10 Common Mistakes in CFD Modeling
Author: Dr. Sharad N. Pachpute (PhD, IIT Delhi)
 There are many common mistakes in CFD Modeling even commercial software like ANSYS, Star CCM, Converge are used.
 Many commercial CFD ANSYS, Converge, etc. have limited CFD support or lack experienced engineers for industrial products. There may be limitations of CFD tools to produce correct results.
 They reduced the maintenance cost and reduced direct support by collaborating with elite channel partners that have less support in academia and productbased industries.
 The Code of CFD is not known like OpenFOAM for customization of problems
 You can refer to the following major common mistakes in CFD Modeling
Lack of Basic Subject knowledge or literature review
Basic subject knowledge is essential to understand assumptions considered in CFD Modeling. Try to learn physical meaning of each term in governing equations before CFD analysis:
 Governing equations in Fluid dynamics: sources of momentum acting on fluids
 Numerical models in CFD solver: Most of CFD simulations are carried our for classical problem of fluid flows. CFD models can not solve numerically at molecular levels.
 Understanding of Turbulent flow physics and its models: understand assumptions considered in modeling of turbulent flow
 Convective Heat and mass transfer:
 Basic of combustion : Assumptions for Turbulent chemistry interaction models
 Multiphase flow and its Modeling : There are several multiphase models in CFD solvers. Models predict the phenomenons using assumptions.
Redundant CFD Simulations
 Some simple problem can be solved analytically instead of time consuming CFD analysis
 CFD analysis helps to find more flow physics instead of average values calculated from analytical or experimental correlations
 Cost of CFD simulations is 2o times more than analytical solutions
 Avoid unnecessary simulations if the project or proposal is not well planned due to incomplete input data
Selection of incorrect CFD domains
Over simplified domain

 Many times hurry in projects, over simplified domain is selected.
 It is important to understand the flow physics are statistically symmetrical
 Ensure no gradient across the symmetry plane
 Ensure flow is periodic or not if periodic boundary is selected
 Make sure that the domain is not affected by location of inlet and outlet boundaries. This case may create unrealistic situation
Incorrect coordinate system

 Use Cartesian coordinate system for flow through slot or rectangular duct or complex geometry which are not modeled in a cylindrical coordinate
 Use the axis symmetrical case for flow through the pipes or coaxial pipe

CFD Domain in Cartesian coordinate system

CFD Domain in Cylindrical coordinate system
Issues in Mesh Quality

Ensure Mesh density in high gradient regions
 Fine mesh to resolve wall bounded shear layer such a way than y+ <1
 High mesh density in region whenever there is a sudden jump in domain or flow

Orthogonal Quality near the walls
 Try to maintain orthogonal quality near the wall
 The mesh less orthogonal quality can be simulated with lower relaxation factors
 Highly skewed mesh would results in incorrect gradients the walls. Which may lead to unbounded solutions or divergence

Sudden Jump in Mesh

Fine mesh at inlet and wall boundaries

Skewed Mesh Cells
 Highly skewed cells may results in incorrect gradient of fluxes near the walls
 It is important avoid skewed cells near the walls
Incorrect Boundary and Input data

Inlet Conditions:
 Ensure the velocity profile at the inlet is uniform or nonuniform.
 Use the fully developed velocity profiles if inlet is the part of long duct
 Use the mass flow rate if velocity profile is important and inlet section is away from the main CFD domain

Outlet Boundary
 Use a fixed pressure boundary (pressure outlet or outlet vent) if pressure field is known
 Use the outflow boundary if gradient at the outlet section is negligible

Example 1: Uniform Velocity at Inlet
Example 2: NonUniform Velocity at Inlet of duct
Incorrect CFD Models

Select the correct turbulence models
 The standard kε model cannot predict the near wall behavior in adverse pressure gradient. Hence, SST kω is recommended
 The standard kω model cannot predict the free expanding flow, The realizable kε model is suitable

Select the correct multiphase models
 The discrete phase model (DPM) can’t used for four way coupling if collision between particles is to be predicted
 The Eulerian Eulerian model doesn’t predict the sharp interface

Select the correct combustion models
 The turbulence –chemistry interaction should be selected based Damkohler number
 The finite rate chemistry model is suitable for slow chemistry combustion
Incorrect CFD Solver Setting

Forced/natural/mixed Convection
 Check the problem is of forced convection, natural or mixed convection. It is determined by Richardson number, Ri = Gr/Re^{2}
 Enable the gravity for the problems of natural or mixed convection. Set the reference density and temperature

Incompressible/Compressible Flow
 Pressure based solver for incompressible or weakly
 Density based solver for compressible flow and high velocity flow ( Mach number > 0.3)

Steady/Unsteady problem:
 Steady solver for statistically steady flow (example: RANS models) or some laminar flows
 Unsteady solver for URANS and LES

Pressure –velocity coupling
 SIMPLE for steady or unsteady solver
 SIMPLEC for unsteady turbulent flow
 PISO for unsteady problem and poor quality mesh
 Pseudo/coupled solver for transient solver if maximum courant number is know

Time steps and number of iterations:
 Select the correct time steps ( local CFL number < 1) For RANS
 Select the correct time steps ( CFL number < 0.2) For LES
 Select more iterations per time step initially. Later it can be reduced in order to get proper convergence of all the variables

