Blind identification of large multivariable systems

A factorial blind identification (FBID) method is proposed in this paper for identification of large multivariable systems with unknown structure. The FBID method is based on the numerically reliable QR-decomposition method and uses a special column-pivoting strategy to efficiently identify the model structure. A row-pivoting strategy can also be added to enhance the numerical reliability. As a result, the FBID method can efficiently, reliably and simultaneously produce information on the number of significant inputs/outputs, the orders, time delays, parameters and quality measures of all the models. The need for a priori model structure information is thus reduced to the minimum.