Incorporating Uncertainty in Mechanics-Based Synthetic Data Generation for Deep Learning–Based Structural Monitoring

This chapter presents a hybrid deep-learning methodology for seismic structural monitoring, damage detection, and localization of instrumented buildings. The proposed methodology develops mechanics-based structural models to generate sample response datasets by accounting for the uncertainty of model parameters that can highly affect the estimation of baseline model nonlinear responses. The mechanics-based models are developed considering uncertainties in the stiffness, strength, and geometry of the baseline numerical model’s characteristics and elements. The baseline model is run multiple times with defined assumptions and variations in the selected parameters of the model. The uncertainty of model parameters is evaluated through the design of experiments methodology by employing the central composite design for sampling. A parameter effect analysis is used to assess the significance of the modeling input parameters on the selected structural output response, such as inter-story drifts. The generated sample response dataset is utilized for training a hybrid data-driven model for feature extraction. To select the damage-sensitive features, a convolutional neural network as the main feature extraction body of the network is used. In addition, wavelet packet–based nodal first temporal moments (energies) are also employed to boost the feature extraction power of the network as a complementary body. This data-driven model is designed to use global story-level noise-contaminated response measurements are employed as input for the data-driven model to perform damage detection and localization in a manner consistent with performance-based design criteria. The performance of the proposed methodology is studied in the context of numerical and experimental case studies developed based on the shake table testing of a concentrically braced frame subject to various input ground motion intensities at the E-Defense facility in Miki, Japan. The results show that the proposed methodology provides high accuracy in classifying and localizing various damage patterns.