10NCEE Tenth U.S. National Conference on Earthquake Engineering Frontiers of Earthquake Engineering July 21-25, 2014 Anchorage, Alaska DYNAMIC CHARACTERISTICS ASSESSMENT OF STEEL BEAM-COLUMN CONNECTIONS WITH FLOOR SLAB M. Kurata 1, Z. Tang 2, K. Minegishi 3 and Y. Shi 2 ABSTRACT The dynamic characteristics of steel buildings and their structural components change with damages induced by severe earthquake loading, while the amount of changes vary with the severity of damage. Nevertheless, the identification of the exact sources of the changes is challenging because the observed changes are not drastic due to the redundancy in building structural systems. Moreover, the combination of various types of structural damage, softening in foundation, and composite actions of different structural materials complicate the identification process of damage. Accordingly, this research examines the influence of different structural damages on the dynamic characteristics of a beam-column connection with concrete floor slab. A full-scale specimen was incrementally damaged with quasi-static cyclic loading and forcedvibration test was conducted at the occurrence of notable damage. In the test, the influences of different type of damages, i.e. crack in floor slab, beam local buckling and beam-end fracture, on the natural frequencies of the specimen were carefully monitored. Test results showed the large influence of floor slab cracking and relatively small influence of beam-end fracture on the natural frequencies when the specimen was locally excited with the source of excitations near the connection. This finding was further examined analytically and numerically. 1 Assistant Professor, Disaster Prevention Research Institute, Kyoto University, Kyoto 611-0011, Japan 2 Researcher, Disaster Prevention Research Institute, Kyoto University, Kyoto 611-0011, Japan 3 Graduate Student Researcher, Dept. of Architectural Engineering, Kyoto University, Kyoto 615-8540, Japan Kurata M., Tang Z., Minegishi K. and Shi Y. Dynamic Characteristics Assessment of Steel Beam-Column Connections with Floor Slab. Proceedings of the 10 th National Conference in Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 2014.
Dynamic Characteristics Assessment Of Steel Beam-Column Connections With Floor Slab M. Kurata 1, Z. Tang 2, K. Minegishi 3 and Y. Shi 2 ABSTRACT The dynamic characteristics of steel buildings and their structural components change with damages induced by severe earthquake loading, while the amount of changes vary with the severity of damage. Nevertheless, the identification of the exact sources of the changes is challenging because the observed changes are not drastic due to the redundancy in building structural systems. Moreover, the combination of various types of structural damage, softening in foundation, and composite actions of different structural materials complicate the identification process of damage. Accordingly, this research examines the influence of different structural damages on the dynamic characteristics of a beam-column connection with concrete floor slab. A full-scale specimen was damaged incrementally with quasi-static cyclic loading and forcedvibration tests were conducted at the occurrence of notable damage. In the test, the influences of different type of damages, i.e. crack in floor slab, beam local buckling and beam-end fracture, on the natural frequencies of the specimen were carefully monitored. Test results showed the large influence of floor slab cracking and relatively small influence of beam-end fracture on the natural frequencies when the specimen was locally excited with the source of excitations near the connection. This finding was further examined analytically and numerically. Introduction Building structures subjected to seismic loads show sudden changes in their natural frequencies as illustrated in recent earthquakes and large-scale shake table testing [1, 2]. The changes in natural frequencies and other vibration characteristics (e.g. acceleration responses, mode shape) implicitly indicate temporal or permanent changes for the physical properties or fixing condition of monitored building structures. However, it is challenging to identify the exact sources of the changes since the observed changes are not drastic mainly due to the redundancy in building structural systems. Moreover, the combination of various types of structural damage, softening in foundation, and composite actions of different structural materials further complicate the identification process of damage. Simulating the influences of different structural damage on the dynamic characteristics of 1 Assistant Professor, Disaster Prevention Research Institute, Kyoto University, Kyoto 611-0011, Japan 2 Researcher, Disaster Prevention Research Institute, Kyoto University, Kyoto 611-0011, Japan 3 Graduate Student Researcher, Dept. of Architectural Engineering, Kyoto University, Kyoto 615-8540, Japan Kurata M., Tang Z., Minegishi K. and Shi Y. Dynamic Characteristics Assessment of Steel Beam-Column Connections with Floor Slab. Proceedings of the 10 th National Conference in Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 2014.
