ASQ RRD Series: Analysis of Survival Data in Engineering, Business, and Medicine

Thu, Apr 8, 2021 12:00 PM – 1:00 PM EDT

Presenter: Dr. Wayne Nelson

This talk is an introduction to survival data analysis in engineering, business, and medicine. It presents basic concepts including the Weibull distribution, its age dependent failure rate, and simple probability plots. The talk shows applications to engineering, business, and medicine,, including

• The reliability of products designed and manufactured by engineers (e.g., toaster life).
• The distribution of time to a bank’s loss of bank accounts and TV service’s loss of subscribers.
• The life distribution of patients under treatment and life of medical devices (e.g., heart pacemakers).
The life distribution of patients and products, e.g., median life, % surviving 5-years or warranty.
• Whether a product failure rate increases or decreases as the population ages. This information is used to determine whether preventive replacement of old units in service reduces in-service failures.
• A prediction of the number of population failures in a coming month, quarter, or year.
During product development, a prediction of the improvement in product life that would result from eliminating one or more failure modes.
• Comparisons of 1) medical treatments, 2) business policies, and 3) product designs, vendors, materials, operating environments, manufacturing methods, etc.

SPEAKER. Dr. Wayne Nelson is an expert on reliability data analysis and accelerated testing. He worked at GE Research & Development for 24 years, and now consults privately. He is a fellow of the Amer. Society for Quality, the Institute of Electrical and Electronic Engineers, and the Amer. Statistical Assoc. ASQ awarded him the Shewhart and Shainin Medals and the Hahn Award for his lifetime achievements, and the Brumbaugh and Youden Prizes for articles on innovative methodology. He was the second person to receive the Lifetime Achievement Award of the 2,000-member IEEE Reliability Society for his innovative contributions to reliability methodology and reliability education.

ASQ RRD series webinar: Robustness Thinking in Design for Reliability – A Best Practice in Design for Reliability

Thu, Mar 11, 2021 12:00 PM – 1:00 PM EST

Presenter: Matthew Hu

Reliability is one of the most important characteristics of an engineering system. Reliability can be measured as robustness over time as a leading key performance indicator (KPI). Robustness thinking is essential to improve quality and reliability proactively by factoring the activities of design for reliability. Nothing can be substituted for thinking. Early robustness development in manufacturing can reduce the variability of those processes with valuable benefits to manufacturing yields, cycle time and costs. Product Development has a huge impact on revenue stream, reliability. It is most cost-effective and less time-consuming to make design insensitive to uncontrollable user environments in upfront design phase. Robustness development in Design for Reliability (DFR) process provides benefits in reduction of early-on physical testing and traditional test-fix-test cycles. Robustness achieved early in development enables shorter cycle times in the later design phases.

Objectives of the presentation
• Define robustness
• Explain product development using Robust Engineering versus traditional product development
• Explain Robust Design for Reliability
• Define Objective Function, Basic Function, and Ideal Function
• Explain how Ideal Function and Two-step Optimization lead to robust technology development and achieve “Better, Cheaper, Faster” product development
• Explain how to conduct a preliminary robustness assessment
• Explain the value of robustness assessment
• LiDar case study in robust autonomous driving technology development

Important Takeaway
• Make design insensitive to uncontrollable user environment (Noise)
• Early development of robustness is key to proactive quality and reliability Improvement
– Capture, front load noise and manage noise
– Gain control of your product performance
– Optimize robustness – avoid all failure modes
• Apply Robust design principles at early stages of product design to “forecast” problems and take preventive action.

1.  A stand‐by redundant system uses two identical units. The failure rate of each unit is 0.0007 failures per hour. What is the system  reliability for 200 hours (Assume the sensing and switching reliability is 0.9).

