Weathering Experimenter's Toolbox
Henry K. Hardcastle III
Atlas Weathering Services Group
It is important for weathering researchers to maintain a collection of tools for addressing complex weathering phenomena. An overview of selected tools for weathering experimentation is presented. These tools include gage characterizations, weathering test methods, approaches for weathering experimentation and weathering philosophies. Many of the tools may be applied to understanding variation in weathering phenomena. Application examples of weathering studies and empirical data are included.
It is important attendees of this seminar walk away with specific tools they can apply at their location. The approaches presented here within represent a collection of tools for the production engineer. Often acquiring a single simple tool from a seminar and applying it successfully in a manufacturing environment offsets the cost of attending. It has been this author's manufacturing experience that acquiring and using a specific tool, procedure, or method from a seminar can often be of greater value in the long run than general concepts presented. It is in this spirit the following presentation is made. This paper presents an overview of tools available to the weathering researcher. Many of these tools not only apply to weathering research but also to a variety of technological applications. These tools - like all others - if used improperly, can result in erroneous decisions with catastrophic results; or likewise, if used with skill, can enhance product performance and customer satisfaction. This treatment only reviews a number of tools available to the engineer. This treatment does not present application and methodology details. Proper use of these tools includes reviewing procedures, cautions and warnings as specified in appropriate standards.
You can often figure out a solution to a difficult problem by knowing what tools are available to apply to the problem. This concept is embodied in the manufacturing line trouble shooter (maintenance engineer) standing next to broken production equipment with his tool box. The trouble shooter opens one drawer after another scanning the contents. A tool is selected and applied to the problem. Often, the tool is replaced and a different tool is selected and tried until an appropriate choice is made. The number of different tools in the collection is important. If the only tool available is a hammer, it is amazing how everything begins to look like a nail. One difference between inexperienced and experienced problem solvers is the number of different tools in the box and a greater familiarity that often results in faster selections of more appropriate tools. This analogy applies to weathering research projects. Weathering phenomena can be considered as an extremely complex and highly variable process. Like manufacturing processes, weathering processes can be effectively studied using simple tools. Ironically, many of the tools used to study manufacturing processes also apply to study of weathering processes. It has been the author's experience that if you know how to listen, the voice of the process will tell you what is going on more quickly and accurately than seminars, conference room discussions, consultants, literature, etc. This presentation is about tools that help researchers listen to the weathering process.
Keeping with the analogy described above, this presentation can be described as an overview of selected tools in a weathering experimenters toolbox. When a difficult weathering project is undertaken, it is hoped this collection may provide a diverse set of tools for weathering researchers to apply to help solve problems and gain understanding. It is also hoped that attendees will contribute additional tools to the collection by a short note to the author. In this way the toolbox can grow and increase in applicability each time it is presented.
Organization of such a collection is difficult. To begin to develop this tool box analogy, we have grouped the tools into "drawers". These drawers are no more than a structure for organization and do not pertain to importance, priority, or relationships.
The drawers have been labeled as follows:
- Test Methods
- Test Philosophies
Within each drawer are a number of tools that will be brought out and reviewed in this presentation. The tools are numbered for presentation organizational purposes only and have no meaning beyond this presentation. The list of tools is as follows:
- Gage Capability and R&R
- Temperature Effect on Colorimeter Measurements
- C&C Study
- Deltas vs. Delta of Deltas
- Round Robin and Intercomparisons
- Weathering Test Method Conitnuum
- Not All Eggs in One Basket
- Surveillance Testing
- Worst Case Approach vs. SLP
- Evolutionary Jumps in Natural Weathering Test Methods
- 90° vs. Other Angels
- n = Replicates
- n = Environments
- n = Exposures
- Blocking Variables
- The Student's t-Test
- Shainin's Six Pack Test
- Single Variable; On-Off
- Single Variable; Which is Better
- Single Variable; Ramp
- Two Variable Squares
- Three Variable Cubes
- Fractional Factorial Arrays
- Fractional Factorial Mean Analysis
- Fractional Factorial ANOVA Analysis
- Confirmation Experiments
- Evolutionary Jumps in Experimental Approaches
- Levels of Context
- Interactions of Formulation, Processing and End Use Environment
- Means vs. Variation; Where the Money Is
- Significance vs. Importance
- Correlation is Not Causation
- Interface of Man Made and Natural Phenomena
1. Gage Capability and R&R
Measurement of weathered surfaces is not always easy. Often, surface degradation results in non-uniformity across a specimen's surface. Small size target areas of some measurement instruments may yield highly variable data when measuring different spots of the same specimen. Weathering researchers should be familiar with variation associated with measurements before making inferences regarding measured weathering data. Sources of variation associated with any measurement include Repeatability and Reproducibility. The R&R study is a powerful tool weathering researchers can use to characterize measurement variation. A fine treatment of R&R application may be found in Larry B. Barrentine's "Concepts for R&R Studies" from ASQC Press. The Barrentine treatment utilizes the revised G.M. long form and describes Repeatability as the variation of measurements of a gage and Reproducibility as the variation in measurements by operators. With some simple modifications, this method can be adapted to characterize variation due to different weathering exposures and devices. "A total measurement system must go beyond R&R and eventually include all sources of measurement variation."
Application of this tool using Barrentine's approach for 20° gloss readings on automotive paint specimens was performed as follows. A set of ten different current automotive coatings with varying degrees of gloss was obtained (designated A-J). The specimens were measured by three different operators to characterize Reproducibility (between operator variation). Each operator measured the set three times to characterize the Repeatability (within operator variation). Data for the measurement trials was recorded as follows.
|X-BAR A||84.4||R-BAR A||6.6||X-BAR B||80.1||R-BAR B||2.7||X-BAR C||79.1||R-BAR C||1.3|
* Beyond UCL for R
For this application we wanted to use the gloss meters to be able to differentiate groups that were as small as four units apart in 20° gloss. From calculations in the modified G.M. long form and the Barrentine Book, the following values were obtained:
Measurement Unit Analysis:
- Repeatability - Equipment Variation (E.V.) = 10.9
- Reproducibility - Appraiser Variation (A.V.) = 2.73
- Repeatability and Reproducibility (R&R) = 11.2
- % E.V. = 100 x [(E.V.) / (Tolerance)] = 272 %
- % A.V. = 100 x [(A.V.) / (Tolerance)] = 69 %
- % R&R = 100 x [(R&R) / (Tolerance)] = 280 %
The values are based on 99% of the area under the normal curve. Using this analysis, we compare the variation attributed to these sources to the tolerance value. We cast a suspect eye on differences closer than approximately 11 units. We know the gloss measurement system does not have sufficient resolution to discern smaller differences at 99% confidence. We can not confidently say that gloss differences smaller than 11 units are due to weathering since our measurement variation appears to be larger than this difference for these specimens using this measurement system.
2. Temperature Effect on Colorimeter Measurements
A researcher's characterization of gage behavior may be critical to successful implementation of a project. It is the weathering researcher's responsibility to qualify the measurement systems used in his/her research. Recent advances in hand held colorimeters and spectrophotometers has led to their widespread use throughout weathering technology. These hand held instruments provide the freedom and flexibility to measure full size operating systems in end use environments. These instruments (as with all measurements instruments) provide opportunities for gross errors in weathering data. An example of a recent study involved hand held spectrophotometer evaluations of samples on exposure in Florida. Unexplainable variations in the data prompted investigation of the instrument. Technicians walked out to backed exposure racks and measured coated panel color while panels were on exposure. Hunter a color values would varied as much as 0.5 units within a week's exposure. It was noted that on some days, the weather was cool and cloudy and on other days the weather was warm and sunny. These conditions seemed to co-vary with the measured a values.
