Water distribution utilities are responsible for supplying both clean and safe drinking water, while under constraints of operating at an efficient and acceptable performance level. The City of Kingston, Ontario is currently experiencing elevated costs to repair its aging buried water main assets. Utilities Kingston is opting for a more efficient and practical means of forecasting pipe breaks and the application of a predictive water main break models allows Utilities Kingston to forecast future pipe failures and plan accordingly. The objective of this thesis is to develop an artificial neural network (ANN) model to forecast pipe breaks in the Kingston water distribution network. Data supplied by Utilities Kingston was used to develop the predictive ANN water main break model incorporating multiple variables including pipe age, diameter, length, and surrounding soil type. The constructed ANN model from historical break data was utilized to forecast pipe breaks for 1-year, 2-year, and 5-year planning periods. Simulated results were evaluated by statistical performance metrics, proving the overall model to be adequate for testing and forecasting. Predicted breaks were as follows, 33 breaks for 2011-2012, 22 breaks for 2012-2013 and 35 breaks for 2013-2016. Additionally, GIS plots were developed to highlight areas in need of potential rehabilitation for the distribution system. The goal of the model is to provide a practical means to assist in the management and development of Kingston’s pipe rehabilitation program, and to enable Utilities Kingston to reduce water main repair costs and to improve water quality at the customer's tap. / Thesis (Master, Civil Engineering) -- Queen's University, 2013-10-21 15:30:10.288

This thesis describes the use of a Real-Time Evolutionary Algorithm (RTEA) to optimise an Artificial Neural Network (ANN) on-line (in this context “on-line” means while it is in use). Traditionally, Evolutionary Algorithms (Genetic Algorithms, Evolutionary Strategies and Evolutionary Programming) have been used to train networks before use - that is “off-line,” as have other learning systems like Back-Propagation and Simulated Annealing. However, this means that the network cannot react to new situations (which were not in its original training set). The system outlined here uses a Simulated Legged Robot as a test-bed and allows it to adapt to a changing Fitness function. An example of this in reality would be a robot walking from a solid surface onto an unknown surface (which might be, for example, rock or sand) while optimising its controlling network in real-time, to adjust its locomotive gait, accordingly. The project initially developed a Central Pattern Generator (CPG) for a Bipedal Robot and used this to explore the basic characteristics of RTEA. The system was then developed to operate on a Quadruped Robot and a test regime set up which provided thousands of real-environment like situations to test the RTEA’s ability to control the robot. The programming for the system was done using Borland C++ Builder and no commercial simulation software was used. Through this means, the Evolutionary Operators of the RTEA were examined and their real-time performance evaluated. The results demonstrate that a RTEA can be used successfully to optimise an ANN in real-time. They also show the importance of Neural Functionality and Network Topology in such systems and new models of both neurons and networks were developed as part of the project. Finally, recommendations for a working system are given and other applications reviewed.

This thesis proposes an innovative wormhole detection scheme to detect wormhole attacks using computational intelligence and an artificial neural network (ANN). The aim of the proposed research is to develop a detection scheme that can detect wormhole attacks (In-band, out of band, hidden wormhole attack, active wormhole attack) in both uniformly and non-uniformly distributed sensor networks. Furthermore, the proposed research does not require any special hardware and causes no significant network overhead throughout the network. Most importantly, the probable location of the wormhole nodes can be tracked down by the proposed ANN-based detection scheme. We evaluate the efficacy of the proposed detection scheme in terms of detection accuracy, false positive rate, and false negative rate. The performance of the proposed model is also compared with other machine learning techniques (i.e. SVM and regularized nonlinear logistic regression (LR) based detection models) based detection schemes. The simulation results show that proposed ANN-based detection model outperforms the SVM and LR based detection schemes in terms of detection accuracy, false positive rate, and false negative rates. / February 2017

