Application of a modified least-squares method to the identification of a multivariable laboratory-scale process

A modified least-squares algorithm based on the bias-compensation principle is presented for the online identification of linear multivariable discrete-time systems contaminated with additive colored noise. This modified least-squares method and the recursive generalized least-squares method are applied to the identification of a laboratory-scale process which simulates industrial material transport problems. Experimental results show that the proposed algorithm is superior to the recursive generalized least-squares algorithm and works quite well on real data