Accelerated Aging Test Methods for Predicting the Long Term Thermal Resistance of Closed-Cell Foam Insulation

The accurate estimation of the thermal performance of insulation products used in buildings over their expected lifetime has been a recognized challenge for over 25 years. This is because the lifetime of such products is long, thermal aging is caused by the diffusion of a multitude of gases, and the insulation product is not homogeneous. The task of developing a standard test method for predicting long term thermal performance which applies to a variety of closed-cell foam products is even more complicated as diffusion processes occur at rates that depend on the type of polymer, the cellular structure, the temperature, the gas type, and its pressure. Both classical approaches to accelerating diffusion controlled phenomena, namely aging at higher temperature and aging a thin slice, present challenges especially if a single method is desired for a variety of cellular foam products such as polyisocyanurate (polyiso) and extruded polystyrene (XPS). Though Europe has favored standard test methods based on aging at elevated temperature, slicing and scaling techniques have been the leading approach in North America. Lately, two prescriptive test methods, ASTM C 1303 in the USA and CAN/ULC-S770 in Canada have emerged in North America. Both are based on accelerating the foam aging process by slicing the foam into thin specimens. Both methods use the projected thermal conductivity at five years of age to represent the insulation s long term thermal resistance value (LTTR). The two methods have many other similarities, such as use of Fickian law for one-dimensional diffusion to calculate aging period, and use of a thin slice from the core and surface areas of the foam. However, they do vary in precisely how the long term thermal resistance is calculated. The C 1303 test method prescribes that the thermal resistance value of a stack of thin slices after aging for a calculated time is the long term thermal resistance value. The S770 test method calls for multiplication of an aging factor to an initial thermal resistance value to determine the five-year value. Though the basic frameworks of the two methods are in place, the precise parameters are still being debated and balloted. This paper compares the two methods, ASTM C 1303-07 and CAN/ULC-S770-03 for their suitability for use as standard test methods by the polyiso and XPS insulation industry for their boardstock products. Mathematical modeling and calculation algorithms developed by Huntsman and described in earlier papers are used along with published thermal aging data to evaluate how effective the two methods will be to meet the various criteria for being an industrially useful method. It looks at the impact on the bias for each of the method for the various parameters still being debated, such as slice thickness and stack composition. This study demonstrates that with the appropriate choice of test conditions, each of the two methods have potential to give low bias with polyiso boards. For XPS boards, ASTM C 1303-07 appears to be the only real choice. INTRODUCTION Heating and cooling of buildings represents a large fraction of the total energy used by mankind. With the increased cost of fuel and concerns about global warming, there is an increased focus on improving the thermal performance of the building system. Decisions about insulation usage are among the most important an architect or specifier will make relative to the operational cost and environmental impact of a building. To achieve an overall building thermal performance, architects and specifiers know the thermal resistance (R-value) needs for a particular building, component, or component system. They need the manufacturer s of building insulation products to supply them with reliable products with high, proven R-values. Among all the insulation products, only polyisocyanurate and extruded polystyrene foam boards are closed celled plastic foams that are designed to capture a gas with much lower thermal conductivity than air. This makes such plastic foam insulation the most efficient insulator among all widely used commercial products. It is the reason why it has the highest growth rate among all insulations globally (1). Due to other requirements on such cellular foam insulations, especially when used in commercial roofing applications, they are not encapsulated in air-tight barriers. This causes the gas composition in the foam to change with time as air diffuses into the foam and the insulating gases diffuse out. The diffusion of some of the gas components is rather slow, leading to gradual shift in R-value over years to even centuries, stopping only when equilibrium concentrations are reached for every gas. This presents a challenge. Certainly architects and specifiers want products that are made using the most advanced materials and process technologies while meeting all the current and anticipated environmental requirements. At the same time, they want to accurately represent the thermal performance. This makes it necessary that there be a relatively short duration, easy to implement and yet technically correct method to measure the thermal performance of closed celled plastics foams for the duration of its use in a building system. As elaborated in a 2002 publication by Huntsman, a prescriptive test method, CAN/ULC S770-00 (hence forth referred as S770) developed in 2000, is exactly such a method for the current polyisocyanurate boards used in commercial