The insertion and removal of autonomous underwater vehicles (AUV) are operations of high cost, duration and risk for operators. A possible solution to reduce these three factors is the replacement of these operators by unmanned surface vehicles (USV). A fundamental step for autonomously removing the AUV is docking, which is the moment when the AUV enters the mobile docking station (DS), which in turn is being towed by the USV. The AUV control algorithm must ensure that the DS is in the field of view (FOV) of the AUV sensors, as well as that the AUV is in the FOV of the DS sensors. Furthermore, the controller must control the AUV precisely to avoid unwanted collision between the AUV and the DS.

A controller widely used in the literature to solve these problems is the model-based predictive controller (MPC). However, most mooring MPCs do not consider DS movement prediction in their optimization problem. Furthermore, several docking MPCs consider that the objective of the MPC is to bring the AUV into the same pose as the DS. However, at the end of the docking maneuver, the AUV must be in a pose and with a speed relative to DS, not the same. Thus, for the docking of an AUV, this relative pose and velocity must be considered in the controller's reference calculation. Therefore, the purpose of this work is to take into account this relative docking pose and the prediction of its movement in the MPC.

Finally, this work also proposes to maintain contributions already made by other works. Such contributions are MPC implementations of thruster limits, ensuring the AUV is at the DS FOV, and ensuring the DS is at the AUV FOV.

To validate the proposal of this work, the solution was tested in a simulation environment in Gazebo, in which the underwater dynamics were computed by the UUV Simulator [1]. The robot operating system (ROS) framework [2] and the Casadi optimization library [3] were
used for the development of the controller and for its connection with the simulator. For the test cases, the parameters and the 3D model of the BlueROV2 Heavy commercial submarine vehicle were used to represent the DS and the AUV in the simulation.

Two test cases were elaborated to demonstrate the efficiency of the prediction of the movement of the docking pose in the MPC, one with a vertical oscillation of the DS and the other with a rotational and horizontal movement. The FOV constraints and the docking pose prediction were successfully tested simultaneously for the test case with vertical oscillation, but because it was too computationally expensive, the AUV FOV constraint was removed for the other test. For both test cases it was concluded that the
docking is reduced and that the error between the AUV pose and its docking pose is smaller when DS motion prediction is performed.

Vehicular Ad hoc Networks (VANETs) constitute a specific type of mobile ad hoc network in which nodes are vehicles capable of sending and receiving messages directly between each other using wireless communication links, enabling the creation of several distributed applications without the dependence on a specific infrastructure for vehicle communication. In urban and highway environments, VANETs have been seen as an opportunity in the development of Intelligent Transportation Systems (ITS), aiming not only to improve traffic flow, security, and well-being but also to reduce pollutant emissions. However, many distributed applications for ITS require groups of vehicles to coordinate their actions consistently, reliably, and with temporal constraints. Nevertheless, in VANETs formed in urban and highway environments, vehicles travel along pathways exhibiting a mobility pattern and highly dynamic communication structure. Therefore, network topology changes, communication link disconnections, and message losses are frequent in a VANET. Hence, meeting the requirements for vehicle coordination in distributed ITS applications based on VANETs represents a challenge. In this context, this work proposes the Vehicular Causal Block Protocol (VCBP) to support the vehicle coordination requirements in distributed ITS applications based on VANETs.The proposed group communication protocol offers the facility of message multicast with guarantees of reliable delivery, temporal constraints, and causal and total ordering of messages. VCBP has been implemented and evaluated in a vehicular network simulation environment, and its
performance has been compared with other protocols available in the literature. The results show that VCBP can provide enhanced message delivery guarantees with a message delivery rate similar to or better than other evaluated approaches.

In the present study, one addresses a daily dynamic optimization problem in oil production through a data-driven approach. To this end, the use of an artificial neural network (ANN) architecture is proposed as a substitute for the phenomenological model that represents the gas-assisted oil well production system. Therefore, it is crucial to include information on the oil flows from each well and also from the top of the riser. However, flow rate measurements are not available in real-time, and what is available is only the total flow rate after separation, which does not allow understanding the behavior of each well individually. Due to the unavailability of individual well flow measurement, the use of a mobile horizon estimator (MHE), supported by a phenomenological model, has proven to be an appropriate solution for estimating these variables, allowing for the provision of data for training and obtaining a substitute model to be used in the dynamic optimization stage. Thus, it was possible to enable the training of the Nonlinear Autoregressive with Exogenous Input (NARX) neural network architecture employed in this research. This choice was based on the finding that the network was able to conveniently predict one-step-ahead. The results obtained from the application of the proposed approach on a single well and on a field consisting of three wells and a riser demonstrated a good performance of the artificial neural network in terms of temporal prediction, in addition to a more efficient computational time in solving the optimization problem when compared to the standard phenomenological model. The solution proposed by this approach opens up several possibilities for implementation in large-scale problems, such as optimizing the daily production of an oilfield composed of multiple wells integrated by different reservoirs and manifolds.

