application of computational techniques during the development of packaging and medical devices
TRANSCRIPT
Application of Computational Techniques during the development of Packaging and Medical Devices (IAG 05)
ADarren Hodson PAR&D, AZC October 2005
Theme• Brief outline of the development
process.• Phases of the product cycle.• Key messages in today’s business environment.
• Applicable computational techniques and rationale for use.
• Modelling approaches and methods.
• Case study• Simulating and verifying material behaviour for medical
devices.• Multi Component systems.• Material investigation.
ADarren Hodson PAR&D, AZC October 2005
Development Process
Phases of the product development cycle
Concept Feasibility Design Verification Validation Launch
• Reduce product development cycle time
• Maintain or enhance quality standards
ADarren Hodson PAR&D, AZC October 2005
Development Process:- Key message
The Product from a Patient Perspective
To meet shorter timelines, we must:-• Show as early as possible that our product will
meet the desired drug performance parameters “Therapeutic effect”.
• Show that the patient can use or interact in the desired manner with the device or packaging.
ADarren Hodson PAR&D, AZC October 2005
Development Process:- Key message
Product development cycle
To deliver the product, we must:-• Show as early as possible that our product can be
manufactured in the required volumes at the desired levels of quality and cost.
• Show as early as possible that our product will meet the desired mechanical, environmental and user performance parameters.
• Commit to plant investment as early as possible but only when the design has been frozen.
ADarren Hodson PAR&D, AZC October 2005
Culture of Up-Front Simulation:-
• Who can perform the simulation.• Resistance in the Ranks.• Doing it right.
”Fereydoon Dadkhan, Delphi Delco Electronics Systems” Ansys solutions Volume 4 No. 1
Time
Res
ourc
e
Using CAE
Design>build>test>analyse
Design>build>test
ADarren Hodson PAR&D, AZC October 2005
Structural Mechanics
• Why Structural analysisFor Medical Devices and Packaging
methods can be used to :-
• Simulate highly non-linear systems.• Rapidly compare design solutions
and provide design guidance.• Verify the design principals of
complex components and inter component interactions
• Verify performance and design sensitivities.
• Verify performance variance in production.
Typical Metering Valve
ADarren Hodson PAR&D, AZC October 2005
Structural Mechanics
Sheldon Imaoka Ansys Inc
• Considerations• Simplicity with
increasing complexity.
• Validate each step of the analysis.
• Expected Output• Understanding of
interactions, (forces, deflection)
• Component Optimisation
ADarren Hodson PAR&D, AZC October 2005
Strategy for use of Advanced Techniques
• Tools best used as comparative tools when system are highly complex and time is limited however confidence in methods must be attained.
• Simulation is an attempt to idealise a very complex non-linear world. We can’t model the Universe.
• Assumptions and approximations have to be made.• Validation is just as important as the simulation itself.
ADarren Hodson PAR&D, AZC October 2005
Development Process
Phases of the product development cycle
Concept Feasibility Design Verification Validation Launch
Simplistic Calculations
Advanced Techniques.
Simplicity and Rapid Comparative Analysis
IncreasedComplexity,
Sensitivities andVariance
Validate methods
FinaliseVerification &Validation
Phases for Simulation:- The Device or packaging
Plant InvestmentDecisions
ADarren Hodson PAR&D, AZC October 2005
Case study 1:- Simulating Medical Devices
Development of a polymeric spring
Moment Vs Rotation (Target)
0
5
10
15
20
25
0 2 4 6 8 10 12 14
Rotation DegreesM
omen
t Nm
m
• Minimise stress/strain.
• Match suggested Torque profile.
• Optimise the design.
• Assess variance of performance and reliability of the design.
• Assess the sensitivities of the design’s performance.
• Compare theoretical results to experimental.
ADarren Hodson PAR&D, AZC October 2005
Case study 1:- Assumptions
• Linear isotropic materials• Are they?
• Application of boundary conditions, point loads, pressure distributions.
• Are they understood?
• How’s the component manufactured?• Variance
– Material Properties and processing– Dimensionality– Assembly conditions
ADarren Hodson PAR&D, AZC October 2005
Case study 1:- Materials
Comparison of Materials
0
50
100
150
200
250
300
350
400
450
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
StrainSt
ress
MPa Plastic 1
Plastic 2Steel
• Polymers are highly Non-linear and Anisotropic
• Process dependent properties• Maximum Strain level important,
hydrostatic stress?• Avoidance of creep
Stress Strain profile For Plastic 2
0
10
20
30
40
50
60
70
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
Strain MPa
Stre
ss
• Linear materials are easier to analyse !
