• Transactions of the Korean Society of Machine Tool Engineers
• /
• v.15 no.5
• /
• pp.123-130
• /
• 2006

In manufacturing, process automation and parameter optimization are required in order to improve productivity. Especially in welding process, productivity and weldablity should be considered to determine the process parameter. In this paper, optimization methodology was proposed to determine the welding conditions using the objective function in terms of productivity and weldablity. In order to conduct this, welding experiments were carried out. Tensile test was performed to evaluate the weldability. Neural network model to estimate tensile strength using the laser power, welding speed, and wire feed rate was developed. Objective function was defined using the normalized tensile strength which represented the weldablilty and welding speed and wire feed rate which represented the productivity. The optimal welding parameters which maximized the objective function were determined.

• Proceedings of the KSME Conference
• /
• 2001.11a
• /
• pp.378-383
• /
• 2001

In order to improve automobile maneuverability and tire durability, it is very important for one to determine a suitable sidewall contour producing the ideal tension and strain-energy distributions. In order to determine such a sidewall contour, one must apply multi-objective optimization technique. The optimization problem of tire carcass contour involves several objective functions. Hence, we execute the tire contour optimization for improving the maneuverability and the tire durability using satisficing trade-off method. And, the tire optimization also requires long cup time for the sensitivity analysis. In order to resolve this numerical difficulty, we apply neural network algorithm.

• Journal of Electrical Engineering and Technology
• /
• v.9 no.1
• /
• pp.170-177
• /
• 2014

This paper presents a new method of classification of the induction machine faults using Time Frequency Representation, Particle Swarm Optimization and artificial neural network. The essence of the feature extraction is to project from faulty machine to a low size signal time-frequency representation (TFR), which is deliberately designed for maximizing the separability between classes, a distinct TFR is designed for each class. The feature vectors size is optimized using Particle Swarm Optimization method (PSO). The classifier is designed using an artificial neural network. This method allows an accurate classification independently of load level. The introduction of the PSO in the classification procedure has given good results using the reduced size of the feature vectors obtained by the optimization process. These results are validated on a 5.5-kW induction motor test bench.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2002.05a
• /
• pp.1086-1092
• /
• 2002

Computer simulation is essential to design the suspension elements of railway vehicle. By computer simulation, engineers can assess the feasibility of a given design factors and change them to get a better design. But if one wishes to perform complex analysis on the simulation, such as railway vehicle dynamic, the computational time can become overwhelming. Therefore, many researchers have used a mega model that has a regression model made by sampling data through simulation. In this paper, the neural network is used a mega model that have twenty-nine design variables and forty-six responses. After this mega model is constructed, multi-objective optimal solutions are achieved by using the differential evolution. This paper shows that this optimization method using the neural network and the differential evolution is a very efficient tool to solve the complex optimization problem.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.58 no.1
• /
• pp.126-131
• /
• 2009

This study was carried out for the reliability of PD(Partial Discharge) pattern recognition. For the pattern recognition, the database for PD was established by use of self-designed insulation defects which occur and were mostly critical in GIS(Gas Insulated Switchgear). The acquired database was analyzed to distinguish patterns by means of PRPD(Phase Resolved Partial Discharge) method and stored to the form with to unite the average amplitude of PD pulse and the number of PD pulse as the input data of neural network. In order to prove the performance of genetic algorithm combined with neural network, the neural networks with trial-and-error method and the neural network with genetic algorithm were trained by same training data and compared to the results of their pattern recognition rate. As a result, the recognition success rate of defects was 93.2% and the neural network train process by use of trial-and-error method was very time consuming. The recognition success rate of defects, on the other hand, was 100% by applying the genetic algorithm at neural network and it took a relatively short time to find the best solution of parameters for optimization. Especially, it could be possible that the scrupulous parameters were obtained by genetic algorithm.

• Journal of Electrical Engineering and Technology
• /
• v.13 no.4
• /
• pp.1724-1731
• /
• 2018

The conventional polynomial neural network (PNN) is a classical flexible neural structure and self-organizing network, however it is not free from the limitation of overfitting problem. In this study, we propose a space search-optimized polynomial neural network (ssPNN) structure to alleviate this problem. Ranking selection is realized by means of ranking selection-based performance index (RS_PI) which is combined with conventional performance index (PI) and coefficients based performance index (CPI) (viz. the sum of squared coefficient). Unlike the conventional PNN, L2-norm regularization method for estimating the polynomial coefficients is also used when designing the ssPNN. Furthermore, space search optimization (SSO) is exploited here to optimize the parameters of ssPNN (viz. the number of input variables, which variables will be selected as input variables, and the type of polynomial). Experimental results show that the proposed ranking selection-based polynomial neural network gives rise to better performance in comparison with the neuron fuzzy models reported in the literatures.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.67 no.8
• /
• pp.1071-1079
• /
• 2018

