Detecção e diagnostico de falhas em sistemas de processos quimicos. Importancia do conhecimento de estados intermediarios de processos dinamicos. Desenvolvimento de uma metodologia baseada em redes neurais

AUTOR(ES)
DATA DE PUBLICAÇÃO

2000

RESUMO

Artificial Neural Networks (ANN) have demonstrated excellent performance in automatic fault detection and diagnosis in many engineering applications. The great problem when using neural networks, is that neural networks perform only as robustly as the data from which they are trained. This means that is necessary to use as much data as possible in order to cover the wider range of possible operational conditions of the processo Most of the published work on fault detection are related to steady state behavior, and the present work was undertaken to study the process dynamics effects on fault detection. A computational system based on neural networks was developed to analyze the influence of the dynamic behavior of a complex chemical plants to detect and to diagnose faults. An efficient artificial neural network which may be trained through large data sets has also been developed. Although backpropagation using a generalized delta mIe (GDR) for gradient calculation has been popularized as a method of training ANN, it is clear that that this methodology is unsuitable for large data systems. For large data systems we found the great efficiency of neural networks using Radial Basis Functions. Finally, is proposed a methodology which permits to group data from simple lumped systems to study and analyze complex systems which makes possible the detection and diagnosis of the large set of possible of fauIts

ASSUNTO(S)

redes neurais (computação) fault of systems (engineering) neural networks (computation) control of chemical processes controle de processos quimicos differential dinamics systems distillation falha de sistema (engenharia) sistemas dinamicos diferenciais destilação

ACESSO AO ARTIGO

Documentos Relacionados