Η παρούσα διατριβή ασχολείται με τη μελέτη, την ανάπτυξη και τη θεμελίωση νέων μεθόδων Υπολογιστικής Νοημοσύνης και Ευφυούς Βελτιστοποίησης. Συνοπτικά οργανώνεται στα ακόλουθα τρία μέρη: Αρχικά παρουσιάζεται το πεδίο της Υπολογιστικής Νοημοσύνης και πραγματοποιείται μία σύντομη αναφορά στους τρεις κύριους κλάδους της, τον Εξελικτικό Υπολογισμό, τα Τεχνητά Νευρωνικά Δίκτυα και τα Ασαφή Συστήματα. Το επόμενο μέρος αφιερώνεται στην παρουσίαση νέων, καινοτόμων οικογενειών των αλγορίθμων Βελτιστοποίησης Σμήνους Σωματιδίων (ΒΣΣ) και των Διαφοροεξελικτικών Αλγόριθμων (ΔΕΑ), για την επίλυση αριθμητικών προβλημάτων βελτιστοποίησης χωρίς περιορισμούς, έχοντας είτε ένα, είτε πολλαπλούς ολικούς βελτιστοποιητές. Οι αλγόριθμοι ΒΣΣ και ΔΕΑ αποτελούν τις βασικές μεθοδολογίες της παρούσας διατριβής. Όλες οι οικογένειες μεθόδων που προτείνονται, βασίζονται σε παρατηρήσεις των κοινών δομικών χαρακτηριστικών των ΒΣΣ και ΔΕΑ, ενώ η κάθε προτεινόμενη οικογένεια τις αξιοποιεί με διαφορετικό τρόπο, δημιουργώντας νέες, αποδοτικές μεθόδους με αρκετά ενδιαφέρουσες ιδιότητες και δυναμική. Η παρουσίαση του ερευνητικού έργου της διατριβής ολοκληρώνεται με το τρίτο μέρος στο οποίο περιλαμβάνεται μελέτη και ανάπτυξη μεθόδων ολικής βελτιστοποίησης για την εκπαίδευση Τεχνητών Νευρωνικών Δικτύων Υψηλής Τάξης, σε σειριακά και παράλληλα ή / και κατανεμημένα υπολογιστικά συστήματα. Η διδακτορική διατριβή ολοκληρώνεται με βασικά συμπεράσματα και τη συνεισφορά της. / The main subject of the thesis at hand revolves mainly around the development and foundations of new methods of computational intelligence and intelligent optimization. The thesis is organized into the following three parts: Firstly, we briefly present an overview of the field of Computational Intelligence, by describing its main categories, the Evolutionary Computation, the Artificial Neural Networks and the Fuzzy Systems. In the second part, we provide a detailed description of the newly developed families of algorithms for solving unconstrained numerical optimization problems in continues spaces with at least one global optimum. The proposed families are based on two well-known and widely used algorithms, namely the Particle Swarm Optimization (PSO) and the Differential Evolution (DE) algorithm. Both DE and PSO are the basic components for almost all methodologies proposed in the thesis. The proposed methodologies are based on common observations of the dynamics, the structural and the spacial characteristics of DE and PSO algorithms. Four novel families are presented in this part which exploit the aforementioned characteristics of the DE and the PSO algorithms. The proposed methodologies are efficient methods with quite interesting properties and dynamics. The presentation and description of our research contribution ends with the third and last part of the thesis, which includes the study and the development of novel global optimization methodologies for training Higher order Artificial Neural Networks in serial and parallel / distributed computational environments. The thesis ends with a brief summary, conclusions and discussion of the contribution of this thesis.

