Automated Assessment of Tall Building Wind-Induced Response Data to Support Long-Term Monitoring Programs

Transcription

1 The 12th Americas Conference on Wind Engineering (12ACWE) Seattle, Washington, USA, June 16-20, 2013 Automated Assessment of Tall Building Wind-Induced Response Data to Support Long-Term Monitoring Programs Tara Weigand a, Tracy Kijewski-Correa a a University of Notre Dame, Notre Dame, IN USA ABSTRACT: While there is incredible value in long-term monitoring programs for tall buildings under wind, these programs generate large volumes of data that can be difficult to mine effectively to generate knowledge that improves the design process and delivers information of value to the owner. As such the authors have begun to develop a suite of automated tools to manage and more effectively mine these large stores of data. This paper presents one such automated framework for parameterizing response data by waveform for the purposes of conducting an occupant comfort assessment. The framework can project, at multiple levels and for multiple modes of an instrumented building, the likely number of occupants who may experience nausea or task disruption based on the data collected in actual motion simulator experiments. A case study is provided to demonstrate the insights that can be obtained from such a tool. KEYWORDS: Occupant comfort, tall buildings, full-scale monitoring, habitability 1 INTRODUCTION Tall buildings represent a substantial investment with high concentrations of life and property, and value that far exceeds traditional buildings. Despite their significance, their design is entirely reliant on scaled model testing and analytical procedures that rarely are systematically validated in full-scale. Recently, full-scale monitoring programs have been attempting to address this shortcoming and have provided valuable feedback to validate design practices [1-5]. However, in order to achieve true systematic validation under a range of wind events, long-term monitoring is essential. In many cases it may take years to observe a wind event originating from an angle other than the prevailing wind direction at the location or to reach the recurrence interval of interest. Over the course of such extended monitoring, substantial quantities of data are generated and, while valuable, their effective mining can prove challenging. Delivering approaches to process such data in an automated fashion is challenging given that many techniques do not lend themselves well to unattended operation. Based on the experiences in the Chicago Full Scale Monitoring Program (CFSMP), which has generated over a decade s worth of data, it is clear that while long term monitoring is valuable, it requires skill to manage and extract critical information from large datasets that are generated in the process. Moreover, as owner-driven solutions for longterm monitoring are now in demand, the ability to deliver information to the owner automatically from measured response data is especially important. For example, the second author s work to develop the SmartSync platform for tall building monitoring, included the development of web interfaces that allowed ready access to key building information, including response statistics, time histories and power spectra, and even extracted frequency and damping values all in near real time [6]. Whether to meet the needs of an owner, an engineer or simply a research team, the means to automatically process and extract unique insights from monitored data is increasingly important. This paper presents one such automated framework that has been developed by the authors to translate data into knowledge to support decision-making in an automated manner: in

2 this case to evaluate the comfort of occupants under wind using in-situ acceleration measurements. 2 AUTOMATED OCCUPANT COMFORT ASSESSMENT Particularly for habitability assessment, which involves delicate matters like occupant perception and comfort, the ability to reliably assess performance without attracting unwanted attention from tenants is critical. For this reason, a framework for pseudo-full scale assessment of occupant comfort was first proposed by Kijewski-Correa and Pirnia [7]. In this approach, full-scale accelerations are mapped to human comfort thresholds derived from extensive motion simulator work [8]. This framework allows the measured accelerations at a given floor of a building to be translated to what an owner desires most: the likely number of tenants that would have been affected, i.e., would experience sensations such as nausea or task disruption. Through this approach, the complex human-structure interaction that dictates whether habitability performance is acceptable can be accounted for in a reliable manner, without directly interviewing or engaging the tenants themselves. Bentz [9] extended this framework to account for the role of torsional response and to project these occupant comfort assessments over the entire building using the projected mode shapes to assess the total number of potentially affected tenants, accounting for occupancy rates dependent on the time of day. In these previous studies, the projected rates were derived solely from the response of the fundamental mode of the structure and relied only on instrumentation at one level, approximating the responses at other floors using the project mode shapes. Moreover, these studies focused only on the fundamental mode. However, in the case of super tall buildings with exceptionally low frequencies, higher modes often participate significantly in the response. Given that motions are perceptible at even lower accelerations for higher frequencies of oscillation, the potential for habitability issues in higher modes becomes increasingly likely. As such, this study will extend the framework to a multimode and multi-level evaluation suitable for automation to support decision making. 3 MULTI-MODAL WAVEFORM CLASSIFICATION The process begins by classifying the waveforms present within the measured acceleration time histories. The process of waveform classification, in the case of occupant comfort, is necessary given the fact that motion simulator investigations have confirmed that Gaussian motions prove most disruptive to occupants over specific durations [10]: these durations are 12 minutes for task disruption and 50 minutes for nausea. With these critical durations and targeted waveform type identified, a procedure to extract them from a recorded time history can be initiated. This procedure is depicted in Figure 1. A bandpass filter bank is applied to the acceleration time history at any instrumented level to isolate each mode. Once isolated, the RMS values and peak factors are extracted from each modal response time history using a moving analysis window of 12 minutes in length. To offset the correlation between windowed observations, the moving analysis window is shifted by ten times the sampling frequency. Response windows with less than 10% of their data points in common are assumed uncorrelated. The uncorrelated segments are then grouped by peak factor and sorted by their RMS accelerations, from highest to lowest. This same procedure is repeated for each mode and the entire process is repeated for the 50 minute moving windows.