An experiment was performed utilizing a commercially available yellow automotive coating on steel substrate. Thermocouples were attached to the painted surface. Initial measurements were made at room temperature. The panel was placed on a hot plate and re-measured at a variety of temperatures. The following graph was obtained.
Even though the color instrument was calibrated at the ambient air temperature of the exposed specimens, the different solar loading under clear and cloudy conditions effected the surface temperature of the specimens resulting in color measurement variation. This phenomenon represented a critical understanding and limitation to the use of this measurement method for field measurements. Characterizing limitations of measurement systems represents an important tool for weathering researchers.
Some of the most critical experimental work for weathering researchers lies with characterizing measurement systems for weathering studies. Quite often measurement systems were designed for more traditional laboratory prepared specimens than the variety present in naturally or artificially weathered specimens. Effects of limited target area, surface non-homo-geneity, surface contaminants (including biologicals), equilibration issues (including dark time reactions), etc., can result in frustrations while trying to answer specific research questions. Just as materials have adverse responses to climates, measurement systems sometimes have adverse or unexpected responses to weathered materials. Similarly, as weathering researchers perform experiments to understand weathering phenomena, researchers may also perform designed experiments on measurement systems interaction with weathered materials. Most tools for weathering research outlined in this presentation may also be useful for characterizing measurement system behavior.
3. C&C Studies
C&C studies (Control and Capability studies) are an important tool in manufacturing processes. C&C concepts may also be applied to weathering processes and represent an important tool for weathering researchers. C&C concepts are especially important for artificial weathering processes and measurement processes.
Statistical control is usually assessed using control charts. Control charts document process output over time and indicate outputs due to special causes ( non-random pattern, trends, or points beyond calculated control limits). Application of control charts to weather data, instrumental measurements, and exposures of materials can characterize these processes for non-random patterns. For instance, a weathering researcher may choose to expose standard reference materials in accelerated exposures along with test specimens. Control charting the degradation of these standards over many exposures may indicate the predictability and consistency of repeated exposures through time. Weathering researchers often perform repeated measures using standard materials on optical measurements instruments. Simply plotting this data on control charts and analyzing the data for the state of statistical control can offer insights regarding the measurement process.
A capability study compares the distribution of repeated measures to specification limits or tolerances. Capability analysis can be easily adapted to accelerated weathering processes. For example, consider a lot of metal halide lamps used for acceleration and specified to cause a degradation of standard material within ± 0.5 D b* units of a target value. If a second set of lamps produces a distribution of degradations outside the ± 0.5 D b* limits, the second set is said to be not capable of meeting the specification requirements. It is important to remember, however, if a process is not in control, it is difficult to predict what the process will do in the future. Thus, process capability cannot be assessed until a process is in control. Cp and CpK indexes express the width of specification limits to the measure of the actual variation of the process.
4. Deltas vs. Delta of Deltas
Modern measurement instruments offer superior long term stability over instruments of only a few years ago. For example, filter wheel colorimety which was susceptible to long term drift has been largely replaced by spectrophotometry using stable holographic gratings. Still, variation due to measurement system effects a weathering researchers ability to discern small differences in test groups (especially important for early predictions).
Probably the most popular tools for reducing the effects of these variations is to use the central limit theorem, increase sample numbers or measurement numbers and report means and grand means. Another tool, however, utilizes blocking with control specimens. In this approach, each material specimen is cut in half. The first half is measured and then exposed. The second half is also measured but retained in a controlled environment archive. The delta between the halves' initial measurement is recorded. At the conclusion of the exposure, the exposed half is re-measured along with the archived half. A delta between the two pieces is again calculated. The two delta values are then compared and a "delta of deltas" is obtained. This delta of deltas blocks a variety of instrument variations and may improve gage capability over very long exposures.
5. Round Robins and Intercomparisons.
Round robins and intercomparisons are invaluable tools for weathering researchers relying on weathered material analysis at different locations using different measurement systems. A round robin or intercomparison is a simple procedure where a set of materials is circulated between and measured on each measurement system involved in a project. Often, customers and vendors circulate a standard set of specimens between locations on a regular basis. Specimen sets should approximate the full range of variation expected in a weathering program (e.g. many appearance measurement devices show highly repeatable results on white or light colored materials, but may show poorer reproducibility with darker materials as signal to noise ratios become smaller). Many times researchers fix mis-communications problems regarding procedural details during the intercomparisons rather than after procedural variations have occurred in destructive weathering tests.
One excellent source for a variety of intercomparisons is The Collaborative Testing Services, Inc. (CTS). CTS is a privately owned company that specializes in inter-laboratory tests for a wide variety of company sectors including rubber, plastics, fasteners, metals, container board, paper, and color. Over 2000 labs from the U.S. and more than 50 other countries participate in CTS programs. The programs allow laboratories to periodically compare the performance of their testing with that of other laboratories.
For The Color and Appearance Collaborative Reference Program, paint chip samples are distributed four times per year to participating laboratories. Gloss and color readings recorded by participating laboratories are reported to the CTS. Results of all laboratories are compared in tables and graphs based on conditions of measurements used and distributed to participants. For Further information contact Collaborative Testing Services, Inc. 340 Herndon Parkway, Herndon, VA 20170. Phone (703) 742-9107.
6. Weathering Test Method Continuum
Organizing the scope of weathering test methods into logical order helps the engineer visualize tools available for weathering tests. The following figure presents one such organization. In general, there appears to be a trade off between confidence and accleration of test methods. Between specific methods and materials, however, this trade off becomes less obivious.
7. Not All Eggs in One Basket
Relying on several test methods rather than a single method provides robust results. Initiating several smaller tests along different points of the testing continuum rather than in one large complex test program using a single methodology, reduces testing risk and does not put all the eggs in one basket. This may be analogous to most 401K programs where risks are spread across a portfolio of different investments. Comparisons between several test methods and multiple replicates within methods also result in a higher level of information.
8. Surveillance Testing
Prudent development engineers precede accelerated testing with identical specimens on natural, real-time exposures. Development engineers often need to show how accelerated results directly relate to natural exposures. Only direct comparison between identical specimens exposed to natural and accelerated methods can achieve confidence in the accelerated methods for specific material types, formulations and lots. After initial product approval by a manufacturer, drifts in formulation, manufacturing process or handling often occur. "Surveillance testing" involves regular sampling from production lots and places samples on natural exposure using a quality control approach. For example, several manufacturers sample production lines once a year and place the samples on natural exposures with a very limited frequency of evaluation. In the unlikely event of customer complaints sometime in the future, reference data from "surveillance testing" is readily at hand. Surveillance data also offers opportunities to anticipate customer dissatisfaction before it arises in the market. Inexpensive, simple, regular surveillance testing provides a level of assurance throughout the product life cycle. Much of the data included in this presentation has been obtained from AWSG surveillance tests.
9. Worst Case Approach vs. SLP (Service Life Prediction)
After considering the most critical weathering variables, engineers should characterize the end-use environment in terms of these variables. This usually leads to the question "Which end-use environment?" An automobile used in metro Detroit, MI area experiences considerably different weathering variables than the same vehicle used in Los Angeles, CA or Orlando, FL. Engineers should design components for robust performance in worst-case end-use markets. By using this design criteria, the probability of satisfactory performance in milder end-use environments (those with milder levels of critical weathering variables) usually increases. Typically components fail fastest in "worst case" end-use environments. Exposure in these areas accelerates failures due to increased levels of critical weathering variables including irradiance, temperature, and moisture. For these two primary reasons; 1) Characterizing performance in "worst case" end use environments and 2) Accelerating the rate of failure due to increased levels of critical weathering variables, industries focus material natural weathering exposure tests on two major US locations; South Florida and Desert Arizona termed "Reference" or "Standard" Environments.
