Damage identification in civil engineering infrastructure under operational and environmental condition

Real-world structures are subjected to operational and environmental condition changes that impose difficulties for detecting and identifying structural damage. In fact, the author believes that separating changes in sensor readings caused by damage from those caused by changing operational and environmental conditions is one of the biggest challenges for transitioning structural health monitoring (SHM) technology from research to practice. The SHM process is posed in the context of the statistical pattern recognition (SPR) paradigm, where vibration-based methods are applied to detect damage in civil infrastructure. Even though this paradigm intends to pave the way for data-based models applicable to systems of arbitrary complexity, the bridge structures are the focus of this dissertation. The objective of this dissertation is to review, develop, and apply several SHM statistical procedures for feature extraction and statistical modeling for feature classification capable of detect damage on structures under unmeasured operational and environmental variations. In the feature extraction step, the auto-regressive (AR) model is focus of special attention due to its simplicity of application and capability to detect damage. Additionally, a novel algorithm for feature extraction is presented that uses the state-space reconstruction to infer the geometrical structure of a deterministic dynamical system from observed time series of a system response at multiple locations. The unique contribution of this algorithm is that it uses a multivariate auto-regressive model of a baseline condition to predict the state space, where the model encodes the embedding vectors rather than scalar time series. Moreover, four machine learning algorithms are presented to remove the effects of operational and environmental variations on the extracted features. These algorithms are desirable because they develop a functional relationship that models how changing operational and environmental conditions influence the underlying distribution of the damage-sensitive features, without direct measurement of the factors such as temperature and humidity. The applicability of the SHM-SPR, along with the reviewed and proposed statistical procedures, is first demonstrated in a base-excited three-story frame structure tested in laboratory environment to obtain standard data sets from an array of sensors under several structural state conditions. Tests were performed with varying stiffness and mass conditions with the assumption that these sources of variability are representative of changing operational and environmental conditions (e.g. changing mass might represent varying live loads and changing temperature will influence stiffness properties on a structure). Damage is simulated through nonlinear effects introduced by a bumper mechanism that induces a repetitive, impact-type nonlinearity. This mechanism intends to simulate, for instance, the cracks that open and close under dynamic loads or loose connections that rattle. Finally, the applicability of the SHM-SPR paradigm is demonstrated in 12-year span data from the real-world undamaged Alamosa Canyon Bridge, near to Truth or Consequences, New Mexico. Herein, the AR models and machine learning algorithms are focus of special attention. The former because their applicability on civil infrastructure is still limited, and the latter because they might be useful for real-world applications, in situations where the operational and environmental variations cannot be measured.