Measuring a company's efficiency is fundamental for decision-making, influenced by the performance of its assets, including technology, industrial capacity, quantity of products and employee qualifications. The efficiency of production systems often depends on large volume production with little variety of products, which is directly linked to the efficiency of processes or bottlenecks. Applying demand forecasting models based on time series is an effective tool to obtain this information. However, to date, no studies have been found that applied these models in a hot rolling process, which raises the opportunity for investigation. The main objective of the dissertation is to develop a prediction model for performance indicators in a non-flat hot rolling process in a steel industry based on time series. The case study demonstrated how the global efficiency index factors impact the rolling process. The results indicated the ARIMA (2,0,2) model as the most appropriate, and its predictions revealed daily values of the OEE global efficiency index between 0.404 and 0.993. The results showed that the L2 Lamination process can work to achieve a challenging working range (0.699 < OEE ≤ 0.891), based on benchmarks from technical literature. The tool developed can be valuable for defining strategies and directing decision-making based on the insights provided by this forecasting model. The research demonstrated applying time series models in the steel industry contributes to management and efficiency improvement strategies.

Video object segmentation (VOS) is a complex computer vision task that involves identifying and separating the pixels in a video sequence based on regions, which can be either the background or foreground of the scene or even specific objects within it. The task must be accomplished consistently throughout the sequence, ensuring that the same object or region receives the same label in all frames. Recent advances in deep learning techniques and high-definition datasets have led to significant progress in the VOS area. Modern methods can handle complex video scenarios, including multiple objects moving over dynamic backgrounds. However, these methods rely heavily on manually annotated datasets, which can be expensive and time-consuming to create. Alternatively, self-supervised methods have been proposed to eliminate the need for manual annotations during training. These methods utilize intrinsic properties of videos, such as the temporal coherence between frames, to generate a supervisory signal for training without human intervention. The downside is that self-supervised methods often demand extensive training data to effectively learn the VOS task without supervision. In this work, we propose Superfeatures in a Highly Compressed Latent Space (SHLS), a novel self-supervised VOS method that dispenses manual annotations while substantially reducing the demand for training data. Using a metric learning approach, SHLS combines superpixels and deep learning features, enabling us to learn the VOS task from a small dataset of unlabeled still images. Our solution is built upon Iterative over-Segmentation via Edge Clustering (ISEC), our efficient superpixel method that provides the same level of segmentation accuracy as top-performing superpixel algorithms while generating significantly fewer superpixels. This is especially useful for processing videos, where the number of pixels increases over time. Our proposed SHLS embeds convolutional features from the frame pixels into the corresponding superpixel areas, producing ultra-compact image representations called superfeatures. The superfeatures comprise a latent space where object information is efficiently stored, retrieved, and classified throughout the frame sequence. We conduct a series of experiments on the most popular VOS datasets and observe interesting results. Compared to state-of-the-art self-supervised methods, SHLS achieves the best performance on the single-object segmentation test of the DAVIS-2016 dataset and ranks in the top five on the DAVIS-2017 multi-object test. Remarkably, our method was trained with only 10,000 still images, outstanding from the other self-supervised methods, which require much larger video-based datasets. Overall, our proposed method represents a significant advancement in self-supervised VOS, offering an efficient and effective alternative to manual annotations and significantly reducing the demand for training data.

Products and byproducts derived from oil, such as diesel, can incorporate water during the production, transportation, and storage process, as well as the biodiesel present in diesel. The presence of water in these products has always been a problem for the industry, causing serious damage to the equipment where they will be used and contributing to corrosion in the tanks responsible for their storage. The water content in fuels can be classified into three categories: soluble water, emulsified water, and free water. Soluble water occurs when the portion of water is able to be dissolved in the fuel, depending on thermodynamic parameters, such as temperature, and the chemical composition of the same. Emulsified water is a mixture of water and diesel in which the water droplets are dispersed in a diesel matrix. The water droplets are so small that they do not sediment, but remain suspended in the diesel. Emulsified water can be formed at any process in the diesel production line. Free water occurs due to the difference in density between the fuel and water, producing a mixture of distinct and perfectly distinguished phases. This work deals with the development of a real-time monitoring and analysis methodology that will compose an engineering solution capable of identifying the presence of free water in storage tanks, making the proper disposal in an intelligent and autonomous way. For this purpose, a test bench was built that simulates the storage process of diesel/biodiesel, composed of: (i) storage tank, (ii) disposal tank, (iii) cleaning tank, (iv) instrumented optical unit operating in the near infrared and ultraviolet range, in addition to acquiring images, (v) electromechanical devices for controlling and actuating the operating modes of the developed plant, and (vi) expert software for controlling, acquiring, and analyzing the acquired data. The optical unit is capable of analyzing and classifying in real time the fluid present in storage tanks by near-infrared spectroscopy and UV spectrophotometry, as well as the use of computer vision techniques for fluid analysis through image recognition.

Over the last decade, the use of solar energy to generate electricity has shown
increasing numbers.In general,solar-electric energy conversion systems are carried out
by flat photovoltaic modules, previously installed, provided based on the local latitude
of installation and in fixed structures. In this sense, the use of mobile structures or
structures with different topology has enabled the increase in the production and study
of solar energy.
This dissertation revisited the literature on biomimetics, the sun as a source of
energy and the different technologies used in the manufacture of solar cells, as well as
their functioning, which contribute to a broader understanding of the constitution of this
device.
In order to contribute to research in photovoltaic energy and the study of solar
irradiance, this work proposes a fly eye topology solar detector with the potential to
measure local insolation. The equipment allows you to observe the dynamics of solar
radiation during the different seasons of the year.
Confrontations were established between the results obtained in this dissertation
referring to the seasons of the year,in this way,to verify the average level of irradiance in
each season and the relationship with the inclination angle of the solar cell in the device
that were used as a radiation sensor.

