Characterizing the Effect of Weathering Variables Using Accelerated Fractional Factorial Experiments
Atlas Material Testing Technology LLC decided to investigate new approaches to developing weathering methods. The close analogy between manufacturing and weathering processes indicated we might find method development approaches for weathering tests in traditional manufacturing method development. Some approaches widely used in manufacturing let the process itself direct method development rather than committees or individuals. We decided to try one of these approaches and let the weathering process itself direct weathering test method evolution by characterizing the effect of different variables on the weathering process.
Letting the weathering process direct the research and development of new methods seems to represent a viable alternative. Using fractional factorial weathering experiments to characterize weathering variables' effects on materials appears effective for engineering weathering research methods. For certain research efforts, this approach may be quicker, less expensive and more effective than many of the alternatives popular in weathering technology today.
Using Fractional Factorial Experiments for Weathering Investigations As part of our R&D strategy, we continued an ongoing investigation of the analogy between production and weathering processes. In production and manufacturing environments, engineers query manufacturing processes to understand root causes of variation. Engineers learn how to accurately and repeatably control manufacturing processes by characterizing the effect of different process variables and then controlling the important variables. The query takes the form of research questions embodied in designed experiments. These complex research questions initiate a virtual dialog between the researcher and the process.
(Applying Traditional Manufacturing Process Tools to the Weathering Process)
Researchers may obtain answers leading to powerful method developments by asking the weathering process how different weathering variables affect it. We selected a traditional manufacturing process tool to communicate with the weathering process: a fractional factorial experiment design. Fractional factorial designs allow many variables to be characterized with relatively few experiment trials. These designed experiments are often referred to as "Screening Experiments" because they can effectively screen out or separate the few important variables from the trivial many. These designs may also uncover important interactions between variables if properly planned. Figure 1 shows a diagram illustrating the use of screening experiment designs.
Materials and Dependent VariablesAlthough we studied four different materials in this effort, this paper presents results obtained only on a commercially available automotive coating described in Figure 2. Gloss represents an important characteristic of this material, and gloss weathering degradation in automotive coatings is widely studied throughout the industry1,2. Recent efforts to understand weathering effects on automotive coating gloss have included chemical analysis of the clearcoat using PAS FTIR (photoacoustic) spectroscopy3,4. We decided to study the gloss weathering behavior and several of the PAS FTIR characteristics in this effort. Figure 3 shows the gloss-degradation function observed in natural, real-time exposures, and Figure 4 shows an example of the PAS FTIR spectra obtained after natural, real-time exposures for the study material.
Exposure and Independent VariablesWe utilized a natural accelerated weathering test apparatus known as EMMAQUA™. This device is widely known in the weathering industry and described in the literature5. Recent studies indicate weathering of automotive coatings on Emma-type exposures may simulate the chemistry observed in natural weathering more closely than many other alternatives for accelerated weathering tests6,7. A collection of 16 EMMAQUA devices was utilized in this effort to answer the research questions.
The general research question(s) addressed in this case study can be stated as follows: "Of these specific weathering classes and settings, which of the weathering variables have the greatest effect on the physical and chemical weathering degradation, and which merit focus of additional R&D efforts for weathering method development?" Table 1 details the specific variables and settings designed into the screening experiment.
We wanted the screening experiment to direct the focus of our research to the most important variables for research and development. Full factorial natural (un-accelerated) weathering experiments would have required considerably more research resources and time than the fractional factorial accelerated approach. Other approaches (outlined in the introduction) may have delayed obtaining this important information indefinitely.
Design of Experiment The general theory of implementing fractional factorial experiments will be left to other publications and details pertaining only to this weathering experiment will be presented herewithin8-10. This experiment utilized the L16 design as shown in Figure 511. The columns show variables from Table 1, and the body of the orthogonal array indicates the individual variable settings for the 16 trials of the experiment.