10. Evolutionary Jumps in Natural Weathering Test Methods
Advances in weathering test methods can be visualized as levels of evolutionary jumps in weathering test method sophistication:
Level 1-Exposures in reference environments represent a first step in accelerating degradation from traditional end use markets. Comparison between South Florida and Phoenix, AZ offers an effective technique for understanding these environments with respect to solar radiant exposure, temperature and moisture.
Level 2-Modern exposure facilities employ racks attached by a single axle to fixed vertical members. The axle allows rack pivoting to the appropriate angles from horizontal as needs arise. Unbacked, backed, under glass or other exposure enhancements then clamp to the pivoted frame. Although simple in retrospect, the pivoting frame advancement provided basis for considerable development of natural weathering test methods. This humble, simple improvement in rack design has allowed more advances in conventional weathering technology than any other single advancement. Static Exposures easily relate back to full system end use exposures. Increases in critical weathering variables can significantly affect the rate at which materials degrade. Simple optimization of location, backing and angles can modestly accelerate degradation rates.
Level 3-Data from the static exposure angles discussed shows the effect of angle on increasing critical weathering variables. A dynamic exposure, which varies orientation in response to the seasonal variation in the sun's path, can increase critical weathering variables throughout the year.
Level 4-The evolution in exposure methods from static to variable angle exposure racks dramatically increased levels of radiant energy deposited on specimens. A similar jump in exposure development occurred with automatic tracking mounts that follow the sun's path from sunrise to sunset. These sun tracking mechanisms dramatically increase solar irradiance and represent the next milestone in natural weathering acceleration methods.
Level 5-After obtaining maximum acceleration from normal incidence, sun tracking exposures, engineers often want to accelerate UV degradation beyond these methods using NATURAL sunlight. Inventors developed an elegant solution to concentrate the image of several suns onto a single target area of the test material. This method became known as the "EMMA," an acronym for Equatorial Mount with Mirrors for Acceleration. Standards for this test method include ASTM G90, ASTM D4364 A, ISO 877 and SAE J1961.
Level 6-The next quantum leap in evolution of natural accelerated weathering test methods (like the pivoting rack, follow-the-sun trackers and EMMAqua concentrators) may come from hybrid test methods, utilizing combinations of different natural, accelerated, and artificial weathering test methods.The design for special exposures is only limited by the engineer's ability to link back or correlate to the reference environments, imagination, and funding. Remaining within the paradigms represented by standards is important for comparisons of performance between different vendors, materials, processes, quality control issues, and the like. For research and development of materials and processes, however, understandings often come from experimentation outside the standards requirements and utilize novel test method approaches.
Note-Methods of acceleration and increasing critical weathering variables may present considerable risks for correlating testing results back to full system end use exposures. Engineers should only use accelerated test methods in conjunction with test methods presented so far. ASTM G 90 clearly states "No accelerated exposure test can be specified as a total simulation of natural field exposures."
11. 90° vs. Other Angles
In early weathering research, paint systems were applied to vertical walls facing south and exposed for performance evaluations. Soon after, it was noted that changing the angle from 90° to 45° South produced faster degradations. This increase in degradation rate was significant and important. Since that development, researchers have proposed a number of other exposure angles often described as more optimal for faster degradation rates. A variety of explanations includes ratio of direct to diffuse component of sunlight, cooling of night sky, condensation on surfaces, etc. Horizontal, at latitude and variable angle have all been recommended over the 45° exposure primarily for faster degradation rates. For some materials, changing standard exposures from 45° to these other angles can result in important increases in degradation rates. However, for others, changing from 90° to 45° results in a quantum leap in degradation rates but changing from 45° to other "more optimal" angles offers only marginal increases in degradation rates.
For example, identical pieces of polystyrene reference material were randomly selected from a single lot. Two pieces were placed on different angles of exposure in Florida simultaneously. The degradation in yellowness index was measured after eight and eighteen weeks exposure from May to August 1999. The following results were obtained.
The differences between the 90° and other angles of exposure were significant and important. The differences between other angles may have been significant but were much less important. Visual analysis indicated the difference between 90° and other angles was important - but not important within angles other than 90° . Changing from 90° to other angles of exposure may represent an important tool for weathering researchers, whereas changing between other angles may offer less acceleration depending on the material and environment.
12. Sample Size: Never One
With application of statistical analysis to weathering experiments, the age old question of appropriate sample size surfaces. Collecting answers to the "How many Samples" question has provided this author with a great deal of entertainment. Some of the answers recorded include; "At least 30", "Five if you want a warm Fuzzy. Thirty if you want to be sure.", "It Depends.", "Ten, but they must be randomly selected.", "As many as possible given the economic constraints.", etc.
One of the reasons the sample size question is sometimes difficult to answer is that it points to a critical issue weathering researchers must address, "Does the central research question deal with location of a results distribution, the dispersion of a results distribution or both?" For instance, will an additive change the mean failure time of a coating from six months to six years (a question of location), will an additive change the range of coating failure times from ± six months to ± six years (a question of dispersion) or will the coating performance with the additive be significantly different than without the additive (a question of both location and dispersion). Oftentimes weathering researchers focus predominately on research aspects regarding location alone (means).
The practice of exposing a sample size of one (n = 1) does not acknowledge the possibility of variation in the natural weathering process. Single sample weathering exposure submissions are based on the premise that the variation within and between the material, processing and the environment is known and small with respect to changes observed. The confound, however, is that this premise cannot be established without submitting sample sizes greater than one! There is a quantum leap in the level of information obtained in going from a single sample to a sample size greater than one. Obviously, as sample sizes increase beyond one, the data becomes more robust and if randomly sampled, soon becomes useful in describing the distribution of the parent population, and can be applied to inferential statistical analysis and can be transformed into a normal distribution with the central limit theorem.
To begin to answer the appropriate sample size question at it's simplest level, one may understand the considerations involved by examining a simple z-test. Application of a z test points to 4 basic considerations;
- The minimum difference between two groups to be detected (e.g. a minimum difference of 5 delta L* units or 0.05 delta L* units between the test group and the control group)
- The inherent variation in each of the two groups (e.g. s = 0.05 or s = 0.50)
- The acceptable risk of saying the two groups are the same when actually they are different.
- The acceptable risk of saying the two groups are different when actually they are the same.
In summary: effect size, variance and risk tolerance.
| (mean1 - mean2) |
Z = ---------------------------
Sp (1/n1 + 1/n2) 1/2
Compare with Z table value AT A CONFIDENCE LEVEL
As effect size, variance and risk tolerance change, a different sample size is dictated. As discrimination levels get smaller, n needs to increase. As variation inherent in the populations gets larger, n needs to increase. As tolerance of risk of making incorrect decisions gets smaller, n needs to increase. By back calculating from statistical tests, researchers can estimate the appropriate sample size without using "rules of thumb." These calculations require researchers to first characterize the inherent variation or dispersion in the population which, by definition, requires sample sizes greater than one.
The simple practice of including randomly selected replicates in a weathering exposure greatly increases the information value of the data. Consider the following weathering data of a commercially available blue automotive coating exposed to weather for three months in Florida.
A simple comparison may lead to the interpretation that backed exposures lead to greater change. Adding additional replicates to this weathering study, however, may lead to a different interpretation.
