Analysis of mid-infrared spectrum characteristics of sandstone with different acidification degrees based on fusion...

Rock dissolution induced by acidic groundwater poses a significant threat to the stability of geotechnical engineering. Therefore, it is crucial to develop a robust spectral prediction model to accurately evaluate the degree of rock acidification. Initially, red sandstone samples were immersed in hydrochloric acid solutions of different concentration for 1 h, 3 h, 5 h, 24 h, and 72 h, respectively. Fourier variation mid-infrared spectroscopy was employed to analyze the spectral characteristics of samples, assessing the acidification degrees. To mitigate environmental interference and eliminate redundant information, Savitzky-Golay (S-G) smoothing, normalization, and Principal Component Analysis (PCA) were applied to preprocess the spectral data of differently acidified rock samples. Subsequently, K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Random Forest (RF) algorithms were compared, and a fusion model of mid-infrared spectral prediction models for red sandstone samples with varying degrees of acidification was established. The fusion model was proposed to integrate the strengths of multiple models, enabling precise characterization of the acidification degrees of red sandstone. The results indicate that as concentration of hydrochloric acid solutions increases and soaking time extends, the reflectance spectral intensity of red sandstone samples decreases, confirming the sensitivity of spectral characteristics to acidification. The proposed fusion model achieves an accuracy of 95% in detecting the acidification degree of red sandstone, surpassing independent RF, KNN, and SVM models. This provides a valuable reference for non-destructive and real-time monitoring of rock engineering stability affected by acidic groundwater intrusion.