(Asking the Process the Research Question)
We performed the 16 trials simultaneously using 16 different machines to achieve nuisance variable blocking. The array of machines required extensive modification to achieve the variable settings of the experiment design. Technicians qualified the 16 machines before and during the exposure to maintain the variable settings as closely as possible throughout the exposure. The exposure continued to 1318 MJ/m2 UV for the 10 mirror element machines (approximately four-year Florida equivalent at 5°). This paper presents the gloss and PAS FTIR results after completion of the exposure.
Table 2 shows the raw data obtained from each of the 16 experiment trials. The gloss values show two important characteristics indicative of a successful screening experiment; reasonable ‘within-trial' repeatability and good trial-to-trial variation. Researchers performed exhaustive quality checks of exposure conditions and specimen measurements before beginning analysis of the data.
Results and Analysis Gloss and PAS FTIR measurements indicated the effect of the variables studied on accelerated weathering of the automotive coating. We performed the gloss measurements in-house and relied on experts in the field to perform the PAS FTIR characterizations on a professional basis12. Traditional analysis of fractional factorial screening experiments typically involves two types of analysis: a visual graphic technique and ANOVA. We used both for all the dependent variable measurements. The measurement analyses answered the research question with fairly close agreement. We interpreted this agreement as a clear indication by "the voice of the weathering process," of which variables are important to pursue with focused research efforts for method development.
(Listening to What the Process Said was Important)
Mean AnalysisOne visual graphic technique uses a comparison of means to visually indicate the effect of each variable setting in the experiment. The L16 fractional factorial array prescribed eight trials with each variable set high and eight trials with each variable set low. Comparison of a mean of the eight trials set high to a mean of the eight trials set low for each variable indicates a relative effect of each variable compared to the others. Graphics in Figure 6 for gloss, Figure 7 for change in absorbance at 1550, Figure 8 for change in absorbance at 3300 and Figure 9 for change in absorbance at 1780 show these mean comparisons.
ANOVA AnalysisOne statistical technique uses an F statistic to compare the variation attributed to the input variables to the variation due to experiment error or variation between trials compared to variation within trials. The L16 fractional factorial array in this experiment included several blank columns that can be pooled to help estimate experiment error. Once main variable effects are determined to be insignificant in the results, these variables can also be pooled into the error estimate. Since the gloss measurements included two replicates, the gloss ANOVA analysis has higher degrees of freedom. Table 3 shows the F-ratios and degrees of freedom for each variable. Table 3 indicates the significant variables whose calculated F-ratios exceeded critical F-ratios as well as Rho values indicating relative importance for significant variables.
Confirmation Trials Fractional factorial screening experiments require confirmation trials. Aliasing inherent in highly saturated fractional factorial designs may confound main effects results with interactions. This experimental design did not fully saturate the array with variables and any significant interactions would have appeared in the columns left blank. Still, before refocusing R&D efforts on a new objective and adjusting research priorities, prudence dictated we carefully confirm the weathering processes behavior with independent exposures under different conditions.
(Confirming What We Thought We Heard the Process Tell Us)
Method Development Based on Screening Experiment ResultsOnce we knew the different variables effects on automotive coating weathering behavior, we had a powerful tool to focus research efforts. The screening experiment not only indicated what weathering variables to pursue, but also which to assign a lower priority in the overall research strategy. This knowledge allowed a Pareto-type adjustment of research projects based on experiment data. Staff could quickly develop an engineering approach to the new method once the data indicated a clear target. We then began designing the nighttime soak cycle into the EMMAQUA exposure.
Initial objectives focused on eliminating the labor-intensive manual repositioning of specimens from exposure target areas to soak tank and back to target areas each night. Although acceptable for small-scale, limited screening experiments and the like, manual repositioning represents increased labor costs, greater opportunity for operator-introduced error, and test variability for commercial-size accelerated weathering tests.
Increasing the soak frequency represented another design objective. Variable frequency allows adjustment of soak events to UV exposure ratios observed in natural real-time sub-tropical exposures. Virtually every night in South Florida, specimens become immersed in layers of water due to condensation. Good correlation between a new method and sub-tropical exposure may require adjusting the ratio of immersion time to UV exposure and cycle frequency. Flexibility for custom research projects as well as the ability to evolve the method beyond initial visions required additional open-ended design objectives for a new method.