Most of the real world problems in water resources involve nonlinear formulations in their solution construction. Obtaining optimal solutions for large scale nonlinear optimization problems is always a challenging task. The conventional methods, such as linear programming (LP), dynamic programming (DP) and nonlinear programming (NLP) may often face problems in solving them. Recently, there has been an increasing interest in biologically motivated adaptive systems for solving real world optimization problems. The multi-member, stochastic approach followed in Evolutionary Algorithms (EA) makes them less susceptible to getting trapped at local optimal solutions, and they can search easier for global optimal solutions. In this thesis, efficient optimization techniques based on swarm intelligence and evolutionary computation principles have been proposed for single and multi-objective optimization in water resource systems. To overcome the inherent limitations of conventional optimization techniques, meta-heuristic techniques like ant colony optimization (ACO), particle swarm optimization (PSO) and differential evolution (DE) approaches are developed for single and multi-objective optimization. These methods are then applied to few case studies in planning and operation of reservoir systems in India. First a methodology based on ant colony optimization (ACO) principles is investigated for reservoir operation. The utility of the ACO technique for obtaining optimal solutions is explored for large scale nonlinear optimization problems, by solving a reservoir operation problem for monthly operation over a long-time horizon of 36 years. It is found that this methodology relaxes the over-year storage constraints and provides efficient operating policy that can be implemented over a long period of time. By using ACO technique for reservoir operation problems, some of the limitations of traditional nonlinear optimization methods are surmounted and thus the performance of the reservoir system is improved. To achieve faster optimization in water resource systems, a novel technique based on swarm intelligence, namely particle swarm optimization (PSO) has been proposed. In general, PSO has distinctly faster convergence towards global optimal solutions for numerical optimization. However, it is found that the technique has the problem of getting trapped to local optima while solving real world complex problems. To overcome such drawbacks, the standard particle swarm optimization technique has been further improved by incorporating a novel elitist-mutation (EM) mechanism into the algorithm. This strategy provides proper exploration and exploitation throughout the iterations. The improvement is demonstrated by applying it to a multi-purpose single reservoir problem and also to a multi reservoir system. The results showed robust performance of the EM-PSO approach in yielding global optimal solutions. Most of the practical problems in water resources are not only nonlinear in their formulations but are also multi-objective in nature. For multi-objective optimization, generating feasible efficient Pareto-optimal solutions is always a complicated task. In the past, many attempts with various conventional approaches were made to solve water resources problems and some of them are reported as successful. However, in using the conventional linear programming (LP) and nonlinear programming (NLP) methods, they usually involve essential approximations, especially while dealing withdiscontinuous, non-differentiable, non-convex and multi-objective functions. Most of these methods consider multiple objective functions using weighted approach or constrained approach without considering all the objectives simultaneously. Also, the conventional approaches use a point-by-point search approach, in which the outcome of these methods is a single optimal solution. So they may require a large number of simulation runs to arrive at a good Pareto optimal front. One of the major goals in multi-objective optimization is to find a set of well distributed optimal solutions along the true Pareto optimal front. The classical optimization methods often fail to attain a good and true Pareto optimal front due to accretion of the above problems. To overcome such drawbacks of the classical methods, there has recently been an increasing interest in evolutionary computation methods for solving real world multi-objective problems. In this thesis, some novel approaches for multi-objective optimization are developed based on swarm intelligence and evolutionary computation principles. By incorporating Pareto optimality principles into particle swarm optimization algorithm, a novel approach for multi-objective optimization