ºÝºÝߣ

ºÝºÝߣShare a Scribd company logo

Applied Reliability Symposium 2009 M Turner

?Download as PPT, PDF?
?
Mark Turner CRP
Mark Turner CRP

This is the PowerPoint slides from my presentation at the 2009 International Applied Reliability Symposium.

1 of 34
Downloaded 134 times
Improving Organization Competitiveness. How Quantitative Accelerated Life Testing Can Present a Value Proposition. Mark Turner, C.R.P. Flextronics International 5:22 PM
Introduction Reliability impacts the organization¡¯s bottom line. Accelerated life tests are common, but are they always successful? Reliability engineers have a responsibility to deliver accurate and meaningful data. Our results must deliver real value. - Achieving this helps our organizations to become more competitive, and can drive up margins. - This is vital in a challenging economic climate.
Offer a refined accelerated life test methodology that: Is technically sound. Delivers accurate, meaningful results. Significantly reduces reliability demonstration test overheads. Reduces product time to market. Drives up margins. Enables our organizations to become more competitive. Drives warranty costs down and profits up! Presentation aims $$$
Two distinct approaches: Accelerated Life Testing (ALT). Identifies a product¡¯s life characteristics Reliability Demonstration Testing (RDT). Demonstrates whether a minimum reliability target has been achieved. Assumes an exponential distribution. A Parametric Binomial Test approach is often adopted. How do we select between each approach Obtaining Reliability Metrics
RDT Methodology Runtime=[c+(0.667x#Failures)]xMTBF Weibull++ Design of Reliability Tests utility can also be used.
A product has a target MTBF of 250,000 hours. We need to conduct a reliability demonstration test to a confidence level of 70%. Runtime = [1.2+(0.667x0)]x250,000 = 300,000 hrs. Using 200 test units, test duration = 1,500 hrs. During the test a unit fails. Runtime=[1.2+(0.667x1)]x250,000 = 466,750 hrs. Revised test duration = 2,334 hrs. An RDT Example
Capacity. Lab space or walk-in chambers. Cost. Test Samples. Manpower. Data acquisition equipment. Power demands. But customers often demand a reliability test. Reliability has a direct impact on customer satisfaction, retention, and ultimately our long term success. Common RDT Problems
Custom designs. Modular designs. Platform and derivative approach. Product Design Strategies
Accelerated Life Tests Activation Energy = 0.7eV. Activation energy of each failure mechanism will differ! Not all components within a system will experience an equal failure rate acceleration for a given set of conditions. The system may be subject to a constant acceleration. The activation energy of each component failure mechanism may differ significantly.
An MTBF Estimate will help to identify the most unreliable components. FIT rates may not be totally accurate. But identifies the ¡°weak links¡± Provides an opportunity for reliability improvement. Indicates that reliability vs. stress may be non- linear. Failure Rates
Acceleration Modeling Elevated stress levels introduce life acceleration. The ALT aim: Experience failures at high stress conditions. Identify acceleration factor. Extrapolate life at high stress to expected life under end-usage conditions. Model accuracy Dependent on number of stress levels within the test. 2 levels = linear extrapolation. >2 levels can identify non-linear life-stress relationships.
Relies on having an understanding of the product design and the physics of failure. Is not based any life or activation energy assumptions! A better question to ask is: What are the most likely components or factors that could cause the product to fail? A Proposed ALT Methodology What activation energy should I use?
A Proposed QALT Methodology Provides flexibility in industrial applications. This methodology will be reviewed in detail, during which an example will be presented¡­ N N Y
Running Example (1) Power supply platform design. Temperature identified as main stressor. ALT aim: Establish the design¡¯s mean life and failure distribution. Design RDT for future derivative products. ALT limitations: Only two thermal chambers are available.
The QALT Methodology Flow (1) Identify limited life components. Estimate their expected life. Manufacturers data. Stress levels. Standards based reliability prediction. Thermal analysis. Computer simulation. N N Y Conduct Physics of Failure Review
The QALT Methodology Flow (2) Identify stressors, performance limits & design margin. Determine degradation potential ¨C avoid phase change! Increasing stress N Y N Determine Stress Limits Conduct Failure Analysis
Running Example (2) PoF Review: Several limited life components: electrolytic capacitors and some semiconductors. Preliminary thermal review. Conducted on 5 units. All component temperatures monitored. 110 100 90 80 70 60 50 40 0 10 20 30 Time (hours) Temperature (?C) Thermal Profile
Running Example (3) Failure points: 99 ¡ãC , 100 ¡ãC , 102 ¡ãC , 103 ¡ãC and 105 ¡ãC. Mean = 101.8¡ãC, Standard deviation = 2.39¡ãC. Power analysis. Nominal usage temperature (23¡ãC). Output power increased in 1W increments. Failure points: 50W, 51W, 52W, 52W and 53W. Mean = 51.6W, standard deviation = 1.14W. Test repeated at midpoint of 40¡ãC & 102¡ãC= 71¡ãC, and at 102¡ãC.
Running Example (4) The performance variation can then be plotted. Temperature (¡ãC) Power (W) Failure distribution Maximum stress limit. Remaining below this ensures that any product failures are the result of life acceleration and not due to excessively high stress levels. Destruct Region Design Margin Specification Limits
