System Identification of a Steel Frame Using Defective Sensor Data

Output-only system identification methods assist engineers in determining operational modal properties of constructed structural systems. Sensors and data acquisition systems have a finite reliability; even state-of-the-art technologies are susceptible to missing packets, erroneous values, and other malfunctions. It is expected that missing data points degrade the estimation accuracy of important structural features such as natural vibration properties or damage metrics. The goal of this paper is to quantify corresponding structural information losses in modal identification. The main hypothesis of this paper is that despite missing data points caused by measurement issues, there is a substantial amount of structural information hidden within the observed data. Structural modal properties can be estimated with sufficient accuracy even with significant data loss through the use of the STRIDE algorithm. This study investigates the response of a three-story steel frame captured by a network of accelerometers. Systematic sensor failures are implemented artificially to study their effect on modal property estimation. It is shown that structural mode shapes can be computed with a certain accuracy even when 90% of the measurements for a particular story are lost.