A parallel computing framework for performing structural-acoustic optimization with stochastic forcing

Research output: Contribution to journalArticle

Abstract

Structural-acoustic optimization procedures can be used to find the optimal design for reduced noise or vibration in many real-world scenarios. However, the time required to compute the structural-acoustic quantity of interest often limits the size of the model. Additionally, structural-acoustic optimization using state-of-the-art evolutionary algorithms may require tens of thousands of system solutions, which add to the limitations for large full-scale systems. To reduce the time required for each function evaluation, parallel processing techniques are used to solve the system in a highly scalable fashion. The approach reduces the analysis time by solving the system using a frequency-domain formulation and distributing solution frequencies amongst processors to solve in parallel. To demonstrate, the sound radiated from a curved panel under the influence of a turbulent boundary layer is minimized in the presence of added point masses, which are varied during the optimization procedure. The total mass is also minimized and the Pareto front relating the trade-off between added mass and reduced noise is determined. Solver scaling information is provided that demonstrates the utility of the parallel processing approach.

Original languageEnglish (US)
JournalStructural and Multidisciplinary Optimization
DOIs
StateAccepted/In press - Jan 1 2019

Fingerprint

Parallel processing systems
Parallel Computing
Forcing
Acoustics
Optimization
Parallel Processing
Function evaluation
Processing
Added Mass
Turbulent Boundary Layer
Evolutionary algorithms
Pareto Front
Boundary layers
Evaluation Function
Acoustic waves
Demonstrate
Frequency Domain
Evolutionary Algorithms
Vibration
Trade-offs

All Science Journal Classification (ASJC) codes

  • Software
  • Control and Systems Engineering
  • Computer Science Applications
  • Computer Graphics and Computer-Aided Design
  • Control and Optimization

Cite this

@article{247bd30da327485f99c2b84f023ce18c,
title = "A parallel computing framework for performing structural-acoustic optimization with stochastic forcing",
abstract = "Structural-acoustic optimization procedures can be used to find the optimal design for reduced noise or vibration in many real-world scenarios. However, the time required to compute the structural-acoustic quantity of interest often limits the size of the model. Additionally, structural-acoustic optimization using state-of-the-art evolutionary algorithms may require tens of thousands of system solutions, which add to the limitations for large full-scale systems. To reduce the time required for each function evaluation, parallel processing techniques are used to solve the system in a highly scalable fashion. The approach reduces the analysis time by solving the system using a frequency-domain formulation and distributing solution frequencies amongst processors to solve in parallel. To demonstrate, the sound radiated from a curved panel under the influence of a turbulent boundary layer is minimized in the presence of added point masses, which are varied during the optimization procedure. The total mass is also minimized and the Pareto front relating the trade-off between added mass and reduced noise is determined. Solver scaling information is provided that demonstrates the utility of the parallel processing approach.",
author = "Shepherd, {Micah R.} and Campbell, {Robert Lee} and Hambric, {Stephen A.}",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/s00158-019-02389-2",
language = "English (US)",
journal = "Structural and Multidisciplinary Optimization",
issn = "1615-147X",
publisher = "Springer Verlag",

}

TY - JOUR

T1 - A parallel computing framework for performing structural-acoustic optimization with stochastic forcing

AU - Shepherd, Micah R.

AU - Campbell, Robert Lee

AU - Hambric, Stephen A.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Structural-acoustic optimization procedures can be used to find the optimal design for reduced noise or vibration in many real-world scenarios. However, the time required to compute the structural-acoustic quantity of interest often limits the size of the model. Additionally, structural-acoustic optimization using state-of-the-art evolutionary algorithms may require tens of thousands of system solutions, which add to the limitations for large full-scale systems. To reduce the time required for each function evaluation, parallel processing techniques are used to solve the system in a highly scalable fashion. The approach reduces the analysis time by solving the system using a frequency-domain formulation and distributing solution frequencies amongst processors to solve in parallel. To demonstrate, the sound radiated from a curved panel under the influence of a turbulent boundary layer is minimized in the presence of added point masses, which are varied during the optimization procedure. The total mass is also minimized and the Pareto front relating the trade-off between added mass and reduced noise is determined. Solver scaling information is provided that demonstrates the utility of the parallel processing approach.

AB - Structural-acoustic optimization procedures can be used to find the optimal design for reduced noise or vibration in many real-world scenarios. However, the time required to compute the structural-acoustic quantity of interest often limits the size of the model. Additionally, structural-acoustic optimization using state-of-the-art evolutionary algorithms may require tens of thousands of system solutions, which add to the limitations for large full-scale systems. To reduce the time required for each function evaluation, parallel processing techniques are used to solve the system in a highly scalable fashion. The approach reduces the analysis time by solving the system using a frequency-domain formulation and distributing solution frequencies amongst processors to solve in parallel. To demonstrate, the sound radiated from a curved panel under the influence of a turbulent boundary layer is minimized in the presence of added point masses, which are varied during the optimization procedure. The total mass is also minimized and the Pareto front relating the trade-off between added mass and reduced noise is determined. Solver scaling information is provided that demonstrates the utility of the parallel processing approach.

UR - http://www.scopus.com/inward/record.url?scp=85072126881&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85072126881&partnerID=8YFLogxK

U2 - 10.1007/s00158-019-02389-2

DO - 10.1007/s00158-019-02389-2

M3 - Article

AN - SCOPUS:85072126881

JO - Structural and Multidisciplinary Optimization

JF - Structural and Multidisciplinary Optimization

SN - 1615-147X

ER -