An E-Defense shake-table test of a 3-story reinforced concrete (RC) disaster management center was performed in December 2019 as part of the Tokyo Metropolitan Resilience Project Subproject C [1]. An organized session for this study will be held at the 17 th World Conference on Earthquake Engineering, and will cover various topics ranging from project background, test details, performance assessment, structural health monitoring, and results from a blind prediction competition. This paper serves as a complementary source of information for interested conference participants who wish to learn more about the project.
The Tokyo Metropolitan area, while a major global economic center and home to over 35 million people, is a seismically active region. To reduce socioeconomic impacts from major seismic events, the "Tokyo Metropolitan Resilience Project" was initiated in 2017 and consisted of three subgroups; (i) social-sciences (subproject A), (ii) natural sciences (subproject B), and (iii) engineering (subproject C).
One objective of subproject C was the development of rapid damage assessment systems to avoid unnecessarily displacing people from safe buildings or to quickly identify dangerous buildings. Furthermore, such systems may be able to detect hidden damage, which was one of many reasons for the CTV Building collapse in New Zealand [2]. Another objective was to evaluate the effectiveness of resilient structural and non-structural detailing for buildings with post-disaster functions due to poor performance of such buildings in recent events (e.g. city hall buildings [3] and school gymnasiums intended to be used as emergency shelters [4]). There are other objectives in subproject C which have been/will be addressed by other research studies.
A shake-table test of a disaster management building was performed at E-Defense to address the two objectives discussed. Following requirements for Japanese buildings with post-disaster functions [5], the specimen must have a base shear coefficient greater than 0.55 and remain elastic with interstory drifts less than 0.33% at design-level shaking. The building must also meet life-safety requirements at 1.5 times designlevel shaking. To achieve these objectives, spandrel walls were casted to be monolithic with RC frame elements to provide additional stiffness and strength, with special detailing provided to ensure a favorable deformation mode. The building was fitted with non-structural elements, and various instrumentation and scanning equipment were used to assess the building's performance. The key objectives of this test were to: i)
Evaluate the effectiveness of the spandrel wall detailing to satisfy performance objectives for buildings with post-disaster functions.
ii) Evaluate the effectiveness of more resilience non-structural element detailing.
iii) Test the reliability and robustness of various structural health monitoring techniques.
The RC specimen layout and dimensions are shown in Fig. 1. Spandrel wall elements were casted to be monolithic with the frame elements with 50 mm wide gaps present at hanging/standing wall ends on 2F and 3F (Fig. 2a) to ensure that plastic hinges form at intended locations and to prevent bar buckling failure. Similar detail was used for RF hanging-walls and 1F wing-wall bases, except that the gaps were filled with concrete to allow the walls to act in compression. The gap detailing had been tested previously [6,7], while the concrete filled-gap detailing was new. Reinforcing details of frame elements parallel to the loading direction are shown in Fig. 3. Note that reinforcing in the RF hanging-wall and 1F wing-wall base was terminated 50 mm before its ends, and that the lower reinforcing layer for B1 beams was also terminated before intersecting G2 and G3 beams. Details for other frame elements are available from Fukai et al [8].
The 17th World Conference on Earthquake Engineering 2i-0020
The 17th World Conference on Earthquake Engineering
Four types of non-structural components were attached to the building as follows:
i) Windows on south-west bay on 2F and 3F (Fig. 1e);
ii) Ceilings on west bay on 2F and 3F (Fig. 1e);
iii) Exterior tiles on south-east bay over entire height (Fig. 1e); and iv) Piping on roof.
Two types of windows were installed in the building; fixed and sliding. The 2F fixed window used two panes of glass (Fig. 4a) while the 3F fixed window used a single pane (Fig. 4b). As the window sash used for both fixed windows were identical, the edge clearance is larger for the 2F fixed window and thus the drift capacity of the fixed window with two glass panes should be greater. The sliding windows were identical on both floors to observe the influence of drift demands on the seismic performance of sliding windows.
