Large Scale, Multidisciplinary Optimization of a Wehicle System in a Scalable, High Performance Computing Environment
Srinivas Kodiyalam, R. J. Yang, Lei Gu, Cheng-Ho Tho
September 12, 2001
Abstract
The focus of this paper is on large-scale MDO of a vehicle system for Safety, NVH (noise, vibration and harshness) and Weight, in a scalable HPC environment. The computational complexity comes from addressing multiple safety modes including, frontal crash, offset crash, side impact and roof crush, in addition to the NVH discipline. A combination of high performance computing utilizing several hundred processors, Kriging metamodeling based design response approximation techniques, formal MDO strategies, and engineering judgement are effectively used to obtain superior design solutions with significantly reduced elapsed computing times. The reduction in large-scale MDO solution times through HPC is significant in that it now makes it possible for such technologies to impact the vehicle design cycle and improve the engineering productivity.
Introduction
Today more than ever before the manufacturing industry, in general, and automotive industry, in particular, is challenged with reducing the design cycle time while lowering costs and improving product quality and reliability. These constitute a set of complex as well as conflicting requirements that requires the use of formal and structured approaches to design, analysis and optimization. Simulation based design (SBD) involving collaboration and communication tools, formal multidisciplinary design optimization (MDO) procedures, and advanced virtual environments in conjunction with the well established CAD and CAE tools provide for such structured approaches to product and process design.
The increasing levels of high capability and cost effective HPC is contributing towards the widespread usage of high fidelity simulation models and tools as well as newer methods and technologies within the manufacturing industry. The computing speed required for large scale applications such as in solutions of CAE analysis (CFD, Explicit FEA) and optimization (MDO, Stochastic) problems are moving beyond MFLOPS (mega) into GFLOPS (giga = 109) and up. Typically, in such applications, there maybe millions of state variables or several hundreds of design variables and the engineering analysis must be repeated many times and rapidly to perform a systematic search through the design space in order to impact the product design cycle. Distributed, shared memory computing systems with several hundreds or thousands of processors having fast access to a large memory, are examples of scalable systems that enable solutions of such large scale applications.
The focus of this paper is on one such large-scale application, MDO of a vehicle system for safety and NVH (noise, vibration and harshness) and weight, that is performed on a scalable, HPC environment. The application of MDO to automotive vehicle design for safety and NVH has been of significant interest over the last several years [Yang et al. 1994, Chargin and Miura, 1999, Stander, 1999, Schramm, Schneider and Thomas, 1999, and Sobieski, Kodiyalam and Yang, 2000]. Sobieski, et. al., (2000) report a very significant reduction in elapsed computing time for such large-scale MDO problems – from 9 months to 1 day – through the efficient use of shared memory multiprocessor systems. The present work is an extension of work reported by Sobieski, et. al., (2000) and Yang, et. al. (2001) with a substantial increase in computational complexity of the MDO problem.
Download The Complete Paper
(A one-time registration will be requested)
