Flávio Vasconcelos da Silva

Desenvolvimento e aplicação experimental de controladores fuzzy e convencional em um bioprocesso

Este trabalho propoe a realizacao de um estudo comparativo do desempenho de controladores Fuzzy e convencional PID aplicados ao controle de temperatura de um processo de precipitacao de bromelina do extrato aquoso de residuos de abacaxi. Uma analise quantitativa da nao-linearidade do processo foi realizada baseada na metodologia de curva de reacao, aplicada em diferentes momentos da batelada, caracterizando o sistema por possuir diferente sensibilidade as acoes de controle ao longo do tempo. O controlador convencional foi sintonizado a partir da aplicacao das equacoes de Ziegler-Nichols aos parâmetros do processo obtidos nos instantes iniciais do experimento, seguido de sintonia fina por tentativa-e-erro. A sintonia do controlador Fuzzy consistiu na alteracao do universo de discurso, na base de regras e na disposicao das funcoes de pertinencia, utilizando-se para isto o conhecimento obtido na analise das curvas de reacao obtidas. Foi observado um melhor desempenho do controlador Fuzzy, apresentando menor valor da integral de erro absoluto multiplicado pelo tempo (ITAE), maior recuperacao de atividade enzimatica e menor consumo de energia eletrica para o resfriamento do sistema.

A fuzzy–split range control system applied to a fermentation process

In this study it was proposed the application of a fuzzy-PI controller in tandem with a split range control strategy to regulate the temperature inside a fermentation vat. Simulations were carried out using different configurations of fuzzy controllers and split range combinations for regulatory control. The performance of these control systems were compared using conventional integral of error criteria, the demand of utilities and the control effort. The proposed control system proved able to adequately regulate the temperature in all the tests. Besides, considering a similar ITAE index and using the energetically most efficient split range configuration, fuzzy-PI controller provided a reduction of approximately 84.5% in the control effort and of 6.75% in total demand of utilities by comparison to a conventional PI controller.

Neural network model predictive control of a styrene polymerization plant: online testing using an electronic worksheet

The batch styrene polymerization process presents transient and nonlinear temperature behavior. In this work, manual control and open loop experiments were carried out in order to build a process knowledge database. Initially, a cascade feedback control loop was implemented by manipulating the thyristor unit of the electrical heater in the thermal fluid tank. Aiming at the MPC development, algebraic equations of a neural network and its adjusted parameters were implemented in an electronic worksheet. Every five seconds, the worksheet was updated with measurements (reactor temperature, thermal fluid temperature and thyristor position) by means of the OLE for the Process Control protocol (OPC). The one-step-ahead temperature prediction was then employed in the objective function of the worksheet solver which used Visual Basic Applications programming. The manipulated variable action was then calculated and sent to the process. A hybrid controller (cascade feedback and MPC) outperformed the pure strategies since the time-integral performance indexes, IAE and ITAE, were reduced by around 22 % and 32 %, respectively. Methodology for the Model Predictive Control presented in this study was considered feasible because the solver of Microsoft Office Excel (2007) is very friendly, easy to understand and ready to implement using VBA.

Experimental assessment and design of multiple model predictive control based on local model networks for industrial processes

This paper presents the experimental assessment of a class of multiple model predictive controllers based on linear local model networks. The control design is established on a clear and easily understandable structure where local models are used to describe the nonlinear process in several different operating points. Thus, simplifying the model predictive control (MPC) algorithm by eliminating the need of a nonlinear optimization strategy, reducing it to a well-grounded quadratic programming dynamic optimization problem. The investigated methodology was tested in an experimental neutralization pilot plant instrumented with foundation fieldbus devices. The steps of models selection, experimental system identification and proper model validation were addressed. In addition, an open-source system used for control calculation was presented. Results regarding the control problem showed that the MPC based on local relevant models was capable of smooth setpoint tracking despite system nonlinearities and a reduced demand of the final control element was attained. It was shown that the technique is easily implementable and can be used to achieve improvements in the control of nonlinear processes at the cost of little modification to the linear MPC algorithm.

Neural network based control of an absorption column in the process of bioethanol production.

Deseja-se recuperar o etanol perdido por evaporacao durante o processo de fermentacao da cana-de-acucar. Para tanto, faz-se uso de uma coluna de absorcao. O controle da concentracao de etanol no efluente gasoso da coluna e realizado pela manipulacao da vazao de solvente, sendo esta determinada pelo controlador nao linear proposto, baseado em um modelo inverso de redes neurais (controlador ANN). Foram feitas simulacoes adicionando-se um sinal de ruido a medida de concentracao de etanol na fase gasosa. Quando perturbacoes degrau foram inseridas na mistura gasosa afluente, o controlador ANN demonstrou desempenho superior ao controle por matriz dinâmica (DMC). Um dispositivo de seguranca, baseado em um controlador feedback convencional, e um filtro digital foram implementados a estrategia de controle proposta para agregar robustez no tratamento de disturbios ocorridos no ambiente operacional. Os resultados demonstraram que o controlador ANN e uma ferramenta robusta e confiavel no controle de uma coluna de absorcao.