Parking for the public in urban environments, such as large office complexes or shopping centers, becomes increasingly larger and more complex to meet the needs of its users. Navigating and locating these environments are a recurring problem faced by drivers, either to locate a vacancy or to find the parked vehicle. Efforts to improve this uncomfortable situation have been made, but the biggest problem for guiding drivers lies in covered and closed parking lots, where geolocator signs do not arrive. Currently, a significant number of systems for guidance in parking lots in covered environments performs the location and signaling to park in available spaces, generally based on systems to indicate regions with available spaces and light signals. However, integration between systems that allow to assist more effectively the location of spaces and, in addition, as they link to vehicles once parked in order to help the driver locate his vehicle when coming back. The present work uses an IoT-based approach to build an effective system for locating and identifying, in real time, covered parking spaces without relying on external signage and associating space with the specific vehicle in order to facilitate its location upon the driver's return. The developed system makes use of fusion mechanisms of properly indexed sensors, a communication system suitable for the specific environment, in this case, covered parking lots, including computer systems able to treat and interpret data and interface with users that facilitate the management and tracking of vacancies in a covered environment.

In the oil and gas industry, in offshore and inshore applications, it is of fundamental importance to combat the phenomena and/or effects that affect the oil exploration and production process. High power ultrasound can be of great value to researching solutions and developing the industrial and technological sector as it is an excellent solution to combat carbonation in pipelines, accelerating the precipitation of particles during the flow of oil. This thesis work proposes a high-power ultrasonic reactor as an active mechanism in the precipitation of calcium carbonate. Ultrasound delays the nucleation of crystals, making it difficult for particles to join during flow. Therefore, there will be a lot of smaller particles closer to the ultrasonic reactor compared to other regions along the pipeline. In this work, he shows the laboratory tests for the development and calibration of the ultrasonic reactor, the distribution of the ultrasonic field and how regions of higher ultrasonic pressure from a simulation software. It is also presented a proposal for an experiment carried out in Rio de Janeiro, at NIDF - Interdisciplinary Nucleus of Fluid Dynamics, in the multipurpose line of studies of calcium carbonate. The tests showed relevant results in reducing the size of the carbonate crystals and the pressure in the completion valve remained practically stable up to 0.107 bar up to 120 minutes.

Digital twins are commonly referred as virtual copies of a physical asset. However, there is no common definition for this term that is widely adopted. Along with other terms in the context of Industry 4.0, digital twins are increasingly being discussed. It is noteworthy that the creation and development of a digital twin is dependent on the physical counterpart being mirrored. Products, services, industrial processes and physical phenomena are some of the instances that can be digitized for various purposes such as optimization, lifecycle management and performance analysis. The need to develop generalizable and scalable solutions for digital twins is increasing. Therefore, this work proposes a modular, scalable and generalizable framework for digital twin creation based on the concept of multi-agent systems to support processes with various specificities of different industries in a configurable constructive structure. Five classes of containers were used in this work: sensor, actuator, conveyor, operator and storage modules. These modules were designed to cover all kinds of entities present in an industrial process. The digital twin of an essential oil steam distillation plant was created to demonstrate the proposed methodology as an instance of validation of the proposed creation method. Distillation is carried out in batches with discrete batches of plant material. As it is a batch process, it presents both continuous and discrete dynamics. So, this process is a good demonstration of a generic digital twin. The proposed digital twin concept can be the basis for the creation of several others due to its modular, scalable and generalizable characteristics.

With the insertion of the Internet of Things concept in the industrial context, challenges related to distributed 

automation systems have arisen, including with regard to communication, which is the main core of the performance 

of these systems. Possible solution paths for communication can mitigate the obstacles encountered. In computational 

terms, today's distributed automation systems resemble hybrid systems such as the partitioned synchronous 

distributed systems model, SPA. Membership services manage communication groups between processes belonging 

to a system and, together with a multicast service, form a group communication service that is a building block of 

fault-tolerant distributed systems. In this sense, this dissertation introduces a membership algorithm for the SPA 

model capable of managing multiple groups of processes and which has in its structure defect detection services 

based on a perfect defect detector P; and consensus available for the SPA. The formally proven properties of the 

algorithm were presented, as well as the elements of its simulation and evaluation, developed in a framework for 

simulation and evaluation of distributed systems, HDDSS.

Sepsis is a severe disease that affects millions of people around the world, and its early detection is fundamental to improve the treatment effectiveness. Recently, several models have been proposed to classify sepsis-positive patients in advance or to identify the probability of the disease occurrence in the future. In both cases, the data input is usually composed of time series of vital signs or other clinical variables. The current research shows an innovative approach for early detection of sepsis by representing a patient as a moving particle in an N-dimensional space, where N is the number of the adopted vital signs. A Sepsis Point is established, which corresponds to the position occupied by a patient if he became positive for the disease. The position, velocity, and acceleration vectors of the patients relative to the Sepsis Point are calculated. These vectors are used to generate the Kinematic Variables, which are imputed in artificial neural network models for early detection of sepsis. The accuracies achieved by the Kinematic Approach were compared to the accuracies achieved by the same models using traditional vital signs as input. It was discovered that the Kinematic Approach resulted in greater accuracy models, proving this research’s hypothesis. Thus, the Kinematic Approach is expected to open new approaches for developing more accurate early detection sepsis models.

