V&V of Computational Software to Predict the …vextec.com/wp-content/uploads/2016/10/VEXTEC_Frontiers...22 Virtual Twin® of a cardiac / defibrillator lead design –Estimate fatigue

1

V&V of Computational

Software to Predict the

Durability of Cardiac

Pacemaker Leads

May 19, 2015

Sanjeev Kulkarni, Animesh Dey, Robert

G. Tryon

© Copyright 2015 VEXTEC Corporation - All rights reserved

2

Talk Outline

1. Introduction

2. VLM Related to VVUQ

3. VEXTEC VLM Overview

4. Cardiac Leads Example

5. Nitinol Stent Example

6. MDDT Pilot Program


3

Who is VEXTEC?

Founded in 2000: Over $25 million from the

United States Department of Defense

Innovative Research programs for

Technology Development

Proprietary Software and Seven Patents:

Virtual Life Management® (VLM®)

generates VIRTUAL TWIN®

Customers: Federal Government and

Industries (Aerospace, Automotive,

Electronics, Energy, MEDICAL DEVICES)

Value Proposition: Help companies

improve products and reduce cost

• New products to market quickly

• Improve reliability of existing products

• Reduce physical and prototype testing

requirements

• Forecast product durability and

manage product life cycle risk

Business Model: Hybrid – Consulting

Services, Software Licensing and Training

VEXTEC accepted into FDA’s

Medical Device Development

Tool (MDDT) pilot Program


4

Uncertainty Management

Virtual Twin® representation of Uncertainty

Propagation across multiple levels of a system

• A probabilistic multi-disciplinary uncertainty

management analytical tool that links

computational models

• How will changing the input uncertainty of

the analysis impact the uncertainty in the

results?

• Can the uncertainty be updated based on

actual usage and observed damage state?

• What is the sensitivity to uncertainty?

Model 1

Model 2

Model 3


5


How general is this Uncertainty Management

Tool?

• General Tool –

– Virtual (In Silico) Clinical Trial

– Post Market Surveillance

– Therapy Effectiveness and Limits

Virtual

Patient

Device or

Therapy

Physician or

Process


6


Simulate the motion of a large number of

interacting spherical particles through tube

with a taper – Collaboration with CD-ADAPCO.

• Tapered Tube – Human Vasculature - Virtual

Patient

• Interacting Spherical Particles – Drug

Eluting Beads Delivered To Target Site –

Device / Therapy

• Random Motion of Spherical Particles –

Process

VLM framework captures the random

statistical nature of the problem to generate

the associated uncertainty scenarios.


7

COU: MP35N Cardiac Leads

• Simulation of two test conditions –

Displacements A” and B”

• Cycling between maximum displacement and

0 displacement

• Cycling to 1E10 cycles and Runout

(suspension) if no failure

• Measured difference in maximum Von-Mises

stress between Conditions 1 and 2

Condition 1: A” Von Mises Stress; Max Stress=XX ksi

Condition 2: B” Von Mises Stress; Max Stress=YY ksi

Cond1: A”+/-a”Cond2: B”+/-b”


8

Virtual Twin® of a cardiac / defibrillator lead design – Estimate fatigue life

• Sensitivity of uncertainty in input variables and

• Sensitivity of modeling approximations .

Flow of the design analysis:

• Manufacturing process model provides residual stress

• Structural analysis provides stresses model

• Microstructural material model predicts fatigue

Sources of uncertainty:

• Coil winding uncertainty

• Predicted residual stress uncertainty

• Structural geometry uncertainty

• Loads and boundary conditions uncertainty

• FEA mesh size uncertainty

• Material microstructure uncertainty

Cumulative Probability Distribution Function (CDF):

• Cycles to failure considering all uncertainties vs. actual test results

• Fatigue durability results are highly sensitive to uncertainty in residual

stress

• Simulations performed at different residual stress levels

• Good correlation for a calibrated value of residual stress



9

Yes

Revise

Appropriate

Model or Experiment

Reality of Interest

(Component, Subassembly, Assembly, or System)

Conceptual

Mode

Mathematical Model

Acceptable

Agreement?

Next Reality of Interest in the Hierarchy

Abstraction

Mathematical

Modeling

Physical Model

Computational Model Experiment Design

Simulation Results Experimental Data

Simulation Outcomes Experimental Outcomes

Physical

Modeling

Implementation

Calculation

Uncertainty Quantification

Implementation

Experimentation

Uncertainty Quantification

Preliminary

Calculations

Quantitative

Comparison

No

Validation

Code Verification

Calculation

Verification

ASME

V&V 10 -

2006 Fig 4

VEXTEC VLM VEXTEC VLM


10

ASME V&V 40 – Draft – Figure 2

VLM can be used to explore the region outside the Validation Domain

as well as at the edges (or even beyond) the CM&S assumptions


11

Execute pre-

defined

M&S and

V&V plan

YES

Is the CM&S

credible for

COU?

