Accelerated Life Testing 

 

Life Estimation Techniques

What is the life expectancy of our product? Can Wyle perform a test that will provide a good estimate of its life?

These are questions that are asked of Wyle very often. The Federal Aviation Administration (FAA) is demanding to know the life or the mean-time-between-failures (MTBF) of new devices, instruments and equipment for aircraft. All branches of the Department of Defense, Army, Navy and Air Force, similarly require that the manufacturers of certain systems provide life data. The same is true for a variety of other Government agencies including NASA. Foreign organizations are demanding life or MTBF data of American products. So, it is no wonder that manufacturers are interested in having an independent third party such as Wyle Laboratories provide an estimate of the life of their products.

The life of a product can be estimated by four methods:

  • Method A: Prediction methodologies using generic data and models (before the product is manufactured).
  • Method B: Assessment methodologies using actual product failure data (after the product has been on the market).
  • Method C: Laboratory simulation of normal usage and environmental conditions (can be done on prototype units, but takes a long time).
  • Method D: Laboratory simulation using accelerated stresses and environments requiring mathematical methods of correlating with normal usage and environmental conditions.

The chart that follows lists the four methods for estimating the reliability and some of the advantages and disadvantages of each. Some manufacturers utilize more than one method. It is very common to use Method A during design and development and use Method B after the product or system is in the marketplace where actual failure history is accumulating. Method C, chronologically, could follow Method A since it would probably be used only in the event that Method B cannot be implemented. Likewise Method D would only be justifiable if Method B is not possible. In some instances, a Governmental agency may request life information from a manufacturer whose product is in the prototype stage and has no history. In such a case, Method D may be the only method possible.

Methods For Estimating Reliability

Method A

Prediction Using One or More of the Methods and Data Bases

Method B

Assessment or Demonstration using Actual Product
Historical Data

Method C

Laboratory Simulation of Normal Usage and Environmental Conditions

Method D

Laboratory Simulation Using Specifically Designed Accelerated Aging Methods

Advantages

-Generally simple
-Relatively inexpensive
-Provides an idea of how the product will behave in application

Advantages

-Requires customers to keep records of the time the device or system has been operating and to keep records of failures
-Reliability can be validly computed

Advantages

-Can be done by the manufacturer
-Test is run at normal time and stress (unaccelerated)
-Total number of failures during test time used to compute failure rate

Advantages

-Accelerated aging shortens test time
-Requires fewer tests than Method C
- One-Stress version used in the nuclear industry to qualify systems

Disadvantages

-Value lies in the ability to compare various concepts and potential modifications and/or redesigns
-Good estimator of a product reliability before the product has accrued failure history
-Not as good as estimates based on failure history

Disadvantages

-Customers usually will not cooperate without incentive or may not have warranty data on product

Disadvantages

-Requires large numbers of test articles/devices
-Requires long time to complete test
-Usually costly

Disadvantages

-Process of testing and test control are tedious and complex
-Based on assumption that failure modes in accelerated test are the same as in unaccelerated testing. (Verify by testing at more than one stress level.)
-Could require 6 months to run two test groups at three stress levels

Method D was formulated by R. Giuntini


Method D is not a single method but a type of method that must be developed for each individual product. To best do this, Wyle would first perform a Product Life Evaluation similar to what we do for the nuclear industry. This would consist of first, performing a product review of the engineering drawings, parts list, materials list, and any other relevant information; second, ranking the failure potentials of the parts and materials; third, determining the life-limiting factors that are stress-related; and fourth, developing an accelerated aging plan and methods unique to the product.

The method or combination of methods suitable for one product will be different from one for another product. For example, the life of nonmetallics could be modeled by means of the Arrhenius equation for temperature degradable materials with known activation energies. There are other products where Miner's rule may be appropriate.

Certain electronic devices and other products can be aged by increasing the environmental stresses of vibration and temperature in test chambers while the devices are powered. For this type of program, several small groups of test articles would have to be subjected to several different accelerated conditions.

A simple graphic below illustrates the nature of the accelerated aging program. This diagram shows two environmental stresses operating to reduce the life (accelerate the aging) of the test articles. Further, it shows that the life is inversely proportional to the stress (i.e. high stress, short life and low stress, long life). In this diagram, the normal usage life, which is what we are looking for, is indicated on the plane that was extracted from the three dimensional model.

To derive each of the two curves shown in (1) Vibration Vs. Life plane and (2) Life EstimationTemperature Vs. Life plane requires at least two sample groups of test articles (i.e. products) to be tested at different stress levels.

At the higher stress levels for both temperature and vibration, five or six test articles would be required for each. At the lower stress levels, 10 to 12 would be required. For both temperature and vibration, a total of around 36 test articles would be used.

A test could take a few days or a few weeks to complete. The complexity, size, and composition of the individual test articles are constraining factors in the test duration and cost because of the limited capacities of the test chambers.

After the planned test data is derived, it will be mathematically modeled and analyzed to produce the estimate of the product's life under normal usage and environment conditions.

HIGHLY ACCELERATED LIFE TESTING (HALT)

HALT is a multi-stress test methodology that accelerates the aging of a test article (i.e. device or product) through sequential step increases of the stress environments well beyond those that the product will encounter in normal shipping, storage, and usage.

HALT should not be mistaken with what some engineers refer to as "shake and bake." HALT is much different. For one thing, the test machine is a triaxial random vibration and thermal cycling test system. The table of the machine is composed of a number of irregularly-shaped polygonal sections which operate independently from one another.

The HALT philosophy itself is totally different from those presented in various Government standards. It is a process that forces design and process maturation by accelerating stresses that cause failures typical of the product to occur much sooner. HALT is performed on prototypes or first articles so that weaknesses and flaws can be uncovered and corrected prior to production of the marketable products. HALT shortens the time to market.

The accompanying flowchart, which can also be read as a .pdf, provides a logic diagram for the implementation of the HALT process. An examination of the flowchart will indicate the things that must be done in an expeditious manner to get results as rapidly as possible so the flaws can be corrected on the prototype and the test can be continued.

The first two blocks of the flowchart are vital. Every product is unique and every HALT HALT processtest plan and test procedure must capture that uniqueness to achieve maximum utility. In addition to the test procedure, the final document, the Test Report provides a record of every aspect of the test as it was conducted with complete traceability including analysis, interpretation, thorough documentation with photographs and accompanying video of HALT process. A HALT typically can be accomplished in two to four 8-hour days.

As a rule, a HALT program will consist of seven (7) tasks shown in the following flowchart. Of course, the individual tasks must be tailored to the product to be tested. Each task has an accelerated stress-duration profile associated with it.

A HALT program cannot be adequately addressed with a few paragraphs and diagrams. For more in-depth information regarding HALT or other reliability engineering tools, methods, and processes, contact:
service@wyle.com