This work focuses on studying and proposing a circuit configuration and reducing the number of controlled switches for single-phase, transformerless uninterruptible power supplies (UPS) using double-conversion topology. In this context, the inclusion of an additional static switch, strategically positioned to enable a new circuit configuration forming the converter, will allow the battery bank to be discharged using two arms.

The classic UPS configuration with double-conversion topology uses four arms in total. Three of them are responsible for forming the rectifier and inverter, while the fourth arm is dedicated to charging the battery bank in mains mode. Normally, classic double-conversion UPSs are described with a static switch responsible for changing the system's operating mode, of which there are three (mains mode, battery mode, bypass mode). With the addition of another static switch, the fourth arm is used in battery mode to reduce the ripple of the battery bank current during discharge, improving the efficiency of the converter. This means that the UPS uses all six controlled switches in both operating modes.

For the proposal, the dynamic models and/or topological state analyses are presented, and the PWM strategies and control strategies are developed. As far as PWM strategies are concerned, a sine-delta modulation is applied which varies according to configuration and operating mode, and can be a function of the polarity of the phase currents or just the pole voltages. The control system uses the classic continuous control technique in the frequency domain, while the mains signals (voltage and current) are kept in sync with the load via the PLL structure.

The telecommunications technologies have advanced to the point of enabling networks specialized in data transmission in the vicinity of the body. This communication method, known as Body Area Network (BAN), is evolving rapidly, requiring greater portability and longer battery life. This, in turn, drives the need for reduced energy consumption. For this reason, solutions have been studied over the past three decades to improve energy eficiency in signal transmission. Among these technologies, the Human Body Communication (HBC) technique stands out in recent academic research. HBC is a wireless communication method for body networks that offers advantages in terms of consumption, security, and interference compared to other radiation-based methods. This advantage arises from using the human body as the signal transfer channel. In this work, we describe the design and post-layout simulations of a receiver for HBC with capacitive coupling, developed in 180 nm CMOS technology, capable of demodulating BFSK signals. An architecture is employed that performs frequency-phase conversion, exploiting the phenomenon of injection locking. The designed receiver incorporates an input amplifier to raise the received signal to the levels required for injection locking in a Voltage Controlled Oscillator (VCO). With injection locking, a phase difference will be present between the VCO signal and the injected signal. This phase difference is identified by a phase detector, determining whether the VCO signal is delayed or advanced. After filtering and comparing the output signals of the phase detector, the demodulation of the BFSK signal is obtained. Post-layout simulations of the circuit demonstrate the receiver's capability to operate in the range between 2 40 and 60 MHz, with data rates of up to 5 Mbps, power consumption between 0.74 mW and 1.58 mW, and communication eficiency between 148 pJ/bit and 316 pJ/bit, with power supply voltages of 1.5 V and 1.8 V, respectively.

This dissertation delved into periodically segmented waveguides, where segmented
guides refer to those with non-trivial geometries where propagation mechanisms are
not obvious. The study explored the behavior of their propagation modes and polarization
through numerical simulations using the 2D Finite Element Method. The
formulation for waveguide mode analysis was employed, enabling the investigation of
the impact of variations in the core geometry (radius, width, gap, and core period)
on their guiding properties for TE and TM modes. The dissertation also examined
the influence of materials (silica, silicon carbide, and silicon nitride) composing the
waveguide on the dispersion relation. This involved analyzing how material properties,
such as refractive index and absorption characteristics, affect the propagation
features of modes within the waveguide

Due to the increase in the amount of data available globally, Machine Learning
techniques are becoming increasingly popular, both in industry and academia. In the
field of materials science, simulation models have already led to several relevant
developments, including the design of new materials. A phenomenological simulation,
in general, requires the analytical description of the physical processes involved and
requires great computational power for its execution. to perform an extrapolation of
the roughness pattern of the Ballistic deposition model. From this premise, this work
has as main objective to use Machine Learning models to the problem of particle
deposition in films, to perform an extrapolation of the roughness pattern of the
Ballistic deposition model. Applying machine learning models and methodologies,
we intend to learn the normalized roughness pattern for substrate sizes that are
computationally viable to simulate, in order to carry out an extrapolation of these
to sizes that have not yet been simulated, and also to large sizes, that would require
enormous computing power.We used the Root of the Mean Square Error as a
validation parameter, in order to evaluate how successful the machine learning
was, in addition, the methodology as a whole. From the learning, we performed
roughness extrapolations to sizes close to “infinity” and from these, we extract the
roughness exponent α, which is a constant with computational value still uncertain.
In conclusion, we can highlight as preliminary results obtained, the roughness curves
for lengths that tend to the thermodynamic limit, something that has not yet been
observed in the literature, and from these, we were able to extract the exponent α
for such lengths, where this approaches the value expected by the theory.

This dissertation presents a methodology for predicting solar irradiance data using short-term
time series models (72-hour horizon), with hourly discretization, for a photovoltaic park located
in Jequié-BA. This tool will help to integrate photovoltaic technology into the electrical system,
as it will help the grid operator in the tasks of energy dispatch planning, contracting auxiliary
and reserve services, managing maintenance actions and grid congestion, and general management
of the park. The model proposed in this dissertation uses Box and Jenkins modeling, an
iterative procedure that fits Auto-Regressive Integrated Moving Averages (ARIMA) models to a
data set. The Auto-Regressive (AR) and Periodic Auto-Regressive (PAR) models were considered
as forecasting methods, whose parameters were adjusted based on hourly measurements
of irradiance, carried out over a period of one year. In order to compare the performance of the
AR and PAR models in predicting hourly values of solar irradiance, tests were carried out considering
variations in the order of these models. The forecasts of days with different luminosity
characteristics were also evaluated. Regarding the order of the models, tests were initially carried
out considering AR and PAR models of order 1 (AR(1) and PAR(1), respectively). In this
case, the adjustment of the model parameters takes into account that the prediction of irradiance
in a given hour depends only on the occurrence of irradiance in the previous hour. It should be
noted that, for the PAR (1) model, 24 parameters were adjusted, that is, an order 1 model for
each hour of the day. Tests were also performed using the AR model of order p and the PAR
model of order pm. In the case of the PAR(pm) model, the model order varies according to the
forecast time. The performance of the considered forecasting methods was evaluated through
error analysis.

There are different methodologies applied to the simulation of the inrush phenomenon in power transformers. These are composed of a series of interdependent procedures involving the technique used for modeling the saturation curve, the numerical method employed in the discretization of the system, and the solution technique used to solve the equations that describe it. Depending on the choice of these techniques and methods, the simulation results may present problems of accuracy, numerical stability, and/or computational efficiency. Therefore, given that the inrush phenomenon is responsible for introducing unusual current and voltage levels in the system, it is important to adopt appropriate methodologies for its simulation, in order to calculate these quantities properly. Thus, a comparative analysis is proposed between five known methodologies, in order to distinguish their performance by identifying their limitations. Methodologies 1, 2 and 3 refer to the curve modeling via piecewise linearization and discretization via trapezoidal, BDF 2 and parabolic approximation methods. On the other hand, methodologies 4 and 5 refer to modeling the saturation curve via operating points and via true non linear equations, with discretization via the trapezoidal method. For this purpose, a system containing a generator, switch, transmission line, and transformer was modeled, adopting the circuit parameters and saturation curve data referring to a known transformer. The comparative analysis was separated into two groups, the first one containing methodologies 1, 2 and 3, which differ only in relation to the numerical method used in the discretization of the system. The second group, containing the methodologies 1, 4 and 5, which differ in relation to the applied modeling technique of the saturation curve. The comparison between methodologies was based on the variation of circuit system parameters and saturation curve parameters. For the first group, it was possible to perceive that depending on the magnitude of the system parameters, the solutions using the BDF numerical method of order 2 are less susceptible to inaccuracies and instabilities, unlike the trapezoidal and parabolic approximation methods. For the second group, it was noticed that modeling techniques such as piecewise linearization and modelling through operating points are based on subjective parameters that can cause problems of inaccuracy and inefficiency, unlike modeling through true nonlinear equations, which has a more objective nature. Finally, among the evaluated methodologies, it is concluded that there is a primacy with respect to the most appropriate numerical method for discretization of the system and the saturation curve modeling technique, namely, the BDF method of order 2 and modeling via true nonlinear equations, respectively.

