Search

This paper presents the use of a radial basis function artificial neural network to estimate sensor readings exploring the analytical redundancy via auto-association. However, in order to guarantee optimal performance of the network, the training and optimization processes have been modified. In the conventional training algorithm, even if a stop criterion, such as summed squared error, is reached, one or more of the individual performance metrics, including: i) accuracy; ii) robustness; iii) spillover and iv) filtering of the neural network, may not be satisfactory while validating sensor measurements. Essentially, the proposed modification in the training algorithm is based on seeking to ensure that one or more of the metrics are met. This paper describes the proposed algorithm including all of its mathematical foundation. Afterward, a data set of a water injection pump for an oil and gas processing unit was used to train the RBF network using the conventional and the modified algorithm, and the performance of each was evaluated. Furthermore, the AAKR model is applied to the same dataset as a quality reference parameter. Finally, a comparison analysis of the developed models is presented for each of the performance metrics, as well as for overall effectiveness, demonstrating that the main advantage of the proposed approach is to obtain the estimation results equivalent or superior to the AAKR with shorter runtime and the disadvantage of having higher complexity during the model training.

Frequency domain multiplexing (FDM) allows simplifying the data acquisition in vector control when the resolver is used as a position sensor. However, the crosstalk reduction in the demultiplexing process is still a difficult task. This article proposes an improvement of the FDM-based acquisition system through synchronous sampling and Differential Evolution (DE). The resolver signals and the PWM carrier are synchronized so that the resolver outputs are zero when the PWM carrier reaches its valleys. Thus, the fundamental component of the stator currents can be easily estimated from the signals sent to the acquisition system. These estimated current signals and an angle tracking observer (ATO) are used to estimate the angular position. DE and a discretized model of the ATO are applied in order to improve the angle estimation transient response and to reduce crosstalk. Conditions to apply the proposed multiplexing approach will be defined. Simulation and experimental results show that the proposed approach allows an accurate estimation of the current fundamental components and the angular position when FDM is used, with better performance respect to other approaches in the literature. This is the first time that a DE is used to tune an ATO.

Transportation electrification is gaining notoriety in recent years as to fulfill the aim of reducing the fossil fuels consumption and to increase energy efficiency. Accordingly, integrated starter alternator is a compelling method to convert conventional vehicles into hybrid ones. A deal of benefits can be reached by using ISA systems. In general, it provides vehicles performance enhancement and also reduces fuel consumption. This paper aims to present simulation results and preliminary experimental achievements with a power electronics-based prototype driving a permanent magnet synchronous machine (PMSM). The ISA operation is demonstrated throughout providing cranking torque and assistant torque for constant power and constant torque conditions and afterward the generation mode. Details of the hardware design and control strategies, which include an ADALINE neural network for properly find the mechanical position and the application of a metaheuristic technique to tune the controllers of the motor control, have been also addressed.

The electrocardiogram (ECG) is basic equipment used in the diagnosis of cardiac illness. However, in non-developed countries, most of the population does not have access to medical tests, and many hospitals do not even have these ECGs. On the other hand, the electrical signals generated by the heart and acquired by the ECG have low power and are affected by electromagnetic interference (EMI), mainly produced by the electrical system. Filtering EMI when frequency varies is a challenging task. Within this context, this work aims to produce an easy-to-use low-cost ECG with good electromagnetic disturbances rejection. The proposed noise rejection system is composed of two moving average filters and a phase-locked-loop, namely 2MAV-PLL. The system operates with a low sampling frequency and attenuates the EMI noise present in the ECG signal regardless of the amplitude, obtaining a filtered signal with a 44-dB signal–noise ratio (SNR) between the frequencies of± 10 Hz of the fundamental frequency. Simulation and experimental results prove that the ECG system can attenuate the EMI using relatively low sampling frequency, giving adequate information for health professionals to properly evaluate an electrocardiogram.

Recently, methods based on Convolutional Neural Networks (CNN) achieved impressive success in semantic segmentation tasks. However, challenges such as the class imbalance and the uncertainty in the pixel-labeling process are not completely addressed. As such, we present a new approach that calculates a weight for each pixel considering its class and uncertainty during the labeling process. The pixel-wise weights are used during training to increase or decrease the importance of the pixels. Experimental results using different datasets like aerial, terrestrial, and ultrasound images show that the proposed approach leads to significant improvements in three challenging segmentation tasks in comparison to baseline methods. It was also proved to be more invariant to noise. The approach presented here may be used within a wide range of semantic segmentation methods to improve their robustness.

