Upload
lamkhanh
View
224
Download
0
Embed Size (px)
Citation preview
© 2016 HBM
HBM Prenscia: Public
© 2016 HBM
Simulating an Imperfect World
Peter Lavelle, Application Engineer
© 2016 HBM
HBM Prenscia: Public
2
1. Introduction
2. Simulation and the Design Process
3. Simulation Accuracy & Precision
4. Simulation Fidelity
5. Validation & Verification
Agenda
www.altairhyperworks.com/industry/Automotive
© 2016 HBM
HBM Prenscia: Public
3
“simulation”Simulation is the imitation of the operation of a real-world process or system over
time. The act of simulating something first requires that a model be developed.This model represents the key characteristics or behaviours/functions
of the selected physical or abstract system or process.[https://en.wikipedia.org/wiki/Simulation]
Introduction
“imperfect”Damaged, containing problems, or not having something.
[http://dictionary.cambridge.org]
The Matrix, Warner Bros, 1999
© 2016 HBM
HBM Prenscia: Public
4
Why describe the world as “imperfect”? Imperfect compared to what…
Engineering is about Maths and Physics!
An Imperfect World
E = mc2
http://worldartsme.com/images/funny-einstein-clipart-1.jpg
Periodic
Table
Electricity
© 2016 HBM
HBM Prenscia: Public
5
Electricity
But do we really know everything…...or is engineering about “knowing enough”?
An Imperfect World
E = mc2
http://worldartsme.com/images/funny-einstein-clipart-1.jpg
Periodic
Table
Higgs-Boson?
Unsolvable?
Quantum?
Empirical!
Statistical!
© 2016 HBM
HBM Prenscia: Public
6
…and is our world really “perfect”?
An Imperfect World
www.mirror.co.uk/news/uk-news/uk-weather-watch-terrifying-moment-4950461
www.thisismoney.co.uk/money/cars/article-2192993/How-compensation-pothole-damage-car.html
www.skynet.be/diensten/reizen/dossier/853527/overladen/slideshow
© 2016 HBM
HBM Prenscia: Public
7
…and do we know how imperfect it actually is?
An Imperfect World
HBM nCode GlyphWorks, Worked Example
https://en.wikipedia.org/wiki/Electronic_color_code
https://en.wikipedia.org/wiki/Mars_Exploration_Rover
www.belvalves.com/engineering-excellence/design/fea/
www.slideshare.net/DemianBorba1/ux-super-powers-with-projectcomet
Difficult to
measure
New & Unknown
Environments
Noise & Interference
Tolerance (Ω) ± 5%Statistical
Variation
© 2016 HBM
HBM Prenscia: Public
Simulation and the Design Process
http://www.altairhyperworks.com/industry/Automotive
© 2016 HBM
HBM Prenscia: Public
9
Discrete Event Systems
• Monte Carlo, Timing level simulation for circuits…
Particles
• Snooker balls, Galaxies & Space Travel, Proteins…• 00
Agent-based / Behavioural
• Market modelling, Financial modelling, Human Interactions…
Differential Systems (single variable, ODEs)
• Circuit simulation, Structural mechanics, Chemical kinetics…
Differential Systems (multiple variable, PDEs)
• Heat, Elasticity, Electrostatics, Multi-Body Physics…
Computer Games, Virtual Reality
• Entertainment, Education, Training (eg. flight simulator)…
Types of Simulation
http://medicalxpress.com/news/2014-07-proteins-scientists-drug-discovery-tool.html
www.simscale.com/blog/2016/05/update-052016-conjugate-heat-transfer/
© 2016 HBM
HBM Prenscia: Public
10Simulation and the Design Process
Design Stage Simulation Tools
SpecificationsMarket modelling
Human interactions and behaviour (Agent-based)
Conceptual
Design
Computer Aided Design (CAD)
Topology Optimisation
Engineering
Design
Finite Element Analysis (FEA)
Computational Fluid Dynamics (CFD)
Fatigue and Durability
Vibration (NVH)
Thermo-Dynamics
Multi-Body Dynamics
Multi-Physics Simulations
ManufacturingProcesses : Stamping, Casting, Forging, Welding…
Production : Planning, Scheduling, Factory design & efficiency…
Products Virtual / Augmented Reality
Evaluation Reliability, Weibull, FMEA, Predictive Maintenance
© 2016 HBM
HBM Prenscia: Public
Simulation Accuracy & Precision
http://knowyourmeme.com/photos/918810-funny-error-messages
© 2016 HBM
HBM Prenscia: Public
12Accuracy & Precision
“accuracy”The degree to which the result of a measurement, calculation,or specification conforms to the correct value or a standard.
