Experimental study on mechanical properties of red-bed mudstone under the action of water–rock hydro-mechanical coupling

• A graphic processing method using Hough Transform and convolutional neural networks was proposed. • An experimental visualization platform was built to explore red-bed crack evolution. • Water damaged cracks were observed in samples when the Felicity ratio FR was lower than 1.2. • Landslide early warning systems in red-bed areas should consider multiple factors. Red-bed mudstone, a key engineering material commonly encountered in infrastructure construction in rain-prone regions, exhibits critical instability under water exposure and excavation. Yet research linking its hydro-mechanical (HM) coupling behavior to materials design optimization remains limited, hindering its reliable application. This study aimed to elucidate the HM-coupled mechanical properties and fracture evolution of red-bed mudstone, providing a scientific basis for enhancing its anti-water-instability performance through targeted design. A novel experimental setup integrating digital radiography (DR) and multi-stage loading was developed to characterize materials performance, in which contrast agent angiography visualized fracture progression in cubic specimens under multi-stage loading, with synchronous recording of acoustic emission (AE) signals, stress data, and fluid flow rates. Specialized graphic processing enabled quantitative analysis of damage mechanisms, directly supporting materials design insights. Results showed that water-induced secondary fractures formed when the Felicity Ratio (FR) fell below 1.2, a quantifiable threshold for guiding microstructural adjustments to improve water resistance. This platform offers a tool for characterizing geomaterial performance, while the FR-based insights enable materials design optimization of red-bed mudstone for engineering applications. Such knowledge contributes to devising more effective disaster mitigation strategies for red-bed geological formations.