By Colin Murphy, RRC, Robert Mills and Steve Bunn INTRODUCTION |- IHE FOLLOWING ARTICLE IS THE THIRD IN A SERIES OF THREE ADDRESSING SIMULATED HAIL retesting and hail (impact) damage to shingles. The first (published in the January issue of Interface) provided an overview of the history of general shingle testing. The second (published in the February issue of Interface) pro- _ vided an overview of the history of shingle hail resistance testing and the development of repeatable test proce¬ dures. This article summarizes the test data and conclusions generated from a recently completed, five-year test program conducted by Exterior Research & Design, LLC. A comprehensive testing program to simulate hail damage to shingles is not new. Test protocols were developed over 30 years ago, indicating there are system design, product manufacturing and installation issues to reduce hail-induced damage to shingles. Unfortunately, past testing and test method development have not led to a method recognized by the roofing and insurance industries or building code authorities. The primary burden of repair and replacement of hail dam¬ aged shingles has fallen to the insurance carriers, and there¬ fore, ultimately to those who pay the premiums. The roofing industry—both manufacturers and installers—has benefitted from these natural events. The roofing industry has had no financial incentive to develop products and promote build¬ ing standards that would reduce the effects of hail damage. Insurance carriers have recently responded to this major loss category by cooperating with and encouraging testing agen¬ cies to develop a recognized test standard for impact dam¬ age. In addition to the efforts of the insurance carriers, state government agencies have been active in promoting legisla¬ tion or executive order directives to insurance carriers to offer premium discounts to policy holders who install shin¬ gle roof systems proven to be more resistant to hail damage. Some carriers are not waiting for government executive orders or legislation, and instead are offering discounts to promote the use of more impact-resistant installations. Discounts in some states range from 17 to 21% of a home¬ owner’s premium, providing the shingle and its application meet the requirements of the insurance carrier. For example, some discounts are promised by removal of the original roof system in re-roof applications. In addition, some carriers are developing their own test facilities to gain first-hand knowl¬ edge of shingle performance. The testing procedures for determining the effects of impact to shingles have been in development since the 1960s. Sidney H. Greenfeld, working under the sponsorship of the U.S. Department of Commerce, was an early researcher in this field of research, publishing findings of a simulated hail laboratory study. The study examined impact resistance of organic and fiberglass-reinforced shingles, as well as various built-up roof assemblies. The study, titled “Hail Resistance of Roofing Products,” 1 was published in 1969. The Greenfeld test method included the production of artificial hailstones from ice balls, which were launched at roofing specimens using a compressed air gun. Greenfeld conducted the testing over various substrates, including 3/8″ plywood, 1/2″ plywood and 1″ x 6″ tongue and groove deck¬ ing. Substrates were further varied by altering the underlay – ment layer. Greenfeld published a series of conclusions based upon his test program. He concluded that: ▼ There are three distinct areas where shingles are vulnera¬ ble to simulated hail damage. These occur at overlaps, tabs and edges. ▼ Impact resistance increases with substrate rigidity and hardness. ▼ There is increased impact resistance with increased shin¬ gle weight. ▼ Fiberglass mats offer increased impact resistance to com¬ parable organic mats. ▼ Shingle aging decreases impact resistance. The American Society for Testing and Materials (ASTM) published Test Method D 3746 2 in 1978. This test protocol utilizes steel missiles, which are fired at test specimens to simulate hailstone impact. In 1986, the Factory Mutual Research Corporation 4 • Interface July 1998 Modified FM 4 4 70 testing procedure. (FMRC) amended its Test Method 4470 3 to incorporate a requirement for hail testing. The protocol called for specifi¬ cally-sized and weighted steel balls to be dropped from established heights. In 1992, the resultant impact energies were given two classifications: “Moderate Hail” classification when an impact energy of 8 ft-lbf (1 1 Joules) is utilized, and “Severe Hail” when an impact energy of 14 ft-lbf (19 Joules) is utilized. While the test protocol does not address all vari¬ ables that could affect hail performance, the method is easily repeatable in the study of hail resistance. The 4470 test pro¬ cedure was developed primarily for low-slope roofing prod¬ ucts and does not specifically address testing of shingle assemblies. The Swiss Federal Laboratories for Materials Testing and Research (EMPA) published an article titled, ‘The Hail Resistance of Plastic Components of the Building Shell,” 4 