Sharif Digital Repository

In this paper, a new method is introduced for identification of transformer winding fault through transfer function analysis. For this analysis, vector fitting and probabilistic neural network are used. The results of transfer functions estimation through vector fitting are employed for training of neural network, and consequently, probabilistic neural network is used for classification of faults. The required data for fault type identification are obtained by measurements on two groups of transformers (one is a classic 20 kV transformer, and the other is a model transformer) under intact condition and under different fault conditions (axial displacement, radial deformation, disc space... 

This study is concerned with investigation of pressure oxidative leaching of entire molybdenum of a molybdenite concentrate. Effects of oxygen pressure, stirring speed, pulp density, acid concentration, and temperature on the leaching rate of molybdenum were studied. A three-layer feed-forward artificial neural network was applied to model the effect of the abovementioned parameters on the leaching ability. The leaching efficiency was considered as a target value for modeling. The quantified leaching efficiencies obtained by applying different parameters demonstrated a good agreement with neural network predictions  

An artificial neural network (ANN) model was used to predict the weight loss of MgO-C composites at different temperatures and graphite contents. The general idea of ANN modeling was presented and after that the empirical weight loss data were used for both model verification and assessment of the oxidation rate predictions. The model was proved to have an astounding power in predicting kinetic parameters of the oxidation process. Graphite oxidation was, for example, found to be controlled by alternative diffusion steps. Plotting the Arrhenius law curves for graphite oxidation indicated a distinguishable slope change at a critical temperature which is related to the graphite content. This... 

In this work, the modified Flory-Huggins coupled with the free-volume concept and the artificial neural network models were used to obtain the osmotic pressure of aqueous poly(ethylene glycol) solutions. In the artificial neural network, the osmotic pressure of aqueous poly(ethylene glycol) solutions depends on temperature, molecular weight and the mole fractions of poly(ethylene glycol) in aqueous solution. The network topology is optimized and the (3-1-1) architecture is found using optimization of an objective function with batch back propagation (BBP) method for 134 experimental data points. The results obtained from the neural network in obtaining of the osmotic pressure of aqueous... 

In this work the effects of backwash pressure, duration of acid and sodium hydroxide backwashing, sodium hydroxide concentration, and the duration of forward washing on performance of permeate flux recovery (PFR) were investigated. A two-level fractional factorial design (FFD) was used to design the experiments. The ability of back propagation neural network (BPNN) and radial basis function neural network (RBFNN) in predicting the performance of cleaning-in-place (CIP) of hydrophilic polyethersulfone (PES) membrane were investigated. It is found that BPNN has better ability in predicting the PFR performance than RBFNN. The best architecture of BPNN was a network consisting of 1 hidden layer... 

Embryogenesis, regeneration and cell differentiation in microbiological entities are influenced by mechanical forces. Therefore, development of mechanical properties of these materials is important. Neural network technique is a useful method which can be used to obtain cell deformation by the means of force-geometric deformation data or vice versa. Prior to insertion in the needle injection process, deformation and geometry of cell under external point-load is a key element to understand the interaction between cell and needle. In this paper the goal is the prediction of cell membrane deformation under a certain force, and to visually estimate the force of indentation on the membrane from... 

In this article, I have tried to design an intelligent system which can separate and classify perfect and defective circular plates according to their size. After preprocessing, specifications of defects and size are determined through image processing, and finally, a system is proposed through perceptron neural networks methods, neuro fuzzy method, and Lvq to separate these products on basis of their size and defects. In the designing of this system, when input and its related intend is obvious before training network, perceptron neural networks give more exact results. If input and its related output have been clarified but the output have been related to some sub-inputs, lvq method is... 

In this paper, synchronization of chaotic systems with unknown parameters and unmeasured states is investigated. Two nonidentical chaotic systems in the framework of a master and a slave are considered for synchronization. It is assumed that both systems have uncertain dynamics, and states of the slave system are not measured. To tackle this challenging synchronization problem, a novel neural network-based adaptive observer and an adaptive controller have been designed. Moreover, a neural network is utilized to approximate the unknown dynamics of the slave system. The proposed method imposes neither restrictive assumption nor constraint on the dynamics of the systems. Furthermore, the... 

PPG based heart rate (HR) monitoring has recently attracted much attention with the advent of wearable devices such as smart watches and smart bands. However, due to severe motion artifacts (MA) caused by wristband stumbles, PPG based HR monitoring is a challenging problem in scenarios where the subject performs intensive physical exercises. This work proposes a novel approach to the problem based on supervised learning by Neural Network (NN). By simulations on the benchmark datasets [1], we achieve acceptable estimation accuracy and improved run time in comparison with the literature. A major contribution of this work is that it alleviates the need to use simultaneous acceleration signals.... 

Localization and quantification of structural damage and estimating the failure probability are key outputs in the reliability assessment of structures. In this study, an Artificial Neural Network (ANN) is used to reduce the computational effort required for reliability analysis and damage detection. Toward this end, one demonstrative structure is modeled and then several damage scenarios are defined. These scenarios are considered as training data sets for establishing an ANN model. In this regard, the relationship between structural response (input) and structural stiffness (output) is established using ANN models. The established ANN is more economical and achieves reasonable accuracy in... 

