Fábio Meneghetti Ugulino de Araújo

Fuzzy Wavelet Neural Network Using a Correntropy Criterion for Nonlinear System Identification

Recent researches have demonstrated that the Fuzzy Wavelet Neural Networks (FWNNs) are an efficient tool to identify nonlinear systems. In these structures, features related to fuzzy logic, wavelet functions, and neural networks are combined in an architecture similar to the Adaptive Neurofuzzy Inference Systems (ANFIS). In practical applications, the experimental data set used in the identification task often contains unknown noise and outliers, which decrease the FWNN model reliability. In order to reduce the negative effects of these erroneous measurements, this work proposes the direct use of a similarity measure based on information theory in the FWNN learning procedure. The Mean Squared Error (MSE) cost function is replaced by the Maximum Correntropy Criterion (MCC) in the traditional error backpropagation (BP) algorithm. The input-output maps of a real nonlinear system studied in this work are identified from an experimental data set corrupted by different outliers rates and additive white Gaussian noise. The results demonstrate the advantages of the proposed cost function using the MCC as compared to the MSE. This work also investigates the influence of the kernel size on the performance of the MCC in the BP algorithm, since it is the only free parameter of correntropy.

A Modified Matricial PSO Algorithm Applied to System Identification with Convergence Analysis

Recently, several evolutionary computation techniques have been used in research areas such as parameter estimation of linear and nonlinear dynamic processes. This motivates the use of algorithms such as the particle swarm optimization (PSO) in the aforementioned fields of knowledge. However, little is known about the convergence of this algorithm, and mainly the analyses and studies have focused on experimental results. Therefore, the objective of this work is to propose a structure for the PSO that better analyze the convergence of the algorithm analytically. For this, the PSO is restructured to assume a matrix form, reformulated as a piecewise linear system. There was a convergence analysis of the algorithm as a whole, using an almost sure convergence criterion applicable to switched systems. Subsequently, traditional parameter identification algorithms were combined with the matricial PSO (MPSO), so as to make the identification results as good as or better than identifying only using the PSO or only the traditional algorithms. The obtained functions, after the identification, using the MPSO algorithm combined with the conventional identification algorithms, presented a better generalization and proper identification. The conclusions reached were that the hybridization permits a minimum performance and also contributes to improve the results obtained with the traditional algorithms, allowing the system representation in a higher range of frequencies.

Model Reference Adaptive Control with inverse compensation applied to a pH plant

In this paper, the Model Reference Adaptive Control (MRAC) with dead-zone compensation is proposed to improve two MRAC variations: MRAC with fixed adaptive gain (γ) and MRAC with variable γ based on η-adaptive optimization algorithm. The proposed MRAC is often used to compensate the dead-zone effect on valve that is not taken into consideration in the usual MRAC. It was also implemented a compensation in inverse dynamic on dead-zone. The nonlinearity inverse model is estimated by method of Recursive Least Squares (RLS). All comparisons and validations were made based on results collected from a simulation and developed in accordance with two models: a simplified Hammerstein model and a phenomenological model of Wiener. To verify the proposed controllers with inverse compensation performance we applied to a pH neutralization process. Simulations have been presented to confirm the effectiveness of the technique.

A nonlinear system identification approach based on Fuzzy Wavelet Neural Network

This paper concerns the use of an alternative Fuzzy Wavelet Neural Network (FWNN) to model the input-output maps of nonlinear dynamic systems. The analyzed structure uses only wavelet functions in the consequent part of its fuzzy rules. The advantages and disadvantages of using this FWNN in model identification tasks are listed considering a comparative study performed with other FWNN structures found in literature. The evaluations are carried out using a real multisection liquid storage tank with abrupt transitions between its sections. The analysis is based on usual criteria such as: mean quadratic error, number of training epochs, number of adjustable parameters, quadratic error variance, among others. The results indicate that the modified FWNN structure maintains the capability of generalization and other important characteristics presented by traditional networks FWNN, despite the reduction in the complexity of the structure.

Hierarchical Fuzzy Control

Growing demands for comfort, reliability, accuracy, energy conservation, safety and economy have fueled interest in proposals that can contribute to facilitate high performance control systems design. In terms of vibrations active control, it may represent, for example, a good relationship between the maximum reduction in vibrations transmission between two systems and the minimum energy expended in order to accomplish this reduction [1].

