This section addresses unique situations that I have encountered when qualifying test methods that assess product reliability.
It is, by no means, exhaustive. As new situations arise, we can add to this page.
Shaker table testing
Known "good" and "poor" reliability materials/construction
The test:
Pouches filled with liquid, hung in the shaker table mechanism (actually a box), subjected to continuous vibration in environment heated to 40 C.
Run multiple shaker table loads, each with randomized placement of "good" and "poor" materials.
Use several operators to place the specimens within each shaker table load. Randomly assign operators to specimens.
The environment:
Chamber (shaker table) heated to 40 C
Table vibrates (non-randomly)
The shaker table is instrumented to monitor and record temperature at several locations within the table throughout the test.
The failure mode:
Pouch leakage of some sort, usually at a seal (film to film, film to flange, film to other coupled component).
The analysis:
Use parametric survival (reliability) analysis. (May also consider non-parametric methods if needed.)
Reproducibility
Check to see if the "Operator" term is significant in the model. If not, then the test system Reproducibility may be acceptable. (For those operators in the study.)
Repeatability
If only one test system is available:
Check to see if the distribution parameters for the parametric reliability fit for "known good" specimens are consistent from one table load to another.
Check to see if the distribution parameters for the parametric reliability fit for "known poor" specimens are consistent from one table load to another.
If they are, then the test system Repeatability may be acceptable.
Can check this by using "table load" as an explanatory variable in a reliability-based regression model.
Can check and/or visually verify this by calculating the point and interval estimates for the reliability parameters (e.g. for a Weibull distribution, the shape and scale parameters) are equivalent across table loads. Can show this visually by plotting point and interval estimates for each parameter across all table loads.
If more than one test system (shaker table) is available:
Then we can compare reliability across test systems to see if each product reliability is the same across test systems. NOTE: It is not the product reliability that is being tested; rather, it is the sensitivity of each test system to product failure. Ideally we would want the sensitivity to product failure (reliability) to be the same across all test systems.
Discrimination between 'good' and 'poor' specimens.
Check to see if the analysis shows a difference in reliability (survival) between the 'known good' specimens and 'known poor' specimens
Drop test - repeated dropping of a liquid-filled bag (pouch) to screen for time (number of drops) to failure
Known "good" and "poor" reliability materials/construction ARE NOT AVAILABLE
The test:
Pouches filled with water (at a particular temperature).
Pouches are dropped from a specific height. Each pouch is dropped until it bursts. The number of drops until burst are recorded, as is the location and mode of failure.
The environment:
Bags dropped when filled with with room temperature (tap) water.
Bags dropped when filled with with water heated to 100 F.
The height of the drop.
This was determined from a preliminary experiment where the parameters of a two-parameter Weibull model were estimated for reliability at different heights. A balance was sought between height of the drop and reduction in uncertainty of the Weibull shape and scale parameter estimates.
Temperature of the test environment:
Temperature of the room.
The room in which the test system resides is at ambient conditions.
Temperature of the water fill.
For the particular products in the study, the hot water does not accelerate the primary failure mode of interest, but instead decelerates the failure mode. This is believed to be due to the change (decrease) in rigidity of the plastic film at the higher water temperature.
The failure mode:
Usually a break in a seal, sometimes where a film is pleated (for one particular product) or perhaps where two films are joined.
In one particular example, the bags (specimens) being dropped were found to have different reliability curves based on the temperature of the liquid.
Bags dropped when filled with room temperature (tap) water had worse reliability.
Bags dropped when filled with water at 100 F had better reliability.
In this case, the tap vs. hot water liquid may be used in place of product with "known poor" and "known good" reliability.
The analysis:
Otherwise, the same approach for the analysis would be used as for scenario 1 (shaker table testing).
Reproducibility, Repeatability, and Discrimination are checked in the same manner.
Leakage through a valve
The test:
Based on ISO 8669-2.
Applies pressurized liquid to challenge a closed valve.
Repeated several times with defined numbers of valve cycling in-between.
The environment:
This test applies usage rate acceleration.
Performed at ambient conditions.
The failure mode:
Leakage through the valve.
The analysis:
The test itself is meant to characterize reliability vs. cycles of valve actuation.
To test this Test Method, we would use different Technicians (Operators), each with their own sample of valves. We could also divide the overall sample of valves into "test lots".
We would check Reproduciblity of the test method by checking to see if statistically significant differences between "Operators" exist.
To check Repeatability of the test method, we could compare the values of the parameter estimates from a parametric reliability fit to each test lot to see if these parameter estimates differ significantly between test lots.
If specimens are available from both "known good reliability" and "known poor reliability" populations are available, then it may be possible to see how well the test method can discriminate between poor and good product.
Hot oil test in (hot oil) ovens for pouch reliability
The test:
Pouches are filled with vegetable oil which simulates certain properties (germane to product reliability) of the effluent collected by the pouches.
After being filled with oil, each pouch specimen is hung on a rack. Each rack can hold multiple pouches. An oven has several shelves (usually three), and each shelf hold several racks of oil-filled pouches.
The environment:
The oven is held at 40 C. There is no humidity control, but the oven may be instrumented to measure humidity in the oven environment during the test. The oven is also instrumented to monitor and record temperature during the test at several different locations throughout the oven's interior. The reservoir at each position on a rack is instrumented to record the time of failure (when the oil leaks out of a pouch) of a pouch.
The failure mode:
Pouches are filled with vegetable oil which simulates effluent collected by the pouches, and which has been known to induce a failure mode based on delamination of pouch film edge seals (vs. tensile strength of the seals).
The analysis:
Much the same as for the shaker table.
Reproducibility
Use several operators / technicians and look for differences between operators.
Repeatability
For only one oven, use a number of oven-loads and test lots. Compare results between test lots and between oven loads.
For several ovens, can use a number of oven loads for each and a number of test lots. Look for differences in product reliability estimates between ovens.
Discrimination
If specimens are available from both "known good reliability" and "known poor reliability" populations are available, then it may be possible to see how well the test method can discriminate between poor and good product.