The Automotive Liquid Reducing Agent, known as ARLA 32 (aqueous solution containing 32.5% urea) is injected into the exhaust gas flow of Diesel vehicles that have Selective Reduction Catalyst (SCR) to reduce emissions of nitrogen oxides ( NOx). As a guarantee of effectiveness in reducing NOx and that the SCR is not deteriorated, it is essential to control the chemical composition of the ARLA. Adulteration of this compound by dilution in water has often been found, due to the high availability and low cost of the diluent element. This work proposes a new optical detector for in situ analysis of ARLA 32 adulterated with water in fuel stations, through spectroscopy in the near infrared. The preliminary tests carried out used a commercial LED in the near infrared (1050 nm, 1300 nm and 1450 nm) and a silicon photodetector, with a non-linear behavior of the transmitted radiation in relation to the water concentration. Thus, a dedicated spectrometer prototype was developed, using a wavelength of 1650 nm, in addition to the implementation of the InGaAs photodetector and a pulsating feeder circuit through a quartz crystal. A precision voltmeter with 0.1 mV resolution (display) was incorporated to read the measurements. Offset and sensitivity adjustment buttons were also implemented to calibrate the optical instrumentation. ARLA samples adulterated with specific percentages of water concentration were tested, showing the direct relationship of adulteration with the concentration of water in the sample and its molar absorptivity. Additional tests were carried out aiming at validating the equipment in the detection of adulterations. Thus, bulk ARLA samples collected from 5 different filling stations in the Metropolitan Region of Salvador were investigated for water adulteration. Urea values above 32.5% were found in all analyzed samples, which can be justified by the evaporation of demineralized water during storage. The results confirmed that the portable prototype developed for field measurement of ARLA samples at fuel stations, presents a quick analysis to identify and quantify the ARLA adulteration with water, and an affordable cost to provide diesel vehicle owners and inspection agencies with an innovative alternative for quality control of ARLA 32.

Thermography is a technique for graphically recording the temperatures of a body, with the intention of distinguishing areas with different temperatures. Bodies with temperatures above -273C (absolute zero on the Kelvin scale) are capable of emitting infrared radiation. This characteristic allows to study the behavior of the temperature in different objects, structures and surfaces over time. Applications involving thermography cover the areas of security and military applications, being used in border surveillance, search and rescue, maritime patrols and coastal surveillance, wildlife; through studies related to understanding the thermal physiology of animals; care related to health and veterinary medicine; involving the diagnosis of various diseases, work-related injuries, studies of behaviors and diagnoses in animals; and the engineering area, being applied in inspections of electrical and mechanical equipment, inspections of buildings, conformities in air conditioning systems, management and maintenance of installations; this is not an exhaustive list. The biggest limitation for the development of studies and applications involving thermography is related to the high cost of your equipment. Pocket thermal cameras like Fluke PTi120 and FLIR C2 cost an average of R $ 6,000.00 and R $ 3,000.00 respectively. Thermal cameras with additional technologies and features like the FLIR T1020 HD and Fluke TiX 580 cost on average US $ 41,500.00 and US $ 14,000.00, respectively. The price variation is related to the application that the equipment will have, its accuracy, camera resolution, sensor resolution, temperature measurement limits and embedded technologies produced by each manufacturer. In this sense, the proposal of this project is to present the development of a low cost mechatronic thermographic device (DMT) responsible for the production of thermal images of objects, covering the stages of construction of the physical structure, data acquisition, movement system, electronics and control, electrical system and graphic interface for device control and image formation. The DMT has an accuracy of ± 1C, at room temperature between 0C and 50C, having a useful reading area of 20x22 cm, producing images with 320x360 pixels, and capable of reading objects with temperatures between 0C and 300C.

The Laboratory of Certification of Components of Systems of Photovoltaic Solar Energy (LABSOLAR) located in the Technological Park of Bahia aims to certify photovoltaic modules. Therefore, one of the equipment that it needs in its facilities is a steady-state solar simulator of large area used for continuous exposure tests in controlled environments. The Laboratory of Optical Properties (LAPO) of the Federal University of Bahia (UFBA) perceived a great opportunity to develop this equipment considering its high cost in the international market. Thus, a low-cost equipment was developed applying mechatronics techniques using commercial spotlights installed in a structural aluminum frame adapted in an automotive elevator. Metallic MSR lamps were used as a radiation source. This unprecedented lamp in this type of equipment allows to reproduce the AM1.5G reference spectrum and spectra characteristic from clear and cloudy sky present on deserts or temperate climates. The final result of this work was a Continuous Solar Simulator with BCA classification in relation to spectral correspondence, non-uniformity and temporal instability respectively, in accordance with the IEC 60904-9 standard. Another important development of this work was the creation of a single non-uniformity meter. It is capable to make radiation maps very fast enabling the calibration process faster and more dynamic from simple parameters configuration. Finally, it is noteworthy that the steady-state solar simulator still has potential for improvement to reach international class, and further increase and deepen the knowledge in this area.