The QALT Methodology Flow (3) Multiple stressors may be used to accelerate a product¡¯s life. If equipment constraints exist, identify the most suitable stressors. Establish impact of each factor. Estimate acceleration factor using PoF. Avoid acceleration factors >10-15. In design environments Taguchi arrays are ideal. N N Y Conduct DOE
The QALT Methodology Flow (4) Rule of thumb: Component life halves for every 10-15 ¡ã C increase above 40 ¡ã C. N N Y During the DOE each limited life item should be monitored. What impact each stress has on component life. Determine exact relationship. Also consider indirect stress effects. Conduct DOE Define Operational Limits & estimate Acceleration Factor
Running Example (5) Design of Experiment. Additional stressors: input power cycling, input voltage variation and output power. 2 3 DOE, focusing on how these stressors affected product wear-out. How each limited life component performs during each run. Which operating conditions are significant. What interactions exist between stressors. Input voltage & output power selected from ALT. This is combined with temperature.
Running Example (6) ALT setup Chamber conditions inverse. Acceleration factor based on most critical component ¨C the weak link! Nominal output power, constant input voltage Increased output power, power cycled Increased output power, constant input voltage ¡­ #1 #2 #3 #4 #n Chamber 1A Chamber 1B High Stress 1 High Stress 2 Nominal output power, power cycled ¡­ #1 #2 #3 #4 #n ... #1 #2 #3 #4 #n Chamber 2A Chamber 2B ¡­ #1 #2 #3 #4 #n Temperature ( ¡ã C) 90 80 70 60 50 40 Acceleration Factor 0 5 10 15 20 25 30 35 1A 1B 2A 2B
The QALT Methodology Flow (5) It is reasonable to assume a Weibull shape parameter Beta of 1.0. Product population is assumed to reside in bathtub curve constant failure rate region. Calculate sample quantity using Parametric binomial approach. Divide time duration by acceleration factor estimate. N N Y First pass approximation that can be refined later. Define Sample Size & QALT duration
Sample size. Dependent on thermal chamber capacity (80 units). Sample allocation. Inverse ratio of acceleration factor. Chamber 1A : 15 units. Chamber 1B : 12 units. Chamber 2A : 23 units. Chamber 2B : 30 units. Reliability demonstration (70% confidence level): Chamber 1A : 332 hours. Chamber 1B : 252 hours. Chamber 2A : 503 hours. Chamber 2B : 663 hours. Running Example (7) The accelerated life test should be continued to validate the conclusions¡­ Parametric Binomial test duration without acceleration required a duration of 3,750 hours
Derived Product QALT Design Field data from the derived product yields a Weibull shape parameter beta of 1.3. Technology was unchanged, but some minor changes were made. Accelerated RDT is conducted on further derived products. Same thermal chambers and sample quantity. Acceleration factors will remain unchanged. Chamber 1A : 15 units, RDT duration 949 hours. Chamber 1B : 12 units, RDT duration 719 hours. Chamber 2A : 23 units, RDT duration 1439 hours. Chamber 2B : 30 units, RDT duration 1896 hours.
Derived Product Accelerated RDT Product derived from initial platform design. Life data analysis identified that acceleration factor for chamber 2B (constant input voltage, nominal output power) was 6.2, not 5.66. Derived products will be subjected to 70% confidence RDT, rather than repeat QALT. 250,000 hour MTBF demonstration would require 130 units for a duration of 16 weeks, if no failures occur. This can be reduced using the acceleration factor of 6.2, and running the RDT at 65 ¡ãC.
Reliability demonstration test options: Assume maximum RDT capacity is 500 units. Derived Product Accelerated RDT One product to market every 6 weeks. OR Five products to market every 4 weeks!
Can be used to reduce ALT duration. Higher stress levels increase acceleration factor! To extrapolate between stress levels and usage condition, a degradation mechanism has to exist. Step Stress Tests t 1 t 2 t 3 Stress 3 Stress 2 Stress 1 f 1 ,d 1 f 2 ,d 2 f 3 ,d 3 F 1-3 : Probability of failure at stress levels 1-3. d 1-3 : Probability of degradation at stress levels 1-3.
Analysis Models Constant & step stress tests: Cumulative damage model. Enables the generation of contour plots. Indicates likelihood of all failure modes being identical. Requires Alta and Weibull++. Note that this results in the Beta from each individual batch being reported in Weibull++, not the overall Beta. Cyclic stress tests: General Log-Linear model. Eta Beta
Summary Reliability demonstration tests. Can take a long time to conduct. Can be expensive due to sample demand. QALT is ideal for platform designs. Enables tests to be based on PoF, rather than activation energy assumption. Identified acceleration factor can be used in derived product reliability demonstrations. Reduces test duration and/or sample needs. Enables organizations to be more competitive.
Where to Get More Information M Turner, ¡°A practical application of quantitative accelerated life testing in power systems engineering,¡± Soon to be published in the IEEE Transactions on Reliability. Reliasoft RS 521 Quantitative Accelerated Life Testing.
Mark Turner Senior Staff Reliability Engineer, Flextronics Medical Systems. (469) 229-2530 [email_address] Creating Value That Increases Customer Competitiveness
Questions Thank you for your attention. Q & A?