Two detailing variants were used for each non-structural component type; one being more seismically resilient than the other. With regards to ceilings, the one located on 2F (attached to 3F slabs) was fixed to the surrounding beams (Fig. 5a) without seismic bracing (Fig. 5b). The ceiling on 3F (attached to roof slab) had gaps around its edge (Fig. 5c) and was seismically braced (Fig. 5d). The latter ceiling was thought to be more seismically resistant and thus was installed where the acceleration demand was expected to be greater.
Two different materials were used to attach tiles to the south-east exterior bay; (a) mortar and (b) chemical adhesive. The mortar was applied over the western half of the bay, while the chemical adhesive was applied over the eastern half. The mortar layer was approximately 10 mm thicker, and thus a discontinuity of the tile surface could be observed as shown in Fig. 6. The chemical adhesive was more flexible than the mortar layer, and therefore the exterior tiles attached using the chemical adhesive were expected to remain attached to the frame at higher drift demands.
Two pipe systems in L-shape configurations were attached to the roof as shown in Fig. 7a. The pipes were weighed down using concrete supports instead of being bolted down as is common practice in Japan to avoid damaging roof weather proofing material. These supports had two or four-legs as shown in Fig. 7b, the latter of which was expected to be more stable and hence have lesser displacement.
Different types of accelerometers, such as those shown in Fig. 8a, were installed on all floors. This was to evaluate (i) the accuracy of cheaper sensors compared to more expensive variants and (ii) the global building performance following an approach proposed by Kusunoki et al. [9].
Laser transducers (Fig. 8b) were installed at three locations on each floor to measure interstory drift; one next to each exterior frame at the mid-width of the east-side bay and a third near the center of each floor. The transducer and the target were attached to steel H sections bolted to the bottom and top slab, respectively. The center location had two transducers in parallel to capture rotation effects.
Displacement potentiometers (Fig. 8c) were attached to the exterior of the northeast bay at column bases and joints between beams/hanging walls/standing walls and wing walls. These were also used on windows to capture window drift.
The specimen was subjected to an artificial record five times. The artificial record with a 1.0 scale factor was representative of the Japanese Building Code design spectra for ordinary buildings. The input record acceleration history and response spectra for a scale factor of 1.0 are shown in Fig. 9. The scale factors adopted for each excitation were as follows:
i) 0.2 (for confirming serviceability performance objectives were satisfied for frequent shaking events);
ii) 1.0 (for confirming that the building could remain mostly elastic with a peak inter-story drift of less than 0.33% at design-level shaking);
iii) 1.5 run #1 (for evaluating the building's performance at 1.5 timed design-level shaking); iv) 1.5 run #2 (this was repeated to evaluate if the building could withstand an aftershock of equal intensity to its design demand as required by the Japanese Building Code); and v) 1.6 (for observing the building's performance in its fully inelastic range and its capability to withstanding multiple significant seismic events).
White noise excitations were applied before and after each of the five major excitations. These were used to track changes in the building's dynamic properties resulting from each shaking event.
The 17th World Conference on Earthquake Engineering
This section covers initial damage observations and evaluations. No detailed results were discussed as these were still being processed at the time of writing. Building damage surveys were performed (i) before testing, (ii) after the 1.0-scaled excitation, (iii) after the first 1.5-scaled excitation, and (iv) after the 1.6-scaled excitation. After (ii) and (iii), the building damage was evaluated following the "Guideline for Post-Earthquake Damage Evaluation and Rehabilitation of RC Buildings" [10].
Before testing, initial cracks were observed at the beam-wing wall joints adjacent to the outer (C2) columns. These cracks were small (0.2 mm or less in the beams) and were thought to not significantly influence building strength. No damage was observed in columns or slabs.
