The use of levees is one of the most prevalent and effective strategies for flood protection. However, owing to the ageing of levees, inconsistent reinforcement efforts, and complex geological conditions, hazards such as piping frequently arise during flood seasons, which lead to significant and often irreparable damage. This study investigates backward erosion piping (BEP) in the foundations of double-structured levees via a back-propagation (BP) neural network optimized by a genetic algorithm (GA). The primary contributions of this study include: 1) the construction of a training dataset through numerical simulations of BEP in heterogeneous aquifers and validation of the dataset against laboratory sandbox piping tests to verify its reliability; 2) the extraction of head H and permeability coefficient K data from Groups Ⅱ, Ⅲ, and Ⅳ in the BEP laboratory tests, augmentation of the dataset, and optimization of the GA–BP model to characterize test results in Group Ⅰ, where the results demonstrate that the optimized model more accurately characterizes areas where the K≤1.0 cm/s; and 3) the use of the optimized GA-BP model to characterizes the development of a BEP channel. The results indicate that the model accurately captures the general trends. However, minor discrepancies remain in the characterized channel location and size compared with the actual conditions. In conclusion, this study offers an effective tool for characterizing BEP and demonstrates the potential of the GA–BP network model for practical applications in this field.

To accurately locate the levee seepage inlet, laboratory model tests and Fluent numerical simulations based on the tracer method were employed, and the effects of tracer dosage, seepage scale, and tracer injection point location on tracer transport pattern were analyzed. The results indicate that increasing the tracer dosage significantly raises the peak tracer mass fraction on the wall, shortens the penetration time, and accelerates tracer transport velocity. Enlarging the seepage scale accelerates the tracer loss rate, reduces the peak tracer mass fraction on the wall, and transforms the tracer distribution shape from circular to elliptical, with its major axis aligning along the line connecting the tracer injection point and the seepage inlet. This alignment can be used to infer the direction of the seepage inlet. Reducing the distance between the tracer injection point and the seepage inlet decreases the peak tracer mass fraction on the wall, prolongs the penetration time, and significantly expands the tracer distribution area.

To investigate the impact of ship-generated waves on the erosion characteristics of the bank slope along the Pinglu Canal, a large-scale laboratory flume model testing system was designed to analyze the surface wave pressure variations and hydraulic erosion characteristics of the bank slope under five different conditions: mesh flexible revetment, square permeable concrete revetment, hexagonal permeable concrete revetment, gravel revetment, and unprotected slope. The results show that the temporal variation of the surface wave pressure on the bank slope is periodic, with the unprotected slope exhibiting the maximum negative wave pressure. The gravel revetment, due to its structural irregularity, demonstrates superior energy dissipation performance. Moreover, under high water levels and high ship speeds, a secondary wave peak effect occurs. The surface wave pressure distribution on the bank slope rapidly attenuates on both sides near the static water level. The ship speed has a significantly greater effect on wave pressure and wave height compared to water level changes. Analysis of surface erosion on the bank slope reveals that the unprotected slope mainly experiences three types of erosion forms, namely slope cracking, soil collapse, and slope scouring. Except for the gravel revetment, all other revetments show fine particle precipitation. The internal soil pressure and surface wave pressure exhibit consistent dynamic response characteristics. Based on the overall ecological benefits and erosion control effects, the gravel revetment is the optimal solution.

An improved permeation device was developed for the experimental investigation of the influence of different skeleton particle compositions with the consistent content of fine particle ranging (0.075 to 1 mm) on critical conditions of piping, erosion pattern and loss pattern of particles in the granular soil. The results show that for samples with different particle gradations, the hydraulic gradient and the seepage velocity exhibit a linear relationship before piping occurs, which essentially complies with Darcy’s Law. Skeleton particles in the size ranges of 1-2mm, 2-3mm, and 3-5mm impede the development of piping, with particles in the 1-2mm range demonstrating a stronger blocking effect on piping voids than particles in other two size ranges. For skeleton particles of different gradations, the larger the heterogeneity coefficient, the larger the lower limit of the critical hydraulic gradient for the sample, making it more difficult for fine particles to be mobilized, and thus delaying the onset of piping. Conversely, the influence of the skeleton particle gradation on the upper limit of the critical hydraulic gradient of the sample is relatively minor.

Using a stress-controlled dynamic permeability deformation test system, experiments were conducted on gap-graded cohesionless soil to investigate the suffusion behavior under confining pressures of 30, 60, and 90 kPa. The study explored the suffusion behavior of the soil and analyzed the variation characteristics of fine particle loss, volume change, and critical hydraulic gradient under different confining pressures. The results show that, at lower confining pressures, fine particles are more prone to being lost, leading to increased pore connectivity and a significant rise in seepage velocity. In contrast, higher confining pressures compress soil pores, reducing the continuity of flow paths and causing a significant decrease in the permeability coefficient. At low hydraulic gradients, seepage velocity exhibits nonlinear variations; however, as the hydraulic gradient increases, it gradually becomes linear, generally conforming to Darcy’s law. Increasing confining pressure raises the critical hydraulic gradient required to initiate suffusion, demonstrating a positive correlation between the confining pressure and critical hydraulic gradient. The effect of confining pressure on fine particle loss differs at various stages, but its overall impact on the total amount of fine particle loss is not significant. Under higher confining pressures, the soil particles bond more tightly, resulting in less deformation of the specimens. Under the confining pressure of 30 kPa, the volume change curve exhibits a distinct step-like pattern.

In order to explore the influence of hydraulic gradient on the seepage erosion in soil, fused silica sand was used as the transparent soil material, and the pore solution was configured with No. 15 white oil and n-dodecane in a certain proportion. Based on the seepage erosion system developed by PIV-PLIF(particle image velocimetry and planar laser induced fluorescence), the influence of hydraulic gradient on the flow velocity distribution of pore solutions and the movement of fine particles in seepage erosion damage was explored. The results show that with the increase of the hydraulic gradient, the flow velocity of tracer particles also increases and the irregular motion phenomenon intensifies, and the flow velocity tends to be stable until the sample fails. Under the same hydraulic gradient, the macroscopic flow velocity is slightly larger than the microscopic flow velocity, and the gap between the two is also narrowing with the increase of the hydraulic gradient. When the hydraulic gradient reaches the critical hydraulic gradient, the fine particles in the particle skeleton begin to lose, and the permeability coefficient also begins to change. If the water head continues to increase, a large number of fine particles will be lost, resulting in the continuous development and expansion of the seepage channel, and finally the critical hydraulic gradient of failure will be reached.