This paper presents a new method of plant identification with correlation technique supplying a test signal which is first proposed by Huffman and named
I-sequence by the authors. This sequence is a non-periodic finite length pulse train whose autocorrelation function is much similar to that of white noise. Impulse response of the plant is obtained by making the crosscorrelation function between the measured output and the test signal of
I-sequence.
Of the conventional identification method with correlation technique, the
M-sequence method is most well-known, while the
I-sequence method has the following features for adaptive control comparing to
M-sequence method.
(a) The
I-sequence method makes it possible to determine a part (not the whole) of impulse response, corresponding to the length of
I-sequence. Therefore, the time required to perform identification is shorter and the number of data to be handled is much smaller in general.
(b)
M-sequence method requires to know a priori the effective length of impulse response. On the other hand
I-sequence one is available even when the effective length is quite unknown.
Comparison between
I-sequence and
M-sequence methods is wholy discussed and is partly shown by the table. The accuracy of
I-sequence identification increases by the suitable selection of the pulse width, the number of pulses and the total energy of pulses of
I-sequence. As for the identification error, the following two cases are discussed. One is the case when a noisy disturbance is forced to supply at plant input, the other is when identification is performed repeatedly.
At last, the above mentioned features of
I-sequence method are demonstrated by showing the experimental results of simulation by the analogue computer.
抄録全体を表示