Use of vibration techniques to determine the rotational stiffness of timber joints Paul Crovella 1, Dr. George Kyanka 2 Abstract A method for determining the rotational stiffness of timber joints using the natural frequency of vibration of the beam is presented. Reviewed research shows that joint stiffness is important for frame behavior when the relative value of joint stiffness exceeds the beam flexural stiffness (α >1). Lab results show that vibration could predict joint stiffness of both hardwood (within 17%) and softwood (within 16%) beams when restrained in a metal fixture (α>1). A beam tested in a metal fixture with sequentially weakened joints shows that the stiffness decreased with a matching decrease in frequency. Two softwood frames with all timber joinery, one with tight and one with loose joints, were tested. Results show that tight wood joinery can produce values of α>1. The stiffness values based on vibration ranged from 2-31% of the measured values (α>1). Finally, a discussion of the considerations for field application of the method is included. Keywords timber joints, non-destructive testing, rotational stiffness, vibration 1. INTRODUCTION 1.1. Overview There is a growing recognition of the need for a more sustainable approach to construction. With this awareness comes an interest in studying existing buildings as a resource for redeveloping the built environment. Existing buildings can be deconstructed and the individual structural members reused, however “the greenest building is the one that is already built” (Carl Elefante). Successful reuse of entire buildings will depend on the ability to determine their current structural condition. To be able to structurally model an existing timber frame, the member properties (strength, stiffness, and dimensions) of the elements must be known, as well as the boundary conditions (strength and stiffness of joints) for each member. While extensive research has been performed to nondestructively evaluate the member material properties (e.g. Ross et al, 2004, Kasal et al, 2009), the research on in-situ determination of the member joint properties is relatively limited. In identifying critical research needs for the American Society of Civil Engineers Structural Engineering Institute, Ron Anthony wrote “Under design loads, seldom do wood members fail in a structure unless they are severely deteriorated. Failures generally occur at connections. Yet we have a wealth of knowledge about wood properties, but not the behavior of connections. Unfortunately, connections are critical in structure performance,…,and yet we do not have a reliable means to assess their condition or capacity.” (Anthony, 2008) 1 Paul Crovella, Department of Sustainable Construction Management and Engineering, State University of New York – College of Environmental Science and Forestry, US, plcrovella@esf.edu 2 Dr. George Kyanka, Department of Sustainable Construction Management and Engineering, State University of New York – College of Environmental Science and Forestry, US, ghkyanka@esf.edu 1.2. Literature Review The determination of wood properties in-situ has been studied by many researchers. Ross and Pellerin (1994) summarize the work done on the approaches and accuracy of assessment of basic wood properties for in-place members. Their work shows that while “relationships between NDT parameters and performance of wood members have been established and verified”, much of what limits NDT methods on-site are the boundary conditions of the wood member. Much work has been done on using low frequency vibration of wood members with known boundary conditions to detect anomalies in stiffness (and potentially strength). As a first step in developing a structural building health monitoring system, Cai found identification of the vibration flatwise fundamental frequency of floor joists as an effective method (r2 = 0.68 to 0.96) to determine the static flatwise MOE (Cai et al, 2000). Soltis determined that measuring natural frequency was more effective than damping ratio at determining stiffness loss in floor systems, and that the difference between damped and undamped frequencies in the floors was negligible due to the low damping ratio (Soltis et al, 2000). Simply supported floors tested by Wang (2005) showed differences of less than 3% between frequencies measured by forced and free vibration. Fewer researchers have worked on determining joint properties in-situ. Destructive testing of traditional timber joints to determine stiffness and strength has been researched by Brungraber (Brungraber, 1985) and others (Bulleit, 1999), (Schmidt, 1997), (Burnett et al, 2003). Prediction of tensile strength for different traditional timber joints with specified geometry ratios has been adopted by the Timber Frame Engineers Council based on work by Miller and others (2010). Anthony has found x-ray techniques effective to reveal in-situ joint geometry and advanced decay. Morlier, Bos, and Castéra (2006) used modal analysis to predict wood joint stiffness in a small scale wood portal frame. However none of these methods are appropriate for rapid evaluation of existing structural assemblies. 