Medidas complementares ao monitoramento in situ da qualidade da água podem ser obtidas por meio de sensoriamento remoto, sendo clorofila-a e sólidos suspensos alguns dos parâmetros que podem ser estimados. Este trabalho teve como objetivo explorar técnicas de processamento de imagens, análises estatísticas e de inteligência artificial com o objetivo de predizer e modelar as concentrações de clorofila-a e sólidos suspensos totais na Represa do Lobo/SP. Metodologicamente, foram realizadas coletas em campo, em três diferentes datas, para amostragem de água e posterior análise laboratorial. Os resultados limnológicos foram analisados, modelados e comparados com imagens processadas do satélite Sentinel-2. Análises de regressão e redes neurais artificiais (RNA) foram exploradas para gerar modelos de predição para a área de estudo. Os resultados indicam que métodos de regressão podem não ser adequados para capturar as relações lineares e/ou não-lineares entre os compostos de interesse e as respostas espectrais da água recebidas pelo satélite, indicando a capacidade das redes neurais em modelar relações mais complexas. Através da integração da resposta que o sensor MSI do satélite Sentinel-2 coletou nas regiões do visível ao infravermelho médio e de RNAs foi possível modelar a concentração de clorofila-a, com valores de R² superiores a 0,65 e de RMSE inferiores a 2,5 μg/L, e gerar mapas que permitam seu monitoramento temporal e análise espacial na área de estudo. Os resultados para SST não foram satisfatórios devido à complexidade óptica do ambiente analisado, bem como as baixas concentrações de SST na represa. Portanto, a integração de dados de sensoriamento remoto no mapeamento de corpos d\'água com a aplicação de redes neurais na análise de dados é uma abordagem promissora para prever clorofila-a e sólidos suspensos, bem como suas variações temporais e espaciais. / Complementary measures to in situ monitoring of water quality can be obtained through remote sensing, with chlorophyll-a and suspended solids being some of the parameters that can be estimated. The objective of this work was to explore techniques for image processing, statistical analysis and artificial intelligence with the objective of predicting and modeling the concentrations of chlorophyll-a and total suspended solids in the Lobo Reservoir/SP. Methodologically, field samples were collected in three different dates for water sampling and laboratory analysis. The limnological results were analyzed, modeled and compared with processed images of the Sentinel-2 satellite. Regression analysis and artificial neural networks (ANNs) were explored to generate prediction models for the study area. The results indicate that regression methods may not be adequate to capture linear and/or nonlinear relationships between the compounds of interest and the spectral responses of water received by the satellite, indicating the ability of neural networks to model more complex relationships. Through of the integration of response wich the MSI sensor of Sentinel satellite collected in the visible and near-infrared regions and of the ANN analysis was possible modeling the chlorophyll-a concentration, wich R² values highers of 0.65 and RMSE less 2.5, and create predict maps wich allow your temporal monitoring and spatial analysis in the study area. The TSS results were unsatisfactory because of the optic complexity of analysed ambient, as well as your small TSS concentrations in the Lobo Reservoir. Therefore, the integration of remote sensing data in the mapping of water bodies with the application of neural networks in the data analysis is a promising approach to predict chlorophyll-a and suspended solids as well as their temporal and spatial variations.