Three-phase induction motors are crucial in advancing industrial technology and are responsible for a large portion of electrical energy consumption in this sector. Understanding the performance of these electrical machines when operated by frequency converters is essential for students, professionals, and researchers who work with this type of motor. In this context, this work presents a study on the performance of three-phase induction motors, considering pre-established variations in the frequency and voltage of the supply signal. Through the tests carried out and the data obtained, it was possible to analyze the behavior of the engine's performance and parameters under these conditions of frequency and voltage variation. It was necessary to develop electronic circuits for the inverter, using the IRAMX16UP60A integrated circuit, as well as a conditioning circuit based on Hall effect transducers. The tests were carried out, and the corresponding calculations were applied to determine the parameters of the tested engine. The results were presented in tables describing each frequency and voltage situation, and graphs were drawn based on these experimental data. Finally, the results provide a better understanding of the behavior of the three-phase induction motor in these specific conditions, highlighting the importance of the appropriate selection of frequency and voltage to optimize the motor's performance and guarantee its energy efficiency.

In this dissertation, the photovoltaic systems of the Guanambi and Uruçuca campuses of IFBaiano were analyzed based on information from installation projects, technical visits, consultation with stakeholders, and mainly, by evaluating the data available on the Internet fed by Fronius frequency inverters. Initially, the attributes of impacts on energy generation and the losses that reduce the generation of electrical energy in a photovoltaic system were addressed. For each of these factors, proposals are created to quantify them in order to estimate how much energy would have been generated and compare with actual production. During each stage it is possible to identify how productive these two installations are and with the calculation of figures of merit it is possible to compare them objectively, evaluating how much better one project was proposed in relation to the other. In the chapter “Methodology for estimating the generation of a photovoltaic system”, the values obtained in the calculations are re-presented in tabular form. The method by which the impact factors on energy generation were found throughout the development of the dissertation was defined as the Detailed method and the method obtained by the joint analysis of the two Photovoltaic Systems was defined as Approximate. The description of this second method assumes that the real results of these two generations are not known and its objective is to make the estimation of energy generation from any photovoltaic system simple and objective. The estimates were compared with the energy generated in order to check whether the final results are reliable. The result of the Approximate Method was satisfactory in that it differed by less than 10% between the actual and estimated values. Finally, a case study is presented for the application of the methodology developed with the aim of estimating energy generation in a supposed photovoltaic system. The calculation steps were guided assertively, quickly, providing quality data for decision making in a short period of time.

Ocean science technology plays a key role in military and exploration of the marine environ-
ment, a fact that has been encouraging the development of new methods of processing and
analysis of underwater acoustic signals (UAS). In the specific case of the passive SONAR
(SOund NAvigation and Ranging) system, most techniques used in detection and classification
of vessels in military operations with submarines are based on the short-time Fourier transform
(STFT). However, this spectral analysis method has limitations regarding time-frequency (TF)
resolution, which directly affects the performance in estimating the characteristic parameters of a
vessel, as well as monitoring its dynamic behavior. This doctoral work proposes a spectral analysis
tool with noise reduction based on the technique called Hilbert-Huang transform (HHT). The
main objective of the proposed method is to be used in conjunction with the Detection of Envelope
Modulation on Noise (DEMON), traditionally used in the analysis of passive SONAR signals.
For that, a sequence of three studies was carried out. The aim of the first study was the choice
of the variant HHT to be used instead of STFT, which is used in traditional DEMON analysis.
In the following, the investigation and selection of a noise reduction technique to mitigate this
serious problem associated with the analysis of experimental passive SONAR signals was carried
out and, ultimately, the implementation of the proposed tool, based on the chosen HHT variant
and a intelligent classification system. The results of applying this new method to simulated and
experimental passive SONAR signals, indicate a better performance when compared to those
obtained with the tradicional DEMON analysis, by presenting greater precision and efficiency in
the analysis of static and dynamic parameters of vessels. When applied to experimental signals,
the proposed method proved to have a higher frequency resolution, with an average spectral
width of about 28 times smaller, in addition to presenting an average signal-to-noise ratio of
87,8 dB bigger.

One of the main problems of cellular mobile communication systems is the interference. The growing demand for connectivity can make this problem even more challenging, creating scenarios with high device density and massive connections. Thus, it is important to investigate and understand different interference models in wireless communication channels. The Epidemic Interference (EI) model was recently proposed to represent the situation in which many devices begin transmitting simultaneously, causing a rapid increase in interference levels on a shared channel. Based on this interference model, this thesis presents a study of the effect of EI on digital communication systems in terms of signal synchronization for coherent demodulation, outage probability, symbol error rate, and the theoretical capacity of the channel affected by EI. In order to evaluate the effect of EI on the received signal synchronization, the expressions for the theoretical limit of the variance of the carrier phase estimation error are derived. These expressions are evaluated by numerical methods for cases in which
the received signal is an unmodulated carrier and a digital phase modulated signal. The impact of EI is also analyzed by the outage probability and symbol (and bit) error probability expressions, which are derived for amplitude, phase and quadrature digital modulation schemes. Finally, the ergodic capacity of the channel in the presence of EI is evaluated. In addition to the numerical expression, approximate analytical expressions for the capacity are proposed. Due to the lognormal distribution of the interference power, the expressions presented in this thesis are assessed by the Gauss-Hermite quadrature. Computer simulations were performed to validate these expressions. For the analyzed cases, it is observed that the presence of EI increases the phase estimation error, the outage probability, and the symbol (and bit) error rates. Also, EI decreases the channel capacity. The partial results also show that the performance of the studied systems depends both on the mean power of EI and on the spread of the interfering power distribution.

This work presents an experimental study of two control techniques based on the Optimal Control theory, applied to an unmanned aerial vehicle (UAV). The first technique, based on the quadratic cost function, is the LQR controller. The second technique optimizes the norm Hinfinity taking into account the pole allocation and using an approach based on LMIs. The mathematical modeling of the vehicle is presented, along with experiments to carry out the identification and validation of the model's parameters. Furthermore, practical tests are carried out in order to investigate the performance of each controller, with and without the presence of external disturbances. The test platform used is the quadrotor AR.Drone 2.0 and the algorithms are implemented in the Robot Operating
System (ROS).

This thesis presents a novel compensation approach for receptance-based second-order systems with long input delay and unknown dead-zone input nonlinearity. The approach is based on a filtered Smith predictor for systems with long input delay based on the receptance model. An innovative discrete-time adaptive strategy approach for dealing with unknown input dead-zone is also based on a receptance model realization. The receptance-based second-order model is used in several fundamental applications, including active control vibration of mechanical vibrating systems. A filtered prediction error that considers the effects of time delay can be applied to pursue state feedback design for active vibration control purposes in the proposed approach. The dead-zone compensation is treated by a discrete-time state observer, which is based on displacement, velocity, and control effort signals to estimate the unknown parameters used in an adaptive algorithm. The main contribution of this work is to combine an unknown dead-zone adaptive mechanism and a receptance-based time delay compensation in a unified
design. Some numerical examples illustrate the effectiveness of the innovative
proposed approach.

