top of page

Research

The integration of energy harvesting, embedded sensing, addititive manufacturing, multifunctional materials and data-driven modeling has given rise to novel technologies that are fundamentally transforming our capacity to approach the realization of smart cities' vision. Our lab conducts cross-disciplinary research to develop intelligent systems capable of implementing self-sensing and multifunctional abilities, as well as delivering long-term autonomous monitoring data to enhance infrastructure resilience and sustainability while mitigating the impacts of hazards on the built environment. We are interested in advancing the development of novel sensing systems tailored for monitoring the built environment, along with data analytics platforms proficient in the swift transformation of collected data into practical insights. Our research portfolio covers a wide range of areas including:
 

  • Intelligent and energy-efficient structural monitoring

  • Self-powered sensing with embedded computational intelligence

  • Real-time data-driven modeling for the built environment

Smart Bridge Monitoring Employing Deep Learning and Unmanned Aerial Vehicle

According to the 2021 Infrastructure Report Card from the American Society of Civil Engineers, 42% of the nation bridges are 50 years old and 7.5% of bridges are in poor condition.  This follows from the fact that the average age of the nation's infrastructure is rising, with many of them approaching the end of their designed lifespan. As such, significant attention should be devoted to develop new platforms to continuously monitor our existing infrastructure. Conventional bridge monitoring techniques mainly use vision-based inspection methods that suffer from several shortcomings, such as dealing with limited accessible areas, the performance being highly dependable on the inspector’s experience, the inspector’s safety, and time consumption. To tackle the noted challenges, this project aims to develop a data-driven framework for smart bridge monitoring through the integration of deep learning and unmanned aerial vehicle (UAV) using images collected from reinforced concrete bridges. The proposed computer vision-based framework would lead to autonomous bridge monitoring, addressing the drawbacks of traditional inspection methods.

Fig.1.png

High-dimensional Data Analytics in Civil Infrastructure Systems

Recent developments in sensing and monitoring techniques have led to the generation of high-dimensional data in the field of civil engineering. Among the different high-dimensional data analytics techniques, tensor decomposition methods have acquired a notable interest in the civil engineering community over the past few years for missing data imputation. Traditional tensor decomposition methods like CANDECOMP/PARAFAC (CP) and Tucker have been recently adopted for data recovery in civil infrastructure systems. Nevertheless, these decomposition techniques have several limitations, including lack of scalability and flexibility when dealing with high-dimensional data. This study explores the applicability of a new algorithmic framework that utilizes a more efficient tensor decomposition method called tensor-train for data imputation in civil infrastructure systems. The performance of the proposed framework is evaluated using real traffic data set constructed as higher-order tensors. The results indicate that the proposed framework can effectively impute high-dimensional missing data while being accurate even in the presence of large rates of missing data.

Fig.2.png
Fig_edited.jpg
Fig_edited.png

Structural Response Prediction of Concrete Members Utilizing Machine Learning

The design of concrete structures requires considering several key mechanical properties of the material, such as compressive and tensile strengths, flexural strength, and elastic modulus. Nonetheless, the mechanical properties of concrete are known to have strong nonlinear relations between the constituents and the macroscale material characteristicsTherefore, the development of reliable models is of interest to explore material mechanical properties in a way that optimizes cost and time. The potential of machine learning has been harnessed in this project to model such properties. A machine learning-based framework is proposed to predict failure mode, as well as shear and flexural strengths of ultra high performance concrete (UHPC) beams using a comprehensive dataset on tests reported on UHPC beams with different geometric, fiber properties, loading and material characteristics. The results indicate the effectiveness of the framework for structural response prediction of UHPC beams. In addition, a computational framework utilizing machine learning is developed to examine spalling in fire-tested reinforced concrete columns where the governing simplified expressions are derived to capture spalling.

Fig.6.png
Fig_edited.png

Mechanical Characterization of Eco-friendly 3D Printed Concrete 

Significant improvements have been recorded on the technology of additive manufacturing in construction, which has revolutionarily reshaped the research and industry of concrete in recent years. Due to inefficiencies of 3D printing technology, unavoidable defects such as voids, low interlayer bond, bulking, necking, and print precision may arise. Following the integration of engineered cementitious composites in 3D printing, the most recent research interests of 3D printed concrete have been shifted towards recycling aggregates and the incorporation of by-products. This study explores the mechanical performance of the eco-friendly 3D printed concrete by utilizing silica fume, glass fiber, and ground waste rubber tire (replacing fine aggregates) in the mix proportions. Numerical simulations are performed to simualte the mechanical performance of the fabricated 3D printed concrete cubes and beams. A machine learning-based model is developed to predict the compressive and flexural strengths of the aggregate-mixed 3D printed concrete.     

Fig_edited_edited_edited_edited_edited.p
Fig_edited.png

Energy availability for wireless sensor networks to collect and communicate data has been a major concern for infrastructure/structural health monitoring (I/SHM) systems. Recently, self‐powered sensors have become a reality by overcoming the gap between the achievable scavenged energy and the energy required for sensing, computing, and communication. In this study, a novel I/SHM platform is developed based on the integration of energy harvesting, self-powered sensing, communication and data analytics. A new sensing mechanism is proposed based on the integration of self-powered sensors and an energyaware pulse communication protocolThe integrated sensing systems is unique due to its ability to harvesting energy from the environment to operate and wirelessly transmit a binary pulse when its measurement exceeds a defined threshold. However, the communicated data are in a binary format, resulting in discrete and asynchronous event‐based information at the I/SHM processor. To tackle this challenge, an algorithmic framework employing data analytics is developed to interpret self-powered sensor data. The effectiveness of the smart I/SHM platform for energy-lean health monitoring of civil infrastructure and aerospace vehicles is evaluated through numerical and experimental studies. 

Fig_edited.jpg
Fig_edited_edited.png
Fig.23.png

Energy-efficient Structural Health Monitoring Integrating Self-powered Sensing and Data Analytics 

Structural Condition Assessment of Civil Structures with Wireless Sensor Data 

Numerous studies have been carried out to develop algorithms to extract and interpret sensor information for structural health monitoring (SHM) purposes, with most of them relying on continuous time‐history data from a physical response. The effectiveness of traditional SHM methods strongly depends on the availability of continuous time-history data, and such data availability is substantially affected when dealing with discrete data. Interpretation of discrete data over a domain resembles a pattern recognition problem due to its pixelated nature. An image-based pattern recognition approach using anomaly detection is proposed to represent sensor nodes’ responses as a pattern and to identify abnormality and failure patterns. To implement the approach, the arrangement of sensor nodes and the distribution of binary values generated from structural response are considered as a pattern/image. Each pattern is then treated as a matrix and represented by specific features. The results indicate that the proposed image-based pattern recognition approach can effectively identify the presence and location of damage with discrete binary data collected from a network of self-powered sensors.

Fig_edited.jpg
bottom of page