Optimal Design of Digital Low Pass Finite Impulse Response Filter using Particle Swarm Optimization and Bat Algorithm

In this paper, the traditional metaheuristic Particle Swarm Optmization (PSO) and the Bat Algorithm (BA) are used to optimal design digital low pass (LP) Finite Impulse Response (FIR) filters. These filters have a wide range of applications because of their characteristics. They are easy to be designed, they have guaranteed bounded input-bounded output (BIBO) stability and can be designed to present linear phase at all frequencies. Traditional optimization methods based on gradient are susceptible to getting trapped on a local optima solution when they are applied to optimize multimodal problems, such as the FIR filter design. Here, to overcome this drawback, the aforementioned metaheuristics are adopted to obtain the coefficients of low pass FIR filters of order 20 and 24. The performance of BA and PSO algorithms are compared with the classical Parks and McClellan (PM) filter design algorithm, which is a deterministic procedure. For this comparison is considered the filters pass band and stop band ripples, transition width and statistical data. The simulation results demonstrate that the proposed filter design approach using BA algorithm outperforms PM and PSO.

Hybrid fuzzy and robust controller applied to a DC-DC boost converter

This paper purposes a hybrid control method applied to a DC-DC Boost converter in order to maintain a constant output voltage and ensure the supply of good power quality to the load. The hybrid controller consists of a fuzzy supervisor which mixes control signals of two types of controllers: robust and fuzzy. Each one of these controllers has good performance in certain operational conditions. The supervisor determines the ideal proportion of the action provided by the controllers in each operational condition. The simulations show that fuzzy and robust control signals can be mixed to provide a good power quality to the load ensuring fast transient response and zero steady-state error due to input voltage and load variations.

Nonlinear Predictive Control System for Stiction Compensation in Electropneumatic Control Valves

This paper presents the implementation of a system that deals with static friction (stiction) in electropneumatic control valves, one of the most common nonlinearities that causes problems such as limit cycles and consequently wear of the valve and its moving parts, as well as losses in production and maintenance costs. This system is composed of a nonlinear predictive controller with adjustable constraints and an online database for estimation of the stiction parameters. The predictive controller uses constraints on the valve speed during its excursion, as well as constraints on the control signal to bring the valve to the desired position and slip it when necessary. The strategy adopted also showed robustness, being able to cope with changes in the spring and stiction parameters, which caused mismatch between the model and the controller and consequently loss of performance or even instability.

Controle Fuzzy-GA para Supressão de Vibrações com Acompanhamento de Referência

Resumo Neste trabalho são apresentados três controladores nebulosos, do tipo Takagi-SugenoKang, utilizados para suprimir vibrações mecânicas em um sistema eletro-mecânico baseado no princípio da alavanca. Buscando-se um melhor desempenho do sistema controlado e a simplificação do processo de sintonia desses controladores foram utilizados algoritmos genéticos na otimização de suas funções de pertinência, de entrada, sua memória associativa nebulosa e suas funções Sugeno, de saída. Esse sistema de controle demonstra então, a aplicação de duas técnicas de inteligência artificial para resolver um problema em que técnicas de controle linear podem não propiciar uma solução adequada ou exigir um alto grau de complexidade para que essa solução seja alcançada. Foram realizadas várias simulações digitais para se avaliar o desempenho do sistema controlado, avaliou-se a supressão de vibrações no que diz respeito ao rastreamento de referência na presença de distúrbios, comprovandose a eficácia do algoritmo proposto na otimização de controladores nebulosos. Foi feita uma comparação entre os resultados obtidos com os diversos controladores.

Redes Neurais Artificiais para Controle de uma Planta de Nível

Resumo Este trabalho pretende mostrar a utilização de Redes Neurais Arti ciais para o controle de uma planta de nível. A idéia é substituir controladores PID já existentes por redes neurais arti ciais que, inicialmente, copiem a dinâmica dos mesmos. Os resultados apresentados mostram o desempenho dos controladores neurais quando aplicados à planta simulada e à planta real. Index Terms Controle Neural, Backpropagation, Controle de Nível.