In our research, we proposed a novel invariant in 2-D image contour recognition based on Hopfield-Tank neural network. At first, we searched the feature points, the position of feature points where are included high curvature and corner on the contour. We used polygonal approximation to describe the image contour. There have two patterns we set, one is model pattern another is test pattern. The Hopfield-Tank network was employed to perform feature matching. In our results show that we can overcome the test pattern which consists of translation, rotation, scaling transformation and no matter single or occlusion pattern.

The use of neural networks in the field of development geology is in its infancy. In this study, a neural network will be used to identify flow units in Happy Spraberry Field, Garza County, Texas. A flow unit is the mappable portion of the total reservoir within which geological and petrophysical properties that affect the flow of fluids are consistent and predictably different from the properties of other reservoir rock volumes (Ebanks, 1987). Ahr and Hammel (1999) further state a highly "ranked" flow unit (i.e. a good flow unit) would have the highest combined values of porosity and permeability with the least resistance to fluid flow. A flow unit may also include nonreservoir features such as shales and cemented layers where combined porosity-permeability values are lower and resistance to fluid flow much higher (i.e. a poor flow unit) (Ebanks, 1987). Production from Happy Spraberry Field primarily comes from a 100 foot interval of grainstones and packstones, Leonardian in age, at an average depth of 4,900 feet. Happy Spraberry Field is unlike most fields in that the majority of the wells have been cored in the zone of interest. This fact more easily lends the Happy Spraberry Field to a study involving neural networks. A neural network model was developed using a data set of 409 points where X and Y location, depth, gamma ray, deep resistivity, density porosity, neutron porosity, lab porosity, lab permeability and electrofacies were known throughout Happy Spraberry Field. The model contained a training data set of 205 cases, a verification data set of 102 cases and a testing data set of 102 cases. Ultimately two neural network models were created to identify electrofacies and reservoir quality (i.e. flow units). The neural networks were able to outperform linear methods and have a correct classification rate of 0.87 for electrofacies identification and 0.75 for reservoir quality identification.

Forest growth models are important to the forestry community because they provide means for predicting future yields and exploring different forest management practices. The purpose of this thesis is to develop an individual tree forest growth model applicable for the province of Nova Scotia. The Acadian forest of Nova Scotia is a prime example a mixed species forest which is best modelled with individual tree models. Individual tree models also permit modelling variable-density management regimes, which are important as the Province investigates new silviculture options. Rather than use the conventional regression techniques, our individual tree growth and yield model was developed using neural networks. The growth and yield model was comprised of three different neural networks: a network for each survivability, diameter increment and height increment. In general, the neural network modelling approach fit the provincial data reasonably well. In order to have a model applicable to each species in the Province, species was included as a model input; the models were able to distinguish between species and to perform nearly as well as species-specific models. It was also found that including site and stocking level indicators as model inputs improved the model. Furthermore, it was found that the GIS-based site quality index developed at UNB could be used as a site indicator rather than land capability. Finally, the trained neural networks were used to create a growth and yield model which would be limited to shorter prediction periods and a larger scale.

This thesis investigates the application of pattern recognition techniques to rock type recognition using monitoring-while-drilling data. The research is focused on data from a large electric blasthole drill operating in an open-pit coal mine. Pre-processing and normalization techniques are applied to minimize potential misclassification issues. Both supervised and unsupervised learning is employed in the classifier design: back-propagation neural networks are used for the supervised learning, while self-organizing maps are used for unsupervised learning. A variety of combinations of drilling data and geophysical data are investigated as inputs to the classifiers. The outputs from these classifiers are evaluated relative to the rock classification made by a commercially available rock type recognition system, as well as relative to independent labelling by a geologist. Classifier performance is improved when drilling data used as inputs are augmented with geophysical data inputs. By using supervised learning with both drilling and geophysical data as inputs, the misclassification of coal, as well as of the non-coal rock types, is reduced compared to results of current commercial recognition methods. Moreover, rock types which were not detected by the previous methods were successfully classified by the supervised models. / Thesis (Master, Mining Engineering) -- Queen's University, 2007-11-28 15:22:17.454 / I would like to thank the financial support provided by the George C. Bateman and J. J. Denny Graduate Fellowship, as well as funding from the Natural Sciences and Engineering Research Council of Canada (NSERC) provided via NSERC grant support to Dr. Daneshmend.

This work presents a new approach for detecting and isolating faults in nonlinear processes using independent neural network models. In this approach, an independent neural network is used to model the proton exchange membrane fuel cell nonlinear systems using a multi-input multi-output structure. This research proposed the use of radial basis function network and multilayer perceptron network to perform fault detection. After training, the neural network models can give accurate prediction of the system outputs, based on the system inputs. Using the residual generation concept developed in the model-based diagnosis, the difference between the actual and estimated outputs are used as residuals to detect faults. When the magnitude of these residuals exceed a predefined threshold, it is likely that the system is faulty. In order to isolate faults in the system, a second neural network is used to examine features in the residual. A specific feature would correspond to a specific fault. Based on features extracted and classification principles, the second neural network can isolate faults reliably and correctly. The developed method is applied to a benchmark simulation model of the proton exchange membrane fuel cell stacks developed at Michigan University. One component fault, one actuator fault and three sensor faults were simulated on the benchmark model. The simulation results show that the developed approach is able to detect and isolate the faults to a fault size of ±10% of nominal values. These results are promising and indicate the potential of the method to be applied to the real world of fuel cell stacks for dynamic monitoring and reliable operations.

