This paper presents a sequential structural damage detection algorithm that is based on a statistical model for the wavelet transform of the structural responses. The detector uses the coefficients of the wavelet model and does not require prior knowledge of the structural properties. Principal Component Analysis is applied to select and extract the most sensitive features from the wavelet coefficients as the damage sensitive features. The damage detection algorithm is validated using the simulation data collected from a four-story steel moment frame. Various features have been explored and the detection algorithm was able to identify damage. Additionally, we show that for a desired probability of false alarm, the proposed detector is asymptotically optimal on the expected delay.
This paper proposes a series of novel Damage Sensitive Features for earthquake damage estimation. The features take into account input (ground motion) and output acceleration (structure response) measurements. The Continuous Wavelet Transform is applied to both acceleration signals in order to obtain both time domain and frequency domain resolution. An algorithm that has been proposed for Maximum Entropy Deconvolution is applied to the Continuous Wavelet Transforms in order to obtain a matrix that relates the output wavelet coefficients to the input ones. The Damage Sensitive Features are then derived through statistical processing of the resulting matrix. This algorithm has been applied on data acquired from shake table tests where the structures were subjected to progressive damage. The proposed features are compared to response quantities that are indicative of damage (such as the hysteretic energy dissipated) and show high correlation with the extent of damage. The data utilized has not been pre-processed, illustrating the robustness of the algorithm against sensor noise. The proposed algorithm has several advantages: Minimal input and knowledge of the structure is required. More information on the structure's state is extracted through use of both the input and output signals than when only output signal is considered. Only two acceleration measurements are required to obtain a damage forecast utilizing primarily the strong motion recordings, resulting in easier sensor deployment. The use of strong motion recordings allows for information delivery immediately after an earthquake without additional data collection.
In a previous paper an algorithm was developed for estimating the slope at locations of ambient vibration measurements along the deformed shape of a single column subjected to strong earthquake motion. These slope estimates were used in obtaining permanent drift values for the single column and those were correlated to various levels of damage. The algorithm was illustrated with applications to single column tests performed at the University of Nevada, Reno and the University of California, Berkeley. In this paper, the same columns are first used to simulate slope values along the deformed shape of the columns in order to determine the best estimate of the displacement distribution. This information can then be used to determine the optimal number and location of sensors needed to provide a reliable permanent drift in a single column. Sensitivity studies are also performed to evaluate the effect of plastic hinge length over or underestimate as this value is usually inferred from empirical equations. The rotation algorithm is then extended to estimate residual drift from rotation measurements taken from multiple sensors in order to correlate them to structural damage. The resulting drift values can then be related to various damage states and can be used for rapid damage assessment immediately following a major earthquake. The main advantage of the proposed approach is that low-cost accelerometers can be used to obtain the information needed for rapid damage assessment.