Lateral load testing and analysis of manufactured homes

Lateral load testing and analysis of manufactured homes Koerner, Brian D. 1, Schmidt, Richard J. 2, Goodman, James R. 3, and Richins, William D. 4 ABSTRACT The lateral-load behavior of a single-wide manufactured home was studied by both experimental and numerical methods. Experimental testing consisted of the application of lateral loads using a pressurized air bag. The response of the home was measured with a detailed instrumentation package. A finite element model of the home was built to predict the behavior of the structure under these loads. The results from the full-scale tests were used to validate the finite element model. This research demonstrates that the finite element method can be effectively applied to the analysis of manufactured homes to predict both global and local behavior mechanisms. However, physical testing of components and interfaces from the home is needed to provide reliable data for the analysis model. INTRODUCTION The research reported in this paper is the result of a joint effort by researchers at the University of Wyoming (UW), the Idaho National Engineering and Environmental Laboratory (INEEL), and EDM International (a structural engineering consulting firm in Ft. Collins, CO). The INEEL and the Manufactured Housing Institute (MHI) sponsored this research program to address the nation s need to improve the ability of manufactured housing to withstand high winds. INEEL, UW, EDM, and MHI teamed together to perform a series of full-scale static load tests and structural analyses on a singlewide manufactured home. The tests provided information to advance the process of structural modeling and simulation of manufactured housing subjected to high winds. With data from the full-scale tests, the research team has developed and calibrated finite element modeling techniques for eventual use in the design and analysis of improved manufactured housing. This paper includes a brief review of the lateral-load test program, a description of the finite element model, and a comparison of the finite element results with those from the experimental tests. Additional details of the full-scale tests can be found in a companion paper (Richins, et al, 2) and in the complete test report (Richins et al, 1999). A more general overview of the research effort, as well as an examination of research needs in the area of affordable housing is contained in (Schmidt et al, 2). A detailed report of the finite element analysis is contained in (Schmidt et al, 1999). LATERAL LOAD TESTS A structurally complete manufactured home was tested using uniformly distributed and concentrated loads. The purpose of these tests was to measure displacements and forces on the home and its structural components at design-level lateral loads, such as those due to wind. The forces and displacements consist of tiedown strap forces, global displacements, slip at the interface between components, and wall and ceiling racking deformations. A 14 ft by 62 ft (4.3 m by 18.9 m) single-wide home, manufactured by Oakwood Homes of Fort Morgan, CO, was installed and tested at EDM s research facility in Ft. Collins, CO. The home was installed on concrete masonry piers and anchored using steel tiedown straps to a concrete footing according to manufacturer s recommendations. A floor plan of the home is shown in Figure 1. 1 Structural Engineer, Nelson Engineering, Jackson, WY 832 USA 2 Professor, Dept. of Civil and Architectural Engineering, University of Wyoming, Laramie, WY 8271 USA 3 Research Professor, Dept. of Civil and Architectural Engineering, University of Wyoming, Laramie, WY 8271 USA 4 Idaho National Engineering and Environmental Lab, Idaho Falls, ID 834 USA

Figure 1 Floor plan of the test home Inside the home, scissors trusses formed a cathedral ceiling over the central portion of the home, including the kitchen, living room, and dining room. All other areas of the home had a flat ceiling with horizontal bottom chord trusses. Longitudinal walls support the roof trusses. The home contained two shear walls, one at each end wall, and several interior partition walls (see Fig. 1). Shear walls and partition walls are similar in materials and assembly, but differ in the way they are placed in the home during construction. A partition wall is nailed to the floor and ceiling to provide a divider between rooms. Partition walls may be placed on the floor sheathing, between joists. Since a shear wall is intended to distribute the lateral force in the home, it is always located over a floor joist. Also, it is attached to the floor using lag bolts and nails. The spacing between fasteners in shear walls is smaller than that in partition walls. All walls are sheathed with gypsum board on the inside of the home. Exterior walls have gypsum board on the interior surface with exterior siding on the other side. The gypsum board wall sheathing and oriented strand board (OSB) floor sheathing are attached to the wood framing members using glue and staples. The glue provides a rigid