Industrial environments are typically characterized by high levels of interference. Therefore, the standards for industrial wireless sensor networks (WirelessHART, ISA 100.11a and IEEE 802.15.4e) have defined a mode of operation based on time division and multichannel, in which time intervals and channel pairs are assigned to links representing a communication between two nodes. In IEEE 802.15.4e, this mode of operation is called Time Slotted Channel Hopping (TSCH). In this work, a distributed algorithm to define such an assignment for a given network is described. The algorithm is efficient, scalable and was developed for the Signal-to-Interference-plus-Noise Ratio (SINR) model, currently considered the most suitable for analyzing algorithms for wireless networks when interference is carried out. in consideration. In particular, the algorithm provides deterministic communication on the network. Previous approaches to this problem are mainly centralized, based on a simple (or none) interference model or do not consider multiple physical channels. In this work, we describe the algorithm and present results of the simulation, where we evaluate the number of rounds necessary to calculate the schedules and the size of the schedules produced. The described algorithm applies to the Internet of Things, characterized by the high scale and presence of interference.

This work presents the study and design of vibration analyzer self-powered by energy harvesting, using piezoelectric and magnetic components to convert mechanical energy into electric energy. A prototype vibration analysis instrument was developed to monitor the operating state of machines and motors. This system is powered by a Lithium battery which by its own is recharged by the generator, providing the configuration of self-sufficient to the system. The hybrid structure proposed in this work analyzes and demonstrates the gains and obtained from the miscegenation of two generator models. The results obtained from simulations and tests in the experimental platform show an effective solution for power generation for low energy consumption devices and instruments.

Anechoic chambers are indicated for the determination of antenna radiation diagrams, among other quantities. This is because cameras of this type are produced with materials that absorb electromagnetic waves, reducing the occurrence of reflections. However, anechoic chambers are often too costly for small businesses or university centers to acquire. In the state of Bahia, for example, there is only one semi-anechoic chamber and none completely anechoic. In order to overcome this difficulty, several techniques are currently being developed with the purpose of suppressing the effect of echoes in measurements performed in echo environments. This has proven to be possible through the use of digital signal processing techniques based on classical concepts such as deconvolution, impulse response and filtering. In this work, a technique for obtaining the radiation diagram in non-anechoic environments was studied. Briefly, the method consists of making measurements taken at different initial positions of the transmitting and receiving antennas and, through a similarity identification algorithm, reconstructing the real radiation diagram. The reconstructed diagram was then evaluated from the comparison with the radiation diagram obtained in an anechoic chamber. The obtained results indicated that the proposed technique is promising for the reconstruction of antenna radiation measured in echo environments. A better functioning of the technique was observed especially when peaks / attenuations due to reflections occurred in only one specific position.

The generation of electric energy from solar energy is an alternative to the traditional energy matrix and, increasingly, its participation in the energy generation market is increasing. Solar irradiation depends on several factors, such as: locality, season, atmospheric composition, cloud cover and surface shape. As the intensity of the solar radiation is variable, a survey of its values is necessary to guarantee a better use of the generation system throughout the year. In this work, a photovoltaic data acquisition and sensing system was implemented to characterize the local insolation, based on three solar cell technologies, namely: monocrystalline silicon, polycrystalline silicon and thin film. During the calibration and validation tests, the saturation of the silicon cell signals was observed when the external resistive load was greater than 30 ohm. This saturation can be avoided by the use of an acrylic bulkhead, with two hole matrices, which served as a collimator and attenuator (attenuation ratio of -12.0 ± 0.5 dB). The developed system was able to determine the normalized efficiency of the solar cells during the day, in different seasons of the year. Thus in the winter the cell that has the highest normalized efficiency was CIGS, whereas in the spring this domain falls, remaining at the beginning and at the end of the day. In the summer the silicon cells show higher normalized efficiency when compared to the CIGS cell.

Alcoholic distillation, one of the main productive stages of the sugar and alcohol industry, still shows timid results in terms of automatic control and optimization. However, advanced control approaches give us good prospects for improving safety, efficiency and productivity in this type of process. In particular, predictive control techniques, widely known as MPC (Model Predictive Control) allow the operation of the distillation columns considering the complexity of the system, such as multiple inputs and outputs and numerous disturbances that can compromise the quality of the products. In addition, strategies based on time compensators, such as the filtered Smith Predictor (FSP), can incorporate formulations that allow to increase the robustness of the closed loop control system. From this perspective, this work presents the application of an advanced control structure in two layers. The first layer consists of implementing a predictive control system. In particular, the infinite horizon (IHMPC), MIMO FSP and DTCGPC were designed, capable of maintaining the alcoholic distillation process under the conditions defined as a reference. For the second layer, a dynamic optimizer in real time was developed, capable of calculating the best operation points of the process using a reliable model, considering economic and productivity conditions. For that, a computational platform was also developed that allows the advanced simulation of the system as well as its dynamic behavior in a Software-in-the-Loop environment as a way to guarantee the fidelity of the results and propose the application of these strategies in a real scenario. OLE (Object Linking and Embedding) Automation integrates Matlab and Aspen Hysys software to simulate practical OPC (Open Platform Communications) scenarios commonly found in the industry. Controllers are evaluated, in simulation scenarios, from control performance indexes (ITAE, ITSE, SSC) and the computational performance of each strategy, as a quantitative metric for comparing the proposed structures. Furthermore, the economic performance of the optimization layer is compared with the classic PID control strategy still applied in the sugar and alcohol industry. This study showed that the dynamic optimization associated with predictive controllers increased the production of the distillation column, showing itself to be more robust and stable compared to the approaches commonly used in the current scenario of ethanol distilleries. In view of the particular benefits of each strategy, this work aims to propose different control solutions for different scenarios of the sugar and alcohol industry, in order to maintain a more efficient, economical and clean production. In addition, it is intended to evolve the practical application of advanced control strategies that take into account the stability of the system, the efficiency of production and the robustness in closed loop.

