Parameter estimation for spatially interconnected descriptor systems using Sequentially Semi-Separable matrices

An algorithm is presented that efficiently estimates the parameters of spatially interconnected descriptor systems. The descriptor system plays an important role in representing the equations of motion of many mechanical structures in state-space form. In this context, descriptor systems offer a clear advantage over normal state-space systems since physical interpretation is preserved. The parameters of the interconnected descriptor systems are estimated by minimizing the Output-Error. To do so, the well-known Gauss-Newton optimization method is adapted to the descriptor form. Furthermore, the optimization algorithm exploits the matrix structure of the interconnected systems by using Sequentially Semi-Separable matrices, such that the computational complexity is linear in the number of subsystems. Two examples demonstrate fast convergence and linear computational complexity.