Melas material simulation model used
ADarren Hodson PAR&D, AZC October 2005
Design Evolution:- “In Silico”
Comparison of Spring rates
0
20
40
60
80
100
120
0 2 4 6 8 10 12 14
Rotation Degrees
Mom
ent N
mm
Bspring v8 LinearBspring V8 nonlinNew spring v1New Spring 2Target rateNew2_nonlinearNew_3_nonlinNew_4 nonlinearnew 5 nonlinVersion 6
Initial design
Final design
• Rapid comparison of designs
• Rapid prototype and test
ADarren Hodson PAR&D, AZC October 2005
Design Assessment:- PDS variance analysis
• Gaussian distributions assumed.
0.051Mfact0.0081.3mmThickness0.0080 mmOffset 20.0080 mmOffset 10.0080.1mmAssembly
VarianceNominal
Variance of Stress Strain profile
0
10
20
30
40
50
60
70
80
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
Strain MPa
Stre
ss Nominal MaximumMinimum
ADarren Hodson PAR&D, AZC October 2005
Design Assessment:- PDS variance analysis
152SimulationsMelasMaterial Model
LHSMonte Carlo
-Material
3915Elements
1 Hour paRun Time
ADarren Hodson PAR&D, AZC October 2005
Design Assessment:- PDS variance analysis
Graph of variance of Torque Profile from Nominal
-30
-20
-10
0
10
20
30
40
50
0 1 2 3 4 5 6 7 8
Degrees Rotation
Varia
nce
%normalised minnormalised maximum
• Number of simulations?
• Accuracy of simulations?
• Acceptable performance?
• Controllable process?
• Assembly?
ADarren Hodson PAR&D, AZC October 2005
Analysis convergence:- Cost factors
12
34
Solution time
Number of elements
Number of nodes
0
20000
40000
60000
80000
100000
120000
140000
160000
180000
Solution number
Analysis Convergence
Solution timeNumber of elementsNumber of nodes
• Meshes that are too coarse may not yield sufficiently accurate results.
• Time vs Accuracy, Gain ?Accuracy
Cost,Time
0
10
20
30
40
50
60
70
80
90
1 2 3 4
Strain %Stress %Moment %Size %
ADarren Hodson PAR&D, AZC October 2005
Case study 1:- Experiment vs. FEA
• Understanding of Material ?• Experimental Technique ?• Component Interactions ?• FEA technique ?
Initial Experimental data
0
5
10
15
20
25
30
35
0 1 2 3 4 5 6
Rotation Degrees
Mom
ent N
mm Experimental Data
Experimental DataFEA Data
Hysteresis
Pre Load @ 3 Degrees
Loss of Pre load
ADarren Hodson PAR&D, AZC October 2005
Case study 1:- Creep adjusted FEA data
Creep adjusted FEA data+/- 19% variance
0
2
4
6
8
10
12
14
16
18
0 1 2 3 4 5 6
Rotation Degrees
Mom
ent N
mm
feaExperimental data after 168 hrsExperimental dta after 25 hrs
ADarren Hodson PAR&D, AZC October 2005
Case study 1:- Simulation of Primary Creep
)/()1(1
.432. TCCC
cr etC −+= σε
32.1
.CC
cr tC σε =
crε
Time10 hrs
Creep Curves
0
0.005
0.01
0.015
0.02
0.025
0 100 200 300 400 500 600 700 800 900 1000
time hours
cree
p st
rain
3mpa
6mpa
9mpa
12mpa
15mpa
18mpa
21mpa
24mpa
27mpa
30mpa
Primary Creep Eqn Time Hardening:-
04 =C
ADarren Hodson PAR&D, AZC October 2005
Case study 1:- Further Work
• Improve understanding of Hysteresis.
• Explore use of appropriate creep equations.
• Explore use of alternative material models. Viscoelastic !!
ADarren Hodson PAR&D, AZC October 2005
Case study 2:- Multi Component System’s
Idealised geometry:- Compound cylinders
• Devices consist of multi-component systems
Key Issues• Combination of materials,
non-linear? • Variance of
performance ?• Reliability of the
design?• Stress Relaxation and
Creep?
Loading+/- Tol
Loading
ADarren Hodson PAR&D, AZC October 2005
Case study 2:- Multi Component System’s
• “Little information is available for predicting the behaviour of multi-component systems.”
Further issues• Agreement on thermoplastic material
models?• Agreement of simulation methodology?
• Rapidly changing stress state ?
