Employing Knowledge on Causal Relationship to Assist Multidisciplinary Design Optimization

2019 ◽  
Vol 141 (4) ◽  
Author(s):  
Di Wu ◽  
Eric Coatanea ◽  
G. Gary Wang
Keyword(s):  
Qualitative Analysis ◽  
Design Optimization ◽  
Causal Relationship ◽  
Design Problem ◽  
Multidisciplinary Design Optimization ◽  
Power Converter ◽  
Multidisciplinary Design ◽  
Concept Design ◽  
Causal Graph ◽  
The Impact

With the increasing design dimensionality, it is more difficult to solve multidisciplinary design optimization (MDO) problems. Many MDO decomposition strategies have been developed to reduce the dimensionality. Those strategies consider the design problem as a black-box function. However, practitioners usually have certain knowledge of their problem. In this paper, a method leveraging causal graph and qualitative analysis is developed to reduce the dimensionality of the MDO problem by systematically modeling and incorporating the knowledge about the design problem into optimization. Causal graph is created to show the input–output relationships between variables. A qualitative analysis algorithm using design structure matrix (DSM) is developed to automatically find the variables whose values can be determined without resorting to optimization. According to the impact of variables, an MDO problem is divided into two subproblems, the optimization problem with respect to the most important variables, and the other with variables of lower importance. The novel method is used to solve a power converter design problem and an aircraft concept design problem, and the results show that by incorporating knowledge in form of causal relationship, the optimization efficiency is significantly improved.

Application of Multidisciplinary Design Optimization to the Power Stage Design of a Power Converter

1993 ◽  
Author(s):  
Gregory Kott ◽  
Gary A. Gabriele ◽  
Jacob Korngold
Keyword(s):  
Design Optimization ◽  
Design Problem ◽  
Multidisciplinary Design Optimization ◽  
Power Converter ◽  
Multidisciplinary Design ◽  
Successful Implementation ◽  
Total System ◽  
Global Approximation ◽  
Stage Design ◽  
Power Stage

Abstract This paper describes the application of multidisciplinary design optimization to the power stage design of a power converter. The decomposition of the power stage design into an electrical and a loss subsystem is developed. The Sequential Global Approximation method is the non-hierarchic algorithm used to optimize the power stage design problem. Results of the non-hierarchic formulation compared to the non-decomposed formulation show a decrease of 63 percent in total system iterations required to converge to the optimal solution. Local and global move limits of 28 percent were found to provide the best performance for this problem. The successful implementation and results of applying multidisciplinary design optimization to power stage design allows the extension of the research to incorporate other disciplines. Our goal is to include all disciplines to completely model the design of a power converter. The details of power stage design problem formulation are provided to be used as a test problem in multidisciplinary design optimization research.

Dimension Reduction and Decomposition Using Causal Graph and Qualitative Analysis for Aircraft Concept Design Optimization

2017 ◽  
Author(s):  
Di Wu ◽  
Eric Coatanea ◽  
G. Gary Wang
Keyword(s):  
Qualitative Analysis ◽  
Dimension Reduction ◽  
Design Optimization ◽  
Design Problem ◽  
Concept Design ◽  
Causal Graph ◽  
Design Structure ◽  
Decomposition Strategies ◽  
The One ◽  
Optimization Efficiency

With the increasing design dimensionality, it is more difficult to solve Multidisciplinary design optimization (MDO) problems. To reduce the dimensionality of MDO problems, many MDO decomposition strategies have been developed. However, those strategies consider the design problem as a black-box function. In practice, the designers usually have certain knowledge of their problem. In this paper, a method leveraging causal graph and qualitative analysis is developed to reduce the dimensionality of the MDO problem by systematically modeling and incorporating knowledge of the design problem. Causal graph is employed to show the input-output relationships between variables. Qualitative analysis using design structure matrix (DSM) is carried out to automatically find the variables that can be determined without optimization. According to the weight of variables, the MDO problem is divided into two sub-problems, the optimization problem with respect to important variables, and the one with less important variables. The novel method is performed to solve an aircraft concept design problem and the results show that the new dimension reduction and decomposition method can significantly improve optimization efficiency.