has been developed. To obtain efficient Pareto-frontiers, along with proper selection scheme and diversity preserving mechanisms, an efficient elitist mutation strategy is proposed. The developed elitist-mutated multi-objective particle swarm optimization (EM-MOPSO) technique is tested for various numerical test problems and engineering design problems. It is found that the EM-MOPSO algorithm resulting in improved performance over a state-of-the-art multi-objective evolutionary algorithm (MOEA). The utility of EM-MOPSO technique for water resources optimization is demonstrated through application to a case study, to obtain optimal trade-off solutions to a reservoir operation problem. Through multi-objective analysis for reservoir operation policies, it is found that the technique can offer wide range of efficient alternatives along with flexibility to the decision maker. In general, most of the water resources optimization problems involve interdependence relations among the various decision variables. By using differential evolution (DE) scheme, which has a proven ability of effective handling of this kind of interdependence relationships, an efficient multi-objective solver, namely multi-objective differential evolution (MODE) is proposed. The single objective differential evolution algorithm is extended to multi-objective optimization by integrating various operators like, Pareto-optimality, non-dominated sorting, an efficient selection strategy, crowding distance operator for maintaining diversity, an external elite archive for storing non- dominated solutions and an effective constraint handling scheme. First, different variations of DE approaches for multi-objective optimization are evaluated through several benchmark test problems for numerical optimization. The developed MODE algorithm showed improved performance over a standard MOEA, namely non-dominated sorting genetic algorithm–II (NSGA-II). Then MODE is applied to a case study of Hirakud reservoir operation problem to derive operational tradeoffs in the reservoir system optimization. It is found that MODE is achieving robust performance in evaluation for the water resources problem, and that the interdependence relationships among the decision variables can be effectively modeled using differential evolution operators. For optimal utilization of scarce water resources, an integrated operational model is developed for reservoir operation for irrigation of multiple crops. The model integrates the dynamics associated with the water released from a reservoir to the actual water utilized by the crops at farm level. It also takes into account the non-linear relationship of root growth, soil heterogeneity, soil moisture dynamics for multiple crops and yield response to water deficit at various growth stages of the crops. Two types of objective functions are evaluated for the model by applying to a case study of Malaprabha reservoir project. It is found that both the cropping area and economic benefits from the crops need to be accounted for in the objective function. In this connection, a multi-objective frame work is developed and solved using the MODE algorithm to derive simultaneous policies for irrigation cropping pattern and reservoir operation. It is found that the proposed frame work can provide effective and flexible policies for decision maker aiming at maximization of overall benefits from the irrigation system. For efficient management of water resources projects, there is always a great necessity to accurately forecast the hydrologic variables. To handle uncertain behavior of hydrologic variables, soft computing based artificial neural networks (ANNs) and fuzzy inference system (FIS) models are proposed for reservoir inflow forecasting. The forecast models are developed using large scale climate inputs like indices of El-Nino Southern Oscialltion (ENSO), past information on rainfall in the catchment area and inflows into the reservoir. In this purpose, back propagation neural network (BPNN), hybrid particle swarm optimization trained neural network (PSONN) and adaptive network fuzzy inference system (ANFIS) models have been developed. The developed models are applied for forecasting inflows into the Malaprabha reservoir. The performances of these models are evaluated using standard performance measures and it is found that the hybrid PSONN model is performing better than BPNN and ANFIS models. Finally by adopting PSONN model for inflow forecasting and EMPSO technique for solving the reservoir operation model, the practical utility of the different models developed in the thesis are demonstrated through application to a real time reservoir operation problem. The developed methodologies can certainly help in better planning and operation of the scarce water resources.

