CFD Mgrs 02 Proj Man 2

2.
1
CFD PROJECT PLANNING
2006 ANSYS, Inc. All rights reserved. ANSYS, Inc. Proprietary

CFD for Managers Fluent User Services Center
February 2008 www.fluentusers.com
CFD PROJECT PLANNING

Task definition
Resources
Timing and deliverables
Design review
Assumed part of bigger engineering project, but same

principles apply to stand-alone consultancy
2008 ANSYS, Inc. All rights reserved. 6-2 ANSYS, Inc. Proprietary
TASK DEFINITION
How can we break a CFD job into activities?
How can work be shared?
How do we measure progress?
What are the inputs and outputs for each task?
Where can bottlenecks occur?
Generic CFD project plan

CFD project activities are serial in nature
Not all activities are full-time: share time across projects
TASK RESOURCE TIME
Define objectives Consultative
Define model features / physics Modeler
Define geometry Engineer / consultative
Define inputs (prelim calcs) Engineer / modeler
Import / build geometry CAD op / modeler
Build mesh Modeler
Specify physical model Modeler
Specify boundary conditions Modeler
Solve equations Modeler, part-time
Post-process / export data Modeler / consultative
Review results Eng / mod / cons
Write report Modeler
Store / archive Modeler
Objectives - 1
Objective should never be to do a CFD model -
unless you just want some colour pictures!
Discuss the objectives with the modeler, explaining

the engineering context and the accuracy required
Get the modelers agreement to provide the specific

information that you need
Objectives - 2
Clear and precise quantitative questions:
What is the pressure change from A to B?
What are the heat transfer coefficients on surface C?
What are the forces acting on body D?
What is the air-fuel unmixedness at plane E?
Qualitative studies are OK, if you know what to do

with the results!
Define model features/physics

Many physical modelling options provided in Fluent
Selection by default adequate for many users
Adopt cook-book approach for similar jobs
Turbulence model, fluid properties should always be considered
Non-standard physical models require UDFs - add

greatly to project workload
Geometric information
Identify source of geometry to be modeled
Select an appropriate calculation domain
Omit irrelevant geometrical detail
Do not omit details which you know to have an

important effect in reality
Inputs and preliminary calcs

Good input data for CFD calculation are vital to success
Flow rates, velocity distributions, boundary pressures, temperatures
Usually necessary to make estimates or hand calculations for

some quantities
Turbulence data hardly ever known
Discuss source of input data with modeler
Geometry building
Maybe the biggest single task in a CFD job
CAD import can be effective, but clean-up can be time-

consuming
Ask the CFD modeller to explain his requirements to the CAD

operator before the CAD model is created
Mesh generation
Often the biggest single task, but increasingly automated
Plan the mesh - type, overall structure, size requirements
Decide how mesh is to be built up from geometry, and required

sequence of operations
Physics and boundary conditions

Menu selection and data input process
The most common source of errors in carrying out a job
Good preparation essential - prepare data and do preliminary

calculations before starting to input boundary conditions
Solution
Can be easy or time-consuming, but not normally a full-time job
Solver optimised for standard flow equations on good quality mesh
Large source terms in equations tend to make solution more difficult:

eg. buoyancy, chemical heat release, radiation
Resist temptation to tweak solver controls (under-relaxation) unless

necessary
Post-processing and data export

Post-processing necessary to
verify solution quality
understand qualitative aspects of flow
determine answers to objective questions
produce material for reports
Data can sometimes be exported directly for further analysis

eg. boundary conditions for stress analysis
Reporting
Conform to requirements of QA system
State objectives
Identify source of data and preliminary calculations
Identify code, version, modelling options selected and reasons
why
State important results
Provide details of archive/data storage
Storage/archive
Media incompatibility a major challenge
Magnetic media (tapes, disks, cartridges) have life limitations
CD or DVD-ROM recommended current option
Store one copy with project file
Resources
Human resources
CFD skills and experience
Management
Computers
Software
What makes a good CFD modeler?
Good fluids engineering background

recognises and understands fluid behaviour
relates to engineering objectives and context
able to read and interpret drawings/CAD!
Basic understanding of CFD principles
Able to think and work in 3D
Trained and experienced in use of software
Who can do CFD?

Many activities do not require specialist
Software improvements and on-line support are aimed at de-skilling CFD
Definition of problem and analysis of results requires fluid dynamics

know-how
Use of CAD operators and stress analysts is feasible, but close

supervision by fluid dynamicist is almost essential
Management of CFD
CFD is not an exact science:
Basic equations correct, but turbulence modelling approximate
Finite mesh size limits accuracy of solution
Experience and know-how needed to interpret solution and judge

quality
Important role for technical management
Make effective use of Fluent Support and other external resources
Computers
Affordable computer hardware now up to the job
Clusters, workstations, top end PCs (Win-32, Win-64, Linux)
1M+ computational cells the norm
Typical need is 1Gb RAM for 1M cells
BUT REMEMBER
Design the job to fit the computer!
Software
The choice is obvious . . .
Timing and Deliverables