A Multidisciplinary Design Optimization Approach to Relating Affordability and Performance in a Conceptual Submarine Design

2010 ◽  
Vol 26 (04) ◽  
pp. 273-289 ◽  
Author(s):  
N. Vlahopoulos ◽  
C. G. Hart
Keyword(s):  
Design Optimization ◽  
Multidisciplinary Design Optimization ◽  
Geometric Model ◽  
Multidisciplinary Optimization ◽  
System Level ◽  
Multidisciplinary Design ◽  
Manufacturing Cost ◽  
Optimization Approach ◽  
And Performance ◽  
The Impact

A multidisciplinary design optimization (MDO) framework is used for a conceptual submarine design study. Four discipline-level performances—internal deck area, powering, maneuvering, and structural analysis—are optimized simultaneously. The four discipline-level optimizations are driven by a system level optimization that minimizes the manufacturing cost while at the same time coordinates the exchange of information and the interaction among the discipline-level optimizations. Thus, the interaction among individual optimizations is captured along with the impact of the physical characteristics of the design on the manufacturing cost. A geometric model for the internal deck area of a submarine is created, and resistance, structural design, and maneuvering models are adapted from theoretical information available in the literature. These models are employed as simulation drivers in the discipline-level optimizations. Commercial cost-estimating software is leveraged to create a sophisticated, automated affordability model for the fabrication of a submarine pressure hull at the system level. First, each one of the four discipline optimizations and also the cost-related top level optimization are performed independently. As expected, five different design configurations result, one from each analysis. These results represent the "best" solution from each individual discipline optimization, and they are used as reference for comparison with the MDO solution. The deck area, resistance, structural, maneuvering, and affordability models are then synthesized into a multidisciplinary optimization statement reflecting a conceptual submarine design problem. The results from this coordinated MDO capture the interaction among disciplines and demonstrate the value that the MDO system offers in consolidating the results to a single design that improves the discipline-level objective functions while at the same time produces the highest possible improvement at the system level.

Power Converter Design With Coupled Thermal Analysis Using Multidisciplinary Design Optimization

1994 ◽  
Author(s):  
Gregory Kott ◽  
Gary A. Gabriele ◽  
Jacob Korngold
Keyword(s):  
Thermal Analysis ◽  
Design Optimization ◽  
Multidisciplinary Design Optimization ◽  
Power Converter ◽  
Multidisciplinary Design ◽  
Total System ◽  
Global Approximation ◽  
Stage Design ◽  
Power Stage ◽  
Percent Decrease

Abstract This paper describes the application of multidisciplinary design optimization to the power stage design of a power converter. Multidisciplinary design is used to integrate the electrical, loss, and thermal analyses into one system problem. The Sequential Global Approximation method, a non-hierarchic algorithm, is used to optimize the power stage design problem. The code used for the thermal analysis, COSMOS/M, runs externally to the Sequential Global Approximation algorithm. A comparison of the results of the non-hierarchic formulation and the non-decomposed formulation shows a 67 percent decrease in total system iterations and a 12 percent decrease in total finite element analyses required.

Multidisciplinary Design Optimization of Link Mechanisms by Combining Decomposition and Mini-Max Relaxation

2000 ◽  
Author(s):  
Kikuo Fujita ◽  
Noriyasu Hirokawa ◽  
Masafumi Inoue
Keyword(s):  
Design Optimization ◽  
Design Problem ◽  
Multidisciplinary Design Optimization ◽  
Beam Theory ◽  
Coupled System ◽  
Optimization Method ◽  
Multidisciplinary Design ◽  
Strength Analysis ◽  
Link Mechanism ◽  
Conventional Optimization

Abstract This paper proposes an efficiency improvement method for the multidisciplinary design optimization of link mechanisms. While behavior of a link mechanism is dominated by kinematic relationships, it is restricted under various other conditions such as structural strength, etc. Thus, the design problem of a link mechanism is obviously multidisciplinary. Further, since the motion of a link mechanism is sequentially time-dependent, the optimization problem can be formulated as a mini-max type one. Under these characteristics, this paper proposes a design optimization method of link mechanisms by combining non-hierarchic coupled system decomposition and mini-max relaxation. Then, after the multidisciplinary design problem of a link mechanism used in hydraulic shovels is modeled and formulated based on skeleton based kinematic analysis and beam-theory based strength analysis, its design optimization with the proposed method is demonstrated. The numerical results show more than about four times of speedup and robust performance as compared with a conventional optimization method.

Sign in / Sign up

Export Citation Format

Share Document