Practical Aspects of Industrial Multivariable Process Identification

Estimation of the transfer function, time moments, and model parameters of a flow process

“…As e ( k ) is a linear function of the parameters up, 6,. parameter estimates can be obtained from the data by the standard linear least-squares solution, minimizing the criterion function (Box and Jenkins, 1976;Eykhoff, 1979;Swaanenburg et al, 1985):…”

“…The choice of the model order n is not critical, provided that it is sufficiently high to represent the dynamics of the system considered (Swaanenburg, 1985). High values of n are always safe, but require a large number of samples N , because the variance of the parameter estimates increases when n / N increases.…”

“…1. The solution of the linear least-squares problem is analytical and can be found from the data by means of existing, numerically robust algorithms (Swaanenburg et al, 1985).…”

“…2. The solution can be computed on-line by using a recursive form of the least-squares algorithm (Swaanenburg et al, 1985).…”

Identification and Control of an Industrial Binary Distillation Column: A Case Study

“…Different system identification approaches have already been extensively applied in the case of distillation columns [8±11] and compared in the literature [12,13]. However, most were done either in a simulation context [10±13] or on a pilot plant [9,14,15]; only a few real applications have been reported so far [8,16,17]. The aim of this paper is not to present a new process identification method, but to describe the results of the practical application to an industrial binary distillation column of the classical black-box model identification methodology.…”