Drug development is a long and expensive process. It is often not until potential drug candidates are administered to humans that accurate quantification of their pharmacokinetic characteristics is achieved. The goal of developing quantitative structure-pharmacokinetic relationships (QSPkRs) is to relate the molecular structure of a chemical entity with its pharmacokinetic characteristics. In this thesis artificial neural networks (ANNs) were used to construct in silico predictive QSPkRs for various pharmacokinetic parameters using different drug data sets. Drug pharmacokinetic data for all studies were taken from the literature. Information for model construction was extracted from drug molecular structure. Numerous theoretical descriptors were generated from drug structure ranging from simple constitutional and functional group counts to complex 3D quantum chemical numbers. Subsets of descriptors were selected which best modeled the target pharmacokinetic parameter(s). Using manual selective pruning, QSPkRs for physiological clearances, volumes of distribution, and fraction bound to plasma proteins were developed for a series of beta-adrenoceptor antagonists. All optimum ANN models had training and cross-validation correlations close to unity, while testing was performed with an independent set of compounds. In most cases the ANN models developed performed better than other published ANN models for the same drug data set. The ability of ANNs to develop QSPkRs with multiple target outputs was investigated for a series of cephalosporins. Multilayer perceptron ANN models were constructed for prediction of half life, volume of distribution, clearances (whole body and renal), fraction excreted in the urine, and fraction bound to plasma proteins. The optimum model was well able to differentiate compounds in a qualitative manner while quantitative predictions were mostly in agreement with observed literature values. The ability to make simultaneous predictions of important pharmacokinetic properties of a compound made this a valuable model. A radial-basis function ANN was employed to construct a quantitative structure-bioavailability relationship for a large, structurally diverse series of compounds. The optimum model contained descriptors encoding constitutional through to conformation dependent solubility characteristics. Prediction of bioavailability for the independent testing set were generally close to observed values. Furthermore, the optimum model provided a good qualitative tool for differentiating between drugs with either low or high experimental bioavailability. QSPkR models constructed with ANNs were compared with multilinear regression models. ANN models were shown to be more effective at selecting a suitable subset of descriptors to model a given pharmacokinetic parameter. They also gave more accurate predictions than multilinear regression equations. This thesis presents work which supports the use of ANNs in pharmacokinetic modeling. Successful QSPkRs were constructed using different combinations of theoretically-derived descriptors and model optimisation techniques. The results demonstrate that ANNs provide a valuable modeling tool that may be useful in drug discovery and development.

<p>In this thesis a novel method for controlling a manipulator with arbitrary number of Degrees of freedom is proposed, the proposed method has the main advantages of two common controllers, the simplicity of PID controller and the robustness and accuracy of adaptive controller. The controller architecture is based on an Artificial Neural Network (ANN) and a PID controller.</p><p>The controller has the ability of solving inverse dynamics and inverse kinematics of robot with two separate Artificial Neural Networks. Since the ANN is learning the system parameters by itself the structure of controller can easily be changed to</p><p>improve the performance of robot.</p><p>The proposed controller can be implemented on a FPGA board to control the robot in real-time or the response of the ANN can be calculated offline and be reconstructed by controller using a lookup table. Error between the desired trajectory path and the path of the robot converges to zero rapidly and as the robot performs its tasks the controller learns the robot parameters and generates better control signal. The performance of controller is tested in simulation and on a real manipulator with satisfactory results.</p>

<p>Erfarenheter från tidigare försök på Korsnäs AB visar att det är väldigt svårt att på matematisk väg förutsäga vad som händer under framställningen av pappersmassa i en kontinuerlig kokare.</p><p>Målet med detta examensarbete var att undersöka möjligheterna att med hjälp av neurala nätverk underlätta regleringen genom att prediktera ligninhalten hos pappersmassan tre och en halv timme innan den aktuella flisen är färdigkokt.</p><p>På grund av den, med produktionstakten, varierande tidsförskjutningen mellan olika givarsignaler löstes problemet med en enkel, lokal modell per produktionstakt. Alla ingående modeller minimeras med avseende på både antalet noder i det gömda lagret och antalet ingångar, vilket gav en slutlig lösning med fyra enkla modeller uppbyggda av framåtkopplade neurala nätverk, var och ett med ett gömt lager innehållandes tre noder.</p><p>Prediktionen av ligninhalten påvisade till slut goda egenskaper, med avseende på hur väl prediktionen följer den verkliga kappatalsanalysatorn.</p>

<p>The purpose of this dissertation is to chart the development of an ambivalent female identity in the Gothic genre, as exemplified by Ann Radcliffe&rsquo / s late eighteenth century fictions. The thesis examines the social and literary context of the emergence of the Gothic in English literature and argues that it is intimately tied up with changes in social, political and gender relations in the period.</p>

In this thesis a novel method for controlling a manipulator with arbitrary number of Degrees of freedom is proposed, the proposed method has the main advantages of two common controllers, the simplicity of PID controller and the robustness and accuracy of adaptive controller. The controller architecture is based on an Artificial Neural Network (ANN) and a PID controller. The controller has the ability of solving inverse dynamics and inverse kinematics of robot with two separate Artificial Neural Networks. Since the ANN is learning the system parameters by itself the structure of controller can easily be changed to improve the performance of robot. The proposed controller can be implemented on a FPGA board to control the robot in real-time or the response of the ANN can be calculated offline and be reconstructed by controller using a lookup table. Error between the desired trajectory path and the path of the robot converges to zero rapidly and as the robot performs its tasks the controller learns the robot parameters and generates better control signal. The performance of controller is tested in simulation and on a real manipulator with satisfactory results.

