Diagnosis of Concrete Infrastructure Deterioration Using Near-Infrared Spectroscopy - Estimation of Deterioration Depth -

Nondestructive analysis is important to assess aging concrete infrastructure, and several nondestructive analytical techniques have been developed. Near-infrared spectroscopy (NIRS) is one technique that has recently been applied with good results. It has several characteristics that make it beneficial for use in field studies, including being simple, fast, and low-cost.
We investigated the applicability of NIRS to assess concrete deterioration and found it can estimate the factors affecting concrete deterioration with good accuracy and precision. However, the light intensity in the near-infrared region is weak and cannot penetrate deep inside concrete, so it is not possible to directly observe deterioration deep inside concrete. One method used to address this problem is to drill into the concrete and insert fiber optics to measure spectra, although this undermines the benefits of NIRS. To apply NIRS to concrete deterioration deep inside concrete, we must determine how to estimate deep deterioration based on surface observations.
In this study, we investigated the applicability of NIRS to the estimation of deterioration deep inside concrete using spectra obtained from surface observations. We used the phenolphthalein method for the carbonation depth and the silver nitrate solution method for the chlorination depth as reference methods.
The spectrometers used in the study were MicroNIR 1700/2200 (JDSU, USA), which use a linear variable filter (LVF) as the dispersing element and a 128-pixel uncooled InGaAs linear diode array as the detector. The light source was a pair of integrated vacuum tungsten lamps. The wavelength range was 908–1676 nm at 6.2 nm intervals for the MicroNIR 1700 and 1150–2150 nm at 8.1 nm intervals for the MicroNIR 2200.
Partial least squares regression (PLSR) with a full cross-validation method was used to develop and validate calibration models using The Unscrambler software v9.8 (CAMO Software, Oslo, Norway). The wavelength ranges of 1100–1676 nm for the MicroNIR 1700 and 1353–2150 nm for the MicroNIR 2200 were used for calibration development, because the spectra were noisy at the edges of the wavelength ranges of both spectrometers.
The best results were R2=0.80 and SECV=1.04 mm for carbonization using MicroNIR 1700, and R2=0.80 and SECV=1.97 mm for chlorination using MicroNIR 2200. Although the results were inferior to those of surface deterioration estimation, they were still sufficient for qualitative analysis.