The introduction of distributed micro and mini-generation systems in the electrical system has grown a lot in recent years. Despite the advantages of this growth, new challenges to assure the indicators of efficiency and quality of electric energy have stimulated the researchers. This happens because the massive insertion of distributed generation - when it is not well conducted - can cause voltage deviations, in addition to problems in the protection of the electrical system. Because of this, several studies have been conducted aiming to understand these impacts and the possible solutions to reduce them. In this context, photovoltaic generators are important as they present the greatest growth in installations of this kind. One of the objectives of research has been the proper control of “intelligent inverters”. Through some functions of these instruments, such as volt-var and volt-watt control, it is possible to mitigate voltage deviations that arise due to the elevated introduction of photovoltaic systems in the grid. Thus, this work aims to optimize the volt-var and volt-watt control curves of the inverters in order to minimize technical losses and voltage deviations. For that, the project foresees the simulation of mass introduction of photovoltaic systems through the software OpenDSS in a real grid located in the Brazilian Northeast, for analysis and optimization of curves through Genetic Algorithm. The results show that the optimized parameterization of the curves can be an effective method to reduce or eliminate voltage deviations while minimizing the negative effects of the control method used.

Robust systems are required for autonomous driving on non-uniform terrain commonly found
in open-pit mines and developing countries. To help narrow the gap in this kind of application,
this work proposes a perception system for autonomous vehicles and advanced driver assistance
specialized on unpaved roads and off-road environments capable of navigating through rough
terrain without a predefined trail. As part of this system, the Configurable Modular Segmentation
Network (CMSNet) framework is proposed facilitating the creation of different architectures
arrangements. Some CMSNet configurations were ported and trained to segment obstacles and
trafficable ground on a new collection of images from unpaved roads and off-road scenarios
containing adverse conditions such as night, rain, and dust. It was also performed: an investiga-
tion regarding the feasibility of applying deep learning to detect regions where the vehicle can
pass through when there is no clear track boundary; a study of how our proposed segmentation
algorithms behave in different severity levels of visibility impairment; and an evaluation of field
tests carried out with semantic segmentation architectures conditioned for real-time inference.
The new dataset (named Kamino) has almost 12,000 new images collected from an operated
vehicle with various sensors, including eight cameras capturing synchronized sequences from
different points of view. The Kamino dataset has a high number of labeled pixels compared to
similar publicly available collections. It includes images collected from an off-road proving
ground exclusively assembled for testing the system that emulates an open-pit mine scenario
under different adverse conditions of visibility. To achieve embedded real-time inference and
allows field tests, many layers of the CMSNet CNN networks were methodically removed and
fused using TensorRT, C++, and CUDA. Empirical experiments on two datasets validated the
effectiveness of the proposed system.

This work aims to investigate and optimize two-dimensional waveguides with 90° bend and crossing 2x2 compounds of silicon and silica. The following custom metaheuristics are analyzed: Memetic Algorithm, Particle Swarm, Gray Wolf and the hybrid (Greedy, Random and Adaptive Search Procedure together with Simulated Annealing) (a) modeling occurs by the Element Method Finite – 2D, in which the optimization objectives are to reduce power losses and diversify the possibilities of device configurations, in order to maximize power transmission efficiency. (b) The geometric parameters of the structures are optimized, considering particular characteristics of the restrictions of the structures and the algorithms used, so the results were studied, for the understand the interaction between the geometry of structures and their associated performances. Transmission efficiency values above 98% were achieved.

This work proposes rigorous methods for hyperparameter tuning of machine learning for classifying images in construction. For this, computational methods called HyperTuningSK and AutoHyperTuningSK are proposed. These algorithms use statistical techniques for recommending hyperparameters, such as Analysis of Variance and the Scott-Knott clustering algorithm. The approach uses statistical experimental design concepts, such as analysis of variance and the Scott-Knott clustering algorithm. In addition, four case studies were used: façade vegetation detection, gutter integrity detection, machinery classification, and crack classification. The results showed that the hyperparameters affect the performance of image classification. It is also worth noting that the hyperparameter configurations adjusted by HyperTuningSK resulted in different recommendations depending on the neural architecture used. In this sense, the adagrad025 combination achieved a HyperScore = AAA for the Densenet121 architecture. The results of the data augmentation analysis show that two transformations were the most recommended by HyperTuningSK: width shift and height shift. Moreover, the AutoHyperTuningSK algorithm recommended an adagrad optimizer and a learning rate of 0.02220 in the experiments for the fourth case study. This combination achieved a maximum accuracy of 99.48%, that is, the correct classification of 3,979 images (4,000 in total) in the test dataset. The results for tuning data augmentation hyperparameters also confirm the efficiency of the proposed approach using the AutoHyperTuningSK-DA algorithm. In this regard, the recommended combination achieved an average accuracy of 99.2% in the test experiments.

In this dissertation, the behavior of purely dielectric absorber metamaterials models based on water in relation to certain geometric aspects and different materials in their composition were analyzed numerically, in order to understand the physical aspects of absorption, for absorption levels above 90% in a broadband frequency in the microwave range. The absorbent metamaterials were analyzed in a large frequency band, between 4.99 GHz - 29.99 GHz, using the resins found in the current literature, they are: Resin [1], Resin [21], PMMA [1], PMMA, [4], Plexiglas [3], Photopolymer [9], UV-PR [5], VC810 [9], TPU [7], FR4 [23], FR4 [24], FR4-LPM [6], and UV-PR [5]. For them, Transversal Electric (TE) and Magnetic (TM) modes were analyzed at different angles of incidence (15º, 30º, 45º and 60º) of electromagnetic beams. Finally, it was observed for some of the proposed structures, absorption above 90% for large frequency broadband and ultra frequency broadband, being able to observe the behavior of water at different temperatures (27 ºC, 40 ºC, 60 ºC, 80 ºC and 100 ºC). Such structures are easy to build to be manufactured, have a low production cost and the possibility for technological application in optical communications.

Partial discharges are transient phenomena that usually occur involving an insulating portion involved in an electric field and immersed in a gaseous environment. The level of this activity has been correlated in recent years to the state of the insulating materials used in electrical systems. Particularly in the segment of electric machines, partial discharges can provide crucial information for monitoring the condition of the equipment, bringing data on the presence of defects in the insulation, its evolution and even location.

As these are signals that mostly occur in the MHz spectrum, the capture, treatment and interpretation of information are complex, requiring robust acquisition systems and computers to filter out noise and components of unwanted frequency of the signals, preserving the relevant features for analysis. In possession of this information, it is possible to use the normative literature to, by comparison, assess what behavior observed in the signal in question and seek to point out the approximate location of a possible defect.

This work discusses a comparative analysis of three partial discharge data filtering mechanisms and compares its final performance with a known diagnosis of a real test object. In order to achieve this goal, an online partial discharge signal collection was performed from a generator that had the coupling capacitors installed on the machine terminals. The raw data was then treated by software by removing unwanted frequency components and eliminating any presence of noise. This package of information is used to classify defects based on the shape of the graphs and their normative reference. Finally, in possession of each suggested diagnosis for each filtering modality, the information is compared with the actual situation of the machine that was opened for maintenance approximately 6 months after data collection.

The method developed in this work can assist in the study, analysis and treatment of the data of partial discharges, trying to provide other mechanisms of separation of the relevant information for the analyst who will compile the information and make the diagnosis of the state of the machine starting from reliable information.

Efficient path planning considering a sequence of goals in a manufacturing scenario is still
challenging for nonholonomic robots. In this paper, we present an approach to improve the
efficiency of path planning considering a sequence of goals for robots of type differential
drive and skid-steer. Besides using a near-optimal sampling-based algorithm applied in
a map with real inflating obstacles, virtual obstacles are added to the map to help the
generation of smoother paths that lead the robot from the start pose to its goals poses
faster, with fewer mechanical efforts, and less power consumption.