[https://en.oxforddictionaries.com/definition/accuracy]
“precision”Refinement in a measurement, calculation, or specification,
especially as represented by the number of digits given.[https://en.oxforddictionaries.com/definition/accuracy]
https://manoa.hawaii.edu/exploringourfluidearth/physical/world-ocean/map-distortion/practices-science-precision-vs-accuracy
© 2016 HBM
HBM Prenscia: Public
13Sources of Inaccuracy
Accuracy
Software
Bug
Geometry
& Meshing
Statistical
Variability
User
Error
Assumptions
& Boundary
Conditions
Computational
Precision
Measurement
Precision
Manufacturing
& Materials
© 2016 HBM
HBM Prenscia: Public
14
A software bug is an error, flaw, failure or fault in a computer program or system that causes it to produce an incorrect or unexpected result, or to behave in unintended ways.
[https://en.wikipedia.org/wiki/Software_bug]
Software Bug
Stay aware of announcements & communications from software companies
Update your software regularly
Contact Support and report bugs to the software companies
Validate simple calculations by hand to check you get the answers you expect
Try turning it off and on again…
https://www.linkedin.com/pulse/top-5-windows-errors-lisa-j-whelan
© 2016 HBM
HBM Prenscia: Public
15
Ways to avoid/reduce the likelihood of User Errors...
User Error
• Software training courses
• Worked Examples
• User manuals
• Online help & forums
• Demonstration videos
• Contact support hotlines
https://uk.pinterest.com/thatkelseygirl/call-center-humor/
© 2016 HBM
HBM Prenscia: Public
16
The “Dunning-Kruger effect” is a cognitive bias in which low-ability individuals suffer fromillusory superiority, mistakenly assessing their ability as much higher than it really is.
Their research also suggests corollaries: high-ability individuals may underestimate their relative competence and may erroneously assume that tasks which are easy for them are also easy for others.
User Error
[https://en.wikipedia.org/wiki/Dunning-Kruger_effect]
WISDOM(Knowledge + Experience)
High
Know-Nothing
CO
NF
IDE
NC
E
Guru
Low
Peak of “Mt. Stupid”
Slope of Enlightenment
Valley of Despair
Plateau of Sustainability
© 2016 HBM
HBM Prenscia: Public
17
By using the correct resources and attending training courses, the impact of the “Dunning-Kruger effect” can be significantly reduced and the risks associatedwith “Mt. Stupid” are mitigated, but not eliminated!
“Despair” will be reduced, the path to “Enlightenment” becomes both fasterand simpler, allowing the “Plateau” to be reached much sooner…
User Error
WISDOM(Knowledge + Experience)
High
Know-Nothing
CO
NF
IDE
NC
E
Guru
Low
“Mt. Stupid” reduced
& reached sooner
“Enlightenment” happens sooner
“Despair” reduced
& reached sooner
“Plateau” reached sooner
© 2016 HBM
HBM Prenscia: Public
18
When setting up a simulation, engineers must make assumptions and choose boundary conditions for the calculations (eg. RBE2 or RBE3?)
Engineers need to ensure the appropriate modelling techniques are used, and should therefore understand the physics and maths being used to drive the simulation.
Assumptions & Boundary Conditions
• Best-practice guides and theoretical training are crucial; available fromboth software vendors and independent providers (eg. NAFEMS)
• Advice can be found in industry standards and approved methods
http://www.quickmeme.com/meme/36jk4r
© 2016 HBM
HBM Prenscia: Public
19
A truly “accurate” geometry would require every individual atom to be defined and for the mesh to represent the crystal lattice!
Therefore, geometries are a compromise between different simulation methods, the limitations of mathematical analysis and final manufacturing techniques…
Geometry
Engineers Simulation Geometry
Design CAD “idealised”
Analyst FEA++ “solvable”
Manufacture Processes “feasible”
• Mathematical singularities need to be avoided by analysts. This can result ingeometry being altered purely for simulation purposes leaving a model that isno longer “accurate” when compared to the original design
• Simulation geometries are often simplified using things like shells(instead of solids) or symmetry to reduce calculations
• Manufacturing has it’s own limitations, tolerances and inaccuracies
http://machinedesign.com/fea-and-simulation/good-solid-modeling-bad-fea
© 2016 HBM
HBM Prenscia: Public
20
Manufacturing processes are not perfect, and each carries its own tolerances, geometric precision, and potential for inaccuracy when compared to simulation…
Manufacturing Tolerances
Machining tools and equipment have tolerances
Casting can leave flaws, porosity and inclusions
Welding is often considered an “art” performed by skilled workers
Residual stresses are left over from the manufacturing processes
Surface finishes can vary (eg. from worn out cutting tools or casting methods)
http://www.flexbar.com/shop/pc/SURFACE-ROUGHNESS-STANDARDS-SET-p4484.htm
© 2016 HBM
HBM Prenscia: Public
21
Often overlooked and expensive to obtain; good materials data is vitalfor performing an accurate simulation…
Materials Data
Data Book / Standards
Supplier (general)
Supplier (batch)
Material Testing (coupon)
Material Testing (component)
Online Did you read off a pixelated image?
