• Proceedings of the KSME Conference
• /
• 2007.05b
• /
• pp.2552-2557
• /
• 2007

Shape optimization of an internal cooling passage with staggered dimples on single surface is performed and performances of surrogates are evaluated in this paper. Optimizations are performed so that turbulent heat transfer can be enhanced compromising with pressure loss due to friction. The three-dimensional governing differential equations have been solved to find the overall Nusselt number and friction factor which are related to the objective functions of this problem. Three design variables were selected among the dimensionless geometric variables. Basic surrogate models such as second order polynomial response surface approximation (RSA), Kriging meta-modeling technique, radial basis neural network (RBNN), and derived press based averaged (PBA) surrogate model are constructed. The optimal points are searched from the above constructed surrogates by sequential quadratic programming (SQP). It is shown that use of multiple surrogates can increase the robustness in prediction of better design with minimum computational cost.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.06a
• /
• pp.163-168
• /
• 2005

It is one of the most important things to determinate optimized cutting conditions which satisfy productivity and cost simultaneously in production and CAPP systems. These days many researchers have figured out the optimizing way for solutions of multi-object function to find the approach methods using algorithm such as genetic algorithm or tabu search, etc., instead of mathematical methods. The main creation of objective function is proposed by empirical method but which is difficult to set it up and to analysis. In this paper, an optimization method of cutting condition is shown using the ANN and GA for the multi-objective function in high speed machining.

• ETRI Journal
• /
• v.35 no.6
• /
• pp.1168-1171
• /
• 2013

We propose an improvement of particle swarm optimization (PSO) based on the stabilization of particle movement (PM). PSO uses a stochastic variable to avoid an unfortunate state in which every particle quickly settles into a unanimous, unchanging direction, which leads to overshoot around the optimum position, resulting in a slow convergence. This study shows that randomly located particles may converge at a fast speed and lower overshoot by using the proportional-integral-derivative approach, which is a widely used feedback control mechanism. A benchmark consisting of representative training datasets in the domains of function approximations and pattern recognitions is used to evaluate the performance of the proposed PSO. The final outcome confirms the improved performance of the PSO through facilitating the stabilization of PM.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.1 no.1
• /
• pp.35-43
• /
• 2001

In this paper, we propose a design method for GBSB (generalized brain-state-in-a-box) based associative memories. Based on the theoretical investigation about the properties of GBSB, we parameterize the solution space utilizing the limited number of parameters sufficient to represent the solution space and appropriate to be searched. Next we formulate the problem of finding a GBSB that can store the given pattern as stable states in the form of constrained optimization problems. Finally, we transform the constrained optimization problem into a SDP(semidefinite program), which can be solved by recently developed interior point methods. The applicability of the proposed method is illustrated via design examples.

• The KSFM Journal of Fluid Machinery
• /
• v.10 no.3 s.42
• /
• pp.39-46
• /
• 2007

This study presents a numerical procedure to optimize the shape of staggered dimple surface to enhance turbulent heat transfer in a rectangular channel. The RBNN method is used as an optimization technique with Reynolds-averaged Navier-Stokes analysis of fluid flow and heat transfer with shear stress transport (SST) turbulence model. The dimple depth-to-dimple print diameter (d/D), channel height-to-dimple print diameter ratio (H/D), and dimple print diameter-to-pitch ratio (D/S) are chosen as design variables. The objective function is defined as a linear combination of heat transfer related term and friction loss related term with a weighting factor. Latin Hypercube Sampling (LHS) is used to determine the training points as a mean of the design of experiment. The optimum shape shows remarkable performance in comparison with a reference shape.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.29 no.6 s.237
• /
• pp.809-814
• /
• 2005

The goal of this paper is to minimize the weight of the T-structure supporting steering system in reducing the vibration level on steering wheel which could be amplified by the resonance. Presently, requirements for reducing noise, vibration and harshness (NVH) in automotive area are more stringent than ever. One of them is the vibration of steering system which occurs sometimes at high speeds or when the engine is idling. Besides, the reduction of weight is also one of requirements for improvement of vehicle performance. This paper used the micro genetic algorithm as an optimization method to satisfy above two requirements. The whole T-structure assembly including steering column was used for frequency analysis.

• 한국전산유체공학회:학술대회논문집
• /
• 2007.04a
• /
• pp.159-162
• /
• 2007

This study presents a numerical procedure to optimize the shape of a staggered dimpled surface to enhance the turbulent heat transfer in a rectangular channel. A optimization technique based on neural network is used with Reynolds-averaged Navier-Stakes analysis of the fluid flow and heat transfer with Shear Stress Transport turbulence model. The dimple depth-to-dimple print diameter ratio, channel height-to-dimple print diameter ratio, and dimple print diameter-to-pitch ratio are chosen as design variables. The objective function is defined as a linear combination of terms related to heat transfer and friction loss with a weighting factor. Latin Hypercube Sampling is used to determine the training points as a mean of the Design of Experiment. Optimal values of the design variables were obtained in a range of the weighting factor.

• Journal of computational fluids engineering
• /
• v.14 no.4
• /
• pp.31-40
• /
• 2009

Although optimization by sequentially refining metamodels is known to be computationally very efficient, the metamodel that can be used for this purpose is limited to Kriging method due to the difficulties related with sample points selections. The present study suggests a novel method for sequentially refining metamodels using Pareto Frontiers, which can be used independent of the type of metamodels. It is shown from the examples that the present method yields more accurate metamodels compared with full-factorial optimization and also guarantees global optimum irrespective of the initial conditions. Finally, in order to prove the generality of the present method, it is applied to a 2D transonic airfoil optimization problem, and the successful design results are obtained.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.32 no.12
• /
• pp.954-962
• /
• 2008

This study presents a numerical procedure to optimize the shape of cylindrical cooling hole to enhance film-cooling effectiveness. The RBNN method is used as an optimization technique with Reynolds-averaged Navier-Stokes analysis of fluid flow and heat transfer with shear stress transport turbulent model. The hole length-to-diameter ratio and injection angle are chosen as design variables and film-cooling effectiveness is considered as objective function which is to be maximized. Twelve training points are obtained by Latin Hypercube Sampling for two design variables. In the sensitivity analysis, it is found that the objective function is more sensitive to the injection angle of hole than the hole length-to diameter ratio. Optimum shape gives considerable increase in film-cooling effectiveness.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.4
• /
• pp.1738-1764
• /
• 2019

With the advent of big data, deep learning technology has become an important research direction in the field of machine learning, which has been widely applied in the image processing, natural language processing, speech recognition and online advertising and so on. This paper introduces deep learning techniques from various aspects, including common models of deep learning and their optimization methods, commonly used open source frameworks, existing problems and future research directions. Firstly, we introduce the applications of deep learning; Secondly, we introduce several common models of deep learning and optimization methods; Thirdly, we describe several common frameworks and platforms of deep learning; Finally, we introduce the latest acceleration technology of deep learning and highlight the future work of deep learning.