in 1988. The paper included a test procedure for determining impact resistance utilizing plastic balls launched from a com¬ pressed air gun at different velocities. The plastic balls are fired at close range to reduce variables, which could alter the impact energy. In 1993, Haag Engineering developed a further test proto¬ col utilizing frozen ice balls to simulate hailstones fired from a compressed air gun, as described in an article titled, “Hail and Composition Shingles.”5 The author, Scott J. Morrison, P.E., further refined the analysis of impact damage by desat¬ urating the test specimens for observation of the shingle mat, rather than simply examining the shingle surface under magnification. The Haag protocol introduced a newly defined mode of shingle failure, rupture of the shingle mat. Most recently, in 1996, Underwriters Laboratories, Inc. published Test Standard UL 22 18 6, an impact resistance test utilizing different sized steel balls to simulate varying levels of impact. This protocol utilizes steel balls dropped from various heights, developing different categories or ratings of impact resistance. Most notably, some members of the insur¬ ance industry and government bodies have begun to classify roofing materials for impact resistance by adopting the UL protocol and criteria. Test Standard 2218 appears to have been adopted as the insurance standard providing the shin¬ gle producers with a clear performance requirement. The test protocol creates scaled performance for production of product to meet various geographic needs. Insurance carriers in Texas, Kansas, and Oklahoma are offering insurance pre¬ mium discounts to policy holders who install roofing materi¬ als based upon the results of UL 2218 testing and applica¬ tion of those materials in compliance with a specified instal¬ lation guideline. With the various impact methods utilized, there is a lively discussion as to whether simulated impact test methods pro¬ duce repeatable results and which method(s) most accurately simulate actual hail impact. Researchers who have used both ice balls and/or steel or plastic missiles have offered conflict¬ ing conclusions regarding repeatability. Exterior Research & Design, LLC (ERD) Test Program In 1994, Exterior Research & Design (ERD) began a test program to determine impact hail resistance of shingles based upon changes in specimen construction, configuration, age, and condition. The test program was comprised of two facets, field analysis of shingle products impacted by hail in one major hail event, and laboratory testing of a series of shingle products manufactured in the United States. Field Analysis: On September 7, 1991, a major hailstorm hit Calgary, Alberta, resulting in extensive damage to roofing. The war¬ ranty cards from one shingle manufacturer were reviewed to locate specific projects that were potentially within the path of severe hail. Field staff first identified the materials, con¬ structions and locations of the projects provided from the warranty cards. File warranty data and field installations did not necessarily match, resulting in the need for field inspec¬ tion for confirmation. The team then tracked the path of the storm by reviewing weather service data, as well as inter¬ viewing building owners and their neighbors to determine the intensity and path of the storm. Confirmation of this data was obtained by proof of actual damage such as pho¬ tographs of car or building damage or photographs of actual hail. From warranty cards and replacements of thirty-nine roofs, sixteen shingle installations were identified as being within, July 1998 Interface • 5 24 Shingle vs Surface Damage New Shingles 4 Year Old Shingles SBS modified shingles were more reesistant to surface damage. SBS modified shingles were less resistant to impact damage. Surface Damage vs Mat Damage Oxldlzed/Modlfled vs Damage Surface damage is less indicative of mat damage for modified shingles. or directly adjacent to, the path of the most severe hail. Five of the sixteen projects were roofed with 20-year strip shingles, manufac¬ tured from fiberglass mats and oxidized asphalt. Of these five projects, one could not be con¬ firmed to be within the path of the most severe hail. This roof, along with a large number of the thirty-nine clearly outside the area of the most severe hail, were replaced by an insurance carrier as a total loss due to the hail event. The remaining four 20-year strip shingle projects were confirmed to be within the path of the most severe hail, and were replaced as the result of an insurance loss. In all cases, the new shingle applications were re-roofs over the damaged shingle assemblies. Nine of the sixteen projects were roofed with Styrene Butadiene Styrene (SBS) modified asphalt shingles manufactured with 2-lb. fiber¬ glass reinforcement. Of these nine projects, two could not be confirmed to be within the direct path of the storm. Of the remaining seven projects, only one was replaced by an insurance carrier as a direct result of the hail event. The one replaced roof was installed over an existing 20-year strip shingle roof installed