Abstract
This paper describes the identification problem of a linear continuous process in the operating condition, by applying the least-square method with an aid of a digital computer.
At first, a discrete model with unknown parameters is fitted to a continuous process. The identification is performed by estimating these parameters so that the sum of square of difference between the model and process output becomes minimum. The sampling period of the input and output data is determined at an appropriate value in reference to a specified estimating function of the output. The necessary length of data depends on the noise added in the output.
Practical computing procedure and program developed are explained briefly. The results of the identification of various processes which are simulated by using an analog computer are sufficient enough. For example, when the signal to noise ratio between the input and the noise added in the output is 10, the identification is performed in the good accuracy by using a length of data equal to 20 times the settling time of the process.