1.3. Theory The fundamental frequency for a uniform beam is governed by the beam’s geometry, material properties, and the joint stiffness (Eq. 1). If the geometry, material properties, and natural frequency can be determined non-destructively, the stiffness of the joints can be determined. This equation is derived by combining expressions for the displacement of beam with the equation of motion of a harmonic oscillator. The solution of this combined equation results in the general form sine, cosine, and hyperbolic sine and cosine. The exact shape is found by applying boundary conditions to the solution. This solution can then be expressed in terms of the frequency that matches the boundary conditions. For a uniform beam, the resulting equation is: fn = Kn (EI/ w)1/2 2πl2 (1) Where fn = Natural frequency of mode n (cycles per second), Kn = Constant dependent on the boundary conditions and mode, w= uniform mass per unit length (including beam mass), E = modulus of elasticity, I = second moment of area, l= length of beam The value of Kn for the first mode of vibration is 9.87 for a simply supported beam, and 22.4 for a beam fixed at each end. (Figure 1) A useful expression for describing the joint stiffness between the two extreme values is the “normalized frequency” which represents a linear variation between 0 for a simply supported beam, and 1.0 for a fixed-fixed beam. Figure 1 –Fundamental Frequency of a beam depending on boundary conditions (adapted from McGuire, 1995) If the beam properties and fundamental frequency are determined, the constant Kn can be calculated as shown in Equation 2 Kn = fn 2πl2 (EI/ w)1/2 (2) The value of Kn is not the value of the rotational stiffness of the joint, but rather a value that can be correlated to the rotational stiffness of the joint. An important point arises here: When considering a beam with semi-rigid joints, whether the joint stiffness will be significant in the analysis depends on the ratio of the joint stiffness to the bending stiffness of the beam. As shown in Figure 2, when the beam flexural or the joint rotational stiffness is large in relation to the other, the effect of changing joint stiffness has a minimal effect on frequency. When the joint stiffness is between 1 and 100 times that of the flexural stiffness, a change in joint stiffness produces a relatively large change in frequency. 1 0.9 Normalized frequency 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0.001 0.1 10 1000 100000 α = kL/EI Figure 2 – Effect on natural frequency of varying rotational joint stiffness relative to beam stiffness Research by Leichti and others (1995) looks at the difficulty of creating a moment resisting connection in wood, and suggests that based on published tests, a 50% rigid connection is a possible maximum value for wood connections due to the deformation of wood around the embedded fasteners. From Figure 2, this corresponds to values on the baseline curve that fall below 0.5 normalized frequency. 2. METHODOLOGY 2.1. Experimental Apparatus The frequency was determined using an accelerometer placed on the beam to produce an electrical signal due to the free vibration of the member. The accelerometer was a quartz shear type single-axis accelerometer (500 mV/g). The accelerometer was attached through a 9 m (30 ft) foot cable to a USB powered signal conditioner. This signal conditioner is powered through the USB port of the computer, and provides a constant current input to the accelerometer. The voltage at which this current flows is measured by the signal conditioner, and then amplified and output to the computer via a stereo jack. The sound card of a personal computer accepted the stereo jack signal as an input. A digital audio editor “Audacity 1.2.6" was used to capture, edit, and save the signal as a wav format file. The signal was then processed using a Fast Fourier Transform (FFT), with code based on an applet from Evan Mroz. The resulting frequency response spectrum output file was imported to Excel® to identify the maximum spectral component for the first mode natural frequency. 