During several decades the stock market has been an area of interest forresearchers due to its complexity, noise, uncertainty and nonlinearity of thedata. Most of the studies regarding this area use a classical stochastics method,an example of this is ARIMA which is a standard approach for time seriesprediction. There is however another method for prediction of the stock marketthat is gaining traction in the recent years; Artificial Neural Network (ANN).This method has mostly been used in research on the American and Asian stockmarkets so far. Therefore, the purpose of this essay was to explore if ArtificialNeural Network could be used instead of ARIMA to predict the Swedish stockmarket (OMXS30). The study used data from the Swedish Stock Marketbetween 1991-07-09 to 2018-12-28 for the training of the ARIMA model anda forecast data that ranged between 2019-01-02 to 2019-04-26. The forecastdata of the ANN was composed of 80% of the data between 1991-07-09 to2019-04-26 and the evaluation data was composed of the remaining 20%. TheANN architecture had one input layer with chunks of 20 consecutive days asinput, followed by three Long Short-Term Memory (LSTM) hidden layers with128 neurons in each layer, followed by another hidden layer with RectifiedLinear Unit (ReLU) containing 32 neurons, followed by the output layercontaining 2 neurons with softmax activation. The results showed that theANN, with an accuracy of 0,9892, could be a successful method to forecast theSwedish stock market instead of ARIMA.

Diabetes Mellitus is a major health problem which affects about 200 million people worldwide. Diabetics require their blood glucose levels to be kept within the normal range in order to prevent diabetes-related complications from occurring. Blood glucose measurement is therefore of vital importance. The current glucose measurement techniques are, however, painful, inconvenient and episodic. This document provides an investigation into the use of near-infrared spectroscopy for continuous, non-invasive measurement of blood glucose. Artificial neural networks are used for the development of multivariate calibration models which predict glucose concentrations based on the near-infrared spectral data. Simulations have been performed which make use of simulated spectral data generated from the characteristic spectra of many of the major components of human blood. The simulations show that artificial neural networks are capable of predicting the glucose concentrations of complex aqueous solutions with clinically relevant accuracy. The effect of interference, such as temperature changes, pathlength variations, measurement noise and absorption due other analytes, has been investigated and modelled. The artificial neural network calibration models are capable of providing acceptably accurate predictions in the presence of multiple forms of interference. It was found that the performance of the measurement technique can be improved through careful selection of the optical pathlength and wavelength range for the spectroscopic measurements, and by using preprocessing techniques to reduce the effect of interference. Although the simulations suggest that near-infrared spectroscopy is a promising method of blood glucose measurement, which could greatly improve the quality of life of diabetics, many further issues must be resolved before the long-term goal of developing a continuous non-invasive home glucose monitor can be achieved.

This research provides a robust solution based on artificial intelligence (AI) techniques to overcome the challenges in Software Defined Networks (SDNs) that can jeopardise the overall performance of the network. The proposed approach, presented in the form of an intelligent agent appended to the SDN network, comprises of a new hybrid intelligent mechanism that optimises the performance of SDN based on heuristic optimisation methods under an Artificial Neural Network (ANN) paradigm. Evolutionary optimisation techniques, including Particle Swarm Optimisation (PSO) and Genetic Algorithms (GAs) are deployed to find the best set of inputs that give the maximum performance of an SDN-based network. The ANN model is trained and applied as a predictor of SDN behaviour according to effective traffic parameters. The parameters that were used in this study include round-trip time and throughput, which were obtained from the flow table rules of each switch. A POX controller and OpenFlow switches, which characterise the behaviour of an SDN, have been modelled with three different topologies. Generalisation of the prediction model has been tested with new raw data that were unseen in the training stage. The simulation results show a reasonably good performance of the network in terms of obtaining a Mean Square Error (MSE) that is less than 10−6 [superscript]. Following the attainment of the predicted ANN model, utilisation with PSO and GA optimisers was conducted to achieve the best performance of the SDN-based network. The PSO approach combined with the predicted SDN model was identified as being comparatively better than the GA approach in terms of their performance indices and computational efficiency. Overall, this research demonstrates that building an intelligent agent will enhance the overall performance of the SDN network. Three different SDN topologies have been implemented to study the impact of the proposed approach with the findings demonstrating a reduction in the packets dropped ratio (PDR) by 28-31%. Moreover, the packets sent to the SDN controller were also reduced by 35-36%, depending on the generated traffic. The developed approach minimised the round-trip time (RTT) by 23% and enhanced the throughput by 10%. Finally, in the event where SDN controller fails, the optimised intelligent agent can immediately take over and control of the entire network.