Replicate exposures represent one of the most important tools for weathering researchers. Replicates may include random samples of a short production run, several production runs, or several years production runs, a single manufacturing line, a manufacturing plant or several plants. By tracking the levels of sampling (levels of production context), the weathering researcher may gain valuable insights into the causes of weathering variation due to production variations.
14. n = Environments
Customers typically use products in a variety of end use environments. Characterizing material degradation in a single environment often yields meager information with respect to actual customer experience. Single environment exposures do not provide the researcher with information about material degradation rate variation in end use. This shortsightedness can be catastrophic if the single environment does not represent worst case service conditions for the specific materials and failure modes. Degradation character changes from environment to environment for many materials. The researcher gains information regarding variation in degradation due to different end use conditions by exposing materials in different environments. Formulation changes to enhance degradation properties in Florida may have different consequences in Arizona. A coil coated materials exposure in Arizona and Florida shows differences in degradation as follows:
15. n = Exposures
Traditionally, weathering researchers have looked at the issue of sample size as the number of replicate pieces of material placed on a single exposure. Often, more valuable information is obtained by considering multiple exposures and blocking other variables. For instance, simultaneous exposures of a single material at a variety of exposure angles, backed condition, treatments, formulations, etc., may characterize degrees of variation (dispersion) in a materials weathering response. It is important that the researcher block other variables. Here researchers obtain information regarding variation between exposure types.
For example, we wanted to investigate the effect of exposure angle on Polystyrene yellowness index. We simultaneously exposed 2 replicates of Polystyrene chips on different exposure angles in Florida. The following results were obtained. The data indicated the variation we could expect for the different exposure angles in Florida was smaller than expected except for the 90° exposure. n = number of replicates = 2. n = number of environments = 1. n = number of exposures = 6.
16. Blocking Variables
Blocking variables represents a required tool for most weathering experiments. Blocking techniques provide a method for normalizing nuisance factors in experiments. The term "Blocking" comes from Agricultural Science experiments that had to be performed against a background of natural variation much like many weathering experiments today. In the agricultural application, experiments had to be performed in soils that varied with location. Early Ag researchers developed blocks of land where soil conditions were more homogeneous. Other experimental variables could be applied and the results attributed to these test variables rather than the soil variability.
In weathering research, we must block a plethora of nuisance variables in order to understand the effect of the variables being studied. For example, if a weathering researcher is trying to study the effect of formulation changes on material weatherability, it may be important to block production variables by using a single manufacturing line during a single manufacturing run to produce the test specimens with different formulations. This may block variations due to line and run. Inventive application of blocking techniques can allow weathering researchers to untangle a large number of root causes of material degradation. These blocking techniques represent the foundation for many weathering experimental designs.
17. Student's t-Test
A t-Test is a simple statistical tool weathering researchers often use to determine if two groups significantly differ. A rigorous treatment is left to more esoteric and involved statistical publications. Weathering researchers should be familiar with the theoretical underpinnings of these types of hypothesis tests. Student's t-tests are often used when a large number of replicates is not available, normally distributed mean values are available and an easy methodology is desired. An example of a t-Test applied to weathering results follows:
Two batches of automotive paint were obtained from a single supplier with the same formulation. We wanted to see if batch to batch differences in raw materials, manufacturing processes, and application processes could result in significant differences in weathering. Six specimens were randomly selected from the first batch and six specimens were randomly selected from the second batch. We wanted a high degree of confidence that the weathering was truly different if we rejected the null hypothesis (a > 0.95). The specimens were measured initially for color and placed side by side on Florida 5° South exposure racks for six months. After six months, the specimens were re-measured with the following results;
|Batch 811||? ?* after 6 months||Batch 921||? ?* after 6 months|
These results were then analyzed using the Student's t-Test in the following manner:
Step 1. Null Hypothesis (research question)
The mean D b* is the same for batch 811 and batch 921 after 6 months weathering in Florida, 5° backed South.
Step 2. Calculation of sample values
|Batch 811||Batch 921|
|n1 = 6||n2 = 6|
|Mean1 = 1.055||Mean2 = 0.823|
|S1 = 0.068||S2 = 0.034|
|(n1 - 1)S12 + (n2 - 1)S22|
|SP||(---------------------------------------------) 1/2||= Estimate of Pooled Variance|
|n1 + n2 - 2|
SP = 0.054
Step 3. Calculation of Test Statistic
|mean1 - mean2|
|t =||-----------------------------------||= 7.469|
|SP ( n1 + n2 )1/2|
Step 4. Determination of rejection region
For á = 0.05, n1 + 2 - 2 = 10 degrees of freedom
tcrit. = 1.8.125
Step 5. Conclusions
The calculated t value (step 3) falls in the rejection region and therefore chance is an unlikely explanation for the observed difference between the weathered D b* value for the different batches. The null hypothesis is rejected and it is concluded that the weathering degradation is different for the two sets.
Even though the two paint samples were from the same vendor, of the same formulation, and certified to be equivalent, their weathering behavior was significantly different when exposed side by side in Florida. Since the exposure variables were blocked (same exposure rack, same exposure period, same measurements, same mounting, etc.) we may infer that significant weathering differences were due to variation in formulation, manufacture, or application of the coating. The Student's t-Test has provided a powerful tool for discrimination that includes both the location of the means and dispersion or variation within the two groups.
18. Shainin's Six Pack Test
Many current weathering studies use parametric statistical tools to analyze results. Researchers should not overlook non-parametric tools for weathering data analysis. Non-parametric analysis can be especially useful in weathering studies involving appearance issues and visual evaluations for acceptability ranking. Also, often non-parametric analysis does not require normally distributed data to be effective. Dorian Shainin has developed a system of unique techniques for process analysis in manufacturing environments. Shainin developed the concept known as "The Red X" source of variations in manufacturing processes. Shainin's logical approaches to production processes may also represent valuable tools for investigating weathering processes. One such tool includes Shainin's "6 pack test." Results are rank ordered. The ranking is then associated with input variables. The probabilities associated with a specific order are then evaluated.
For example, two different lots of automotive paint were obtained from a single supplier with a single formulation. The research question was to see if lot to lot differences could result in different weathering characteristics. Three randomly selected specimens from each lot were exposed side by side, backed in Florida at 5° South for six months. After weathering, the specimens were measured for color. The D b* values were ranked. The ranking and corresponding batch numbers associated with each D b* values are as follows:
|Rank Order of Six Specimens||? b* value after 6 mos., Florida, backed, 5°||Batch Numbers|
There are 20 different equally likely rank orders for these 6 specimens. Similarly there is only a one in twenty chance that the three "C" specimens would all rank above the "T" specimens. The likeli hood of this particular order from chance alone is 1/20 or 0.05. This handy little tool is an effective way to identify vital variations with out requiring normality or distribution tables. Readers are highly encouraged to seek out additional Shainin techniques for weathering processes, as well as, manufacturing process solutions.
19. Single Variable: On-Off
Single variable or "on-off" trials probably represent the earliest levels of experiment. This type represents "Experiment" in it's simplest application, forms the foundation for all other levels of experimentation, enjoys wide use in the weathering industry today, and represents a good starting point for developing a lexicon describing weathering experimentation evolution. In this level, the experimenter wants to understand the effect of a single variable on a system. The system is tested with and without the variable applied. Other conditions are kept constant or "blocked".
This type of experiment answers questions such as:
- Will doubling line speed sacrifice critical weathering properties?
- Will addition of photo stabilizer improve impact strength retention after three years of Florida exposure?
- Will adding new lubricant "X" effect gloss retention on Arizona exposure?
- Will increasing filler content by 10% effect loss of color retention on Florida exposure?
- Will increasing the process temperature by 30° C effect the weathering color stability?
