• Structural Engineering and Mechanics
• /
• v.36 no.2
• /
• pp.165-179
• /
• 2010

Maximum deflection in a beam is a design criteria and occurs generally at or close to the mid-span. Neural networks have been developed for the continuous composite beams to predict the inelastic mid-span deflections (typically for 20 years, considering cracking, and time effects, i.e., creep and shrinkage, in concrete) from the elastic moments and elastic mid-span deflections (neglecting instantaneous cracking and time effects). The training and testing data for the neural networks is generated using a hybrid analytical-numerical procedure of analysis. The neural networks have been validated for four example beams and the errors are shown to be small. This methodology, of using networks enables a rapid estimation of inelastic mid-span deflections and requires a computational effort almost equal to that required for the simple elastic analysis. The neural networks can be extended for the composite building frames that would result in huge saving in computational time.

• Computers and Concrete
• /
• v.3 no.6
• /
• pp.439-454
• /
• 2006

This paper develops an artificial neural network (ANN) model for uniformly loaded restrained reinforced concrete (RC) slabs incorporating membrane action. The development of membrane action in RC slabs restrained against lateral displacements at the edges in buildings and bridge structures significantly increases their load carrying capacity. The benefits of compressive membrane action are usually not taken into account in currently available design methods based on yield-line theory. By extending the existing knowledge of compressive membrane action, it is possible to design slabs in building and bridge decks economically with less than normal reinforcement. The processes involved in the development of ANN model such as the creation of a database of test results from previous research studies, the selection of architecture of the network from extensive trial and error procedure, and the training and performance validation of the model are presented. The ANN model was found to predict accurately the ultimate strength of fully restrained RC slabs. The model also was able to incorporate strength enhancement of RC slabs due to membrane action as confirmed from a comparative study of experimental and yield line-based predictions. Practical applications of the developed ANN model in the design process of RC slabs are also highlighted.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.527-531
• /
• 1998

This study introduce to the fuzzy membership filter to cancel a high frequency noise of welding current. And ART2 which has the competitive learning network classifiers the signal patterns for the filtered welding signal. A welding current possesses a specific pattern according to the existence or the size of a welding gap. These specific patterns result in different classification in comparison with an occasion for no welding gap. The patterns In each case of 1mm, 2mm, 3mm, and no welding gap are identified by the artificial neural network. These procedure is an off-line execution. In on-line execution, the identification model of neural network for the classified pattern is located on ahead of the welding plant. And when the welding current patterns pass through the neural network in the direction of feedforward. it is possible to recognize the existence or the size of a welding gap.

• The Korean Journal of Applied Statistics
• /
• v.26 no.5
• /
• pp.847-856
• /
• 2013

We propose a modified learning process for generalized neural network using a learning algorithm by Liu et al. (2001). We consider the effect of initial weights, training results and learning errors using a modified learning process. We employ an incremental training procedure where training patterns are learned systematically. Our algorithm starts with a single training pattern and a single hidden layer neuron. During the course of neural network training, we try to escape from the local minimum by using a weight scaling technique. We allow the network to grow by adding a hidden layer neuron only after several consecutive failed attempts to escape from a local minimum. Our optimization procedure tends to make the network reach the error tolerance with no or little training after the addition of a hidden layer neuron. Simulation results with suitable initial weights indicate that the present constructive algorithm can obtain neural networks very close to minimal structures and that convergence to a solution in neural network training can be guaranteed. We tested these algorithms extensively with small training sets.

• Journal of Power Electronics
• /
• v.5 no.2
• /
• pp.129-141
• /
• 2005

A high-performance robust hybrid speed controller for a permanent-magnet synchronous motor (PMSM) drive with an on-line trained neural-network model-following controller (NNMFC) is proposed. The robust hybrid controller is a two-degrees-of-freedom (2DOF) integral plus proportional & rate feedback (I-PD) with neural-network model-following (NNMF) speed controller (2DOF I-PD NNMFC). The robust controller combines the merits of the 2DOF I-PD controller and the NNMF controller to regulate the speed of a PMSM drive. First, a systematic mathematical procedure is derived to calculate the parameters of the synchronous d-q axes PI current controllers and the 2DOF I-PD speed controller according to the required specifications for the PMSM drive system. Then, the resulting closed loop transfer function of the PMSM drive system including the current control loop is used as the reference model. In addition to the 200F I-PD controller, a neural-network model-following controller whose weights are trained on-line is designed to realize high dynamic performance in disturbance rejection and tracking characteristics. According to the model-following error between the outputs of the reference model and the PMSM drive system, the NNMFC generates an adaptive control signal which is added to the 2DOF I-PD speed controller output to attain robust model-following characteristics under different operating conditions regardless of parameter variations and load disturbances. A computer simulation is developed to demonstrate the effectiveness of the proposed 200F I-PD NNMF controller. The results confirm that the proposed 2DOF I-PO NNMF speed controller produces rapid, robust performance and accurate response to the reference model regardless of load disturbances or PMSM parameter variations.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.60 no.4
• /
• pp.862-870
• /
• 2011