3 The 12th Americas Conference on Wind Engineering (12ACWE) Seattle, Washington, USA, June 16-20, 2013 Sample Time History Apply independent filters to isolate each mode Sample Response Window Extract Peaks Peak Factor Peak Factor = 2.7 milli-g Figure 1. Example of mode isolation for a given time history and subsequent peak factor estimation for a sample response window.

4 The results from this procedure then undergo waveform classification whereby segments with peak factors in the ranges , , and greater than 4.05 are classified as sinusoidal, Gaussian, and burst waveforms, respectively, according to the convention proposed by Pirnia [11] based again on the motion simulator experiments of Burton [10]. Once the waveforms are classified, a response waveform composition chart is generated for each instrumented level. This chart demonstrates the waveform composition in each mode. One of these waveform composition charts is shown in Figure 2 for one of the CFSMP building s x-sway response under a stationary wind event. Independent of an occupant comfort analysis, this chart demonstrates which modes experience strong lock-in (dominantly sinusoidal response) under the action of wind. Figure 2. Waveform classification chart for x-sway modes using 12 minute moving windows 4 OCCUPANT COMFORT VISUALIZATION After each recorded time history is broken down into its waveform response classifications, an occupant comfort analysis can easily be performed by relating the observed acceleration levels to the occupant disruption rates from the motion simulator work of Burton et al. [10]. For each mode isolated within the time history, the responses classified as Gaussian, according to their peak factors, are sorted into minimum RMS acceleration levels based on the highest occupant perception threshold they exceed. These thresholds are frequency dependent and thus defined uniquely for each mode. The number of times this minimum RMS threshold is reached over the course of the event is recorded. For each threshold at a given frequency of oscillation, the num-

5 The 12th Americas Conference on Wind Engineering (12ACWE) Seattle, Washington, USA, June 16-20, 2013 ber of human respondents adversely affected in experimental motion simulator experiments is reported. For each mode, the number of times the response exceeds these predefined thresholds indicates the number of instances that reported percentage of occupants would be affected during the event. An example of such a report that would be generated for a client is shown in Figure 3 at various instrumented levels of one of the CFSMP buildings. Here the x-axis corresponds to the number of occurrences and the y-axis corresponds to the magnitude of the RMS acceleration thresholds triggering task disruption. The higher the RMS acceleration level, the greater number of occupants that may be disrupted. During this case study event, there are several instances of task disruption (identified with the 12 minute moving window) at instrumented Level 6. Six different RMS acceleration thresholds were surpassed in mode 8, which corresponds to six instances of task disruption. At Level 5, the RMS accelerations crossed the minimum RMS threshold three distinct times in mode 4, which translates to 3 instances of task disruption affecting a projected 15% of occupants each time. The lower levels all remained below the minimum RMS acceleration threshold, which indicates that no occupants would likely be affected at these levels. It is important to note that these rates are indicative of tolerance levels, not perception levels. Though the occupants are not projected to be disrupted by this event, they may still perceive the motion of the building. The results yielded by this automated framework can also provide a visualization of the worstcase scenarios as illustrated in Figure 4. This plot shows the largest percentage of occupants affected by task disruption (instances captured by 12 minute moving window) or nausea (instances captured by 50 minute moving window) at any point during the event. The results remain separated by mode in order to provide valuable insights into which modes most significantly influence occupant comfort at a given level of the building. During the event in this case study, mode eight clearly dominates the tolerance limits with task disruption rates of 75% and nausea rates of 52%. This demonstrates the importance of doing a multi-modal analysis for tall buildings rather than focusing solely on the fundamental mode. In this example, the occupied floors show no susceptibility to occupant discomfort, which provides important information the owner in evaluating the performance of the building: in this case, performance was completely satisfactory from the perspective of occupant comfort. 5 CONCLUSIONS This paper presented an automated framework for identifying waveform responses in multiple modes at various instrumented levels of a tall building and then related these waveforms, as a function of their frequency and amplitude, to observed rates of task disruption and nausea from motion simulator studies. This serves as an important decision support tool for determining whether a building meets habitability standards, pinpointing which levels and even which modes are contributing to potential disruption of occupants. This tool can be enhanced through the addition of a population density map to determine the actual number of people affected per level, thus providing a more salient representation of occupant comfort throughout the building. The automated waveform characterization framework, independent of the human comfort assessment tool, has proven furthermore to be useful as a feature extraction engine used in conjunction with other data mining techniques operating on the large databases collected in long-term monitoring programs to seek underlying patterns in the data. This represents but one type of automated analysis that has been developed by the authors for mining and decision support in their monitoring programs, flanking other tools created for response assessment and dynamic property extraction, all automated through secure web interfaces.

