Integração de processos em tempo real para monitoramento e controle : aplicação para planta de PET

AUTOR(ES)
DATA DE PUBLICAÇÃO

2003

RESUMO

This work deals with the implementation of a soft-sensor for a PET (Polyethylene terephtalate) industrial process based on neural networks. The soft-sensor acts as redundant measure current for the polymer viscosity in the liquid phase; and, in case of mechanical faults, it is able to diagnoses the faults on the viscometer and can replace it.The viscosity control is realized through to two Proportional-Integral-Derivative (PID) controllers in cascade configuration. The viscosity controller is the master and vacuum controller of the high polymerizer is the slave, manipulating the vacuum valve. The use of neural network in this work is suitable due to kinetic complexities related to polymeric system. In this paper, it was employed the retro-propagation algorithm with only an intermediary layer for a feed-forward net. The soft-sensor algorithm was implemented in the supervisory system (SETCIM-Aspentech), where the database is actualized in real time. The supervisory system is executed in a Alpha-Server 1000, that is interconnected with PLC s (Programmable Logic Controllers). All data in the Setcim are organized in fundamental units (records). The applications are able to read and write data that use (among other options) one rich Application Programming Interfaces (API), supported by Remote Procedure Call (RPC). This became possible the development of several soft-sensors. The soft-sensor implementation for viscosity control was realized through requisitions of experiments with the participation of all the factory sectors related to product quality. An algorithm of restrictions was developed in the PLC, for assure the process safety, while the soft-sensor is controlling the viscosity.The use of a soft-sensor as process variable estimator for controllers is an interesting strategy; it assures a robust control and, in the presence of some disturbance, identifies which the process variables sensitize such variation. Process engineering group verified this conclusion after the attendance of on-line tests using the soft-sensor. Nowadays, the soft-sensor operates as viscosity estimator in the control of one production plant line with excellent results

ASSUNTO(S)

polimerização viscosidade redes neurais (computação)

ACESSO AO ARTIGO

Documentos Relacionados