attachment of the sheathing to its frame and contributes to increased stiffness and improved serviceability by reducing cracking in walls and along component interfaces. Staples are used to provide short term attachment of the sheathing panels until the glue cures. The floor system of the manufactured home consists of several different parts. A conventionally framed floor joist system is placed on two steel beams used as chassis rails that run the full length of the home. Perpendicular to these chassis rails are outriggers. The outriggers provide support for the floor system and transfer the load from the floor to the chassis rails, which in turn are supported by concrete masonry piers and anchored by steel tiedown straps. Three air bag tests were performed. The first air bag test was performed to check out the instrumentation system and remove initial slack from the building. Next, the tiedown straps were pretensioned and a second test was performed. For the third test, greased steel plates were placed between the chassis rails and the masonry piers to act as bearings. The objective of this change in the installation was to reduce the friction between the chassis and the piers, thus improving the ability to do computer modeling. The second air bag test is referred to here as the high friction test and the third test, with the greased bearing plates, is referred to as the low friction test. Both of these tests involved monotonic ramp loading up to approximately 3 psf (1.44 kpa) of air bag pressure. Instrumentation focused on four response mechanisms: tiedown strap forces, global displacements, wall racking, and component interface slip. Figure 2 shows a typical cross section of the home with instrumentation. Global displacements of the home at this section, which happened to be at the tongueend of the home, were recorded with LVDTs on channels 9, 1, 11, and 12 of the data acquisition system. Global displacements were measured at six such sections along the length of the home. Also at the section Figure 2 - Typical cross section with instrumentation

shown, shear wall racking was measured by channel 33 using a cable extensiometer with an extension wire, and wall to wall interface slip was recorded on channels 8 and 6 with linear potentiometers. Ceiling-to-wall interface slip was measured on channels 4 and 6, while channels 46 and 48 measured wall-to-floor interface slip. Finally, tiedown straps extended from the top of the chassis rail on the windward side of the home to the base of the airbag support frame. Each tiedown strap was attached to a load cell, such as that on channel 16 (Fig. 2), for direct measurement of strap force. Tiedown straps were spaced at 1 ft (3. m) intervals along the length of the home; masonry piers were placed at 8 ft (2.4 m) intervals. FINITE ELEMENT MODEL The finite element mesh for the manufactured home, shown in Fig. 3, contained 3693 nodes and 219 elements. Quadratic (8-node) orthotropic shell elements were used to model the individual wall, floor, and ceiling components of the structure. Since these components in a manufactured home consist of OSB panels, wood framing, and gypsum board, fastened with rigid glue and staples, full composite behavior of these subassemblies can be assumed; this assumption has been verified by physical tests (Goodman and Klasi, 1996). The model also includes space truss elements for the roof framing and tiedown straps, and 3-node Lagrangian space frame elements for the chassis rails and outriggers. Finally, the interface (connection) between the individual components is modeled with specially designed link elements. The link elements model the inelastic interface between shell elements and retain displacement compatibility with the shells. The link element, shown in Fig. 4, contains six nodes, each with four degrees of freedom (three translations and one rotation). To reduce model complexity, the finite element model was laid out on a regular grid, upon which the node points were located. To align structural components to this grid, it was necessary to shift the locations of some components as well as door and window openings. Grid spacing was 2. ft (.61 m) in the x and z directions (in the horizontal plane) and 1. ft (.46 m) in the y direction (vertically). This grid spacing was chosen based on a convergence study performed in previous research (Jablin and Schmidt, 1996). The models were analyzed using FINITE, a general-purpose structural analysis system for static and dynamic, nonlinear analysis (Dodds and Lopez, 198). Linear analyses were performed in this research. Since the physical testing of components and joints is still pending, material properties for the various finite elements were based on estimates from previous work (Jablin, 199; Creighton, 1997). In those studies, data for the orthotropic shell elements was generated from a laminate theory for the wall, floor, and ceiling components. Link element properties were taken from physical test data of mocked-up joint specimens. Properties for the other elements were selected as