Investigation of the role of fire retardant treatment in the failure of wooden trusses

Size: px
Start display at page:

Download "Investigation of the role of fire retardant treatment in the failure of wooden trusses"

Transcription

1 Investigation of the role of fire retardant treatment in the failure of wooden trusses Ronald W. Anthony 1 and Michael J. Drerup 2 Abstract An investigation of failed metal plate connected wood trusses provided an opportunity to study the long-term effects of fire-retardant treatment (FRT) on dimension lumber. Specimens of lumber that were originally treated with mono-ammonium phosphate were removed from the roof framing of a multi-unit residential building in the state of New Jersey in the United States after nearly 20 years of service. Approximately 170 samples were removed from the building, yielding a sufficient number of test specimens for a statistically robust data set. The samples were prepared and mechanically tested in bending and tension, and the results were compared to previously published research, code provisions, and strength values for new, untreated lumber. Testing was performed on both full-size lumber and small clear specimens in accordance with applicable ASTM standards. The results of the laboratory analysis indicate that the FRT wood removed from building had approximately half the expected strength of untreated wood of the same species. These laboratory results, when compared to previously published data for less aged FRT lumber, suggest that the strength loss is progressive over time rather than a simple reduction at the time of manufacture. Keywords metal plate connected wood trusses, fire retardant treatment, strength testing 1. BACKGROUND AND INITIAL STRUCTURAL INVESTIGATION After a strong wind and rain storm in the summer of 2006, a residence hall in the state of New Jersey in the United States (Figure 1) was inspected by building staff for damage. During the walkthrough, separations between ceiling tiles in the dining room of the building were observed and the ceiling appeared to be sagging. Structural engineers were requested to conduct more detailed investigations, which revealed that the wood roof trusses above the dining room had been severely compromised by a large number of failed members. The fracture surfaces typically were brash with relatively smooth surfaces and very little splintering. Wood that is in good condition typically exhibits irregular, splintered fractures when it fails in tension or bending; brash fractures indicate a breakdown of the wood fibers that give the wood its strength. Markings on the truss components confirmed that they had received a proprietary fireretardant treatment (FRT) that used mono-ammonium phosphate. 1 Ronald W. Anthony, Anthony & Associates, United States, woodguy@anthony-associates.com 2 Michael J. Drerup, Exponent, Inc., United States, mdrerup@exponent.com

2 Figure 1 - Partial exterior view of building. Position of trusses beneath the roof is visible due to the snow melting pattern. Observations and measurements also confirmed that the dining room trusses exhibited excessive downward deflections, or sag, and that the masonry walls supporting the trusses were tilting outward, consistent with the outward force that would be exerted by a compromised truss (Figure 2). Based on the initial structural investigation, shoring was installed to stabilize the roof. Figure 2 - Oblique view of dining room roof. Deflection is somewhat visible due to variations in the light reflection pattern from the roof shingles. Structural analysis of the wood roof trusses above the dining room (Figure 3) was performed based on field measurements and a review of available drawings, which determined that the roof framing was designed and constructed in accordance with codes applicable at the time of its construction in the late 1980s. Based on this analysis, the roof framing should have been able to support normal loads that occur over the life of the building, notably snow loads, which can be significant. The analysis was repeated, taking into account the failed members documented during the investigation. This second analysis determined that the compromised trusses could no longer support code-required loading and that they may be in a state of imminent collapse. The analysis was consistent with the observed member fractures, downward deflection of the trusses, and outward tilting of the masonry support walls.

3 Figure 3 - Dining area truss geometry. The ridge vertical tension members of each truss (labeled) exhibited the most failures. Other portions of the building, notably residential wings A and B (Figure 4), also used FRT wood trusses and likely experienced similar service conditions over the life of the building. Detailed inspection of representative portions of these trusses was performed revealing numerous failed truss members, but relatively fewer than in the dining room roof. Approximately 80 percent of the dining room trusses exhibited one or more fractured members. Approximately 60 percent of the documented trusses in the residential wings exhibited at least one fractured member. Also, there were no indications of global structural instability in the residential wings, such as ceiling deflection and tilting support walls. Figure 4 - Aerial view of building, shaded to indicate different structural zones. The residential halls (A and B) and the dining room roofs were framed with metal plate connected trusses. The C-Wing roof was framed with steel trusses, and was not part of the investigation. The most severe distress was observed in the dining room trusses.

