Quantitative evaluation of steel corrosion induced deterioration in rubber concrete by integrating ultrasonic testing, machine learning and mesoscale simulation

Significance 

Reinforced concrete structures are highly susceptible to chloride-induced steel corrosion, especially in marine environments. This has become a major hindrance to the widespread adoption of reinforced concrete structures as the corrosion severely affects the structural durability and is costly to maintain. Thus, it is urgent to develop effective strategies to prevent chloride diffusion and steel corrosion and its associated deterioration effects on structures exposed to the marine environment.

Rubber concrete developed by recycling waste rubber is an ecosystem-friendly and cost-effective construction material. It exhibits superior resistance to steel corrosion induced by chloride and related concrete deterioration effects, mainly due to improved resistance to chloride penetration, smaller expansion and improved cracking and electrical resistance of the rubber concrete. Despite the significant amount of research on rubber concrete, little attention has been paid to the corrosion of steel in reinforced concrete, which can cause spalling, cracking, brittle failure, delamination and possible collapse of the concrete.

Promoting the use of rubber concrete in the marine environment and securing marine infrastructure requires accurate monitoring of the corrosion degree of the rebar and quantitative evaluation of the rubber concrete deterioration degree. Although these can be achieved through electrochemical, reaction control and empirical methods, they have various limitations. Non-destructive detection methods, such as ultrasonic testing, have emerged as alternative and efficient solutions. Nevertheless, the material complexity makes it difficult to use non-destructive methods to evaluate the concrete deterioration degree.

Advanced quantitative evaluation and data processing methods based on non-destructive technologies have been established to overcome these inherent challenges. Specifically, machine learning algorithms have successfully evaluated and predicted concrete properties as they can accurately represent relationships among variables. Equipped with this knowledge, Professor Jinrui Zhang and Dr. Mengxi Zhang from Tianjin University in collaboration with Professor Biqin Dong from Shenzhen University and Professor Hongyan Ma from Missouri University of Science and Technology, proposed a new approach integrating mesoscale simulation, machine learning and ultrasonic testing to quantitatively and accurately evaluate corrosion-induced rubber concrete deterioration. The work is currently published in the journal, Cement and Concrete Composites.

In their approach, the research team prepared the reinforced concrete specimens with different rubber contents: 0%, 10% and 20% and subjected them to corrosion experiments accelerated electrochemically. The experiments were monitored through ultrasonic testing to produce corrosion and deterioration data used to train six machine learning models. The corrosion degree was predicted based on the obtained ultrasonic traits as well as ultrasonic, steel corrosion and concrete mixture parameters. The models were compared in terms of accuracy and robustness to select the most feasible one for practical applications.

The researchers demonstrated that all the machine learning models with the exception of the linear model, were able to predict the degree of corrosion robustly and accurately under the interference of the size and outlier amplitude of the training set. High rubber content in the concrete minimized not only the corrosion effects of chloride ions but also improved the corrosion resistance of the concrete structures. The recorded low damage ratios at similar steel corrosion degrees showed that incorporating rubber could effectively mitigate steel corrosion and deterioration.

Being a non-destructive method, the ultrasonic amplitude was considered the most critical monitoring index in evaluating the corrosion degree. Therefore, its measurement accuracy must be guaranteed to achieve the desired performance accuracy. The excellent accuracy and robustness of the five models showed the superiority of the proposed concepts and their applicability in practical scenarios.

Furthermore, the corrosion-induced deterioration process is computed by the mesoscale simulation method based on the corrosion degree developed by authors, so that the damages of specimens with different rubber contents are quantitatively evaluated. Further prove that the steel corrosion induced deterioration in reinforced concrete can be effectively mitigated by the incorporation of rubber, evidenced from lower damage ratios at the same steel corrosion degree.

In summary, the study successfully evaluated the corrosion degree of reinforced steel concrete with various contents of rubber concrete. It provided novel and robust non-destructive methods for evaluating corrosion-induced degradation of concrete structures. In a joint statement to Advances in Engineering, the authors said that the proposed models would enhance the safety and structural integrity of marine infrastructure.

Reference

Zhang, J., Zhang, M., Dong, B., & Ma, H. (2022). Quantitative evaluation of steel corrosion induced deterioration in rubber concrete by integrating ultrasonic testing, machine learning and mesoscale simulationCement and Concrete Composites, 128, 104426.

Go To Cement and Concrete Composites

Check Also

Reliability of structural wall shear design for tall reinforced-concrete core wall buildings - Advances in Engineering

Reliability of structural wall shear design for tall reinforced-concrete core wall buildings