roofing (2). S770 Test Method S770 test method is based on acceleration of aging by slicing and scaling (3). It was developed under the auspices of the Underwriters Laboratories of Canada (ULC) by a task group comprising of representatives from National Research Council of Canada (NRC) and all sectors of the plastic foam industry. It advocates that thermal performance of cellular foam insulation should be defined as the time-weighted average of R-value over 15 years at room temperature at the used thickness. This is primarily because the average useful life of a typical roof on a commercial building is 15 years. Field aging data, though limited, has confirmed that laboratory aging at room temperature adequately represents aging on a roof with all the temperature/humidity variations through the years (4, 5). For all of this paper, room temperature is understood to be a temperature of 22C ± 5C (72F ± 9F) and a relative humidity of 50% ± 20% and R-value measurement is understood to be done according the ASTM C 518 or C 177 with an average test temperature of 24C ± 2C (75F ± 4F) with a temperature difference of 22C ± 2C (40F ± 4F). It has been mathematically derived that for long term typical Rvalue aging (i.e., an exponential decay of R-value over years), the time-weighted average over 15 years is equivalent to the actual R-value after approximately 5 years (6). The S770 test method uses a slicing and scaling based methodology to predict this 5 year aged R-value in weeks to months. The S770 standard and other publications (3, 6) describe the exact procedure in detail but it entails the following three steps: (a) Determination of the mean initial thermal resistance of the product, R Product, initial (b) Determination of the aging factor as the ratio between the thermal resistivity of a 6-12 mm thick slice at the specified time of aging (R Slice, aged ) to its initial value (R Slice, initial). The specified time of aging is found using the formula Time of aging in days = 5 x 365 [ Thickness of slice / Thickness of full thickness product ] 2 (1) The traditional approach to the slicing method has been to cut the thin slice from the core of the cellular board, but the S770 method requires cutting slices of the same thickness from both the surface and core. This is done to account for the long standing position that cellular insulations are not homogeneous in the thickness direction. Thus two aging factors, one for the core layer and another for the surface are determined. (c) LTTR is calculated using the larger of the two aging factors, surface or core. Thus, R Product, initial x R Slice, aged LTTR S770 = ------------------------------------------(2) R Slice, initial The S770 method states that the initial thermal resistance of the product, R Product, initial shall be determined within 7-14 days after the production date and all the thin slices must be prepared on the same day as the measurement of initial R-value of the product. In addition, the method states that initial R-value of the slices (R Slice, initial) must be measured within 2 hours of cutting the slice as the thin slices age very rapidly. There are other conditions such as if slices are stacked to measure Rvalue, the core stack cannot contain surface slices or vice-versa; during aging, both surfaces of the slices should be exposed to the ambient air; slice thickness should be uniform and so on. The test is considered invalid if the two aging factors, surface and core, are different by more than 12%. In this case the product is too heterogeneous and slicing and scaling principles do not apply to the product. A 2003 revision of the S770 method, CAN/ULC-S770-03, put an additional requirement that R Slice, initial can not be more than 12% lower than R Product, initial and that R Product, initial shall now be determined within 3-14 days after the production date. CAN/ULC-S770 is incorporated by reference in the 2001 and subsequent editions of the Canadian standards for polystyrene, polyisocyanurate and spray polyurethane foam insulation: CAN/ULC-S701, Standard for Thermal Insulation, Polystyrene, Boards and Pipe Covering: CAN/ULC-S704, Standard for Thermal Insulation, Polyurethane and Polyisocyanurate, Boards, Faced: and CAN/ULC-S705.1, Standard for Thermal Insulation Spray Applied Rigid Polyurethane Foam, Medium Density Material-Specification. In the U.S., the S770 methodology was incorporated into the 2002 and subsequent editions of ASTM C 1289, Standard Specification for Faced Rigid Cellular Polyisocyanurate Thermal Insulation Board. As evidenced from some recent publications, comparison of results from S770 test data with actual aged R-value measured on full thickness product for polyiso boards and comparison of S770 data with historical aged R-value data for XPS boards suggest that the LTTR predicted by the S770 method is higher than the actual/historical aged data, i.e., there is a positive bias (8, 9). A study conducted under the auspices of Polyisocyanurate Insulation Manufacturers Association (PIMA) suggest that for the set of polyiso boards used in the study, the average bias is approximately +6% (8). Various XPS manufacturers have reported an over-prediction of 10-25% in their product literature. The ULC Thermal Insulation Committee has recognized the limitations of the current test method and has reconvened a task group to address the issue of bias in the next revision. In the mean time, the ASTM C1303 task group has approved a new test method ASTM C 130307, Standard Test Method for Predicting Long-Term