Facial expression imitation with applications in the design of human robot interaction (HRI) systems is an active area of research. In this study, we propose an approach for real-time imitation of human facial expression by a humanoid social robot 'Alice'. Artificial neural network (ANN) and Kinect sensor are used for recognition and classifying of the facial expressions like happiness, sadness, fear, anger and surprise; with the Alice humanoid robot imitating the comprehended expressions. Results and experiments demonstrate the effectiveness of the approach. © 2016 IEEE  

Various systems with nonlinearity can be modeled by utilizing uncertain linear-in-parameter models. In this paper, the uncertain parameters are in the form of Z-number coefficients. Fuzzy equations are utilized to represent the models of the uncertain nonlinear systems. The solutions associated with fuzzy equations are considered to be controllers while the desired references are outputs. The existence condition associated with the solution is laid down. Two various structure of neural networks are applied for approximating solutions of fuzzy equations with Z-number coefficients. The suggested techniques are validated by implementing an example. © 2018 The Authors. Published by Elsevier B.V  

In this research, a new metaheuristic optimization algorithm, inspired by biological nervous systems and artificial neural networks (ANNs) is proposed for solving complex optimization problems. The proposed method, named as neural network algorithm (NNA), is developed based on the unique structure of ANNs. The NNA benefits from complicated structure of the ANNs and its operators in order to generate new candidate solutions. In terms of convergence proof, the relationship between improvised exploitation and each parameter under asymmetric interval is derived and an iterative convergence of NNA is proved theoretically. In this paper, the NNA with its interconnected computing unit is examined... 

For simulating the surgical procedures in a virtual environment, it is necessary to propose a suitable virtual environment model which can reflect the real physical tool-tissue interaction behavior and a proper human user interface appropriate dynamic model. In this study, a linear Kelvin-Voigt and a nonlinear Hunt-Crossley models have been utilized to describe human hand dynamics during interaction with a haptic interface. Using the Lyapunov stability criteria, an adaptive neural network based controller being designed for guaranteeing the stability of the entire system, considering a nonlinear model for the environment, an inertia for a virtual tool, a constant time delay for data... 

We are introducing a new type of membrane, called neuro-skin or neuro-membrane. It is comprised of neurons embedded in a plastic membrane. The skin is smart and adaptive and is capable of providing desirable response to inputs intelligently. This way, the neuro-skin can be considered as a new type of neural network with adaptivity and learning capabilities. However, in this paper, only the response of neuro-skins to a dynamic input is studied. The membrane is modelled by nonlinear dynamic finite elements. Each finite element is considered as a cell of the neuro-skin which has a neuron. The neuron is the intelligent nucleus of the element. So, the finite elements are called finite... 

Accurate gross tumor volume (GTV) segmentation in esophagus CT images is a critical task in computer aided diagnosis (CAD) systems. However, because of the difficulties raised by the contrast similarity between esophageal GTV and its neighboring tissues in CT scans, this problem has been addressed weakly. In this paper, we present a 3D end-to-end method based on a convolutional neural network (CNN) for this purpose. We leverage design elements from DenseNet in a typical U-shape. The proposed architecture consists of a contractile path and an extending path that includes dense blocks for extracting contextual features and retrieves the lost resolution respectively. Using dense blocks leads to... 

Statistical bias correction methods are inferred relationships between inputs and outputs. The constructed functions are based on available observations, which are limited in time and space. This study investigates the ability of regression models (linear and nonlinear) to regionalize a domain by defining a minimum number of training pixels necessary to achieve a good level of bias correction performance. Linear regression is used to divide northern South America into five regions. To correct the biases of temperature and precipitation, an artificial neural network (ANN) model was trained with selected pixels within each region and then used to reproduce bias-corrected temperature and... 

We are introducing a new type of membrane, called neuro-skin or neuro-membrane. It is comprised of neurons embedded in a plastic membrane. The skin is smart and adaptive and is capable of providing desirable response to inputs intelligently. This way, the neuro-skin can be considered as a new type of neural network with adaptivity and learning capabilities. However, in this paper, only the response of neuro-skins to a dynamic input is studied. The membrane is modelled by nonlinear dynamic finite elements. Each finite element is considered as a cell of the neuro-skin which has a neuron. The neuron is the intelligent nucleus of the element. So, the finite elements are called finite... 

In this paper, an advanced integrated pixel-level method based on the deep convolutional neural network (DCNN) approach named DeepLabv3+ is proposed for weak interlayers detection and quantification. Furthermore, a database containing 32,040 images of limestone, dolomite, loess clay, and red clay is established to verify this method. The proposed model is then trained, validated, and tested via feeding multiple weak interlayers. Moreover, robustness and adaptability of the proposed model are evaluated, and the weak interlayers are extracted. Compared with the fully convolutional network (FCN)-based method and traditional image techniques, the proposed model provides higher accuracy in terms... 

Semantic segmentation is one of the most important tasks in the field of computer vision. It is the main step towards scene understanding. With the advent of RGB-Depth sensors, such as Microsoft Kinect, nowadays RGB-Depth images are easily available. This has changed the landscape of some tasks such as semantic segmentation. As the depth images are independent of illumination, the combination of depth and RGB images can improve the quality of semantic labeling. The related research has been divided into two main categories, based on the usage of hand-crafted features and deep learning. Although the state-of-the-art results are mainly achieved by deep learning methods, traditional methods...