typical for the materials used in their construction. Complete tables of the material property data used in this study can be found in (Schmidt et al, 1999). The finite element model was analyzed for two sets of material properties in a limited parameter study to examine the effects of changing material properties. The first analysis used material properties from previous research, as described above. In the second analysis, wall, floor, and ceiling elements were stiffened to reduce deflections in the model. In the first analysis, exterior walls, partition walls, and shear walls all had different properties. For the second analysis, all walls were given the same properties. The partition wall material stiffnesses were increased by two orders of magnitude and used for all wall elements. Properties of the links, floor, and ceiling were not changed. These modified properties were chosen arbitrarily to examine the sensitivity of the model to changes in material stiffness. The modified properties do not represent any specific construction material or method. Figure 3 Manufactured home finite element mesh Wall - shell element Interface - link element 4 6 Floor - shell element 1 2 3 (a) Component interface (b) Link element Figure 4 - Detail of interface between components

In the full-scale tests, two different support conditions were examined. For the first support condition, the manufactured home was placed on concrete masonry piers, as in a conventional installation. During the physical tests with the highfriction support conditions, the piers were observed to tip slightly under lateral load. No attempt was made to model friction or tipping of the piers. The second support condition involved use of greased plates between the home and the piers. The greased plates were intended to reduce friction between the home and the piers, thus enabling simplified modeling that did not include friction or tipping of the piers. The second of these two support conditions was modeled by finite elements using simulated roller supports beneath the chassis rails at the pier locations. COMPARISON OF RESULTS In this section, results of two finite element analyses are compared to those from the lateral load tests. In the comparisons that follow, the labels High Friction and Low Friction identify the experimental results. The labels FEM and FEM Modified identify the initial and the second analyses, respectively. Again, the initial analysis used material properties from previous research and the second analysis used properties with artificially high stiffness for the wall elements. Global Displacements The global displacements from the high-friction test and finite element analyses are shown in Fig.. The global displacements during the full-scale testing involved two distinct parts: rigid-body translation and racking deformation. The rigid-body translation can be attributed to flexibility (initial slack) in the tiedown system. During installation of the home, a steel strap was wrapped in a coil around a steel pin to secure the system. As lateral load was introduced into the system, the coil was tightened around the steel pin allowing the strap to lengthen under relatively low load. The manufactured home continued to translate until the steel strap was wrapped tightly against the pin. No attempt was made with the finite element model to represent slack or initial gaps in the foundation system. Instead, the finite element model responded immediately to load by deformation of the home. Therefore, when comparing the displacements of the fullscale tests to the computer model, it is necessary to compare only the response quantities associated with deformation response of the home. Global Displacement -.4 -.4 -.3 Displacement (inches) -.3 -. -.2 -. -.1 -. 8 7 6 High Friction, Bottom Transducer, 3 psf FEM, Top Tansducers, 3 psf FEM Mod., Top Transducers, 3 psf 4 3 Station (inches) High Friction, Top Transducer, 3 psf FEM, Bottom Transducers, 3 psf FEM Mod., Bottom Transducers, 3 psf 2 1 Figure Global Displacements

Examination of Fig. shows that the racking deflection, given by the difference between the displacements measured by the top transducers and those measured by the bottom transducers, is reasonably well predicted by the initial finite element model. The modified finite element model, with its artificially high wall stiffnesses, substantially under predicts global racking deformation. In the test at station 12, the top transducer recorded less horizontal displacement than the bottom transducer. This behavior was observed in both the high-friction and the low-friction tests. The reason for this anomalous behavior is not known. However, several possibilities exist, including localized flexibility in the floor diaphragm due to a splice in the chord, placement of the partition wall at that location between two floor joists, or some initial gaps between components that may have contributed to unusual flexibility at that location. Tiedown Strap Forces The tiedown forces are shown in Fig. 6. Results from the FEM contained higher