building structures and their structural components require careful considerations in numerical modeling. Simple assumptions of removing structural elements or reducing member stiffness may lead to overestimation or underestimation of the performance of damage detection algorithms. However, it is not rare to find the inclusion of unrealistic damage to building models in the validation and performance evaluation processes for system and damage identification algorithms. This is in part due to the lack of large-scale experimental data that evaluate the dynamic characteristics of structural components in sub-assemblage level. In steel structures, major structural components that sustain severe seismic damage are beam-column connections. While large-scale test data of steel beam-column connection under quasi-static loading can be found elsewhere [e.g. 3], seismically induced changes in the dynamic characteristics of steel beam-column connection are not well examined. Thus, the objective of this research is to examine the influence of various structural damages on the dynamic characteristics of a beam-column connection with floor slab in large-scale structural testing. First, a full-scale beam-column connection was damaged incrementally with quasi-static cyclic loading, and forced-vibration tests were conducted at the occurrence of notable damage in floor slab and steel beams. Secondly, influences of different types of damage, i.e. crack in concrete floor slab, beam local buckling and beam-end fracture, on the natural frequencies of the connections were carefully examined. Then, using the same specimen, influences of beam fracture on the dynamic characteristics was separately examined by incrementally welding the fractured sections near the column. Finally, the findings from tests were further examined with an analytical study. Forced-Vibration Tests of Beam-Column Connection Subjected to Cyclic Loading When subjected to severe earthquake loading, beam-column connection in steel buildings resists against story drift and sustains various types of damage in steel beam and floor slab. These damage were reproduced in quasi-static loading test and, as damage progressed notably, the dynamic characteristics at different damage extents were identified by applying dynamic excitations using a modal shaker. Specimen and Loading System The specimen was a typical beam-column connection of a mid-rise steel building designed following Japanese design guidelines for steel structures [4]. The specimen had a cruciform shape with the height of 3000 mm and the width of 6000 mm (Figure 1). Cold-formed steel tubes (BCR 295, 350 350 19 mm) and hot-rolled wide flange beams (SN400B, 400 200 8 13 mm) were welded with through-diaphragms, which are widely used in Japan. The yielding stress of the column, beam flange and beam web identified in coupon test were 400, 296, and 336 MPa, respectively. Floor slab with the thickness of 120 mm and the length of 3800 mm rested over steel beams and headed shear studs of D19 were used to achieve composite action in steelconcrete. The width of the floor slab was defined as 1000 mm following the effective width of RC slab recommended in Japanese design guideline for composite structures [5]. A reinforcement of D10 was provided for longitudinal bars in one layer distributed evenly. As shown in Figure 1, the specimen was pin-supported at the column bottom and rollersupported at the beam ends, assuming these points correspond to the points of inflection. Quasi-
static loading was applied to the column top. A modal shaker (APS Dynamics 113) firmly attached to the left beam at the outside of the RC slab was the source of dynamic excitations which applied vertical forces to the left beam and induced rotational vibrations of the connection. In the forced-vibration tests, the same boundary conditions with the quasi-static loading test were used to evaluate the dynamic characteristics of the beam-column connection itself when it was excited locally in a building. Quasi-Static Cyclic Loading Test Loading Protocol Figure 1. Specimen, loading system and measurement locations. Figure 2(a) shows the test loading protocol in terms of drift angle, which was defined as the lateral displacement of the column top over the column height. Up to the drift angle of 3%, the specimen was loaded for three times at each drift angle. At the drift angle of 4%, cyclic loading was repeated for 5 times until both the left and right beams sustained bottom and web fracture. Hysteresis Behavior and Damage Progress Figure 2(b) shows the hysteresis behavior of the specimen in terms of jack force and drift angle. The beam end of the specimen yielded approximately at the drift angle of 1% and the force reached its maximum value at the drift angle of 3%. The progress of damage in terms of the drift angle is marked in the loading protocol. The observed damage in the concrete floor slab and steel beams are summarized in Figure 2 (c) and (d). Crack damage in floor slab continued to progress as deformation increased. While significant damage in floor slab was observed even for small amplitude loading of 0-3%, the influence of the floor slab damage on the strength of the specimen was negligible. The strength of the specimen deteriorated mainly with the progress of damage in the steel beams. Local buckling of bottom flanges initiated in the first cycle at the drift angle of 3%, while their influence on the strength of the specimen was slight. When small cracks initiated around the tails of the scallops in the first and second cycles at the drift angle of 4%, the strength of the specimen started to decrease, and dropped significantly in the 4th cycle with the formation of large fracture of the entire width of the bottom flange and the 2/3rd of the web.