6 Level 6 Mode 4 Mode 6 Mode 8 Level 5 Level 4 Level 3 Level 2 Level 1 Figure 3. Example of instances of task disruption and percentage of occupants affected per instance.

7 The 12th Americas Conference on Wind Engineering (12ACWE) Seattle, Washington, USA, June 16-20, 2013 Figure 4. Maximum percentages of occupants with task disruption or nausea at any one instance during the event. 6 ACKNOWLEDGEMENTS The authors gratefully acknowledge the support of the National Science Foundation (NSF) through Grants CMS and CMS that founded and expanded the Chicago Full- Scale Monitoring Program. Additional financial support from the Chicago Committee on High Rise Buildings is also humbly acknowledged. There are also a number of collaborators who, over the last decade, have contributed significantly to these efforts. These include those at Skidmore Owings and Merrill LLP in Chicago, most notably William Baker and Bradley Young, our colleagues at the Boundary Layer Wind Tunnel Laboratory at the University of Western Ontario, led by Dr. Nicholas Isyumov, past students at the University of Notre Dame in both the NatHaz

8 and DYNAMO Labs who helped to process and curate the data generated by this project, and research assistant professor Dr. Dae Kun Kwon, also at the University of Notre Dame, who has been instrumental in maintaining this network for over a decade. Finally, none of this work would be possible without the support, enthusiasm and cooperation of the building owners and management, particularly in the building engineering and rooftop operations divisions. 7 REFERENCES 1 Q.S. Li, Y.Q. Xiao, C.K. Wong, Full-scale monitoring of typhoon effects on super tall buildings, Journal of Fluids and Structures, 20 (2005) Q.S. Li, K. Yang, C.K. Wong, A.P. Jeary, The effect of amplitude-dependent damping on wind-induced vibrations of a super tall building, Journal of Wind Engineering and Industrial Aerodynamics, 91 (2003) Q.S. Li, J.R. Wu, S.G. Liang, Y.Q. Xiao, C.K. Wong, Full-scale measurements and numerical evaluation of windinduced vibration of a 63-story reinforced concrete tall building, Engineering Structures, 26 (2004) J.M.W. Brownjohn, Lateral loading and response for a tall building in the non-seismic doldrums, Engineering Structures, 27 (2005) T. Kijewski-Correa, J. Kilpatrick, A. Kareem, D.K. Kwon, R. Bashor, M. Kochly, et al, Validating the windinduced response of tall buildings: A synopsis of the Chicago full-scale monitoring program, Journal of Structural Engineering, 132 (2006) T. Kijewski-Correa, D.K. Kwon, A. Kareem, A. Bentz, Y. Guo, S. Bobby, SmartSync: An integrated real-time structural health monitoring and structural identification system for tall buildings, Journal of Structural Engineering, Special Issue on State-of-the-Art in Structural Identification, In press (2013). 7 T. Kijewski-Correa, D. Pirnia, Pseudo-full-scale evaluation of occupant comfort in tall buildings, In 11th Americas Conference on Wind Engineering, San Juan, Puerto Rico, M. Burton, P. Hitchcock, K. Kwok, R. Roberts,, Acceptability curves derived from motion simulator investigations and previous experience with building motion, In Proceedings of the 10th Americas Conference on Wind Engineering, Baton Rouge, LA, A. Bentz, Dynamics of tall buildings: full-scale quantification and impacts on occupant comfort, PhD, University of Notre Dame, M. Burton, Effects of Low-Frequency Wind Induced Building Motion on Occupant Comfort, PhD, The Hong Kong University of Science and Technology, J.D. Pirnia, Full-Scale Dynamic Characteristics of Tall Buildings and Impacts on Occupant Comfort, MSCE, University of Notre Dame, S. Lagomarsino, Forecast models for damping and vibration periods of buildings, Journal of Wind Engineering and Industrial Aerodynamics, 48 (1993)