Intelligent and Rapid Recognition Method for Moisture Content in Clayey Shield Tunnel Muck Based on VIM-OFE

The rapid and accurate detection of moisture content in clayey shield tunnel muck is a key issue for improving the muck disposal efficiency and reducing environmental impact. In this paper, an intelligent identification method based on VIM-OFE was proposed, and experiments were conducted using three clayey muck samples from different geological strata. A dynamic analysis window was applied to Sample 1 to extract the spectral intervals highly correlated with the moisture content, specifically 1438—1455 nm and 1916—1926 nm. The resulting prediction models achieved high accuracy within these intervals. The study further examined the impact of absorption peak area and depth features on the model performance and compared the proposed method with iPLS, SiPLS and CSMW in terms of predictive accuracy and computational efficiency. Results show that VIM-OFE demonstrated superior prediction performance while maintaining a comparable computation time. Both the single-soil and mixed-soil moisture prediction models were developed, with the mixed model achieving high accuracy in Samples 1 and 3(R²＞0.890), indicating strong generalization capability. Additionally, the influence of particle size and mineral composition on near-infrared(NIR) spectral response was analyzed, and strategies such as outlier optimization and the development of a regional spectral database were proposed to enhance the model robustness and adaptability. The proposed method offers a reliable technical approach for on-site rapid detection of moisture content in clayey shield tunnel muck and shows promising application potential in engineering practice.