The diploma thesis deals with the image registration using evolutionary algorithms from metaheuristic optimization techniques which are considered recent and widely used. The frst part of the thesis contains theoretical description of components in image registration and later focuses on the ultrasound images. In that part, the thesis explains chosen evolutionary algorithms. Three optimization methods have been implemented, in particular genetic algorithm, particle swarm optimization and frey algorithm. Chosen similarity metrics for optimization are sum of squared dierences, cosine similarity and correlation coefcient. The main part of thesis includes testing of proposed methods with the evaluation of obtained results. These parameters are later used for optimization of real ultrasound sequences obtained by the contrast imaging.

<div>In this thesis work, a novel approach for the design of a fractional order proportional integral</div><div>derivative (FOPID) controller is proposed. This design introduces a new time-varying FOPID controller</div><div>to mitigate a voltage spike at the controller output whenever a sudden change to the setpoint occurs. The</div><div>voltage spike exists at the output of the proportional integral derivative (PID) and FOPID controllers when a</div><div>derivative control element is involved. Such a voltage spike may cause a serious damage to the plant if it is</div><div>left uncontrolled. The proposed new FOPID controller applies a time function to force the derivative gain to</div><div>take effect gradually, leading to a time-varying derivative FOPID (TVD-FOPID) controller, which maintains</div><div>a fast system response and signi?cantly reduces the voltage spike at the controller output. The time-varying</div><div>FOPID controller is optimally designed using the particle swarm optimization (PSO) or genetic algorithm</div><div>(GA) to ?nd the optimum constants and time-varying parameters. The improved control performance is</div><div>validated through controlling the closed-loop DC motor speed via comparisons between the TVD-FOPID</div><div>controller, traditional FOPID controller, and time-varying FOPID (TV-FOPID) controller which is created</div><div>for comparison with all three PID gain constants replaced by the optimized time functions. The simulation</div><div>results demonstrate that the proposed TVD-FOPID controller not only can achieve 80% reduction of voltage</div><div>spike at the controller output but also is also able to keep approximately the same characteristics of the system</div><div>response in comparison with the regular FOPID controller. The TVD-FOPID controller using a saturation</div><div>block between the controller output and the plant still performs best according to system overshoot, rise time,</div><div>and settling time.</div>

The PID controller is regarded as a dependable and reliable controller for process industry systems. Many researchers have devoted time and attention to PID controller tuning and they all agree that PID controllers are very important for control systems. A PID equation is very sensitive; its parameters must always be varied following the specific application to increase performance, such as by increasing the system’s responsiveness. PID controllers still have many problems despite their importance for control systems in industries. The problem of big overshoot on the conventional gain tuning is one of the serious problems. Researchers use the PSO algorithm to try and overcome those problems. The tuning of the MIMO PID controller based on the PSO algorithm shows many disadvantages such as high-quality control with a short settle time, steady-state error, and periodical step response. The traditional PSO algorithm is very sensitive and it sometimes affects the quality of good PID controller tuning. This research has proposed a new equation for improving the PSO algorithm. The proposed algorithm is the combination of linearly decreasing inertia weight and chaotic inertia weight, after which a control factor was introduced as an exponential factor. This was very useful for simulations as it is adjustable. The Matlab simulation results of the experiments show that the simulations as it is adjustable. The Matlab simulation results of the experiments show that the new proposed equation converges faster and it gives the best fitness compared to linear inertia weight and oscillating inertia weight and other old equations. The MIMO PID controller system that consists of four plants was tuned based on the new proposed equation for the PSO algorithm (LCPSO). The optimized results show the best rise time, settling time, time delays, and steady-state compared to the systems that are tuned using the old equations. The exploration was directed at considering the impact of using the PSO calculation as an instrument for MIMO PID tuning. The results obtained in the examination reveal that the PSO tuning output improved reactions and can be applied to various system models in the measure control industry. The results for the MIMO PID controller tuned using PSO were assessed using integral square error (ISE), integral absolute error (IAE), and the integral of time expanded by absolute error (ITAE). The five well-known benchmark functions were also used to endorse the feasibility of the improved PSO and excellent results in terms of convergence and best fitness were attained. / Electrical and Mining Engineering / M. Tech. (Electrical Engineering)