4 The difference in performance between the residential wings and the dining room was attributed to a more robust truss design and support structure, which was confirmed through structural analysis. Although all trusses were adequately designed based on applicable requirements, the truss designs in the residential wings had greater load path redundancy and a higher factor of safety, allowing the trusses to better accommodate material strength loss without structural failure. Based on the brash nature of the fracture surfaces and the analytical findings that the trusses were adequately designed, the second phase of the investigation focused on evaluating strength loss of the wood used in the trusses. 2. WOOD MATERIALS INVESTIGATION AND TESTING Chemicals used for FRT in the United States in the late 1980s through early 1990s are known to degrade wood in the environment of elevated temperature and humidity, lowering the wood s strength (LeVan and Winandy 1990). Such conditions are commonplace in attic spaces during summer months in the eastern United States. Mono-ammonium phosphate is a chemical treatment known to degrade both plywood and solid lumber resulting in a loss of strength over time. A characteristic of wood treated with mono-ammonium phosphate is that the degradation occurs as darkening of the wood and breakdown of the wood fibers over time. In some cases, the degradation is not visually apparent. A variety of failure types were observed in various locations, including at truss plates, in the top chord of trusses, in truss web members (Figure 5), and at a pullout of wood from truss plate connectors (Figure 6). As noted, the damaged truss members typically had brash fracture surfaces. A brash fracture surface is generally smooth with little or no splintering and is indicative of wood that has become embrittled. Embrittled wood has less ductility making it more sensitive and prone to failure at discontinuities (stress risers) such as knots or connections. The original roof framing documents specified FRT wood. Since it is well established in the literature that FRT can cause wood to lose strength over time, which can result in the failure of wood members under even benign loading, loss of wood strength was considered a primary factor in the failure of the truss members. To determine whether substantial strength loss had occurred, approximately 170 representative samples of the FRT lumber were collected from the residential wing trusses for laboratory analysis. Figure 5 - Brash tension failures in vertical truss components above the dining room.

5 Figure 6 - Pullout of wood from truss plate connector above the dining room. The type of failure gives clues to the condition of the wood, as indicated by brash or splintering failures in the wood or failure of the truss plates to hold the wood in place. Non-degraded wood typically exhibits a splintering failure with a jagged, uneven surface. Brash failures exhibit more of an abrupt, smooth surface because of the breakdown of the wood fibers that give the wood its strength. Most of the failure surfaces were brash. The extent of failure observed throughout the building was indicative of a systemic problem. Built in the late 1980s, the roofs of the building had been subjected to wind and snow loads for approximately twenty years, but seldom does the frequency of failures observed occur without one or more other factors contributing to the failures. It is these factors that were of interest in this investigation. The wood species indicated by the grade stamps on wood samples taken throughout the building was southern yellow pine (Pinus spp.). The grade was typically No. 2 or No. 3, although some samples were stamped No. 1. A few were also marked DNS for dense. Some lumber had no grade stamp. Much research has been conducted on the effects of FRT on wood in service. It is not the intent of this paper to provide a technical discussion of the mechanism of degradation; it is, however, worth noting the variables that affect the rate of degradation, which can vary from building to building. The factors include: Wood species Type of product (lumber versus plywood) Lumber dimensions Type of chemical treatment used The temperature at which the wood product was dried after treatment How the wood is loaded (tension, compression, bending, etc.) In-service temperature In-service relative humidity Figure 7 shows the fresh cut surface of a wood truss member removed from the dining room. The dark color in the cross section is due to the effect of the fire-retardant chemical. From this cut surface, there is no indication of any material weakness. The other end of the lumber, however, which consisted of the fracture, shows a brash surface (Figure 8). A fracture surface from non-degraded wood would have more of a splintering failure surface. Brash failures, such as this, are indicative of weak lumber, regardless of how the piece is loaded.

6 Figure 7 - Fresh-cut surface of sample T-11 from the dining room showing penetration of the FRT chemical throughout the cross section. Figure 8 - Fractured surface of sample T-11 from the dining room. A summary of the effects of FRT on wood strength is provided in a study published by LeVan and Winandy (1990). They reported the findings of several research and testing programs on FRT material. It is difficult to compare most test results or, in this case, extend them directly to the conditions at this building due to the lack of a uniform test protocol during the testing period in the 1980s and 1990s. The comparison is further complicated when the composition or process used for proprietary treatments, such as mono-ammonium phosphate, are not disclosed. Nonetheless, published values of strength loss for these two treatments support the observations of the field investigation of the treated wood components at the facility. Strength losses reported by LeVan and Winandy for wood tested in tension parallel to grain that were treated with mono-ammonium phosphate ranged from 9 to 25 percent. The duration of the test, temperature and relative humidity conditions, and whether the testing applied to both plywood and lumber are not provided in their report. Similarly, for compression parallel to grain, the range of strength change is from a decrease of five percent to an increase of two percent. For bending, the strength loss ranged from 5 to 19 percent. To gain a better understanding of the actual strength loss at the residence hall, mechanical testing was conducted on material removed from the building.