The primary goal of artificial intelligence research is to develop a machine capable of learning to solve disparate real-world tasks autonomously, without relying on specialized problem-specific inputs. This dissertation suggests that such machines are realistic: If No Free Lunch theorems were to apply to all real-world problems, then the world would be utterly unpredictable. In response, the dissertation proposes the information-maximization principle, which claims that the optimal optimization methods make the best use of the information available to them. This principle results in a new algorithm, evolutionary annealing, which is shown to perform well especially in challenging problems with irregular structure. / text

Αντικείµενο της διατριβής είναι η ανάπτυξη νέων µεθοδολογιών εκµάθησης και σύγκλισης των Ασαφών Γνωστικών ∆ικτύων που προτείνονται για τη βελτίωση και προσαρµογή της συµπεριφοράς τους, καθώς και για την αύξηση της απόδοσής τους, αναδεικνύοντάς τα σε αποτελεσµατικά δυναµικά συστήµατα µοντελοποίησης. Τα νέα βελτιωµένα Ασαφή Γνωστικά ∆ίκτυα, µέσω της εκµάθησης και προσαρµογής των βαρών τους, έχουν χρησιµοποιηθεί στην ιατρική σε θέµατα διάγνωσης και υποστήριξης στη λήψη απόφασης, καθώς και σε µοντέλα βιοµηχανικών συστηµάτων που αφορούν τον έλεγχο διαδικασιών, µε πολύ ικανοποιητικά αποτελέσµατα. Στη διατριβή αυτή παρουσιάζονται, αξιολογούνται και εφαρµόζονται δύο νέοι αλγόριθµοι εκµάθησης χωρίς επίβλεψη των Ασαφών Γνωστικών ∆ικτύων, οι αλγόριθµοι Active Hebbian Learning (AHL) και Nonlinear Hebbian Learning (NHL), βασισµένοι στον κλασσικό αλγόριθµό εκµάθησης χωρίς επίβλεψη τύπου Hebb των νευρωνικών δικτύων, καθώς και µια νέα προσέγγιση εκµάθησης των Ασαφών Γνωστικών ∆ικτύων βασισµένη στους εξελικτικούς αλγορίθµους και πιο συγκεκριµένα στον αλγόριθµο Βελτιστοποίησης µε Σµήνος Σωµατιδίων και στον ∆ιαφοροεξελικτικό αλγόριθµο. Οι προτεινόµενοι αλγόριθµοι AHL και NHL στηρίζουν νέες µεθοδολογίες εκµάθησης για τα ΑΓ∆ που βελτιώνουν τη λειτουργία, και την αξιοπιστία τους, και που παρέχουν στους εµπειρογνώµονες του εκάστοτε προβλήµατος που αναπτύσσουν το ΑΓ∆, την εκµάθηση των παραµέτρων για τη ρύθµιση των αιτιατών διασυνδέσεων µεταξύ των κόµβων. Αυτοί οι τύποι εκµάθησης που συνοδεύονται από την σωστή γνώση του εκάστοτε προβλήµατος-συστήµατος, συµβάλλουν στην αύξηση της απόδοσης των ΑΓ∆ και διευρύνουν τη χρήση τους. Επιπρόσθετα µε τους αλγορίθµους εκµάθησης χωρίς επίβλεψη τύπου Hebb για τα ΑΓ∆, αναπτύσσονται και προτείνονται νέες τεχνικές εκµάθησης των ΑΓ∆ βασισµένες στους εξελικτικούς αλγορίθµους. Πιο συγκεκριµένα, προτείνεται µια νέα µεθοδολογία για την εφαρµογή του εξελικτικού αλγορίθµου Βελτιστοποίησης µε Σµήνος Σωµατιδίων στην εκµάθηση των Ασαφών Γνωστικών ∆ικτύων και πιο συγκεκριµένα στον καθορισµό των βέλτιστων περιοχών τιµών των βαρών των Ασαφών Γνωστικών ∆ικτύων. Με τη µεθοδο αυτή λαµβάνεται υπόψη η γνώση των εµπειρογνωµόνων για τον σχεδιασµό του µοντέλου µε τη µορφή περιορισµών στους κόµβους που µας ενδιαφέρουν οι τιµές των καταστάσεών τους, που έχουν οριστοί ως κόµβοι έξοδοι του συστήµατος, και για τα βάρη λαµβάνονται υπόψη οι περιοχές των ασαφών συνόλων που έχουν συµφωνήσει όλοι οι εµπειρογνώµονες. Έτσι θέτoντας περιορισµούς σε όλα τα βάρη και στους κόµβους εξόδου και καθορίζοντας µια κατάλληλη αντικειµενική συνάρτηση για το εκάστοτε πρόβληµα, προκύπτουν κατάλληλοι πίνακες βαρών (appropriate weight matrices) που µπορούν να οδηγήσουν το σύστηµα σε επιθυµητές περιοχές λειτουργίας και ταυτόχρονα να ικανοποιούν τις ειδικές συνθήκες- περιορισµούς του προβλήµατος. Οι δύο νέες µέθοδοι εκµάθησης χωρίς επίβλεψη που έχουν προταθεί για τα ΑΓ∆ χρησιµοποιούνται και εφαρµόζονται µε επιτυχία σε δυο πολύπλοκα προβλήµατα από το χώρο της ιατρικής, στο πρόβληµα λήψης απόφασης στην ακτινοθεραπεία και στο πρόβληµα κατηγοριοποίησης των καρκινικών όγκων της ουροδόχου κύστης σε πραγµατικές κλινικές περιπτώσεις. Επίσης όλοι οι προτεινόµενοι αλγόριθµοι εφαρµόζονται σε µοντέλα βιοµηχανικών συστηµάτων που αφορούν τον έλεγχο διαδικασιών µε πολύ ικανοποιητικά αποτελέσµατα. Οι αλγόριθµοι αυτοί, όπως προκύπτει από την εφαρµογή τους σε συγκεκριµένα προβλήµατα, βελτιώνουν το µοντέλο του ΑΓ∆, συµβάλλουν σε ευφυέστερα συστήµατα και διευρύνουν τη δυνατότητα εφαρµογής τους σε πραγµατικά και πολύπλοκα προβλήµατα. Η κύρια συνεισφορά αυτής της διατριβής είναι η ανάπτυξη νέων µεθοδολογιών εκµάθησης και σύγκλισης των Ασαφών Γνωστικών ∆ικτύων προτείνοντας δυο νέους αλγορίθµους µη επιβλεπόµενης µάθησης τύπου Hebb, τον αλγόριθµο Active Hebbian Learning και τον αλγόριθµο Nonlinear Hebbian Learning για την προσαρµογή των βαρών των διασυνδέσεων µεταξύ των κόµβων των Ασαφών Γνωστικών ∆ικτύων, καθώς και εξελικτικούς αλγορίθµους