Contribuições ao problema de detecção automática de oscilações em malhas de controle

Abstract

The oscillatory control loop behavior may be caused by different reasons, which are inherent to the industrial processes. Such behavior increases the costs and decreases the quality of the final products. This subject has being widely studied, however even robust algorithms present false results when the analyzed signals are highly influenced by noise, nonperiodic perturbations, or very low frequencies. This work discusses the algorithm Oscilation Detection and Caracterization, a well-known algorithm used to detect and characterize oscillations based on the measurement of the period of the autocorrelated signal obtained from the inverse Fourier transform of the power spectral density. Several drawbacks of the algorithm are overcome through the application of our proposals. The root cause problem is also addressed, comparing the energy of the oscillatory signals calculated in the power spectral density plot. The contributions of this work are highlighted through its application to data of four different industrial processes