This study is concerned with the analyses of modern electric power-grids designed to support large supply-demand considerations in metro areas of large cities. Hence proposed are methods to determine optimal performance of the associated distribution networks vis-á-vis power availability from multiple resources (such as hydroelectric, thermal, wind-mill, solar-cell etc.) and varying load-demands posed by distinct set of consumers of domestic, industrial and commercial sectors. Hence, developing the analytics on optimal power-distribution across pertinent power-grids are verified with the models proposed. Forecast algorithms and computational outcomes on supply-demand performance are indicated and illustratively explained using real-world data sets. This study on electric utility takes duly into considerations of both deterministic (technological factors) as well as stochastic variables associated with the available resource-capacity and demand-profile details. Thus, towards forecasting exercise as above, a representative load-curve (RLC) is defined; and, it is optimally determined using an Artificial Neural Network (ANN) method using the data availed on supply-demand characteristics of a practical power-grid. This RLC is subsequently considered as an input parametric profile on tariff policies associated with electric power product-cost. This research further focuses on developing an optimal/suboptimal electric-power distribution scheme across power-grids deployed between multiple resources and different sets of user demands. Again, the optimal/suboptimal decisions are enabled using ANN-based simulations performed on load sharing details. The underlying supply-demand forecasting on distribution service profile is essential to support predictive designs on the amount of power required (or to be generated from single and/or multiple resources) versus distributable shares to different consumers demanding distinct loads. Another topic addressed refers to a business model on a cost reflective tariff levied in an electric power service in terms of the associated hedonic heuristics of customers versus service products offered by the utility operators. This model is based on hedonic considerations and technoeconomic heuristics of incumbent systems In the ANN simulations as above, bootstrapping technique is adopted to generate pseudo-replicates of the available data set and they are used to train the ANN net towards convergence. A traditional, multilayer ANN architecture (implemented with feed-forward and backpropagation techniques) is designed and modified to support a fast convergence algorithm, used for forecasting and in load-sharing computations. Underlying simulations are carried out using case-study details on electric utility gathered from the literature. In all, ANN-based prediction of a representative load-curve to assess power-consumption and tariff details in electrical power systems supporting a smart-grid, analysis of load-sharing and distribution of electric power on smart grids using an ANN and evaluation of electric power system infrastructure in terms of tariff worthiness deduced via hedonic heuristics, constitute the major thematic efforts addressed in this research study. / Includes bibliography. / Dissertation (Ph.D.)--Florida Atlantic University, 2019. / FAU Electronic Theses and Dissertations Collection

Last few decades have witnessed the use of artificial neural networks (ANN) in many real-world applications and have offered an attractive paradigm for a broad range of adaptive complex systems. In recent years ANN have enjoyed a great deal of success and have proven useful in wide variety pattern recognition or feature extraction tasks. Examples include optical character recognition, speech recognition and adaptive control to name a few. To keep the pace with its huge demand in diversified application areas, many different kinds of ANN architecture and learning types have been proposed by the researchers to meet varying needs. A novel hybrid learning approach for the training of a feed-forward ANN has been proposed in this thesis. The approach combines evolutionary algorithms with matrix solution methods such as singular value decomposition, Gram-Schmidt etc., to achieve optimum weights for hidden and output layers. The proposed hybrid method is to apply evolutionary algorithm in the first layer and least square method (LS) in the second layer of the ANN. The methodology also finds optimum number of hidden neurons using a hierarchical combination methodology structure for weights and architecture. A learning algorithm has many facets that can make a learning algorithm good for a particular application area. Often there are trade offs between classification accuracy and time complexity, nevertheless, the problem of memory complexity remains. This research explores all the different facets of the proposed new algorithm in terms of classification accuracy, convergence property, generalization ability, time and memory complexity.

