A detailed and comprehensive description of the state of the art in the field of flood forecasting opens this work. Advantages and shortcomings of currently available methods are identified and discussed. Amongst others, one important aspect considers the most exigent weak point of today’s forecasting systems: The representation of all the fundamentally different event specific patterns of flood formation with one single set of model parameters. The study exemplarily proposes an alternative for overcoming this restriction by taking into account the different process characteristics of flood events via a dynamic parameterisation strategy. Other fundamental shortcomings in current approaches especially restrict the potential for real time flash flood forecasting, namely the considerable computational requirements together with the rather cumbersome operation of reliable physically based hydrologic models. The new PAI-OFF methodology (Process Modelling and Artificial Intelligence for Online Flood Forecasting) considers these problems and offers a way out of the general dilemma. It combines the reliability and predictive power of physically based, hydrologic models with the operational advantages of artificial intelligence. These operational advantages feature extremely low computation times, absolute robustness and straightforward operation. Such qualities easily allow for predicting flash floods in small catchments taking into account precipitation forecasts, whilst extremely basic computational requirements open the way for online Monte Carlo analysis of the forecast uncertainty. The study encompasses a detailed analysis of hydrological modeling and a problem specific artificial intelligence approach in the form of artificial neural networks, which build the PAI-OFF methodology. Herein, the synthesis of process modelling and artificial neural networks is achieved by a special training procedure. It optimizes the network according to the patterns of possible catchment reaction to rainstorms. This information is provided by means of a physically based catchment model, thus freeing the artificial neural network from its constriction to the range of observed data – the classical reason for unsatisfactory predictive power of netbased approaches. Instead, the PAI-OFF-net learns to portray the dominant process controls of flood formation in the considered catchment, allowing for a reliable predictive performance. The work ends with an exemplary forecasting of the 2002 flood in a 1700 km² East German watershed.

Enantioselective indicator displacement assays (eIDAs), was transitioned to a high-throughput screening protocols, for the rapid determination of concentration and enantioselectivity (ee) of chiral diols and α-hydroxycarboxylic acid. To improve the design of our previously established receptor based on o-(N,N-dialkylaminomethyl)arylboronate scaffolds for eIDAs. The rigidity of the receptor, which pertinent from the formation of an intramolecular N-B dative bond was investigated. o-(Pyrrolidinylmethyl)phenylboronic acid its complexes with bifunctional substrates such as catechol, [alpha]-hydroxyisobutyric acid, and hydrobenzoin was studied in detail by x-ray crystallography and ¹¹B NMR. Our structural study predicts that the formation of an N-B dative bond, and/or solvolysis to afford a tetrahedral boronate anion, depends on the solvent and the complexing substrate present. To simplify the operation of eIDAs, we introduced an analytical method, which utilize a dual-chamber quartz cuvette, which reduces the number of spectroscopic measurements from two to one and introduced artificial neural networks (ANNs) which simplifies data analysis. In a second example a high-throughtput screening protocol for hydrobenzoin was developed. The method involves the sequential utilization of what we define herein as screening, training, and analysis plates. Several enantioselective boronic-acid based receptors were screened using 96-well plates, both for their ability to discriminate the enantiomers of hydrobenzoin and to find their optimal pairing with indicators resulting in the largest optical responses. The best receptor/indicator combination was then used to train an ANN to determine concentration and ee. To prove the practicality of the developed protocol, analysis plates were created containing true unknown samples of hydrobenzoin generated by established Sharpless asymmetric dihydroxylation reactions, and the best ligand was correctly identified. The system was extended to pattern recognition for the rapid determination of identity, concentration, and ee of chiral vicinal diols. A diverse enantioselective sensor array was generated with three chiral boronic acid receptors and pH indicators. The optical response produced by the sensor array, was analyzed by two pattern recognition algorithms: principal component analysis (PCA) and ANNs. The PCA plot demonstrated good chemoselective and enantioselective separation of the analytes, and ANNs was used to accurately determine the concentration and ee of five unknown samples. / text

