Analysis of a discontinuous least squares spectral element method (Q1610539)

A reinitialized partial moment technique, which is interpreted as a finite impulse response filtering approach, is employed for handling time derivatives of systems of differential equations. A least-squares technique is applied to the identification problem of a continuous-time model from sampled data of input-output measurements. An original methodology, based on an instrumental variable techniques, is developed to eliminate the bias resulting from the correlation between regressors and equation error. Some numerical results are presented to demonstrate the efficiency and the feasibility of the proposed methodology in the case of large noise to signal ratio.