In mobile radio frequency telecommunication devices, the antenna is often subject to

varying surroundings. These change the antenna input impedance and lead to impedance

mismatch with the following blocks, especially the power amplifier. This can cause various

unwanted effects such as nonlinear behavior and possibly damage. Amongst other

possibilities, one solution is an automatic impedance matching system, which acquires

information about the actual antenna input impedance and controls a variable impedance

matching network accordingly to minimize mismatch. The complete information about the

antenna impedance can be gained using a reflectometer. These devices usually comprise various analog hardware components with their respective nonidealities and possibly

energy consumption. These hardware components also need physical space and thus

obstruct miniaturization.

An interesting way to reduce nonidealities and space consumption is by moving a

maximum of the functionality into software. The challenge, in this case, is the high frequency

of the signals, leading to extreme processing requirements. To counteract this, sampling

effects in the inevitable analog-to-digital converter can be exploited. Physical and

mathematical constraints for this operation are found and the design theory of such a

system is presented. Then the validity of the concept is evaluated by computer simulations

and automated measurements on a discrete electronic setup against known references.

Also, once the reflection coefficient information is available in software, its further use for

controlling the variable impedance matching network is of relevance. One possible

algorithm is outlined for future reference.

The validity of the concept and of the according to design theory was proven. Mostly, the

measurement errors were below 2%. In severe mismatch cases, the under-sampling scheme

stopped working, but these cases present a high standard deviation and are thus easily

recognizable. The reason is the non-linear behavior of the unprotected power amplifier.

Various promising remedies are outlined, such as protecting the amplifier, reducing the

transmitted power when necessary, and filtering in the signal path for the generation of the

sample clock. As the system is planned to be part of an automatic impedance matching

system, however, extreme mismatch cases should normally not occur.

The study shows that it is possible to build a vastly software-defined reflectometer for

ultra-high frequency operation using standard, inexpensive hardware components. This

means that it should also be uncomplicated to integrate into a frontend integrated circuit

for any telecommunication system.

The use of emerging technologies through the so-called industry 4.0 has provided greater productivity and efficiency. The intelligent production process involves people, machines and equipment, logistic systems and suppliers through a real-time communication and control system that connects them. The continuous monitoring of industrial activities stands out in the optimization of the life cycle, in work safety, assisting to contract industrial insurance, allowing for greater flexibility, precision, reliability, predictability and adaptability to the means of production. In some industrial sectors, equipment and devices are often subject to deterioration by corrosion, which compromises their mechanical and structural property. In extreme conditions, it can cause plant operation failures, human risks and high maintenance costs. One of the challenges in detecting corrosion occurs in equipment with coated metallic material or with thermal insulation. The Pulsed Eddy Current (PEC) technique is well suited for this kind of testing, as it enables inspection on thermally insulated metal parts without the need for coat removal or stopping system operation. However, the magnetic field interaction inside the evaluated specimen generates PEC signals that are not of simple interpretation. Given the above, a processing system was developed electronic, to generate, acquire and process PEC signals for properly identify corrosion in thermally insulated pipes. The system includes analog (current driver and data acquisition) and digital (signal processing, feature extraction and decision support) sub-systems. The proposed signal processing chain comprises feature extraction using discrete Fourier and Wavelet transforms, information compression by Principal Component Analysis and decision support through intelligent classification techniques, using Multi Layer Perceptron - MLP and Extreme neural networks Learning Machine - ELM. The results obtained from tests on a carbon steel tube used in the petrochemical industry indicate the high efficiency of the proposed method. By combining the neural network and data compression, it was possible to obtain an intelligent portable system for rapid assessment and decision support in the field. The best designed neural discriminator had a total efficiency of 99.7% and an average processing time of approximately 2.4 ms.

The growing interest in new metamaterials and hypercrystals is justified by the need for materials that meet the constant evolution of market demand, brings new questions and answers that reflect on the best ways to use the characteristics of good absorption and / or reflection. This work evaluates the behavior of hypercrystals, analyzing their transmission, reflection and absorption, which are important for applications in technologies such as solar cells. As simulated structures in the region visible to the infrared from 400 to 1600 nm, they are composed of layers of metals and dielectrics with variation in their parameters, geometries and configurations that allow to obtain better results. The results were moderate in the Transversal Electric (TE) and Magnetic Transversal (TM) modes, where the applied measures show satisfactory results of absorption and reflection along the entire wavelength, demonstrating new possibilities of hypocrisy structures and the possibility of using or structure in manufacturing and testing photonic devices.d Applications of Hypercrystals

This paper discusses control strategies for the connection of the three phase inverter to the grid through LCL filters. Three different methods are presented: conventional dq, multivariate and optimized PI. The first two proposals consider simplified modeling for the system plant and therefore neglect the resonance phenomenon caused by the use of high order filters. Closed-loop stability is guaranteed for this simplification through the use of extra cushioning strategies. The optimized method uses non-parametric models of the plant and the control is performed using fifth order controllers. The constants of these controllers are determined with the solution of a restricted optimization, which guarantees the decoupling and stability of the system. It was possible to verify that the conventional strategy guarantees the stability of the closed-loop plant, but the power decoupling is not achieved. This objective is achieved with the controls, multivariate and optimized PI, presenting stability and decoupling. Among these two techniques, only the optimized one was able to decouple the powers, even with variations of the plant parameters.

This work proposes a new approach for the Finite Horizon Robust Model-Based Predictive Control (FH-RMPC) problem of linear discrete systems with time-varying polytopic uncertainties. The basic idea of this approach is to find, in each sampling time, the optimal state feedback control law based on Riccati difference equations (RDE) that minimizes an objective function with finite horizon and takes in account input and output constraints. The closed loop stability, known as one of the challenging themes of FH-RMPC strategy with constraints, is based on monotonic property of the RDE associated with finite horizon Linear Quadratic (LQ) control problem. In this proposed method it is not necessary that the usual terminal weighting matrix be fixed to achieve asymptotic stability. The stability analysis is done considering a finite set of Lyapunov functions and also considering parameter dependent Lyapunov functions. Finally, the efficiency of the proposed approach is illustrated by numerical examples and comparing to the IH-RMPC strategy through error-based performance indexes.

In Elastic Optical Networks - EON, the bandwidth of an optical path is variable and the overlap of the virtual topology in a physical topology must be designed to optimize the use of the spectrum. EONs, under static traffic, are typically designed using Mixed Integer Linear Programming (MILP) in order to minimize spectrum usage.

In this dissertation, a formulation for protection in EONs is proposed, which, based on the concept of squeezing and fixing the bandwidth, including shared risk groups, returns with the minimization of the maximum number of slots used in any physical link in the network. The route for each demand in the physical topology is determined by equilibrium equations together with restrictions of the physical layer in the formulation, so that pre-calculated routes are not necessary and the modulation format of each established optical path can be chosen with transmission quality needed. The first results evaluate the effectiveness of the MILP formulation for a small network when the connections are under different requirements of the Service Level Agreement - SLA and are provided by an appropriate protection scheme and different modulation formats. Subsequently, simulations are carried out on slightly larger real networks, such as the Via Network and the National Education and Research Network (RNP).
Case studies are carried out to analyze the basic properties of the proposed formulation.

Four formulations of Mixed Integer Linear Programming (MILP) will be presented to solve the problem of traffic maximization and spectrum routing, modulation and allocation (Max-RMSA) in elastic optical networks (EONs). The first two models are formulations without using the Grooming technique with and without predefined optical paths (NPSP - Non Predefined Shortest Paths) and (PSP - Predefined Shortest Paths), in which NPSP provides the global optimal solution, as it contains all the possibilities of network paths between each of the pairs of nodes, whereas the PSP provides, in general, a near-optimal solution, as it only has a set of predefined paths, thus reducing complexity and its search space. The last two models are also for solving Max-RMSA, however considering the use of the Grooming technique. Several simulations were carried out for each MILP and the results were analyzed in networks of small instances. For moderately large networks such as the National Science Foundation Network (NSFNET) only PSP formulations will be used, due to the complexity of NPSP formulations in networks with a higher number of instances, and a good solution can be found with a reasonable simulation time.

