CAD package for industrial process identification W. C. Amaral”, L. V. R. Arruda*, S. S. Buena+*, A. S. Barbara+, M. A. Silva+, H. J. Almeida Jr.+ and M. C. Castanho” ‘LJNICAMP-FEE, Department of Computer Engineering and industrial Automation, PB 6101, 13081 Campinas, SP, Brazil ‘CT1 -Automation Institute, PB 6162, 13081 Campinas, SP, Brazil *l3S/CNRS - LASSY, 250 Avenue A. Einstein, SPI - BAT4, 06560, Valbonne, France (Received 20 September 1991; revised I I June 1992) A CAD system for off-line industrial process identification is presented. The system has two modules. In the first data acquisition is performed. The second performs the process identification procedure. The functions and facilities of these methods are described, which include: the configuration of the data acquisition procedure to be performed, preliminary data analysis, model structure and parameter estimation method selection, time varying parameter estimation and model validation. An industrial application in petrochemi- cal process identification is presented. (Keywords: CAD; process identification; petrochemical industry) The necessity of improved efficiency in the operation of plants of higher quality products, increased yields and decreased energy consumption, has led industries to adopt advanced process control strategies’. The design and implementation of control systems based on such strategies require a deep knowledge of the process to be controlled, which usually requires a model description. However, building the model from physical laws is in most cases difficult, involving complex numeri- cal methods for dealing with their associated mathemati- cal representations. On the other hand, an enormous development in both theory and practice of system iden- tification has taken place during the last few years. Thus another possibility for process modelling consists of using the existing process identification methods, such as step response techniques, frequency response techniques, Fourier and spectral analysis, correlation techniques and those techniques based on parameter estimationz. System identification algorithms usually require a con- siderable amount of processing power and memory capa- bility, which are available today on microcomputers at a relatively low cost. These machines are also character- ized by the convenient facilities they provide for imple- menting efficient man-machine and graphics interfaces. This paper describes an interactive and robust soft- ware package based on parameter estimation methods that is configured via a user-friendly environment. It was developed as a tool to be used by a large Brazilian refin- ery for off-line industrial process identification. The programs were written in C language; the execu- table file has approximately 300 kbytes and runs on an IBM compatible microcomputer with an arithmetic co- 0959-1524/92/03015547 0 1992 Butterworth-Heinemann Ltd processor. It has 16 channels of A/D and eight channels of D/A process interfaces with 10 bits resolution, and eight digital interface channels. This software package is composed of two indepen- dent modules designated MAdEP (inverse of Process Excitation and Data Acquisition Module) and MIP (inverse of Process Identification Module). The functions performed by these modules are shown schematically in zyxwvu Figure 1. These tasks show the phases to be followed in a typical procedure for system identification, and are des- cribed in more detail below. The paper is organized as follows: the main features of the man-machine interface are described first. The next section describes the MAdEP module, followed by the MIP module. Finally, an industrial application is shown and the conclusions and future research directions are discussed. Man-machine interface The man-machine interface was developed in a menu- oriented fashion. It uses context dependent function keys, as shown in Figure 2. Windows are employed for data input-output, so that the data fields are always initiated with default values which can be user modified. The data furnished by the user are automatically checked, to avoid invalid values. The results obtained during the identification pro- cedure are displayed through text or graphical windows. Contextual help, activated by pressing a specific func- tion key, gives the casual and novice user all the infor- J. Proc. Cont. 1992, I/o/2, No 3 155