W. Stephen Tait, Ph.D.
Chief Science Officer & Principal Consultant,
Pair O Docs Professionals, LLC
Corrosion Corner
Is corrosion testing
cost-effective?
Hello everyone. Sometimes it’s tempting to take a shortcut
by either skipping or abbreviating corrosion testing
during new product or derivative product development.
Such shortcuts are sometimes justified by budget constraints
and/or lack of resources, plus the belief that a formula has no
corrosive ingredients or that small changes in formula chemistry
and/or package materials won’t affect spray package corrosion.
Last month’s Corrosion Corner discussed when corrosion testing
should be conducted on new products and derivative products
(line extensions). It provides the background for this two-part
series on the cost effectiveness of corrosion tests. Let’s start the
new discussion with a few definitions.
Package service lifetime
Package material corrosion has the potential to reduce spray
package service lifetime. Package service lifetime is defined as the
filled-package-age when:
• Packages leak product or propellant
• Valves leak propellant; or
• Partially full packages cease to spray
In other words, service lifetime is the length of time during which
spray packages and valves function properly.
Statistical confidence & statistical risk
Statistical confidence is based on the theoretical normal distribution.
Statistical confidence increases as the number of replicate
samples increases and statistical risk is 100% minus the percent
statistical confidence.
Correlation/empirical probability
I use the terms “correlation” and “empirical probability” interchangeably.
Correlation is the percentage of times a test prediction/
26 Spray February 2020
estimation is the same as the actual corrosion. Think of correlation
as a grade on how well a corrosion test predicts the actual
corrosion—higher correlations mean more accurate predictions.
Corrosion test objectives
Corrosion tests have four basic objectives to:
1. Determine if there will be spray package corrosion
2. Determine what types of corrosion will occur
3. Determine the rates for each type of corrosion
4. Use the rates to estimate the package service lifetime with a
given formula
Typical spray package service lifetimes range from 2–5 years,
depending on the type of product.
Corrosion test statistical confidence
& statistical risk
High statistical confidence means a low risk. Figure 1 provides an
empirical relationship between risk as a function of the corrosion
test percent completion. It was developed from many decades of
spray product corrosion test data and its correspondence to actual
package corrosion.
The Y-axis for Figure 1 is the % correlation (as risk) between
predictions from corrosion testing and the actual corrosion
in commercial spray packages. The X-axis has the test percent
completion for storage stability tests at constant temperature and
electrochemical corrosion tests.
Electrochemical tests are significantly shorter than storage tests.
Hence, the percent-completion on the X-axis in Figure 1 allows
direct comparison between the risk for completed storage corrosion
tests and electrochemical corrosion tests.
Notice in Figure 1 that the empirical no-corrosion-testing risk is
approximately 62% that corrosion will occur. This risk is for all
types of corrosion, however not all types of corrosion cause package
or product failure. In other words, Figure 1 is not the risk for
package failures.
The risk for (100%) completed storage testing is around 7% and
the corresponding risk for completed electrochemical testing is less
than 1%. Hence, the risk associated with electrochemical tests is
significantly lower with a significantly shorter test length than the
corresponding one-year storage test.
The risk for abbreviated storage testing ranges from approximately
30% after 25% completion, to an approximately 12% risk
after 75% completion—three and nine months of storage testing,
respectively.
In other words, the probability of surprise corrosion in commercial
spray packages is high with no corrosion testing (~62%) and
remains high with abbreviated storage test data. Indeed, a 7% risk for
a completed storage test is considered by some to also be too high.
Part 1