In this paper, we introduce a new topology of Radial Basis Function-based Polynomial Neural Networks (RPNN) that is based on a genetically optimized multi-layer perceptron with Radial Polynomial Neurons (RPNs). This study offers a comprehensive design methodology involving mechanisms of optimization algorithms, especially Fuzzy C-Means (FCM) clustering method and Particle Swarm Optimization (PSO) algorithms. In contrast to the typical architectures encountered in Polynomial Neural Networks (PNNs), our main objective is to develop a design strategy of RPNNs as follows : (a) The architecture of the proposed network consists of Radial Polynomial Neurons (RPNs). In here, the RPN is fully reflective of the structure encountered in numeric data which are granulated with the aid of Fuzzy C-Means (FCM) clustering method. The RPN dwells on the concepts of a collection of radial basis function and the function-based nonlinear (polynomial) processing. (b) The PSO-based design procedure being applied at each layer of RPNN leads to the selection of preferred nodes of the network (RPNs) whose local characteristics (such as the number of input variables, a collection of the specific subset of input variables, the order of the polynomial, and the number of clusters as well as a fuzzification coefficient in the FCM clustering) can be easily adjusted. The performance of the RPNN is quantified through the experimentation where we use a number of modeling benchmarks - NOx emission process data of gas turbine power plant and learning machine data(Automobile Miles Per Gallon Data) already experimented with in fuzzy or neurofuzzy modeling. A comparative analysis reveals that the proposed RPNN exhibits higher accuracy and superb predictive capability in comparison to some previous models available in the literature.

• Proceedings of the KOSOMBE Conference
• /
• v.1998 no.11
• /
• pp.207-208
• /
• 1998

Classification of extracellularly recorded action potential into each unit is an important procedure for further analysis of spike trains as point process. We utilize feedforward neural network structures, multilayer perceptron and radial basis function network to implement spike classifier. For the efficient training of classifiers, nonlinear energy operator that can trace the instantaneous frequency as well as the amplitude of the input signal is used. Trained classifiers shows successful operation, up to 90% correct classification was possible under 1.2 of signal-to-noise ratio.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.12
• /
• pp.3855-3867
• /
• 2022

Name separation (splitting full names into surnames and given names) is not a tedious task in a multiethnic country because the procedure for splitting surnames and given names is ethnicity-specific. Malaysia has multiple main ethnic groups; therefore, separating Malaysian full names into surnames and given names proves a challenge. In this study, we develop a two-phase framework for Malaysian name separation using deep learning. In the initial phase, we predict the ethnicity of full names. We propose a recurrent neural network with long short-term memory network-based model with character embeddings for prediction. Based on the predicted ethnicity, we use a rule-based algorithm for splitting full names into surnames and given names in the second phase. We evaluate the performance of the proposed model against various machine learning models and demonstrate that it outperforms them by an average of 9%. Moreover, transfer learning and fine-tuning of the proposed model with an additional dataset results in an improvement of up to 7% on average.

• Journal of Broadcast Engineering
• /
• v.23 no.5
• /
• pp.614-621
• /
• 2018

This paper proposes a method of restoring corrupted depth image captured by depth camera through unsupervised learning using generative adversarial network (GAN). The proposed method generates restored face depth images using 3D morphable model convolutional neural network (3DMM CNN) with large-scale CelebFaces Attribute (CelebA) and FaceWarehouse dataset for training deep convolutional generative adversarial network (DCGAN). The generator and discriminator equip with Wasserstein distance for loss function by utilizing minimax game. Then the DCGAN restore the loss of captured facial depth images by performing another learning procedure using trained generator and new loss function.

• Structural Engineering and Mechanics
• /
• v.64 no.4
• /
• pp.437-447
• /
• 2017

Explicit expressions for rapid prediction of inelastic design quantities (considering cracking of concrete) from corresponding elastic quantities, are presented for multi-storey composite frames (with steel columns and steel-concrete composite beams) subjected to service load. These expressions have been developed from weights and biases of the trained neural networks considering concrete stress, relative stiffness of beams and columns including effects of cracking in the floors below and above. Large amount of data sets required for training of neural networks have been generated using an analytical-numerical procedure developed by the authors. The neural networks have been developed for moments and deflections, for first floor, intermediate floors (second floor to ante-penultimate floor), penultimate floor and topmost floor. In the case of moments, expressions have been proposed for exterior end of exterior beam, interior end of exterior beam and both interior ends of interior beams, for each type of floor with a total of twelve expressions. Similarly, in the case of deflections, expressions have been proposed for exterior beam and interior beam of each type of floor with a total of eight expressions. The proposed expressions have been verified by comparison of the results with those obtained from the analytical-numerical procedure. This methodology helps to obtain the inelastic design quantities from the elastic quantities with simple calculations and thus would be very useful in preliminary design.