The challenge of transporting large blocks of energy over long distances with low cost and high efficiency has attracted several studies for the application of transmission lines above 1000 kV. In this work, an analysis of the energization switching in UHV (Ultra-High-Voltage) lines is performed for lines with voltage levels between 1000 kV and 1200 kV, long lines above 900 km and considering the variation of the reactive compensation. This analysis seeks to find the most severe overvoltage levels, looking for methods to mitigate such disturbances in the energization of the lines. Besides the overvoltage analysis, the work also compared different configurations for the line compensation, defining, therefore, the one that presented the best configuration among them. The overvoltages resulting from the simulated operations were obtained with the PSCAD (Power System Computer Aided Design) software, commonly used for the simulation of electromagnetic transients in power systems. In the end, a comparison for the suppression of these overvoltages was performed involving three possible methods: surge arresters, pre-insertion resistor, and the controlled switching of the lines.

This work presents a Maximum Power Point Tracking (MPPT) method in photovoltaic systems combining an Articial Neural Network (ANN) with a classic tracking technique known as Incremental Conductance (InC). Despite being widely applied, the InC technique generally fails to track the Maximum Power Point (MPP) under partial shading conditions due to the existence of local maximum points on the P-V characteristic curve. For this reason, an ANN has been trained in order to provide an initial reference voltage to the system that ensures that the InC tracking starts in a region which will converge to the MPP. To help in simulations, a solar cell is modeled and, based on parameters from a commercial solar module, the model of a photovoltaic panel is built using MATLAB software. The main methods of MPPT are described, pointing out the positive and negative sides as well as the respective implementation algorithms. The average small signal model is applied to the Boost converter, which considers small variations around an operating point, to obtain a linear model. With the delevoped model, it is possible to calculate the transfer function which re ects the in uence of the switching duty cycle of the DC-DC converter on the bus voltage that a photovoltaic arrangement is connected and thus insert a compensator to improve the efciency of the system. The main concepts of an ANN are presented and neural network training is performed using the backpropagation algorithm, which is based on the error calculated in the output layer to adapt the synaptic weights of the neuron layers. Then, the Constant Voltage, Perturb and Observe and Incremental Conductance methods are simulated. The results are discussed and compared taking into account the efciency of the search under variation of radiation and temperature and also under the efect of partial shading. The results of the proposed technique are presented considering dierent cases of arrangements and partial shading in order to demonstrate the efciency of the technique.

The use of digital signal processing as an aid for diagnosing failures in manufac-
turing processes has been very promising, as it allows to increase the efficiency of such
processes and ensures product quality and the safety of installations. This work proposes
the use of Hilbert-Huang transform (HHT) as a tool for feature extraction of ultrasound
signals obtained in an experimental procedure on welded joints using the TOFD techni-
que. HHT is a time-frequency decomposition that uses base functions for transformation
that are estimated from the signals of interest. Such base functions are obtained from
decomposition algorithms such as EMD, EEMD and CEEMDAN. Once the experimental
signals were obtained and the HHT applied for feature extraction, a classifier based on
multilayer perceptron neural network was used and showed a satisfactory performance,
reaching efficiency product above 90% in all considered cases. When EEMD and CE-
EMED were used as decomposition algorithms, the efficiency product achieved 98.5%
and 97.4%, respectively, whereas the EMD algorithm was used, the value reached 95.3%.
When HHT was associated with PCA, the classifier was able to discriminate the different
defect classes with 91.0% for HHT-EMD, 93.0% for HHT-EEMD and 92.7% for HHT-
CEEMDAN. These results are comparable to those found in the literature in which use
other pre-processing techniques, whose highest values obtained for a similar dataset were
94.8% using the DFT and 97.5% for DFT with PCA.

The first theory created for the analysis of power in electric circuits under nonsinusoidal steady-state conditions

was proposed by Budeanu in 1927. It is based on an extrapolation of the classical method used for sinusoidal

steady-state conditions.

Since it does not explain adequately some aspects of the behavior of non-active

power flow, it has allowed the proposition of several other theories for almost the last

100 years.

Among the most recent propositions is the one by Castro-Nuñez, which uses, in

order to model the non-active power and the multivector aspect of the electric power, a

mathematical tool called Clifford Algebra or Geometric Algebra. However, this proposal

still does not reach its goal since its results diverge in the time-domain.

In this work, a new approach for the transformation between the time and

Clifford domains is presented, which is capable of reproducing the results from the

analysis of the instantaneous power in the time-domain. Also, it is proposed, for the first

time, a rotating operator, different for each frequency present in the circuit. Therefore,

the time-domain is perfectly reproduced in the Clifford-domain, a characteristic that is

not present in the proposal of other authors.

In order to validate the proposed method, four single-phase circuits in nonsinusoidal steady-state conditions

were solved in the time-domain and in the Clifford domain. These circuits contain fundamental frequency and

third harmonic voltage sources together with RLC loads. Also, a circuit containing a fundamental frequency

voltage source, a linear load, and a non-linear load consisting of a single-phase fullwave rectifier, was solved.

The autonomous action of a robot is directly related to the way it perceives the environment around it. One of the main tasks of an autonomous robot is to be able to understand the complexity of the environment in which it is inserted. This complexity is due to the shape of the objects that are present in the environment, such as tables, chairs, boxes, walls, among others. For this, sensors such as: LiDAR, RGB and RGB-D cameras are used to perceive the world outside the robot. The method that perceives the external environment is known as SLAM - Simultaneous Localization and Mapping. In SLAM, the environment is modeled using sensors and an environment map is created iteratively as the robot moves. This work addresses the mapping of three-dimensional environments on a two-dimensional map using LiDAR and RGB-D sensors. For this, the LiDAR SICK LMS111 sensor and the Microsoft Kinect RGB-D sensor coupled to the mobile terrestrial robot HUSKY A200 were used. Using GAZEBO, a 3D model from the Robotics Laboratory of the Federal University of Bahia, the 3D model of the HUSKY robot and the SICK and Kinect sensors, the SLAM algorithms were used: HECTOR SLAM and RTAB-Map to produce an occupation map two-dimensional over a fixed path. The methods were able to represent the objects contained in the environment, which for trajectory planning activity is essential to have a consistent and true to reality map.

This work presents a methodology for properly sizing the LSOT (Low Saturation Onset Transistor). The LSOT is a four-transistor network that emulates a MOS device with much lower saturation onset voltage by compensating the reverse saturation component of the main transistor drain current. The main transistor must have the same dimensions of the device to be replaced and, theoretically, the compensation device should be equal. However, the mobility degradation due to the transversal electrical field as well as other second order effects lead to current overcompensation in the structure, so that the DC output characteristic of the equivalent device presents an undesirable hump. A systematic procedure to dimension the compensation transistor is thus proposed, in order to smooth the LSOT current-voltage characteristic. A few guidelines are also suggested to dimension the other two transistors and slight modifications in the LSOT topology are tested, aiming to optimize circuit power and area. At last, we analyze the applicability of the LSOT as a succedaneum of one of the output transistors of a regulated cascode mirror. All circuits have been tested through a simulation tool.

This work presents new architectures of analog multipliers in CMOS technology, for the application as synaptic elements in cellular neural networks (CNN). The proposed networks adopt two voltage-mode inputs and current-mode output, and are based on the principle according which the MOSFET acts as a gate-source voltage squarer in strong inversion and saturation. Although such principle is employed by numerous CMOS multipliers with voltage-mode inputs, the here described circuits are innovative concerning the signal application methods. One of these methods avoids the use of reference voltage generators. The circuit performance is analyzed in several aspects through simulation in 130 nm CMOS technology using symmetric power-supply of +0.6 V. In all designs a special care has been taken to reduce the area with respect to prior works of the same research group.

