A linear estimation algorithm for ARMAX models with time dependent coefficients

An approach that models a nonlinear process based on multi-input/single-output measurements is developed. The approach uses a stochastic time-varying autoregressive moving average models that incorporate a number of exogeneous measurable inputs (TARMAX). The TARMAX model coefficients are explicit functions of time and are expressed as a linear combination of a set of pre-selected functions. The modeling approach is shown to be suitable to a milling process and a strictly linear method for evaluating the TARMAX model coefficients is presented. The model estimation approach does not require initial guess parameter values and is suitable for microcomputer implementation. The performance of the estimation algorithm is verified through numerical simulation examples