PurposeDefine

COUAssess

Model

Risk

Establish

Credibility

Threshold

Establish

Work

plan for

VV

Is the plan

Achievable ?

Document M&S

And VV plan

and findings

No

No

YES

PIRTExisting

VV DataM&S

Plan

ASME V&V 40 – Draft – Figure 3

VLM can play a Key Role in each step

of the Credibility Assessment

Strategy especially in the

Execution Step – where

Uncertainty Quantification is

important


12

ASME V&V 40 – Draft – Table1

Draft refers to “Uncertainty” 31 times and “Sensitivity” 12 times – both

implying increased level of credibility and VLM can address the

uncertainty management and design sensitivity considerations

Credibility Factors

Verification Validation

Applicability

Code Solution Model ComparatorOutput

Assessment

So

ftw

are

Qu

ali

ty A

ss

ura

nc

e

Nu

me

ric

al A

lgo

rith

m

Ve

rifi

cati

on

Dis

cre

tiza

tio

n E

rro

r

Us

e E

rro

r

Nu

me

ric

al S

olv

er

Err

or

Sys

tem

Co

nfi

gu

rati

on

Sys

tem

Pro

pe

rtie

s

Bo

un

da

ry C

on

dit

ion

s

Go

ve

rnin

g E

qu

ati

on

s

Sa

mp

le C

ha

rac

teri

za

tio

n

Co

ntr

ol O

ve

r Te

st

Co

nd

itio

ns

Me

as

ure

me

nt

Un

ce

rta

inty

Eq

uiv

ale

nc

y o

f in

pu

t a

nd

ou

tpu

t ty

pe

s

Rig

or

of

Ou

tpu

t C

om

pa

ris

on

Re

leva

nc

e o

f th

e

Qu

an

titi

es o

f In

tere

st

Ap

pli

ca

bil

ity t

o

the

Co

nte

xt

of

Us

e


13

How the VLM Process Compares with

Conventional Methods?

VLM approach

predicts failures

Fatigue strength is traditionally

determined by testing VEXTEC’s view of component,

grains & damage

Conventional view of component10

60,000 75,000 lbs. persquare inch (psi)

Load Appliedto Shaft

Load BearingCapacity of Shaft

Percent

of Shafts

Shaft Failures


14

VLM Predicts How, When, Where, and Why

Damage Occurs

Design & Stress Life & Where & Why

Standard Industry Analysis VLM Analysis


http://www.ansys.com/

http://www.ansys.com/

http://www.mscsoftware.com/

http://www.mscsoftware.com/

http://www.simulia.com/index.html

http://www.simulia.com/index.html

15

VLM: Grain – FEA – Component – Fleet

ComponentDesign

Configuration

Material

Configuration

VLM

Computational

Processing

Mapping the

Elements

Component

Simulation

Fleet

Simulation


16

VLM: Grain – FEA – Component – Fleet

Tooth Life: 15,932 cycles

Failure Cause: Defects

VLM Integration for

Entire Component

1st Virtual Twin

Gear Simulated

Component Life:

14,334 cycles

17,561 24,793

27,943

22,229

25,34218,961

22,113

Repeat Sequence

for Each Tooth

Integrate VLM

Results with FEA

Run 1,000 SimulationsVT1, VT2, VT3 … VT1,000


17

Grain Size

Particle Size

J. E. Schaffer: Masters Thesis, Purdue, 2007

Residual Stress

Cardiac Leads – Material Model


18

Near-initiation, cracked TiN particle

Chevron crack-initiating feature

J. E. Schaffer: Masters Thesis, Purdue, 2007

Nucleation

Crack front arrest at microstructural features.

Small Crack Growth

Striation spacing at crack front



19



20


• Simulation of two test conditions –

Displacements A” and B”

• Cycling between maximum displacement and

0 displacement

• Cycling to 1E10 cycles and Runout

(suspension) if no failure

• Measured difference in maximum Von-Mises

stress between Conditions 1 and 2

Condition 1: A” Von Mises Stress; Max Stress=XX ksi

Condition 2: B” Von Mises Stress; Max Stress=YY ksi

Cond1: A”+/-a”Cond2: B”+/-b”


21

Cardiac Leads – VLM Simulation


22

Virtual Twin® of a cardiac / defibrillator lead design – Estimate fatigue life

• Sensitivity of uncertainty in input variables and

• Sensitivity of modeling approximations .

