Experimental Study on the Mechanism of Overtopping Failure and Breach Development in Homogeneous Earth Dams

Abstract

According to statistics, between 1954 and 2021, China experienced 3558 dam failures in reservoirs, with flood overtopping accounting for 51.04% of these incidents. Once an earth-rock dam fails, it not only directly threatens the lives and property of surrounding residents and disrupts normal living order, but also damages infrastructure such as farmland, transportation, and power systems, resulting in enormous economic losses. To investigate the mechanisms of overtopping failure and breach evolution in homogeneous earthen embankments during flood seasons, this study conducted seven sets of laboratory model tests with the Changkai Embankment in Fuzhou City, Jiangxi Province, as a prototype. The tests considered various operational conditions, including different crest widths, embankment heights, channel water depths, and river flow velocities. The test results are as follows: Overtopping failure of earth embankments can be categorised into three distinct stages. The breach formation process can be categorised into three stages: vertical erosion (stage I), breach expansion (stage II) and breach stabilisation (stage III). River water levels and inflow rates were identified as pivotal factors influencing the final morphology of the breach and the flow velocity within it. Conversely, the height of the dike was found to have little influence on the shape of the breach and the flow velocity. The breach width ranges from 6 cm to 12 cm. An increase in water depth, corresponding to a greater difference in water levels on both sides of the river, has been observed to result in a deeper breach and faster widening rate. Elevated water levels have been shown to increase the potential energy of the water, which is subsequently converted into greater kinetic energy during breach formation. This, in turn, increases the flow velocity at the breach. However, a negative correlation has been observed between inflow velocity and flow at the breach. This paper combines the material properties of the embankment to discuss the overtopping failure mechanism and the breach evolution law of homogeneous earth embankments. This provides a basis for preventing and controlling embankment failure disasters.

1. Introduction

The embankment is the last line of defence against flooding and is a key part of China’s flood control infrastructure. According to statistical data, by the end of 2022, a total of 330,600 km of grade 5 embankments and above had been constructed across the country, protecting 682 million people and 629 million mu of cultivated land. However, the majority of embankment projects were constructed in the early years of the People’s Republic of China. The limitations imposed by the prevailing level of construction technology, in addition to more than half a century passing, have resulted in embankment projects experiencing a succession of potential hazards [1,2]. Given that the country has experienced some of the world’s most severe flooding, it is crucial to understand the mechanisms and progression of dike breaches. This knowledge is essential for effectively implementing dike engineering measures to mitigate the impact of such disasters [3,4].

Researchers at home and abroad have conducted systematic research into the mechanisms of embankment bursts and disaster prevention and control, achieving significant theoretical and technical progress [5]. In the context of engineering practice and physical problem analysis, integrating numerical calculations with visual representations has been shown to be a highly effective approach [6]. In numerical simulation, Wang et al. utilised a two-digit model to study the evolution process of a dam-break flood in a specific area [7]. The inundation data acquired during the progression of the dam-break flood were analyzed. Salamtalab et al. proposed a numerical model based on the shallow water equations [8]. A two-dimensional finite element framework was used to solve the problem, and an enhanced approximate HLLC Riemann solver was implemented to accurately treat the discontinuity in the water flow during the dam-break process. Qiu et al. constructed a coupling model of fluid dynamics and the discrete element method to accurately simulate the coupled motion characteristics of fluid and particle phases in a dam-break flood [9]. A variety of typical flood evolution modes were observed as a result. Akgun et al. utilised Flow 3D Hydro software (Flow3D Version 10.0) to conduct 2D and 3D simulations of dam breakage [10]. They then made a systematic comparison between the two models in terms of their differences and the applicability of each model in simulating the evolution of dam break flow. Aiming et al. used the Volume of Fluid (VOF) method to carry out a three-dimensional numerical simulation of a dam break [11]. This provided support for the quantitative analysis, prevention, and control of dam break disasters. With the rapid development of artificial intelligence, researchers have also developed a series of numerical models to study dam-break flow in more depth [12]. Seyedashraf et al. employed machine learning techniques to model and analyze the hydrodynamic behavior of dam-break flows, including variations in water level and shock wave propagation [13,14]. Additionally, Vosoughi et al. and Nguyen et al. employed neural network learning to predict the water level in dam-break flows and gain insight into their characteristics [15,16].