Recently, methods based on Convolutional Neural Networks (CNN) achieved impressive success in semantic segmentation tasks. However, challenges such as the class imbalance and the uncertainty in the pixel-labeling process are not completely addressed. As such, we present a new approach that calculates a weight for each pixel considering its class and uncertainty during the labeling process. The pixel-wise weights are used during training to increase or decrease the importance of the pixels. Experimental results show that the proposed approach leads to significant improvements in three challenging segmentation tasks in comparison to baseline methods. It was also proved to be more invariant to noise. The approach presented here may be used within a wide range of semantic segmentation methods to improve their robustness., Comment: 15 pages, 9 figures, 3 tables

Frequency-domain multiplexing was proposed to reduce the number of signals to be digitized in vector control using a resolver as position sensor. Signals from resolver and current sensors can be multiplexed through analog adders. This approach has many advantages over using analog switches. On the other hand, the conventional demultiplexing system involves filtering and delay compensation, while the rotor angle is obtained after the signal demultiplexing. However, this conventional approach is difficult to implement because the required filters must have good stopband rejection, without adding a significant phase delay. This paper proposes an improved demultiplexing system whose structure is simpler than the conventional approach. A type-III angle tracking observer estimates the rotor angle and the resolver outputs directly from the multiplexed signals. The estimated resolver outputs and stopband filters are used to get the current signals. Thus, the demultiplexing of resolver signals and the delay compensators used in the conventional approach were eliminated. Experimental results prove that the proposed approach gives accurate estimations of the rotor angle and current signals. Besides, conditions for the resolver excitation frequency, the inverter switching frequency, and for current synchronous sampling were defined for the application of frequency-domain multiplexing.

This paper presents comparative analysis among digital current controllers for single-phase grid connected photovoltaic inverters with L-filter. The considered controllers are proportional and integral controller (PI), proportional and integral plus resonant controller (PI+Res), deadbeat control (DBC) and model predictive control (MPC). The considered parameters for analysis are the total harmonic distortion (THD) for the injected grid current, phase and amplitude error in steady-state and during power reference step change. Simulations were made through Matlab/Simulink® platform to guide novel projects.

The present work deals with the simulation of a photovoltaic distributed generation system at the micro generation level, using a DC-AC power converter responsible for injecting active power into the grid and acting as a shunt active power filter, mitigating the harmonics of local nonlinear loads. For this, the system has a reference current generator that employs an algorithm based on the synchronous reference frame. The current reference has a distorted portion, corresponding to the distortion power of the nonlinear load added to the active portion to be injected to the grid. A PI plus multi-resonant controller are inserted to reproduce correctly the current reference. In order to enhance stability and waveform quality a genetic algorithm is adopted to find the best gains for the controllers. Simulations throught MatLab/Simulink® are presented to confirm the feasibility of the proposal.

Direct torque control (DTC) allows having a speed and torque control with good dynamic response. However, DTC uses a square root function to get the motor stator flux. This function is difficult to implement in microcontrollers or FPGAs. Hence, this paper proposes the use of SPLINE interpolation to implement the square root function with low computational cost, in order to get the magnitude of the stator flux. In SPLINE, the function is divided in intervals, and each interval is approximated using polynomial function. Besides, different discretization techniques to implement the integrator used in flux estimation will be tested in order to show their effects in flux and torque estimation. Simulations show that the proposed approach gives good estimations of torque and flux.

Photovoltaic generation systems under partial shading conditions are difficult to optimize using conventional maximum power point tracking (MPPT) algorithms. Most of the techniques developed for these conditions fail to track dynamically the MPP and leads to energy losses during normal operation. This paper presents hybrid MPPT algorithms combining global and local MPPTs to extract the most available energy from the system under any condition. The methods are compared through simulations using Matlab/Simulink®, where the tracking factor (TF) and power characteristics during time are evaluated.