Diagnosis of Knowledge Management based on the measurement of performance indicators in Brazilian civil construction companies in the building sector, whose main activities are: execution of residential works and industrial and commercial buildings. Use of knowledge management using performance indicators in the field of civil construction, aiming to study its application effectiveness and convergence of the identified results. The available literature is focused on the conceptual field, without effective proof of application and validation in practice. A self-administered questionnaire for data collection was elaborated, with subsequent application of the Diffuse Analytical Hierarchical Method (Fuzzy-AHP) to choose and make decisions about the best Knowledge Management model for the Brazilian civil construction sector. The use of Fuzzy-AHP to diagnose the importance and performance of Knowledge Management for Brazilian Civil Construction Organizations in the Buildings sector ratified the need for a structured approach so that organizations can remain in constant learning, productive and competitive. Performance gaps and opportunities for improvement in the field of knowledge were identified as knowledge management is centered on controls that do not necessarily provide opportunities for improvement for organizations, being used only as a source of data and information to support research. reaction to the failures that occurred, without visualization by all hierarchies, mainly the tactical and operational levels. It is important to emphasize the implications of this research related mainly to the nature of qualitative work. As it involves subjective elements of analysis and considerations, the present situation and feeling of the participants reflectedand considerations, the situation and present feeling of the participants reflected considerably in the contributory actions of this research. Similarly, there are the limitations of Knowledge Management itself, as it includes intangible assets, which are difficult to measure, especially if related to tacit knowledge. Despite a vast body of literature on Knowledge Management, the article demonstrates these factors within a Brazilian context. An essential point is that the study is based on feedback from construction professionals and managers, who carry out their analyzes and conclusions from practical points of view, without relevant theoretical knowledge of the relationship between the process variables. The Diffuse Analytical Hierarchical Method applied to the theme of Knowledge Management through Performance Indicators in Civil Construction addresses the degrees of imprecision arising from difficulties and ambiguities of human perception.

The thesis deals with the analysis of the feedback transverse linearization as a control for a unicycle robot's navigation. The path to be followed by a robot will be given through a set of curves whose distance is Euclidean. Feedback transverse linearization makes this path attractive and time-invariant, it means that, if the robot is initialized on the path with an appropriate orientation, it will remain on the path for all the time. The limitation of the technique is analyzed when the angle between the segments of lines that compose the path is obtuse, in this case, the robot follows an opposite path to the desired. There was no evidence in the literature of limitation of transverse feedback linearization when the angles are obtuse. It is mathematically demonstrated the existence of a set of controller gain values that guarantee that the robot will follow the desired path, if the angle formed between the straight segments of the path is acute. Dynamic Transverse Feedback Linearization is proposed as a solution to this limitation. For the purpose of exemplifying the proposed control, the transverse feedback linearization with the rapidly exploring random trees star is integrated, modifying the path planner in order to meet the limitation of the transverse feedback linearization. Here, such linearization is compared with dynamic transverse feedback linearization. Transverse feedback linearization is applied to the Kobuki robotic platform in order to experimentally show the robot convergence in a direction opposite to the desired one.

The use of photovoltaics is still considered to be challenging because of certain reliability issues and high dependence on the global horizontal irradiance (GHI). GHI forecasting has a wide application from grid safety to supply-demand balance and economic load dispatching. On the other hand, load forecasting is important for smart grids and the electricity market in terms of the meeting the demand and distribution of electrical energy. The current research develops a methodology for more precise forecasting results. Two different approaches are developed and applied for GHI forecasting and electric load forecasting. Given a data set, a multi-layer perceptron neural network (MLPNN) is a strong tool for solving the forecasting problems. Furthermore, noise detection and feature selection in a data set with numerous variables are of crucial importance to obtain the desired results. This work employs density-based spatial clustering of applications with noise (DBSCAN) and non-dominated sorting genetic algorithm II (NSGA II) algorithms for noise detection and feature selection, respectively. Tuning the neural network is another important issue that includes choosing the hidden layer size and activation functions between the layers of the network. Previous studies have utilized a combination of different parameters based on trial and error, which seems to be inefficient in terms of accurate selection of the desired features and also tuning of the neural network. In this research, two different methods namely, particle swarm optimization (PSO) algorithm and genetic algorithm (GA) are utilized in order to tune the MLPNN, and the results of one-hour ahead forecasting of the GHI are subsequently compared. The methodology is validated using the hourly data for Elizabeth City located in North Carolina, USA, and the results demonstrated a better performance of GA in comparison with PSO. For electric power demand forecasting, this research proposes a hybrid algorithm for improving the forecasting accuracy where a non-dominated sorting genetic algorithm II (NSGA II) is employed for selecting the input vector, where its fitness function is a multi-layer perceptron neural network (MLPNN). Thus, the output of the NSGA II is the output of the best-trained MLPNN which has the best combination of inputs. The result of NSGA II is fed to the Adaptive Neuro-Fuzzy Inference System (ANFIS) as its input and the results demonstrate an improved forecasting accuracy of the MLPNN-ANFIS compared to the MLPNN and ANFIS models. In addition, genetic algorithm (GA), particle swarm optimization (PSO), ant colony optimization (ACO), differential evolution (DE), and imperialistic competitive algorithm (ICA) are used for optimized design of the ANFIS. Electricity demand data for Bonneville, Oregon are used to test the model and among the different tested models, NSGA II-ANFIS-GA provides better accuracy.

