Microcomputer based interactive process identification package

Abstract

A systems approach for solving problems of process control engineers in process diagnostics, graphics and computational power is applied here. A popular microcomputer is used to develop an interactive but powerful process identification package. Stochastic time series analysis methods are employed for identification and prediction of processes. Box and Jenkins technique is implemented as it is interactive and requires little knowledge about the process itself. Adequate and parsimonious transfer function models obtained from this method are used in forecasting using leading indicators. As an alternative to Box and Jenkins, the automatic technique of Pandit and Wu is implemented. From Pandit and Wu’s procedure transfer function, open/close loop relationships and dual-input single-output models can be obtained. Both techniques are tested using real data. Identification routines are developed in FORTRAN, graphics, communication and integration is achieved using multi-purpose Symphony integrated package. The whole system is developed in menu driven and user friendly environment for immediate use in process industries.