Introduction

RecurDyn/AutoDesign is an efficient design optimization tool for RecurDyn, which integrates the design of experiments, meta-model techniques and numerical optimization techniques.

Unlike other design optimization software, AutoDesign can successfully integrate the meta-model and numerical optimization techniques in the context of automated design optimization process. It requires nearly minimum sampling points for constructing the initial meta-models and improves the fidelity of meta-models by automatic switching the regression models during optimization process. It can overcome the numerical singularity for insufficient data when constructing meta-model. It even overcomes the analysis failure of RecurDyn during optimization process. Also, the optimization techniques are automatically selected for the given design formulation. Thus, the user just defines his design problem.

AutoDesign provides the following six menus:

 is the Design Parameters selection, which can select all parametric values in RecurDyn. It even includes:

•     FE Nodes in RecurDyn/FFLEX

•     Gain values in RecurDyn/CoLink

 is the Analysis Response selection, which helps one to select the dynamic responses as the performance index for design optimization. The user can select all analysis responses by using all expressions provided in RecurDyn.

 is the Design Study module, which can provide the effect analysis, design variable screening, correlation analysis and variation analysis charts. It provides level-balanced DOE methods such as :

•     Full Factorial Design

•     Plackett-Burman Design

•     2-Level Orthogonal Array

•     3-Level orthogonal Array

•     Bose’s Orthogonal Array

•     Level-Balanced Descriptive design

The last one is a strength-I orthogonal array design. Unlike other design tools, AutoDesign can automatically generate DOE table for the user-defined number of design variables.

 is the Design Optimization module, which can formulate the design optimization problem by using the pre-defined design parameters and analysis responses. As AutoDesign is a meta-model based optimization tool, It provides the meta-model techniques such as :

•     Simultaneous kriging

•     Radial basis Functions

•     Conservative RSM

Unlike other kriging techniques, the simultaneous Kriging solves only one sub-optimization problem when constructing meta-models for multiple performance indexes. So, it is called as Simultaneous kriging to be distinguished from the conventional kriging methods. It is a unique technique in AutoDesign. Also, in this module, rotatable DOE methods and space filling DOE methods are provided for meta-models. They are efficient DOE methods and Classical DOE methods. For practical design, it recommends efficient DOE methods such as:

•     Discrete Latin-hypercube design

•     Incomplete Small Composite Design-I (ISCD-I)

•     Incomplete Small Composite Design–II (ISCD-II)

In the classical DOE methods, a Generalized Small Composite Design (GSCD), Faced Central Composite Design and Box and Behnken designs are provided. Among them, ISCD-I, ISCD-II and GSCD are unique methods in AutoDesign. To help one’s choice for DOE methods and Meta-model method, AutoDesign marks the recommended methods.

 is the DFSS and Robust Design Optimization module. Most of design optimization tools provides Taguchi method for a robust design and uses the MonteCarlo simulation for checking the DFSS (Design For Six Sigma) criterion. Thus, they don’t satisfy the practical designer’s requests that find the design to satisfy the DFSS criterion. From the view of design optimization, AutoDesign can solve the robust design problem including 6-sigma constraints. Like Design Optimization module, it constructs the meta-models. Then, it internally evaluates the variance of responses for the random design variables and noise variables. If one defines the robust optimization problem and 6-sigma constraints, AutoDesign solve the approximate problems. In the sub-menu for the DFSS/Robust design formulation, one can define the design variables as deterministic or random variables. Also, he/she can select variable or constant. For example, a random constant becomes a noise factor in Taguchi method. It is noted that 6-sigma constraints can make no feasible region in the design space, even though optimizers try to satisfy 6-sigma constraints as possible as they can. In this case, most of design tools run until maximum iteration is reached. Then, they give that there is no feasible solution. AutoDesign, however, gives a feasible k-sigma design by automatic relaxation of 6-sigma constraints to guarantee feasible region. The value of k can become a nearest real value less than 6.

 is the Simulation History module. When one uses AutoDesign, all the analysis results are stored in the simulation history. Thus, one can use the design study results when initial constructing meta-models for design optimization and DFSS/Robust design optimization modules. Also, when the first try for design optimization is failed, one can use these results for re-optimization process, which can reduce the number of analyses. Finally, in this module, one can create the optimum design result as a new analysis model file.