Experimental Evaluation of Drive-by Health Monitoring on a Short-Span Bridge Using OMA Techniques 121 similar damping ratios as the half-power bandwidth method; however, it can again be seen that the methodology struggles to calculate damping in the lower frequency range. 6.4 Summary of Uncoupled OMA System Identification The results from the uncoupled analysis of the bridge identified that the system violated the time invariant assumption, as nonlinearities believed to be introduced by breathing cracks and heavy tractor-trailers caused the apparent first resonant frequency and damping ratio to fluctuate between 11.5 Hz–16.5 Hz and 1%–5%, respectively. It was observed that the FDD procedures produced narrower variations in the principle frequency, indicating the FDD technique was more robust against apparent time variant system properties. It was also determined, however, that the NExT procedure used for calculating damping under the FDD technique was significantly influenced by initial coefficient estimates and the magnitude of frequency excitations; thus indicating the half-power bandwidth method may be more consistent for identifying damping under the given framework. The results from the uncoupled analysis of the RAM truck also indicated that the vehicle violated the time invariant assumption, as nonlinearities believed to be introduced by nonlinear wheel properties at the front of the vehicle caused notable harmonics and frequency fluctuations in the range of the unsprung natural frequency. In a similar manner to the direct bridge analysis, the FDD technique was robust against the apparent time variant effects, effectively eliminating the frequency bin distribution observed at the front of the truck during the PP analysis. The NExT procedure was once again significantly influenced by initial coefficient estimates and the magnitude of frequency excitations. 7 Coupled Vehicle-Bridge System Identification 7.1 Preprocessing The same preprocessing procedures used in Sect. 6.1 were used for the coupled analysis when considering direct bridge data. When processing vehicle data, however, time histories were divided into time frames of before the vehicle entered the bridge span and while the vehicle was on the subject bridge span. Figure 11 demonstrates how the full vehicle time histories were divided. On-bridge data encompassed the period of time from half-a-second before the front wheels entered the subject bridge span to half-a-second after the rear wheels exited the span; this correlated to an approximately 2.5 s time window. The half-a-second buffer before and after ensured that the subject time frame fully captured the truck occupation time and allowed multiple Hann windows to be employed for the average and short-time analyses. Before-bridge data encompassed Fig. 11 Example of coupled vehicle time history highlighting before-bridge and on-bridge time frames
RkJQdWJsaXNoZXIy MTMzNzEzMQ==