Fusion of Rebound Number and Ultrasonic Pulse Velocity Data for Evaluating the Concrete Strength Using Bayesian Updating

Abstract

Reinforced concrete (RC) is one of the most widely used construction material. However, failure in old RC structures is primarily due to degradation and deterioration of concrete, while failure in new structures is due to various other reasons. Most of the current RC structure stock is reaching its service life limit and many are still being used beyond their anticipated life span. Structural health monitoring (SHM) becomes an essential step to decide whether the structure is to be repaired, retrofitted, replaced, or allowed to continue without any action. The concrete strength assessment at any stage of the RC structure is necessary for SHM. Non-destructive testing (NDT) methods are widely accepted techniques and commonly used tools for SHM. Different NDT tools are used to understand different properties of the structure such as homogeneity, strength, presence of crack, and carbonation. Two different NDTs may deliver complementary information or conflicting results on the same parameter. Combining the measurements of these different techniques will incorporate various effects of the property on concrete strength resulting in better assessment. This paper focuses on using Bayesian updating for fusing data of rebound number (RN) and ultrasonic pulse velocity (UPV), which are two most commonly used NDT tools to evaluate the strength of concrete. Bayesian inference helps to quantify the epistemic uncertainty (in measurements) and integrate both UPV and RN measurements into a combined estimate. To account for the uncertainty, the measurement error model (MEM) is calibrated through Bayesian updating using the Markov Chain Monte Carlo (MCMC) algorithm. A framework of data fusion is discussed in detail considering three prior cases (non-informative, partially informative and fully informative), which are adopted based on IS 456:2000 specifications. A robust estimation of the estimated strength density functions (PDFs) is obtained, which becomes unconditioned on the MEM parameters obtained from the Bayesian calibration of a MEM. Robust estimates from two different NDT tools are then fused to a single (probabilistic) estimate of the crushing strength of concrete.