over an underlayment and a plywood substrate. The remaining six projects remained in service after the hailstorm. Further investigation revealed that these six projects had nominal 1″ tongue-and-groove deck substrate and slope of 4″: 12″ or greater. Laboratory Testing: The laboratory testing included both new shingle samples and four-year-old field-weath¬ ered shingle samples. The field samples were obtained from sites within the Seattle Metropolitan Area. The Seattle area was cho¬ sen due to the lack of hailstorms within the locality. The shingle samples were taken from four-year-old projects because the shingles of this age were easily identified by warranty cards on file. Older projects had insufficient documentation to confirm the precise installa¬ tion dates. Each project was sampled in a con¬ sistent manner. Samples were taken from the east and west elevations, visually inspected to be free from surface damage, and were carefully removed to prevent damage during extraction. A total of nine brand name shingles were sam¬ pled. In all cases, the samples were identified by warranty cards, and an interview was held with the roofing installer as a cross check of the warranty card data. The same nine brand name shingles were purchased new from local distributors for com¬ parison with weathered product. The nine dif- 6 • Interface July 1998 ferent shingles included: ▼ Three strip shingles, SBS modified asphalt, mat weight 2 lb. ; ▼ One strip shingle, oxidized asphalt, mat weight 1.54 lb. ; ▼ One strip shingle, oxidized asphalt, mat weight 1.93 lb. ; ▼ One dimensional shingle, oxidized asphalt, mat weight 1.61 lb. ; ▼ One dimensional shingle, oxidized asphalt, mat weight 1.58 lb. ; ▼ One dimensional shingle, oxidized asphalt, mat weight 1.86 lb. ; and ▼ One dimensional shingle, SBS modified asphalt, mat weight 2 lb. The FMRC Test Standard 4470 was chosen for the labora¬ tory testing for consistency in the impact energy created by the steel ball dropped from the same height. Moreover, when testing began in 1994, the UL Test Method 22 18 had not yet been published. Specifically, the severe hail require¬ ment from 4470 was chosen because the impact energy of 14 ft-lbf (19 Joules) equated to hailstones sized 1-1/2″ to 2” in diameter, a size known to consistently damage shingle roof systems. Tests were conducted to examine a number of vari¬ ables: ▼ New shingles and four-year-old “aged” shingles,- ▼ Strip and dimensional shingles (laminated versus non¬ laminated),■ ▼ Oxidized asphalt and SBS modified asphalt shingles to compare the different types of impregnating asphalt,- ▼ Substrate slope comparing 2″: 12″ to “1”:12″ (additional testing, after completion of the primary test program, also included slopes of 6″: 12″ and 9″: 12″); ▼ Lise of 15# underlayment felt comparing constructions with and without the use of an underlayment; ▼ Decks constructed from 15/32″ plywood and 1″ nominal tongue and groove, both attached at 24″ centers,- ▼ Simulated impact locations (shingle tabs versus over¬ laps) with impact at center span,- ▼ Dimensional vs. non-dimensional shingles and damage within top and bottom plies of the dimensional shin¬ gles. All test specimens, new and aged, were “conditioned” in accordance with the protocol published in ASTM D 3161 7 after installation to the test decks. The impact test consisted of dropping a 2″ diameter steel ball through a plastic tube from a height of 17’ 9-1/2″. Sixteen tests were performed for each variable under examination. The simulated hail impact locations were marked on the test specimen. Upon completion of the testing, the impacted area was removed and examined through 5x magnification. Three ananalysts qualitatively evaluated the damage on a threepoint scale of 0 to 2. A recording of “0” indicated no dam¬ age, “1” indicated moderate damage and “2” indicated a sur¬ face fracture or other pre-defined severe damage. During the analysis, there was complete agreement between the three parties for approximately 85% of the samples. When there were differences in opinion, the group of three met to arrive at a mutual consensus on the numeric value. After the surface damage had been numerically recorded, the area of impact was desaturated to expose the fiberglass mat. The damage to the fiberglass mat was qualitatively evaluated in the same manner. For the mat damage, recordings of “0” indicated no damage, “1” indicated moderate damage and “2” indicated a tearing or rupture of the mat. All of the relevant test data were input into a spreadsheet. The data were then imported into a statistical analysis pro¬ gram to determine the correlation of the different variables to the observed damage. The statistical analysis yielded the following results: ▼ Surface damage and mat damage are nearly proportional for oxidized asphalt shingles, with slightly greater dam¬ age to the mat than visually observed on the surface,- ▼ Surface damage and mat damage for SBS modified asphalt shingles are in general agreement,- however, the visual observation of mat damage is greater than visual observations of surface damage,- ▼ SBS modified asphalt shingles