5 Drift angle (%) Case8 1.5 Fracture 0 Case1 Case2 Case3 Case4 Case5 Case6 Case7 0.0 σ/σ -5 0.5% 1% 1.5% 2% 3% 4% (a) Fracture Vibration test -1.5-5 0 5 Drift angle (b) (c) (d) Figure 2. Quasi-static cyclic loading test: (a) loading protocol; (b) hysteresis behavior; (c) progress of floor slab damage; (d) progress of beam damage. Forced-Vibration Test Forced-vibration tests were conducted at the occurrence of notable damage in the floor slab and steel beams. Considering a practical testing environment after earthquake events, the specimen was unloaded to zero jack force prior to the tests. Figure 2(a) shows the eight major damage cases where the dynamic characteristics of the specimen were examined. Case 1 to 5 include the damage in floor slab and yielding of steel beams. Case 6 to 8 include both the progress of floor slab damage and severe steel beam damage (i.e. local buckling and fracture). Measurement Plan and Excitations The locations of accelerometers for the forced-vibration test are A1 to A8 in Figure 1. MEMS accelerometers with the frequency range of DC to 300 Hz (Silicon Design 2012-002) were firmly attached to the bottom flange of the beam. Acceleration data were collected through a wireless measurement network using Civionics Narada wireless modules. The excitation was a white noise acceleration with the band of 20 to 150 Hz. Considering the peak frequency of the specimen corresponding to the vibration mode with beam bending observed around at 80 Hz, the
sampling frequency of the measurement system was set to 400 Hz. With the limitation in sampling data point in wireless sensing system, the sampling duration was set to 18 seconds. Test Results Figure 3 (a) and (b) shows the power spectrum density of the acceleration records measured at the location A4 for different damage cases defined in the previous section. The acceleration data were band-pass-filtered with the range of 30 to 150 Hz. In the range of 30 to 150 Hz, only one peak appeared consistently around the frequency of 70 to 80 Hz as surrounded with a dotted circle in the plots. While the peak was not very clear due to the relatively low power of the modal shaker and the large damping characteristics of the specimen in the high frequency range, slight reduction of the peak frequency was observed as damage progressed. Figure 3(c) summarizes the reduction of frequency from the undamaged condition for eight damage cases. When the amplitude of cyclic loading increased to the drift angle of 3% (i.e., Case 1 to 6), the peak frequency significantly decreased by 15%. The changes from case 1 to 5 involved the crack in the floor slab and the yielding of the beams, and the corresponding reduction in the peak frequency was 9%. The progress of concrete damage and the local buckling of the bottom flanges and web further decreased the peak frequency from case 5 to 6 by 6%. However, the formation of large fracture in the bottom flange and web of one beam-end (from Case 6 to 7) only decreased the peak frequency by 3%. Case 8 was conducted right after the fracture of both beams without unloading the specimen and did not show the reduction of frequency from Case 7; large force externally applied to the specimen likely affected the result. PSD (db) 10 0-10 Case 1 Case 2 Case 3 Case 4 PSD (db) 10 0-10 Case 5 Case 6 Case 7 Case 8-20 Figure 3. 0 20 40 60 80 100 Frequency (Hz) Freq. reduction (%) 20 15 10 5 0 (a) Deformation increased and slab damage progressed (c) Forced-vibration test results for cyclic loading: (a) power spectrum density of acceleration for case 1 to 4; (b) power spectrum density acceleration for case 5 to 8; (c) reduction of frequency by cyclic loading. -20 1 2 3 4 5 6 7 8 Cases 0 20 40 60 80 100 Frequency (Hz) (b) Fracture at both beam-ends Fracture at one beam-end Local buckling Constant deformation
Discussions Simulated Beam Fracture Test In the quasi-static cyclic loading test, the reduction in the peak frequency was observed consistently when the amplitude of the loading increased and the damage of the floor slab increased. However, the peak frequency seemed to change slightly with the progress of the beam damage. To examine this finding, the same specimen, which was retrofitted after the first test by placing new floor slab around the column (1250 mm 1000 mm 120 mm) made of steel fiber reinforced cementitious composite (SFRCC) and bottom flange made of a haunch plate and tested again for another cyclic loading test, was used to simulate beam fracture by cutting (with gas torching). Note that the SFRCC slab around the column remained almost undamaged during the cyclic loading test. Fracture Patterns Figure 4(a) shows the fracture patterns studied in the test. In the plot, the symbols X(+) and X( ) mean the relative location of the sections to the column in the X direction defined in Figure 2(a). First, the beam-end in X( ) direction was cut from the edge of bottom flange to the 2/3rd of the web (Patter 1 to 4) and then the beam-end in X(+) direction was cut from the center of the bottom flange to the 2/3rd of the web (Pattern 5 to 8). (a) (b) Figure 4. Simulated beam fracture test: (a) layouts of simulated fracture at beam ends; (b) reduction in natural frequency. Test Results For each fracture pattern, the forced-vibration test described in the last section was repeated. Figure 4(b) shows the reduction of the peak frequency for different fracture patterns. The reduction in the frequency was slight and only 3.4% from Pattern 1 to Pattern 8 (62 Hz to 59.9 Hz). In addition, the reduction was not monotonic from Pattern 1 to 8 due to the measurement error associated in the vibration test. Finite Element Simulation To further study the influence of the simulated beam fracture, a finite element (FE) model of the Freq. reduction (%) 4.0 3.0 2.0 1.0 0.0-1.0 Test Simulation 1 2 3 4 5 6 7 8 Patterns