Η παρούσα διπλωματική εργασία πραγματεύεται την χρήση τεχνητών νευρωνικών δικτύων για την δημιουργία ενός αξιόπιστου μοντέλου πρόβλεψης της μέγιστης ημερήσιας συγκέντρωσης του τροποσφαιρικού όζοντος σε αστικό περιβάλλον. Η ανάλυση και η πρόβλεψη των επιπέδων της ατμοσφαιρικής ρύπανσης, είναι σημαντικά θέματα της ατμοσφαιρικής και περιβαλλοντικής έρευνας, λόγω του αντίκτυπου που έχει η ατμοσφαιρική ρύπανση στην υγεία και την ποιότητα ζωής του ανθρώπου. Ένας από τους πιο σημαντικούς ρύπους είναι το τροποσφαιρικό όζον (O3), και ειδικότερα το όζον του κατώτερου οριακού στρώματος της ατμόσφαιρας (επιφανειακό ή surface ozone). Επομένως, η πρόβλεψη της μέγιστης συγκέντρωσης όζοντος στις πυκνοκατοικημένες αστικές περιοχές, είναι μεγάλης σπουδαιότητας για τον έλεγχο και τη βελτίωση της ποιότητας της ατμόσφαιρας. Μέχρι σήμερα, αν και διάφορα μοντέλα πρόβλεψης όζοντος έχουν ερευνηθεί, υπάρχει ακόμα η ανάγκη για ακριβέστερα μοντέλα, έτσι ώστε να αναπτυχθούν αποτελεσματικές στρατηγικές πρόληψης και ελέγχου σε περιπτώσεις που οι οριακές τιμές όζοντος ξεπερνιούνται πέρα από κάποιο συγκεκριμένο χρονικό διάστημα (επεισόδιο ρύπανσης). Χρησιμοποιήθηκαν χρονοσειρές συγκέντρωσης όζοντος και άλλων ρύπων που οδηγούν στην δημιουργία του, από το σταθμό της “Λυκόβρυσης” και του “Αμαρουσίου”, καθώς και μετεωρολογικά δεδομένα που σχετίζονται με την δημιουργία, καταστροφή και διασπορά ή διάχυση του όζοντος. Κατά τη βελτιστοποίηση του νευρωνικού δικτύου, δόθηκε έμφαση στην κατά το δυνατόν ακριβέστερη πρόγνωση των αυξημένων τιμών της συγκέντρωσης Ο3 (επεισόδια ρύπανσης για [Ο3] >=180 μg/m3), για τις οποίες απαιτείται η λήψη εκτάκτων μέτρων. Τα Τεχνητά Νευρωνικά Δίκτυα, όπως αποδείχθηκε, αποτελούν μια πολύ καλή εναλλακτική λύση ως προς τις παραδοσιακές στατιστικές τεχνικές, καθώς για την εκπαίδευσή τους χρησιμοποιούνται διαθέσιμα στοιχεία προηγούμενων μετρήσεων αλλά και λόγω της ικανότητάς τους να χειρίζονται δεδομένα με μη γραμμικές σχέσεις μεταξύ τους. Αρχικά δοκιμάστηκαν διαφορετικές αρχιτεκτονικές νευρωνικών δικτύων επιτρέπουν την κατηγοριοποίηση των πρότυπων δεδομένων εισόδου, όπως οι χάρτες Kohonen, τα δίκτυα επιβλεπόμενης κατηγοριοποίησης LVQ και τα στοχαστικά RBF δίκτυα με σκοπό το μοίρασμα των δεδομένων εισόδου σε κατηγορίες σύμφωνα με τα επίπεδα όζοντος. Στη συνέχεια ελέγχθηκε η δυνατότητα πρόβλεψης με την χρήση πολυστρωματικών αντιληπτήρων πρόσθιας τροφοδότησης (Multilayer Feed-Forward) και εποπτευόμενης μάθησης μέσω διόρθωσης σφάλματος με ανατροφοδότηση (Back Propagation), γνωστά ως ΜultiLayer Perceptrons ή MLPs που έδωσαν και τα καλύτερα αποτελέσματα. Η απόδοση ενός μοντέλου πρόβλεψης της συγκέντρωσης του όζοντος της χαμηλής τροπόσφαιρας μπορεί να κριθεί από το ποσοστό των επεισοδίων που θα προβλεφθούν σωστά από το μοντέλο, σε αντιδιαστολή με το ποσοστό των ψεύτικων-λανθασμένων συναγερμών (δηλ. προβλέψεις επεισοδίων ρύπανσης που δεν συνέβησαν στην πραγματικότητα). Ο δείκτης απόδοσης SI, είναι ο δείκτης υψηλότερης σπουδαιότητας καθώς αντιπροσωπεύσει την απόδοση στην πρόβλεψη των επεισοδίων αλλά και στην γενικότερη συμφωνία μεταξύ των προβλεφθέντων και παρατηρηθέντων δεδομένων. Στο σταθμό της “Λυκόβρυσης” ο δείκτης απόδοσης SI κυμάνθηκε από 0,943 έως 0,831 και ο λόγος των ψεύτικων συναγερμών FA από 0 έως 0,014 ενώ στο σταθμό του “Αμαρουσίου” ο δείκτης απόδοσης SI κυμάνθηκε από 0,777 έως 0,68 και ο λόγος των ψεύτικων συναγερμών FA από 0,211 έως 0,234. / This study deals with the use of artificial neural networks for ground-level ozone modeling in the Athens area. Forecasting next day’s maximum hourly ozone concentration is an important topic of air quality research nowadays. The continuing worldwide environmental problem suggests the need for more accurate forecasting models. Development of such models is a difficult task as the meteorological variables and the photochemical reactions involved in ozone formation are complex. Meteorological variables and concentrations of ozone and ozone precursors, from two monitoring stations “Lykovrisi” and “Marousi”, are used as inputs in order to obtain the best estimate of the next day’s maximum hourly ozone concentration. The violation of the European public information threshold (ozone episode), defined by control authorities, of 180 μg/m3 is successfully predicted in most cases. Neural networks seem to be very well situated since they allow for nonlinear relations among input variables. Several architectures of Neural Networks were tested but Multi-Layer Perceptrons (MLPs) came up with the best results. At “Lykovrisi” monitoring station the Success Index (SI) that is able to represent performance in forecasting exceedances as well as the overall goodness between predicted and measured data, varied from 0.943 to 0.831 and the fraction of False Alarms (FA) that represents predicted episodes that didn’t happen to the overall number of predicted episodes, varied from 0 to 0.014. At “Marousi” monitoring station the SI varied from 0.777 to 0.68 and the fraction of False Alarms (FA) from 0.211 to 0.234.