βελτιστοποιώντας συγκεκριµένες αντικειµενικές συναρτήσεις για κάθε εξεταζόµενο πρόβληµα. Τα νέα βελτιωµένα Ασαφή Γνωστικά ∆ίκτυα µέσω των αλγορίθµων προσαρµογής των βαρών τους έχουν χρησιµοποιηθεί για την ανάπτυξη ενός ∆ιεπίπεδου Ιεραρχικού Συστήµατος για την υποστήριξη λήψης απόφασης στην ακτινοθεραπεία, για την ανάπτυξη ενός διαγνωστικού εργαλείου για την κατηγοριοποίηση του βαθµού κακοήθειας των καρκινικών όγκων της ουροδόχου κύστης, καθώς και για την επίλυση βιοµηχανικών προβληµάτων για τον έλεγχο διαδικασιών. / The main contribution of this Dissertation is the development of new learning and convergence methodologies for Fuzzy Cognitive Maps that are proposed for the improvement and adaptation of their behaviour, as well as for the increase of their performance, electing them in effective dynamic systems of modelling. The new improved Fuzzy Cognitive Maps, via the learning and adaptation of their weights, have been used in medicine for diagnosis and decision-making, as well as to alleviate the problem of the potential uncontrollable convergence to undesired states in models of industrial process control systems, with very satisfactory results. In this Dissertation are presented, validated and implemented two new learning algorithms without supervision for Fuzzy Cognitive Maps, the algorithms Active Hebbian Learning (AHL) and Nonlinear Hebbian Learning (NHL), based on the classic unsupervised Hebb-type learning algorithm of neural networks, as well as a new approach of learning for Fuzzy Cognitive Maps based on the evolutionary algorithms and more specifically on the algorithm of Particles Swarm Optimization and on the Differential Evolution algorithm. The proposed algorithms AHL and NHL support new learning methodologies for FCMs that improve their operation, efficiency and reliability, and that provide in the experts of each problem that develop the FCM, the learning of parameters for the regulation (fine-tuning) of cause-effect relationships (weights) between the concepts. These types of learning that are accompanied with the right knowledge of each problem-system, contribute in the increase of performance of FCMs and extend their use. Additionally to the unsupervised learning algorithms of Hebb-type for the FCMs, are developed and proposed new learning techniques of FCMs based on the evolutionary algorithms. More specifically, it is proposed a new learning methodology for the application of evolutionary algorithm of Particle Swarm Optimisation in the adaptation of FCMs and more concretely in the determination of the optimal regions of weight values of FCMs. With this method it is taken into consideration the experts’ knowledge for the modelling with the form of restrictions in the concepts that interest us their values, and are defined as output concepts, and for weights are received the arithmetic values of the fuzzy regions that have agreed all the experts. Thus considering restrictions in all weights and in the output concepts and determining a suitable objective function for each problem, result appropriate weight matrices that can lead the system to desirable regions of operation and simultaneously satisfy specific conditions of problem. The first two proposed methods of unsupervised learning that have been suggested for the FCMs are used and applied with success in two complicated problems in medicine, in the problem of decision-making in the radiotherapy process and in the problem of tumor characterization for urinary bladder in real clinical cases. Also all the proposed algorithms are applied in models of industrial systems that concern the control of processes with very satisfactory results. These algorithms, as it results from their application in concrete problems, improve the model of FCMs, they contribute in more intelligent systems and they extend their possibility of application in real and complex problems. The main contribution of the present Dissertation is to develop new learning and convergence methodologies for Fuzzy Cognitive Maps proposing two new unsupervised learning algorithms, the algorithm Active Hebbian Learning and the algorithm Nonlinear Hebbian Learning for the adaptation of weights of the interconnections between the concepts of Fuzzy Cognitive Maps, as well as Evolutionary Algorithms optimizing concrete objective functions for each examined problem. New improved Fuzzy Cognitive Maps via the algorithms of weight adaptation have been used for the development of an Integrated Two-level hierarchical System for the support of decision-making in the radiotherapy, for the development of a new diagnostic tool for tumour characterization of urinary bladder, as well as for the solution of industrial process control problems.

