Transcript
Accelerated Tes,ng for Design Improvement of Harmonic Surgical Blades ASTR Presenta,on September 2014 Presenter: Nabeel Jadeed September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
1
Nabeel Jadeed • B.S. Materials Science & Engineering – University of Cincinna,
• Currently a Quality Engineer (5yrs) at Ethicon in Cincinna,, OH (Johnson & Johnson) • 16yrs Experience in the medical device field – Wide range of product experience (staplers, trocars, clip appliers, energy-‐based devices) in manufacturing, quality, and R&D
• ASQ Cer,fied Quality Engineer, Reliability Engineer, and Six Sigma Black Belt September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
2
Background
• Harmonic Blades are used for spine soZ ,ssue access. • Instrument blade can become cracked due to inadvertent contact of ac,ve blade by the surgeon with metal objects in surgical field.
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
3
Harmonic Technology Basics
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
4
Harmonic Technology Basics • Energy transformed into mechanical mo,on, longitudinal vibra,on at ultrasonic frequencies • Mechanical mo,on cuts and coagulates simultaneously Blue Areas Denote Nodes
Antinodes
• Nodes:
• An,-‐nodes
– Zero movement – Maximum stress and strain September 5, 2014
– Maximum movement – Minimum stress and strain
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
5
Blade Cracking • Each Type of Blade Has a Sensi,ve Area Where Metal to Metal Contact Will Most Easily ini,ate a crack. • The Red Areas in FEA show the Most Sensi,ve Areas of Each Blade.
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
6
Problem • Tradi,onal life data analysis proves difficult to compare alternate designs – Time consuming due to long failure ,mes with some samples taking days of abuse without failing – Inconclusive due to high variability in life data Difficult for new product teams to make ,mely design improvement decisions September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
7
A Beber Way • Accelerated Life Tes,ng – Evaluate life using an accelera,ng stress factor to obtain failures more quickly without inducing unrealis,c failures modes1. – In the case of the Harmonic blade, we can use simulated metal contact at increasing levels of contact pressure to accelerate failures. September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
8
Test Procedure • Main stressor iden,fied as contact force of metal object at areas of high stress on the ,p of the blade. • Points of high localized stress are iden,fied with FEA • The points are validated by comparing product returns • A Carbide “hiber” is posi,oned at these points to simulate contact with metal while the blade is vibra,ng September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
9
Force – Deflec,on Curves
• Force-‐deflec,on curve are generated as input to the test fixture • Unique to different designs depending on overall length and loca,on of high stress area. 0.7
L / Blue y = 22.432x + 0.0501 R² = 0.9948
0.6
Force [lbf]
0.5
C / Red y = 21.618x + 0.0514 R² = 0.9944
0.4
L
0.3
C Linear (L)
0.2 0.1 0 0
0.005
0.01
0.015
0.02
0.025
Deflec,on [in] September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
10
Test Procedure • With main stressor iden,fied and force-‐deflec,on curves generated suitable stress points were needed. • These suitable stress points determined through experimenta,on: Carbide – Balance between instantaneous cracking and excessive ,me to failure (>300 sec)
probe
