Abstract: Urban emergency flood control boxes have been increasingly used as modular deployable facilities for temporary flood defense. Compared with conventional sandbag barriers, such boxes can be assembled more efficiently. Depending on the availability of materials at the site, the box may be filled with water, sand, or soil. Different filling media alter not only the self-weight and anti-sliding stability of the system, but also the internal lateral pressure transmitted to the walls. As a result, the deformation pattern and stress distribution may vary substantially under different retained water depths. Existing studies have mainly focused on hydraulic performance or on single filling conditions, whereas the coupled influence of filling medium and water-retaining conditions on mechanical response evolution has not been fully clarified. To address these issues, this study investigates the mechanical response evolution and structural optimization of a polypropylene flood control box filled with different media. The prototype polypropylene box had dimensions of 900 mm × 650 mm × 600 mm. In the numerical model, the box material was treated as an elastic solid with corresponding density, elastic modulus, Poisson’s ratio, and allowable stress. The external load acting on the upstream face was represented by the combined action of hydrostatic pressure and flow-induced hydrodynamic loading. The internal load of the filling medium was simplified as equivalent lateral pressure. For water filling, the pressure was assumed to follow a hydrostatic distribution. For sand- and soil-filled cases, reduced lateral pressure coefficients were introduced to account for the fact that the internal side pressure generated by granular media is lower than that of an equivalent liquid column with the same density. The simulations were conducted under a series of retained water depths from 0 to 0.6 m in order to reveal the evolution of deformation, strain, and stress. The reliability of the numerical model was verified by flume experiments. The flood control box was installed along the centerline of a glass flume, and strain gauges were arranged on the upstream face to record structural responses under stepwise increases in retained water depth. The measured strain data were converted into stress values and then compared with the numerical predictions under the same loading conditions. The comparison showed good agreement in both magnitude and variation trends. The first part of the analysis focused on anti-sliding stability. The results show that the density of the filling medium plays a decisive role in the required filling height. For a given retained water depth, higher-density media provide greater self-weight and frictional resistance, thereby reducing the minimum filling height required to satisfy anti-sliding conditions. At the representative retained water depth of 0.6 m, the corresponding filling heights adopted in this study were 0.74 m for water, 0.46 m for sand, and 0.37 m for soil. The numerical results further reveal that the structural response of the flood control box exhibits a clear staged evolution with increasing retained water depth. When the retained water depth does not exceed the filling height, the internal lateral pressure provided by the filling medium remains dominant over a large portion of the wall, and the box mainly bends toward the downstream side. Under this condition, the maximum deformation and peak von Mises stress are concentrated near the bottom of the downstream face, and their magnitudes change only slightly as the water depth increases. In contrast, once the retained water depth exceeds the filling height, the upper part of the box enters a partially unfilled state. The external water load on the upstream face then becomes dominant in the upper zone, causing the direction of the resultant lateral load to reverse. The critical region shifts from the bottom of the downstream face to the bottom of the upstream face, and both deformation and stress increase sharply. Therefore, the most unfavorable working condition is not simply associated with a high water level itself, but with the combination of high retained water depth and insufficient filling height, which produces a pronounced load imbalance between the outer and inner sides of the box. Taking the soil-filled case as an example, the mechanism of response transition can be interpreted through the distribution of net lateral pressure. When the retained water depth is lower than or equal to the filling height, the lateral pressure generated by the filling medium exceeds the external water pressure over most of the wall, so the wall is pushed outward and the downstream face remains the main tensile side. When the retained water depth becomes greater than the filling height, the increase in external water pressure, together with the absence of counteracting internal pressure in the upper zone, makes the upstream face the new critical side. This transition is accompanied by an abrupt increase in both maximum deformation and stress, indicating a marked deterioration in structural safety. Based on this mechanical diagnosis, four structural optimization schemes were proposed and compared: thickening the top region, thickening the bottom region, increasing the depth of the stress-relief groove, and reducing the groove width. The comparison shows that all four schemes provide some degree of local improvement, but their effects are not equally stable across different media. Among them, the bottom-thickening scheme, denoted as P2, performs best. Under a retained water depth of 0.6 m, the peak von Mises stress for the original box was 12.859 MPa in the water-filled case, 13.625 MPa in the sand-filled case, and 13.899 MPa in the soil-filled case. After adopting the P2 scheme, the corresponding peak stresses were reduced to 7.990, 9.355, and 9.576 MPa, respectively. The stress reduction exceeded 30% in all three cases, showing that targeted reinforcement of the bottom region is an effective and robust optimization measure. In summary, this study clarifies the coupled effects of filling medium and retained water depth on the anti-sliding stability and structural safety of flood control boxes. These findings provide a mechanical basis for rational medium selection, structural design refinement, and engineering application of modular flood control boxes in urban emergency flood-fighting practice.