FUZZY-PID CONTROLLER APPLIED TO A REFRIGERATION SYSTEM

A refrigeration system exhibits a dynamic behavior on which the variables are interdependent and subjected to oscillation, hence, implicating necessity of changes on operating conditions and undesirable energy expenses. These characteristics ratify the importance of adequate dimensioning and equipment selection to find pre-defined operating conditions such as, the maximum cooling capacity and the evaporating and condensing temperatures. The application of fuzzy control in industrial processes is growing fast in the last decades, mainly in processes whose first principle models require complex methods to be simulated. In these cases, the fuzzy controllers’ capacity of acting correctly based only on expert knowledge and on the capacity of inter-relating all the variables of the process are attractive features. This work presents the experimental development and evaluation of fuzzy-PID controllers for the maintenance of the evaporating temperature in a chiller. The system was submitted to load and set-point disturbances accomplishing an analysis based upon error parameters and transient response. The results showed that fuzzy controllers were adapted satisfactorily.

Scilab/Scicos: An Alternative Tool for Real-Time Monitoring and Advanced Control of Fieldbus Industrial Systems

Abstract The present work deals with real-time data acquisition and advanced control evaluation utilizing the open source scientific platform Scilab. Implementation and visualization of online data with the Scicos toolbox and utilization of OPC technology were discussed. The feasibility and effectiveness of the proposed methodology was shown by means of application of fuzzy controllers to a bromelain enzyme precipitation process instrumented with Foundation Fieldbus devices. Results confirmed Scilab/Scicos suitable for HMI and control systems applied in small industrial applications.

Development of a Predictive Control Based on Takagi-Sugeno Model Applied in a Nonlinear System of Industrial Refrigeration

Refrigeration systems exist in different branches of industry and are characterized as great energy consumers with considerable nonlinear behavior. Several studies have promoted energy costs reduction and minimization of nonlinearities effects in such systems. Model predictive control has been successfully used to stabilize processes in the presence of such nonlinearities; therefore, its application in refrigeration systems is considered promising. In the present study, Takagi–Sugeno models were developed and validated in order to predict the evaporating and secondary fluid temperatures (TE and TP) based on the ANFIS technique (Adaptive Network-based Fuzzy Inference Systems) for a vapor-compressor chiller equipment. The prediction performance of resulting models was analyzed and accessed based on the variance accounted for criteria. These models were then used as the basis for prediction models in several generalized predictive controllers (GPC) denoted here as GPC-ANFIS controllers. Different predictive controllers were designed for different local rules (Fuzzy rules) and the global control action was assumed as the weighted sum of local controllers. Experimental tests considered two distinct controllers, namely the GPC-ANFISTE (evaporating temperature control by means of compressor speed variation) and GPC-ANFISTP (propylene glycol temperature control by means of compressor speed variation), were performed. The experimental tests for setpoint tracking (±1°C) considering 3000 W of constant heat load showed satisfactory results with setpoint deviation around ±0.3°C. Therefore, the ANFIS technique demonstrated to be able to provide reliable predictive models to be used in generalized predictive control algorithms.

Nonlinear System Identification of a Refrigeration System

Applications of advanced control algorithms are important in the refrigeration field to achieve low-energy costs and accurate set-point tracking. However, the designing and tuning of control systems depend on dynamic mathematical models. Approaches like analytical modeling can be time-consuming because they usually lead to a large number of differential equations with unknown parameters. In this work, the application of system identification with the fast recursive orthogonal least square (FROLS) algorithm is proposed as an alternative to analytical modeling to develop a process dynamic model. The evaporating temperature (EVT), condensing temperature (CDT) and useful superheat (USH) are the outputs of interest for this system; covariance analysis of the candidate inputs shows that the model should be single-input–single-output (SISO). Good simulation results are obtained with two different validation data, with average output errors of 0.0343 (EVT model), 0.0079 (CDT model) and 0.1578 (USH model) for one of the datasets, showing that this algorithm is a valid alternative for modeling refrigeration systems.

Identification and Online Validation of a pH Neutralization Process Using an Adaptive Network-Based Fuzzy Inference System

In this study, the application of adaptive neuro-fuzzy inference system (ANFIS) architecture to build prediction models that represent the pH neutralization process is proposed. The dataset used to identify the process was obtained experimentally in a bench scale plant. The prediction model attained was validated offline and online and demonstrated as able to precisely predict the one step-ahead value of effluent pH leaving the neutralization reactor. The input variables were the current and one past value of the acid and base flow rates and the current value of the output variable. Variance accounted for (VAF) indices greater than 99% were achieved by the model in experiments in which the disturbances in the acid and basic solutions flow rates were applied separately. For tests with simultaneous disturbances, conditions never seen in the training and suffering from reactor level oscillations, the prediction model VAF index was still approximately 96%. The validations demonstrated the capability of ANFIS to build precise fuzzy models from input–output datasets. R2 values achieved were always larger than 0.96.