Test Specimen September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
11
Design Comparison • New Design – More cross-‐sec,onal area at high stress areas – Blade waveguide op,mized for less stress over control blade, i.e. more balanced – Minimize stress risers at high points of stress
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
12
Life Test Data Commercially available blade (control) benchmarked against new design Control New Design State End Time State End Time (sec) (F or S) (sec) Force (lbf) (F or S) 0.0769 F 66.7 -‐ -‐ -‐ -‐ -‐ -‐ 0.0769 F 68.7 F 59.5 0.0769 F 68.8 S 360 0.0769 F 160.4 S 360 0.151 F 37 F 34.2 0.151 F 48.6 F 35.2 0.151 F 66.5 F 39.9 0.3 F 10.4 F 20 0.3 F 27.1 F 21 0.3 F 40.4 F 28.6 0.4853 F 16.1 F 17.5 0.4853 F 23.5 F 29
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
13
Data Analysis – Model Assump,ons • ReliasoZ ALTA® used for ALT analysis • Determining best fit for life distribu,on: – Likelihood values compared for both control and new: New Design Weibull Beta = 1.513228 K = 0.204756 n = 1.596630 LK Value = -‐43.343574 Exponen-al K = 0.242417 n = 1.696578 LK Value = -‐44.309122 Lognormal Std = 0.721576 K = 0.192337 n = 1.349473 LK Value = -‐42.533399 September 5, 2014
Control
Lognormal best fit for new design
Weibull Beta = 2.943067 K = 0.096996 n = 0.904398 LK Value = -‐50.445923 Exponen-al K = 0.101116 n = 0.861578 LK Value = -‐57.582688 Lognormal Std = 0.394616 K = 0.109249 n = 0.864066 LK Value = -‐50.575704
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
Based on similar LK values, lognormal used for consistency
14
Data Analysis – Model Assump,ons • Lognormal used for life distribu,on • Inverse Power Law, common for single, non-‐ thermal stresses2, was used for Life-‐ Stress model ReliaSoft AL TA 7 - www.ReliaSoft.com
Use Level Probability Lognormal
– No evidence that shape parameters differ at each stress level based on Likelihood ra,on test September 5, 2014
U n re l i a b i l i ty
99.000
Use L ev el Control\Data 1 Inv erse Power L aw L ognormal .05 F=12 | S=0 Data Points Use L ev el L ine New Design\Data 1 Ey ring L ognormal .01 F=9 | S=2 Data Points Use L ev el L ine
50.000
10.000
5.000
1.000 10.000
100.000
1000.000
10000.000
100000.000
1000000.000
Nabeel Jadeed Ethicon Endo-Surgery 8/15/2014 4:36:08 PM 1.000E+7
Time Control\Data 1: Std=0.3946; K=0.1092; n=0.8641 New Design\Data 1: Std=0.6850; A=-1.7023; B=0.0805
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
15
Model Diagnos,cs • The standardized residuals should follow a normal distribu,on if the assumed life-‐stress model and distribu,on are a good fit3. ReliaSoft AL TA 7 - www.ReliaSoft.com
Standardized Residuals
99.000
Standard Residuals ARIES-2\Data 1 Inv erse Power L aw L ognormal Residual L ine 0.0769 F=1 | S=2 Residuals 0.151 F=3 | S=0 Residuals 0.3 F=3 | S=0 Residuals 0.4853 F=2 | S=0 Residuals
P ro b a b i l i ty
– No excessive devia,on of residuals from expected values
HK105-2\Data 1 Inv erse Power L aw L ognormal Residual L ine 0.0769 F=4 | S=0 Residuals 0.151 F=3 | S=0 Residuals 0.3 F=3 | S=0 Residuals 0.4853 F=2 | S=0 Residuals
50.000
10.000
5.000
1.000 -10.000
Nabeel Jadeed Ethicon Endo-Surgery 8/6/2014 9:45:12 AM -6.000
-2.000
2.000
6.000
10.000
Residual ARIES-2\Data 1: Std=0.7216; K=0.1923; n=1.3495 HK105-2\Data 1: Std=0.3946; K=0.1092; n=0.8641
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
16
Model Diagnos,cs • Cox-‐Snell residuals should follow an exponen,al distribu,on if the assumed life-‐stress model and distribu,on are a good fit3.
