mesh dependency - monterrey institute
DESCRIPTION
Every time a CFD problem is solved, no influence fromthe mesh configuration is desired in the final results.During our study, different ways of constructing meshesfor a heat transfer of a car interior were analyzed in anattempt to find an optimal model that could allow anacceptable precision when comparing the results with theones obtained from experimentation. Having the intentionof understanding the influence of the mesh over theresults, all possible ways of CFD meshes construction wereanalyzed, as well as the variables that could be modified.TRANSCRIPT
MESH DEPENDENCY
Ivan Dario Arroyave Zuluaga Automotive Engineering Research Center
(CIMA) Tecnológico de Monterrey Campus Toluca
México
Opportunity Statement / Expected Outcome
Current State In a OEM, Cabin & Thermal models are meshed for analysis using tetrahedral mesh with generic requirements recommended by the Software Vendor. CFD engineers have encountered significant differences in the simulations when the grow rate and type of element is changed.
Desired State
Find the required mesh type and mesh requirements for each analysis creating Best practices.
Robust Design
• We are looking for a new general methodology to build grids for CFD problems; we would like to do our mesh design robust for different types of elements. Hence we selected robust design to carry out our analysis.
• This is the first approach to mesh dependency problems; we want to understand not only how to optimize the mesh construction, but also figure out if Robust Design is a good tool to do this optimization.
Background
Background
Background
Background
Opportunity Statement / Expected Outcome
Defroster Windshield Physical
test patterns
Defroster Side Windows
Physical test patterns
Expected Outcome (Specific physics test):
• Understand sensitivity of meshing characteristics of the determined factors in the event
of thermal analysis
Constrains:
• Software capability, computational cost
Opportunity Statement / Expected Outcome
Develop Concept: Current Best Practices
Develop Concept The following parameters were considered critical for mesh construction
using Hypermesh
4 1
1mm 0.6mm
A. Tetra Number of uniform layers
D. First layer thickness
B. Tetra Growth rate
1.4 1
F. BL Growth rate:
C. Number of BL layers
• YES • NO
1.6 1.1
6 0
Noise/Tetra to Polyhedral
A
B
C,D,F
Parameter Diagram
System
Fluid Mesh process
Control Factors:
Tetra Number of uniform layers
Tetra Growth rate
Number of BL layers
First layer thickness
BL Growth rate
Noise Factors:
(e.g.) Tetra to polyhedral
Noise Factor 1
Noise Factor 2
Noise Factor 3
Outputs:
fit against
physical test
results multiple
response
(R1,R2,R3)
Symptoms:
Meshing time
Solve time
Quality Mesh
Input:
2D Mesh, Boundary
condition
Control Factor Strategy
A - Tetra Number of uniform layers TNUL1 1 Layers
TNUL2 2 Layers
TNUL3 3 Layers
TNUL4 4 Layers
B -Tetra Growth rate TGR1 1.1 rate
TGR2 1.2 rate
TGR3 1.4 rate
TGR4 1.6 rate
C -Number of BL layers NBL1 0 Layers
NBL2 2 Layers
NBL3 4 Layers
NBL4 6 Layers
D– First layer thickness FLT1 0.6 mm
FLT2 0.8 mm
FLT3 0.9 mm
FLT4 1 mm
F - BL Growth rate BGR1 1. rate
BGR2 1.2 rate
BGR3 1.3 rate
BGR4 1.4 rate
• Tetra Number of uniform layers
levels comprehend benchmark
observed typical values
• Tetra Growth rate levels is
selected for its current best
practices and the lower limit that
allows software.
• Number of BL layers , First layer
thickness And BL Growth rate is
chosen to explore different ways to
achieve fill out space without
interference between one surface
and its opposite.
Noise Factor Strategy Full factorial, two cases • First case one factor:
– Tetra to Polyhedral (2 levels)
• Second case Three Factors – Noise factor 1 (2 levels) – Noise factor 2 (3 levels) – Noise factor 3 (3 levels)
Response Strategy Separated Analysis • R1. Numerical-fit physical results
– % Defroster area 25 minutes. – % Defroster area 35 minutes.
• R2. Qualitative-fit physical results • R3. Solving time
Due to the parameters and parameters levels, an L16 orthogonal array was
chose. The objective is to fit the response to experimental value, the
nominal is best formulation is selected.
Optimization Details response 1
Noise factor polyhedral
Response Plots for Means
Response Plots for S/N
Results for one Response
• Optimal levels and factors for Means
– Factor A Level 3
– Factor B Level 4
– Factor C Level 1
• Optimal level and factors for S/N
– Factor A level 2
– Factor B level 2
– Factor C level 1
Conclusions • Hypermesh is versatile enough to carry out an experimental mesh
dependency for CFD Thermal Analysis.
• This analysis should be carried out for each phenomena.
• Variables (e.g. iteration convergence) from the specific solver should
be considered as a response, in order to get general mesh
construction rules.
• This study is the first step to create rules for optimal mesh
generation process.
• Robust Design could be a useful tool to analysis Mesh Dependency
when the quality of mesh does not interfere with the convergence
speed.
Thank you