2.2. Beam testing 2.2.1. Validation The first phase of the research studied the ability of frequency measurements to predict the rotational stiffness of softwood and hardwood beams connected to a rigid test frame. The first test used a softwood beam, Tsuga canadensis, 18.4 cm x 18.4 cm x 3.0 m (7.25 in x 7.25 in x 120 in) clamped at each end to a continuous steel beam. The steel beam had a flexural stiffness much greater than the wood beam, allowing the wooden member to be vibrated relative to the rigid steel frame. The center of the wood beam was instrumented with the accelerometer and then the beam was struck by hand and allowed to freely vibrate. The beam properties (MOE, I, density) and measured frequency were used to calculate rotational stiffness of the joints. This was compared to the rotational stiffness as measured by center-point loading the beam and measuring the deflection. The deflection of a beam with equal semi-rigid joints can be determined in closed form (Pilkey, 2005), and the measured deflection was used to determine the rotational stiffness. The experimental set-up for this work is shown in Photo 1 Photo 1 – Beam in test apparatus and data acquisition equipment Next, a similar size beam of hardwood, Quercus rubra, was studied under the same two conditions (simply supported and clamped). The results of frequency testing the two beams are shown in Figure 3. 1 0.9 0.8 Normalized frequency 0.7 0.6 0.5 Frequency Softwood 0.4 Frequency Hardwood 0.3 0.2 0.1 0 0.001 0.01 0.1 1 10 100 1000 10000 100000 α = kL/EI Figure 3 – Comparison of joint stiffness measured by frequency to deflection – clamped and simply supported. 2.2.2 Beam with Varying Joint Stiffness Testing of joints with intermediate rotational stiffness was done by using the softwood beam in the fixture, but progressively cutting through the clamps at each end to decrease the stiffness of the restraints (Photo 2). Photo 2 - Image of metal plates being cut through to reduce their stiffness 1 0.9 0.8 Normalized Frequency 0.7 0.6 0.5 Frequency - softwood 0.4 0.3 0.2 0.1 0 0.001 0.01 0.1 1 10 100 1000 10000 100000 α = kL/EI Figure 4 - Comparison of joint stiffness measured by frequency, to stiffness measured by deflection as joints are weakened 2.3 Frame Testing Next, two timber “H” frames were fabricated with Pinus strobus to test the effect of creating joint stiffness using only timber joinery. The frames had identical overall dimensions: two 17.75 cm x 17.75 cm x 3.65 m posts connected to a 17.75 cm x 17.75 cm x 6.10 m beam (two 7 in x 7 in x 12 ft posts connected to a 7 in x 7 in x 21 ft beam). One of these frames was fabricated with mortise and matching tenons of identical size (0.000 m clearance - tight). The holes in the tenon were placed 3.2 mm (0.125 in) closer to the tenon cheeks than the holes in the mortise to the face of the mortise. When assembled with a wood dowel the tenon cheeks were forced against the mortise by the 3.2mm (0.125 in) difference (“Draw bored”). The other frame was fabricated with mortise and tenon joints with a uniform 3.2 mm (0.125”) clearance between all surfaces (loose). Sample dimensions are shown in Figure 5. The frames were fabricated using a CNC joinery machine at New Energy Works in Farmington, NY, and tested at the State University of New York in Syracuse as shown in Photo 3. Photo 3 – “H” frame assembled for testing in lab Figure 5 – Sample joint dimensions (loose) The tests compared the stiffness of the joints measured by deflection to the stiffness measured by vibration. Four boundary conditions were applied during the tests: First the beam was set in place, but no pins were installed (simply supported). Second, a single pin was installed at each mortise. Third both pins were installed in each mortise. Finally, with both pins installed in each mortise, the columns were loaded with a 910 kg (2000 lb) load. The results for the loose and stiff frames are shown in Table 1 for frequency and in Table 2 for stiffness. Figure 6 shows the results compared to the baseline curve. Table 1 – Frequency Measurements (Hz) for “H” Frames Loose Frame Loose Frame Tight Frame Tight Frame Equivalent Deflection Vibration Equivalent Deflection Vibration Simply supported 6.80 6.31 6.94 6.60 1 pin installed 6.77 6.98 9.32 9.00 2 pins installed 7.28 7.36 9.54 9.58 7.40 9.71 9.73 Boundary Conditions 2 pins w/load on No data taken column Table 2 – Joint Stiffness (kg-m/rad) for “H” Frames w/% error Loose Frame Loose Frame Tight Frame Tight Frame Deflection Equivalent Vibration Deflection Equivalent Vibration Simply supported 0 -2940 * 0 -760* 1 pin installed -5.1 480 * 21,900 15,200 (-31%) 2 pins installed 2,070 3,250 * 26,400 27,200 (+3%) 3,580 * 29.700 30,200 (+2%) Boundary Conditions 2 pins w/load on No data taken column * α < 1 no % error given 1 0.9 0.8 Normalized frequency 0.7 0.6 0.5 Loose Frame Tight Frame 0.4 0.3 0.2 0.1 0 0.001 0.01 0.1 1 10 100 1000 10000 100000 α = kL/EI Figure 6 – Stiffness of loose and tight H frames with various boundary conditions. 3. DISCUSSION 3.1 Effectiveness of method as a function of α The three sets of tests share common characteristics: There is a much higher degree of precision when α > 1. In the case of the frames, this method predicts actual joint stiffness within 2-31% for α >1. Although this range is large, for structural analysis the stiffness can vary over orders of magnitude. This degree of precision will allow for a known degree of certainty during analysis. For the samples with α < 1, the ability of the method to predict the stiffness is inadequate. However when analyzing a structure, knowing that the joint stiffness is low in relation to the member stiffness will allow the designer to assume no joint stiffness, and use a pinned model. This is an acceptable assumption for this case. Further, the values for the frame joints tested with double pins and with a single pin indicate that the method can distinguish the decrease in stiffness due to the missing/failed pin (e.g. damaged joint) at higher stiffnesses (18% strength loss vs. 12% error). Contrary to popular opinion, a timber joint can be created with sufficient stiffness to have a significant effect on beam stresses due to bending. 