T.Hyde et al Journal of Strain Analysis, V 31 No. 6 Nov 1996.
Classical Case
ADarren Hodson PAR&D, AZC October 2005
Case study 2:- Multi Component System’s
• Lame’ equations can be used to describe static case for linear elastic materials
ADarren Hodson PAR&D, AZC October 2005
Case study 2:- PDS Variance analysis
ParametersNominal Tolerance STDEV
RD1 6.0 0.05 0.016667RD2 7.0 0.05 0.016667RD3 7.0 0.05 0.016667RD4 10 0.05 0.016667RD5 10.0 0.05 0.016667RD6 11.5 0.05 0.016667Internal pressure 10.0 1 0.333333
R
MethodRun Time per sim (Sec) 198Materials LinearNumber of simulations 250.0Monte Carlo LHSRsponse surface Simulations 10000.0
• Variance of performance?
• Location of maximum stress?
• Run time approx 10 Hours
ADarren Hodson PAR&D, AZC October 2005
Case study 2:- PDS Variance analysis
LHS Results Response surface fit• Variance of performance easily assessed.
• Peak stress evaluated.
• Number of core simulations is important.
ADarren Hodson PAR&D, AZC October 2005
Case study 2:- PDS Variance analysis
16
• Correlation between output parameters ?
• How does creep or relaxation influence this peak stress over time?
PDS simulation Results LHSNominal Tolerance STDEV
Hoop Stress location 1 0.382 39.423 13.141Hoop Stress location 2 -0.985 34.215 11.405Hoop Stress location 3 -0.705 2.923 0.974Hoop Stress location 4 -2.725 22.718 7.573Hoop Stress location 5 -1.110 1.754 0.585Hoop Stress location 6 -1.785 2.839 0.946
PDS simulation Results Response SurfaceNominal Tolerance STDEV
Hoop Stress location 1 0.344 42.825 14.275Hoop Stress location 2 -1.008 36.561 12.187Hoop Stress location 3 -0.742 29.981 9.994Hoop Stress location 4 -2.755 23.389 7.796Hoop Stress location 3 -1.113 1.850 0.617Hoop Stress location 4 -1.786 2.991 0.997
ADarren Hodson PAR&D, AZC October 2005
Case study 2:- Multi Component System’s
Static t=0
32.1
.CC
cr tC σε =
Primary Creep Eqn Time Hardening:-
• Applicability of Time hardening equation?
• Availability of Viscoelastic models and data?
• We need “push button tools” but appropriate science behind them
Deflection of R1 over time
ADarren Hodson PAR&D, AZC October 2005
Case study 2:- Multi Component System’s
• Stress state changes over time.
• How does this influence failure rate of system?
• Time hardening?
ADarren Hodson PAR&D, AZC October 2005
Case study 2:- Further Work
• Carry out PDS creep simulation
• Implement Viscoelastic models.
• Develop more robust Stress relaxation/creep simulation methods.
• Develop experimental verification test apparatus.
ADarren Hodson PAR&D, AZC October 2005
Summary of Presentation
• CAE strategy needs careful consideration to ensure effective use and maximum return.
• Experimental validation is just as important as the simulation itself.
• Tools best used as comparative ones when confidence has been gained.
• Developing high performance plastic components is not easy.
• Highly advanced simulations can be resource intensive.
Key Messages:-
ADarren Hodson PAR&D, AZC October 2005
AcknowledgementsAstraZeneca
G Dean, L Crocker and R Mera NPL
Bryan Deacon, Ticona UK Ltd.
Ansys, Sheldon Imaoka
ADarren Hodson PAR&D, AZC October 2005
References:-• “Design optimisation of an electro scalpel”,Ansys solutions V1 No. 2
1999• “Redesign of a medical stent “:- Ansys solutions V1 No. 2 1999• “Quality based design with probabilistic methods” Dr. Stefan Reh,
Dr. Paul Lethbridge, Dale Ostergaard, Ansys Solutions Volume 2 Number 2.Spring 2000
• “ The Ansys probablistic design system”, Dr. Stefan Reh, Volume 3 number 1
• “Easier ways to make a packet”,Professional Engineering 24 Feb 1999.
• “The macroscopic yield behaviour of polymers” Ram Raghava, Robert M Caddell, Gregory S. Y. Yeh. J Material Science 8 1973 225-232
• Observations on the creep of two material structures . T.Hyde et al Journal of Strain Analysis, V 31 No. 6 Nov 1996.
• Sheldon Imaoka, www.ansys.net/ansys