High-performance fiber reinforced concrete as a repairing material to normal concrete structures: Experiments, numerical simulations and a machine learning-based prediction model. (30th October 2019)

• Record Type: Journal Article
• Authors: Jiao, Pengcheng; Roy, Manish; Barri, Kaveh; Zhu, Ronghua; Ray, Indrajit; Alavi, Amir H.
• Is Part Of: Construction & building materials. Volume 223(2019)
• Page Start: 1167
• Page End: 1181
• Publication Date: 2019-10-30
• Subjects: High-performance fiber reinforced concrete (HPFRC) -- Normal concrete (NC) -- Debonding behavior -- Machine learning -- Prediction model -- Direct shear test; Building materials -- Periodicals; 624.18
• DOI: 10.1016/j.conbuildmat.2019.07.312
• Languages: English
• ISSNs: 0950-0618

Abstract

Highlights: High-performance fiber reinforced concrete (HPFRC) is reported as repairing material to normal concrete (NC). HPFRC and NC samples are manufactured and experimentally calibrated. HPFRC-NC debonding test is conducted under direct shear load. A machine learning model is developed to predicted the shear debonding behavior of HPFRC-NC. Abstract: High-performance fiber reinforced concrete (HPFRC) has been reported as a repairing material to normal concrete (NC) structures due to its predominant mechanical performance. Here, we investigate the debonding behavior between HPFRC and NC subjected to direct shear loading. HPFRC specimens are fabricated and experimentally calibrated to determine the compressive and bending (i.e., flexural) strengths. HPFRC-NC samples are fabricated using two bonding strategies, i.e., mechanical surface treatments with and without chemical agent. Direct shear loading is applied to test the HPFRC-NC debonding behavior. A finite element (FE) model is developed to predict the direct shear debonding response. The FE model is validated by the experimental observations and then used to characterize the debonding behavior with various geometric and material parameters, as well as bonding interface treatments. Subsequently, a robust machine learning model is developed to formulate the shear debonding strength of HPFRC-NC with those influencing parameters. Design examples are presented to illustrate the efficiency of the proposed machine learning model in describing the debonding response of HPFRC-NC. A sensitivity analysis is further conducted to investigate the contribution of the chosen predictors to the debonding behavior of HPFRC-NC. The reported HPFRC and machine learning-based prediction model provide powerful tools to address repairing issues in various existing normal concrete structures.