Avec l’arrivée des objets connectés, la reconnaissance de l’activité physique connait une nouvelle ère. De nouvelles considérations sont à prendre en compte afin d’aboutir à un meilleur processus de traitement. Dans cette thèse, nous avons exploré le processus de traitement pour la reconnaissance de l’activité physique dans un environnement non contrôlé. Les activités physiques reconnues, avec seulement une centrale inertielle (accéléromètre, gyroscope et magnétomètre), sont dites élémentaires. Les autres types d’activités dépendantes d’un contexte sont dites « basés sur le contexte ». Nous avons extrait la transformée en cosinus discrète (DCT) comme principal descripteur pour la reconnaissance des activités élémentaires. Afin de reconnaitre les activités physiques basées sur le contexte, nous avons défini trois niveaux de granularité : un premier niveau dépendant des objets connectés embarqués (smartphone, smartwatch et samrt TV). Un deuxième niveau concerne l’étude des comportements des participants en interaction avec l’écran de la smart TV. Le troisième niveau concerne l’étude de l’attention des participants envers la TV. Nous avons pris en considération l’aspect imperfection des données en fusionnant les données multi capteurs avec le modèle de Dempster-Shafer. A ce titre, nous avons proposé différentes approches pour calculer et approximer les fonctions de masse. Afin d’éviter de calculer et sélectionner les différents descripteurs, nous avons proposé une approche basée sur l’utilisation d’algorithmes d’apprentissage en profondeur (DNN). Nous avons proposé deux modèles : un premier modèle consiste à reconnaitre les activités élémentaires en sélectionnant la DCT comme principal descripteur (DNN-DCT). Le deuxième modèle consiste à apprendre les données brutes des activités basées sur le contexte (CNN-brutes). L’inconvénient du modèle DNN-DCT est qu’il est rapide mais moins précis, alors que le modèle CNN-brutes est plus précis mais très lent. Nous avons proposé une étude empirique permettant de comparer les différentes méthodes pouvant accélérer l’apprentissage tout en gardant un niveau élevé de précision. Nous avons ainsi exploré la méthode d’optimisation par essaim particulaires (PSO). Les résultats sont très satisfaisants (97%) par rapport à l’apprentissage d’un réseau de neurones profond avec les méthodes d’optimisation classiques telles que la descente de Gradient Stochastique et l’optimisation par Gradient accéléré de Nesterov. Les résultats de nos travaux suggèrent le recours à de bons descripteurs dans le cas où le contexte n’importe peu, la prise en compte de l’imperfection des données capteurs quand le domaine sous-jacent l’exige, l’utilisation de l’apprentissage profond avec un optimiseur permettant d’avoir des modèles très précis et plus rapides. / With the arrival of connected objects, the recognition of physical activity is experiencing a new era. New considerations need to be taken into account in order to achieve a better treatment process. In this thesis, we explored the treatment process for recognizing physical activity in an uncontrolled environment. The recognized physical activities, with only one inertial unit (accelerometer, gyroscope and magnetometer), are called elementary. Other types of context-dependent activities are called "context-based". We extracted the DCT as the main descriptor for the recognition of elementary activities. In order to recognize the physical activities based on the context, we defined three levels of granularity: a first level depending on embedded connected objects (smartphone, smartwatch and samrt TV . A second level concerns the study of participants' behaviors interacting with the smart TV screen. The third level concerns the study of participants' attention to TV. We took into consideration the imperfection aspect of the data by merging the multi sensor data with the Dempster-Shafer model. As such, we have proposed different approaches for calculating and approximating mass functions. In order to avoid calculating and selecting the different descriptors, we proposed an approach based on the use of deep learning algorithms (DNN). We proposed two models: a first model consisting of recognizing the elementary activities by selecting the DCT as the main descriptor (DNN-DCT). The second model is to learn raw data from context-based activities (CNN-raw). The disadvantage of the DNN-DCT model is that it is fast but less accurate, while the CNN-raw model is more accurate but very slow. We have proposed an empirical study to compare different methods that can accelerate learning while maintaining a high level of accuracy. We thus explored the method of optimization by particle swarm (PSO). The results are very satisfactory (97%) compared to deep neural network with stochastic gradients descent and Nesterov accelerated Gradient optimization. The results of our work suggest the use of good descriptors in the case where the context matters little, the taking into account of the imperfection of the sensor data requires that it be used and faster models.

