COMPOSITION CONTROL BY TEMPERATURE INFERENCE USING FUZZY CONTROLLERS IN THE ETHYLBENZENE PRODUCTIVE PROCESS
Paulo Romero de Araujo Mariz1; Emanuella Francisca Lacerda Vieira1; Jonas Laedson Marinho da Silva Santos1; Marcelo da Silva Pedro1; Walter Yanko de Aragão Brandão1; Leopoldo Oswaldo Alcázar Rojas1; Arioston Araújo de Morais Júnior2
1 Universidade Federal da Paraíba; 2 Universidade Federal de Santa Catarina
doi:10.20906/CPS/CILAMCE2017-0240
Resumo
The nonlinearity, coupling, slow dynamics and interactivity are process characteristics that cause the need to develop new technologies and advanced control algorithms, capable to minimize these particularities. In these cases, fuzzy controller is suited to be employed, especially for multiple inputs and multiple outputs systems (MIMO), which there are significant interactions between the manipulated variables and controlled. In this paper, the methodology applied in the ethylbenzene productive process (EB), that is produced via alkylation reaction of benzene, was the employ of the fuzzy control algorithms and classical PID (proportional-derivative-integrative) technology and comparing their performance. After the chemical reaction, which occurs in two continuous stirred tank reactor (CSTR), the EB follows to the separation stage in two multicomponent distillation columns, C1 and C2, and should be recovered in the top stream of the second distillation column (C2), producing a high purity component (≥99.9%). However, since the composition variable is an analytical quantity, its measurement and control are not performed directly, and the use of inferential temperature control is common to keep the compositions in its Setpoint. Through the techniques of selecting variables - decomposition of singular values (SVD), Relative Gain Array (RGA) and Non Square Relative Gain Array (NRG) - the better pairs of manipulated variables (MV) and process variables (PV) were selected, after performing the control loop sensitivity analyses in the distillation columns C1 and C2. The selection step of variables proved satisfactory since every techniques of variable selection (SVD, RGA and NRG) corroborated with the results. Finally, the nonlinear fuzzy controller was compared to the classical PID, presenting relevant minimization of transient disturbances, being employed criteria for evaluating the integral of the error to verify the quality of the controllers adjustment.
Palavras-chave: Ethylbenzene process; Fuzzy controller; SVD; RGA; NRG; Inferential control