exhibited less damage to both the surface and the mat when compared to all oxi¬ dized asphalt shingles tested within the series,- ▼ Weathered shingles always exhibited greater surface and mat damage when compared with new samples of the same product,- ▼ The top mat layer of dimensional shingles exhibited significantly greater mat damage than the bottom layer where two layers are bonded together,- ▼ Simulated hail impact at shingle overlaps recorded higher level damage than impact over shingle tabs,- ▼ Shingles installed over 15/32″ plywood recorded mar¬ ginally greater damage than those installed over 1″ tongue and groove decking with identical support and underlayment. ▼ Shingles installed over 15 lb. felt recorded marginally greater damage than those installed without an under¬ layment. In further testing, shingles installed over a double layer of 30 lb. felt recorded significantly greater damage than those installed without underlayment. ▼ Shingles installed at a slope of 2”: 12″ recorded margin¬ ally greater damage than those installed at a slope of 4″: 12″; however, preliminary testing at slopes of 6″: 12″ and “9”: 12″ indicated a greater reduction in both surface and mat damage,- ▼ Mat damage to lighter weight shingles (210-249 lb/100 ft2) was marginally greater than heavier weight shingles (250-275 lb/100 ft 2). ▼ There was no apparent correlation between shingle weight and surface damage. ▼ There was no apparent correlation between mat weight and mat or surface damage. Almost without exception, the ERD test results agree with the conclusions published by Sidney H. Greenfeld nearly 30 years ago. The ERD testing supports the conclusions that particular zones of a shingle are more sensitive to impact than others, shingles installed over harder or more rigid sub¬ strates show increased impact resistance, and weathering decreases impact resistance. The only exception to the ERD and Greenfeld correlation was the relation that increased mat weight correlated to increased impact resistance. Shing- July 1998 Interface • 7 Severe bail: surface fracture damage. Ies tested by Creenfeld were primarily organic shingles, not fiberglass. This is likely a factor in the lack of correlation. The manufacture of shingles has also changed, resulting in many more variables than existed thirty years ago. These factors may also play a role in this lack of correlation. Additional conclusions drawn from the ERD test program are as follows: ▼ While the surface damage to oxidized asphalt shingles is a good indicator of mat damage, determination of mat damage in SBS modified asphalt shingles solely from visual observation of surface damage is difficult. As a result, to properly document damage to SBS modified asphalt shingles, the shingle must be desaturated to bet¬ ter identify the full extent of the damage. This issue becomes relevant as more SBS modified asphalt shingles enter the market and more insurance authorities turn to the LIL 2218 procedure as a basis of premium reduction. The new UL 2218 test procedure requires a visual exam¬ ination of the shingle surface,- however, it does not require shingle desaturation for examination of the mat. The UL 2218 may not provide the most accurate mea¬ sure of impact damage when testing SBS modified asphalt shingles. ▼ The tested SBS modified asphalt shingles exhibited greater impact resistance than the tested oxidized asphalt shingles. SBS modified asphalt shingles better conform to the underlying substrate than oxidized asphalt shingles, thereby reducing bridging and con¬ tributing to their increased performance, especially at overlaps. ▼ The top plies of dimensional shingles exhibited greater damage than single and bottom layers even on harder substrate materials On softer substrates, particularly in re-roofs over existing shingles, damage to the bottom Severe hail: surface fracture damage. layer can be significant. ▼ There is a minimal improvement in impact resistance when installing shingles over slope increases from 2″: 12″ to 4″: 12″. However, preliminary testing indicates signifi¬ cant improvement at slopes of 6″: 12″ and greater. ▼ Regardless of shingle weight, type, mat weight and impact location, re-roofing over existing shingle roofs decreases the impact resistance of the new shingle roof assembly. The primary mode of shingle failure caused by impact is a failure of the fiberglass mat in tension. When impact occurs, the resultant deformation of the shingle stretches (elongates) the fibers of the mat. When the ultimate elongation of the fibers is exceeded, the mat ruptures. Any factor that increas¬ es the magnitude of shingle deformation also contributes to an increased fiber elongation and greater potential for mat rupture. These factors include soft substrates, flexible sub¬ strates, soft underlayments, multiple shingle plies, and tab and overlaps. As of this date, not all variables have been sufficiently addressed. Additional testing is required to examine impact resistance when shingles are greater than four years old, installed at slopes of 6″: 12″ and greater, and installed over other substrates. Preliminary