On-line parameter estimation of a multivariable discrete data system

The estimation of parameters of process dynamics is an important phase in the modeling of systems and in the design of many instrumentation systems. This problem has been taken up by many researchers and many methods and algorithms have been formulated. To the best of the author´s knowledge, a compact and simple recursive method is not available in the literature for a fixed memory type estimator from the measurements of input and output data sets. In this proposed paper a simple and straightforward on-line method for parameter estimation of multivariable discrete-data systems is presented. The parameters are determined taking into consideration the present and n + m − 1 past measurements for a nth-order process with m inputs. To reduce the computational burden, a recursive scheme is detailed for updating the estimate of the parameters as soon as the new data arrives. This recursive scheme can be started with arbitrary parameter values. But to have an easy starting procedure, a simple starting matrix with a recursive computational procedure is also given. This method is suitable for situations involving tracking of slowly time-varying parameters that remain unchanged at least for n + m iterations.