LS-Dyna与MF耦合教程

MASERATI car body optimization
Courtesy of

Multi-physics Design Chain using modeFRONTIER
Car body and chassis design optimization purposes: ? optimization performed considering only physical parameters ? chassis components designed using only thickness value as parameter Design models: ? 3 parallel analysis have been made using ANSYS and LS-Dyna ? static and dynamic behavior of car body have been examinated
Courtesy of

Model presentation
The object of the analysis is a 3D car body of a MASERATI model. Our purpose is related to achieve a complete knowledge about the static and dynamic behavior of the car body in its working conditions. Accordingly, the design chain consists of 3 parallel analysis: ? torsional analysis using ANSYS ? modal analysis using ANSYS ? crash analysis using LS-Dyna In order to retrieve a huge number of designs, an IBM SP4 cluster is exploited, using the IBM LoadLeveler queue system to take the greatest possible advantage of the MPI capabilities of ANSYS and LS-Dyna. The entire design chain exporation is managed by the multi-objective design environment tool modeFRONTIER, which allows a fast search of the input parameters and the objective functions.
Courtesy of

Design models
Torsional Analysis
Modal Analysis
Courtesy of
Crash Analysis

Components to be optimized
There are 46 components of the chassis considered in the design chain analysis. The elements to be optimized can be classified as following: ? specific components defined as local reinforcements ? components defined as local rigid elements ? components requiring a greater thickness ? components of crosspiece beam The parameter chosen to build the analysis is the components thickness, as its influence on the car structural response is very consistent. Futhermore, the thickness parametrisation advantage is that the FEM model ( build in ANSYS ) is always the same for each design, so that the run time of a single design chain step is also reduced. Courtesy of

Components to be optimized – first set
Thickness range values:
LOWER BOUND: 0,5 mm UPPER BOUND: 6 mm
Courtesy of

Components to be optimized – second set
Thickness range values:
LOWER BOUND: 0,7 mm UPPER BOUND: 3 mm
Courtesy of

Torsional analysis
A FEM model is built in ANSYS for a static structural analysis. The load conditions are set up in order to examinate the torsional response of the car body in the normal working conditions. The goals to reach are respectevely: ? minimize the rotational displacements of the car body ? achieve a body stiffness of 150000 Kgm/rad ? minimize the equivalent rotation ? smooth the elastic torsional line
Courtesy of

Modal analysis
The modal analysis is performed considering the first 30 vibration proper modes of the structure. The FEM model is the same as that used in the torsional analysis. The purpose is about the maximization of the first natural frequency of the car body, in order to increase the safety conditions of the car with respect to the applied loads. The deformed shape of the car body is very similar to the torsional one.
Courtesy of

Crash analysis
The crash analysis is performed using LS-Dyna solver. The goals of the optimization are respectevely to minimize the displacement of the rigid body and maximize the internal energy adsorped in the frontal crash.
Courtesy of

Computation resources – IBM SP4 at CINECA
All the processors are POWER 4 - 1300 MHz. The configuration has 16 nodes with 32 processors each and with a memory of 64 GByte per node, except one with 128 GBytes. The physical nodes are partitioned in order to improve parallel performance: each partition is connected with 2 interfaces to a dual plane switch. Because of that the user have access to: ? 32 partitions with 8 processors and 16 GB of RAM each ? 14 partitions with 16 processors and 32 GB of RAM each ? 2 partitions with 16 processors and 64 GB of RAM each
Courtesy of

Computation resources – IBM LoadLeveler 3.1 at CINECA
? LoadLeveler is a batch job scheduling application and program product of IBM. ? Provides the facility for building, submitting and processing batch jobs within a network of machines ? LoadLeveler scheduling matches job requirements with the best available machine resources ? In a multi-user production environment, the use of LoadLeveler is intended to promote overall improved system performance, turnaround time and equitable resource distribution for all users ? LoadLeveler permits each machine in the pool to be configured differently. For example: a machine may be configured to be used for batch work only during offwork hours. ? Can schedule serial or parallel (MPI) jobs
Courtesy of

Design chain inside modeFRONTIER environment: the workflow.
Design parameters
Input variables Chassis components thickness Output variables Displacements, stresses, elastic torsional line, virbration proper modes, deformation energy
Design goals
Reduce chassis thickness in safety working conditons
Optimization’s set-up data
Sampling phase: Random DOE sequence – nr initial individuals: 128 Exploration phase: DOE sequence

Logic flow
Application node template
LoadLeveler command file
JOB SUBMISSION
Courtesy of

Logic flow – ANSYS run
ANSYS run script
ANSYS input file template
Courtesy of

Logic flow – LS-Dyna run
LS-Dyna input file template
LS-Dyna run script
Courtesy of

Post processing inside modeFRONTIER environment - OBJECTIVES
Multi-history chart of the objective functions Correlation chart of the objective functions
Frequency chart of the objective functions
Effective robustness visualization using modeFRONTIER Chart Desktop
Courtesy of
RESULTS

Post Processing inside modeFRONTIER environment – PARALLEL CHART
Parallel chart of the objective functions There are two trends of possible good solutions, due to the dynamical behaviour of che chassis in the frontal crash test. In fact, an high adsorbed energy produces great displacements of the car body, otherwise for a low value. In order to evict the problem, modeFRONTIER then creates a design ranking through MCDM theory, based on an utility function defined by user preferences about relationships between designs ad single objectives.
Courtesy of

Post Processing inside modeFRONTIER environment – CORRELATION CHART
Correlation chart between input variables and objectives
The correlation chart allows to visualize and calculate the inluence of each parameter Courtesy of on the evaluation of the objective functions.
After the evaluation of 128 designs, it comes up that one of the most important elements of the chassis is the car flank, because of the strategic position of this component in the car body, so that its thickness value influences all of the objective functions.

Post Processing inside modeFRONTIER environment – RUN TIME STATISTICS
The three analysis have been accomplished with different queues managed by Load Leveler. Therefore, the time lenght of the jobs depends on the system requirements set up in the Load Leveler command file. In fact, for the crash analysis, the use of a parallel queue of 16 processors nodes leads to a greater idle time than a serial analysis like the the torsional one. Load Leveler command file for the crash parallel analysis Run time statistics for MODAL analysis
Run time statistics for CRASH analysis
Run time statistics for TORSIONAL analysis
Courtesy of