Data compression of very large-scale structural seismic and typhoon responses by low-rank representation with matrix reshape Yongchao Yang 1 , Satish Nagarajaiah 1,2, * ,† and Yi-Qing Ni 3 1 Department of Civil and Environmental Engineering, Rice University, Houston, TX 77005, USA 2 Department of Mechanical Engineering, Rice University, Houston, TX 77005, USA 3 Department of Civil and Environmental Engineering, The Hong Kong Polytechnic University, Hong Kong SUMMARY The intrinsic low-dimensional structure, which is implicit in the large-scale data sets of structural seismic and ty- phoon responses, is exploited for efficient data compression. Such a low-dimensional structure, empirically, stems from few modes that are active in the structural dynamic responses. Originally, limited to the sensor and time- history dimension, the structural seismic and typhoon response data set generally does not have an explicit low- rank representation (e.g., by singular value decomposition or principal component analysis), which is critical in multi-channel data compression. By the proposed matrix reshape scheme, the low-rank structure of the large- scale data set stands out, regardless of the original data dimension. Examples demonstrate that the developed method can significantly compress the large-scale structural seismic and typhoon response data sets, which were recorded by the structural health monitoring system of the super high-rise Canton Tower. Copyright © 2015 John Wiley & Sons, Ltd. Received 19 May 2014; Revised 16 October 2014; Accepted 19 January 2015 KEY WORDS: Matrix reshaping; Very large scale data; data compression; Low rank representation; Seismic re- sponse; Typhoon response; Large scale SHM 1. INTRODUCTION Seismic monitoring systems have been widely instrumented in civil structures to record the strong ground motions and structural responses, with the objective of continuously monitoring and assessing structural behaviors and performance during earthquakes, as well as for rapid post-earthquake assess- ment and long-term analysis of structures [1–7]. Currently, for example, the California Strong Motion Instrumentation Program (CSMIP 2013) [8] has embedded in California advanced earthquake monitor- ing devices in more than 600 ground motion stations and 200 infrastructures including buildings, dams, bridges, and so forth. Recently, many landmark suspension bridges and high-rise buildings have been equipped with structural health monitoring (SHM) systems with networks of sensors for continuous monitoring of structural ambient vibration and during extreme events, such as earthquake, typhoon. For example, the Tsing Ma Bridge (1997) in Hong Kong, the Canton Tower (2010) in Guangzhou, China, the Stonecutters Bridge (2009) in Hong Kong has been embedded with more than 280, 800, and 1500 sensors [9–11], respectively. Along with these dense sensor networks continuously collecting structural responses, the data-intensive issue arises: the acquired data are overwhelmingly voluminous, whose transfer, storage, retrieval (especially accessed by remote users), and management remain chal- lenging, especially in wireless sensor networks with limited communication bandwidth and battery power supply [12,13]. *Correspondence to: Satish Nagarajaiah, CEVE and MEMS, Rice University. † E-mail: Satish.Nagarajaiah@rice.edu STRUCTURAL CONTROL AND HEALTH MONITORING Struct. Control Health Monit. (2015) Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/stc.1737 Copyright © 2015 John Wiley & Sons, Ltd.