MODELING OF AUTO-REGRESSIVE

MOVING-AVERAGE(ARMA) BASED ON

HYBRID OPTIMIZATION STRATEGY

GUO Jing  DONG Yanliang  ZHAO Keding  YU Jinying

(School of Mechatronic Engineering, Harbin Institute of Technology, Harbin 150001 )

Abstract: Estimating parameters of auto-regressive moving-average(ARMA) model is the focus of ARMA. The disadvantages of least-squares algorithm and its generalization algorithm which are used in estimating parameters of ARMA are aimed at. Simulated annealing genetic algorithm based on hybrid optimization strategy is used in estimating parameters of ARMA and it can overcome the disadvantages of the traditional methods. Based on the new algorithm, a new method of modeling ARMA is presented by determinating the autoregressive orders p and moving-average orders q in ARMA model. Finally, the precision data ARMA model of a mechanical system is built by the new technology and by the traditional modeling method. The new technology proves effective and high-precision by comparing the two models.

Key words: Auto-regressive moving-average(ARMA) model Hybrid optimization strategy Simulated annealing genetic algorithm

CLC No: TP183 O329

Received 20060528, received in revised form 20061029