The GA-tuned MLPNN reported a normalized root mean square error (nRMSE) of 0.0458 and a normalized mean absolute error (nMAE) of 0.0238 for hourly GHI forecasting and obtained values of error indicators for one-hour-ahead demand forecasting are 107.2644, 1.5063, 65.4250, 1.0570, and 0.9940 for RMSE, RMSE%, MAE, MAPE, and R, respectively.

Mainly when applied in the underwater environment, sonar simulation requires modelling complex acoustic physics and simultaneously rendering time-efficient data. Simulation of sonar operation allows evaluating algorithms and control systems without going to the real underwater environment; that reduces the costs and risks of in-field experiments. Existing methods focus on a basic implementation of one sonar device type with high computational cost, where most of the sound properties are disregarded. Towards a high fidelity virtualization with real time constraints, our proposed system is able to reproduce the operation of two main types of imaging sonars: mechanical scanning imaging sonar and forward-looking sonar. The underwater simulated scene is performed based on three frameworks: (i) Gazebo handles with hydrostatic and hydrodynamic forces, (ii) OpenSceneGraph renders the ocean visual effects, and (iii) ROCK framework manages the sonar in the virtual underwater scenario and exports simulation resources. The underwater simulated scene is then processed by a hybrid rendering pipeline on GPU: Rasterization computes the primary reflections between the sonar and observable surfaces based on deferred shading, while only the reflective areas are ray-traced. Also, the parallel ray-box routines on GPU also accelerated the intersection tests on ray-tracing algorithm. This approach launches few rays when compared to a full ray-tracing, achieving a significant performance gain without quality loss in the final image. Resulting reflections are then characterized as two sonar parameters: Echo intensity and pulse distance, being all calculated over insonified objects in the 3D rendered scene. Sonar-intrinsic parameters, such as speckle noise, transmission loss, reverberation and material properties of observable objects are also considered as part of the final acoustic image. Our evaluation demonstrated that the proposed system is able to operate close to real-world devices in terms of computation time. Also, our method demonstrated the effectiveness to render complex scenes with high quality when compared with real-world sonar images of similar scenes.

This work presents a predictive control strategy with stability assurance for systems composed of unstable and integrating modes. Such dynamic characteristics make it difficult to control the system, especially when the controller synthesis seeks to guarantee the stability of the system in closed loop. Here, a permanent error-free control law based on a single optimization problem is proposed, and closed-loop stability is achieved by adopting an infinite prediction horizon, which together with the imposition of a terminal constraint set associated with the system of unstable and integrating states, guarantees their viability. The results obtained from the simulated scenarios in two systems, CSTR (Reactor Continuous Tank Reactor) and rotational inverted pendulum, point to an efficient solution to systematically improve the control mesh.

Classical mechanics, the oldest of the physical sciences, has been since its inception the tool used by humans to study its relationship with the environment in which it operates. Galileo, and then Newton, laid the groundwork for this to be done scientifically. Two procedures are used to study human movement: quantitative and qualitative analysis. Quantitative analysis describes the movements of the body or its parts in numerical terms, and qualitative analysis seeks to describe a movement in non-numerical terms. Motion analysis of the human body studies the objective variation of position within the space in which it develops and over a certain period of time. Over time, motion analysis began to be used as an evaluation of the evolution of physical therapy treatments. To analyze motion, multiple camera devices are required, as a single camera may not capture details of the movement being performed, especially when the user is not facing the camera frontally. To use a system composed of multiple cameras, it is necessary to apply a calibration process in order to obtain the parameters used to unify the different coordinate systems of each camera. Currently, there are devices with multiple cameras, but high cost, but not accessible. In this sense, the present work proposes the use of multiple kinect RGB-D cameras, performing a stereo calibration, by pairs of cameras, in order to determine the best configuration that contains minimal occlusion and self-occlusion problems inherent to the use of a single camera. In addition, the process of defining camera location and position aims to minimize interference between the devices used. To evaluate system performance, the composite skeleton from the fusion of the skeletons provided by each kinect was used, using the positions of the upper limb and head joints. The results show the best angular capture distances that correspond to a lower occurrence of the aforementioned problems, and the resulting fusion skeleton is more robust than when a single skeleton is used to evaluate the movement.

The electrocardiogram (ECG) was invented by Willian Einthoven in 1902 and is the most used clinical procedure in the detection and diagnosis of arrhythmias, infarcts and even noncardiac anomalies. Because of its simplicity compared to other available techniques, it is often used as a research tool in long-term population clinical trials and in experimental drug trials that may affect cardiac functions. The cardiac monitor is one of the electromedical equipment available to display the ECG signals on the screen and as such must present technical compliance to the regulatory requirements of each country where it is marketed. ABNT NBR IEC 60601-2-27: 2013 is the Brazilian technical standard that establishes the safety and performance requirements of the electrocardiographic monitoring equipment and describes the compulsory tests for its certification in our territory. The procedures for performing calibration tests suggested by the standard are inherently manual and their systematic execution can induce sources of error to the measurements, influencing the judgment of the functionality of the apparatus. The present work consisted in the construction of an integrated and automated system for the calibration of cardiac monitors that presented, in the validation, uncertainties smaller than 1 \% and made possible the execution of the tests in a simplified, robust and reliable way.