When was the data generated?
How consistent is their product?
Includes manufacturing stresses…
Pay for peace of mind!
Do they have material test certification?
http://www.clipartkid.com/money-clip-art-BZpa5B-clipart/
© 2016 HBM
HBM Prenscia: Public
22
Materials themselves are inherently variable, and for most engineering purposes theyare modelled empirically to account for inconsistencies on an atomic level
Materials Data
Solids
Grain shapes and sizes, crystal defects and dislocations…
Liquids & Gasses
Brownian motion affecting enthalpy, entropy…
http://dtrinkle.matse.illinois.edu/MatSE584/metamod/preface.html
http://jimmyakin.com/2012/05/brownian-motion-explained.html
HBM nCode, Practical Fatigue Theory (Training)
© 2016 HBM
HBM Prenscia: Public
23
Assumptions and boundary conditions are often derived from measured data which itself is a source of inaccuracy and imprecision, carried through to simulation results…
Measurement Accuracy & Precision
When collecting data to be used as part of a simulation, it is important to choose the appropriate sensor type/sensitivity and acquisition hardware settings in order to understand and be aware of tolerances, limits and potential errors…
Sample
Points
Nyquist sampling rate
= 2 x Frequency
https://www.hbm.com
https://en.wikipedia.org/wiki/Electronic_color_code
© 2016 HBM
HBM Prenscia: Public
24
Even the best data acquisitions kits rely on being handled correctly…
Measurement Accuracy & Precision
Attachment (eg. strain gauge bonding)
Positioning (eg. hotspot)
Calibration (eg. thermal)
Interference & Noise (eg. electrical spikes)
Old / Broken sensors (eg. drift)
Raw data should be treated with caution, understood and ‘cleaned’ before use…
HBM nCode GlyphWorks
http://bradburybrothers.com/does-your-air-conditioner-have-faulty-wiring/
© 2016 HBM
HBM Prenscia: Public
25Significant Figures & Decimal Places
0.978V x 0.187A = 0.182886W ???
0.183W ???
0.18W ???
0.9775 ≤ 0.978 < 0.9785 0.1865 ≤ 0.187 < 0.1875
0.9775 x 0.1865 = 0.18230375 0.9785 x 0.1875 = 0.18346875
0.18230375 ≤ 0.978 x 0.187 < 0.18346875
© 2016 HBM
HBM Prenscia: Public
26
Question: How many significant figures should you use after multiplying “float x double” ???
Computational Accuracy
“float” 8 s.f.π = 3.14159265358979323846…
float (π) = 3.1415926XXXXXXXXXXX…
“double” 16 s.f.π = 3.14159265358979323846…
double (π) = 3.141592653589793XXXX…
IEEE 754 : Standard for Binary Floating-Point Arithmetic
© 2016 HBM
HBM Prenscia: Public
27
Many of the previous sources of inaccuracy are themselves subject tostatistical variability…
Statistical Variability
Source Statistical Variability
Assumptions &
Boundary Conditions
Physical differences in the user (eg. size & weight)
Different habits and usage profiles (eg. extreme users)
Environments will vary (eg. hot summer)
MeasurementsHardware tolerances (+/-) mean that measurements are never completely
exact and the ‘true’ value will always be obscured by background noise
ManufacturingProduction tolerances (+/-) mean that some components will be slightly
larger or smaller than the design, and could contain flaws
Materials Inherent statistical variation, therefore modelled empirically
ComputingDifferent computers will produce different ‘floating point digits’
therefore giving slightly different answers
http://sittingsafely.com/adjustment/
© 2016 HBM
HBM Prenscia: Public
28Sources of Inaccuracy
Accuracy
Software
Bug
Geometry
& Meshing
Statistical
Variability
User
Error
Assumptions
& Boundary
Conditions
Computational
Precision
Measurement
Precision
Manufacturing
& Materials
© 2016 HBM
HBM Prenscia: Public
29Usefulness -vs- Accuracy
Having explored the variety of inaccuracies that can affect a simulation, it is
important to consider how much time and effort (aka. money!) must be spent
to achieve 100% accuracy, and how useful this might actually be…
Ac
cu
rac
y
Time Accuracy
Us
efu
lne
ss
Harty, D. The Myth of Accuracy, The Journal of the Engineering Integrity Society, January 1999
Usefulness = Accuracy / [-ln(Accuracy - 1]
Accuracy = 1 - e(-duration)
© 2016 HBM
HBM Prenscia: Public
Simulation Fidelity
https://minecraft.net
© 2016 HBM
HBM Prenscia: Public
31
“fidelity”The degree of exactness with which something is copied or reproduced
[https://en.oxforddictionaries.com/definition/fidelity]
Simulation Fidelity
“real answer”
Mesh Density
Re
su
lt
• Compromises between fidelity of the model’s geometry/mesh and the available
computing power (time/cost) must be made for any simulation.
