arXiv subjects

Shamim N. Pakzad

Publications and source records attributed to Shamim N. Pakzad.

9 recordsLinked to original sources

arXiv2022

Identifying Damage-Sensitive Spatial Vibration Characteristics of Bridges from Widespread Smartphone Data

The knowledge gap in the expected and actual conditions of bridges has created worldwide deficits in infrastructure service and funding challenges. Despite rapid advances over the past four decades, sensing technology is still not a part of bridge inspection protocols. Every time a vehicle with a mobile device passes over a bridge, there is an opportunity to capture potentially important structural response information at a very low cost. Prior work has shown how bridge modal frequencies can be accurately determined with crowdsourced smartphone-vehicle trip (SVT) data in real-world settings. However, modal frequencies provide very limited insight on the structural health conditions of the bridge. Here, we present a novel method to extract spatial vibration characteristics of real bridges, namely, absolute mode shapes, from crowdsourced SVT data. These characteristics have a demonstrable sensitivity to structural damage and provide superior, yet complementary, indicators of bridge condition. Furthermore, they are useful in the development of accurate mathematical models of the structural system and help reconcile the differences between models and real systems. We demonstrate successful applications on four very different bridges, with span lengths ranging from about 30 to 1300 meters, collectively representing about one quarter of bridges in the US. Supplementary work applies this computational approach to accurately detect simulated bridge damage entirely from crowdsourced SVT data in an unprecedentedly timely fashion. The results presented in this article open the way towards large-scale crowdsourced monitoring of bridge infrastructure.

physics.app-ph↗

arXiv2022

Crowdsourcing Bridge Vital Signs with Smartphone Vehicle Trips

A key challenge in monitoring and managing the structural health of bridges is the high-cost associated with specialized sensor networks. In the past decade, researchers predicted that cheap, ubiquitous mobile sensors would revolutionize infrastructure maintenance; yet many of the challenges in extracting useful information in the field with sufficient precision remain unsolved. Herein it is shown that critical physical properties, e.g., modal frequencies, of real bridges can be determined accurately from everyday vehicle trip data. The primary study collects smartphone data from controlled field experiments and "uncontrolled" UBER rides on a long-span suspension bridge in the USA and develops an analytical method to accurately recover modal properties. The method is successfully applied to "partially-controlled" crowdsourced data collected on a short-span highway bridge in Italy. This study verifies that pre-existing mobile sensor data sets, originally captured for other purposes, e.g., commercial use, public works, etc., can contain important structural information and therefore can be repurposed for large-scale infrastructure monitoring. A supplementary analysis projects that the inclusion of crowdsourced data in a maintenance plan for a new bridge can add over fourteen years of service (30% increase) without additional costs. These results suggest that massive and inexpensive datasets collected by smartphones could play an important role in monitoring the health of existing transportation infrastructure.

cs.CY↗

arXiv2021

Transfer Learning for Input Estimation of Vehicle Systems

This study proposes a learning-based method with domain adaptability for input estimation of vehicle suspension systems. In a crowdsensing setting for bridge health monitoring, vehicles carry sensors to collect samples of the bridge's dynamic response. The primary challenge is in preprocessing; signals are highly contaminated from road profile roughness and vehicle suspension dynamics. Additionally, signals are collected from a diverse set of vehicles vitiating model-based approaches. In our data-driven approach, two autoencoders for the cabin signal and the tire-level signal are constrained to force the separation of the tire-level input from the suspension system in the latent state representation. From the extracted features, we estimate the tire-level signal and determine the vehicle class with high accuracy (98% classification accuracy). Compared to existing solutions for the vehicle suspension deconvolution problem, we show that the proposed methodology is robust to vehicle dynamic variations and suspension system nonlinearity.

cs.LG↗

arXiv2021

RL-Controller: a reinforcement learning framework for active structural control

To maintain structural integrity and functionality during the designed life cycle of a structure, engineers are expected to accommodate for natural hazards as well as operational load levels. Active control systems are an efficient solution for structural response control when a structure is subjected to unexpected extreme loads. However, development of these systems through traditional means is limited by their model dependent nature. Recent advancements in adaptive learning methods, in particular, reinforcement learning (RL), for real-time decision making problems, along with rapid growth in high-performance computational resources, help structural engineers to transform the classic model-based active control problem to a purely data-driven one. In this paper, we present a novel RL-based approach for designing active controllers by introducing RL-Controller, a flexible and scalable simulation environment. The RL-Controller includes attributes and functionalities that are defined to model active structural control mechanisms in detail. We show that the proposed framework is easily trainable for a five story benchmark building with 65% reductions on average in inter story drifts (ISD) when subjected to strong ground motions. In a comparative study with LQG active control method, we demonstrate that the proposed model-free algorithm learns more optimal actuator forcing strategies that yield higher performance, e.g., 25% more ISD reductions on average with respect to LQG, without using prior information about the mechanical properties of the system.