Flow of the design analysis:

• Manufacturing process model provides residual stress

• Structural analysis provides stresses model

• Microstructural material model predicts fatigue

Sources of uncertainty:

• Coil winding uncertainty

• Predicted residual stress uncertainty

• Structural geometry uncertainty

• Loads and boundary conditions uncertainty

• FEA mesh size uncertainty

• Material microstructure uncertainty

Cumulative Probability Distribution Function (CDF):

• Cycles to failure considering all uncertainties vs. actual test results

• Fatigue durability results are highly sensitive to uncertainty in residual

stress

• Simulations performed at different residual stress levels

• Good correlation for a calibrated value of residual stress



23

• Define geometry and stress with FEA

• Define microstructure with metallography

• Run Simulation

• Calibrate to residual stress profile with experimental data



24

• Virtual Design of

Experiments

• Applied displacement

• Residual stress profile

• Inclusion size

distribution

• Top durability drivers • Residual Stress – 10%

improvement yields 60% improvement in mean total life

• Inclusion Size – 50% smaller inclusion size yields 25% improvement in mean total life

• Inclusion Density- 50% lower inclusion density; No significant impact

• Trade-off threshold exists between residual stress and inclusion size

…



25

Cardiac Leads: Summary and Outcomes

Summary

• Simulated fatigue buckling test under 2 load conditions

• Virtual DOE consisted of 9600 individual coil simulations

Outcomes / Next Steps

• Sensitivity study around particle size, density and residual stress

• Determined residual stress to be a calibrated value - new knowledge

• Developed Insights - Design alternatives, Material substitution, Vendor management

• Potential - Sensitivity analysis, Design trade studies, Supplier controls, Design optimization

• Add Realism – Coiling Simulation for Residual Stresses

VDOE Results for Residual Stress


26

• Evaluate Endoscopic Nitinol Stent to

calculate the effect of metallurgical

cleanliness on the fatigue life.

• Stent subject to fatigue cycles at a high

level of displacement to simulate

coughing..

• Two different materials (2 suppliers) with

different inclusion sizes and population

densities were evaluated using VLM.

COU: Endoscopic NiTi Airway Stent


27

• Check out versus

– Published data

– Life

– Mechanisms

– Initiation type and size

• Data available for 3 mean

strains and various

alternating strains

Lin, Z et. al.; JMEP ASM International; Nov 2010Specimens cut from stents

Fracture Surface- failure at inclusion

Airway Stent – Material Model


28

Airway Stent – Material Model


29

Simulated Mean Cycles to Stent Failure

Airway Stent – VLM Simulation


30

Simulated -3σ Cycles to Stent Failures and is an

indicator for early failure

Airway Stent – VLM Simulation


32

VEXTEC has been selected into the Medical Device Development

Tools (MDDT) Pilot Program

MDDT – Cardiac Leads

Chronology• Medical Device Development Tools (MDDT) –

Draft Guidance - FDA – November 2013

• MDDT Pilot Program – Announced August 2014 – Began Accepting Nominations September 2014

• VEXTEC – Submitted in November 2014 and FDA sent Acceptance in December 2014

Highlights• VEXTEC VLM at a mature stage of

development

• FDA recognizes that VLM meets a key public health need – Major Efficiencies to be gained in Device Development and Evaluation Time

• VEXTEC currently working with OEMs, Software Vendors and FDA staff

• Seeking Partners / Collaborators.


33

Summary / Takeaways

• Virtual Life Management and Virtual Twin represent a general

framework that incorporates systems realism including

uncertainty management and design sensitivity

• The framework supports key aspects of VVUQ standards –

ASME V&V 10 – 2006 and proposed ASME V&V 40

• A Cardiac Pacemaker Leads implementation was shown as an

example with key insights

– Ability to Virtually Simulate 10000 Bench Tests

– Study sensitivity of Microstructural Parameters and Residual

Stresses on Component Life

– Evaluate of Significant Outcomes – Design Alternatives,

Material Substitution and Vendor Management

– Demonstrate Potential for Major Gains in terms of time and

cost of Design Development and Evaluation

• Nitinol Airway Stent example compares results for two different

material suppliers

• VLM Methodology into FDA’s Medical Device Development Tool

(MDDT) Pilot program


34

VLM software is applicable throughout a product’s life cycle,

constantly growing in capability and value

Thank you


Documents

V&V of Computational Software to Predict the …vextec.com/wp-content/uploads/2016/10/VEXTEC_Frontiers...22 Virtual Twin® of a cardiac / defibrillator lead design –Estimate fatigue