3.2 Challenges for extending work to the field. A number of factors complicate the application of the method under field conditions: Determination of MOE and density, assumption of equal stiffness at each joint, attachment of materials and braces, and vibration of other parts of the structure are to be considered. 3.2.1 Determination of MOE and Density The determination of MOE (and density) might be performed in two ways in a field scenario: A visual grade assigned, or a small wood sample taken and tested. In the case of the H frame material, identification of the species and grade (Eastern White Pine, Grade 2) yields an MOE= 6200 MPa (900,000 psi) (AFPA, 2005). The increment bore samples yielded a specific gravity of 0.33. Using the specific gravity to determine the MOE as described in the Wood Handbook (FPL, 2010), the clear sample MOE would be 6800 MPa (990,000 psi). The ASTM procedure for correcting the value for defects (ASTM, D245) would reduce this to 80% of the clear sample value, 5400 MPa (790,000 psi). The Wood Handbook lists 22% as the average coefficient of variation of MOE and 10% as the average COV of specific gravity for all species. Table 3 shows the percent change in stiffness for MOE values calculated with various methods. Table 3 – Values of stiffness using different methods for determining MOE MOE Value (MPa) Stiffness (kg-m/rad) Deflection 5,700 30,200 0 Species and visual grade EWP timber, #2 6,200 24,200 20% Measure density and ASTM 245-00 5,400 33,300 10% 4,500 -7,000 23,60036,500 21% MOE Method MOE of 22% COV, density by 10% COV Stiffness % error Theoretically, a transverse vibration of the timber could also yield an MOE if the joints had no stiffness in the transverse direction. An alternative to using the increment bore to determine the density, would be the approach used by Minster, et al (2006) for MOE determination based on bore testing. 3.2.2 Assumption of equal stiffness at each joint –Machine fabrication of the joints ensured that the stiffnesses of each end of the beam were similar. However, multiple researchers have modeled the frequency response for a beam with unequal stiffness joints at each end, and the mode shape and frequency can be calculated for these conditions. In an example by Gorman (1975), a beam with a normalized frequency of 0.22, and individual spring stiffness of α=2.678 and α=1.53, (one spring 73% stiffer than the other) the natural frequency of vibration would be equivalent to that of a beam with equal springs of α = 1.85 (a 14% error compared to the arithmetic average α stiffness). 3.2.3 Attachment of materials and braces – The vibration characteristics of beams with point loads and distributed mass have also been modeled by researchers. Vibration characteristics for continuous beams pinned at multiple points (such as connections due to knee braces) have also been determined. However the multiple span beam does not allow for determining joint stiffness with a single vibration measurement. This will require modal analysis. 3.2.4 Vibration of other parts of the structure – As joints stiffness increases, more of the deformation is transferred through the joint to the surrounding structure. As Leichti (1999) notes, typically the ratio of column to beam dimensions is such that the columns are stiffer than the beams. Depending on the relative stiffness of these members, their deformations can influence the vibration of the adjacent member. While this effect may require modal analysis to determine the final behavior, a sensitivity analysis of the effect of loading the columns of the H frame was performed. The frequency of the beam vibration and the joint stiffness was measured as a load was placed on the column. The load was increased in increments of 227 kg (500 pounds) from ~0 to 910 kg (2000 lbs). For the tight frame, this caused the joint stiffness to increase by 12%. 