Electronics Reliability Prediction Using the Product Bill of Materials Cheryl Tulkoff Jim Lance National Instruments
Outline Basic Definitions and Background Case Study Going Forward
Definitions Reliability Prediction Process used to estimate constant failure rate (λ) of useful product life
Definitions MTBF: Mean Time Between Failures Reliability of a component or assembly that can be repaired and put back in service MTBF = 1/λ where λ = failure rate, typically # of failing units per million hours
Common MTBF Misconceptions Minimum, guaranteed time between failures Correlation between service life & λ Can have a very reliable but short-lived device: missile Includes assembly and construction factors (quality)
Survival Based on the Exponential Failure Law Reliability is the probability of zero failures (survival). Probability Distributions (Exponential, Binomial, Normal, Weibull) The Exponential Distribution is fairly simple and can get you close with less parameters. R = exp (-T λ) = exp (-T / MTBF)
Example Calculated Survival
MTBF Calc Assumptions Perfect Design All stresses/use data known Failures are random Any part failure causes a system failure Parts models are up to date and accurate
Reliability Prediction: Industry Standards Mil Specs MIL-HDBK-217F Telcordia (Bellcore) SR-332 Prism (System Reliability Center) Mixed Others.
Some Software Providers / Options Relex Reliasoft Asent (Raytheon) RelCalc (T Cubed) Lambda Consultants (Ops A La Carte, DfR, others)
Why try to predict reliability at all? Compare to competitor s products Compare product design from one revision to the next Tool for design improvement Identify design weaknesses or gaps
Product Case Study Case Study Details Data Acquisition product in market for several years with design revisions Relex Software using 217Plus Model MTBF calc d with and without use data
Case Study: MTBF w/o Use Data Calculation Parameters Temp = 30C Temp Dormant = 23C Environment = GSI (Ground Stationary Indoors) Operation Profile = Industrial Duty Cycle = 100% Vibration Level = 0 Cycling Rate = 184 Max Lambda by Component Type Calculated Failure Rate = 3.46 MTBF = 33 years Probability of Survival 1 year = 97%
Case Study: MTBF with Use Data Calculation Parameters Temp = 30C Temp Dormant = 23C Environment = GSI (Ground Stationary Indoors) Operation Profile = Industrial Duty Cycle = 100% Vibration Level = 0 Cycling Rate = 184 Max Lambda by Component Type Calculated Failure Rate = 3.06 MTBF = 37.3 years Probability of Survival 1 year = 97.4%
Case Study: MTBF with Use Data & Duty Cycle Calculation Parameters Temp = 30C Temp Dormant = 23C Environment = GSI (Ground Stationary Indoors) Operation Profile = Industrial Duty Cycle = 100% Vibration Level = 0 Cycling Rate = 184 Max Lambda by Component Type Calculated Failure Rate = 0.77 MTBF = 148 years Probability of Survival 1 year = 99.3%
RMA Data Year 12 Month Base Returns % Survival 2004 2005 2006 2007 2008 1165 3157 3282 3052 3113 3 38 24 26 19 99.7% 98.8% 99.3% 99.0% 99.3% Overall Average Survival = 99.2% Calculated Survival = 99.3% Issues: Can not be certain of field environments. Not certain actual duty time per unit (Calculations 100% Duty) Out of 19 failures (2008) only 30% had component issues. Other types of failures include (DOA, Calibration, Unknown, etc). Component failures likely use driven (abnormal circuit conditions).
RMA Data Sampled Data from 2008 The ceramic cap was not among the larger calculated lambda components. The failure was among other parts that failed in the circuit most likely due to unusual spike in current during use. Actual Failures versus Calculated = Field Failures = Calculated Lambda None of the higher lambda components showed up in the data.
Recommendations It is difficult to represent field failures with calculated MTBF models. It is important for consumers to know how MTBFs were generated and what the limitations are for those calculations.
What next? Our customers expect us to provide MTBF values for our products. Continue to educate our customers and provide the most consistent numbers we can. Monitor RMA for biggest impact reliability issues from the field.
Closing Questions How well does the predicted number match actual product return rates from the field? Does the model predict which components will contribute the most to reliability issues in the field? In our experience, a resounding NO! to both questions So, is MTBF good for anything practical? References Reliability for the Technologies Second Edition, Leanard A. Doty, Industrial Press Inc., 1989