Time estimation
Milestones
Managing the processes
The unforeseen
Job completion
Time estimation
Time-consuming activities are skills-based
plan according to the individual
Base estimates on experience of similar job
Uncertainties usually occur due to geometry creation and mesh

generation
Slippage most frequently results from rework due to inadequate,

incomplete or changing data
Milestones
Major milestones used to measure progress

Data and geometry complete
Mesh complete
Solution converged
Project complete
Final deliverables
Fluent data files
Answers to objective questions
Report
Managing the processes - 1

Attempt to define standard approach
QA procedure, best practice guidelines
Resist temptation for modeler to try the effects of different

options!
Data matching encourages this tendency
Grid refinement, validation and calibration should be carried

out separately, not as a normal part of an engineering project
Managing the process - 2

Get the modeler talking to the engineers
explanation uncovers errors
Manage data exchange between the modeler and other team

members
use engineering co-ordination memo or similar configuration control
system
Dont get blinded by science!
The unforeseen
Turbulence models are inexact
Fluid dynamics is non-linear

chaotic behaviour can result!
Even the best software can contain bugs

impossible to test all combinations of options
Critical changes can occur at a late stage
so allow reasonable contingency - it is often necessary to repeat

parts of work
Job completion
Reporting and archiving essential tasks
generally high turnover of CFD staff
difficult to pick up someone elses model without notes
Typically 20% of project remains after converged solution obtained

post-processing, report, review, presentation
DESIGN REVIEW
Purpose
Content
Reviewers
Purpose of Design Review

To verify that:
calculations were fit for purpose (ie. appropriate to original objectives)
model conforms to intent, and free from obvious errors
results have been interpreted correctly
To review and ratify engineering implications of results
Design review content

Presentation of model
Physical models, critical areas of mesh, boundary conditions
Presentation of results
Overall fluid behaviour, key objective results
Summary of engineering implications
Critical discussion
Reviewers
Engineering and CFD know-how required
hard in many organisations to provide both
key questions relate to quality, uncertainty, and best practice
engineering judgement has important role
ANSYS Fluent can provide resource to contribute to either side

of review process
Summary
Guidelines offered for project managers in handling CFD tasks
CFD is not fundamentally different from many other technically

challenging tasks
beware of the specialist!
ANSYS support and consultancy services are available to help

CFD Mgrs 02 Proj Man 2

Uploaded by

Copyright:

Available Formats

CFD Mgrs 02 Proj Man 2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CFD Mgrs 02 Proj Man 2

Uploaded by

Copyright:

Available Formats

2.

2006 ANSYS, Inc. All rights reserved. ANSYS, Inc. Proprietary

CFD PROJECT PLANNING

Assumed part of bigger engineering project, but same

Generic CFD project plan

Discuss the objectives with the modeler, explaining

Get the modelers agreement to provide the specific

Qualitative studies are OK, if you know what to do

Define model features/physics

Non-standard physical models require UDFs - add

Select an appropriate calculation domain

Omit irrelevant geometrical detail

Do not omit details which you know to have an

Inputs and preliminary calcs

Usually necessary to make estimates or hand calculations for

Discuss source of input data with modeler

CAD import can be effective, but clean-up can be time-

Ask the CFD modeller to explain his requirements to the CAD

Plan the mesh - type, overall structure, size requirements

Decide how mesh is to be built up from geometry, and required

Physics and boundary conditions

The most common source of errors in carrying out a job

Good preparation essential - prepare data and do preliminary

Can be easy or time-consuming, but not normally a full-time job

Solver optimised for standard flow equations on good quality mesh

Large source terms in equations tend to make solution more difficult:

Resist temptation to tweak solver controls (under-relaxation) unless

Post-processing and data export

Data can sometimes be exported directly for further analysis

Magnetic media (tapes, disks, cartridges) have life limitations

CD or DVD-ROM recommended current option

Store one copy with project file

What makes a good CFD modeler?

Good fluids engineering background

Basic understanding of CFD principles

Able to think and work in 3D

Trained and experienced in use of software

Who can do CFD?

Definition of problem and analysis of results requires fluid dynamics

Use of CAD operators and stress analysts is feasible, but close

Experience and know-how needed to interpret solution and judge

Important role for technical management

Make effective use of Fluent Support and other external resources

The choice is obvious . . .

Timing and Deliverables

Base estimates on experience of similar job

Uncertainties usually occur due to geometry creation and mesh

Slippage most frequently results from rework due to inadequate,

Major milestones used to measure progress

Managing the processes - 1

Resist temptation for modeler to try the effects of different

Grid refinement, validation and calibration should be carried

Managing the process - 2

Manage data exchange between the modeler and other team

Dont get blinded by science!

Fluid dynamics is non-linear

Even the best software can contain bugs

Critical changes can occur at a late stage

so allow reasonable contingency - it is often necessary to repeat

Typically 20% of project remains after converged solution obtained