7 3. LABORATORY INVESTIGATION Based on the failures observed at the facility and what is known about the type of treatment used, it is reasonable to conclude that the lumber had lost strength since installation. While some attempts have been made to quantify the loss of strength due to these treatments, such as the study by Lyon et al. (1988), there are a number of factors that preclude making an accurate estimate of the remaining strength of the lumber at the building. Most estimates would be based on research that was limited to a few years of exposure rather than the nearly 20 years of service of products installed at the building. One hundred seventy samples approximately 76 inches (193 cm) long were cut from 2-inch by 4-inch (38 x 89 mm) members of trusses in the residential wings. One hundred fifty seven samples were shipped to the Civil Engineering Department at Colorado State University in Fort Collins, Colorado for testing. The samples were separated into two groups to conduct bending and tension tests. Bending and tension tests were selected so that comparisons could be made to the findings of Lyon et al. and to gain an understanding of the strength of the 2 x 4 lumber used in the trusses at the building. Lyon et al. conducted bending and tension tests on untreated specimens of small southern yellow pine and specimens treated with mono-ammonium phosphate. Their bending tests were conducted following the procedures of ASTM D 143, and their tension tests were conducted following the procedures of ASTM D Prior to conducting tests similar to Lyon et al., full-scale bending and tension tests were conducted on the lumber removed from the building. The full-scale bending tests were conducted in accordance with ASTM D 198. Figure 9 shows the failed surface of a full-scale bending specimen. Note the brash failure surface similar to that shown in Figure 8 of a failed truss member observed during the field investigation at the building. Figure 9 - Brash failure surface of a full-scale laboratory bending test. After conducting the full-scale bending tests, small specimens were cut from an undamaged area of the full-scale 2 x 4 lumber. The small bending specimens were tested following the procedures used by Lyon et al. using the test setup shown in Figure 10. Table 1 provides the summary of the bending test results for the material. The mean bending strength (modulus of rupture [MOR]) for the small bending specimens was 59 percent of the mean value recorded by Lyon et al. for new FRT southern yellow pine. The mean bending strength for the small bending specimens was 48 percent of the untreated southern yellow pine strength reported by Lyon et al. Clearly, the FRT lumber in the building had lower average strength than comparable untreated southern yellow pine specimens.

8 Figure 10 - Test setup for small bending test, based on ASTM D 143. Table 1 - Summary data for bending tests. Test type Full-scale bending Small bending Test Standard ASTM D198 ASTM D143 Number of Specimens Mean MOR (psi) Minimum MOR (psi) Full-scale tension tests were conducted in accordance with ASTM D 198. Figure 11 shows the failure of a full-scale tension specimen. Note the similarity of the fracture surface to that shown in Figure 5 of a failed truss member observed during the site investigation at the building. Figure 11 - Failure of a full-scale tension specimen. Note the brash characteristic of the fracture surface. Although the majority of the full-scale tension specimens failed between the grips (Figure 11), some failed in the grips that were attached to the lumber to apply the tensile load (Figure 12). These failures

9 were similar to connection failures observed during the site investigation at the building (Figure 6). Although failure in the grips can occur in tension specimens not treated with FRT, the characteristics of the failures are similar to the in-situ failures observed in the building and are indicative of the effect of FRT on lumber strength. Note the pull-out of the lumber from the toothed-plate connector in Figure 6. The pull-out is due to shear failure between the growth rings of the lumber. Also note the similarity of the separation at the growth rings to that shown in the grip failure from a tension test shown in Figure 12. Figure 12 - Grip failure of a full-scale tension test observed from the specimen end showing shear failure between the growth rings. The maple wood blocks were glued and fastened to the FRT lumber to provide a bearing surface for applying the tensile load. After conducting the full-scale tension tests, small specimens were cut from an undamaged area of the full-scale 2 x 4 lumber. The small tension specimens were tested following the procedures used by Lyon et al. Table 2 provides a summary of the tension test results for the material. The mean tensile strength for the small tension specimens was 50 percent of the mean value recorded by Lyon et al. for newly treated southern yellow pine. The mean tensile strength for the small tension specimens was 41 percent of the untreated southern yellow pine strength reported by Lyon et al. Clearly, the FRT lumber in the building had less tensile strength than untreated southern yellow pine specimens. Table 2 - Summary data for tension tests. Test type Full-scale tension Small tension Test Standard ASTM D198 ASTM D3500 Number of Specimens Mean MOR (psi) Minimum MOR (psi) Mechanical tests showed that the lumber removed from the roof trusses at the building exhibited lower mean strength than both untreated southern yellow pine, and specimens of similar size that had been treated with mono-ammonium phosphate and tested by Lyon, et al. A strength loss by itself, however, may not render the lumber defective. It is the magnitude of the strength loss that dictates whether the lumber is able to carry to loads required by the applicable building code. To assess whether the 2 x 4 FRT lumber used in the trusses was defective, the strength distribution was compared to the allowable design values specified in the National Design Specification for Wood

