an error criterion based on error propagation, ensuring the high fidelity of the converged result. Convergence results are given for a general class of fixed point problems with complex dependence structures between multiple auxiliary linear systems. The proposed algorithm is applied to the solution of a system of coupled partial differential equations. The speedups obtained are significant, and the output of the method can be considered high-fidelity when compared to the reference solution.