DIFFERENT CONFIGURATIONS OF FUZZY COGNITIVE MAPS CONTROLLERS APPLIED IN INDUSTRIAL PROCESSES
Márcio Mendonça1; Lucas Botoni de Souza1; Lillyane Rodrigues Cintra1; Rodrigo Henrique Cunha Palácios1; Asmaa Mourhir2
1 UTFPR-CP; 2 Al Akhawayn University
Baixar PDF doi:10.20906/CPS/CBA2018-0889
Resumo
In this work, we present different configurations of Fuzzy Cognitive Maps (FCMs) controllers of three industrial pro-cesses. A first-order heat exchanger (Heatex) with a classic FCM compared with a Fuzzy Logic Controller (FLC) and the original PI controller designed by MatWorks®,®, using the Integral Absolute Error (IAE), the Integral Squared Error (ISE), settling time, overshoot and control signal. The second process is a fourth order alcoholic fermenter, using FCM as an adjustment mechanism for a PID controller tuned heuristically, using the same comparison parameters than the Heatex. The last process is a mixer tank, with two methods for Dynamic Fuzzy Cognitive Maps' (DFCMs) weight adjustment: the Hebbian learning algorithm (DFCM-Heb) and a weight-scheduling (DFCM-WS) configuration. In this process, the most satisfactory method (DFCM-Heb) was embedded in a PIC18F4520 to verify the low computational complexity of the DFCM. For the Heatex, the PI con-troller still the best option, however, the three controllers presented similar results. In the alcoholic fermenter process, the FCM-PID mechanism obtained the most satisfactory responses
Palavras-chave: Heat exchanger; alcoholic fermenter; mixer tank; Fuzzy Cognitive Maps; adaptive control