No description available.

In this Master’s degree work artificial neural networks and back propagation learning algorithm for human faces and pattern recognition are analyzed. In the second part of work artificial neural networks and their architecture and structures models are analyzed. In the third part of article the backpropagation procedure and procedures theoretical learning principle are analyzed. In the fourth part different kinds of ANN methods and patterns extracting methods in recognition, learning and classification use were researched. In this part RGB method for patterns features extraction was described. In the fifth part the requirements specification, prototype model, use case diagram, system architecture, programs modules and objects project for software realization were created. In the same part backpropagation procedures running principle was realized. After the project part was completed, a face and patterns recognition system was created. In the sixth part the created software system was tested. According to the testing results software’s recognition rate is 82,5 % using supervised learning and 82,8 % using unsupervised learning. We found using the FAR and FRR rates the ERR rate, which was 40 %. While doing the testing with changed human characteristics, the system showed 84,6 % recognition rate. This rate shows very good work of the system by a little bit changed human characteristics. Systems realization was evaluated by users as very good one. In the seventh part software’s... [to full text]

Thesis (M.Sc. (Computer Science))--North-West University, Potchefstroom Campus, 2008.

Thesis (M.Sc. (Computer Science))--North-West University, Potchefstroom Campus, 2008.

This thesis presents a framework for an intelligent interface for the access of robotsupported remote laboratories through the Internet. The framework is composed of the student assessment and coaching system, the experimentation scenario, and the associated graphical user interface. Student assessment and coaching system is the main feature of a successful intelligent interface for use during remote experimentation with a robot-supported laboratory setup. The system has a modular structure employing artificial neural networks and a fuzzy-rule based decision process to model the student behaviour, to evaluate the performance and to coach him or her towards a better achievement of the tasks to be done during the experimentation. With an experimentation scenario designed and a graphical user interface, the system is applied to a robotic system that is connected to the Internet for the evaluation of the proposed framework. Illustrative examples for the operation of the each module in the system in the context of the application are given and sensitivity analysis of the system to the change in parameters is also done. The framework is then applied to a mobile robot control laboratory. The user interface and the experimentation scenario is developed for the application, and necessary modifications are made to the student assessment and coaching system in order to support the experiment.

Modeling of flow in porous and fractured media is a very important problem in reservoir engineering. As for numerical simulations conventional Navier-Stokes codes are applied to flow in both porous and fractured media. But they have long computation times, poor convergence and problems of numerical instabilities. Therefore, it is desired to develop another computational method that is more efficient and use simple rules to represent the flow in fractured media rather than partial differential equations. In this thesis Lattice Boltzmann Automaton Model will be used to represent the single phase fluid flow in two dimensional synthetic fractures and the simulation results obtained from this model are used to train Artificial Neural Networks. It has been found that as the mean aperture-fractal dimension ratio increases permeability increases. Moreover as the anisotropy factor increases permeability decreases with a second order polynomial relationship.

This paper inspired by simulation by Artificial Neural Networks (ANNs) applied recently for evaluation of phonics methodology to teach children "how to read". A novel approach for teaching a mathematical topic using a computer aided learning (CAL) package applied at educational field (a children classroom). Interesting practical results obtained after field application of suggested CAL package with and without associated teacher's voice. Presented study highly recommends application of a novel teaching trend based on behaviorism and individuals' learning styles. That is to improve quality of children mathematical learning performance.