This work aims to establish a technological basis for an ultrasound non-destructive evaluation decision support system, that may be portable to a microcontrolled platform. Extreme learning machines, a fast training class of artificial neural network, are used as pattern classifier. The purpose is to obtain a dedicated system capable of recording the input data, training and operating the classifier system, and finally producing an integrity status indication for supporting the operator decision. A case study on integrity evaluation of weld beads in steel plates was considered. The basics of ultrasonic non-destructive testing and artificial neural networks are introduced. Further aspects are considered in order to accomplish the referred objective, such as: a comparison between a fast-training neural network and traditional ones; a method to reduce the feature set dimensionality and the proposed approach for system realization in microcontrollers. The results indicate that the system realizes the training step of a classifier in the microcontroller to carry out the classification of weld defects in similar accuracy performance as reported works in the literature.

Introduction: Time-delay systems are presents in several industrial sectors and they has also beeing affected by 4.0 connectivity trends whose changes impose relevant control challenges in order to achieve robust performance due time varying deadtime effect included in control loops. Objective: propose simple and robust stability criteria to systems affected by time varying delay and controlled by predictors using a frequency approach based on small gain theorem. Methods: based on small gain theorem some criteria are proposed using predictors in order to compensate a constant portion of time varying deadtime and the others criteria are obtained directly on second order systems described by receptances. Unstructured uncertainty descriptions are used as a way to simplify robust analysis and the results are illustrated through simulations and an experimental case. Results: the criteria obtained provide limited values to maximum deadtime based on model information and uncertainties bounds. The robust stability can be recovered through the robustness filter during operation by adjusting the filter tuning. The criteria are simple to apply and can be verified using computational mathematical tools. Besides that, there is no needs to assume any information about time delay change rate. Conclusion: simulation and experimental results illustrate the usefulness of this approach as an alternative to traditional methods, which usually disregard the use of predictors and their impact on robustness margins and time varying deadtime.

This Thesis presents an improved version of the Advanced Compact MOSFET model (ACM model) for application in hand calculations, required in the design of CMOS analog integrated circuits. The here accomplished research gave rise to semiempirical sub-models to be introduced into the basic model of the MOSFET drain current. These sub-models take into account second order effects, thus increasing the accuracy of the ACM model in the design stage of preliminary device dimensioning, which precedes the simulation stage. Methodologies have also been developed for the extraction of modeling parameters and have been applied to simulated DC characteristics from a CMOS 130 nm technology. The chief aim of this remodeling approach is to minimize the time spent in analog circuit design, by allowing reducing the number of iterations between the stages of hand calculation and simulation. Systematic methodologies have been proposed to the design of some basic CMOS analog cells, using the improved and original versions of the ACM model. The application of these methodologies to several design examples made it possible to compare simulation results, using the dimensions obtained from design by hand, with the specifications and hence to verify the reliability of the improved ACM model.

This work is a contribution to the realization in CMOS technology of an analog circuit, belonging to the biomorphic class of cellular neural networks, for image processing. Aiming at the reproduction of more complex operations than usual, a compact cell architecture was employed in the implementation of a network featuring two layers and mutual coupling, originated from an extension of the traditional CNN. To obtain the parameters that configure the network to perform the desired functions, a training methodology was developed, taking inspiration from some existing techniques addressed to this category of neuronal network and including modifications either to improve its performance in general or to adapt the process to some relevant characteristics of the various types of image operations considered. For verification purposes, the circuit is applied in simulations involving bipolar functions and image filtering, exhibiting similar results to the ones generated from an ideal model.

This project aims the recognition and classification of hand gestures using
electromyographic signals (EMG) obtained through a Myo® armband, which has eight
medical electrodes. Each electrode provides signal information regarding muscle
contraction performed in the execution of the movement. These signals will be
processed to obtain a standard "signature" to enable artificial neural network (ANN)
training. These signatures will be extracted for each type of movement, and then a
suitable mapping method is applied to classify the EMG signals. A comparison
between the Resilent Backpropagation (Rprop) and Levenberg-Marquardt (LM)
algorithms is performed on all trained datasets. Five motions were chosen for
classification and data sets with up to seven characteristics were used. Among the 144
tests performed, the best result was 88.2% to Rprop and 88.4% to LM. However, in
general, the Rprop algorithm presented better performance in most of the tests when
compared to LM, the training time being the most significant difference between both.
This work is expected to expand research in the area of Assistive Technology (AT) to
achieve alternative forms of social inclusion and improve the autonomy of people with
disabilities or reduced mobility of the upper limbs.

The Partial Eigenvalue Assignment (PEVA) is a technique of linear control state feedback,
which aims to modify the transient response of a linear time-invariant system (LTI) by
changing the eigenvalues of an open loop system. The reduced number of variables used
by PEVA, in a ratio directly proportional to the number of undesired eigenvalues, turns
it a pole allocation methodology computationally feasible and efficient in control designs
for large-scale systems (LSS). However, due to the conventional allocation for reaching a
specific set of coordinate points on the complex plane, it is challenging to apply PEVA in
multi-objective control project, commonly required in several areas of engineering. To reduce
this difficulty, a new approach for PEVA using the regional allocation of eigenvalues
is desired and studied. A new theorem based on linear matrix inequalities (LMI) to calculate
feedback matrix is created to solve PEVA-regional problems, which can be computed
on MATLAB (Mathworks, Massachusetts, USA) script language programs. This theorem
is built using elements of the classical theory strict-PEVA, the theorem of D-stability,
and similarity transformation. After calculating feedback matrix, spectral analysis of the
open-loop and closed-loop system are made after building up: Complex-plane eigenvalue
localization charts; comparative tables of open-loop and closed-loop system’s eigenvalues;
and transient response with zero inputs. Applied this technique for different control
projects and executed analysis step, it can be observed that theorem can allocate the
eigenvalues to desired regions and avoid the phenomenon of spillover with an accuracy of
four decimal digits. The good results of regional-PEVA theory proposed and its success
in guaranteeing control objectives enable this masters dissertation offers an excellent base
for futures researches on linear multi-objective control projects.

The Doubly-Fed Induction Generator allows efficient energy generation where turbine

speed has a non-controlled behavior when using a bidirectional converter. The stator flux

oriented control is a vector control technique that controls the rotor speed and

reactive power through power converters in back-to-back configuration.

This text presents an analysis of a field oriented control of the rotor currents using a Model

Based Predictive Control. Its control law is given by the minimization of a cost function

that considers the control effort and the error between the predicted outputs and its future

references To obtain the prediction model, a linearized incremented model is given with

the support of a feed-forward decoupler. An anti-windup technique is also studied as an

costless alternative than the regular optimal algorithms that considers input restrictions.

With the set of strategies proposed in this document, it is possible to minimize nonlinearities

and coupling effects while controlling DFIG's current loop using a low cost

predictive algorithm

No electric power system, transmission lines are a fundamental role in guaranteeing or transporting energy between generators and consumers. Allied to its position, its extension and exposure to the same equipment, these equipments are subject to a short circuit, increasing the importance of their protection system.

In order to guarantee and protect the protection systems, simulations of faults and analyzes of records in COMTRADE are relevant, such as support ferramenta, allowing the repetition of oboects or simulation of new ones.

This work presents a strategy for the simulation of single-phase faults in a transmission line by means of ATP; a result of this simulation em um arquivo COMTRADE e posterior avaliação deste no diz respect à localização da falta. Foram simulated reais cases for comparison of the COMTRADE records originais. Além disto, foram simulated novos ao longo da LT in all phases of the system.

For a given estimate of location, it is used or a method of compensating for sequência zero (k0), which uses voltage and current data from two ends of the transmission line. Additionally, it is used or negative sequência method, which uses these dice of both ends you give linha. I would like to review the bibliography of both the methods and implementations of the Octave scripts.