The genetic algorithm (GA) is a popular random search and optimization method inspired by the concepts of crossover, random mutation, and natural selection from evolutionary biology. The real-valued genetic algorithm (RGA) is an improved version of the genetic algorithm designed for direct operation on real-valued variables. In this work, a modified version of a genetic algorithm is introduced, which is called a modified genetic algorithm with micro-movement (MGAM). It implements a particle swarm optimization(PSO)-inspired micro-movement phase that helps to improve the convergence rate, while employing the e'cient GA mechanism for maintaining population diversity. In order to test the capability of the MGAM, we firrst implement it on five generally used test functions. Then we test the MGAM on two typical nonlinear dynamical systems. The performance of the MGAM is compared to a basic RGA on all these applications. Finally, we implement the MGAM on the most important application, which is the plasma physics-based model of the solar wind-driven magnetosphere-ionosphere system (WINDMI). In order to use this model for real-time prediction of geomagnetic activity, the model parameters require up-dating every 6-8 hours. We use the MGAM to train the parameters of the model in order to achieve the lowest mean square error (MSE) against the measured auroral electrojet (AL) and Dst indices. The performance of the MGAM is compared to the RGA on historical geomagnetic storm datasets. While the MGAM performs substantially better than the RGA when evaluating standard test functions, the improvement is about 6-12 percent when used on the 20D nonlinear dynamical WINDMI model.

Indiana University-Purdue University Indianapolis (IUPUI) / As our population ages, treating bone and joint ailments is becoming increasingly important. Both osteoporosis, a bone disease characterized by a decreased density of mineral in bone, and osteoarthritis, a joint disease characterized by the degeneration of cartilage on the ends of bones, are major causes of decreased movement ability and increased pain. To combat these diseases, many treatments are offered, including drugs and exercise, and much biomedical research is being conducted. However, how can we get the most out of the research we perform and the treatment we do have? One approach is through computational analysis and mathematical modeling. In this thesis, quantitative methods of analysis are applied in different ways to two systems: osteoporosis and osteoarthritis. A mouse model simulating osteoporosis is treated with salubrinal and knee loading. The bone and cell data is used to formulate a system of differential equations to model the response of bone to each treatment. Using Particle Swarm Optimization, optimal treatment regimens are found, including a consideration of budgetary constraints. Additionally, an in vitro model of osteoarthritis in chondrocytes receives RNA silencing of Lrp5. Microarray analysis of gene expression is used to further elucidate the mode of regulation of ADAMTS5, an aggrecanase associated with cartilage degradation, by Lrp5, including the development of a mathematical model. The math model of osteoporosis reveals a quick response to salubrinal and a delayed but substantial response to knee loading. Consideration of cost effectiveness showed that as budgetary constraints increased, treatment did not start until later. The quantitative analysis of ADAMTS5 regulation suggested the involvement of IL1B and p38 MAPK. This research demonstrates the application of quantitative methods to further the usefulness of biomedical and biomolecular research into treatment and signaling pathways. Further work using these techniques can help uncover a bigger picture of osteoarthritis's mode of action and ideal treatment regimens for osteoporosis.

Indiana University-Purdue University Indianapolis (IUPUI) / A gradient free function optimization technique, namely particle swarm optimization (PSO) algorithm, is utilized in parameter identification of the electrochemical model of a Lithium-Ion battery having a LiCoO2 chemistry. Battery electrochemical model parameters are subject to change under severe or abusive operating conditions resulting in, for example, Navy over-discharged battery, 24-hr over-discharged battery, and over-charged battery. It is important for a battery management system to have these parameters changes fully captured in a bank of battery models that can be used to monitor battery conditions in real time. In this work, PSO methodology has been used to identify four electrochemical model parameters that exhibit significant variations under severe operating conditions. The identified battery models were validated by comparing the model output voltage with the experimental output voltage for the stated operating conditions. These identified conditions of the battery were then used to monitor condition of the battery that can aid the battery management system (BMS) in improving overall performance. An adaptive estimation technique, namely multiple model adaptive estimation (MMAE) method, was implemented for this purpose. In this estimation algorithm, all the identified models were simulated for a battery current input profile extracted from the hybrid pulse power characterization (HPPC) cycle simulation of a hybrid electric vehicle (HEV). A partial differential algebraic equation (PDAE) observer was utilized to obtain the estimated voltage, which was used to generate the residuals. Analysis of these residuals through MMAE provided the probability of matching the current battery operating condition to that of one of the identified models. Simulation results show that the proposed model based method offered an accurate and effective fault diagnosis of the battery conditions. This type of fault diagnosis, which is based on the models capturing true physics of the battery electrochemistry, can lead to a more accurate and robust battery fault diagnosis and help BMS take appropriate steps to prevent battery operation in any of the stated severe or abusive conditions.

No description available.

No description available.

No description available.