• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 1995.10a
• /
• pp.17-24
• /
• 1995

This work proposes an optimality criteria applicable to the optimum design of plane frames. Stress constraints as well as displacement constraints are treated as behavioural constraints and thus the first order approximation of stress constraints is adopted. The design space of practical reinforced concrete frames with discrete design variables has been found to have many local minima, and thus it is desirable to find in advance the mathematical minimum, hopefully global, prior to starting to search a practical optimum design. By using the mathematical minimum as a trial design of any search algorithm, we may not full into a local minimum but apparently costly design. Therefore this work aims at establishing a mathematically rigorous method ⑴ by adopting first-order approximation of constraints, ⑵ by reducing the design space whenever minimum size restrictions become "active" and ⑶ by the of Newton-Raphson Method.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.34C no.12
• /
• pp.61-69
• /
• 1997

This paper proposes a new method for improving the performances of the neural network for optimization using a hyubrid of gradient descent method and dynamic tunneling system. The update rule of gradient descent method, which has the fast convergence characteristic, is applied for high-speed optimization. The update rule of dynamic tunneling system, which is the deterministic method with a tunneling phenomenon, is applied for global optimization. Having converged to the for escaping the local minima by applying the dynamic tunneling system. The proposed method has been applied to the travelling salesman problems and the optimal task partition problems to evaluate to that of hopfield model using the update rule of gradient descent method.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.48 no.11
• /
• pp.1464-1471
• /
• 1999

This paper introduces complex-valued competitive learning neural network for spatio-temporal pattern recognition. There have been quite a few neural networks for spatio-temporal pattern recognition. Among them, recurrent neural network, TDNN, and avalanche model are acknowledged as standard neural network paradigms for spatio-temporal pattern recognition. Recurrent neural network has complicated learning rules and does not guarantee convergence to global minima. TDNN requires too many neurons, and can not be regarded to deal with spatio-temporal pattern basically. Grossberg's avalanche model is not able to distinguish long patterns, and has to be indicated which layer is to be used in learning. In order to remedy drawbacks of the above networks, unsupervised competitive learning using complex umber is proposed. Suggested neural network also features simultaneous recognition, time-shift invariant recognition, stable categorizing, and learning rate modulation. The network is evaluated by computer simulation with randomly generated patterns.

• Proceedings of the KIEE Conference
• /
• 1993.07b
• /
• pp.968-970
• /
• 1993

In this paper, the loss components of IPMSM(Inteiror Permanent Magnet Synchronous Motor) is derived. To maximize the efficiency of the motor, a design method that optimizes the design variables is proposed Objective function consists of stator winding loss, core loss, and mechanical loss. Simulated annealing is used as the optimization method which is appropriate for finding the global minimum of nonlinear function with many local minima. Through the simulation of the motor characteristics, the prominence of the proposed design method is verified.

• Nuclear Engineering and Technology
• /
• v.53 no.11
• /
• pp.3772-3783
• /
• 2021

This work proposes a new efficient meta-heuristic optimization algorithm called Gamma Ray Interactions Based Optimization (GRIBO). The algorithm mimics different energy loss processes of a gamma-ray photon during its passage through a matter. The proposed novel algorithm has been applied to search for the global minima of 30 standard benchmark functions. The paper also considers solving real optimization problem in the field of nuclear engineering, radioisotope identification. The results are compared with those obtained by the Particle Swarm Optimization, Genetic Algorithm, Gravitational Search Algorithm and Grey Wolf Optimizer algorithms. The comparisons indicate that the GRIBO algorithm is able to provide very competitive results compared to other well-known meta-heuristics.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.6
• /
• pp.83-90
• /
• 1994

In this paper, we propose a new learning algorithm, ASOFM(Adaptive Self Organizing Feature Map), to solve the defects of Kohonen's Self Organiaing Feature Map. Kohonen's algorithm is sometimes stranded on local minima for the initial weights. The proposed algorithm uses an object function which can evaluate the state of network in learning and adjusts the learning rate adaptively according to the evaluation of the object function. As a result, it is always guaranteed that the state of network is converged to the global minimum value and it has a capacity of generalized learning by adaptively. It is reduce that the learning time of our algorithm is about $30\%$ of Kohonen's.