- Will adding regrind decrease tensile strength after exposure?
- Application of this experimental design is simple. The experimenter prepares two sets of samples for trial, sample set "A" without the variable applied, sample set "B" with the variable applied (often referred to as "The Input Variable" or the "Independent Variable"), and exposes the two sets side by side to the weather. After a period of exposure, the experimenter measures the two sets and analyzes the measured values (often referred to as the "Output Variables" or "Dependent Variables"). The analysis looks for differences between the two sets.
The following graphic shows a representation of this simple variable design.
For example, a red automotive coating was subject to a 45° south exposure for 12 months in Arizona. Specimens were placed on plywood backing. One specimen was exposed to a 30 second water spray every hour during the day. The other sample did not have water spray. The results obtained are as follows:
20. Single Variable: Which is Better?
Formulators may often use the single variable approach in weathering experiments for alternative vendor evaluations or "Drop In" formulation component qualification. In this design, one includes multiple pairs instead of a single pair of experimental samples.
This type of experimental design answers questions such as:
- Whose tin mercaptide stabilizer has the least effect on weathering, Vendor "A", Vendor "B" or Vendor "C"?
- Which impact modifier has the best long-term leathering performance, the current vendor or a candidate vendor?
- Is there one TiO2 grade that offers better weathering performance in my system?
- Which colorant pigment system offers better performance in my system, powder pre-blend or liquid color?
- Which is the best solvent system for my coating from these alternatives, Methylacetate, Acetone or PCBTF?
- Application of this design is similar to the single variable. The experimenter prepares a separate sample set for each alternative candidate (usually including a standard set to compare performance against as a control). The experimenter measures the output variable(s) of all sets after exposure and analyzes for significance and importance.
The following graphic shows a representation for this widely used design.
21. Single Variable: Ramp
The previous two approaches were interested in "does it" or "does it not"- presence or absence of a characteristic. The next level of sophistication involves a different research question. Process engineers often use a "ramp" experimental design to determine "how much" of an input variable effects performance. This design involves amounts of an input variable and represents a higher level of context than simple presence or absence. This design is the most widely used tool for optimizing formulation component levels today.
The following graphic shows a representation of a single variable design with multiple levels of the variable.
Most Weathering Research utilizes this type of approach for dosage studies. The following experiment changes the dosage of UV and measures the materials' response to the increasing levels
22. Two Variable: Squares
A jump in the sophistication of weathering experiments occurs when designs utilize more than one input variable. Rarely, if ever, does a product's manufacturing or in-service environment involve a single variable. Two variable experiments involve different research questions than those of simpler designs including investigation of interactions between input variables. Often these experiments are represented as "Square" designs.
This type of experimental design answers questions like:
- Which has a bigger effect on color change; UV irradiance level or condensation level?
- Do exposure irradiance and film thickness interact or are the effects of one on the weathering performance independent of the other?
- If higher process temperatures cause unacceptable yellowing after four years in Florida and if higher thermal stabilizer levels cause unacceptable yellowing after four years in Florida, will both higher process temperatures and higher stabilizer levels produce unacceptable yellowing after two years, four years, as an additive effect, as a multiplicative effect or at all?
- How much improvement is there in tensile strength after exposure if I increase photo stabilizer "A"? How much improvement will there be if I increase photo stabilizer "B"? How much improvement if I increase "A" with "B"?
- How much acceleration is achieved by increasing the irradiance level in a Xenon arc Weather-Ometer from 0.35 W/m^2 at 340nm to 0.6 W/m^2? How much acceleration is achieved by increasing Black Panel Temperature from 75° C to 85° C? How much acceleration is achieved by increasing irradiance level while increasing Black Panel Temperature?
One of the most popular applications of these "Square" designs involves identifying a low and high setting for each independent variable. Four trials are then conducted according to the following array;
|Trial 1||Both Variables At Low Setting|
|Trial 2||Variable "A" at Low Setting||Variable "B" at High Setting|
|Trial 3||Variable "A" at High Setting||Variable "B" at Low Setting|
|Trial 4||Both Variables at High Setting|
Multiple "identical" samples or replicates are often included in each trial to characterize the background variance with in each trial. Output from each trial may be graphed on the same coordinate system for easy comparison with two low settings and two high settings compared for each variable. These graphs are read differently than traditional Cartesian coordinate systems. Often, intermediate input variable settings between the high and low settings are added to this design as additional trials and reveal intermediate topography. Sometimes this analysis is referred to as a "response surface".
The following graphic shows a representation of this design.
These designs are also referred to as "Full Factorial" designs if each factor or input variable is tested at each setting. These designs are also said to be orthogonal or balanced: two trials with "A" set low, two trials with "A" set high, two trials with "B" set low, two trials with "B" set high. The othogonality is sometimes best understood by visual examination of the geometric layout of the "Square" design. The orthogonal characteristics of these designs result in experimental efficiency and power and advance these designs to a higher level of sophistication than the traditional weathering experiments widespread throughout industry today.
As an example, we were interested in determining if angle of exposure or daytime spray had a greater effect on color change of a Red Automotive Coating after 12 months exposure in Arizona. We also wanted to see if there was an obvious interaction between these variables. We selected the low and high setting for the angle variable as 45° and 5° respectively. We selected the low and high setting for the spray variable as either no spray or with spray. After 12 months exposure, in Arizona, the following data were obtained.
It is obvious from this graphical analysis that presence of spray had a greater effect on the 12 month Delta E values. There does not appear to be an obvious interaction between the angle and spray variables in this experiment.
23. Three Variable: Cubes.
A simple enhancement to the two variable "Square" concept is to add a third variable resulting in a "Cube" experimental design. This simple advancement drives experimental sophistication to a new level of context. These designs are more appropriate for weathering experiments since very rarely are only two weathering variables acting simultaneously on a product in end use. Within this three-dimensional experimental volume resides the classic weathering interaction of temperature, irradiance and moisture variables. A two dimensional experimental design does not have the inherent level of context necessary to characterize temperature-irradiance-moisture interaction phenomenon. It is befuddling to this author why the vast majority of weathering research is conducted at less sophisticated levels of experimentation to gain understanding of systems that operate at high levels of context.
Experimenters ask similar types of questions of three variable cube designs as for the two variable square designs;
- For this thermoplastic, how do intensity, duration and frequency of irradiance interact to cause failure and which is most important?
- What effects do melt temperature differences, gate pressure differences and dwell time differences have on cold temperature impact after five years exposure?
- Of the three variables; exposure in Arizona vs. Florida, exposure in backed vs. unbacked condition, and exposure with water spray vs. without water spray, which has the biggest effect on gloss change. What is the effect of interactions between any two or all three of these variables?
- Which has a bigger effect on impact after three years Florida exposure, 10% differences in impact modifier level, 10% differences in lube package levels, or 10% differences in process aid levels; and is there an interaction between these three formulation components?
- If I change irradiance levels by 20%, temperature levels by 20%, and moisture levels by 20% in my Xenon Weather-Ometer, can I identify a weak link in my formulation? Specifically, must I stabilize for all three variables or can I just stabilize for one and, due to the synergistic effect of these three variables, stabilize all three for the price of stabilizing just one?
Application of this "cube" design follows similar procedures as for square designs. Eight Trials (23) are performed with individual variables set to high and low settings independently of other variable settings. The design is full factorial (each variable at each setting of high and low) and othogonally balanced.
The following graphic shows a representation of a three variable design.
