Earticle

Thought Extreme Learning Machine-Radial Basis Function(ELM-RBF) can be used easily and can complete learning phase at very fast speed and provide more compact network than classical Extreme Learning Machine(ELM), it still has some room to improvement. Micro-Genetic Algorithms(uGA) improved calculated speed while inherits the Genetic Algorithms advantage of good for optimization and overall search. Considered on these, the paper designed a optimization strategy for ELM-RBF neural network based on uGA. In particular, based on classical RBF-ELM, we use real-u GA algorithm to optimize ELM-RBF hidden layer neurons center and biases value. Experiments results show that ELM-RBF-uGA has better recognition and prediction performance than classical ELM-RBF.

Simple genetic algorithm(SGA) has been used to optimize a lot of structural optimization problems because it can optimize non-linear problems and obtain the global solution. But, because of large evolving populations during many generations, it takes a long time to calculate fitness. Therefore this paper applied micro-genetic algorithm(μ -GA) to structural optimization and compared results of μ -GA with results of SGA. Additionally, the Paper applied μ -GA to gate optimization problem for injection molds by using simulation program CAPA.

산업현장에서 생산원가의 절감에 관심이 않아지면서 넓은 원자재에서 크기가 다양한 작은 부재를 절단하여 소재를 얻는 경우, 넓은 원자재에 소재를 배치하는 상태에 따라 원자재의 사용비율은 달라지게 된다. 같은 양의 제품을 만들 때 원자재의 사용량을 줄일 수 있으면 그 만큼 원가를 절약할 수 있다. 따라서 본 논문에서는 원자재에 소재를 배치하는 기본 논리와 방법을 고찰하고, 적당하며 실용적인 방법을 제시하며, 이러한 기반을 토대로 Micro-Genetic Algorithms($\mu$GAs)을 이용한 2차원적 부재의 최적화 배치 효율을 측정하여 기존의 배치알고리즘과 비교하였다. 위의 과정을 컴퓨터 시뮬레이션을 이용하여 구현하였으며 기존의 연구보다 좋은 배치효율을 얻을 수 있었다. 이렇게 하여 얻어진 배치의 들과는 최적에 가까운 상태를 얻을 수 있음을 확인하였으며 원자재의 사용을 최소화 할 수 있었다.

The error back-propagation(BP) algorithm is widely used for finding optimum weights of multi-layer neural networks. However, the critical drawback of the BP algorithm is its slow convergence of error. The major reason for this slow convergence is the premature saturation which is a phenomenon that the error of a neural network stays almost constant for some period time during learning. An inappropriate selections of initial weights cause each neuron to be trapped in the premature saturation state, which brings in slow convergence speed of the multi-layer neural network. In this paper, to overcome the above problem, Micro-Genetic algorithms(.mu.-GAs) which can allow to find the near-optimal values, are used to select the proper weights and slopes of activation function of neurons. The effectiveness of the proposed algorithms will be demonstrated by some computer simulations of two d.o.f planar robot manipulator.

Recently the trends toward reduction in size of industrial products have increased the application of micro drilling. But micro drilling has still much difficulty so that the needs for active control which give adaptation to controller are expanding. In this paper initial cutting condition was determined for some sorkpieces by experiment and GA-based Fuzzy controller was devised by genetic algorithms and fuzzy inference. The fuzzy inference has been applied to the various prob- lems. However the determination of the membership function is one of the difficult problem. So we introduce a genetic algorithms and propose a self-tuning method of fuzzy membership function. Based on this intelligent control, automation of micro drilling was carried out like the cutting process of skilled machinist.

기상수치예보모델의 강수물리과정은 강수 발생과 연관된 입자의 낙하속도, 부착 및 자동전환, 입자크기분포 등의 과정을 다룬다. 하지만 수치예보모델의 미세물리과정과 모수에는 상당한 불확실성이 내포되어 있다. 수치예보모델의 불확실성을 줄이기 위하여 일반적으로 모수 추정을 사용한다. 이 연구에서는 모수 추정을 위한 최적화 알고리즘으로 마이크로 유전알고리즘과 하모니탐색 알고리즘을 사용하고 우리나라에서 발생한 강수사례에 대해 통합모델의 강수물리과정에서 사용하는 모수를 최적화하였다. 두 알고리즘의 서로 다른 특성으로 인해 최적화 과정 중의 차이가 보였다. 마이크로 유전알고리즘은 440회 수행 후 약 1.033의 적합도로 수렴하였고 하모니탐색 알고리즘은 60번 수행 후 약 1.031의 적합도로 수렴하였다. 이를 통해 하모니탐색 알고리즘이 마이크로 유전알고리즘보다 더 빨리 최적의 모수를 탐색하는 것을 알 수 있었다. 따라서 계산비용이 방대한 기상수치예보모델의 최적화 문제에서 빠른 시간 내에 최적의 모수를 탐색해야 한다면 하모니 탐색 알고리즘이 더 적합하다는 것을 확인하였다.

The microphysical processes of the numerical weather prediction (NWP) model cover the following : fall speed, accretion, autoconversion, droplet size distribution, etc. However, the microphysical processes and parameters have a significant degree of uncertainty. Parameter estimation was generally used to reduce errors in NWP models associated with uncertainty. In this study, the micro- genetic algorithm and harmony search algorithm were used as an optimization algorithm for estimating parameters. And we estimate parameters of microphysics for the Unified model in the case of precipitation in Korea. The differences which occurred during the optimization process were due to different characteristics of the two algorithms. The micro-genetic algorithm converged to about 1.033 after 440 times. The harmony search algorithm converged to about 1.031 after 60 times. It shows that the harmony search algorithm estimated optimal parameters more quickly than the micro-genetic algorithm. Therefore, if you need to search for the optimal parameter within a faster time in the NWP model optimization problem with large calculation cost, the harmony search algorithm is more suitable.