Masoud Sanayei

Train-induced ground vibrations can propagate into pile foundations, potentially causing undesirable vibration in nearby buildings, laboratories housing vibration-sensitive equipment, and manufacturing facilities for high-precision processes. This paper presents an innovative method for predicting building vibration from free-field ground vibration measurements at locations away from the tracks during train pass-bys. The proposed method accounts for site-specific soil profiles and train-soil-pile-structure interactions and is implemented in four steps. In Step 1, train-induced vibration transmission into the ground is estimated using an axisymmetric finite element model that simulates wave propagation through layered soils from the tracks to free-field ground locations. Step 2 estimates free pile head vibration using a three-dimensional finite-element model that captures the ground-borne transmission of track inputs through soil layers to the pile. Step 3 estimates vibration at the junction of the pile head and depot column base using a finite-element model to estimate the pile head impedance and an analytical impedance model for the depot structures supported by the pile. In Step 4, estimates of column-base vibration that transmits into over-track buildings are compared to measured column-base vibration levels obtained during train pass-bys. The method was applied at a metro depot in China, where tracks were in close proximity to columns supporting over-track buildings. Ground and column base vibration levels were measured during multiple train pass-bys. The estimated vibration levels at the base of depot columns closely agreed with the measured vibration levels at the columns during six-car train pass-bys. It demonstrated the potential effectiveness of this hybrid method for assessing vibration transmission into structures atop existing railway tracks. By integrating field measurements, finite element simulations, and analytical impedance models, the proposed hybrid method provides a framework for evaluating the transmission of the train-induced vibration to nearby building structures.

Transfer learning techniques for structural health monitoring in bridge-type structures are investigated, focusing on model generalizability and domain adaptation challenges. Finite element models of bridge-type structures with varying geometry were simulated using the OpenSeesPy platform. Different levels of damage states were introduced at the midspans of these models, and Gaussian-based load time histories were applied at mid-span for dynamic time-history analysis to calculate acceleration data. Then, this acceleration time-history series was transformed into grayscale images, serving as inputs for a Convolutional Neural Network developed to detect and classify structural damage states. Initially, it was trained and tested on datasets derived from a Single-Source Domain structure, achieving perfect accuracy (1.0) in a ten-label multi-class classification task. However, this accuracy significantly decreased when the model was sequentially tested on structures with different geometry without retraining. To address this challenge, it is proposed that transfer learning be employed via feature extraction and joint training. The model showed a reduction in accuracy percentage when adapting from a Single-Source Domain to Multiple-Target Domains, revealing potential issues with non-homogeneous data distribution and catastrophic forgetting. Conversely, joint training, which involves training on all datasets except the specific Target Domain, generated a generalized network that effectively mitigated these issues and maintained high accuracy in predicting unseen class labels. This study highlights the integration of simulation data into the Deep Learning-based SHM framework, demonstrating that a generalized model created via Joint Learning utilizing FEM can potentially reduce the consequences of modeling errors and operational uncertainties unavoidable in real-world applications.

In recent years, the construction of the overtrack building complexes including offices, schools, residences, stores, and restaurants above metro depots has increased in China. Train-generated vibration transmits through ground columns into and across transfer structures into the building’s columns and upper floors. The vibration has the potential to impact people’s quality of life, work in office buildings, and the proper function of vibration-sensitive equipment in laboratories. An efficient computational model has been developed that characterizes the dominant mode of vibration transmission through each structural element including those in transfer structures in building designs where ground and building columns are not aligned. Impedance matrices are combined for the elements along the transmission path from inputs at ground columns with axial wave transmission into the transfer structures’ girders and beams with bending wave transmission and the building columns with axial wave transmission. Floors are characterized by point input impedances. Model inputs are vibration levels at ground column bases during train pass-bys. Model predictions were compared with measured vibration levels within a 4-story overtrack building on a transfer structure in the Qianhai depot in China. The agreement between predicted and measured vibration levels was quite good. The proposed model can be used to accurately predict overtrack vibration levels for buildings supported on a transfer structure with a small computational cost in order to identify potential vibration concerns prior to construction.

Recent research on structural health monitoring has shown the applicability of Kalman filtering for fusion of models and data of dynamic systems. In this study, the augmented Kalman filter (AKF) is combined with a substructuring approach for the estimation of strain time histories at unmeasured locations of a rollercoaster connection, with the purpose of predicting remaining fatigue life. The AKF allows for response estimation in the absence of direct measurements of operational loads and in the presence of noisy measurements at limited locations of the instrumented system. The proposed research implements for the first time the AKF and a substructure with exclusively in-service strain measurements from a full-scale structure. Given that uniaxial fatigue life assessment methods often prove insufficient for large in-service structures with complex geometry and subjected to multiaxial nonproportional loadings, the critical plane method is here employed for estimation of fatigue life. The applicability of a substructuring approach combined with strain estimates from the AKF is demonstrated and evaluated with operational strain measurements obtained from an in-service rollercoaster support structure. It is shown that the proposed methodology and fine-tuning of the AKF parameters result in a close match of predicted and measured strains.