Application and effectiveness analysis of grouting reinforcement technology in building tilt correction

Abstract

This study presents an AI‑enhanced integrated framework combining theoretical analysis, laboratory experiments, and field validation for correcting building tilt through grouting reinforcement. The methodology encompasses physics‑informed machine learning for slurry diffusion, artificial neural network‑based rheology, and hybrid PSO‑BPNN settlement prediction. Low‑pressure grouting tests with varying compositions (raw ash 0.25‑0.35, sand 0.3, soil 0.35‑0.5) and lime pile tests were conducted. Field implementation on a 14‑story residential building (initial inclinations: 1.77‰ east‑west, 7.09‰ north‑south) used systematic grouting (42 mm holes, 3‑6 m spacing, 0.8‑1.5 MPa pressure) with 14‑day monitoring. Results show foundation bearing capacity of 320 kPa (100% above design), tilt rates reduced below 3‰, and settlement controlled under 15 mm/day. Bayesian optimization retrospectively identified optimal parameters (depth 9.8 m, raw ash 0.32). All AI/ML techniques were applied retrospectively; the field correction itself used conventional monitoring and control. However, this digital twin framework was not operational during the field correction; it is presented as a future research direction. A digital twin (GeoMCP) framework is proposed for future adaptive control. The close correlation between AI‑enhanced predictions and field performance validates this scientifically grounded, generalizable methodology for tilt correction in buildings of up to 14 stories.