Product Validation Testing is a critical and expensive endeavor. The part build process must align with the latest prototype process (for Design Validation (DV)) or production process (for Product Validation (PV)) and be fully documented for posterity. Depending on the product and application, the validation test plan consists of a battery of tests, some of which are lengthy – often six months or more. Because of this, DV and PV test plans are invariably on the critical path for a customer program. Test failures that require fixing the design and repeating DV or PV jeopardize project timing and company profits. Further, they jeopardize the customer program along with your company’s reputation.
For these reasons the best policy is that you should not start validation unless you expect to pass. And you cannot expect to pass unless you’ve given your parts a chance to fail on test exposures that are similar to those that your product will see on test and in the field. This is often done during the design verification or design confirmation phase of a program.
These test exposures take different forms. In each case we are less concerned with accurate and lengthy field-correlated tests than we are in quickly finding and fixing failure modes. Four accelerated testing methods are provided here from the most qualitative to the most quantitative. In each case, test parameters are informed by project team experience and risk assessments (e.g., DFMEA).
- A HALT (Highly Accelerated Life Test) usually takes less than a week to run and is a purely qualitative activity using just a few parts/assemblies. The tests are run very early in the design cycle using specially designed HALT chambers. They utilize both steady state and stepped profiles of temperature, vibration, humidity and electrical power. Hit it with a sledge hammer (figuratively), fix the failure modes that arise, and continue on until you approach the limits of the technology.
- A Proportional Overstress Test (aka MEOST) is a semi-quantitative step-stress methodology run on 4 to 6 pcs that applies increasing levels of stress in succeeding steps until failures are observed. The selection of stress values for each environment (temp, vibe, volts, etc.) and the duration of each step are such that they are applied “proportionally” to field exposure, so that any particular environment is not over-represented on test. Stress and time both increase with each step and are thus normalized and combined into a stress-time metric that serves as the independent variable for a modified “Weibull” plot. As such the methodology is best used for A-B comparisons and failure mode identification than as a tool for predicting life.
- Damage models and physics of failure models are related and may be used to develop an accelerated version of an accepted legacy test. An accelerated test profile is designed to have equivalent damage (usually thermal or mechanical) as the legacy test. Analytical models that align with the primary physics of degradation are employed to calculate damage. Tests developed as such have the advantage of running on existing test equipment, are designed to exercise the failure mode(s) that keeps the project team up at night, and are still quite short. E.g., a 3-month PTC test reduced to 9 days. Though quantitative, the correlation to life in the field is usually not firmly grounded due to excessive acceleration.
- An ALT (Accelerated Life Test) is a quantitative methodology for estimating reliability (R/C) at customer use-level conditions by running tests to failure in the lab at multiple elevated stress levels. At each stress level four to eight parts are run to failure or suspension. A life/stress model (e.g., Arrhenius or Power Law) is then used with the corresponding life distributions to estimate the distribution of time-to-failure at customer use conditions.
In this phase, the cost to iterate (fix) the design is relatively low since the test is short and it’s early in the program. Hard tools are not yet cut and significant capital investments have not been made. And since you have given your parts a chance to fail early in the development cycle and your design has improved as a result, you have some reason to expect to be successful in validation.
As you move along into validation and consider DV and PV test plans, the more apt question at that point is: Am I giving the right parts a chance to fail on the right tests? In short this means that in validation we wish to test representative parts on tests that have been rigorously correlated to field use conditions. Otherwise, you risk making ill-informed design and program decisions and inaccurate assessments of product reliability.
I will cover this in my next article, Validation Testing – Right Parts on the Right Tests.
Leave a Reply