values than were observed during the low friction test. This can be attributed to the friction that still existed in the full-scale test. Even in the low friction test with the greased plates, there is still some strap force lost to friction. In the physical tests, the total lateral load on the building was 13,9 lb (62.1 kn). The total horizontal component of the strap forces was 69 lb (3.7 kn) in the highfriction test and 12,2 lb (4.3 kn) in the low-friction test. Hence, about 13% of the horizontal load on the building was resisted by pier friction in the low-friction test. In some cases, the modified FEM exhibited higher strap forces than the original FEM. While in other cases, it was just the opposite. As with the comparison to global displacements, the initial FEM analysis more accurately predicted strap forces than the modified analysis. 3 Strap Forces 3 Air Bag Load (psf) 2 1 1 2 3 3 4 4 Strap Force (lbs) Channel 2 Channel 3 Channel 4 Channel Channel 6 Channel 7 Channel Ch. 2 FEM Ch. 3 FEM Ch. 4 FEM Ch. FEM Ch. 6 FEM Ch. 7 FEM Ch. 8 FE Figure 6 Tiedown Strap Forces Wall Racking As expected, increasing the stiffness in the walls decreases the racking strains. Some walls in the finite element model exhibit close to the same racking strain as was found experimentally. Nevertheless, it is imperative to look at the slope of the load-deflection plots, rather than the specific data points at 3 psf (1.44 kpa). A majority of the data obtained from the full-scale tests displays little response until the lateral load reached from psf (.24 kpa) to 1 psf (.48 kpa). This is due to the slack in the system. The FEM contains no such slack, so it responds immediately to load. Hence, the slopes of the lines are better indicators of model performance than are the data values themselves. The racking strains in the front shear wall (at the tongue end of the home) are shown in Fig. 7. Those in the kitchen partition wall at station 482 are shown in Fig. 8. In both cases, the initial analysis greatly over-predicts wall racking, even though global displacements are reasonably accurate. This suggests a mismatch in material properties between the wall elements and the component interfaces. Certainly, physical testing is needed to develop appropriate material properties for these components.

3 Channel 33 Front Shear Wall At Sta '-" Racking Strain 3 Lo ad (ps f) 2 1 - -.1.1.2.3.4..6.7.8 Strain (in/in) Low Friction High Friction FEM FEM Modified Figure 7 Front Shear Wall Racking Strains 3 Channel 3 Kitchen Wall At Sta 4'-2" Racking Strain 3 Load (psf) 2 1 - -.1.1.2.3.4..6.7.8.9 Strain (in/in) Low Friction High Friction FEM FEM Modified Figure 8 Partition Wall Racking Strains Interface Slip Interface slip results for the high-friction and low-friction tests were nearly identical. Prediction of interface slip by finite element analysis was reasonably successful, considering the small contribution made to overall response by the interfaces. The initial analysis resulted in good matches to measured response. Since it stiffened the walls of the structure and not the interfaces, the modified analysis produced markedly higher slips than did the initial analysis. In many cases, such as the response shown in Fig. 9, the interface responded to load immediately. In others, such as that in Fig. 1, no interface slip occurred until a relatively high lateral load was on the building. This suggests that a static friction mechanism in the interface had to be overcome before the interface could respond to load. Precompression between two components, caused by tightly installed lag screws, could cause such a mechanism.

3 Channel Master Bedroom Wall (Sta. 62'-") - Ceiling Interface Slip 3 Load (psf) 2 1 -.1.1.2.3.4. Displacement (in) Low Friction High Friction FEM FEM Modified Figure 9 Master Bedroom Wall Interface Slip 3 Channel 8 Bedroom #2 - South W all Interface Slip 3 Load (psf) 2 1.1 -.1 -.2 -.3 -.4 -. -.6 -.7 -.8 Displacem ent (in) Low Friction High Friction FEM FEM Modified Figure 1 Bedroom #2 Wall Interface Slip CONCLUSIONS The full-scale tests of the manufactured home provided excellent results. The tests provided accurate data, produced in a controlled laboratory setting. The data yielded plots that aided in finite element model refinement and verification. After testing this home, it is apparent that some of the channels can be moved to other locations to record potentially significant response. For instance, the research team did not instrument the building to measure interface separation. Only slip along the interface was monitored. Separation of one component from another due to tension across the interface could also be significant at certain locations on shear walls. Several trends can be seen in this data. It is apparent that interface slip is not affected by the pier support conditions. This is important to consider when modeling the manufactured home in a finite element program. Even though the chassis rail and piers may have a high coefficient of friction between them, this friction had little effect on interface slip.

