2005 Volume 31 Issue 4 Pages 266-272
The partial fraction decomposition method is proposed as a new method for identifying time-delay systems. Many chemical plants include a large or small time-delay factor in their process dynamics. However, it is very difficult to estimate the time-delay exactly by using conventional methods.
In the proposed method, a suitable transfer function of the plant is built from operational input-output data with the usual identification method. The transfer function usually includes redundancy, or consists of a characteristic part and a redundant part, because the order of the plant is not known exactly. The redundant transfer function is decomposed into first-order partial transfer functions to extract the characteristic part from the model. Thereby, not only the redundant part but also errors included there in are removed. Consequently, the transfer function of the plant can be rebuilt precisely. The processing algorithm is very simple and calculations are fast and easy.
In this paper, two examples are treated to evaluate accuracy for identification. One is to model a process of which the transfer function is well known. The other is to model an industrial distillation column from operational data. Once exact time-delay is estimated, the plant can be expressed with a very simple transfer function.