Despite considerable progress in recent years in the numerical simulation of dike bursts, such methods rely on specific theoretical frameworks, such as shallow water equations and continuum mechanics theory, or simplified assumptions, such as ignoring microscopic particle migration and the effects of unsaturated seepage on the dam [17,18]. The reliability of the results must be verified by physical tests, which cannot replace the fundamental role of physical testing in revealing the burst mechanism in depth. Due to the scarcity and sparseness of field data on historical events, and the high level of uncertainty associated with it, laboratory testing of dam-break flow is widely considered a valuable benchmark for verifying numerical models [19]. In the field of physical experimentation, Lu et al. [20] monitored surface and internal deformation fields throughout the entire process of water storage, operation, and overtopping failure in homogeneous earthen dams using indoor small-scale model tests. They revealed the temporal patterns of displacement changes, providing theoretical support for failure mechanisms. Chen et al. [21] simulated induced breaches in natural dams through 64 flume tests, analyzing the effects of dam material permeability and riverbed type on breach patterns and downstream landforms. Yang et al. [22] investigated the influence of sand content in bimodal materials on the breach process and downstream landforms for landslide dams with erodible beds through eight flume tests, proposing a new breach pattern. Their research focused on heterogeneous bimodal material landslide dams, with sand content as the core variable, emphasizing downstream landform evolution. In contrast, this study centers on homogeneous earthen dams, examining how operational parameters influence breach morphology and velocity evolution. Gao et al. [23], based on 1:100 large-scale model tests and FLOW-3D simulations, revealed the seepage failure mode (flowing soil) and saturated-line variation patterns of tailings dams under flood conditions, quantifying the downstream inundation extent and flow velocity distribution post-breach. Wang et al. [24] deduced an approximate analytical solution by considering the friction effect of levee break flow in a horizontal triangular channel through indoor physical experiments. This revealed how river channel morphology regulates the motion characteristics of the break front. Xu et al. [25] conducted a systematic analysis of the landslide dam failure mechanism. This was achieved by carrying out multiple sets of laboratory tests, which made it possible to quantify the strength of the various influencing factors.

However, existing physical model experiments (e.g., Chen, Yang, Gao, Xu et al.) have focused on heterogeneous natural dams, landslide dams, and tailings dams, without delving into failure modes beneath homogeneous earth dams. While Lu et al. did not consider overtopping failure beneath homogeneous dams, their considerations primarily centered on the influence of deformation factors. In a comprehensive comparison, none of the physical experiments systematically designed external factors such as water level, dam height, or inflow rate as core independent variables. Instead, they primarily emphasized intrinsic factors like material properties, structural types, and deformation fields.

In light of the above research deficiencies, this paper adopts physical model testing as its core methodological approach. Data is collected using high-precision test instruments, including a pore water pressure gauge and an earth pressure sensor. Comparative analyses are conducted on the vertical expansion rate of the breach, velocity distribution in the breach section, dynamic changes in the earth pressure of the dam body, and the response law of pore water pressure under different embankment heights, initial water levels, and inflow conditions. Furthermore, the dynamic response mechanism of each control factor throughout the levee breach initiation, development and stability process is systematically explored. Combining the test results with the internal mechanism of a dike breach provides a comprehensive understanding of the underlying processes. This serves as the theoretical basis for the design of pre-disaster prevention and control measures in dike projects. Additionally, it provides technical support for formulating emergency response plans in the event of a breach disaster.

