Navigating Expensive and Complex Design Spaces Using Genetic Algorithms

1999 ◽  
Author(s):  
David C. Zimmerman
Keyword(s):  
Function Approximation ◽  
Design Space ◽  
Optimization Technique ◽  
Computational Cost ◽  
Approximation Procedure ◽  
Network Approach ◽  
Neural Network Approach ◽  
Design Problems ◽  
Complex Design ◽  
Single Function

Abstract The overall objective of this study is to formulate and study a generic procedure for navigating expensive and complex design spaces. The term generic is meant to imply that the procedure would be equally valid in exploring design problems in a multitude of fields. The term expensive design space implies that the computational cost, or burden, associated with a single function is considered “large”. What is desired is a methodology which can identify “promising regions” of the design space using as few function evaluations as possible. To approach this problem, a neural network approach is developed to serve as an inexpensive and generic function approximation procedure. The genetic algorithm was selected as the optimization technique based on its ability to search multi-modal, discontinuous, mixed parameter, and noisy design spaces.

Multi-Objective Optimization of an Autonomous Underwater Vehicle

2009 ◽  
Vol 43 (2) ◽  
pp. 48-60 ◽  
Author(s):  
M. Martz ◽  
W. L. Neu
Keyword(s):  
Design Space ◽  
Autonomous Underwater Vehicle ◽  
Underwater Vehicle ◽  
Design Problems ◽  
Single Measure ◽  
Design Variables ◽  
Complex Design ◽  
Final Design ◽  
Undersea Vehicle ◽  
Hierarchy Process

AbstractThe design of complex systems involves a number of choices, the implications of which are interrelated. If these choices are made sequentially, each choice may limit the options available in subsequent choices. Early choices may unknowingly limit the effectiveness of a final design in this way. Only a formal process that considers all possible choices (and combinations of choices) can insure that the best option has been selected. Complex design problems may easily present a number of choices to evaluate that is prohibitive. Modern optimization algorithms attempt to navigate a multidimensional design space in search of an optimal combination of design variables. A design optimization process for an autonomous underwater vehicle is developed using a multiple objective genetic optimization algorithm that searches the design space, evaluating designs based on three measures of performance: cost, effectiveness, and risk. A synthesis model evaluates the characteristics of a design having any chosen combination of design variable values. The effectiveness determined by the synthesis model is based on nine attributes identified in the U.S. Navy’s Unmanned Undersea Vehicle Master Plan and four performance-based attributes calculated by the synthesis model. The analytical hierarchy process is used to synthesize these attributes into a single measure of effectiveness. The genetic algorithm generates a set of Pareto optimal, feasible designs from which a decision maker(s) can choose designs for further analysis.

NEURAL NETWORK-BASED DERIVATION OF EFFICIENT HIGH-ORDER RUNGE–KUTTA–NYSTRÖM PAIRS FOR THE INTEGRATION OF ORBITS

2011 ◽  
Vol 22 (12) ◽  
pp. 1309-1316
Author(s):  
I. TH. FAMELIS
Keyword(s):  
Neural Network ◽  
Kepler Problem ◽  
Optimization Technique ◽  
Network Approach ◽  
Specific Test ◽  
Neural Network Approach ◽  
Orbital Problems ◽  
Free Parameters ◽  
Runge Kutta ◽  
Specific Tolerance

We use a neural network approach to derive a Runge–Kutta–Nyström pair of orders 8(6) for the integration of orbital problems. We adopt a differential evolution optimization technique to choose the free parameters of the method's family. We train the method to perform optimally in a specific test orbit from the Kepler problem for a specific tolerance. Our measure of efficiency involves the global error and the number of function evaluations. Other orbital problems are solved to test the new pair.

scholarly journals Hybrid Neural Network Approach Based Tool for the Modelling of Photovoltaic Panels

2015 ◽  
Vol 2015 ◽  
pp. 1-10 ◽  
Author(s):  
Antonino Laudani ◽  
Gabriele Maria Lozito ◽  
Francesco Riganti Fulginei ◽  
Alessandro Salvini
Keyword(s):  
Neural Network ◽  
Hybrid Approach ◽  
Computational Cost ◽  
Point Of View ◽  
Reduced Form ◽  
Network Approach ◽  
Neural Network Approach ◽  
Hybrid Neural Network ◽  
The Neural Network ◽  
Two Parameters