A.0.991      B. 0.983      C. 0.979         D,  0.965

2. Which of the following is true if all the subsystems in a series system have a constant failure rate?

A. The failure rate of the system is constant

B. The failure rate of the system will increase as more subsystems are added

C. The failure rate of the system is the sum of the subsystem failure rates

D.  All of the above

3. What is the reliability of this system?

A. 0.9191

B. 0.9244

C. 0.9297   

D. 0.9856

4. A parallel system has three subsystems each with a reliability of R.

The system reliability can be calculated as

A. 3R

B. R3

C.  1 ‐ (1 ‐ R)3  

D.  1 ‐ (1 ‐ R3)

5. To place confidence limits on a prediction which of the following is true?

A. The Chi Square distribution is used

B. The  F distribution is used

C. The t distribution is used

D.  A prediction is probabilistic, therefore confidence does not apply

6. 50 electronic devices have been tested for 3,000 hours without failures.

What is the approximate MTBF of this   device at 90% lower confidence ?

A. 65150 hours  

B. 25500 hours 

C.  6500 hours 

D. 6500 hours

7. Which one is not the reliability prediction technique?

A. Weibull plot

B. Duane plot  

C.  Uniform Precision Design   

D. Fix effectiveness Model?

8. In success testing, how many samples need to operate for one lifetime without failure to demonstrate 95% confidence with  99% reliability?

A. 298 samples   

B.  90 samples   

C.   59 samples   

D.   458 samples

9. The following data is used for thermal stress evaluation of ICs using Arrhenius Equation.

What is the acceleration factor ?

Wearout Activation Energy is  in eV

  • Ea = 0.5 eV k is Boltzmann’s Constant,
  • 8.617 x 10-5 eV / K
  • T1 is Temperature in degrees C = 70 deg C (343⁰K)
  • T2 is Junction temperature during test in degrees C= 125 deg C  (398⁰K)

A. 10.4

B.   5.2     

C.  9.5   

D.   10.0

10. The reliability of a system consisting of two units in parallel is 0.96.

If the reliability of each component is increased by 10%, what is the percentage increase in the reliability of the system?

A. 10%

B.  5%  

C.  3.33% 

D.  2.66%

ASQ RRD Series: Weibull – Special Topics in Weibull Analysis

Feb 11, 2021 12:00 PM Eastern Time (US and Canada)

Presentor: Jim Brenneman

Special Topics in Weibull Analysis: Continuation of Using Weibull Analysis to Solve REAL Engineering Problems
1. What Happens if a Weibull distribution doesn’t fit the data?   (Comparing the Weibull to other possible distributions). With some reminders/surprises.
2. Weibull Confidence Bounds and their use in Reliability. The only confidence bounds  application I have found extremely useful.
3. Regression with Life data  (Modeling S/N data, …).. Not your usual Regression analysis. You can use censored (runout) data and get results that are accurate (based on the data).
4. Sudden Death Testing. Sounds ominous, but can save you big $$ in many reliability tests.

Title: Prognostics and Health Management; Fundamentals, Elements, RUL Determination Techniques

Dec 10, 2020 12:00 PM Eastern Time (US and Canada)

Presenter: Dr. Mohammad Pourgol-Mohammad

The core of U.S. economy, security and quality of life depends on complex engineering systems that range from power plants, energy systems, and pipelines to aircraft, defense, and transportation systems. These complex engineering systems consistent of interconnected and diverse hardware, software, and human elements in dynamic conditions, physical processes, and
environments. Over the past decades, significant advances in sensing and computing have led to an explosion of new data and PHM algorithms designed to monitor component reliability. This webinar will discuss the advancements of PHM techniques and the simulation tools available to estimate the fatigue, corrosion, wear and creep life (RUL) of structures. Several Case studies will be presented.