4. CONCLUSIONS The understanding of the effect of joint stiffness on beam behavior is based on the relative stiffness of the joint and beam. The vibration method does not effectively distinguish low joint stiffnesses (α < 1), but by determining the stiffness to be low, allows the designer to use an appropriate model (pinned joints). In the case of semi-rigid behavior (α >1), the method predicted the stiffness within 12%, on average. Since the stiffness can vary be an order of magnitude in this range, this 12% error can be considered in that context. The method was able to distinguish the missing peg (simulating failure) from the double peg joint (18% loss vs. 12% error). However much more testing would be needed to develop a clear understanding of the average error for the α < 1 range. A number of confounding factors for application in the field will affect the results. The method can not distinguish when joints have unequal stiffness at each end of the beam. The method is affected by the stiffness and load of the members supporting the beam. The method will apply only when attached members and materials can be idealized as loads and restraints. The method will require a field determination of the MOE. As shown in section 3.2, each of the above mentioned factors could be expected to introduce an error of approxiamtely 10-20%. ACKNOWLEDGMENTS Research was supported by donations in-kind from Northcott Wood Turning and New Energy Works. REFERENCES American Forest and Paper Association. (2005). National design specification for wood construction. American Wood Council. Madison, WI. American Society for Testing and Materials. (1993). D 245-00 Standard Practice for Establishing Structural Grades and Related Allowable Properties for Visually Graded Lumber. Annual book of ASTM standards. Vol. 04.10. Philadelphia, PA. Anthony, Ronald W., (2008). Maintenance and Rehabilitation – Assessing and Extending the Life of Existing Wood Structures. Paper presented at the Structural Engineering Institute PreCongress Workshop “Wood Engineering Challenges in the New Millennium: Critical Research Needs” for the annual international meeting of the American Society of Civil Engineers, April 23-24, 2007 in Vancouver, BC, Canada. Audacity 1.2.6 free download available at http://audacity.sourceforge.net/ Blevins, Robert D. (2001). Formulas for Natural Frequency and Vibrations. Krieger Pub Co. 506 pp. Brungraber, R. (1985). Traditional Timber Joinery: A Modern Analysis. PhD dissertation, Stanford University, Palo Alto, CA. Bulleit, W.M., L.B. Sandberg, M.W. Drewek, and T.L. O’Bryant.. (1999). Behavior and Modeling of wood-pegged timber frames. J. of Struct. Engr. 125(1):3-9. Burnett, D., P. Clouston, D. Damery, and P. Fisette. (2003). Structural properties of pegged timber connections as affected by end distance. Forest Products Journal 53(2): 50-57. Cai, Z., M.O. Hunt, R.J. Ross, and L.A. Soltis. 2000. Static and vibration moduli of elasticity of salvaged and new joists. Forest Prod. J. 50(2):35-40. Cai, Z., R.J. Ross, MO. Hunt, and L.A. Soltis. 2002. Pilot study to examine use of transverse vibration nondestructive evaluation for assessing floor systems. Forest Prod. J. 52(I):89-93. Forest Products Laboratory. (2010) Wood handbook - Wood as an engineering material. General Technical Report FPL-GTR-190. Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory: 508 p. http://www.fpl.fs.fed.us/products/publications/specific_pub.php?posting_id=18102 Gorman, Daniel J. (1975) Free Vibrations Analysis of Beams and Shafts Wiley. 386 p. Kasal, B., R. Anthony and M. Drdacky. 2009. In-Situ Evaluation of Historic Wood and Masonry Structures 130 p. http://www.rilem.net/proContents.php?pro=pro062 Leichti, R.J., Hyde, R.A., French, M.L., and Camillos, S.G. The continuum of connection rigidity in timber structures. Wood and Fiber Science, 32(1), pp. 11-19, 2000. McGuire, John, (1995) “Notes on Semi-Rigid Connections” Finite Element Modeling Continuous Improvement. http://femci.gsfc.nasa.gov/semirigid/index.html Miller, J., Schmidt, R., and Bulleit W. (2010) New Yield Model for Wood Dowel Connections J. Struct. Engrg. Volume 136, Issue 10, pp. 1255-1261 Minster, J., M. Micka, and P. Václavik (2006). Determination of elastic modules by means of compression test of a timber core. 5 p. http://www.rilem.net/proContents.php?pro=pro062-009 Morlier, Bos, and Castéra, Diagnosis of a Portal Frame using Advanced Signal Processing of Laser Vibrometer Data Journal of Sound and Vibration Volume 297, Issues 1-2 , 24 October 2006, Pages 420-431 Pilkey, Walter D. (2005). Formulas for Stress, Strain, and Structural Matrices (2nd Edition). John Wiley & Sons. 1571 p. Ross, R.J., B.K. Brashaw, X.Wang, R.H. White, and R.F. Pellerin. 2004. Wood and Timber Condition Assessment Manual. Forest Prod. Soc., Madison, WI. 73 pp. Ross, Robert J., and Ray F. Pellerin, 1994. Nondestructive Testing for Assessing Wood Members in Structures: A Review. Gen. Tech. Rep. FPL_GTR_&) (Rev.). Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory. 40 p. Schmidt, R. J. and R.B. MacKay. (1997). Timber frame tension joinery. Report on research sponsored by the Timber Frame Business Council. Washington, D.C. Soltis, Wang, Ross, and Hunt (2000). Vibration Testing of Timber Floor Systems. Forest Products Journal. Vol. 52, No. 10. Wang, Ross, Hunt, Erickson, and Forsman (2005) Low Frequency Vibration Approach for Assessing Performance of Wood Floor Systems Wood and Fiber Science, 37(3), 2005, pp.371-378.