The results found corroborate the methodology such as ferramenta de apoio na análise and localização de faltas em linhas de transmissão, além de evider a praticidade dos arquivos COMTRADE n this type of analysis that surrounds proteção de linhas de transmissão.

The reconfiguration of radial distribution electrical systems, consisting of the technique of modifying the network topology by opening and / or closing keys, and the allocation of voltage regulators, are important tools for planning and operation of distribution networks, with the objective of minimizing loss of active system power. This technique can be interpreted as a programming problem, due to its discrete and combinatorial nature. The combinatorial explosion of possible solutions to these problems occurs with the increase the number of bars and the number of workable switches, causing a high processing time of algorithms and computational effort, becoming It is unattractive to solve them by the classical optimization method, giving heuristic and metaheuristic techniques. In this paper, we present a methodology to solve problems of reconfiguration and optimal allocation of voltage regulators in radial, three-phase, balanced permanent regime, in order to minimize the value of the active losses, without violating the operational constraints, using the Ants Algorithm. For the problem of reconfiguration and allocation of voltage regulators, algorithms are presented based on Ant Colony Optimization or Ant Colony Optmization (ACO), in the Ant Colony System (ACS) variant in the method of calculating the objective function, and in the heuristic information used by the ants in the construction of a solution. For the TAP adjustment, the power sum method is adapted to obtain better performance and quality in solutions. We present tests carried out in systems of 16, 33, and 70 bars to evaluate the efficiency and robustness of the methodology proposal. The study obtained satisfactory results, presenting a significant of the active losses and improvement of the tension profiles of the tested systems, contributing to the application of the ant colony algorithm as a tool for the solution of these types of problems.

Advances in optical data transmission technology have allowed for great growth

Internet and bandwidth-intensive services. Simultaneously,

There was also a growing demand for faster connection speeds from users.

The latest paradigm for increasing the science of backbone communication networks

It is composed of the Elastic Optical Networks or Spectrum-Sliced Optical Networks (SLICE)

based on orthogonal frequency division multiplexing or Orthogonal Frequency-

Division Multiplexing (OFDM). OFDM-based networks are more aware than

predecessors, based on wavelength division multiplexing or Wavelength

Division Multiplexing (WDM), however routing is more complex because

must take into account modulation, continuity and contiguity of spectrum; originating

so the problem of routing, modulation and spectrum allocation or routing,

Modulation and Spectrum Assignment (RMSA). RMSA resolution is critical

for better use of optical network spectral features, as optimizing this problem

and able to make more bandwidth available for future demands. Whereas

backbone network traffic is constantly growing, and it is also possible to assume

an incremental traffic approach and resolve RMSA for multiple arrival phases

from traffic. This work proposes a new formulation and algorithms capable of solving

o RMSA for sequential periods of incremental traffic. A new formulation

Linear programming was proposed to optimize the maximum spectrum used in the

network, but due to the fact that RMSA is a problem whose time to reach

to a solution and given as a nondeterministic polynomial (NP), the resolution requires a

Long time processing when the problem involves a large number of instances.

This is the case with large networks. In view of this, two proposals were also proposed

heuristics to resolve incremental RMSA in a feasible time. Routing one

by shortest path and another by load balancing routing. Both

heuristics were simulated with stochastically generated incremental multiperiod traffic,

to study the long-term effects of using such heuristics in relation to

spectral resources. Artificial neural networks are also proposed to predict

traffic in order to test realistic traffic values generated by

neural To confirm the viability of the forecast, two proposed models are evaluated.

Both simulations with random traffic and simulations with traffic coming from

have shown similar patterns and provided results that are

to plan optical networks to meet future traffic and

while saving resources.

This paper presents the methodology adopted for the development of the radio frequency interface.

(RF) of a full-duplex transceiver capable of measurement of quantities

through capacitive and inductive sensors built into an oscillator. The supply of

transceiver energy is captured from electromagnetic waves transmitted from a central unit

remote, and batteries are not required for operation. To make this measurement system

feasible, it makes use of full duplex communication. This system consists of a duplexer that

allows separation between transmission and reception while sharing the same antenna

simultaneously; Colpitts oscillator whose supply voltage is obtained from a

rectifier. This oscillator functions as both a measurement system and a transmitter of

RF, whose frequency is modulated by means of a relative humidity sensor. The sizing

of these circuits, the component specifications and the definition of topologies, as well as all

The necessary theoretical basis are described in detail in this paper. The simulations

and measurements developed presented satisfactory results for the purpose of the project. With

In order to validate the proposed techniques, a prototype of this system was developed.

around 27 MHz, more precisely at 26 MHz for transmission and 28 MHz for reception

in the unit of measurement. The operating characteristics of the prototype as a function of distance were

evaluated in a semi-anechoic chamber and to function, the transceiver needs to pick up a

power of 8.5 dBm. Thus, based on the antennas used in the prototype, it was possible to feed it

remotely at a distance of 5 m.

In the planning of radial distribution electrical systems, one of the alternatives to reduce power losses in the lines is the implementation of Distributed Generation (GD). This consists of connecting low power generation plants to inject energy as close as possible to the loads or consumers of the system. According to the primary energy used by the generators, they are classified into 2 groups: renewable or non-renewable. To ensure the benefits of deploying distributed generation you need to do some analysis to choose the best installation locations and the best generator sizing. This analysis becomes a nonlinear combinatorial problem, dependent on the number of nodes in the system, the number of generators to be installed and their capacities. An additional problem is the presence of harmonics in the systems due to nonlinear loads or renewable distributed generators that are connected to the system via electronic power converters. This dissertation work uses a methodology for the optimal allocation and sizing of generators distributed in a radial distribution grid considering harmonics. This methodology is based on the Particle Swarm Optimization (PSO) Particle Cloud metaheuristic optimization method, which together with a harmonic power flow calculation algorithm, seeks an optimal solution aiming to improve efficiency and reliability. minimizing technical losses, voltage drops and total harmonic distortion of the system. The proposed PSO-based algorithm was compared with other optimal GD unit allocation techniques to evaluate its performance. Thus tests were performed in radial distribution systems of different sizes, with linear and nonlinear loads.

In this work, we present or study the effect of changes and temperaments, sudden changes in converter load and resistance to inductor resistance. It applies to solar panels alternating with power converters in order to propose a simplified strategy to mitigate nonlinearities based on anti-winding action. The study is done numerically with Matlab / Simulink using four anti-string strategies and with actual pressure and temperature data, monitored by the National Environmental Data Organization System / National Institute for Space Research (SONDA / INPE). PID driver tuning is based on the root location method with pre-captured specifications and the accommodation time without loss of generality. It is shown as anti-winding techniques, in a modified version, better performance or closed loop when pressure drops due to cloud passing as it attenuates or affects nonlinearities. Moreover, one of its advantages comes from the simplicity of implementation of the techniques used.

An Artificial Neural Network (ANN) is a parallel and distributed network made of nonlinear
processing units arranged in layers.This work consists of the project and hardware
implementation of the architecture of a parameterizable feedforward neural network. The
Verilog hardware descriptions are generated by a Python script. The parameters are the
number of input bits, the number of layers, the number of neurons per layer, the number
of bits for the integer and fraction parts in fixed-point representation and the number of
bits for the activation function look-up table (LUT) representation. This parameterization
allows a minimum hardware configuration to be met at simulation time for a given
problem, thus saving both development time and resources in hardware. As an application,
a hardware was generated to describe a network which was trained offline with the MNIST
database and which classifies images from a camera video streaming in real time. In this
way, this work is appropriate both to support research in related areas that need to embed
an artificial neural network, as well as to develop improvements in the architecture, given
the availability of the developed verification environment.