Following the 1.0-scaled excitation, more cracks were identified at each beam end, though damage was concentrated on a single crack. Minor insignificant cracking had formed on columns and slabs, the latter of which near wall gaps and propagated towards the center of the slab. Based on the damage evaluation procedure [10], the damage was minor on 1F and 2F. The damage on 3F was considered moderate if the hanging-wall crack width was considered. However, this occurred where there were no reinforcing and thus had no influence on strength. If the beam residual crack was considered, the 3F damage would be slight.
Following the first 1.5-scaled excitation, only a few new cracks were observed. However, the largest residual beam cracks widened to 2.0 mm. Horizontal cracks with residual width of 0.3 mm formed at column bases and some crushing occurred at wing-wall corners. Slab cracks intersected, resulting in large cracks spanning the slab width residual widths of 3.5 mm. The damage was evaluated as "significant" on all floors. Photos of final observed damage is shown in Fig. 10. Significant concrete spalling occurred on 1F and 2F beams (Fig. 10a) resulting in reinforcing bars being exposed. While concrete spalling also occurred on 3F (Fig. 10b), the extent was not as significant, and no reinforcing was not exposed. This showed the advantage of the concrete-filled-gap detailing which lessened the compression strains at the soffit of the beam. Significant concrete crushing occurred at the base of the wing-walls (Fig. 10c), and small diagonal column cracks started forming. The slab cracks widened to over 15.0 mm and diagonal cracks formed due to secondary compression struts developing due to a change in the force transfer mechanism. Due to the extensive damage observed, the damage was classified as "significant" and was deemed close to collapse.
Several structural health monitoring approaches were applied to the building to evaluate its performance and safety. One such approach was the use of accelerometer data following Kusunoki [9], and another was using laser scanning and motion tracking technology to capture damage and track building response. Other approaches were also utilized, but only these were described herein.
The approach by Kusunoki [9] involved (i) double integrating accelerometer data to obtain relative displacement response, (ii) simplifying the response into an equivalent single-degree-of-freedom response 2i-0020
The 17th World Conference on Earthquake Engineering using modal analysis methods considering the displaced shape at each time-step, (iii) applying the Wavelet Transform Method and selecting ranks representing the predominant response, and (iv) obtaining the resultant "skeleton curve" from the representative acceleration-displacement response. The skeleton curve obtained from preliminary application of this approach using the experimental data is shown in Fig. 11.
The skeleton curves are useful in several ways. Firstly, the change in the "initial" period could be estimated based on the change of the initial slope with each excitation. Secondly, the overall building ductility demand could be estimated based on the ratio of peak representative displacement to estimated yield displacement. Thirdly, if a "safety limit" is known, one could assess how close the building is to reaching this safety limit, essentially providing an alternative to the damage evaluation procedure [10] followed previously. One important advantage is that this system can be applied in real-time, meaning that the safety of the building could be evaluated almost immediately. This can reduce the time and effort required to evaluate a building's level of safety. Fig. 11 -Derived capacity curves using WTM
Two separate research teams were involved with using laser scanning and motion tracking technology; (i) the University of Washington (UW) and University of Nevada, Reno (UNR) team, and (ii) the Building Research Institute (BRI) from Tsukuba, Japan.
The UW/UNR team were funded by the US National Science Foundation (NSF), and collaborated in the project by collecting response and damage data using multiple non-contact image and lidar based systems. UW researchers employed equipment provided by the NSF-funded Natural Hazards Engineering Research Infrastructure (NHERI) "RAPID" Facility. This included two portable lidar scanners, each with integrated imaging but with different lidar scan resolutions and scan times (Leica ScanStation P50 and Leica RTC360), and a surveying total station (Leica Nova TS161). Data collection activities were performed while the structure was stationary (i.e. prior to testing and at the end of each test day) and included: i)
Use of the total station to locate, with a high level of accuracy, approximately 30 high-contrast targets located on the test specimen, the shake table, and the laboratory structure to provide information about residual deformation of the test specimen;
ii) Use of the Leica RTC 360 scanner to generate a modestly accurate/dense 3D point cloud encompassing the interior and exterior of the specimen with image overlay (Fig. 12a) to qualitatively assess the structural performance and support use of localized, high-resolution point cloud data; and, iii) Use of the Leica P50 to generate high accuracy/density point clouds in regions with significant damage (Fig. 12b) to evaluate the potential for lidar to be used to identify cracks in concrete and masonry structures as well as to determine crack length, width and depth.