Simplification in the acquisition system allows increasing the robustness and the sampling frequency of a motor drive system. Frequency domain multiplexing (FDM) allows reducing the number of signals to be digitized in vector control, simplifying the analog-to-digital system in a motor drive. Nonetheless, the required demultiplexing process is still an open question. This paper proposes a simple demultiplexing technique based on synchronous current sampling. The triangular carrier used in modulation and the resolver excitation signal are synchronized, so that the valley of the carrier happens when the resolver outputs are equal to zero. Thus, the demultiplexed current signals are equal to the samples obtained through synchronous current sampling, eliminating the harmonics in the current signals. The estimated current signals and an angle tracking observer are used to estimate the angular position without the previous demultiplexing of the resolver output signals. Simulation and experimental results show that the demultiplexing process has high accuracy, rejecting the current harmonics and gives a good estimation of the angular position.

This work presents the use of radial basis function artificial neural network to estimate the sensors readings, exploring the analytical redundancy via auto association. However, in order to guarantee good performance of the network the training and optimization process was modified. In the conventional training algorithm, although the stop criteria, such as summed squared error, is reached, one or more of the individual performance metrics, including: i) accuracy; ii) robustness; iii) spillover and iv) filtering matrix of the neural network may not be satisfactory. The paper describes the proposed algorithm including all the mathematical foundation. A dataset of a petroleum refinery is used to train a RBF network using the conventional and the modified algorithm and the performance of both will be evaluated. Furthermore, AAKR model is used to the same dataset. Finally, a comparison study of the developed models will be done for each of the performance metrics, as well as for the overall effectiveness in order to demonstrate the superiority of the proposed approach.

This paper proposes a tracking angular position controller for Permanent Magnet Synchronous Motor (PMSM). Instead of linearizing the dq model of the motor, the proposed approach compensates the non-linear terms of the model in order to use robust linear control techniques. Thus, the control laws are composed by two components: the first term compensate the nonlinearities of the dq model, while the second component is defined by linear controllers to get the desired position and d-axis stator current. A type-II internal model controller is used to control the angular position on the motor. Hence, the motor can track even sinusoidal position references. Simulation and experimental results prove the accuracy of the proposed tracking controller.

This paper proposes a simple nonlinear technique to control the angular position and d-axis stator current of a permanent magnet synchronous motor. Instead of making a linearization according to an operation point, the control laws, i.e., the dq-axis stator voltages are composed by two terms. The first one is to compensate the nonlinearities of the PMSM. The second component of the control law is determined by a linear controller to get the angular position. Thus, the design of the controller becomes easier. Simulation results prove the accuracy and robustness of the proposed angular position controller.

Resolver is an angular position sensor widely used in electric/hybrid vehicles and robotics. However, estimating the angular position and speed from resolver outputs is as difficult task. Most algorithms in literature use trigonometric functions to estimate the angular position. However, these functions are difficult to implement, especially in FPGAs. On the other hand, in speed vector control of three-phase motors, the angular position is only used to estimate the mechanical speed and Park transformation matrix. This paper presents a simple technique to get the mechanical speed and Park matrix without calculating trigonometric functions and without estimating directly the angular position. Filtering and synchronous demodulation allows obtaining the sine and cosine of the angular position, while the compensation of the delay produced by filtering and the estimation of the Park matrix are done using simple trigonometric properties and taking advantage that the number of pole pairs is an integer number. Experimental results show the accuracy of the proposed approach.

Este artigo apresenta resultados da implementacao em hardware de um Controlador Preditivo baseado em Modelo (MPC) com e sem observador de estado, para um sistema turbina gerador, usando linguagem VHDL sintetizados em FPGA. O objetivo deste trabalho e a implementacao de controle preditivo em FPGA, e realizar testes usando cosimulacao (o processo e simulado no ambiente Simulink, enquanto que o controlador e implementado no programa Modelsim), e FPGA-in-the-loop (semelhante a cosimulacao, mas o controlador e embarcado em FPGA). A implementacao digital do MPC e uma tarefa dificil, pois as amostras das variaveis de estado e de saida da planta estao atrasadas em relacao a estimacao da lei de controle e os calculos matematicos necessarios no MPC. Para solucionar o problema de atraso, adicionou-se um contador de eventos ao MPC. Assim, e possivel fornecer no tempo correto os valores das variaveis de estados, atuais e passadas, necessarios para o calculo da lei de controle. Em relacao as operacoes matriciais desenvolveram-se adequacoes de bibliotecas especificas da Altera. Os resultados mostram que a lei de controle do MPC, calculada em VHDL, e semelhante ao obtido em MATLAB.