Thermal Resistance of Closed-Cell Foam Insulation which includes a prescriptive method to complement the long-standing research method (10). ASTM C 1303 Test method Though many research papers had been published on the use of slicing and scaling to predict aged R-value of specific cellular plastic foams, ASTM C1303-95 was the first standard test method for estimating the long-term thermal resistance of unfaced rigid closed-cell plastic foam. Though this test method did not mandate a specific time period, it provided a methodology to calculate time-weighted average R-value for any period. Validation of this method came from a 5 year project at Oak Ridge National Laboratory (ORNL) in Oak Ridge, TN that showed good correlation among in-place polyiso roof insulation R-values with those measured on products stored in a laboratory and those predicted from laboratory slicing and scaling (4). Surprisingly this test method was not adopted as a method to specify LTTR values in any of the closed celled foam insulation standards. Though the exact reason stated by each type of material polyiso and XPS, is somewhat different, a common complaint was that it is too complicated to perform and requires a large number of measurements. A revision of the test method published as ASTM C 1303-00 in year 2000 included some simplifications but still neither the polyiso nor the XPS material standards adopted it (11). In 2003, the ASTM C 1303 task group decided to work towards adding a prescriptive element to the test method and ASTM C 1303-07 is a result of that work (10). Similar to S770, the prescriptive part of the ASTM C 1303-07 (henceforth referred as C1303) defines LTTR as R-value at an age of five year which corresponds to the average thermal resistance over a 15-year service life (Ref 4, 5). Here, LTTR is defined simply as LTTR C1330 = R Slice, aged (3) Where R Slice, aged is the thermal resistivity of a 9 mm or higher thickness slice after aging for a time period defined by Equation (2). The method requires that slices be prepared between 7-14 days after the production date. At present C1303 requires that R Slice, aged data be collected using a stack of core slices only, a stack of surface slices only and a mixed stack of slices representing a cross-section of the product. Stacks of slices are used for thermal resistance measurements in order to minimize errors associated with radiation heat transfer phenomena between the hot and cold plates of the measuring device when separated by small specimen thicknesses. The method outlines many other requirements such as the foam portion of each surface slice be a minimum of 9 mm; slices be uniform in thickness; slicing technique be geared towards minimization of damaged layers of cells; both surfaces of slice be exposed to free air circulation during aging; there be no air gap between slices during R-value measurement of the stack; stack orientation be consistent and so on. The standard method should be consulted for detail on each of these. Recognizing that neither polyiso nor XPS are perfectly homogeneous, C1303 method sets forth a homogeneity qualification to insure that they are homogeneous enough for this test method to produce meaningful results. Aging characteristics of a stack of core thin slices is compared to a stack of surface thin slice specimens using the following formula: 2 [ (k1 / k2)Core (k1 / k2)Surface ] Aging Equivalence = 100 % { 1 ----------------------------------------} (4) [ (k1 / k2)Core + (k1 / k2)Surface ] Where k1 is the thermal conductivity of a stack after aging for (24 ± 0.5) hours per (cm of slice thickness) 2 from the moment the initial cut was made on the full thickness product to make the thin slices and k2 is the same after aging for (30 ± 1) days per (cm of slice thickness). Thus if the slice thickness is 9 mm, k1 and k2 should be measured after (24 x 0.9 = 19.44) hours and (30 x 0.9 = 24.3) days of aging respectively from the moment the initial cut was made on the full thickness product to make the thin slices. Currently, the C1303 method stipulates that if the aging equivalence is between 90% and 110%, the insulation specimen satisfies the homogeneity qualification for the use of the test. If not, accelerated aging results may not be used. The standard also has an alternate product thickness qualification which defines when test results for specimen taken from a particular product thickness shall be considered representative of other product thicknesses more than 1.3 cm (0.5 ) apart and made from identical components and having identical foam morphology. Alternate product thickness qualification is not in the scope of this paper. The research part of ASTM C 1303-07 provides a relationship between thermal conductivity, age and product thickness. The calculation methods can be used to predict the R-value at any specific point in time as well as the average R-value over a specific time period. Many of the requirements of the prescriptive procedure, such as age of the product at the start of testing, homogeneity qualification, slice thickness, stack composition, aging environment and so on are only a guideline for research portion of the ASTM C 1303-07 method. Though there appears to be willingness to adopt the C1303 test method by the polyiso and XPS industry, they are awaiting results of a ruggedness test currently underway at ORNL. The ruggedness test examines the influence of several test variables, most importantly thin slice stack composition, slice thickness, and product homogeneity on the bias with room