基于基因算法的用于系统识别的测试信号的设计

Design of Test Signals Using Genetic Algorithm for System Identification

详细介绍了如何用Matlab实现基因算法生成两个不同的具有最小峰值的测试信号,并且把这两个不同的测试信号输入到一个识别工具箱里,通过检验输出的结果来确定这两个测试信号是否适合用作特殊系统的识别.本文所用基因算法产生的测试信号如预期的一样具有最小峰值,但当把这两个测试信号输入到识别工具箱之后,估计系统的矩阵特征值与原始系统的矩阵特征值有很大的差距.产生这种差异的原因主要有两个,一个可能是产生这两个测试信号的长度不够,另一个可能是这两个测试信号非正交.

Two different test signals are generated by using genetic algorithms implemented in Matlab. Each test signal has a minimum peak factor. These two different test signals are applied to an identification toolbox to determine whether two different test signals can be used for identifying a specific system. Results show that the values of peak factors of these two test signals are as low as expected. However,the eigenvalues of the estimated system which are obtained from the test signals are quite different from the eigenvalues of the original system. Two possible reasons why the test signals generated do not work for a two-input and two-output system are explained in this paper. One possible reason is that both the test signals generated are not long enough;the other possible reason is that the test signals produced are not orthogonal.