This paper presents the design of a Fuzzy speed controller for a permanent magnet synchronous motor (PMSM), based on a MIMO model of the motor, internal model controller (IMC) and Takagi-Sugeno Fuzzy modeling technique. The outputs of the MIMO of the PMSMl are the d-axis stator current and the mechanical speed, which are required to control the speed of PMSM below and above its nominal speed. A MIMO model allows a better representation of the influence between these outputs signals. Takagi-Sugeno Fuzzy controllers and internal model controllers allows reducing the steady-state error. Linear matrix inequalities were used to get the local controllers and set the dynamics of the close-loop plant. Simulation and experimental tests using hardware-in-the-loop prove the accuracy of the proposed controller, even considering variation in the motor parameters.

Space Vector Pulse Width Modulation (SVPWM) is widely used in three-phase inverters. Besides, Field Programmable Gate Array (FPGA) is a powerful tool in the development of digital controllers applied to power electronics. However, the teaching of these topics for undergraduate and postgraduate students is a difficult task. For this reason, this paper presents a new educational platform based on the integration of a Graphical User Interface (GUI) and FPGA for the didactic teaching of SVPWM. The user defines the DC voltage source of the inverter, the magnitude and angle of the reference space vector in the GUI. As a result, the GUI shows the modulation index, working sector, duty cycle and the switching pulses in each phase. Simultaneously, the results of these calculations are sent to the FPGA through serial communication. Thus, the switching patterns are also generated in the FPGA in real time, which can be used for experimental tests. These experimental tests show that the pulses generated for each phase by the FPGA and the GUI are equivalent. Therefore, the proposed didactic platform helps in the teaching process and motivates the use and research of both SVPWM algorithms and FPGAs for undergraduate and postgraduate students.

This paper presents a vector control system for the control of the capacitor voltages of a 7-level Packed-U-Cell (PUC) three-phase inverter. Usually, PUCs in a three-phase system are controlled as isolated systems. As a result, three voltage loop and three load current loops are required to control the capacitor voltage in each PUC. Additionally, conventional approach requires the tracking of sinusoidal references. This paper proposes a different approach. A single voltage loop is designed to control the average voltage of the capacitor. This loop defines the required tf-axis load current, while the required q-axis load current is set to zero. Two current controllers generate the voltage references to get the required dq currents. Those voltages are produced through hybrid modulation (HPWM). As a result, only three control loops are needed to control the capacitor voltages. The proposed approach reduces the computational cost of the PUC controller, makes easier the control tuning, and the current control improves, due to the steady-state dq load current referents are almost constants, which are easier to track than sinusoidal references. Simulation results show that the proposed vector control for a three-phase PUC inverter is accurate and robust against parameter variation.

This paper deals with the design and implementation of an artificial neural network for pattern classification in the FPGA EP2C20F484C7. A perceptron with quadratic decision boundary is used as nonlinear classification system, but using a hardlims as activation function, instead of a sigmoid function. The training algorithm is similar to the used in conventional perceptron. The elimination of the sigmoid function makes simpler the implementation of quadratic perceptrons. As the mentioned FPGA does not do neither float-point nor fixed-point multiplications, the synaptic weights were normalized to integers. The proposed quadratic perceptron is tested in a set of classification problems and compared with multilayer perceptron. Example of experimental implementation of the proposed classification system is shown, including parameters about computational cost.

Este artigo apresenta um algoritmo simplificado da modulação por largura de pulso por vetores espaciais (SVPWM) para um inversor trifásico de dois níveis, o qual pode operar em submodulação e sobremodulação. Em outras simplificações achadas na literatura, as tensões de referência são modificadas e comparadas com um portador triangular para estimar os estados de comutação do inversor. Não obstante, este artigo propõe a modificação do portador, em lugar das referências. Este procedimento reduz o número de operações matemáticas e aumenta a velocidade de execução do algoritmo SVPWM em DSPs ou FPGAs. As ondas de referência são senoidais, ainda no modo de sobremodulação. Resultados de simulação e experimentais demonstram que a simplificação proposta produz o mesmo padrão de chaveamento que o SVPWM convencional, é mais simples e rápida que outras simplificações. ABSTRACT - This paper presents a simplified algorithm of space vector pulse width modulation (SVPWM) for a two-level three-phase inverter, which can operate in undermodulation and overmodulation modes. In other simplifications founded in literature, the reference voltages are modified and compared with a triangular carrier to estimate the switching states of the inverter. However, this paper proposes the modification of the carrier signal instead of the references. This procedure reduces the number of mathematical operations and increases the execution speed of SVPWM algorithm in DSPs or FPGAs. The reference voltages are sinusoidal, even for overmodulation mode. Simulation and experimental results proves that the proposed simplification produces the same switching patterns than conventional SVPWM, is simpler and is faster than other simplifications.