Σε αυτή την εργασία, μελετάμε την κλάση των Υψηλής Τάξης Νευρωνικών Δικτύων και ειδικότερα των Πι—Σίγμα Νευρωνικών Δικτύων. Η απόδοση των Πι—Σίγμα Νευρωνικών Δικτύων αξιολογείται με την εφαρμογή τους σε διάφορα πολύ γνωστά χαρακτηριστικά προβλήματα εκπαίδευσης νευρωνικών δικτύων. Στα πειράματα που πραγματοποιήθηκαν, για την εκπαίδευση των Πι—Σίγμα Νευρωνικών Δικτύων υλοποιήθηκαν και εφαρμόστηκαν Σειριακοί και Παράλληλοι/Κατανεμημένοι Εξελικτικοί Αλγόριθμοι. Πιο συγκεκριμένα χρησιμοποιήθηκαν οι σειριακές καθώς και οι παράλληλες/κατανεμημένες εκδοχές των Διαφοροεξελικτικών Αλγόριθμων. Η προτεινόμενη μεθοδολογία βασίστηκε σε αυτές τις εκδοχές και εφαρμόστηκε για την εκπαίδευση των Πι—Σίγμα δικτύων χρησιμοποιώντας συναρτήσεις ενεργοποίησης «κατώφλια». Επιπρόσθετα, όλα τα βάρη και οι μεροληψίες των δικτύων περιορίστηκαν σε ένα μικρό εύρος ακέραιων αριθμών, στο διάστημα [-32, 32]. Συνεπώς, τα εκπαιδευμένα Πι—Σίγμα νευρωνικά δίκτυα μπορούν να αναπαρασταθούν με ακεραίους των 6-bits. Αυτής της μορφής τα δίκτυα είναι πιο κατάλληλα για την εφαρμογή τους σε «υλικό» (hardware), από νευρωνικά δίκτυα με πραγματικά βάρη. Τα πειραματικά αποτελέσματα μας δείχνουν ότι η διαδικασία εκπαίδευσης είναι γρήγορη, σταθερή και αξιόπιστη. Ακόμα η εφαρμογή των παράλληλων/κατανεμημένων Εξελικτικών Αλγορίθμων για την εκπαίδευση των Πι—Σίγμα δικτύων μας επιδεικνύει αρκετά καλές ικανότητες γενίκευσης των εκπαιδευμένων δικτύων καθώς και προσφέρει επιτάχυνση στην διαδικασία εκπαίδευσης τους. / In this contribution, we study the class of Higher-Order Neural Networks and especially the Pi-Sigma Networks. The performance of Pi-Sigma Networks is evaluated through several well known neural network training benchmarks. In the experiments reported here, Evolutionary Algorithms and Parallel/Distributed Evolutionary Algorithms are implemented for Pi-Sigma neural networks training. More specifically the serial as well as a parallel/distributed version of the Differential Evolution have been employed. The proposed approach is applied to train Pi-Sigma networks using threshold activation functions. Moreover, the weights and biases were confined to a narrow band of integers, constrained in the range [-32, 32]. Thus the trained Pi-Sigma neural networks can be represented by just 6 bits. Such networks are better suited for hardware implementation than the real weight ones. Experimental results suggest that this training process is fast, stable and reliable and the trained Pi-Sigma networks, with both serial and parallel/distributed algorithms, exhibited good generalization capabilities. Furthermore, the usage of a distributed version of the Differential Evolution, has demonstrated a speedup of the training process.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / A identificação de danos estruturais é uma questão de fundamental importância na engenharia, visto que uma estrutura está sujeita a processos de deterioração e a ocorrência de danos durante a sua vida útil. A presença de danos compromete o desempenho e a integridade estrutural, podendo colocar vidas humanas em risco e resultam em perdas econômicas consideráveis. Técnicas de identificação de danos estruturais e monitoramento de estruturas fundamentadas no ajuste de um Modelo de Elementos Finitos (MEF) são constantes na literatura especializada. No entanto, a obtenção de um problema geralmente mal posto e o elevado custo computacional, inerente a essas técnicas, limitam ou até mesmo inviabilizam a sua aplicabilidade em estruturas que demandam um modelo de ordem elevada. Para contornar essas dificuldades, na formulação do problema de identificação de danos, pode-se utilizar o Modelo de Superfície de Reposta (MSR) em substituição a um MEF da estrutura. No presente trabalho, a identificação de danos estruturais considera o ajuste de um MSR da estrutura, objetivando-se a minimização de uma função de erro definida a partir das frequências naturais experimentais e das correspondentes frequências previstas pelo MSR. Estuda-se o problema de identificação de danos estruturais em uma viga de Euler-Bernoulli simplesmente apoiada, considerando as frequências naturais na formulação do problema inverso. O comportamento de uma viga de Euler-Bernoulli simplesmente apoiada na presença de danos é analisado, com intuito de se verificar as regiões onde a identificação dos mesmos pode apresentar maior dificuldade. No processo de identificação de danos, do presente trabalho, são avaliados os tipos de superfícies de resposta, após uma escolha apropriada do tipo de superfície de resposta a ser utilizado, determina-se a superfície de resposta considerando os dados experimentais selecionados a partir do projeto ótimo de experimentos. A utilização do método Evolução Diferencial (ED) no problema inverso de identificação de danos é considerado inerente aos resultados numéricos obtidos, a estratégia adotada mostrou-se capaz de localizar e quantificar os danos com elevada acurácia, mostrando a potencialidade do modelo de identificação de danos proposto. / The identification of structural damage is an issue of fundamental importance in engineering, since a structure is subject to deterioration processes and to the occurrence of damage throughout its useful lifetime. The presence of damage compromises the performance and structural integrity, may put human lives at risk and may result in considerable economic losses. Damage identification and structural health monitoring techniques built on Finite Element Model (FEM) updating are constant in the specialized literature. However, the problem generally rank deficient and the high computational cost, inherent to these techniques, limit or even render their applicability in structures that require a high order model. To circumvent these difficulties, in