testing in the second round has already indicated reduced impact resistance of eight-year-old shingles when compared to the four-year-old samples. Even though more testing is required, the first round of testing has provided a number of considerations for increasing the impact resistance of shingles: ▼ Substrates should be well supported, hard and rigid; ▼ Thin, hard underlayment layers should be considered. Thin, durable “permeable” membranes are available on the market, reducing underlayment thickness to 10-20 mils, as compared to the common 27 mils for 15# and 58 mils for 30#,- ▼ Shingles should be installed with as little bridging as possible,- 8 • Interface July 1998 ▼ Shingles should lie as flat against the substrate as possi¬ ble,- and ▼ Existing roof systems should be removed in re-roof applications to enhance impact resistance of the new shingle installation. Methods for reducing hail-related damage have been known for the last 30 years. The roofing industry, however, has been slow to develop new building materials and alter methods of application. Slow development has been fueled by the classification of hail damage as an insurable event, resulting in increased sales of product and installation paid for by insurance premium dollars. This has provided little incentive to invest in the development of a more hail resis¬ tant shingle roof assembly. However, until recently, there has not been a published test standard recognized by all par¬ ties. State officials and insurance companies are now foster¬ ing change by offering insurance premium discounts to homeowners who install roofs with more impact-resistant products over specified substrates. As a result of this inter¬ vention and the adoption of a new standard, shingle manu¬ facturers are beginning to develop more impact-resistant products. Further testing and study will help create specific guidelines for building products and the construction of more hail-resistant roof assemblies, which, in turn, will help divert significant dollars expended on hail-related reroofing to more pressing issues, relieve overburdened landfills and promote roofing manufacturers to increase the structural life of roof installations. iS 1. Greenfeld, S., “Hail Resistance of Roofing Products,” U.S. Department of Commerce: National Building Standards, 1969. 2. ASTM D3746-85, “Standard Test Method for Impact Resistance of Bituminous Roofing Systems,” ASTM Standards in Building Codes. Vol. 4, 1996. 3. FMRC 4470 “Class I Roof Covers,” Factory Mutual Research Approval Standard, 1986. 4. Flueler, P., Rupp. F., “The Hail Resistance of Plastic Components of the Building Shell,” EMPA Research and Development Working Report No. m/l, 1988. 5. Morrison, S.,”Hail and Composition Shingles,” Haag Engineering Company, 1993. 6. UL 2218, “Impact Resistance of Prepared Roof Covering Materials,” Underwriters Laboratories Inc., 1996. 7. ASTM D3 161-93, “Standard Test Method for Wind- Resistance of Asphalt Shingles (Fan-Induced Method),” ASTM Standards in Building Codes. Vol. 3, 1996. About The Authors Colin Murphy, RRO, RRC, founded Trinity Group Fastening Systems, a company that developed fastening systems for use in roof¬ ing assemblies, in 1981. In 1986, he established Trinity Engineering, focusing primarily on forensic analysis of roof systems, materials analysts, laboratory testing and long-term analysis of in-place roof systems life finn is based in Seattle, IVA Colin joined RC1 in 1986 and became an RRC in 1993 He is cur¬ rently the Director of Region VIE In 1996. he ums honored with the Richard Horowitz Award for excellence in technical writing for Interface. This spring, he was granted the prestigious Herbert Busching Jr. Award by RCI for “significant contribution to the general better¬ ment of the roof consulting industry.” Robert MillsjonirJ Exterior Research & Design, LLC, Trinity Engineering, in 1993 after obtaining a B.S. in Aerospace Engineering from Arizona State University. He has since become a senior engineer with the firm, focusing his knowledge of aerodynamics, material properties and testing and engineering principles on the field of exterior envelope design and analysis Robert has provided his knowledge of roofing design and testing to RCI as an instructor in its Advanced Roof Consulting courses Steve Bunn joined Exterior Research & Design. LLC, Trinity Engineering, m 1995. He earned a B.S. in Aeronautical and Aerospace Engineering from the University of Washington Steve is an engineer with the firm, concentrating on forensic analysis and exterior envelope design and analysis. Region I Region III Region V Region VII Mike DeFrancesco, RRC Dennis McNeil, RRC. RRO, CCS Daniel J. Neuhaus, RRC Colin Murphy, RRC Langhorne, Pennsylvania Homewood, Illinois Denver, Colorado Seattle, Washington (215)757-1450 (708)799-5599 (305) 542-5902 (206) 467-0054 Region II Region IV Region VI Region VIII C. Allen Kidd, RRC, EIT Paul Ridley James Magowan, RRC, FRCI Albert Duwyn, RRC Charlotte, North Carolina Dallas, Texas Lodi, California Mississauga, Ontario (704)555-8285 (214)959-4900 (209)555-4595 (905) 607-7244 July 1998 Interface • 9