specimen was constructed using a commercial structural analysis code SAP2000 (Figure 5(a)). The beams and column were modeled using a shell element, and the slab was modeled by a solid element. The slab was connected rigidly to the top flange of the steel beams. The boundary conditions of the FE model was pin-supported at the column ends and springsupported at the beam-ends. The spring was modeled to account friction forces in the large structural pins at the ends of the rollers (Figure 1). Note that the structural pins were specifically designed to resist large forces in quasi-static tests and thus allow friction force resistance against rotation. The property of the spring (with the lateral stiffness of 80 kn/mm) was tuned to achieved the modal frequency of the FE model close to the peak frequency of the undamaged specimen identified in the test (Pattern 1). In the FE model, the fracture was simply modeled as the separation of the nodes at beam ends and the column surfaces. Figure 5(b) shows the mode shape of the specimen that corresponds to the mode frequency of interest. Note that the column bending shape was not asymmetric due to the location of the pin-supports; one was attached right under the column and the other was attached the side of the column top where the hydraulic jack was attached. The FE simulation showed the changes of the modal frequencies similar to the test (Figure 3(a)). From Case 1 to Case 8, the modal frequencies decreased only by 2%. The error between the simulation and tests were large but both showed very slight reduction in the frequency. Shell Solid Figure 5. (a) (b) Finite element simulation: (a) model; (b) mode shape. Analytical Study for Influence of Damage As observed in the test and simulation, the crack of the slab affected the dynamic characteristics of the connection significantly but the influence of the fracture on the steel beam was relatively minor. To provide rational reasoning to this finding, a simple analytical study was carried out. Formulation of Connection Rotational Stiffness Given the boundary conditions in the test, which restricted the lateral movement of the column top and perturbed the lateral movement of the beam ends with the friction resistance in the structural pins, the specimen mainly vibrated with the rotational movement around the joint of the connection. Thus, in this section, the rotational stiffness of the connection is formulated. The rotational stiffness of the connection is assumed as the sum of the bending stiffness of the left and right beams, and top and bottom columns, each of which is modeled as a simply supported beam or column subjected to rotation at one end. Figure 6(a) shows the analytical model for one beam. One end of the beam which is far from the joint of the connection is
supported by a pin, while the other end is represented with a rotational constraint. For modeling different section properties along length, the beam is divided to three segments, A, B and C. The sections of Segment A and B are composed of floor slab and H-shaped steel beam. Segment A is to simulate the bottom flange fracture at the beam-end. Segment B is to simulate the slab damage. Segment C represents the steel beam section without slab. The bending stiffness of the beam is calculated as, K b =M b /θ b (1) where θ b is the rotation angle of the rotational constraint of the beam when a moment of M b is assigned to the rotational constraint. The rotation angle, θ b, is derived by the principle of virtual work as, b * * * ( ) ( ) ( ) ( ) ( ) ( ) M x M x M x M x M x M x θ = + + b b b b b b dx dx dx EI A A A EI (2) B B B EI C C C where x is the coordinate along the beam; M b * is the bending moment of the beam when a unity bending moment is assigned at the rotational restraint; E i and I i (i=a, B, and C) are the equivalent Young s modulus and the second moment inertia of different sections. The rotational stiffness of the column K c is derived by the same method for the beam but with a single section property along their length. Assuming the left and right beams are subjected to the same level of damage, the rotational stiffness of the beam-column connection, K total, is computed as follows, K totoal =2K b +2K c (3) Beam Damage Cases Besides undamaged case, UD, and four different damage cases, D1 to D4 (Figure 6). UD is used as the baseline to compare with other cases. D1 and D2 represent the damage patterns observed in the quasi-static cyclic loading test. The floor slab in D1 and D2 is damaged for the entire length of Segment A and B. The left and right beams in D1 sustain no damage. The beams in D2 sustain the fracture of the entire bottom flange in Segment A with a short length of 10 mm. The length of the fractured section is determined considering the length of the beam section affected by the fracture, whose opening was nearly equal to zero in the quasi-static loading test. D3 and D4 are unrealistic but considered to illustrate the influence of the damage length to the rotational stiffness. In D3, the floor slab is removed for Segment A. In D4, the entire length of the beam from the segments A to C, sustain damage for the slab and the bottom flange of the steel beam. In the calculation, the floor slab damage is simulated by removing the full concrete section, while the beam damage in the damaged segment is simulated by reducing the Young s modulus to 10E-6 of the original value.