Master of Science / Department of Geology / Matthew W. Totten / The Meramec member of Mississippian age is a proficient oil and gas producing formation within the midcontinent region of the United States. It is produced in Kansas, Oklahoma, and Texas. In Kansas, 12% of the state’s petroleum production comes from Mississippian-aged rocks. Bindley Field, located in central west Kansas, has produced 3,669,283 barrels of oil from one facies within the M2 interval of the Meramec formation. This facies is a grain-supported echinoderm/bryozoan dolostone, of variable thickness. Its sporadic occurrence in the subsurface has made exploring Bindley Field and the surrounding area difficult. The challenge in finding oil in this area is in locating a producible zone of this productive facies. Previously, Bindley Field has been the subject of detailed reservoir characterization studies (Ebanks et al., 1977; Johnson, 1990; Johnson, 1994). These studies helped to contribute to a better understanding of Meramecian stratigraphy in Kansas. The Meramec was divided into four major depositional sequences, with some of those sequences nonexistent in the subsurface, due to aerial exposure and erosion post-deposition. The Meramecian units were further separated into parasequence-scale chronostratigraphic units based on marine flooding events. The primary producing interval in Bindley Field is the Meramec 2 interval which consists of seven lithotypes, and is recognized to have six, meter-scale depositional cycles (Johnson, 1990). As production from this interval increased, more information became available about controls on reservoir quality. There are still areas, however, where core data do not exist, and predicting the productive facies remains challenging. The aim of this study is to create a workflow for evaluating the subsurface using regional core and log data from Bindley Field to create a model of the subsurface distribution of the reservoir facies, which could be extended to data poor areas. Geophysical logs (neutron, gamma ray, guard) along with an artificial neural network (ANN), was used to create an accurate prediction of producing intervals within the subsurface. Values are derived from wire line log data and used to develop the ANN definition of facies distribution within Bindley Field. The ANN model was examined for accuracy and precision using core description and well cuttings from wells within Bindley Field and the surrounding area. Correlations were found between the subsurface geometry of the study area, and the production of oil and gas within the study area. An ANN model with an accuracy of 72% was achieved and applied to wells surrounding the Bindley Field, where reservoir intervals have not been as extensively studied. A total of 87 wells in Bindley Field and the surrounding 50 square mile area where applied to the ANN model. The model predicted that the productive facies thickens gradually to the northwest of Bindley Field. Cross sections as well as an isopach map were created using the prediction data from the ANN. Finally, an analysis for the accuracy of the ANN and the predicted facies was created. The productive facies yielded an accuracy value of 77%.

Historically, artificial neural networks have been loosely defined as biologically inspired computational models. When deciding what sort of network to use for a given task there are two things that need to be considered. The first is the representational power of the given network, that is what class of problems can be solved by this network? Given a set of problems to be solved by neural networks, a network that can solve any of these problems is called a universal approximator. The second problem is the ability to find a desired network given an initial network via a learning rule. Here we are interested in the question of universal approximation. A general definition of artificial neural networks is provided along with definitions for different kinds of universal approximation. We then prove that the recurrent temporal restricted Boltzmann machine (RTRBM) satisfies a general type of universal approximation for stochastic processes, an extention of previous results for the simple RBM. We conclude by examining the potential use of such temporal artificial neural networks in the biological process of perception. / Graduate

Spiking neural networks (SNNs) are a more biologically realistic model of the brain than traditional analog neural networks and therefore should be better for modelling certain functions of the human brain. This thesis uses the concept of deriving an SNN from an accepted non-spiking neural network via analysis and modifications of the transmission function. We investigate this process to determine if and how the modifications can be made to minimize loss of information during the transition from non-spiking to spiking while retaining positive features and functionality of the non-spiking network. By comparing combinations of spiking neuron models and networks against each other, we determined that replacing the transmission function with a neural model that is similar to it allows for the easiest method to create a spiking neural network that works comparatively well. This similarity between transmission function and neuron model allows for easier parameter selection which is a key component in getting a functioning SNN. The parameters all play different roles, but for the most part, parameters that speed up spiking, such as large resistance values or small rheobases generally help the accuracy of the network. But the network is still incomplete for a spiking neural network since this conversion is often only performed after learning has been completed in analog form. The neuron model and subsequent network developed here are the initial steps in creating a bidirectional SNN that handles hetero-associative and auto-associative recall and can be switched easily between spiking and non-spiking with minimal to no loss of data. By tying everything to the transmission function, the non-spiking learning rule, which in our case uses the transmission function, and the neural model of the SNN, we are able to create a functioning SNN. Without this similarity, we find that creating SNN are much more complicated and require much more work in parameter optimization to achieve a functioning SNN.

This thesis deals with the potential application of neural network technology to aviation fuel consumption estimation. This is achieved by developing neural networks representative jet aircraft. Fuel consumption information obtained directly from the pilot's flight manual was trained by the neural network. The trained network was able to accurately and efficiently estimate fuel consumption of an aircraft for a given mission. Statistical analysis was conducted to test the reliability of this model for all segments of flight. Since the neural network model does not require any wind tunnel testing nor extensive aircraft analysis, compared to existing models used in aviation simulation programs, this model shows good potential. The design of the model is described in depth, and the MATLAB source code are included in appendices. / Master of Science

To understand remotely sensed data, one must understand the relationship between radiative transfer models and their predictions of the interaction of solar radiation on geophysical media. If it can be established that these models are indeed accurate, some form of evaluation has to be performed on these models, for users to choose the model that suits their requirements. This thesis focuses on the implementation of a variable linear kernel model, its validation, and to study its application in the prediction of BRDF effects using two different neural networks-- the backpropogation and the radial basis function neural network and finally to draw conclusions on which neural network is best suited for this model. Based on these results the optimum number of kernels for this model is derived.