2. Test Model and Methodology

2.1. Prototype Background

The experiment uses the Changkai Embankment in Fuzhou City, Jiangxi Province, as its prototype. The Changkai Embankment spans 81.8 km, with an average embankment height of 5.5 m, an average elevation of 38.54 m, and a crest width of 10 m. It protects a total area of 100 square kilometers, encompassing 140,000 mu of farmland and a permanent population of approximately 145,000 within its protected zone. On 21 June 2010, the Changkai Embankment was struck by a once-in-50-year torrential downpour. Sustained heavy rainfall caused a breach in the embankment section between Pile Number. 32 + 923 and Pile Number. 33 + 270. Initially measured at 60 m wide when formed on the 21st, the breach rapidly widened laterally due to continuous water erosion and embankment soil collapse, expanding to 348 m by the 22nd. After days of preparatory work and implementation, the breach was successfully closed and sealed at 18:00 on the 27th. The breach affected an area of 85.5 km², with the deepest inundation reaching 34.54 m. The total reservoir capacity was approximately 296 million cubic meters. Nearly 100,000 affected residents were forced to evacuate, resulting in severe economic losses and widespread social impact.

The analysis posits that three principal factors have been instrumental in causing the breach to occur. Firstly, the flood is no longer a matter of historical significance. Subsequent to the main flood season, there has been a continuous historical rare heavy rainfall in the city, resulting in a rapid rise in the water level of the Fuhe River. The high precipitation levels and the consequent saturation of the soil in the early stage are indicative of the substantial water content of the soil. In the subsequent stage, the precipitation primarily results in runoff. The occurrence of heavy rainfall resulted in a flood of unprecedented magnitude in the Fuhe River Basin, with the flood frequency attaining a 50-year milestone. However, the current flood control standard of the Changkai embankment is less than once in 20 years, and it is difficult to resist the super-standard flood. Secondly, the soil quality of the embankment foundation is poor. The embankment body at the breach of the dam is composed of fine sand, while the embankment foundation consists of a sand and gravel layer. The structure is characterised by a single, highly permeable component. Due to the early commencement of the flood season, the water level has been consistently high, resulting in the long-term immersion of the embankment, thereby causing its softening. Thirdly, the location of the dam breach corresponds to the point of head-on impact. The breach embankment section is located on the concave bank of the confluence of the dry port and the main stream of the Fuhe River. Following the confluence, the mainstream of the Fuhe River directly impacts the embankment toe, thereby causing the underwater bank toe to be emptied and steepened. The slope of the embankment toe is characterised by softness and saturation, and the shear strength is found to be low. It is evident that structures may be susceptible to collapse when subjected to the forces of elutriation resulting from elevated water levels and wind waves.

2.2. Test Equipment and Materials

The experiment was conducted within the glass tank of the Water Conservancy Experimental Center of Nanchang University. The glass tank utilised in the experiment had dimensions of 3 m in length, 0.5 m in width, and 0.5 m in height. The component parts of the system under consideration are as follows: firstly, a water storage tank; secondly, a river channel section; thirdly, a breach; fourthly, a water outlet section; and fifthly, a baffle. The river section measures approximately 0.5 m in length, with its width varying according to different test conditions. The experimental glass sink is presented in Figure 1.

The primary function of this test is to simulate the process of embankment failure. In order to ensure that the internal friction coefficient of the sediment particles utilised in the test is comparable to that of the prototype particles, the test soil materials are configured in accordance with the particle gradation of the soil samples obtained from the Changkai embankment. The soil material with a maximum particle size of 20 mm and a median particle size of d50 = 0.75 mm is selected for the construction of the main particles of the test embankment. As shown in Figure 2, the gradation curve is analogous to the soil composition observed at the Changkai embankment site.

