See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/242531069

Identification Based on MATLAB

Article · January 2009

CITATIONS READS

16 2,001

3 authors, including:

Mei Li
China University of Geosciences (Beijing)
2 PUBLICATIONS 23 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mei Li on 30 July 2014.

The user has requested enhancement of the downloaded file.

 ISBN 978-952-5726-06-0
Proceedings of the 2009 International Workshop on Information Security and Application (IWISA 2009)
Qingdao, China, November 21-22, 2009

Identification Based on MATLAB

Mei Li1, Chen Chen1, and Wenlin Liu2
1
Key Laboratory of Geo-detection (China University of Geosciences, Beijing), Ministry of Education, Beijing, China
Email: maggieli@cugb.edu.cn
2
Campus Network Center, Nanchang Hangkong University, Nanchang, China
Email: xiaochen52@qq.com , septsta_cn@sina.com

Abstract—System identification based on MATLAB is Toolbox of MATLAB can simplify the calculation
highly efficient. First of all, a second-order system is built process and improve the efficiency of identification.
with System Simulation Toolbox of SIMULINK, and then
the parameters of the system are identified by use of System
II. HELPFUL IDENTIFICATION PRINCIPLE
Identification Toolbox. The identification report is
generated in the form of web page and the identification The general form of least square identification model
interface is produced by GUIDE (Graphical User Interface is as follows.
Development Environment). The calculation process is very B( z −1 ) C ( z −1 )
simple and the identification system is direct-viewing and A( z −1 ) z (k ) = −1
u (k ) + v( k ) (1)
easy to modify. F (z ) D( z −1 )
v ( k ) is a zero-mean random noise. z ( k ) is the output
Index Terms—system identification, GUIDE, SIMULINK, signal. u ( k ) is the input signal. The identification
MATLAB
process is to seek each parameter, that is, to correct the
I. INTRODUCTION model parameters by use of the difference between
estimated model output and actual output through some
System identification is widely used in engineering kind of algorithm and to get the final optimal model.
and non-engineering areas. There are many identification The work of this paper is to build a system to be
approaches. According to different classification identified and then identify it utilizing the system input
methods they can be divided into non-parametric model and output data. After pre-processing the data and
identification and parametric model identification or choosing an appropriate identification algorithm, the
online identification and offline identification. There are system parameters of the model are obtained by
three principles of identification: least square method, calculating, and then verified [1].
the gradient correction method and the maximum
likelihood method. The recognition system described in III. BUILDING A SYSTEM
this paper is the least square offline parametric
identification system. To build a second-order model to be identified by use
Due to the complexity and diversity of the real system, of SIMULINK[2]. A pseudo-random m-sequence is input
the actual modeling problem from data acquisition to into the system and white noise is added which are
model establishment is difficult to complete by manual shown in Figure 1.
labor because it need repeatedly questing and the amount The system input and output data, which will be used
of calculation is quite huge. The System Identification

Figure 1. Building a second-order system in SIMULINK environmen

in later identification process, will be imported into

workspace through ToWorkplace module in the sink
This work was supported in part by a grant from Key Laboratory of modules library.
Geo-detection (China University of Geosciences, Beijing), Ministry of
Education.

© 2009 ACADEMY PUBLISHER

AP-PROC-CS-09CN004 523
 Compare the predicted output with the measured output;
IV. IDENTIFICATION PROCESS Function resid computes and tests the residuals of the
model; Function pe computes prediction errors; Function
A. Data pre-processing [3] predict computes the k-step ahead prediction; Function
Function iddata will change the input and output data idsim simulates a given dynamic system.
into the format which can be used by some functions of Run the following procedures:
system identification toolbox. After that function dtrend [yh fit]=compare(ze,m);
will remove the trend term of input and output data. fit
zp=predict(m,z);
B. Choosing structural parameters plot(y);
In system identification, an important issue is to hold on
choose rational structural parameters of the model. plot(zp,'r-.');
Considering ARX model structure, the first step of title('estimated values and true values');
choosing model structure is to generate ARX structure legend('true values','estimated values');
parameter by function struc, and then use the function xlabel('time');ylabel('signal amplitude');
arxstruc to compute loss function of the ARX model The results obtained are shown in Figure 3.
structure, that is, the normalized quadratic sum of output
prediction error. The last step is to select structural
parameters with function selstruc based on loss function
[4]
.

C. Parameter estimation
Off-line parametric model identification functions
include ar, arx, ivx, armax and so on. As for ARX model
identification, after choosing structure parameters, input
the following statements:
m=arx(ze,nn);
present(m)
MATLAB will complete relevant attributes
identification of this model, as shown in Figure 2.
Figure 3. The contrast curves of the predicted and actual output

From Figure 3 we can see that the waveforms of the

forecast output and the actual output are basically the
same, and the matching degree is about 75%.

E. Identification report generated by Report Generator

MATLAB Report Generator can turn the data of the
model into various kinds of identification report,
including HTML, PDF, RTF, Microsoft Word and XML
Figure 2. Results of ARX model parameters identification format.

ARX model is A(q) y(t) = B(q)u(t) + e(t) and B(q) is F. Identification interface generated by GUIDE
A(q) GUIDE generates a graphical user interface (GUI) [5-6],
its transfer function. The transfer function of the second- which makes the identification interface simple, intuitive,
−1 −2 shown in Figure 4.
order system we just set up is −1.5q − 0.5q . It can
1 − 1.5q + 0.7 q −2
−1
V. CONCLUSION
be seen that the results of parameter identification is
quite accurate. A second-order system is built and its parameters are
identified by use of SIMULINK and System
D. Model Checking Identification Toolbox of MATLAB. The identification
After the identification results are obtained, it also system takes advantage of many MATLAB functions
needs to verify whether this model is applicable. If not, such as GUI, report generation and a number of easy-to-
the model structure should be changed and the use toolbox, which allow us to quickly finish
parameters should be re-estimated. programming. The program is easy to be understood and
There are 5 main functions for model validation and operate and further practical use.
simulation in the toolbox. Function compare is used to

524
 Figure 4. Identification interface

REFERENCES [4] Xiangdong Wang, Ming Wei, Jianwen Zhao and Xiang
Chen, Parameter Identification Based on System
[1] Boyi Ni and Deyun Xiao, System Identification and Identification Toolbox in Matlab, Journal of Ordnance
Simulation Toolbox under MATLAB Environment, Engineering College. 2008, 20(4): 75-78.
Journal of System Simulation. 2006,18(6): 88 -90. [5] Jianxing Zhou, Xingming Qi and Jinyi Jiao, MATLAB:
[2] Jing Hu and Zhongfan Yuan, Second-order
third-order From introduction to proficiency, Posts &
system modeling and simulation based on SIMULINK, Telecommunications Press, 2008
China measurement Technology. 2004, (2): 20-21. [6] Nannan Li, Qing Wu and Huilin Cao, concise course of
[3] Xiaohui Qi, Qingmin Tian and Hairui Dong, System MATLAB7. Qinghua University Press, 2006
Modeling Based on System Identification Toolbox in
Matlab, Exploitation and Application of Software. 2006,
25(10): 1493-1496.

525

View publication stats