At the end of the experiment, the experimenter has obtained data from:
- 4 trials with "A: set low,
- 4 trials with "A: set high,
- 4 trials with "B" set low,
- 4 trials with "B" set high,
- 4 trials with "C" set low and
- 4 trials with "C" set high
by only performing a total of 8 trials! This experimental efficiency constitutes the advantages of full factorial orthogonal arrays over less sophisticated experimental designs typically used in weathering studies today.
As an example, we wanted to compare the effects of three variables on the weathering behavior of a Red Automotive Coating in the following ways; 1) We wished to observe the effect of day time water spray, 2) We wished to observe the effect of exposure angle and 3) we wished to observe the effect of an Arizona exposure as compared to a Florida exposure. Specimens from the same batch of material were randomly selected and exposed under a unique set of conditions in a blocked manner. The organization of the three variables and the results on the Red Automotive Coating were obtained as follows:
The water Spray and exposure location appear to have had a bigger effect than exposure angle in this experiment. There appears to be an interaction between angle and location. There appears to be an interaction between spray and location. An interaction between angle and spray does not appear obivious.
24. Fractional Factorial Arrays.
Manufacturing Process Engineers widely use fractional factorial screening experiments to screen out the "trivial many" variables from the "important few" variables. A small confirmation experiment (typically two to four additional trials) follows a screening experiment to confirm the results. Once the manufacturing process engineer identifies the important variables, the engineer can focus process improvement efforts in an efficient manner on those variables that the process indicates are important. This analogy described for characterizing and improving manufacturing processes also applies to improving weathering performance. Weathering processes are multi variable complex processes that are highly material dependent. Weathering investigators can use fractional factorial screening experiments to screen out the "trivial many" variables from the "important few" variables. After a screening experiment indicates the most important variables for a particular material, a small confirmation experiment always confirms the results.
Once the weathering investigator identifies the important variables, he can effect the material formulation, processing variables and in-service environments to improve weatherability. Equally important, the investigator can optimize variables identified as "trivial" by the process to reduce manufacturing costs! The materials weathering research efforts can then be focused in an efficient manner on these variables that the process indicates are important.
Fractional factorial experimental design answers questions such as:
- Of the 9 components in this vinyl formulation, which have the biggest effect on yellowing after five years Florida exposure, what is each component's order and magnitude of importance on yellowing and which components can be optimized for cost without sacrificing weathering performance?
- Of the 10 major production line variables the line operator can control, on which should I establish control charting to improve quality of weatherability and approximately what mean and tolerances should I begin with?
- Of the major weathering agents this product will be exposed to (temperature, moisture, irradiance, pollution, abrasion, solvents, biologicals, cycling, etc.) which require research efforts to improve customer satisfaction for weathering performance?
- For the major weathering failure modes I have identified in the FMEA, what are the risks associated with each?
- For my material, which of these many weathering variables can be increased in order to accelerate weathering for test development?
- For this vendor's candidate material, which environmental variables have the biggest effect on the system's weatherability?
Application of fractional factorial screening and confirmation designs is relatively simple to implement but relatively complicated to describe thoroughly. This presentation describes a straightforward ten-step procedure leaving the "why" to more esoteric and involved statistical publications. The reader should become familiar with theoretical underpinnings of these designs. The reader should also practice application of these designs - beginning with inexpensive, simple, non-critical investigations - to gain experience with these techniques. The reader should not become dissuaded by complex, restrictive theoretical considerations that often become barriers to beginning these types of empirical applications in experimentation.
"The purpose of Mathematics is insight, not numbers." Sam Saunders, Ph.D., 02/08/99
The author has used the following ten step procedure effectively in many screening experiments both for investigating manufacturing processes and weathering processes. This ten-step outline may need modification for specific applications and is not exhaustive in detail but will serve to identify the major components and sequence for most uses. This ten-step procedure was used to perform the case study described in this presentation.
10 Step Procedure:
|Step 1||Write a simple, concise research question.|
|Step 2||List the variables to be investigated. Check that variables are truly independent.|
|Step 3||Select high and low settings (if a two level experiment) or high, middle, low (if three level experiment), etc. Check to make sure settings are not so far apart as to cause catastrophic failure if all variables are set high or low simultaneously. Check to make sure variable settings are far enough apart that they can be described as "significantly different."|
|Step 4||Select an appropriate fractional factorial orthogonal array for the number of variables under investigation, L16, L8, L64, etc. Use published arrays. Assign specific variables to specific columns in the design with regards to known interactions and alias concerns of the fractional factorial. Table 1 shows an L16 fractional Factorial Array.|
|Step 5||Perform trials according to the fractional factorial array schedule. Include multiple replicates within each trial as necessary given with in trial variability and gauge variability. Control the input variables to described levels for each trial in the array. Block other variables not designed into the experiment.|
|Step 6||Measure the desired output variables from each trial.|
|Step 7||Quality Control check all work. Recheck that all trials were performed in accordance with the array. Recheck all measurement values. Recheck all data entry. A simple error in variable settings, data entry, measurements, etc., may void the orthogonal basis resulting in erroneous decisions.|
|Step 8||Analyze the output using samples collected, effects graphs, and ANOVA techniques.|
|Step 9||Determine the significance and importance of each variable's effect on the output using the analysis.|
|Step 10||Confirm the conclusions. Run a minimum of two confirmation trials, one trial with the variables indicated as significant and important set to high levels, one trial with significant and important variables set to low levels. Set insignificant variables to cost effective settings for both of these trials. Check that the results of these two confirmation trials confirm predictions of the screening experiment in both direction and magnitude of differences. Additional confirmation trials can be added for additional understanding of main effects, interactions, and aliases.|
|1=Variable at Low Setting |
2=Variable at High Setting
|Temperature||Irradiance||Day Time Spray||Pretreat |
|Night Time Soak||Pretreat |
An L16 Fractional Factorial Array
After the experiments in the fractional factorial array were performed, the following data was obtained on randomly selected Polystyrene reference chips.
|Trial Number||Delta Yellowness Index of Replicate "A"||Delta Yellowness Index of Replicate "B"|
25. Fractional Factorial Mean Analysis
- Visual Review
A visual review was performed by laying out exposed specimens in groups. Groups were formed using the orthogonal array for each variable. For the first layout, all specimens that were exposed to higher temperatures were grouped together. All specimens that were exposed to lower temperatures were grouped together. The two groups were compared for overall appearance. Next, the eight sets of specimens that were exposed to high irradiance were grouped together and compared to a group composed of the eight sets of specimens exposed to low irradiance. This grouping procedure was continued for each variable in the experimental design. The groupings that revealed the most apparent differences between high and low settings were identified.
- Graphical Technique
A similar analysis to the visual grouping was performed using the Delta Yellowness Index values. A mean was calculated for the eight sets of specimens exposed to high temperature. A mean was calculated for the eight sets of specimens exposed to the low temperature condition. These two mean values were plotted on a graph and connected with a line. Next, a mean was calculated for all specimens exposed to high irradiance and a mean was calculated for all specimens exposed to low irradiance. These two means were plotted next to the temperature variable means. This procedure was continued for all the variables included in this design. This graphing technique allowed the effects of each variable to be compared with the effects of all other variables with a single graph. Using this analysis, it was quite simple to determine which variables had the largest effect on Yellowing of the polystyrene and the magnitude of the effect compared to that of the other variables.
Graphical Representation of Mean Analysis For Polystyrene Fractional Factorial Weathering Experiment.
26. Fractional Factorial ANOVA Analysis
For this analysis, we simply compare the effects caused by input variables to the background variation of the experiment. As this ratio approaches one, we say that the effects due to the input variables are not significantly different than the background variation of the experiment or that the effect of input variables is not significant. However, as the F ratio becomes larger and larger, the effects due to the input variables become more different than background experimental noise. A large F ratio indicates the effect of the input variables is significant.