In thiswork, oneinvestigatesthe design ofresonantcontrollersappliedtogridconnectedLCL inverters. The analysisisfocusedontechniquesbasedonthediscretefrequencydomain. It isdemonstratedthatthe system stabilityisdependentontherelationbetweenthefilter'sresonancefrequencyand a criticalfrequencywhosevalueisproportionaltothesamplingfrequency. Thisrelation determines thenecessityoftheso-calledactivedampingtechniquetostabilizethe system. Thisworkalsoverifestheinfluenceofthedelayimposedbythediscreterealizationofthe PWM modulationandthe grid impedancevariationonthe system stability. Finally, oneevaluatestheinfluenceofdiferentcontrol feedbacks ontheinverter'sstability. The results show thelimitationsofthesetechniqueswithregardsthedeterminationofthecontrolgainsandwithregardstheircapability for injectioncurrentswithlowharmoniccontent.

This work proposes a frequency-adaptive method for calculating the associated phase

to the positive sequence component of the mains voltages. This is a procedure

which performs the separation of the positive and negative sequence components in a stationary ab plane.

The performance of the procedure is enhanced by a one-way Fourier filter.

cycle (One-Cycle Fourier - OCF). Positive sequence voltages are used as input

of a phase capture loop in a synchronous reference system

Phase Locked Loop Frame (SRF-PLL). Thus, the structure is composed by an estimator of

positive sequence acting on stationary coordinate axes and an SRF-PLL which in this

called SAPS-PLL (Stationary Axis Based Positive Sequence Estimator PLL).

SAPS-PLL is efficient against harmonic distortion and is capable of rejecting DC components of

voltages under analysis due to the frequency response characteristic of the OCF filter. Adaptability

frequency is obtained by cascading the two SAPS-PLL blocks. In the first

block, the fundamental frequency is calculated by the SRF-PLL. This frequency is provided.

to the second SAPS-PLL block which estimates the phase value of the positive sequence component

of the measured voltages. Frequency adaptability is mathematically demonstrated. The method

is initially tested by synthetically simulated voltages. In addition,

the technique through voltages produced by a programmable power supply capable of emulating

distorted and unbalanced tensions. And finally, the performance of this technique is analyzed taking

based on the effect that the synchronization method has on current control

injected by a simulated distributed generation system in a Simulink / Matlab © environment. The

results prove the effectiveness of the technique.

The development of plasmonic metamaterials surfaces has expanded the possibilities
of manipulating electromagnetic waves in an unprecedented way. A new field of study was
opened for the research of applications and it has been possible by the fundamental
revolution caused by the discovery of the properties of plasmonic metamaterials. This work
presents absorptive surfaces based on this technology, in addition to the proposing of new
devices developed with the aid of iterative searches based on genetic algorithms. Optical
broadband filters with absorption coefficient near the unit are proposed. A device with
asymmetric absorption was demonstrated, built from the superposition of layers of
dielectrics and metals with an absorption asymmetry ratio above 28 and a contrast ratio close
to the unit. In this work, devices capable of modifying their spectral behavior from
controlling the state of the molecular arrangement of structures based on phase change
materials are proposed. These devices were designed, simulated, manufactured and
characterized experimentally and the simulated and experimental results showed a good
agreement. The automated search method was also used in the design of a coupling grating
for an integrated waveguide of silicon nitride, with coupling efficiency of up to 90%. We also
proposed a demultiplexing coupling grating capable of directing a beam from a fiber to two
different waveguides with an efficiency of 25% at λ = 1550nm and 27% efficiency at λ =
2000nm.

This work address the topics of dynamical modeling, parameters identification and simulation of underwater vehicles. Simulation plays an important role in the development of underwater vehicles. Nevertheless, currently used robot simulators, such as Gazebo, do not allow an accurate representation of the underwater vehicle’s dynamics, since the physic engine employed by the simulator is not able to represent the hydrodynamical term of the added mass. This lack of representativeness does not allow a realist test scenario of the underwater environment. This work has as objective to propose a mathematical approach to overcome the inability of the simulator in representing the added mass without changes to the physic engine. To simulate the underwater vehicle, it is necessary to know the parameters of the dynamical model. Therefore, this work has also for objective to identify the parameters of the dynamical model of the autonomous underwater vehicle FlatFish. Two parameters identification techniques were applied: least square method and adaptive identifier. In sequence, experiments were conducted, in basin and in the sea, with the purpose of identifying the parameters of the simplified dynamical model of FlatFish, using data collect experimentally from vehicle’s onboard sensors. Comparative analysis of velocities among vehicle’s measured data and the simulated model actuated upon the same control effort show that the parameters identified by both methods are able to represent the dynamics of the vehicle when actuated in one degree of freedom. However, it was verified that the parameters of the simplified model are not able to represent the vehicle’s dynamics when actuated in several degree of freedom simultaneously. It was established the need of performing parameters identification for the complete dynamical model so that the Gazebo simulation altogether with the proposed technique to represent the added mass, be able to emulate the behavior of the real vehicle. Despite that, the identified parameters can be used in the project of model based controllers and in state estimator filters.

This work presents the analysis, as well as, describes the design of radiometer and termometer based on thermoresistive sensors in feedback configuration, whose drive signal is proportional to square sensor current. At first, mathematical analysis of the sensor behavior is performed, followed by a dynamic analysis of the system to analyze the noise generated behavior during the quantization process of the analog-digital converter. This analysis is ratified through MATLAB simulation results, where a system model capable of analyzing the performance of the system was constructed for both temperature measurement and thermal radiation measurements. It is shown in this analysis that the sensor when operating in constant temperature, is able to infer through the control signal of PI controller the desired measurement variables. The impact of noise from the quantization error can be attenuated by filtering the noise shifted to higher frequencies. The measurement system is implemented by microcontroller and an analog circuit to drive the sensor and conditioning the acquired signal. The results obtained with the prototype validate the observations provided in the model.

Nowadays, our way of life is increasingly more dependent on electric power. This characteristic
makes the transmission lines, responsible for making the connection between the
generating units and consumers, to be on operating conditions 24 hours a day, 7 days a
week. Currently, manned aircrafts are responsible for the inspection of these transmission
lines, which is very expensive and extremely dangerous. Given this perspective, robots have
become more and more real alternatives to reduce costs and increase safety in inspection
activity. In order to seek autonomy in the inspection activity, it is necessary that the robot
identify the obstacles in its course, so that it can continue inspecting the line without
interruptions. Considering this context, this paper presents an approach to characterize
and classify the obstacles found in the line. The applied strategy includes the usage of a
LiDAR for data collection, shape features extraction and classification using the K-Nearest
Neighbor artificial intelligence algorithm. The approach of this work used two datasets to
demonstrate the performance of the proposed method, presenting 100% accuracy when the
images had no occlusion and 98.4% with the dataset showing occlusion in the images. Classifier
analysis was performed with addition of imagen noise types salt&pepper and speckle
with little change in performance with a minimum performance of 97.6% of accuracy.

ROSAS, L. N. R. Safety in Electrical Installations of Assistance Establishments Health. Case Study: Baiano Scenario. 2018. Dissertation - Federal University of Bahia, Salvador Bahia, 2018. The development of medical technologies and procedures has in health care. However, it also generated a strong dependence on health in relation to the proper functioning of electromedical equipment and, consequently, also of a safe, reliable electrical installation with high quality electricity. This one The paper then proposes to present an overview of the main characteristics of the system in health care establishments and to evaluate the scenario of electrical installations of health care establishments in the state of Bahia, through implementation of an evaluation protocol based on existing standards, three health care establishments. It was verified through the evaluations carried out, several points of non-conformity of the electrical installations of the establishments visited, indicating serious problems related to the reliability of electric power supply and the electrical safety of the establishments. There is, therefore, an indication of the need to state's attention to the electrical installation of EAS in order to ensure the insurance of such establishments.