Therefore, if a model is analyzed and interface slip output is desired, boundary conditions should not be a major concern. Also, the largest racking strain was.333. If one were to assume a nominal value of 9, psi (62 MPa) as the shear modulus for the shear wall, this strain would correspond to a shear stress of approximately 3 psi (27 kpa). Once again, this result was the same for both the high-friction and low-friction tests. This reinforces the conclusion that internal deformation, and hence forces, are not highly dependent on external boundary conditions. As expected, boundary conditions had a big influence on global displacements. The home moved twice as far horizontally during the low friction test than the high friction test. Finite element analysis of a manufactured home to obtain both global and localized response is feasible for manufactured homes. However, better component and interface data is needed in order to seed the model. Additional parameter studies are needed to identify which material parameters create the greatest sensitivity. Then physical testing can be emphasized in those key areas. ACKNOWLEDGEMENTS The United States Department of Energy under DOE Idaho Operations Office, Contract DE-AC7-99ID13727, sponsored this work. Financial support received from the Manufactured Housing Institute is also acknowledged. The leadership and guidance of Frank Walter, Technical Director of the MHI is also greatly appreciated. REFERENCES Creighton, John, 1997. Finite Element Structural Analysis for Manufactured Homes Under Lateral Loading, Master of Science Thesis for Colorado State University. Dodds, Robert H., and Lopez, L. A., 198. "A Generalized Software System for Nonlinear Analysis," Advance Engineering Software, Vol. 2, No. 4, ppg. 161-168. Goodman, James R.; Klasi, Melvin L., 1996. Testing of Wall Stiffness in the Crownpointe Manufactured Home, South Dakota School of Mines & Technology, February. Jablin, M. C. and Schmidt, R. J., 1996 Finite Element Modeling of Manufactured Homes, Proceedings, International Wood Engineering Conference, Vol. 3, pp. 17-177, New Orleans, LA. 28-31 October. Jablin, Mark C., 199. Finite Element Modeling of Manufactured Homes Using Interface Elements, Master of Science Thesis for University of Wyoming. Richins, W. D., Lacy, J. M., Larson, T. K., Rahl, T. E., Goodman, J. R., Schmidt, R. J., Koerner, B. D., & Pandey, A. K., 2. Full-Scale Structural Testing of a Single-Wide Manufactured Home, Proceedings World Conference on Timber Engineering, 31 July 3 Aug. Richins, W. D., Rahl, T. E., Lacy, J. M., Larson, T. K., Flood, M. N., Goodman, J. R., Schmidt, R. J., Koerner, B. D., Pandey, A. K., Stewart, A. H., and Walters, F. 1999 "Full-Scale Structural Testing of a Single-Wide Manufactured Home, Oakwood Model 132; 13ft x 62ft," INEEL Test Report, INEEL/EXT-1999-31, March. Schmidt, R. J., Koerner, B. D. and Goodman, J. R. 1999. Methods for Manufactured Home Testing and Analysis, Contract Report, Contract Report to Idaho National Engineering and Environmental Laboratory, Idaho Falls, ID and Manufactured Housing Institute, Arlington, VA, Dec. Schmidt, R. J., Goodman, J. R., Richins, W. D., and Pandy, A., 2. Improved Design of Manufactured Homes for Hazardous Winds, Proceedings World Conference on Timber Engineering, 31 July 3 Aug.