Challenges in Experimental Vibration Analysis for Structural Identification and Corresponding Engineering Strategies

Structural identification (St-Id) of long-span bridges by vibration testing under operation provides a starting point for quantitatively characterizing the in-service mechanical characteristics and behaviours of these complex constructed systems by designing and integrating additional experiments such as local NDE and crawl-speed load tests. The resulting characterization can serve as an effective baseline for structural health monitoring, designing more reliable and cost-effective structural retrofits, or developing and implementing more timely and efficient maintenance procedures. However, there are various uncertainties involved in the experimental and identification processes that impact the reliability of St-Id especially if vibration testing is the only experiment, these serve as a barrier to more widespread applications in civil engineering practice. The prevailing excitations (wind and traffic), environmental conditions (radiation and ambient temperature), experimental hardware (sensors, cabling and data acquisition system), and the execution of the experiment (array density and distribution, data acquisition parameters, on-site quality control, etc) all have a significant influence on the field test data quality and whether this data can be reliably processed for dynamic characteristics. Analytical modelling of complex structural systems and their components for St-Id brings its own significant uncertainties. Recognizing the impact of various uncertainty mechanisms and taking appropriate measures to quantify, bound and mitigate their impacts will greatly benefit bridge owners and engineers. A vibration test on a long-span suspension bridge is leveraged as an example to illustrate a number of possible strategies for coping with the challenges presented above for practical identification of the structural dynamic characteristics of large-scale constructed systems. The design and implementation of a multi-reference field test are first presented to illustrate how uncertainties can be mitigated from an experimental point of view. Next, data pre-processing strategies including data inspection, time window selection, band-pass filtering, averaging and windowing are proposed to reduce data errors and to detect possible causes of outliers in the measured data (higher traffic, higher wind, temperature shocks, construction activity, unusual activity, etc). Three separate St-Id post-processing methods, including Peak-Picking, PolyMAX, and Complex Mode Indicator Function (CMIF), are applied for accurate structural modal parameter identification. Statistical analysis of the St-Id results from a number of time history windows are also performed, providing effective ways to investigate window relevance, data reliability and how they affect St-Id results. The identification results obtained for both the bridge spans and the towers demonstrate that the demonstrated field testing and data processing methods may provide a reliable bridge characterization.