Implementação de sistema nebuloso (“fuzzy”) para controle de oxigênio dissolvido no cultivo de Escherichia coli para expressão de proteínas recombinantes

• Ano de defesa: 2022
• Autor(a) principal: Akisue, Rafael Akira
• Orientador(a): Sousa Júnior, Ruy
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Química - PPGEQ
• Palavras-chave em Português: Oxigênio dissolvido; Controle fuzzy; Escherichia coli recombinante
• Palavras-chave em Inglês: Adaptive Neuro-Fuzzy Inference System; Supersys_HCDC; Dissolved oxygen; Fuzzy control; Recombinant Escherichia coli
• Área do conhecimento CNPq: ENGENHARIAS::ENGENHARIA QUIMICA
• Link de acesso: https://repositorio.ufscar.br/handle/ufscar/16658

Due to low oxygen solubility in water and mechanical limitations of a bioreactor, ensuring an adequate dissolved oxygen supply during a recombinant Escherichia coli cultivation is a major challenge in process control. In light of these facts, a fuzzy dissolved oxygen controller was developed based on a decision tree strategy (presented in the literature) and implemented in the cell culture supervision and monitoring system (SUPERSYS_HCDC). The algorithm was coded in MATLAB and the membership function parameters were optimized by the Adaptive Neuro-Fuzzy Inference System tool. The fuzzy controller was composed of tree independent fuzzy inference systems: Princ1, Princ2 – that determined if there would be and increase or a reduction in air and oxygen flow rates (respectively) – and Delta system, that estimated the size of those variations. E. coli cultivation data previously controlled by the decision tree were employed in the training process of the three fuzzy inference systems. After the training process, the first tests with the fuzzy controller consisted of simulations using neural network models of the process (mimicking E. coli cultivation conditions). Results showed that the fuzzification process of the decision tree was successful, resulting in smoother changes in air and oxygen flow rates comparing with those provided by the previous controller (decision tree). In all simulations the dissolved oxygen concentration mean was kept close to its setpoint value of 30% with an attenuation of dissolved oxygen peaks. In order to prove the viability of the fuzzy controller, the initial tests consisted of Saccharomyces cerevisiae (commercial yeast) and E. coli (PspA4PRO strain) cultures, partially supervised by the fuzzy controller. Both cultivations we carried out in a 5 L, in-house, bioreactor and supervised by the SUPERSYS_HCDC software. On average, for the E. coli cultures, the dissolved oxygen concentration was kept close to 30% and its standard deviation was lower than 6%, pointing towards a softening of the peaks observed in the controlled variable. For the robustness tests the optimized fuzzy controller was put in charge of dissolved oxygen control for the entire E. coli cultivation period. The cultivations were carried out without the induction phase and with different induction periods (using IPTG as the inductor). For the non-induced culture, the final cell mass concentration obtained was 27,50 gdry cell weight /L in 13 hours of cultivation. On average, the fuzzy controller kept the dissolved oxygen concentration at about 30,5% with a standard deviation of 6,30%. The attenuation of air and oxygen flow rates steps resulted in a smoother dissolved oxygen profile. For the high cell density cultivation followed by 4.5 hours of induction phase, the maximum cell mass concentration obtained was 35,00 gdry cell weight /L in 16 hours of cultivation. On average, the dissolved oxygen concentration was kept at about 29,80% with a standard deviation of 5,52%. The results presented in this thesis attest not only an increase in flowmeters lifespan (due to their sensibility of abrupt oscillation in the manipulated variables), but also point towards a possible reduction in the metabolic stress suffer by the E. coli (due to its sensitivity to fluctuation in dissolved oxygen concentration). Fuzzy logic has proven successful for controlling dissolved oxygen concentration in bioreactors, an area that lacks new solution to complex problems.