Once the significance of the individual treatments is understood, it is appropriate to decide which input variables are important. Significance involves a statistical exercise. Importance is a human exercise and often depends on other sources of data beyond the scope the experiment. For instance, in this experiment, a 20% difference in irradiance resulted in a difference in mean Yellowness Index of about 3 units (21.5 at -10% to 24.5 at +10%). In some type of end uses, this magnitude of difference may cause great user consequences including formulation and/or process changes. In other end uses, this magnitude of difference might not be important to the end-use at all. Prior to beginning this study, the end-user of these polystyrene chips was interviewed regarding criteria for importance. The customer identified differences exceeding approximately 2 Delta Yellowness Index units at this experimental level to be important. Based on this information, the significant and important information can be completed as follows:
|Variables Tested That Are |
|Variables Tested That Are |
Significant But Unimportant
|Variables Tested That Are |
Significant And Important
|1. Temperature||1. Day Time Spray||1. Night Time Soak|
|2. Oven Pretreatment||2. Soak - Freeze - Thaw Pretreat||2. Abrasion|
|3.||3. Chemical Pretreatment||3. C-Arc - Pretreatment|
27. Confirmation Experiments
Confirmation trials should always be conducted in conjunction with fractional factorial screening experiments. Confirmation trials should also be considered as a critical part of the screening experiment. This is especially important if high levels of input variable saturation are designed into the orthogonal array and where significant interactions are identified between several input variables. Only confirmation trials can decode alias characteristics of the fractional array. Two confirmation trails were conducted with input variables set as shown.
|#1 - Least Degradation Predicted||#2 - Most Degradation Predicted|
| || |
| || |
| || |
| || |
| || |
The remaining variables investigated in the screening experiment that were identified as not important or not significant were set to optimal levels for cost for both trials as shown.
|Variables optimized for cost|
Conformation Trials Optimized for Cost 28. Evolutionary Jumps in Experimental Approaches
Reviewing the recent history of the weathering discipline (the past 100 years or so) one finds an evolution in the sophistication of weathering experiments. This evolution trends from simpler, single variable experiments towards more stochastic and broader experimental approaches. All the major types of weathering experiments evolved represent important tools development engineers can use for answering different types of questions in product development. Reviewing the levels of sophistication in this evolution provides an understanding of experimental tools available. This presentation reviews tools in the context of weathering experimentation. These tools also apply to most aspects of experimental science and one observes their use in agriculture, health sciences, industrial engineering, service industries, etc.
The on-off trial and single variable ramp experimental designs, variations and combinations represent the majority of weathering experiments performed today. Their application is easy and analysis is clearly communicated to technical and non-technical audiences. The application of these designs to weathering characterization is so widespread that dependence on these designs may represent a limiting paradigm constraining growth in understanding of weathering phenomena today. Exclusive use of these narrow focus designs represents important limitations for product designers and process engineers struggling with manufacturing capability and control issues. Experimentation at these levels of context alone does not provide an efficient approach for comprehensive manufacturing improvement efforts involving the plethora of variables effecting product weathering performance. Supplementing these approaches with experimentation at higher levels of context prior to application of these simple, fundamental designs represents a stochastic approach offering efficiencies for understanding weathering through experimentation.
A jump in the sophistication of weathering experiments occurs when designs utilize more than one input variable. Rarely, if ever, does a product's manufacturing or in-service environment involve a single variable. Two variable experiments involve different research questions than those of simpler designs including investigation of interactions between input variables. Often these experiments are represented as "Square" designs.
A simple enhancement to the two variable "Square" concept is to add a third variable resulting in a "Cube" experimental design. This simple advancement drives experimental sophistication to a new level of context. Very rarely are only two weathering variables acting simultaneously on a product in end use. A two-dimensional experimental design does not have the inherent level of context necessary to characterize temperature-irradiance-moisture interaction phenomenon.
The othogonal characteristics of these designs result in experimental efficiency and power and advance these designs to a higher level of sophistication than the traditional weathering experiments widespread throughout industry today.
The orthogonality of previously discussed designs allows a jump to the next level of sophistication in weathering experimental designs. As we increase the number of variables in a full factorial design, the number of trials required dramatically increases. For any number of variables in a full factorial performed at two setting levels (high and low), the number of trials required equals the number of setting levels raised to the power of the number of variables investigated (for example; 4 variables at high and low settings = 24 = 32 trials = 16 at high setting, 16 at low setting). Fractional factorial designs allow the experimenter to delete some of the trials required by the full factorial design as long as the orthogonality of the design is maintained. The power of the orthogonality can be virtually exchanged for efficiency in the number of trials. The resulting design is termed a Fractional Factorial Screening Experiment.
Clearly, the types of research questions addressed with fractional factorial screening approaches represent a different level of context than single variable and ramp level experiments which make up the majority of weathering experiments performed today.
There has been an evolution in the sophistication of experimental designs for weathering tests. The vast majority of current weathering exposures utilize more fundamental designs effecting few variables. This type of experimental design requires far more trials and thus more cost less information and poorer quality than more sophisticated approaches using screening fractional factorial experiments. Additionally, important interactions between weathering variables are not readily apparent using these simple experimental approaches. Preceding fundamental level, few variable weathering trials with fractional factorial screening and confirmation experiments represents an efficient, stochastic, powerful approach for improving knowledge regarding weathering's n-dimensional hyper volume of environmental effects on materials degradation.
29. Levels of Context
Often weathering researchers become distracted from their primary research goals. Considering the different levels of context involved in the weathering research project may represent an important tool for preventing these distractions. An illustration of different levels of context uses the behavior of Sand. For a researcher to effectively study behavior of sand, it is important he understand what level of context includes the behavior of question. Maintaining primary focus on the level of context of important behavior, with occasional efforts in other levels as needed to explain behavior on the focus level, is prudent. Spinning off the level of primary focus is often too easy, represents blue sky research, and is usually better left to academics than to industrial technologists in commercial settings.
The following four photographs consider the same subject (Sand) from four different levels of context. Different sets of variables effect Sand at different levels of context.
Covalent bond strength, crystal lattice structure, atomic diameters, inclusions, etc. effect Sand at the atomic level.
Surface energies, inclusions, fracture planes, matrix, etc. effect Sand at the agglomerate level.
Boardwalks, sanitary facilities, lifeguards, populations, etc. effect Sand at the commercial level.
Ecologies, plate tectonics, ocean currents, hurricanes, etc. effect Sand at a planetary level.
Studying bond strength of the Si-O complex may offer some insights at the commercial level but should probably not be the primary focus of a study trying to understand human interactions with Sand during summer months in Ft. Lauderdale.
Because of the diversity of weathering phenomena and the ability to study weathering at a large number of different levels of context, researchers often become distracted from original goals. Because of this diversity and complexity of weathering phenomena, researchers should understand which level of context they are working at and make sure efforts at other levels of context lead to understandings at the level under investigation. Using this tool of identifying levels of context can help researchers identify root causes, important variables, interactions, and critical behaviors of a materials response to the environment more effectively. Contextual understanding may help separate the critical from the trivial.
30. Interactions of Formulation, Processing, and End Use Environment.
From a design and product engineer's standpoint, it may be important to consider materials weathering behavior as an interaction between the formulation, processing, and end-use environment. A Venn diagram illustrates this interplay.
Currently much of the work being done in service life prediction is only focusing on the environment and a part of formulation (blending and compounding). Most current approaches leave out formulation variations and processing aspects even though it is generally accepted that many end use failures are rooted in processing and formulation variations.
