How to Select the Right ALT Approach

How to Select the Right ALT Approach

byFred Schenkelberg, CRE, CQE

Senior Reliability Engineer, Ops A La Carte LLCfms@opsalacarte.com // www.opsalacarte.com // (408) 710-8248

ASQ Silicon ValleyQuality Conference

Objectives

• Introduce thought process to select ALT approach

• To discuss trade-off decisions involved

• To enable you to select the right ALT approach for your situation

Presenter BiographyFred Schenkelberg is a Senior Reliability Engineering Consultant at Ops A La

Carte. He is currently working with clients using reliability assessments as a starting point to develop and execute detailed reliability plans and programs. Also, he exercises his reliability engineering and statistical knowledge to design and conduct accelerated life tests.

Fred joined HP in February 1996 in Vancouver, WA. He joined ESTC, Palo Alto, CA., in January 1998 and co-founded the HP Product Reliability Team. He was responsible for the community building, consulting and training aspects of the Product Reliability Program. He was also responsible for research and development on selected product reliability management topics.

Prior to joining ESTC, he worked as a design for manufacturing engineer on DeskJet printers. Before HP he worked with Raychem Corporation in various positions, including research and development of accelerated life testing of polymer based heating cables.

He has a Bachelors of Science in Physics from the United States Military Academy and a Masters of Science in Statistics from Stanford University. Fred is an active member of the RAMS Management Committee and currently the IEEE Reliability Society Santa Clara Valley Chapter President.

• Respect constraints– Time, expenses, sample size, accuracy

• Use use conditions– Environment, frequency, loading

• Minimize assumptions– Statistical, modeling, variability

• Focus on failure mechanisms– Stress, damage, measurement

Guiding Principles

With no constraints

• Build products and deploy, measure time to failure for every unit.

– Real time– Real stresses– Real variability– “Customer” defines failure

Paperclip Experiment 1

• What is use profile?

• Environment

• Probability of success

• Duration

• Failure definition

• Benefits– Perfect results– No assumptions

• Problems– Takes time– No Sales, or– Customer risk

No Constraints

• Use product more often– Increase use cycles per day– Maintain other stresses at use

levels– Measurement system detects

specific failure

• Assumes specific action leads to failure– Wear, fatigue, accumulation– Must understand or assume failure

mechanism

Time Compression

Acceleration Factor (AF)

Time shiftTime shift

Paperclip Experiment 2

• Assume clipping action leads to eventual failure

• How many cycles?

• What forces, angles?

• How many?

Time Compression

• Benefits– Less time– May use a sample– Simple extrapolation

• Problems– Not suitable for time

dependent failure mechanisms

– Nor, 24/7 loaded products

Stress to Life Relationship

Paperclip Experiment 3

• Assume bend angle is related to damage accumulation

• Angles?

• Other variables?

• Benefits– Creates AF model – Product of use specific

• Problems– 1 or 2 stresses or

failure mechanisms– Logistically challenging

Stress to Life Relationship

Acceleration Model• A model exists describing the stress

to life relationship– Arrhenius, Peck, Coffin-Manson, prior

work – The failure mechanism,

use profile, and environment within applicable limits of model

• If failure mechanism doesn’tchange, may use comparative testing in absence of model, giving up on accuracy of result.

Paperclip Experiment 4

• Assume failure mechanism is the same for new product

• Best to run till failure and conduct detailed failure analysis

Acceleration Model

• Benefits– Single stress to apply– Fewer samples

• Problems– Assumptions

multiplying– May miss new failure

mechanism

ALT w/ AF model

• Peck’s Relationship

• Norris-Landzberg

Step Stress

• Variation of acceleration model ALT, with samples experiencing step increases in stress– Quicker than acceleration model method– May require more samples– Requires careful control of stress and failure

detection– Best with accumulated damage failure

mechanisms

Degradation

• Useful when performance decays over time in a measurable way– Often used with an acceleration model– Failure is defined by threshold value– Non destructive measurements permit

repeated measures on samples

ALT Limits

• Extrapolation

• Stresses

• Failure Mechanisms

Wearout

Sample Size

Success Testing

Weibull Success Testing