The 17th World Conference on Earthquake Engineering The UW/UNR team also collected dynamic response data during shake table tests, under both white noise and seismic excitation. The UW team employed the Leica P50 scanner in "line-scan" mode to track the east-west (EW) location of the vertical centerline of the northwest column to generate EW displacement, velocity, and acceleration histories for the column. The UNR team employed two cameras sets; (i) two stateof-the-art monochrome ultra high-speed Fastec cameras and (ii) consumer-grade DSLR Canon cameras (Fig. The 17th World Conference on Earthquake Engineering 12c). These cameras faced the south side of the building on which approximately 40 high-contrast targets were placed as shown in Fig. 12c. Software produced by GOM [11] will be used to perform target-tracking digital image correlation (TT-DIC) of data to generate displacement time-histories for the high-contrast targets. One challenge was dealing with low light levels which were required to avoid potentially interfering with laser sensor readings employed by other research teams. This will require pre-processing of image data to ensure accurate calculation of camera and target locations. Two types of analyses will be conducted using TT-DIC data; (i) determining building frequency and mode shapes using acceleration readings, and (ii) obtaining peak and residual drifts from each excitation. The response and damage data obtained by the UW/UNR team will be compared with similar data collected by other research teams.
The BRI team also conducted similar scans using their own set of equipment. Description of past deployment of such equipment and examples demonstrating the usefulness of acquired data are available in literature [12]. Some initial outputs from their data collection in this project can be seen in Figs. 12d-12f and included (i) evaluating the residual deformation of the building, (ii) identifying exterior damage, such as spalling of tiles, and (iii) tracking the dynamic specimen response.
In addition to the E-Defense test, a blind prediction competition was held to evaluate the accuracy and reliability of numerical modelling approaches and assumptions. Participants were requested to predict the following building response parameters for the 1.0 and 1.5 (run 1) scaled excitation cases: i)
Roof displacement history (relative to base of building);
ii) Peak absolute interstory drift on all floors; and iii) Peak base shear demand.
The blind prediction competition was still running at the time of writing, and thus only the judging criteria was discussed here. Contestants will be ranked for each category based on the calculated error, and the three contestants with the best average rank will be deemed competition winners. If a top-three could not be clearly identified due to multiple teams having the same average rank, then the ranking in the roof displacement history category will be used as a tiebreaker as this is thought to be the hardest to predict. The error for roof displacement history were determined using sum-squared-error, SSE RDH , using the following equation. Where x predicted,i and x measured,i are the predicted and measured roof-displacements at step i, respectively, and n is the number of steps.
Accuracy of peak values of predicted and measured values (R max,predicted,j and R max,measured,j , respectively) were also assessed using sum-square-error considering each floor (denoted by j), though the natural log values were used instead. Base shear was also evaluated using log values, though as only one value was considered for each excitation, the sum-squared-error calculation was not required.
This paper summarized various aspects of a shake-table test of disaster management center performed at E-Defense in December 2019 and serves as a companion paper to an organized session to be held at the 17 th World Conference on Earthquake Engineering. Based on building response and initial damage evaluations, 2i-0020
The 17th World Conference on Earthquake Engineering the building was able to satisfy design objectives from a performance perspective (i.e. minor damage at design-level shaking and ability to survive multiple significant events without collapsing), though the damage at 1.5-times design-level shaking was classified as "significant". Application of various structural health monitoring approaches, such as deriving capacity curves using accelerometer data or using laser scanning and motion tracking technology, demonstrated its effectiveness in rapidly assessing a building's performance and damage.
© The 17th World Conference on Earthquake Engineering -2i-0020 -