In this study, the design methodology as well as network architecture of Support Vector Machine based Polynomial Neural Network, which is a kind of the dynamically generated neural networks, is introduced. The Support Vector Machine based polynomial neural networks is given as a novel network architecture redesigned with the aid of polynomial neural networks and Support Vector Machine. The generic polynomial neural networks, whose nodes are made of polynomials, are dynamically generated in each layer-wise. The individual nodes of the support vector machine based polynomial neural networks is constructed as a support vector machine, and the nodes as well as layers of the support vector machine based polynomial neural networks are dynamically generated as like the generation process of the generic polynomial neural networks. Support vector machine is well known as a sort of robust pattern classifiers. In addition, in order to enhance the structural flexibility as well as the classification performance of the proposed classifier, multi-objective particle swarm optimization is used. In other words, the optimization algorithm leads to sequentially successive generation of each layer of support vector based polynomial neural networks. The bench mark data sets are used to demonstrate the pattern classification performance of the proposed classifiers through the comparison of the generalization ability of the proposed classifier with some already studied classifiers.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.5 no.4
• /
• pp.756-765
• /
• 2001

An effective dexterous motion control method of redundant robot manipulators based on neural optimization network is proposed to satisfy multi-criteria such as singularity avoidance, minimizing energy consumption, and avoiding physical limits of actuator, while performing a given task. The method employs a neural optimization network with parallel processing capability, where only a simple geometric analysis for resolved motion of each joint is required instead of computing of the Jacobian and its pseudo inverse matrix. For dexterous motion, a joint geometric manipulability measure(JGMM) is proposed. JGMM evaluates a contribution of each joint differential motion in enlarging the length of the shortest axis among principal axes of the manipulability ellipsoid volume approximately obtained by a geometric analysis. Redundant robot manipulators is then controlled by neural optimization networks in such a way that 1) linear combination of the resolved motion by each joint differential motion should be equal to the desired velocity, 2) physical limits of joints are not violated, and 3) weighted sum of the square of each differential joint motion is minimized where weightings are adjusted by JGMM. To show the validity of the proposed method, several numerical examples are illustrated.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.35S no.4
• /
• pp.67-75
• /
• 1998

In general, Evoluationary Algorithm(EAs) are refered to as methods of population-based optimization. And EAs are considered as very efficient methods of optimal sytem design because they can provice much opportunity for obtaining the global optimal solution. This paper presents a co-evolution scheme of artifical neural networks, which has two different, still cooperatively working, populations, called as a host popuation and a parasite population, respectively. Using the conventional generatic algorithm the host population is evolved in the given environment, and the parastie population composed of schemata is evolved to find useful schema for the host population. the structure of artificial neural network is a diagonal recurrent neural netork which has self-feedback loops only in its hidden nodes. To find optimal neural networks we should take into account the structure of the neural network as well as the adaptive parameters, weight of neurons. So we use the genetic algorithm that searches the structure of the neural network by the co-evolution mechanism, and for the weights learning we adopted the evolutionary stategies. As a results of co-evolution we will find the optimal structure of the neural network in a short time with a small population. The validity and effectiveness of the proposed method are inspected by applying it to the stabilization and position control of the invered-pendulum system. And we will show that the result of co-evolution is better than that of the conventioal genetic algorithm.

• ETRI Journal
• /
• v.42 no.4
• /
• pp.505-517
• /
• 2020

The increasing size and complexity of deep neural networks (DNNs) necessitate the development of efficient high-performance accelerators. An efficient memory structure and operating scheme provide an intuitive solution for high-performance accelerators along with dataflow control. Furthermore, the processing of various neural networks (NNs) requires a flexible memory architecture, programmable control scheme, and automated optimizations. We first propose an efficient architecture with flexibility while operating at a high frequency despite the large memory and PE-array sizes. We then improve the efficiency and usability of our architecture by automating the optimization algorithm. The experimental results show that the architecture increases the data reuse; a diagonal write path improves the performance by 1.44× on average across a wide range of NNs. The automated optimizations significantly enhance the performance from 3.8× to 14.79× and further provide usability. Therefore, automating the optimization as well as designing an efficient architecture is critical to realizing high-performance DNN accelerators.