A hybrid neural network approach based tool for identifying the photovoltaic one-diode model is presented. The generalization capabilities of neural networks are used together with the robustness of the reduced form of one-diode model. Indeed, from the studies performed by the authors and the works present in the literature, it was found that a direct computation of the five parameters via multiple inputs and multiple outputs neural network is a very difficult task. The reduced form consists in a series of explicit formulae for the support to the neural network that, in our case, is aimed at predicting just two parameters among the five ones identifying the model: the other three parameters are computed by reduced form. The present hybrid approach is efficient from the computational cost point of view and accurate in the estimation of the five parameters. It constitutes a complete and extremely easy tool suitable to be implemented in a microcontroller based architecture. Validations are made on about 10000 PV panels belonging to the California Energy Commission database.

Airfoil Flow Separation and Stall Prediction Using a Neural-Network Approach

2010 ◽  
Author(s):  
Amir Nejat ◽  
Saeed Alem Varzaneh Esfehani ◽  
Amin Doostmohammadi
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Flow Separation ◽  
Lift Coefficient ◽  
Computational Cost ◽  
Base Line ◽  
Control Points ◽  
Network Approach ◽  
Neural Network Approach ◽  
Artificial Neural

An Artificial Neural Network approach is applied to predict the lift coefficient of the airfoil for different angles of attack at the specified Reynolds number for airfoils with different shape characteristics. The Stall of the airfoil is characterized by a sudden drop in the lift force caused by flow separation. Therefore by prediction of the lift coefficient for different angles of attack and via determining its maximum, the stall angle for a specific shape of the airfoil can be determined. The Computational Fluid Dynamics is used to provide the initial data needed for ANN training; the lift coefficient of the airfoil for the range of angles of attack is taken as a target and NACA0012 is considered as a base line model in the present study. The profile of the airfoil is constructed by Be´zier curves with eight control points and these control points along angles of attack are adopted as the training data for the Artificial Neural Network training. The ANN results are compared with those obtained by CFD. The results show the ANN approach can accurately predict the lift coefficient for the preliminary design process at substantially lower computational cost.

Constraint Dominance Determination Methods

2003 ◽  
Author(s):  
James R. Rinderle ◽  
Ashish D. Deshpande
Keyword(s):  
Design Space ◽  
Transformation Method ◽  
Design Problems ◽  
Interval Method ◽  
Identification Methods ◽  
Complex Design ◽  
Range Determination ◽  
Determination Methods ◽  
Function Range ◽  
Critical Aspects

Dominance among constraints exists when the satisfaction of a constraint guarantees the satisfaction of another, rendering the second constraint irrelevant. Identifying dominance not only facilitates numerical solution but may also focus the designer’s attention on critical aspects of the design. A number of dominance identification methods have been described in the literature, including the Constraint Difference Method, the Constraint Transformation Method, and the Necessary-Sufficient Interval Method. We elaborate on the basis for and the character of these methods and we discuss relative similarities, differences, strengths, and weaknesses of the methods. We also discuss computational issues relevant to the application of these methods, most specifically function range determination and interval analysis issues. We observe that the differences among the methods lead to advantages for each method in circumstances that depend on the nature of the constraints and the extent of the design space. These distinct advantages suggest a synergism among the methods in the identification of constraint dominance in complex design problems.

Assessing the Effectiveness of Using Graveyard Data for Generating Design Alternatives

2011 ◽  
Author(s):  
Garrett Foster ◽  
Scott Ferguson
Keyword(s):  
Simplex Method ◽  
Design Space ◽  
Computational Cost ◽  
Initial Point ◽  
Sampling Procedure ◽  
Design Problems ◽  
Search Techniques ◽  
Design Alternatives ◽  
Engineering Design Problems ◽  
Nonlinear Design