More Related Content

Applied Reliability Symposium 2009 M Turner

  • 1. Improving Organization Competitiveness. How Quantitative Accelerated Life Testing Can Present a Value Proposition. Mark Turner, C.R.P. Flextronics International 5:22 PM
  • 2. Introduction Reliability impacts the organization¡¯s bottom line. Accelerated life tests are common, but are they always successful? Reliability engineers have a responsibility to deliver accurate and meaningful data. Our results must deliver real value. - Achieving this helps our organizations to become more competitive, and can drive up margins. - This is vital in a challenging economic climate.
  • 3. Offer a refined accelerated life test methodology that: Is technically sound. Delivers accurate, meaningful results. Significantly reduces reliability demonstration test overheads. Reduces product time to market. Drives up margins. Enables our organizations to become more competitive. Drives warranty costs down and profits up! Presentation aims $$$
  • 4. Two distinct approaches: Accelerated Life Testing (ALT). Identifies a product¡¯s life characteristics Reliability Demonstration Testing (RDT). Demonstrates whether a minimum reliability target has been achieved. Assumes an exponential distribution. A Parametric Binomial Test approach is often adopted. How do we select between each approach Obtaining Reliability Metrics
  • 5. RDT Methodology Runtime=[c+(0.667x#Failures)]xMTBF Weibull++ Design of Reliability Tests utility can also be used.
  • 6. A product has a target MTBF of 250,000 hours. We need to conduct a reliability demonstration test to a confidence level of 70%. Runtime = [1.2+(0.667x0)]x250,000 = 300,000 hrs. Using 200 test units, test duration = 1,500 hrs. During the test a unit fails. Runtime=[1.2+(0.667x1)]x250,000 = 466,750 hrs. Revised test duration = 2,334 hrs. An RDT Example
  • 7. Capacity. Lab space or walk-in chambers. Cost. Test Samples. Manpower. Data acquisition equipment. Power demands. But customers often demand a reliability test. Reliability has a direct impact on customer satisfaction, retention, and ultimately our long term success. Common RDT Problems
  • 8. Custom designs. Modular designs. Platform and derivative approach. Product Design Strategies
  • 9. Accelerated Life Tests Activation Energy = 0.7eV. Activation energy of each failure mechanism will differ! Not all components within a system will experience an equal failure rate acceleration for a given set of conditions. The system may be subject to a constant acceleration. The activation energy of each component failure mechanism may differ significantly.
  • 10. An MTBF Estimate will help to identify the most unreliable components. FIT rates may not be totally accurate. But identifies the ¡°weak links¡± Provides an opportunity for reliability improvement. Indicates that reliability vs. stress may be non- linear. Failure Rates
  • 11. Acceleration Modeling Elevated stress levels introduce life acceleration. The ALT aim: Experience failures at high stress conditions. Identify acceleration factor. Extrapolate life at high stress to expected life under end-usage conditions. Model accuracy Dependent on number of stress levels within the test. 2 levels = linear extrapolation. >2 levels can identify non-linear life-stress relationships.
  • 12. Relies on having an understanding of the product design and the physics of failure. Is not based any life or activation energy assumptions! A better question to ask is: What are the most likely components or factors that could cause the product to fail? A Proposed ALT Methodology What activation energy should I use?