eess.SY↗

arXiv2020

Bridge Structural Health Monitoring using Asynchronous Mobile Sensing Data

This study presents a flexible approach for bridge modal identification using smartphone data collected by a large pool of passing vehicles. With each trip of a mobile sensor, the spatio-temporal response of the bridge is sampled, plus various sources of noise, e.g., vehicle dynamics, environmental effects, and road profile. This paper provides further evidence to support the hypothesis that through trip aggregation, such noise effects can be mitigated and the true bridge dynamics are exhibited. In this study, the continuous wavelet transform is applied to each trip, and the results are combined to estimate the structural modal response of the bridge. The Crowdsourced Modal Identification using Continuous Wavelets (CMICW) method is presented and validated in an experimental setting. In summary, the method successfully identifies natural frequencies and absolute mode shapes of a bridge with high accuracy. Notably, these results are the first to extract torsional mode shape information from mobile sensor data. Moreover, the influence of vehicle speed on the estimation accuracy is investigated. Finally, a hybrid simulation framework is proposed to account for the vehicle dynamics within the raw mobile sensing data. The proposed method is successful in removing vehicle dynamic effects and identifying modal properties. These results contribute to the growing body of knowledge on the practice of mobile crowdsensing for physical properties of transportation infrastructure.

eess.SP↗

arXiv2020

DynNet: Physics-based neural architecture design for linear and nonlinear structural response modeling and prediction

Data-driven models for predicting dynamic responses of linear and nonlinear systems are of great importance due to their wide application from probabilistic analysis to inverse problems such as system identification and damage diagnosis. In this study, a physics-based recurrent neural network model is designed that is able to learn the dynamics of linear and nonlinear multiple degrees of freedom systems given a ground motion. The model is able to estimate a complete set of responses, including displacement, velocity, acceleration, and internal forces. Compared to the most advanced counterparts, this model requires a smaller number of trainable variables while the accuracy of predictions is higher for long trajectories. In addition, the architecture of the recurrent block is inspired by differential equation solver algorithms and it is expected that this approach yields more generalized solutions. In the training phase, we propose multiple novel techniques to dramatically accelerate the learning process using smaller datasets, such as hardsampling, utilization of trajectory loss function, and implementation of a trust-region approach. Numerical case studies are conducted to examine the strength of the network to learn different nonlinear behaviors. It is shown that the network is able to capture different nonlinear behaviors of dynamic systems with very high accuracy and with no need for prior information or very large datasets.

cs.LG↗

arXiv2020

Finite Difference Neural Networks: Fast Prediction of Partial Differential Equations

Discovering the underlying behavior of complex systems is an important topic in many science and engineering disciplines. In this paper, we propose a novel neural network framework, finite difference neural networks (FDNet), to learn partial differential equations from data. Specifically, our proposed finite difference inspired network is designed to learn the underlying governing partial differential equations from trajectory data, and to iteratively estimate the future dynamical behavior using only a few trainable parameters. We illustrate the performance (predictive power) of our framework on the heat equation, with and without noise and/or forcing, and compare our results to the Forward Euler method. Moreover, we show the advantages of using a Hessian-Free Trust Region method to train the network.

stat.ML↗

arXiv2020

Simplified Vehicle-Bridge Interaction for Medium to Long-span Bridges Subject to Random Traffic Load

This study introduces a simplified model for bridge-vehicle interaction for medium- to long-span bridges subject to random traffic loads. Previous studies have focused on calculating the exact response of the vehicle or the bridge based on an interaction force derived from the compatibility between two systems. This process requires multiple iterations per time step per vehicle until the compatibility is reached. When a network of vehicles is considered, the compatibility equation turns to a system of coupled equations which dramatically increases the complexity of the convergence process. In this study, we simplify the problem into two sub-problems that are decoupled: (a) a bridge subject to random Gaussian excitation, and (b) individual sensing agents that are subject to a linear superposition of the bridge response and the road profile roughness. The study provides sufficient evidence to confirm the simulation approach is valid with a minimal error when the bridge span is medium to long, and the spatio-temporal load pattern can be modeled as random Gaussian. Quantitatively, the proposed approach is over 1,000 times more computationally efficient when compared to the conventional approach for a 500 m long bridge, with response prediction errors below $0.1\%$.

eess.SP↗

arXiv2020

Bridge Modal Identification using Acceleration Measurements within Moving Vehicles

Vehicles crossing bridge structures respond dynamically to the bridge's vibrations. An acceleration signal collected within a moving vehicle contains a trace of the bridge's structural response, but also includes other sources such as the vehicle suspension system and surface roughness-induced vibrations. This paper introduces two methods for the bridge system identification using data collected by a network of moving vehicles. The contributions of the vehicle suspension system are removed by deconvolving the vehicle response in frequency domain. The first approach utilizes the vehicle transfer function, and the second uses EEMD method. Next, roughness-induced vibrations are extracted using second-order blind identification (SOBI) method. After these two processes the resulting signal is equivalent to the readings of mobile sensors that scan the bridge's dynamic response. Structural modal identification using mobile sensor data has been recently introduced as STRIDEX algorithm. The processed mobile sensor data is analyzed using STRIDEX to identify the modal properties of the bridge. The performance of the methods is validated on numerical case studies of a long single-span bridge monitored via a network of moving vehicles. The analyses consider three road surface roughness patterns. Results show that the proposed algorithms are successful in extracting pure bridge vibrations, and produce accurate and comprehensive modal properties of the bridge. The study shows that the proposed transfer function method can efficiently deconvolve the linear dynamics of a moving vehicle. EEMD method is able to extract vehicle dynamic response without a-priori information about the vehicle. This study is the first proposed methodology for complete bridge modal identification, including operational natural frequencies, mode shapes and damping ratios using \textit{moving vehicles sensor data}.

eess.SP↗