10 Construction (NFPA 1986). Figure 13 shows allowable design values for various grades after adjustment for being treated with a fire-retardant chemical. Note that a few of the full-scale tension specimens failed at loads below the design value, even for the lowest grade found (No. 3). These specimens were weaker than the design value of any structural grade for 2 x 4 southern pine lumber and would not comply with any building code that referenced the 1986 National Design Specification for Wood Construction. Tensile Strength for Full-Scale Tension Specimens Tensile Strength (psi) Test Specimens No.1 DNS Design Value No.1 Design Value No.2 Design Value No.3 Design Value Percent of Specimens Figure 13 - Tensile strength values for full-scale test specimens (in psi). Lines indicate the design values for No. 1 DNS (Dense), No. 1, No. 2, and No. 3 southern pine 2 x 4 lumber adjusted for fire-retardant treatment. The mechanical test data are somewhat biased in favor of stronger pieces because a number of pieces of lumber were observed during the field investigation that had already failed and could not be removed for mechanical testing. The failed (i.e., weak) pieces would have been below the allowable design value (or they would not have failed under design loads). Further, lumber design values are not based on the expected failure strength determined from mechanical tests, but they include an adjustment factor (essentially a factor of safety) to allow for variability in the material (i.e., an occasional very weak piece). Including the adjustment factor in Figure 13 would result in more pieces falling below the design threshold. The failed truss members and lumber removed from the building for mechanical testing shows that several pieces failed to meet the allowable design value, making them defective (i.e., they are unable to carry the intended loads). The FRT was the primary cause of the reduced strength that resulted in the lumber failing to meet the allowable design values. 4. CONCLUSIONS Based on the investigation of the wood trusses connected by metal plates installed in the residential building, the following conclusions can be drawn: The lumber for the roof truss members was treated with a mono-ammonium phosphatebased FRT. Numerous failures had occurred in truss elements due to degradation of the wood as a result of the FRT. Failures were observed in all wings of the residential building that were built with wood trusses, with the most severe damage observed in the dining room. This is due to the design differences between the dining room and residential areas. The site observations, engineering analysis, and laboratory testing indicated that weakening of the wood truss members by FRT was the cause of failure of the wood trusses installed throughout the structure.

11 The degradation has likely resulted in a strength loss in some of the lumber to a level below what the current building code would require based on the grade stamps. Based on the loss of strength determined by the mechanical tests, the wood components installed in the roof trusses at building that were treated with FRT constituted defective lumber and needed to be replaced. When compared to published test results, the test data indicate that the loss of strength is progressive over time and it cannot be assumed that reinforcement will correct the problem, since the lumber will continue to lose strength. REFERENCES American Society for Testing and Materials (ASTM) (2007). Designation D , Standard Test Methods for Small Clear Specimens of Timber. ASTM International, Bar Harbor, PA. American Society for Testing and Materials (ASTM) (2005). Designation D198 05a, Standard Tests Methods of Static Tests of Lumber in Structural Sizes. ASTM International, Bar Harbor, PA. American Society for Testing and Materials (ASTM) (2003). Designation D , Standard Test Methods for Structural Panels in Tension. ASTM International, Bar Harbor, PA. LeVan, S.L. and Winandy, J.E. (1990). Effects of fire retardant treatments on wood strength: a review. Wood and Fiber Science, 22(1), Lyon, D.E., Bigbee, K.L., and McNamara, W. S. (1988). Evaluation of strength properties of fire-retardant treated wood using the 1983 National Forest Products Association protocol. Forest Products Journal, 38(6), National Forest Products Association (NFPA) (1986). Design Values for Wood Construction, A Supplement to the 1986 Edition. National Forest Products Association, Washington, D.C.