Comparison Of Modelling Between ARX And ARMAX Model In System Identification

Nonchik, Siti Nur Nadhirah (2018) Comparison Of Modelling Between ARX And ARMAX Model In System Identification. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Comparison Of Modelling Between ARX And ARMAX Model In System Identification.pdf - Submitted Version Download (421kB)

Abstract

This research presents the comparison of modelling between ARX and ARMAX in System Identification. System Identification is a methodology to explain the dynamic behavior by building mathematical models using measurement of the system’s input and output signals. The field of system identification is now widely used in most of the industrial projects in which the identification software have a wide circulation in industrial world. There are several type of general models in system identification that consist of AR model, ARX model, ARMAX model, Box-Jenkins model and Output-Error model but the main focus of this project are using ARX and ARMAX model. The aim of this research is able to simulate modelling using ARX model and ARMAX model and to compare the modelling performance of ARX and ARMAX model based on selected performance indicators. Specifically, the performance indicators that were used includes best fit value, final prediction error value and mean square error value. In a completion of the analysis, all simulation is conducted using the ‘ident’ graphical user interface in MATLAB R2015b and the least square method is utilized to estimate the parameters of the ARX and ARMAX models structure in this research. The results generally showed that ARX model structure is slightly better than ARMAX model structure in terms of model best fit, final prediction error and mean square error due to an additional input variable in the model. Thus, ARMAX could not provide better fit value caused by the random disturbance provided. However, by implementation the real data, the results showed that ARMAX model structure is better compared to ARX model. In conclusion, the better performance of both ARX and ARMAX model is still depending on the data distribution.

Actions (login required)

View Item