Advanced search
Start date
Betweenand


Development of software for static and dynamic gross error detection and data reconciliation of chemical and petrochemical processes

Author(s):
Agremis Guinho Barbosa
Total Authors: 1
Document type: Doctoral Thesis
Institution: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Química
Defense date:
Examining board members:
Marlei Barboza Pasotto; Valdir Apolinário de Freitas; Basilino Barbosa Freitas Junior; Gilmar Barreto
Advisor: Rubens Maciel Filho
Abstract

The main goal of this work was the development of software for data reconciliation, gross errors detection and identification, data reconciliation, parameter estimation, and information quality monitoring in industrial units under steady state and dynamic operation. The development of this software was focused on meeting the criteria of modularity, extensibility, and user friendliness. Data reconciliation is a procedure for measurement data treatment in process plants, which is necessary due the fact of the inexorable presence of random, small magnitude errors associated to the values obtained from measurement devices. In addition to the random errors, sometimes data are associated to major magnitude errors that lead to a trend or bias. Errors of this nature can be qualified and quantified through gross errors detection techniques. It is important for optimization routines that data are reliable and error free as much as possible. The task of removal of these errors using previously known models (data reconciliation) is not trivial, and has been studied for the last 40 years in the field of chemical engineering, showing an increasing amount of published works. However, part of these works is devoted to applying data reconciliation over single equipment, such as tanks, reactors, distillation columns, or small sets of these equipments. Furthermore, not much of this published work relies on real operation data. This can be regarded to the dimension of computational work associated to the great number of variables. This work proposes to take advantage of increasing computational capacity and modern development tools to provide an application in which the task of higher dimension systems description is accomplished with ease in order to produce data estimates of superior quality, in a suitable time frame, to control and optimization systems. It is worthwhile mentioning that data reconciliation and gross error detection are fundamental for reliability of the results from supervisory control and optimization routines, and can be used also to process state reconstruction. (AU)