With the potential growth of devices and objects connected to the Internet, the Internet of things (IoT) has gained special attention in several areas of knowledge, being seen as the key technology for ubiquitous computing. Within the context of access control in parking lots, IoT is a promising reference and of great growth potential, by enabling the automation of repetitive processes and increase the safety of users of this site. In this way, this dissertation was developed with the purpose of developing a non-invasive system capable of assisting the control of parking vehicles via wireless mobile use of commonly installed technologies in public and private places, such as and mobile devices that have daily Wi-Fi connectivity user. From this, a system was developed that rules out the need for marking cars and other devices that are not in common use. For the user identification via Wi-Fi was used the NodeMCU ESP8266, which was programmed to function as an access point, allowing the MAC address to be read of the cellular network card at the time it is connected and then, enabling the opening of the automatic gate or gate for the registered user. For to perform a secure closing, a computer vision system was developed able to identify people or vehicles in the gate region. Prioritizing the bass computational cost was created two algorithms, the first using medium filter and the second employing Multi-Layer Perceptron neural network (MLP). From tests it was verified that the median filter as a classifier showed to be efficient for the group of images obtained during the day. However for the group of images captured during the night period, due to the low illumination in the region of the gate where the tests were performed, it was necessary to use an extractor and a more robust classifier. Therefore, we used the Local Binary Patterns (LBP) and another extractor proposed in this work for the RNA. Finally, a success rate of 71.6% was obtained for LBP and 92.6% with the proposed extractor.

This work presents the development of a closed loop control system capable of maintaining a constant dilution ratio between the air / exhaust gas mixture in tests using constant volume particle dilution (CVS) tunnel, in analyzes of particulates emitted by diesel engines. This control is necessary because, as the particulate material collection tests are carried out, the flow of treated air is reduced, due to the accumulation of dirt in its filters, causing its saturation. The project automated the particle collection bench previously developed in the Laboratory of Energy and Instrumentation - LEI of the Federal University of Recôncavo da Bahia - UFRB. As part of the development process, the instrumentation and power drive devices were constructed, revised and calibrated. The system management was performed by an embedded system with free platform microcontroller communicating with MATLAB®. The design and tuning of the controller were done after the mathematical model, based on the experimental data obtained in CVS. The implemented classical controller was evaluated using the basic performance metrics for both trajectory and regulatory (disturbance rejection) testing, with success in both cases.

This work aims to contribute to the development of solutions for the planning of people transportation. The characteristics involved in transportation planning problems are extensively studied in a branch of the Operational Research known as Vehicle Routing Problem (PRV). The PRV is derived in several other specific problems depending on the context in which they are inserted. When it comes to transporting people, as is the case study proposed in this paper, the characteristics of the problem are included in the PRV variation known as the Dial-a-Ride Problem or the Dial-a-Ride Problem (DARP). Most of the studies carried out on DARP are directed to the application in the transport of a certain target public, such as: patients in medical treatment, elderly, wheelchair users, people with special needs (from home to hospitals, clinics, etc. and vice versa ). This work presents a new computational approach based on Fuzzy Logic, Reinforcement Learning and Agents as a solution proposal for the static DARP with heterogeneous vehicle fleet. This approach considers some variables such as: distance and average time, comfort, availability of drivers and vehicles, amount of demand, load capacity, road condition and number of passengers as basis for decision making. A new fuzzy model was developed to assist in decisions about how best to organize requests and choose the best vehicle for each. A Q-Learning Enhancement Learning algorithm is used to generate the routes to meet the requests for transportation. Eight test cases are presented, using real and fictitious data, as a case study for the validation of the proposed solution. The results found, based on established quality criteria, such as: comfort, quantity of requests, average time and distance, indicate that the proposed solution is adequate for most of the analyzed scenarios.

-The deposition of paraffin in ducts is one of the big problems faced by the petrol industry during the production, transportation and storage of petrol. The acumulation of paraffin cristilized in the walls of the pipes, leads to the loss of force in flow on and, consequently the increase in the power necessary for the system of bombardiment, when the decrease of flow rate or a situation of extreme and complete obstruction of tubulation doesn’t happen, causing disorders, waste and decrease of production rythim. Mechanical methods of brushing, washing, scraping, chemical attack, warming, are some of the most common methods that dispose the industry to combat and try to overcome this obstacle in the production and petrol transportation. Research centrers make effort to understand the base parameters of this phenomenon and develop models that offer great precision in obstruction prevision of transportation lines of petrol by the crisitalization and deposit of paraffin. In spite of this, there are still limits not yet overcomed concerning the understanding of the variables and mechanisms involved in the process of deposition of paraffin in ducts. The objective of this research was to verify a possibility of data collection and information, referring to the deposition mechanisms of paraffin in lines of production and transportation of petrol, in a laboratory that would produce in a realistic scale of parameters, the stages and the phenomenon of flow of petrol in the field of oil Industry. Materials and equipments with compatible dimensions with those practised in the field were used. This was due to, as it is believed that, the more the experimental system of parameters resembles the real ones, the date and information collected, without mentioning the final results of the given investigation, reach the highest level of confiability as regards to prevision of the dynamic of the deposit of paraffin in the pipes. Initially, some tests of dewaxing were carried out using the cavitation provoked by ultrasonic transducers joined to tanks. After that, a line of paraffination in realistic scale was projected and build, where parameters related to the flow of oil and paraffination of ducts would be reproduced and measured. The line also served as base for the development of a temperature sensor and for initial tests of the system of dewaxing using ultrasonic necklaces.