Detecção e isolação de falhas utilizando estimação de parâmetrosde modelos contínuos e limites de confiança adaptativos - aplicaçãoa uma Planta Piloto

Abstract

This work describes the development and implementation of a fault detection and isolation method applied to an interative tank pilot plant. This method is based on recursive parameter estimation of a continuos-time mathematical model. Some problems related to the estimation of continuos-time model parameters were found. The main problem was the difficulty of estimating time derivates of noisy signals,and this was solved with a state variables filter. Specific tunings were made to the parameter estimation algorithm thus improving the fault detection and isolation method. A procedure for fault isolation was developed and yielded promising results. This procedure is able to quantify a fault as weak, median or serious. Different tests were made using the plant for algorithm tuning. Typical results of each fault are presented and discussed.