September 5, 2014
Cox-Snell Residua ls
99.000
Cox-Snell Residuals Control\Data 1 Inv erse Power L aw L ognormal Residual L ine 0.0769 F=4 | S=0 Residuals 0.151 F=3 | S=0 Residuals 0.3 F=3 | S=0 Residuals 0.4853 F=2 | S=0 Residuals
95.000
P ro b a b i l i ty
– Residuals are derived from the nega,ve natural log of es,mated reliability of each observa,ons, which should be uniform – Based on plot. no excessive devia,on of residuals from expected values
ReliaSoft AL TA 7 - www.ReliaSoft.com
New Design\Data 1 Inv erse Power L aw L ognormal Residual L ine 0.0769 F=1 | S=2 Residuals 0.151 F=3 | S=0 Residuals 0.3 F=3 | S=0 Residuals 0.4853 F=2 | S=0 Residuals
90.000
50.000
1.000 0.000
Nabeel Jadeed Ethicon Endo-Surgery 8/12/2014 8:08:36 AM 2.000
4.000
6.000
8.000
10.000
Residual Control\Data 1: Std=0.3946; K=0.1092; n=0.8641 New Design\Data 1: Std=0.7216; K=0.1923; n=1.3495
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
17
Model Diagnos,cs • The residuals vs the fibed values should appear random with minimal outliers. ReliaSoft AL TA 7 - www.ReliaSoft.com
New Design
ReliaSoft AL TA 7 - www.ReliaSoft.com
Standardized vs F itted Value
10.000
10.000
Standard - Fitted
6.000
Slight pabern
New Design\Data 1 Arrhenius Weibull 0.0769 F=1 | S=2 Residuals 0.151 F=3 | S=0 Residuals 0.3 F=3 | S=0 Residuals 0.4853 F=2 | S=0 Residuals
Standard - Fitted Control\Data 1 Inv erse Power L aw L ognormal 0.0769 F=4 | S=0 Residuals 0.151 F=3 | S=0 Residuals 0.3 F=3 | S=0 Residuals 0.4853 F=2 | S=0 Residuals
6.000
2.000
Re s id u a l
2.000
Re s id u a l
Control Standardized vs F itted Value
0.000
0.000
-2.000
-2.000
-6.000
-6.000
-10.000 10.000
100.000
Nabeel Jadeed Ethicon Endo-Surgery 8/12/2014 10:21:45 AM 1000.000
Fitted Value
Fitted Value New Design\Data 1: Beta=2.2378; B=0.2739; C=10.3747
Nabeel Jadeed Ethicon Endo-Surgery 8/16/2014 11:01:37 AM 100.000
-10.000 10.000
Control\Data 1: Std=0.3946; K=0.1092; n=0.8641
Abnormal behavior at low stress point: 2 out 3 blades suspended at 300 sec. ASTR 2014, Sep 10 -‐ 12, St. Paul, MN September 5, 2014
18
Life Comparison
Control vs. New Design ReliaSoft AL TA 7 - www.ReliaSoft.com
Lif e vs Stress
10000.000
L ife Control\Data 1 Inv erse Power L aw L ognormal .05 F=12 | S=0 Median L ine 0.0769 Stress L ev el Points Median Point Imposed Pdf 0.151 Stress L ev el Points Median Point Imposed Pdf 0.3 Stress L ev el Points Median Point Imposed Pdf 0.4853 Stress L ev el Points Median Point Imposed Pdf
New Design
L i fe
1000.000
New Design\Data 1 Inv erse Power L aw L ognormal .01 F=9 | S=2 Median L ine 0.0769 Stress L ev el Points Median Point Imposed Pdf 0.151 Stress L ev el Points Median Point Imposed Pdf 0.3 Stress L ev el Points Median Point Imposed Pdf 0.4853 Stress L ev el Points
100.000
Control 10.000 0.010
0.100
1.000
Stress Control\Data 1: Std=0.3946; K=0.1092; n=0.8641 New Design\Data 1: Std=0.7216; K=0.1923; n=1.3495
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
New design predicted to last longer with a probability of 97.88% 19
Benefits of this Method • Faster design itera,on -‐ allows for comparison in a maber of 1-‐2 days. – Previous abempts could take several days for 1 blade with inconclusive results
• Anecdotal informa,on from preclinical use in cadaver indicates improved robustness to cracking
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
20
Benefits of this Method • Test has demonstrated a degree of repeatability based on replicated experiments. – Repeated twice with new samples of the same design – Direc,onally the same conclusions (new design beber than old design)
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
21
Benefits of this Method • Subsequently used to test addi,onal design enhancements (surface treatment)
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
22
Poten,al Weaknesses • Assump,ons: – Constant contact pressure not realis,c, intermibent in actual prac,ce – Metal can contact any part of the blade, not just the high stress areas – Blade FEA model accuracy
• Is higher stress crea,ng unrealis,c failure modes? – Actual use stress difficult to determine September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
23
Ques,ons?
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
24
References 1. "Quan,ta,ve Accelerated Life Tes,ng Data Analysis." Accelerated Life Tes/ng and Data Analysis. ReliasoZ. Web. 6 Aug. 2014. . 2. “Inverse Power Law Rela,onship.” Accelerated Life Tes/ng Data Analysis Reference, ReliasoZ. Web. 6 Aug. 2014. . 3. "Residual Plots in Accelerated Life Tes,ng Data Analysis." Reliability Hotwire 105 (2009). ReliasoZ. Web. 6 Aug. 2014. .
September 5, 2014
ASTR 2014, Sep 10 -‐ 12, St. Paul, MN
25