• Journal of Institute of Control, Robotics and Systems
• /
• v.4 no.4
• /
• pp.448-457
• /
• 1998

A neural network based control algorithm with fuzzy compensation is proposed for the automated adjustment in the production of electronic end-products. The process of adjustment is to tune the variable devices in order to examine the specified performances of the products ready prior to packing. Camcorder is considered as a target product. The required test and adjustment system is developed. The adjustment system consists of a NNC(neural network controller), a sub-NNC, and an auxiliary algorithm utilizing the fuzzy logic. The neural network is trained by means of errors between the outputs of the real system and the network, as well as on the errors between the changing rate of the outputs. Control algorithm is derived to speed up the learning dynamics and to avoid the local minima at higher energy level, and is able to converge to the global minimum at lower energy level. Many unexpected problems in the application of the real system are resolved by the auxiliary algorithms. As the adjustments of multiple items are related to each other, but the significant effect of performance by any specific item is not observed. The experimental result shows that the proposed method performs very effectively and are advantageous in simple architecture, extracting easily the training data without expertise, adapting to the unstable system that the input-output properties of each products are slightly different, with a wide application to other similar adjustment processes.

• Asian Journal of Atmospheric Environment
• /
• v.9 no.2
• /
• pp.158-164
• /
• 2015

The estimation of the concentration of hydroxyl radical (OH) in the atmosphere is essential to build atmospheric models and to understand the mechanisms of the reactions involved in OH. Although water vapor is one of the most abundant species in the troposphere, only a few studies have been performed for the reaction of OH and water vapor. Here I demonstrate an ab initio study on the complex forming reation of OH with $H_2O$ in the gas phase performed based on density functional theory to calculate the reaction rate and the energy states of the reactant and the OH-$H_2O$ complex. The structure of the complex, which belongs to the Cs point group, was optimized at global minima. The transition state was not found at the B3LYP and MP2 levels of theory. Rate constants of the forward and the reverse reactions were calculated as $1.1{\times}10^{-16}cm^3\;molecule^{-1}\;s^{-1}$ and $5.3{\times}10^9\;s^{-1}$, respectively. The extremely slow rates of complex forming reaction and the resulting hydrogen atom exchange reaction of OH and $H_2O$, which are consistent with experimentally determined values, imply a negligible possibility of a change in OH reactivity through the title reaction.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.10 no.3
• /
• pp.270-279
• /
• 2000

This paper describes the on-chip learning algorithm of neural networks using the stochastic pulse arithmetic. Stochastic pulse arithmetic is the computation using the numbers represented by the probability of 1' and 0's occurrences in a random pulse stream. This stochastic arithmetic has the merits when applied to neural network ; reduction of the area of the implemented hardware and getting a global solution escaping from local minima by virtue of the stochastic characteristics. And in this study, the on-chip learning algorithm is derived from the backpropagation algorithm for effective hardware implementation. We simulate the nonlinear separation problem of the some character patterns to verify the proposed learning algorithm. We also had good results after applying this algorithm to recognize printed and handwritten numbers.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.28B no.10
• /
• pp.831-841
• /
• 1991

A combined model of probabilistic and MLP(multi layer perceptron) model is proposed for the pattern classification of EMG( electromyogram) signals. The MLP model has a problem of not guaranteeing the global minima of error and different quality of approximations to Bayesian probabilities. The probabilistic model is, however, closely related to the estimation error of model parameters and the fidelity of assumptions. A proper combination of these will reduce the effects of the problems and be robust to input variations. Proposed model is able to get the MAP(maximum a posteriori probability) in the probabilistic model by estimating a priori probability distribution using the MLP model adaptively. This method minimize the error probability of the probabilistic model as long as the realization of the MLP model is optimal, and this is a good combination of the probabilistic model and the MLP model for the usage of MLP model reliability. Simulation results show the benefit of the proposed model compared to use the Mlp and the probabilistic model seperately and the average calculation time fro classification is about 50ms in the case of combined motion using an IBM PC 25 MHz 386model.