Atlas WSG would like to extend an invitation at this time to manufacturing process owners to join Atlas in designing experiment projects regarding the effects of processing on weatherability. Please contact the AWSG R&D manager after the presentation.
Throughout manufacturing technology today, there is an increasing need to understand how interactions effect processes. The inherent inability of single variable experiments to characterize behavior of complex manufacturing processes has led process engineers to evolve experimental techniques for characterizing these interactions. Similarly, in weathering technology, single variable experiments do not provide the sophistication to characterize the effects of weathering interactions in many materials end use environments. Today, some weathering researchers have recognized the overwhelming role of complex interactions in weathering phenomena and are developing new tools specifically designed to look for and characterize interactions.
Simply put, if we find the difference in response between the levels for one variable is not the same at all levels of the other variables, there is an interaction between variables. This concept is often easier to visualize in a square design;
Often in weathering, very complex interactions between three or more variables may occur. The temperature, moisture, UV interactions represent one classic example that weathering researchers strive to characterize for many materials. By designing weathering experiments to seek out the natural interactions occurring on materials in end use environments, weathering researchers will develop important tools for characterizing materials behavior.
31. Means vs. Variations: Where the Money is
Typically, weathering researchers employ experiments aimed at characterizing mean values. For example, a researcher might wish to determine what the mean change in gloss is as a function of Arizona exposure. Developing experiments to characterize variation (the range of degradations) also represents an important and often under-utilized tool for weathering researchers. Many of the experimental tools outlined in this presentation, and generally used throughout manufacturing, can be easily modified to study variations of weathering phenomena. Also, variables that effect mean characteristics will often have effects on dispersion and can be studied using simple experimental designs. Many of today's robust automotive materials have mean degradation characteristics well within customers satisfaction tolerance. Generally, it is the infrequent dramatic variations from these mean values that result in customer dis-satisfaction. Studies of variation represent an important set of tools for weathering researchers to understand and characterize these events.
Often, identifying and controlling input variables that narrow variation in degradations can represent greater return on research investment than marginal changes in degradation means. The concept of the Taguchi Loss Function may apply to variation in weathering degradations; L=K (Yi - T)2 where K is a constant that converts deviation to a monetary value, Yi is the weathering characteristic of interest for the product and T is the weathering degradation characteristic target.
32. Significance vs. Importance
Many weathering researchers become confused regarding the concepts of significance and importance of weathering study results. Oftentimes researchers can become preoccupied with significance of results and remain unaware of the level of importance of the results, choosing to focus on the significant yet unimportant results of experiments. Significance is a statistical concept that is often involved with hypothesis testing. Significance involves a statistical exercise. Importance is an exercise in human judgment. Importance often depends on other sources of information beyond the scope of the experiment, often a tolerance. Assessment of significance takes into account location and dispersion issues. Simply stated, if two groups of data are subjected to an appropriate statistical hypothesis test (z, t, etc.) and the null hypothesis is rejected for a certain level of confidence, the groups are said to be significantly different for that level of confidence; e.g., "The formulation including additive "A" performed differently than the formulation without the additive and the difference was significant." The characterization of importance, however, speaks to a higher level of context; e.g., "Will inclusion of additive "A" result in a different customer perception than not using the additive?". "Will the effect of customer's perceptions be enough to offset the cost of using the additive?"
Criteria for significance is typically inherent in the experimental design and the results. Experimental variables must pass statistical tests for significance in order to be considered for importance. Researchers may benefit by having a clearly justifiable criteria for importance established before obtaining experimental results and compare significant results with importance criteria. Importance is not exactly a scientific tool, but does represent a tool to help keep research projects on track and addressing critical aspects of weathering phenomena rather than trivial aspects.
33. Correlation is Not Causation
In weathering results presentations, researchers often present correlation plots between two sets of variables (often dependent and independent) and take great efforts to describe the correlation; then make the assumption that since the variables correlate, it means that changes in one variable cause changes in the other variable. Actually, good correlation means only that the two variables co-vary. Correlation does not prove causation. Many times correlation analysis provides clues to uncover actual root causes in weathering phenomena but weathering researchers should not confuse correlation with causation. For instance, we may find a good correlation between people who eat potatoes and people who commit crime but we cannot say eating potatoes causes crime. Likewise, because we find a good correlation between UV dose and color change in end-use environments, we cannot say for sure, just from the correlation, that UV alone causes the color change. There may be other variables that are acting synergistically or co-varying with UV dose to cause the color change including possible thermal and moisture variables.
34. Interface of Man Made and Natural Phenomena
In past presentations, weathering has been defined as "the adverse response of a material to climate". A previous tool outlined in this presentation has described a materials weathering behavior as an interplay between formulation, processing, and end-use environment. Whatever the literal definition, conceptually, weathering phenomena takes place at a very critical interface between the man-made world and the natural world. Weathering researchers would do well to pause and compare these two very large and very different bodies of knowledge.
On one hand, technological aspects of man-made materials embody modern design and manufacturing characteristics. Variation is controlled, tolerances are held, theoretical engineering applied and modern industrial practices form raw materials into finished engineered products within economical constraints. This structured and controlled system is the world of material manufacture.
On the other hand, the variety inherent in natural science effects the environment of exposure. Environmental variations are not generally controlled by man and can be considered n-dimensional. Statistical distributions of nature's environmental variables result in what are called ecological niches in biological sciences. Many aspects of the "n-dimensional hypervolume" pressuring evolution in ecological niches also pressures man-made materials on exposure (Consider a number of different material failure modes at a similar order of magnitude as the number of biological species!). In many ways , biological sciences seem better equipped to deal with natural variation found in environments than the current industrial sciences approaches to weathering.
It is at the interface of these two great concepts; of the man-made and of the natural, that the weathering phenomena plays out. Researchers must acquire the skills to not only deal effectively in both arenas but also to relate them to each other. Interfacing these two great concepts is difficult, at best, given their different understandings, but is made even more difficult by researchers who elect to deal predominately with only one of these two bodies of knowledge. Understanding and characterizing the exposure environment without regard to the manufacturing aspects embodied in the material will not provide the researcher with an accurate perception of weathering phenomena. Controlling and effecting the manufacturing process without accounting for the variety and dimensionality of the "in service" environment will not provide the researcher with predictive understandings of weathering phenomena. Only when the researcher accounts for the real interactions between the numerous variables in both the man-made and the natural will the understanding of this complex interface advance to the next level. This point of view is a very important tool for weathering researchers to understand.
- J.M. Juran and Frank M. Gryna, Juran's Quality Control Handbook, 4th ed., McGraw-Hill, 1988.
- Bob Seymore and Jim Hayes, Critical Processes, Control and Capability - Module 2, Aluminum Company of America, Pittsburgh, 1990.
- Jeffrey T. Luftig, Experimental Design and Industrial Statistics, Level I-IV, Luftig and Associates, Inc., Farmington Hills, MI, 1987.
- Dorian Shainin, "Establishing Uncommon, But Entirely Logical Strategies For Improving Total Quality Management," ASQC Presentation, Ohio State University, Columbus, OH, 1993.
- G. Taguchi and S. Konishi, Orthogonal Arrays and Linear Graphs, Tools for Quality Engineering, ASI Press, Dearborn, 1987.
- Robert C. Duncan et al., Introductory Biostatistics for the Health Sciences, John Wiley & Sons, New York, 1977.
- Richard I. Levin, Statistics for Management, 4th ed., Prentice-Hall, NJ, 1987.
- Douglas C. Montgomery, Design and Analysis of Experiments, 3rd ed., John Wiley & Sons, NY, 1991.