Kwon-Su Jeon

Design Optimization Process Using Artificial Neural Networks, Bayesian Learning and Hybrid Algorithm

The Bayesian learning technique, mapped into feed-forward neural networks, is considered as a system approximation, which, for training, highly non-linear and implicit complex functions. This process is integrated with a Hybrid Algorithm (HA) in the proposed design optimization process. The combination of the Back-Propagation Levenberg-Marquardt (BPLM) algorithm and the Bayesian learning technique shows good and accurate generalization, which creates the meta-model, considered as the fitness and constraints function in the hybrid algorithm. Here, a Genetic Algorithm (GA), hybridized with a gradient-based method, performs the effective and robust evolutionary search and reduces the computation cost. D-optimality is used to select the appropriate points in the design space, to obtain the significant responses. A numerical example and the design of a twomember frame are presented to demonstrate the accuracy and feasibility of the process. The triumph of this method and its successful applications will hopefully contribute to the handling of large and complex systems in Multidisciplinary System Design Optimization

Optimal space launcher design using a refined response surface method

To effectively reduce the computational loads during the optimization process, while maintaining the solution accuracy, a refined response surface method with design space transformation and refined RSM using sub-optimization for the regression model is proposed and implemented for the nose fairing design of a space launcher. Total drag is selected as the objective function, and the surface heat transfer, the fineness ratio, and the internal volume of the nose fairing are considered as design constraints. Sub-optimization for the design space transformation parameters and the iterative regression model construction technique are proposed in order to build response surface with high confidence level using minimum number of experiment points. The derived strategies are implemented to the nose fairing design optimization using the full Navier-Stokes equations. The result shows that an optimum nose fairing shape is obtained with four times less analysis calculations compared with the gradient-based optimization method, and demonstrates the efficiency of the refined response surface method and optimization strategies proposed in this study. The techniques can be directly applied to the multidisciplinary design and optimization problems with many design variables.

Multidisciplinary Design Approach Using Repetitive Response Surface Enhancement and Global Optimization

In this study, repetitive response surface enhancement technique (RRSET) is proposed as a new system approximation method for the efficient multidisciplinary design and optimization (MDO). In order to represent the highly nonlinear behavior of the response with second order polynomials, RRSET introduces a design space transformation using stretching functions and repetitive response surface improvement. The tentative optimal point is repetitively included to the set of experimental points to better approximate the response surface of the system especially near the optimal point, hence a response surface with significantly improved accuracy can be generated with very small experimental points and system iterations. As a system optimizer, the simulated annealing, a global optimization algorithm is utilized. The proposed technique is applied to the several numerical examples, and demonstrates the validity and efficiency of the method. With its improved approximation accuracy, the RRSET can contribute to resolve large and complex system design problems under MDO environment.

A LOGICAL APPROACH FOR EVALUATING PRELIMINARY SHAPE DESIGN OF A VERY LIGHT JET

In this paper, a design approach is presented to determine a reference shape by using decision making process on the preliminary aircraft design. Itemized design criteria and weight factors are proposed. Those are necessary for evaluating alternative shapes of aircraft. The proposed design approach incorporates three decision making steps those using AHP, QFD, and TOPSIS. These wellknown methods help a designer determine both the system requirements and the design criteria. To validate the proposed design approach, a design of Very Light Jet (VLJ) aircraft is applied for finding reference configuration by using the results of conceptual design. According to the results, this procedure can be adopted not only to obtain the airplane shape for satisfying design criteria but also to set the reference design values for corresponding design characteristics such as assessment items, weight factors The contribution of this paper is characterized such that the designer can analyze the system requirements and incorporate them during the design process. Thus, the designer can be able to achieve the design goal while effectively determining alternative shapes in accordance with information provided by the proposed methodology.

Rotor design for the performance optimization of canard rotor/wing aircraft

A program for the sizing and performance analysis is developed for Canard Rotor/Wing (CRW) aircraft which operates in both fixed wing and rotary wing modes. The system characteristics, such as reaction driven rotor system, are analyzed first and then the system design process is defined. The developed program is verified for both fixed wing and rotary wing modes with existing aircraft data and the design optimization process is performed for a reconnaissance mission. For the CRW aircraft optimization for both fixed wing and rotary wing modes, a multi-objective function is constructed using weighting factors. For several design cases with different weighting factors and several design constraints, the optimization analysis is performed and improved results are derived.