With clearly defined objectives, we efficiently designed a new method. The design process itself introduced new enhancements and derivative embodiments in addition to the primary design objectives. Considerable implementation and application work still remains in order to bring the design to commercial viability; however, the process created powerful advancements in natural accelerated weathering test apparatus and methods. Furthermore, this research process developed method design features based on the foundation of data from the weathering process via the screening and confirmation work, rather than a method based on conference-room science.
The Patented MethodThe U.S. Patent and Trademark Office patent application process represents the longest phase of this project; more than several times the length of time required for the screening research and design development process. Fortunately, the time saved by performing efficient screening techniques and focusing development resources more than made up for patent processing time. U.S. Patent number 6,533,452 B1 entitled "Accelerated Weathering Test Apparatus With Soaking Cycle" was awarded March 18, 2003 with 20 claims. Whether or not the method in this embodiment achieves commercialization remains to be seen. More importantly, the design represents an important foundation for future development work regarding natural accelerated weathering test methods. Ultimately, the invention is relatively unimportant compared with the value of querying the weathering process for direction in weathering research efforts. In this sense, although only one inventor was named in the patent, the development represents a joint invention between the author and the weathering process. Characterizing the effect of weathering variables using fractional factorial screening experiments appeared effective for developing weathering research methods.
References1. Bauer, D.R. "Chemical Criteria for Durable Automotive Topcoats," Journal of Coatings Technology, 66, No. 835, 57 (1994).
2. Adamsons, K. "Chemical Depth Profiling of Automotive Coating Systems Using Slab Microtome Sectioning with IR/UV-VIS Spectroscopy and Optical Microscopy," Journal of Coatings Technology, 74, No. 924, 47 (2002).
3. Bauer, D.R.; Peck, M.C.; Carter, R.O. "Evaluation of Accelerated Weathering Tests For a Polyester-Urethane Coating Using Photoacoustic Infrared Spectroscopy," Journal of Coatings Technology, 59, No. 755, 103 (1987).
4. Gerlock, J.L.; Kucherov, A.V.; Nichols, M.E. "On the Use of UVA, HALS, Photooxidation, and Fracture Energy Measurements to Anticipate The Long-Term Weathering Performance of Clearcoat/Basecoat Automotive Paint Systems," Journal of Coatings Technology, 73, No. 918, 45 (2001).
5. ASTM G90-94 Practice for Performing Accelerated Outdoor Weathering of Nonmetallic Materials Using Concentrated Natural Sunlight. 1994 Annual Book of ASTM Standards, vol. 14.02. Philadelphia, PA: American Society for Testing and Materials, 1994.
6. Gerlock, J.L.; Smith, C.A.; Remillard, J.T. "Accelerated Weathering Tests for Automotive Paint Systems: Case For Distorted Weathering Chemistry," Proceedings of the American Chemical Society Division of Polymeric Materials: Science and Engineering, 83, Washington D.C., p. 155 (2000).
7. Bauer, D.L. "Perspectives on Weatherability Testing of Automotive Coatings," Journal of Coatings Technology, 74, No. 924, 33 (2002).
8. Luftig, J.T. Experimental Design and Industrial Statistics, Level I-IV, (1987) Luftig and Associates, Inc.: Farmington Hills, NJ.
9. Montgomery, D.C. Design and Analysis of Experiments, 3rd ed., (1991) John Wiley and Sons: New York.
10. Taguchi, G. Introduction to Quality Engineering, Designing Quality into Products and Processes, American Supplier Institute, Inc.: Dearborn, MI, 1986.
11. Taguchi, G.; Konishi, S. Orthogonal Arrays and Linear Graphs, ASI Press: Dearborn, MI, 1987.
12. Centre National D'Evaluation De Photoprotection, Analytical Study Reports: DF CP 2001-070 and JL RB 2000-407, University Blaise Pascal, Clermont, 2000-01.
This paper was presented at the 1st European Weathering Symposium EWS, Confederation of European Environmental Engineering Societies, September 25 and 26, 2003.