System identification is the process of building mathematical models of dynamic systems from measured input-output data. In structural engineering, it is used to identify modal properties, update finite element models, and characterize structural behavior for monitoring and control.
System identification is the discipline of building mathematical models of dynamic systems from observed input-output data. In the context of structural dynamics and structural health monitoring, it refers to the set of methods used to identify a structure's modal properties (natural frequencies, mode shapes, damping ratios), to update analytical models (typically finite element models), and to characterize structural behavior. The field has its roots in control theory, signal processing, and statistics, and has developed into a mature discipline with applications across civil, mechanical, and aerospace engineering.
System identification in structural engineering is typically classified by the type of data available. Input-output identification uses both measured input (excitation) and output (response) β as in experimental modal analysis (EMA) using impact hammers or shakers. Output-only identification uses only the response, assuming the input is broadband stationary (as in operational modal analysis, OMA). Nonlinear system identification extends the classical linear framework to systems that exhibit nonlinear behavior β including hysteretic structures, base-isolated buildings, and structures with damage. Time-varying identification tracks how the system's properties evolve over time, which is essential for detecting damage or degradation during an earthquake. Bayesian identification provides a probabilistic framework for system identification, quantifying uncertainty in the identified parameters β increasingly important in SHM for decision-making under uncertainty.
System identification is central to model updating and SHM. In model updating, identified modal properties are used to calibrate a finite element model so that its predictions match measured behavior; the updated model can then be used for response prediction, reliability assessment, and damage assessment. In SHM, system identification is used to track structural health over time β by identifying modal properties continuously and using changes in those properties as indicators of damage. Modern SHM systems increasingly rely on automated system identification algorithms that can process large datasets in real time and provide near-real-time damage assessment. The key challenges in structural system identification are (1) the presence of unmeasured or unknown input (especially for output-only methods), (2) the presence of nonlinearity (which can be significant even in nominally linear structures under strong motion), (3) the influence of environmental and operational variability (which can mask damage-induced changes), and (4) the need for uncertainty quantification (to enable probabilistic decision-making). In earthquake engineering, system identification is used both for post-earthquake assessment (identifying whether a structure's properties have changed after an event) and for pre-earthquake monitoring (tracking long-term changes in structural health). In Iran, system identification is increasingly used in SHM projects, particularly for bridges, dams, and important buildings, with growing interest in machine learning and Bayesian methods.