Optimal Gas Generator Design for the Liquid Rocket Engine

The optimal design of gas generator was conducted. Turbo pump system with fuel rich combustion in gas generator was the system considered in this study. The objective function is the specific impulse of main thrust chamber operating with RP-11LOX propellants. Sensitivity analyses to the objective function were used to measure the relative dependence of design variables the objective function. Also optimization method was adopted with two different algorithms; GBM (Gradient based method) and non-GBM. GBM includes SLP, MMFD, and SQP algorithms and the result of GBM were compared with non-GBM result; SA (Simulated Annealing) method. As a result, it was confirmed that SA could yield a global optimization regardless of initial variations. Whereas GBM showed the results were dependent on the initial conditions. And the global optimization could provide a better configuration of gas generator in the design stage. The present study aims at the optimization of gas generator design. Thus optimization methodology was tested with variety of optimization technique applied to the gas generator design with various constraints. Also, the present study compares the results from GBM algorithms with that of SA in order to check the validity of global solution

Development of Repetitively Enhanced Neural Networks (RENN) for Efficient Missile Design and Optimization

An improved approach for design optimization of air intercept missile is developed and presented. A Bayesian learning technique is mapped into Back-propagation neural networks (BPNN) to establish an accurate and effective system approximation, namely an enhanced neural network module. Then, the surrogate models are generated and sent to a hybrid optimizer in which a tentative optimum result is obtained and updated into the training data to refine the response surfaces. This process, which is called Repetitively Enhanced Neural Networks (RENN), is executed repeatedly to refine the response surface until the convergent optimum solution is obtained. A numerical example and a two-member frame design are presented and discuss to demonstrate the accuracy and feasibility of RENN. Eventually, this RENN approach is applied to re-design the air intercept missile-AIM

Rotor Performance Optimization of the Canard Rotor Wing Aircraft

In this study, the sizing and performance analysis program is developed for the canard rotor wing(CRW) aircraft which operates in dual modes (fixed wing mode and rotary wing mode). The developed program is verified for both fixed wing and rotary wing modes using the existing aircraft data and the design optimization formulation is made to perform the reconnaissance mission. For the canard rotor wing aircraft optimization , multi-objective function is constructed to consider both the fixed wing mode and rotary wing mode the weighting factor. For six design cases with different weighting factors and different design constraints, the optimization is performed and improved rotor design results are derived.

Performance optimization of hypervelocity launcher system using experimental data

This study presents the performance optimization of hypervelocity launcher system by using the experimentall data. During the optimization, the RSM (Response Surface Method) is adopted to find the operating parameters that could maximize the projectile speed. To construct a reliable response surface model, 3 full factorial method is used with the selected design variables, such as piston mass and 2 driver fill pressure. Nine test data could successfully construct the reasonable response surface, which used to yield the optimal operational conditions of the system using the genetic algorithm. The optimization results are confirmed by the experimental test with a good accuracy. Thus, the optimization can improve the performance of the facility.

The repetitive optimization design strategy using neural network and hybrid algorithm

In this paper, a Bayesian learning technique, mapped into feed-forward artificial neural networks, is considered as a system approximation, which, for training, highly non-linear and implicit complex functions. This process is integrated with a hybrid algorithm (HA) in the proposed design optimization strategy. The combination of the back-propagation Levenberg-Marquardt (BPLM) algorithm and the Bayesian learning technique shows good and accurate generalization, which creates the meta-model, considered as the fitness and constraints function in the hybrid algorithm. Here, a genetic algorithm (GA), hybridized with a local gradient-based method, performs the effective and robust evolutionary search and reduces the computation cost. D-optimality is used to select the appropriate points in the design space, to obtain the significant responses. A numerical example, the design of a two-member frame and air intercept missile-AIM design optimization problem are presented to demonstrate the accuracy and feasibility of the process.