Sintonia do filtro de Kalman para medição indireta das variáveis de estado no banho eletrolitico

Abstract

A tunning model for Kalman filter based on QR duality principle is presented to measure the state variables of the electrolytic bath in aluminium production cells. The main goal is to establish a set of filter gains that better represents the percentual of alumina in the bath. The filter bandwidth tunning is performed by increasing or decreasing the filter bandpass from the Q and R variations. The design and analysis of the Q and R covariance matrices are performed to find a pattern of the resistance variations that could be associated with the alumina concentration in the bath. The technical solution encloses on-line evaluation of the Kalman filter in order to prove its capableness of response when used to control real production cells. The Standard Kalman is coded upon a scalar form to reduce the use of computing resources when the filter is processed. The line current and pot voltage are directly read from the hardware interface and then converted in a third variable, the resistance, which is used to infer the alumina concentration in the bath. Moreover, the filter implementation goes in the direction of practical aspects limits of the indirect measurement system implementations, its robustness is appraised by observability, roundo® and modeling errors.