Detecting Chloride Degradation in Concrete Structures by Developing a Physics-trained Artificial Intelligence Framework

Patel, Parth; Gupta, Abhinav; Bodda, Saran Srikanth; Sandhu, Harleen Kaur

Permalink

https://hdl.handle.net/2142/121828 Copy

Description

Title

Detecting Chloride Degradation in Concrete Structures by Developing a Physics-trained Artificial Intelligence Framework

Author(s)

• Patel, Parth
• Gupta, Abhinav
• Bodda, Saran Srikanth
• Sandhu, Harleen Kaur

Issue Date

2023

Keyword(s)

• Structural health monitoring
• Simulating concrete degradation
• Artificial intelligence (AI)
• Uncertainty quantification

Abstract

Concrete structures often interact with various chemical and physical degradation agents throughout their service life. It causes severe alterations in the concrete properties, such as a reduction in durability and load- carrying capacity. One such primary and often the most important chemical degradation agent is chloride ion. The chloride ion causes corrosion in reinforcement by diffusing inside the concrete and eventually leads to its cracking. However, the chloride degradation mechanism is only noticeable when concrete spalling and cracking occur. Therefore, the early detection of degradation is essential to maintain the safety of critical infrastructures. It can be achieved by developing a structural health monitoring (SHM) framework using non-destructive testing (NDT) techniques on the structure with chloride degradation. However, the data obtained from NDT techniques are often complex and difficult to interpret. Recent studies use artificial intelligence (AI) to comprehend these complex data through pattern recognition. In the case of SHM applications, the efficacy of AI solutions will highly depend on the type of structures the NDT is performed on. Additionally, training and building a good-performing AI algorithm for SHM requires data for already degraded structures at various stages in their lifespan and degradation process. Since this degradation occurs over many years, it is not practical to collect actual NDT data over multiple years and allow the structure to continue to degrade. Past research show that a multi-step simulation process can be employed to generate data at numerous stages of degradation. It can help provide insight into the structure’s condition due to different amounts of chloride exposure. Since the information is obtained from various sources and is provided in a variety of formats, the information accessible in simulating the chloride degradation process is diverse. Therefore, it is crucial to account for various uncertainties. It can be accomplished using forward uncertainty quantification (UQ) approaches to propagate individual uncertainty through various phases of the concrete degradation process. In this research, we propose an integrated structural health monitoring framework by propagating various uncertainties through a multi-step chloride degradation simulation. The obtained learnings and data are used to develop a physics-driven AI framework. The proposed framework detects non-uniform chloride degradation utilizing a combination of advanced finite element modeling, sensor data, and deep learning.

Type of Resource

text

Language

eng

Handle URL

https://hdl.handle.net/2142/121828

Owning Collections