ABSTRACT Granite Butterfly is a novel about three women—grandmother, mother, and daughter—and the unusual attachments that break apart their family. Tuula Laine is a Rockport, Massachusetts, native of Finnish descent, whose parents moved to Cape Ann for work in the area’s granite quarries. Her life changes one afternoon when her son Henri, a brilliant surgeon who has never seriously dated anyone before, visits with his pregnant girlfriend, Coreen. Tuula immediately senses that Coreen not the right match for him in terms of age, education, or temperament, and as the couple separates and unites over the course of one summer, Tuula witnesses, for the first time, the pattern of desire and abandonment that will define their relationship. By the time Tuula’s granddaughter, Suvi, is fourteen years old, she, too, has established a destructive relationship pattern with Coreen: whenever Coreen and Henri separate, Suvi’s mother clings to her until they develop a bond closer to that of sisters than a mother and child. In the final movement of the novel, this bond, and the bond between Suvi’s parents, is finally put to the test. Granite is cut into precise blocks—dynamite is never used, lest it shatter the stone. In a few short weeks, the Laine family is pulled apart, but unlike with quarrying, there is no way to divide them in a careful manner, no way to detach them that isn’t violent and abrupt, no way to predict, or guide, where they will split.

Erfarenheter från tidigare försök på Korsnäs AB visar att det är väldigt svårt att på matematisk väg förutsäga vad som händer under framställningen av pappersmassa i en kontinuerlig kokare. Målet med detta examensarbete var att undersöka möjligheterna att med hjälp av neurala nätverk underlätta regleringen genom att prediktera ligninhalten hos pappersmassan tre och en halv timme innan den aktuella flisen är färdigkokt. På grund av den, med produktionstakten, varierande tidsförskjutningen mellan olika givarsignaler löstes problemet med en enkel, lokal modell per produktionstakt. Alla ingående modeller minimeras med avseende på både antalet noder i det gömda lagret och antalet ingångar, vilket gav en slutlig lösning med fyra enkla modeller uppbyggda av framåtkopplade neurala nätverk, var och ett med ett gömt lager innehållandes tre noder. Prediktionen av ligninhalten påvisade till slut goda egenskaper, med avseende på hur väl prediktionen följer den verkliga kappatalsanalysatorn.

The owners of a house in today’s society do not know in real-time how much electricity they use. It could be beneficial for any residential consumer to have more control and overview in real-time over the electricity consumption. This could be done possible with a system that monitors the consumptions, micro renewables and the electricity prices from the grid and then makes a decision to either use or sell electricity to reduce the monthly electricity cost for the household and living a "Greener" life to reduce carbon emissions. In this thesis, estimations are made based on artificial neural network (ANN). The predictions are made for air temperature, solar insolation and wind speed in order to know how much energy will be produced in the next 24 hours from the solar panel and from the wind turbine. The predictions are made for electricity consumption in order to know how much energy the house will consume. These predictions are then used as an input to the system. The system has 3 controls, one to control the amount of sell or buy the energy, one to control the amount of energy to charge or discharge the fixed battery and one to control the amount of energy to charge or discharge the electric vehicle (EV). The output from the system will be the decision for the next 10 minutes for each of the 3 controls. To study the reliability of the ANN estimations, the ANN estimations (SANN) are compared with the real data (Sreal ) and other estimation based on the mean values (Smean) of the previous week. The simulation during a day in January gave that the expenses are 0.6285 € if using SANN, 0.7788 € if using Smean and 0.5974 € if using Sreal. Further, 3 different cases are considered to calculate the savings based on the ANN estimations. The first case is to have the system connected with fixed storage device and EV (Scon;batt ). The second and third cases are to have the system disconnected (without fixed battery) using micro generation (Sdiscon;micro) and not using micro generation (Sdiscon) along with the EV. The savings are calculated as a difference between Scon;batt and Sdiscon, also between Sdiscon;micro and Sdiscon. The saving are 788.68 € during a year if Scon;batt is used and 593.90 € during a year if Sdiscon;micro is used. With the calculated savings and the cost for the equipment, the pay-back period is 15.3 years for Scon;batt and 4.5 years for Sdiscon;micro. It is profitable to only use micro generation, but then the owner of the household loses the opportunity to be part of helping the society to become "Greener".

No description available.