Dr. Mohammad Pourgol-Mohammad is a safety/reliability analyst in multidisciplinary systems analysis with Keurig Green Mountain and Associate Professor (adj) of Mechanical engineering at University of Maryland and was an Associate Professor of Reliability Engineering, with Sahand University of Technology (SUT). He received his Ph.D in Reliability Engineering from University of Maryland (UMD), and holds one M.Sc degree in Nuclear Engineering and another in Reliability Engineering from UMD. His undergraduate degree was in Electrical
Engineering. Dr Pourgol-Mohammad has more than 18 year of work experience including research and teaching in safety applications and reliability engineering at various institutions including Johnson Controls, Sahand University of Technology, FM Global, Goodman Manufacturing, UMD, Massachusetts Institute of Technology (MIT), University of Zagreb-Croatia. He is a senior member of ASQ, ASME (currently ASME Safety Engineering and Risk/Reliability Analysis Division (SER2D) Chair), ANS and member of several technical committees and a registered Professional Engineer (PE) in Nuclear Engineering in States of Massachusetts. He is a certified reliability engineer (ASQ CRE), certified six sigma Black Belt (CSSBB) and Manager of Quality

Picture © B. Poncelet

Picture © B. Poncelet

1. The  lifetime  of  a  mechanical  lifter  is  normally  distributed  with  a  mean  of  100  hours  and  a  standard deviation of 3 hours. What is the reliability of the lifter at 106 hours  ? 

A. 0.0228                       B. 0.0570                       C. 0.9430                    D,  0.9772

2. A full factorial design of experiments has four factors. The first factor has two levels, the second factor has three levels, the third factor has two levels, and the final factor has four levels. How many runs are required for this analysis ?

A. 16     B. 48     C. 192     D. 256

3. On the basis of the fault tree, what is the likelihood of the top event occurring?

A. 0.0050            B. 0.2600                C. 0.3000              D. 0.5200

 4. Assuming  perfect switching  and  perfect  starting,  which  of  the  following  systems  has  the  longest  mean life if each system consists of n units with identical reliability ?

A. A series system.      B. A parallel system. 

C.  A k out of n system.    D. A cold standby system.

5. Which of the following is an appropriate use for experimental design ?

A. Establishing product requirements.             

B. Developing a fault-tree analysis.

C. Ensuring the robust design of a product.     

D. Analyzing customer complaint reports.

 6. The primary aim of sequential-life testing is to determine:

A. The probability density function of failures.

B. The mean time between failures (MTBF).

C. Whether a lot meets the reliability goal.

D. Whether the stress-level variation is significant

7. A small sample from a product population is subjected to multiple levels of elevated stress. Which of the following could be used to model the life of the product ?

A. Poisson process.   B. Pascal expansion.

C. Pareto rule.           D. Inverse power law.

8. A component fails on the average of once every 4 years with 75% of the failures observed to occur during stormy weather. If there are 12 hours of stormy weather to every 240 hours of good weather, what are the failure rates for stormy and good weather, respectively ?

A. λ(Stormy) = 3.939 failure/yr, λ(Good) = 0.0656 failure/yr.

B. λ(Stormy) = 4.202 failure/yr, λ(Good) = 0.0525 failure/yr.

C. λ(Stormy) = 6.594 failure/yr, λ(Good) = 0.0458 failure/yr.

D. λ(Stormy) = 20.16 failure/yr, λ(Good) = 0.0403 failure/yr.

9. Given a reliability growth test in progress having accumulated 4 failures during 5000 test hours. Assume a growth rate of 0.3, what is the expected MTBF at 25,000 hours ?

A. 1250 hrs     B. 1895 hrs     C. 2026 hrs     D. 3856 hrs

10. In a certain application, two identical transducers are used to measure the vacuum in a system. The system is considered to have failed if either of the vacuums read by the transducers varies from the standard by more than 10mm Hg. Which of the following is the correct reliability logic block diagram for the transducer assembly ?