  • 13. A Proposed QALT Methodology Provides flexibility in industrial applications. This methodology will be reviewed in detail, during which an example will be presented¡­ N N Y
  • 14. Running Example (1) Power supply platform design. Temperature identified as main stressor. ALT aim: Establish the design¡¯s mean life and failure distribution. Design RDT for future derivative products. ALT limitations: Only two thermal chambers are available.
  • 15. The QALT Methodology Flow (1) Identify limited life components. Estimate their expected life. Manufacturers data. Stress levels. Standards based reliability prediction. Thermal analysis. Computer simulation. N N Y Conduct Physics of Failure Review
  • 16. The QALT Methodology Flow (2) Identify stressors, performance limits & design margin. Determine degradation potential ¨C avoid phase change! Increasing stress N Y N Determine Stress Limits Conduct Failure Analysis
  • 17. Running Example (2) PoF Review: Several limited life components: electrolytic capacitors and some semiconductors. Preliminary thermal review. Conducted on 5 units. All component temperatures monitored. 110 100 90 80 70 60 50 40 0 10 20 30 Time (hours) Temperature (?C) Thermal Profile
  • 18. Running Example (3) Failure points: 99 ¡ãC , 100 ¡ãC , 102 ¡ãC , 103 ¡ãC and 105 ¡ãC. Mean = 101.8¡ãC, Standard deviation = 2.39¡ãC. Power analysis. Nominal usage temperature (23¡ãC). Output power increased in 1W increments. Failure points: 50W, 51W, 52W, 52W and 53W. Mean = 51.6W, standard deviation = 1.14W. Test repeated at midpoint of 40¡ãC & 102¡ãC= 71¡ãC, and at 102¡ãC.
  • 19. Running Example (4) The performance variation can then be plotted. Temperature (¡ãC) Power (W) Failure distribution Maximum stress limit. Remaining below this ensures that any product failures are the result of life acceleration and not due to excessively high stress levels. Destruct Region Design Margin Specification Limits
  • 20. The QALT Methodology Flow (3) Multiple stressors may be used to accelerate a product¡¯s life. If equipment constraints exist, identify the most suitable stressors. Establish impact of each factor. Estimate acceleration factor using PoF. Avoid acceleration factors >10-15. In design environments Taguchi arrays are ideal. N N Y Conduct DOE
  • 21. The QALT Methodology Flow (4) Rule of thumb: Component life halves for every 10-15 ¡ã C increase above 40 ¡ã C. N N Y During the DOE each limited life item should be monitored. What impact each stress has on component life. Determine exact relationship. Also consider indirect stress effects. Conduct DOE Define Operational Limits & estimate Acceleration Factor
  • 22. Running Example (5) Design of Experiment. Additional stressors: input power cycling, input voltage variation and output power. 2 3 DOE, focusing on how these stressors affected product wear-out. How each limited life component performs during each run. Which operating conditions are significant. What interactions exist between stressors. Input voltage & output power selected from ALT. This is combined with temperature.
  • 23. Running Example (6) ALT setup Chamber conditions inverse. Acceleration factor based on most critical component ¨C the weak link! Nominal output power, constant input voltage Increased output power, power cycled Increased output power, constant input voltage ¡­ #1 #2 #3 #4 #n Chamber 1A Chamber 1B High Stress 1 High Stress 2 Nominal output power, power cycled ¡­ #1 #2 #3 #4 #n ... #1 #2 #3 #4 #n Chamber 2A Chamber 2B ¡­ #1 #2 #3 #4 #n Temperature ( ¡ã C) 90 80 70 60 50 40 Acceleration Factor 0 5 10 15 20 25 30 35 1A 1B 2A 2B