the formulation of the damage identification problem, one may use a Response Surface Model (RSM) in place of a FEM of the structure. In the present work, the identification of structural damage considers the update of a RSM of the structure, with the aim at minimizing an error function defined from the experimental natural frequencies and the corresponding natural frequencies prescribed by a RSM. The problem of structural damage identification in a simply supported Euler-Bernoulli beam is studied, taking into account the natural frequencies in the inverse problem formulation. The behavior of a simply supported Euler-Bernoulli beam, in the presence of damage, is analyzed, in order to verify the identification of regions where the damage identification may present greater difficulties. In the damage identification process, in the present work, after a suitable choice of the type of the response surface model, the surface model is derived considering the experimental data selected from an optimal design of experiments. The use of the Differential Evolution (DE) method in the inverse problem of damage identification is considered. Considering the numerical results obtained, the strategy adopted proved to be able to locate and quantify the damage with high accuracy, showing the capability of the proposed damage identification model.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / A identificação de danos estruturais é uma questão de fundamental importância na engenharia, visto que uma estrutura está sujeita a processos de deterioração e a ocorrência de danos durante a sua vida útil. A presença de danos compromete o desempenho e a integridade estrutural, podendo colocar vidas humanas em risco e resultam em perdas econômicas consideráveis. Técnicas de identificação de danos estruturais e monitoramento de estruturas fundamentadas no ajuste de um Modelo de Elementos Finitos (MEF) são constantes na literatura especializada. No entanto, a obtenção de um problema geralmente mal posto e o elevado custo computacional, inerente a essas técnicas, limitam ou até mesmo inviabilizam a sua aplicabilidade em estruturas que demandam um modelo de ordem elevada. Para contornar essas dificuldades, na formulação do problema de identificação de danos, pode-se utilizar o Modelo de Superfície de Reposta (MSR) em substituição a um MEF da estrutura. No presente trabalho, a identificação de danos estruturais considera o ajuste de um MSR da estrutura, objetivando-se a minimização de uma função de erro definida a partir das frequências naturais experimentais e das correspondentes frequências previstas pelo MSR. Estuda-se o problema de identificação de danos estruturais em uma viga de Euler-Bernoulli simplesmente apoiada, considerando as frequências naturais na formulação do problema inverso. O comportamento de uma viga de Euler-Bernoulli simplesmente apoiada na presença de danos é analisado, com intuito de se verificar as regiões onde a identificação dos mesmos pode apresentar maior dificuldade. No processo de identificação de danos, do presente trabalho, são avaliados os tipos de superfícies de resposta, após uma escolha apropriada do tipo de superfície de resposta a ser utilizado, determina-se a superfície de resposta considerando os dados experimentais selecionados a partir do projeto ótimo de experimentos. A utilização do método Evolução Diferencial (ED) no problema inverso de identificação de danos é considerado inerente aos resultados numéricos obtidos, a estratégia adotada mostrou-se capaz de localizar e quantificar os danos com elevada acurácia, mostrando a potencialidade do modelo de identificação de danos proposto. / The identification of structural damage is an issue of fundamental importance in engineering, since a structure is subject to deterioration processes and to the occurrence of damage throughout its useful lifetime. The presence of damage compromises the performance and structural integrity, may put human lives at risk and may result in considerable economic losses. Damage identification and structural health monitoring techniques built on Finite Element Model (FEM) updating are constant in the specialized literature. However, the problem generally rank deficient and the high computational cost, inherent to these techniques, limit or even render their applicability in structures that require a high order model. To circumvent these difficulties, in the formulation of the damage identification problem, one may use a Response Surface Model (RSM) in place of a FEM of the structure. In the present work, the identification of structural damage considers the update of a RSM of the structure, with the aim at minimizing an error function defined from the experimental natural frequencies and the corresponding natural frequencies prescribed by a RSM. The problem of structural damage identification in a simply supported Euler-Bernoulli beam is studied, taking into account the natural frequencies in the inverse problem formulation. The behavior of a simply supported Euler-Bernoulli beam, in the presence of damage, is analyzed, in order to verify the identification of regions where the damage identification may present greater difficulties. In the damage identification process, in the present work, after a suitable choice of the type of the response surface model, the surface model is derived considering the experimental data selected from an optimal design of experiments. The use of the Differential Evolution (DE) method in the inverse problem of damage identification is considered. Considering the numerical results obtained, the strategy adopted proved to be able to locate and quantify the damage with high accuracy, showing the capability of the proposed damage identification model.