This dissertation aims to propose a new and practical method to obtain equivalent fracture network permeability (EFNP), which represents and replaces all the existing fractures located in each grid block for the reservoir simulation of naturally fractured reservoirs. To achieve this, first the relationship between different geometrical properties of fracture networks and their EFNP was studied. A MATLAB program was written to generate many different realizations of 2-D fracture networks by changing fracture length, density and also orientation. Next, twelve different 2-D fractal-statistical properties of the generated fracture networks were measured to quantify different characteristics. In addition to the 2-D fractal-statistical properties, readily available 1-D and 3-D data were also measured for the models showing variations of fracture properties in the Z-direction. The actual EFNP of each fracture network was then measured using commercial software called FRACA. The relationship between the 1-, 2- and 3-D data and EFNP was analyzed using multivariable regression analysis and based on these analyses, correlations with different number of variables were proposed to estimate EFNP. To improve the accuracy of the predicted EFNP values, an artificial neural network with the back-propagation algorithm was also developed. Then, using the experimental design technique, the impact of each fracture network parameter including fracture length, density, orientation and conductivity on EFNP was investigated. On the basis of the results and the analyses, the conditions to obtain EFNP for practical applications based on the available data (1-D well, 2-D outcrop, and 3-D welltest) were presented. This methodology was repeated for natural fracture patterns obtained mostly from the outcrops of different geothermal reservoirs. The validity of the equations was also tested against the real welltest data obtained from the fields. Finally, the concept of the percolation theory was used to determine whether each fracture network in the domain is percolating (permeable) and to quantify the fracture connectivity, which controls the EFNP. For each randomly generated fracture network, the relationship between the combined fractal-percolation properties and the EFNP values was investigated and correlations for predicting the EFNP were proposed. As before, the results were validated with a new set of fracture networks. / Petroleum Engineering

In this thesis, we use evolutionary robotics techniques to automatically design and synthesise behaviour for groups of simulated and real robots. Our contribution will be on the design of non-trivial individual and collective behaviour; decisions about solitary or social behaviour will be temporal and they will be interdependent with communicative acts. In particular, we study time-based decision-making in a social context: how the experiences of robots unfold in time and how these experiences influence their interaction with the rest of the group. We propose three experiments based on non-trivial real-world cooperative scenarios. First, we study social cooperative categorisation; signalling and communication evolve in a task where the cooperation among robots is not a priori required. The communication and categorisation skills of the robots are co-evolved from scratch, and the emerging time-dependent individual and social behaviour are successfully tested on real robots. Second, we show on real hardware evidence of the success of evolved neuro-controllers when controlling two autonomous robots that have to grip each other (autonomously self-assemble). Our experiment constitutes the first fully evolved approach on such a task that requires sophisticated and fine sensory-motor coordination, and it highlights the minimal conditions to achieve assembly in autonomous robots by reducing the assumptions a priori made by the experimenter to a functional minimum. Third, we present the first work in the literature to deal with the design of homogeneous control mechanisms for morphologically heterogeneous robots, that is, robots that do not share the same hardware characteristics. We show how artificial evolution designs individual behaviours and communication protocols that allow the cooperation between robots of different types, by using dynamical neural networks that specialise on-line, depending on the nature of the morphology of each robot. The experiments briefly described above contribute to the advancement of the state of the art in evolving neuro-controllers for collective robotics both from an application-oriented, engineering point of view, as well as from a more theoretical point of view.