Initialization of variables and patching
 Incorrect initialization can lead to unbounded solutions
 Standard or hybrid initialization can be used
 Patching of variable should be correct for multiphase or combustion problems
 Relaxation factors
Issues in Monitoring the Simulations
 Make ensure bounded solutions
 Constant and stable convergence
 Check the residuals of all the variables below the their limit (< 10^6)
 Monitor the key parameters which are not changing significantly and monotonous trends with number of iteration
 Check the net fluxes of mass, total heat transfer, radiation heat transfer, total sensible heat transfer. Makes sure the net fluxes are less than 0.1 % of source values
 Time step and number of iterations affect the convergence /residuals of variables
Incorrect Interpretation of CFD Results

Validation with lab based experimental data
 Postprocess the results in nondimensional form and compare with experimental data assuming all the flow parameters are with experimental conditions
 Avoid direct comparison of CFD result with instantaneous experimental data
 For RANS, the CFD variables are averaged quantities which can be compared with averaged quantities from experimental results
 Qualitative comparison of CFD results is essential for assessment of CFD Models

Expected Maximum deviation between experimental and CFD results for
 Laminar flow Problem : 02 %
 Low speed turbulent flow Problem : 05 %
 High speed turbulent flow Problem : 015 %
 For Complex problems in multiphase flow and reactive flow : 020%
 Example 1: CFD Modelling of a bubble column
 Example 2: Water flow over a triangular weir: From the following figures, it is noted that the variation between experimental and CFD results is less.

Comparison with Power plant experimental data
 Compare the CFD results with power plant data which is obtained over a long operation
 Understand the assumptions considered in the CFD modeling and note practical conditions in power plant. Note what major physics in the modelling are not neglected as objective of projects
 There is always uncertainty in the measured physical quantities. Hence, numerical and experimental can be matched well all time

When you do not have proper experimental data
 Using basic knowledge of fluid mechanical and heat transfer, we can compare with average velocity and surface heat transfer rate
 Using the available simple correlation we can estimate velocity , temperature and pressure drop across the from inlet to outlet
 Intuition of engineering can be applied to judge the predicted CFD results are correct or not
 Example 1: Cooling of electronics chip sets
 In this, experimental date over small chips is difficult to measure over the small ships
 However, based on the values of voltage and current, we generally give as a source iof heat generation to each chips
 Average velocity and temperature at inlet and outlet can be measured and compared with the CFD values
Lack of concrete Conclusion from CFD results
 Due to uncertainty in CFD results, sometimes it is difficult for CFD engineer to provide a conclusion with confidence

Reasons for uncertainty for CFD result
 Simplified geometry for complex problems because of limited resources
 Assumption in CFD Modelling : RANS modelling is used instead of LES to get results in less time
 Uncertainty in CFD Models for complex flows and experimental data
 Uncertainty in experimental data
 Lack of stability in convergence of residuals
 Lack of credibility in CFD results by considering practical situations
 Limited CFD input data from process industries
 Insufficient data for comparison: Inadequate measurement of flow parameters in process industries which results in Limited practical database.
 Differences in the interpretation of data by CFD engineers and practical persons or operators at plants
Don’t simply match your CFD results to satisfy your supervisor or manager without strong technical justifications
A basic understanding of physics (fluid flows and heat transfer) and numerical modeling is essential along with good skills in CFD Tools
CFD Training to Overcome Mistakes
 Training in Computational Fluid Dynamics (CFD) is crucial for overcoming common mistakes and improving the accuracy and efficiency of simulations.
 Here are some key areas to focus on during CFD training to avoid common pitfalls:
Understanding the Basics Subjects
 Governing Equations: Make sure you understand the NavierStokes equations and other relevant fluid dynamics equations.
 Mesh Generation: Learn how to create and refine a computational mesh that balances accuracy and computational cost.
 Boundary Conditions: Properly setting boundary conditions is crucial for realistic simulations.
Software Proficiency
 Learning the Tools: Gain proficiency in popular CFD software like ANSYS Fluent, OpenFOAM, or COMSOL Multiphysics.
 PreProcessing Skills: Develop skills in geometry preparation and mesh generation using tools like ANSYS Meshing or Gmsh.
 PostProcessing Skills: Learn to analyze and visualize results using software such as ParaView or Tecplot.
Advanced Topics in CFD Course
 Turbulence Modeling: Understand different turbulence models (RANS, LES, DNS) and their appropriate applications.
 Multiphase Flow: Learn how to handle simulations involving multiple phases (e.g., liquidgas interactions).
 Heat Transfer: Incorporate heat transfer mechanisms if they are relevant to your study.
 Combustion and Pollution Modeling
 Turbomachinery Modeling
Advanced CFD Course To fill the gaps
 The advanced CFD Course will overcome your mistakes by gaining proficiency in software and advanced topics in the above topics
 This course is as good as a Master’s degree if you finish all assignments with dedicated CFD support teams
 The affordable and valuable course at Rs 3800 for students and USD 80 for overseas students for seven months
 Trusted by hundreds of students from top universities and professionals
 Click for the detailed syllabus
Continuous Learning and Practice
 Workshops and Courses: Attend workshops, webinars, and courses to stay updated on the latest techniques and software updates.
 Community Engagement: Participate in CFD forums and user groups to share experiences and learn from others.
 HandsOn Projects: Apply your knowledge to realworld projects and case studies to gain practical experience.
Documentation and Reporting
 Detailed Documentation: Keep detailed records of your simulation setup, assumptions, and results for future reference and reproducibility.
 Reporting Results: Learn to effectively communicate your findings through reports and presentations.
Focusing on these areas during CFD training can significantly reduce common mistakes and lead to more reliable and accurate simulations.