Cette thèse porte sur la caractérisation de structures mécaniques, au travers de leurs paramètres structuraux, à partir d'observations perturbées par des bruits de mesure, supposés additifs blancs gaussiens et centrés. Pour cela, nous proposons d'utiliser des modèles à temps discret à parties linéaire et non-linéaire séparables. La première permet de retrouver les paramètres recherchés tandis que la seconde renseigne sur la non-linéarité présente. Dans le cadre d'une modélisation non-récursive par des séries de Volterra, nous présentons une approche à erreurs-dans-les-variables lorsque les variances des bruits ne sont pas connues ainsi qu'un algorithme adaptatif du type LMS nécessitant la connaissance de la variance du bruit d'entrée. Dans le cadre d'une modélisation par un modèle récursif polynomial, nous proposons deux méthodes à partir d'algorithmes évolutionnaires. La première inclut un protocole d'arrêt tenant compte de la variance du bruit de sortie. Dans la seconde, les fonctions fitness sont fondées sur des fonctions de corrélation dans lesquelles l'influence du bruit est supprimée ou compensée. / This PhD deals with the caracterisation of mechanical structures, by its structural parameters, when only noisy observations disturbed by additive measurement noises, assumed to be zero-mean white and Gaussian, are available. For this purpose, we suggest using discrete-time models with distinct linear and nonlinear parts. The first one allows the structural parameters to be retrieved whereas the second one gives information on the nonlinearity. When dealing with non-recursive Volterra series, we propose an errors-in-variables (EIV) method to jointly estimate the noise variances and the Volterra kernels. We also suggest a modified unbiased LMS algorithm to estimate the model parameters provided that the input-noise variance is known. When dealing with recursive polynomial model, we propose two methods using evolutionary algorithms. The first includes a stop protocol that takes into account the output-noise variance. In the second one, the fitness functions are based on correlation criteria in which the noise influence is removed or compensated.

In the modern generation, Electric Power has become one of the fundamental needs for humans to survive. This is due to the dependence of continuous availability of power. However, for electric power to be available to the society, it has to pass through a number of complex stages. Through each stage power quality problems are experienced on the grid. Under-voltages and over-voltages are the most common electric problems experienced on the grid, causing industries and business firms losses of Billions of dollars each year. Researchers from different regions are attracted by an idea that will overcome all the electrical issues experienced in the traditional grid using Artificial Intelligence (AI). The idea is said to provide electric power that is sustainable, economical, reliable and efficient to the society based on Evolutionary Algorithms (EAs). The idea is Smart Grid. The research focused on Power Quality Optimization in Smart Grid based on improved Differential Evolution (DE), with the objective functions to minimize voltage swells, counterbalance voltage sags and eliminate voltage surges or spikes, while maximizing the power quality. During Differential Evolution improvement research, elimination of stagnation, better and fast convergence speed were achieved based on modification of DE’s mutation schemes and parameter control selection. DE/Modi/2 and DE/Modi/3 modified mutation schemes proved to be the excellent improvement for DE algorithm by achieving excellent optimization results with regards to convergence speed and elimination of stagnation during simulations. The improved DE was used to optimize Power Quality in smart grid in combination with the reconfigured and modified Dynamic Voltage Restorer (DVR). Excellent convergence results of voltage swells and voltage sags minimization were achieved based on application of multi-objective parallel operation strategy during simulations. MATLAB was used to model the proposed solution and experimental simulations. / Electrical and Mining Engineering / M. Tech. (Electrical Engineering)

The theoretical part of this work deals with evolutionary algorithms (EA), neural networks (NN) and their synthesis in the form of neuroevolution. From a practical point of view, the aim of the work is to show the application of neuroevolution on two different tasks. The first task is the evolutionary design of the convolutional neural network (CNN) architecture that would be able to classify handwritten digits (from the MNIST dataset) with a high accurancy. The second task is the evolutionary optimization of neurocontroller for a simulated Falcon 9 rocket landing. Both tasks are computationally demanding and therefore have been solved on a supercomputer. As a part of the first task, it was possible to design such architectures which, when properly trained, achieve an accuracy of 99.49%. It turned out that it is possible to automate the design of high-quality architectures with the use of neuroevolution. Within the second task, the neuro-controller weights have been optimized so that, for defined initial conditions, the model of the Falcon booster can successfully land. Neuroevolution succeeded in both tasks.