Modeling to Generate Alternatives (MGA) is a technique used to identify variant designs that maximize design space distance from an initial point while satisfying performance loss constraints. Recent work has explored the application of this technique to nonlinear design problems, where the design space was investigated using an exhaustive sampling procedure. While computational cost concerns were noted, the main focus was determining how scaling and distance metric selection influenced alternative discovery. To increase the viability of MGA for engineering design problems, this work looks to reduce the computational overhead needed to identify design alternatives. This paper investigates and quantifies the effectiveness of using previously sampled designs, i.e. a graveyard, from a multiobjective genetic algorithm as a means of reducing computational expense. Computational savings and the expected error are quantified to assess the effectiveness of this approach. These results are compared to other more common “search” techniques; namely Latin hypercube samplings, grid search, and the Nelder-Mead simplex method. The performance of these “search” techniques are subsequently explored in two case study problems — the design of a two bar truss, and an I-beam — to find the most unique alternative design over a range of different thresholds. Results from this work show the graveyard can be used as a way of inexpensively generating alternatives that are close to ideal, especially nearer to the starting design. Additionally, this paper demonstrates that graveyard information can be used to increase the performance of the Nelder-Mead simplex method when searching for alternative designs.

Bayesian Network Structure Optimization for Improved Design Space Mapping for Design Exploration With Materials Design Applications

2017 ◽  
Author(s):  
Conner Sharpe ◽  
Clinton Morris ◽  
Benjamin Goldsberry ◽  
Carolyn Conner Seepersad ◽  
Michael R. Haberman
Keyword(s):  
Machine Learning ◽  
Bayesian Network ◽  
Design Space ◽  
Computational Cost ◽  
Materials Design ◽  
Optimization Techniques ◽  
Machine Learning Algorithms ◽  
Data Set ◽  
Design Problems ◽  
Highly Nonlinear

Modern design problems present both opportunities and challenges, including multifunctionality, high dimensionality, highly nonlinear multimodal responses, and multiple levels or scales. These factors are particularly important in materials design problems and make it difficult for traditional optimization algorithms to search the space effectively, and designer intuition is often insufficient in problems of this complexity. Efficient machine learning algorithms can map complex design spaces to help designers quickly identify promising regions of the design space. In particular, Bayesian network classifiers (BNCs) have been demonstrated as effective tools for top-down design of complex multilevel problems. The most common instantiations of BNCs assume that all design variables are independent. This assumption reduces computational cost, but can limit accuracy especially in engineering problems with interacting factors. The ability to learn representative network structures from data could provide accurate maps of the design space with limited computational expense. Population-based stochastic optimization techniques such as genetic algorithms (GAs) are ideal for optimizing networks because they accommodate discrete, combinatorial, and multimodal problems. Our approach utilizes GAs to identify optimal networks based on limited training sets so that future test points can be classified as accurately and efficiently as possible. This method is first tested on a common machine learning data set, and then demonstrated on a sample design problem of a composite material subjected to a planar sound wave.

A fuzzy neural network approach for modeling the growth kinetics of FeB and Fe2B layers during the boronizing process

2018 ◽  
Vol 106 (6) ◽  
pp. 603 ◽  
Author(s):  
Bendaoud Mebarek ◽  
Mourad Keddam
Keyword(s):  
Neural Network ◽  
Experimental Data ◽  
Fuzzy Neural Network ◽  
Treatment Time ◽  
Calculation Model ◽  
Average Error ◽  
Network Approach ◽  
Neural Network Approach ◽  
Fuzzy Neural ◽  
Kinetics Of

In this paper, we develop a boronizing process simulation model based on fuzzy neural network (FNN) approach for estimating the thickness of the FeB and Fe2B layers. The model represents a synthesis of two artificial intelligence techniques; the fuzzy logic and the neural network. Characteristics of the fuzzy neural network approach for the modelling of boronizing process are presented in this study. In order to validate the results of our calculation model, we have used the learning base of experimental data of the powder-pack boronizing of Fe-15Cr alloy in the temperature range from 800 to 1050 °C and for a treatment time ranging from 0.5 to 12 h. The obtained results show that it is possible to estimate the influence of different process parameters. Comparing the results obtained by the artificial neural network to experimental data, the average error generated from the fuzzy neural network was 3% for the FeB layer and 3.5% for the Fe2B layer. The results obtained from the fuzzy neural network approach are in agreement with the experimental data. Finally, the utilization of fuzzy neural network approach is well adapted for the boronizing kinetics of Fe-15Cr alloy.

Sign in / Sign up

Export Citation Format

Share Document