(a) θ b A B C E A I A E B I B E C I C (b) Slab completely damaged (full length) (d) Slab damaged (length 10 mm) (c) Slab completely damaged (full length) (e) Slab completely damaged (full length) Bottom flange damaged (length 10 mm) Bottom flange damaged (full length) Figure 6. Beam model. (a) UD: no damage; (b) D1: slab damage for full length; (c) D2: slab damage for full length and bottom flange fracture for short length; (d) D3: slab damage for short length; (e) D4: slab damage and bottom flange fracture for full length. Figure 7. Rotational stiffness of connection. Table 1. Stiffness summary (unit: MN m/rad). Slab (full section) Beam (bottom flange) K b K c K total UD No damage No damage 116 219 671 D1 Full length No damage 50.4 219 539 D2 Full length Partial length (10mm) 49.6 219 537 D3 Partial length (10 mm) No damage 115 219 668 D4 Full length Full length 19.0 219 476 Analysis Results Normalized stiffness of connection 1.0 0.8 0.6 0.4 0.2 0 UD D1 D2 D3 D4 Figure 7 shows the rotational stiffness of the connection relative to the undamaged case, UD, and Table 1 summarizes the rotation stiffness of the beam, column and connection for the five cases. When floor slab is damaged in D1, the reduction of the rotational stiffness of the connection is 20%. Compared to that, the reduction from D1 to D2 caused by beam fracture is very small and less than 0.5%. This result is consistent with the findings from the forced-vibration tests. It is trivial but the slab damage with a short length does not affect the rotational stiffness significantly in D3 and the bottom flange fracture with full length significantly affect the rotational stiffness in D4.
The analyses clearly indicate the importance of damage length for properly simulating the influences of floor slab damage and steel beam damage. While this conclusion cannot be simply extended to the dynamic characteristics of beam-column connections with other boundary conditions, e.g. roller-supports at the column top and beam-ends to consider lateral deformation, simulation of damage in steel buildings should not underestimate the influence of slab damage and overestimate the influence of beam fracture with unrealistic assumption in damage length. For instance, an improper assumption of beam fracture for the entire beam length may lead to the overestimation of the performance of system identification and damage detection algorithms for steel buildings. Conclusions The influence of seismic damage on the dynamic characteristics of steel beam-column connections was examined through experimental, numerical and analytical studies. The significant findings of the studies are listed below: 1. When the full-scale specimen of a beam-column connection with floor slab is dynamically excited without the lateral movement at its column top, slab damage reduced the peak frequency by 10 to 15% while the formation of large fracture in the bottom flange and web of one beam-end reduced the peak frequency only by 3%. 2. Supplemental simulated beam fracture tests also showed the small influence of beam fracture on the peak frequency of the connection. The reductions of the peak frequency by the fracture at the entire bottom flange and 2/3rd web at the end of two beams were 3% in the test and 2% in the analysis. 3. An analytical study simulating slab and steel beam damage in a beam-column connection confirmed the importance of damage length for properly evaluating the influence of damage on the vibration frequency. 4. While the finding of this research cannot be simply extended to the dynamic characteristics of a steel beam-column connection with other boundary condition, the authors believe that the simulation of damage in steel buildings should not underestimate the influence of slab damage and overestimate the influence of beam fracture with unrealistic assumption in damage length. References 1. Motosaka, M. Lessons of the 2011 great east japan earthquake focused on characteristics of ground motions and building damage. Proceedings of the International Symposium on Engineering Lessons Learned from the 2011 Great East Japan Earthquake March 1-4, 2012, Tokyo, Japan. 2. Ji X., Fenves G., Kajiwara K., Nakashima M. Seismic damage detection of a full-scale shaking table test structure. Journal of Structural Engineering 2011; 137(6): 14-21. 3. FEMA-355C. State of the art report on systems performance of steel moment frames subject to earthquake ground shaking, FEMA, September, 2000. 4. Architectural Institute of Japan (AIJ). Recommendation for Design of Connections in Steel Structures, AIJ, Japan, 2012. 5. Architectural Institute of Japan (AIJ). Design Recommendations for Composite Constructions, AIJ, Japan, 2012.