A FAST EXPLICIT FETD METHOD BASED ON COMPRESSED SENSING

Linear equations must be solved at each time step as the explicit finite element time-domain (FETD) method is used to solve time dependent Maxwell curl equations, which leads to a huge amount of computational cost in a long period time simulation. A new scheme to accelerate the iteration solution for matrix equation is proposed based on compressed sensing (CS), in which a low rank measurement matrix is established by randomly extracting rows from mass matrix. Meanwhile, to reduce the number of measurements required, a sparse transform is constructed with the help of prior knowledge offered by the solution results of previous time steps. Numerical results of homogeneous cavity and inhomogeneous cavity are discussed to validate the effectiveness and accuracy of the proposed approach.