Lead Investigator

Monica Kohler

Overview

The wired Factor building network is a testbed for observational and predictive modeling that is being used to design a structural monitoring wireless network.  Common limitations to the measurement and detection of many types of useful signals by wireless sensing are that digitizer resolution is 16-bit or lower and cannot record useful low-amplitude vibrations.  Typically, MEMS-type sensors have noise levels that are too high, especially for low-frequency signals from which to infer global structural properties.  The vibration data often cannot be calibrated or validated because there are no analogous wired systems with which to compare the wireless data.  Communications and power limitations often limit the number of nodes, components and sample rates of waveform data being recorded continuously for long periods of time.  Factor building earthquake data provide useful information for wireless system design.  By looking at spectral properties, wave propagation properties, and predictive modeling, we use the low-amplitude vibration data in non-traditional ways to identify inelastic behavior, particularly for detecting structural damage.  We are designing a wireless system with high-resolution, high sample rate devices and low-noise sensors for flexible, multi-tier deployments and reconfiguration based on suspected damage locations.

Approach

Our approach to the structural monitoring problem is two-pronged: (1) we are testing existing wireless hardware and software for multi-tier network operation robustness, and (2) we are computing observed and synthetic wave propagation behavior of the Factor building, in part to make appropriate hardware selections for our wireless package.  For (1), we used the wired Factor network, as well as a wave generator, in a side-by-side evaluation of the reliability and accuracy of an existing wireless package.  For (2), we used the Green’s function approach to characterize wave fields propagating through the building between sets of receivers. 

Systems/Experiments

During 2006, we tested several motes that had been constructed with outdated Crossbow data acquisition boards, and Mica-Z MEMS accelerometers.  We conducted tests using a wave generator spanning frequencies between 0.5 and 30 Hz, and a range of input voltages.  We also tested two nodes on the 13th-16th, and 1st-3rd floors in the Factor building.  We discovered several major problems with the hardware setup that motivates us to begin with new, state-of-the art data acquisition boards and MEMS accelerometers.  We discovered that the boards were producing a large voltage offset that precluded the recording of any useful signals above ambient vibration levels.  We also discovered a DC coupling problem that also affected the quality of recorded signal.  In comparison with relatively low-amplitude ambient vibrations in the Factor building, we discovered that the signal-to-noise ratio of the MEMS sensor was too high to be useful for the large amount of ambient vibration recording we would expect to do during most of our continuous, real-environment monitoring.

Accomplishments

We attempted to fix the known hardware problems with the help of Crossbow staff and diagrams, but realized after a few months that the amount of time repairing the existing unreliable equipment was likely to cost more than developing boxes partially from scratch with new boards.  Consequently we are searching for an appropriate low-noise (especially at frequencies < 5 Hz) MEMS accelerometer and A/D board for our system.  The results of our wave propagation modeling have shown that a building’s Green’s functions can be calculated for low-amplitude earthquake data.  They also indicate that it is now necessary to determine minimum measuring thresholds for a variety of realistically observable, damage-related parameters (e.g., additional torsional motions as a result of stiffness changes due to cracked welds in one region) in order to select appropriately sensitive hardware for the wireless nodes.  We are currently considering the Colibrys Si-Flex 1500 MEMS sensor, and are in discussions with a professional A/D board designer who would accommodate the high-resolution, low-noise, small form factor, low-power requirements.

Future Directions

We will construct a prototype 24-bit vibration card for seismic and structural monitoring.  The card, low-noise MEMS accelerometers, and related components will be packaged into two node boxes patterned on a mote network based on the multi-tier network design managed by the Tenet software.  The prototype nodes will be tested using a wave generator and shake table for controlled input, and in side-by-side recordings of ambient vibrations made by the wired Factor seismic network.  This project comprises a proof-of-concept project that will lead to the long-term goal of building and deploying an array of ~50 nodes in a variety of real structures around southern California for pre-earthquake system identification pending funding from NSF or another agency.

We also plan to determine hypothetical observational thresholds for specific types of damage patterns through ongoing analysis and modeling of the wired Factor network.  These results will be used to determine how to deploy wireless nodes before a large shaking event, and how to reconfigure them after suspected damage has occurred due to a large earthquake.  We will be computing signatures of realistic damage scenarios for a typical urban, steel moment-frame building using a computer model of the building in finite-element dynamic analysis simulations.  We will examine three complementary types of data recorded by building arrays: modal, spectral, and wave propagation properties.  An already-developed structural model of the building will be modified for use in dynamic analysis simulations for nonlinear behavior.  The simulations will use a variety of scenario input ground shaking scenarios applied to a range of damaged building scenarios obtained by altering our model in ways that simulate expected damage events, for example, breaking the welds on a particular floor.  We plan to build a wide-range, online catalog of the resulting damage event signatures that could eventually be used in damage detection tools.

People

Faculty:

• Ramesh Govindan (USC Dept. of Computer Science)
• Paul Davis (UCLA Dept. of Earth and Space Sciences)

Researchers:

• Monica Kohler (UCLA CENS)
• Bob Nigbor (UCLA Dept. of Civil Engineering)

Staff:

• Igor Stubailo (UCLA CENS)

Graduate Student:

• Omprakash Gnawali (USC Dept. of Computer Sciences)

External Research Partnerships

• Faculty: Tom Heaton and Swami Krishnan (Caltech Dept. of Civil Engineering) (current/planned)
• Graduate Student: Case Bradford (Caltech Dept. of Civil Engineering) (past)
• Post-doctoral: Matt Muto (Caltech Dept. of Civil Engineering) (planned)