• It is impossible to get a ‘perfect’ answer because individual atoms would need to be
defined in the geometry and the mesh would have to be infinitely small!
• There is no point creating a fantastic mesh if others inputs are inaccurate!
www.comsol.fr/multiphysics/mesh-refinement
© 2016 HBM
HBM Prenscia: Public
32Simulation Fidelity
Simulation
FidelityVariables
Geometry
Measurement
Accuracy & Precision
Calculation Precision &
Correct Software Use
Statistical
Variability
Manufacturing
Accuracy & Precision
Physics
& Maths
Assumptions
“fidelity”The degree of exactness with which something is copied or reproduced
[https://en.oxforddictionaries.com/definition/fidelity]
© 2016 HBM
HBM Prenscia: Public
33Simulation Fidelity
“fidelity”The degree of exactness with which something is copied or reproduced
[https://en.oxforddictionaries.com/definition/fidelity]
Simulation
FidelityVariables
Geometry
Assumptions
UNDERSTAND
INFORMED
DECISIONS
Measurement
Accuracy & Precision
Calculation Precision &
Correct Software Use
Statistical
Variability
Manufacturing
Accuracy & Precision
Physics
& Maths
© 2016 HBM
HBM Prenscia: Public
Verification & Validation (V&V)
© 2016 HBM
HBM Prenscia: Public
35Verification & Validation (V&V)
“verification”The process of determining that a model implementation and its associated data accurately represent the developer’s conceptual description and specifications.
“validation”The process of determining that a model implementation and its associated data accurately represent the developer’s conceptual description and specifications.
DoD Modelling and Simulation Verification, Validation and Accreditation
(DoD instruction 5000.61; Dec 9, 2009)
http://www.dtic.mil/whs/directives/corres/pdf/500061p.pdf
How to Perform Credible Verification,Validation, and Accreditation for Modeling and Simulation, STC April 2005 (Dr. DA Cook and Dr. JM Skinner)
© 2016 HBM
HBM Prenscia: Public
36Validation & Verification (V&V)
SIMULATIONINPUTS OUTPUTS
PHYSICAL
TESTINGINPUTS OUTPUTS
© 2016 HBM
HBM Prenscia: Public
37Verification
"Engineers are not superhuman.
They make mistakes in their assumptions, in their calculations, in their conclusions.
That they make mistakes is forgivable; that they catch them is imperative.
Thus it is the essence of modern engineering not only to be able to check one's own
work but also to have one's work checked and to be able to check the work of others.”
Henry Petroski
© 2016 HBM
HBM Prenscia: Public
38Validation
There are a wide variety of different validation methods that usually involve physical testing in order to collect data and validate the results of a simulation…
Test tracks Wind tunnels Shaker tables Static loading Durability testing Crash tests Pressure tests Environmental tests
etc…
© 2016 HBM
HBM Prenscia: Public
39Validation
Does the FE-model accurately represent the real component ???
VIRTUAL Strain Gauge REAL Strain Gauge
Cross-Plot
Validation
http://www.caranddriver.com/bmw/2-series
Vehicle image for illustrative purposes only
© 2016 HBM
HBM Prenscia: Public
40
Use statistical methods such as “Design Of Experiments” and “Weibull” analysis to help with optimised testing schedules and for reliability modelling from physical tests…
Validation
© 2016 HBM
HBM Prenscia: Public
41Verification & Validation
How to Perform Credible Verification, Validation, and Accreditation for Modelling and SimulationDr. D A Cook & Dr. J M.Skinner; STC, 18 April 2005http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.640.7396&rep=rep1&type=pdf
© 2016 HBM
HBM Prenscia: Public
www.hbmprenscia.com