  • 24. The QALT Methodology Flow (5) It is reasonable to assume a Weibull shape parameter Beta of 1.0. Product population is assumed to reside in bathtub curve constant failure rate region. Calculate sample quantity using Parametric binomial approach. Divide time duration by acceleration factor estimate. N N Y First pass approximation that can be refined later. Define Sample Size & QALT duration
  • 25. Sample size. Dependent on thermal chamber capacity (80 units). Sample allocation. Inverse ratio of acceleration factor. Chamber 1A : 15 units. Chamber 1B : 12 units. Chamber 2A : 23 units. Chamber 2B : 30 units. Reliability demonstration (70% confidence level): Chamber 1A : 332 hours. Chamber 1B : 252 hours. Chamber 2A : 503 hours. Chamber 2B : 663 hours. Running Example (7) The accelerated life test should be continued to validate the conclusions¡­ Parametric Binomial test duration without acceleration required a duration of 3,750 hours
  • 26. Derived Product QALT Design Field data from the derived product yields a Weibull shape parameter beta of 1.3. Technology was unchanged, but some minor changes were made. Accelerated RDT is conducted on further derived products. Same thermal chambers and sample quantity. Acceleration factors will remain unchanged. Chamber 1A : 15 units, RDT duration 949 hours. Chamber 1B : 12 units, RDT duration 719 hours. Chamber 2A : 23 units, RDT duration 1439 hours. Chamber 2B : 30 units, RDT duration 1896 hours.
  • 27. Derived Product Accelerated RDT Product derived from initial platform design. Life data analysis identified that acceleration factor for chamber 2B (constant input voltage, nominal output power) was 6.2, not 5.66. Derived products will be subjected to 70% confidence RDT, rather than repeat QALT. 250,000 hour MTBF demonstration would require 130 units for a duration of 16 weeks, if no failures occur. This can be reduced using the acceleration factor of 6.2, and running the RDT at 65 ¡ãC.
  • 28. Reliability demonstration test options: Assume maximum RDT capacity is 500 units. Derived Product Accelerated RDT One product to market every 6 weeks. OR Five products to market every 4 weeks!
  • 29. Can be used to reduce ALT duration. Higher stress levels increase acceleration factor! To extrapolate between stress levels and usage condition, a degradation mechanism has to exist. Step Stress Tests t 1 t 2 t 3 Stress 3 Stress 2 Stress 1 f 1 ,d 1 f 2 ,d 2 f 3 ,d 3 F 1-3 : Probability of failure at stress levels 1-3. d 1-3 : Probability of degradation at stress levels 1-3.
  • 30. Analysis Models Constant & step stress tests: Cumulative damage model. Enables the generation of contour plots. Indicates likelihood of all failure modes being identical. Requires Alta and Weibull++. Note that this results in the Beta from each individual batch being reported in Weibull++, not the overall Beta. Cyclic stress tests: General Log-Linear model. Eta Beta
  • 31. Summary Reliability demonstration tests. Can take a long time to conduct. Can be expensive due to sample demand. QALT is ideal for platform designs. Enables tests to be based on PoF, rather than activation energy assumption. Identified acceleration factor can be used in derived product reliability demonstrations. Reduces test duration and/or sample needs. Enables organizations to be more competitive.
  • 32. Where to Get More Information M Turner, ¡°A practical application of quantitative accelerated life testing in power systems engineering,¡± Soon to be published in the IEEE Transactions on Reliability. Reliasoft RS 521 Quantitative Accelerated Life Testing.
  • 33. Mark Turner Senior Staff Reliability Engineer, Flextronics Medical Systems. (469) 229-2530 [email_address] Creating Value That Increases Customer Competitiveness
  • 34. Questions Thank you for your attention. Q & A?

Editor's Notes

  • #2: TX-SX: Title of presentation Presenters Name ARS 2009, San Diego, CA, USA Page