This content is not included in
your SAE MOBILUS subscription, or you are not logged in.
A Neural Network Application to Shape Optimization
Annotation ability available
Sector:
Language:
English
Abstract
A neural network based design methodology is described which generates an approximate optimal design when given the desired overall size of the component. Once the methodology is developed for a particular class of components, near optimal designs can be obtained with minimal computational resources by practicing engineers who have little or no training in formal design optimization methods. The technique developed here uses computationally simple neural networks to simulate the function of a more complex shape optimization capability. These networks serve as approximate mapping functions for the optimal design space. Near optimal values of the design variables, performance measures and the objective function are obtained as outputs from the networks. The only data that must be supplied by the engineer are the values of a few key dimensions which describe the overall size of the part. The methodology is demonstrated for the design of generic suspension components in which the key dimensions are the overall length and width of the parts. Networks were developed using optimal design data obtained from a mathematical programming based shape optimization capability. These networks are capable of designing parts with accuracy suitable to that needed for early, conceptual design studies.
Recommended Content
Technical Paper | Neural Network Based Fast-Running Engine Models for Control-Oriented Applications |
Technical Paper | An Electronic Throttle Simulation Model with Automatic Parameter Tuning |
Technical Paper | Analytical Engine Design Methods: A Review |
Authors
Citation
Botkin, M. and Lust, R., "A Neural Network Application to Shape Optimization," SAE Technical Paper 951102, 1995, https://doi.org/10.4271/951102.Also In
References
- Botkin M. E. “Shape Optimization with Multiple Loading Conditions and Mesh Refinement,” AIAA Journal 28 5 1990 922 927
- Hajela P. Berke L. “Neurobiological Computational Models in Structural Analysis and Design,” Computers & Structures 41 4 1991 657 667
- Swift R. A. Batill S. M. “Application of Neural Networks to Preliminary Structural Design,” Proceedings of the AIAA/ASME/ASCE/AHS/ASC 32nd Structural, Structural Dynamics and Materials Conference Baltimore, MD, AIAA, Washington, DC April 1991 335 343
- Berke L. Hajela P. “Applications of Artificial Neural Nets in Structural Mechanics,” Structural Optimization 4 1992 90 98
- Swift R. A. Batill S. M. “Simulated Annealing Utilizing Neural Networks for Discrete Variable Optimization Problems in Structural Design,” Proceedings of the 33rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference Dallas, TX, Washington DC April 1992 2536 2544
- Hornik K. Stinchombe M. White H. “Multilayer Feedforward Networks Are Universal Approximators,” Neural Networks 2 1989 359 368
- White H. “Connectionist Nonparametric Regression: Multilayer Feedforward Networks Can Learn Arbitrary Mappings,” Neural Networks 3 1990 535 549
- Hornik K. Stinchombe M. White H. “Universal Approximation of an Unknown Mapping and Its Derivatives Using Multilayer Feedforward Networks,” Neural Networks 3 1990 551 560
- Carpenter William C. Barthelemy Jean-Francois M. “A Comparison of Polynomial Approximations and Artificial Neural Nets as Response Surfaces,” Proceedings of the 33rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference Dallas, TX, Washington DC April 1992 2536 2544
- Nilsson Nils J. Learning Machines Morgan Kaufmann San Mateo, CA 1990
- Pao Y.-H. Adaptive Pattern Recognition and Neural Networks Addison Wesley 1989
- Rumelhart David E. McClelland James L. Parallel Distributed Processing 1 MIT Press Cambrige, MA 1989
- Botkin M. E. “3-D Shape Optimization Using Fully-Automatic Mesh Generation,” AIAA Journal 30 7 July 1992 1932 1934
- Vanderplaats G. N. Numerical Optimization Techniques for Engineering Design: With Applications McGraw-Hill New York, NY 1984
- Vanderplaats G. N. ADS - A Fortran Program for Automated Design Synthesis Engineering Design Optimization, Inc. 1987