temperature aging of full thickness product. The five year bias results will be available in the year 2011. BASIC ISSUES The above discussions and a review of the wider literature on the aging of closed cell foam suggests that there has been a tremendous effort by universities, national research laboratories, board insulation industry groups, and test standard bodies to first understand the topic of heat transfer and thermal aging and then to deduce a test method for the purposes of product evaluation, specifications or product comparison (12, 13). At the same time, it is clear that the only test available in USA and Canada at present, the S770 method, has a bias and a proposed new method is in the early stages of validation. Keep in mind, in most cases, 5-year actual aging is not an option due to continuing need to change the product composition and the cellular morphological characteristics to meet the various environment requirements (ozone depletion, green house gas, recycle content and so on) and to benefit from advances in polymer science. Other approaches, namely aging at higher temperature and numerical modeling have been considered and deemed less desirable. The primary limitation of this method has been that a specific increase in temperature does not equally change the diffusion coefficients of all the gases involved in the aging process and whether the elevated temperature could damage the cellular structure of the foam. Given that LTTR is now being defined as the R-value at a specific time, 5 years, it is likely that at least for some types of cellular foam blown with a specific blowing agent, one can find a temperature which overcomes both limitations. After all, acceleration of aging by raising temperature is favored by standard making bodies in many European countries (14). Several numerical models of the gas diffusion process through foam with location dependent diffusion coefficients have been proposed and each have the capability to accurately predict aged R-value of a given polyiso or XPS boards with essentially no bias (15-17). The difficulty with the numerical modeling is the effort required to measure all the input parameters. As an example, Distributed Parameter Continuum (DIPAC) model requires input of over 25 parameters including density, initial blowing agent fraction and polymer index for each blowing agents, effective diffusion coefficients at two temperatures for N2, O2, and each of the blowing agents; and all for core layer and surface layer (15). This is very onerous and unlike any standard test method for any material. This suggests that the two prescriptive methods, S770 and C1303, are the best hope for practical LTTR approaches to be widely used by industry. In this paper we examine the reasons behind the high bias in the S770 method for XPS boards and the smaller bias in polyiso boards and steps that can be taken to reduce both significantly. An objective for this work is to ascertain what has the potential to give a lower bias across the range of polyiso and XPS products. We do all this by using the thermal aging simulation software developed by Huntsman Polyurethanes called Agesim to predict the thermal aging curves for thin slices and full thickness products using the input parameters obtained from published literature for both XPS and polyiso boards (2). This allows us to look at the effect on bias in LTTR of various parameters such as age of the foam, slice thickness, and time since slicing for initial R-value measurements, very expeditiously compared to experimentation and in some cases in ways not possible experimentally. THERMAL AGING SIMULATIONS: AGESIM Huntsman s thermal aging simulation software, Agesim, is described in detail in an earlier publication and should be consulted for a fuller description (2). For each foam specimen, it requires conditions of aging; dimension, temperature of aging and R-value measurements. It also requires input of cell gas partial pressures and effective diffusion coefficients at the temperature of aging for each of the gases involved, namely N2 and O2 (or air), and the blowing agent(s). In order to simulate the aging behavior of products containing a densified skin layer and/or a surface density gradient having different diffusion characteristics than core foam, a one mm thick skin layer with effective diffusion coefficient, Deff, defined as Deff through skin/facer = X * Deff through core (5) X is referred to as the skin factor and is calculated iteratively by fitting measured k-factor aging curves of the laminate with those predicted by Agesim. The solid conduction and the radiative heat transfer components of thermal conductivity are assumed constant in Agesim and all aging is attributed to changes in gas phase conduction due to diffusion of gases. Huntsman has been using the Agesim software for a number of years to accurately predict R-value aging curves for polyiso boards. We have built an extensive database of diffusion coefficients, skin factors, and initial cell gas partial pressures through various foam laminates. As an example, Figure 1 shows the modeled and measured thermal resistance for a laminate board produced in 1998. This is a reproduction of Figure 6 from Reference 2 with additional measured data gathered since 2002. The measured diffusion coefficients and skin factors are listed in Table 3 of reference 2 and re-listed in the Table in section calculations on polyiso boards here. 5.0 5.5 6.0 6.5 7.0 7.5 0 5 10 15 Aging time at RT (years) R -v al ue ( ft 2 .h r. o F/ B tu .in )