Answer A

Title: Strategic Planning in the Midst of a Crisis

Oct 15, 2020 12:00 PM Eastern Time (US and Canada)

Presenter: David Auda

Earlier in the year I ran a webinar to explore strategic planning after the onset of an unanticipated disruptive event, covid. The material content shared the benefits of doing such a plan even after the disaster had struck and laid out steps on how to develop such a plan. I suspect that a good number of businesses and/or leaders simply cobbled up a reaction plan to mitigate losses and then attempted to put together some types of a tactical plan to navigate the duration of this disruption by addressing new risks as they develop. Such a plan would fall into the category of Risk Based Decision Making. It is a plan that most likely highlights waiting for something to happen, making another containment/mitigation plan while hoping that the problem will go away soon. Hope is not a strategy. Risk based Decision Making is rooted in the visceral not the logical mindset. There is of course another side to this, the opportunistic side. So given that the return to normal, the new normal, is not even on the calendar yet, there is still plenty of time to lay out a strategic plan that includes not only sustainability through this prolonged disruption, but explores the potential opportunities that are coupled to it. What would appropriately fall into the category of a growth mindset rather than a fixed mindset.

Picture © B. Poncelet

Picture © B. Poncelet

1. Given the density function obtained  under accelerated conditions:  Determine the density function under normal operating conditions,  given an acceleration factor of 5

 a. (1/150)exp(-t/150) 


c. (1/30)exp(-t/6)  

d. (1/30)exp(-t/150)

2. Which of the following is NOT a type of sample?

a. acceptance sample  

b. SPC sample   

c.  Application sample  

d. Measurement system correlation sample

3. Failure rate derating curves are not  dependent upon:

a. Environmental stresses  

b. Operating life 

c. Failures per hour  

d. Component application levels

4.  FMECA classifies each failure mode according to:

a. Probability    

b. Criticality    

c.  Severity   

d. Unreliability

5. Allocation of functions to personnel and equipment in combinations to achieve the required reliability is defined as  

a. Human Factors allocation  

b. Design Factors allocation                                        

c. The function allocation      

d. Cross-functional allocation

6.  The reliability of the logic diagram is:

  a.  0.1800   b.  0.5918   c. 0.7796   d. 0.8201

7. What does the failure mode and criticality number (Cm) replace in the most common qualitative methods?  

a. Severity number    

b. Risk Priority number

c. Failure Effects Number   

d. Failure mode number

8. Two gages are used to inspect an item, yours and a supplier’s. A correlation sample of n=25 units was inspected using both gages. The  std dev using gage #1 is 20, using gage #2 it is 30. What is critical value of the appropriate  statistic to test the null hypothesis that the  variances of both gages are equal at 5% significance  (1-sided test)?

 a.  1.98   b.  2.25   c. 4.85   d. none of the above

9.  The acceleration factor for increasing vibration from 50 units to 200 units is 6.4. At a vibration level of 200 units,  the time to failure is lognormal with a scale parameter  of 0.8 and  a location parameter of 3.2. What is the reliability under normal conditions at time=100?

a. 0.7136    b. 0.8271    c. 0.8361     d. none of the above

10. The equation below represents:

a. the Weibull distribution transformed for probability plotting

b. the Arrhenius model

c. the lognormal hazard function

d.  the linearized Eyring model

ASQ RRD series webinar: Warranty Analysis…in the Real World

Thu, Nov 12, 2020 12:00 PM – 1:00 PM EST

Presenter:  Dr. Joseph Voelkel

Warranty analysis is offered in many software packages, and the examples in those packages are usually based on simple sets of data. By contrast, this talk presents a case study of how such an analysis was conducted for a more complex problem. Join the speaker on this journey, with its numerous twists and turns.

Dr. Joseph Voelkel is Professor Emeritus (retired July 2020), School of Mathematical Sciences, Rochester Institute of Technology, Rochester, New York, where he had been Chair of the Graduate Statistics Program. In addition to teaching graduate students, he consults for a wide range of clients. His focus ranges from explaining fundamental statistical methods and mentoring Six-Sigma-project teams, to teaching advanced techniques and developing novel methods to solve complex client-specific problems. He is currently consulting and is also engaged in contract work for the Rochester Data Science Consortium at the University of Rochester.

Picture © B. Poncelet