This paper presents a new technique for load torque estimation applied on the position control of a Permanent Magnet Synchronous Motor (PMSM). The estimator makes use of the different properties of mechanical and electromagnetic torque in frequency domain, extracted by a filterbank. This information is processed by two real-time ADALINEs, which estimate the rotor inertia and the load torque. The proposed estimator is quite simple to implement, has a good accuracy and only depends on the electromagnetic torque and mechanical acceleration. That estimator can be used for others AC motors. Simulations in MATLAB/SIMULINK and experimental results show its performance.

A resolver is an angular position sensor used in applications that demands robustness and reliability, such as electric vehicles. However, getting the angular position from resolver output signals is a difficult task, and many algorithms were proposed to achieve that task. This paper describes the application of model predictive control (MPC) to get the angular position from resolver signals. Synchronous demodulation is used to get the envelopes of the resolver outputs and get the estimation error. The structure of the conventional model predictive controller was modified to be used as an angle tracking observer (ATO). Simulations show the performance of the proposed approach. According to the bibliographic review of the authors, it is the first time that model predictive control is applied as an observer to get the angular position from resolver signals.

The emergence of SARS-CoV-2 in 2019 increased the large-scale need for pulmonary ventilators, as the disease affects the respiratory system. Brazil and other countries in the world do not have as many pulmonary ventilators due to the high cost, and at the moment due to their unavailability to purchase. Faced with that problem, different academic institutions and companies around the world have developed emergency pulmonary ventilators based on different operating principles, many of them are based on automating the operation of a handheld lung resuscitator. However, local-regional development of emergency pulmonary ventilators is needed to serve as many patients as possible within a state, and to reduce technology import costs and times. This work aims to design and manufacture a pulmonary ventilator for emergency use, which uses a handheld lung resuscitator, to implement the intelligent FiO2 system, responsible for dosing the amount of O2 in the resuscitator, its control is performed by a Fuzzy controller using the PEEP and SpO2 input variables, for its implementation in a bench using components from the Brazilian market. Results were generated through simulation of the Fuzzy logic for the FiO2 output and a bench test was performed, simulating a patient in which the system was able to perform ventilation and FiO2 control. Key-words: SARS-CoV-2, emergency mechanical ventilator, FiO2, fuzzy logic. O surgimento da SARS-CoV-2 em 2019 aumentou a necessidade em larga escala por ventiladores pulmonares, uma vez que a doença afeta o sistema respiratório. O Brasil e outros países do mundo não possuem tantos ventiladores pulmonares devido ao custo elevado, e no momento atual por sua indisponibilidade de compra. Ante aquele problema, diferentes instituições acadêmicas e empresas no mundo desenvolveram ventiladores pulmonares de emergência baseados em diferentes princípios de operação, sendo muitos deles baseados na automatização da operação de um ressuscitador pulmonar de mão. Contudo, o desenvolvimento local-regional de ventiladores pulmonares de emergência é necessário para atender a maior quantidade de pacientes possíveis dentro de um estado, e para reduzir custos e tempos de importação de tecnologia. Esse trabalho tem o objetivo de projetar e fabricar um ventilador pulmonar de uso emergencial, o qual utiliza um ressuscitador pulmonar de mão, implementar o sistema de FiO2 inteligente, responsável por dosar a quantidade de O2 no ressuscitador, seu controle é realizado por um controlador Fuzzy utilizando as variáveis de entrada de PEEP e SpO2, para sua implementação em bancada utiliza componentes do mercado brasileiro. Foram gerados resultados por meio de simulação da lógica Fuzzy para a saída de FiO2 e foi realizado teste em bancada, simulando um paciente o qual o sistema conseguiu realizar a ventilação e o controle de FiO2. Palavras chaves: SARS-CoV-2, ventilador de uso emergencial, FiO2, lógica fuzzy.