2.3. Test Scheme and Similarity Design

In this experiment, one side of the transparent glass trough is employed as the river channel, and the water level is elevated by the baffle. The location and size of the embankment are arranged under different test conditions, and the DV camera, positioned parallel to the glass trough, is set up for video recording for subsequent conclusion analysis. Following the conclusion of the arrangement process, the test is initiated. The water level on one side of the river channel is elevated by the baffle, and the water level is measured using the transparent measurement sticker. At the designated water level for each operational condition, the inflow is gradually introduced, and the velocity data at the breach is measured using a velocity meter that has been installed in advance. Subsequent to the test, a video recording is utilised for comprehensive analysis.

The model and the prototype demonstrate geometric and gravitational similarity. In accordance with the prevailing parameters of this experiment, the horizontal geometric scale has been designated as 1:100. As the embankment height is employed as a variable, there are divergent scales, as shown in

Table 1. Geometric dimensions of the prototype and test embankment.

Parameter	Changkai Embankment	Test Embankment 1	Test Embankment 2	Test Embankment 3	Horizontal Sale
top width (cm)	1000	10	10	10	1:100
Height (cm)	550	15	20	25	/
Vertical Scale	/	1:36	1:28	1:22	/

.

The reliability of the conclusion of this test is confirmed by the actual measurement and analysis of the data. The parameters that were measured during the course of this experiment included the shape of the breach, the flow velocity, and the water level. The following instruments are utilised: The LGY-III multifunctional intelligent flow meter and DV camera.

This study employs the Sony HDR-CX450 as the video capture device and the LGY-III multifunctional intelligent flow velocity meter as the flow velocity acquisition device. The DV camera is positioned on the side parallel to the glass flume and on the downstream side to accurately capture details during embankment overtopping and failure. Its sensor size is 1/5.8 inches; it supports 30× optical zoom, enabling detailed capture of dam failure events; Supporting 2.01–3.00 megapixels.

The LGY-III multifunctional intelligent flow velocity meter (Φ15 mm) represents an advanced, practical technology widely adopted in the water conservancy industry for scenarios such as hydraulic model testing. It meets the multi-point flow velocity monitoring requirements of this study, with a measurement range of 1–300 cm/s (covering the flow velocity range in this study); Start-up velocity: ≤1 cm/s (enables precise capture of low-velocity operational data); Sampling interval: Freely adjustable from 1 to 99 s (set to 1 s for this study); Velocity calculation model: Utilizes the standard propeller-type velocity calculation formula.

$$V=K\times {\displaystyle \frac{N}{T}}+C$$

(1)

where C represents the measured flow velocity, K denotes the flow sensor calibration coefficient, T indicates the set sampling time; N represents the propeller rotation speed during the sampling period; where K = 2.88 in this test; and C signifies the correction factor, C = 1.00. The instrument is positioned as shown in Figure 3.

In this experiment, the influence of dam elevation, river water level and inflow rate on levee breach is analysed under the condition of the same top width. This analysis is based on data regarding breach shape, water level and flow rate. The results of this analysis will deepen our understanding of the mechanism of levee breach, and facilitate the prevention of such breaches, thus avoiding loss of human and financial resources. The exploration of the mechanism of the breach is also beneficial in terms of the subsequent rescue operation, with the aim of minimising any loss. The experiment comprises a treatment group and six additional groups, thus yielding a total of seven distinct test conditions. The soil samples utilised in the test were obtained from the soil of a nearby embankment site. The specific test conditions are delineated in

Table 2. Test conditions of embankment breach.

Test Condition	Top Width (cm)	Height (cm)	Water Level (cm)	Inflow Rate (cm3/s)
1	10	15	8	60
2	10	15	10	60
3	10	15	12	60
4	10	20	8	60
5	10	25	8	60
6	10	15	8	80
7	10	15	8	100

.

3. Test Results and Analysis

3.1. Evolution of Breach Phenomena

In this experiment, conditions 1, 2 and 3 constitute the water level treatment group, conditions 4 and 5 constitute the embankment height treatment group, and conditions 6 and 7 constitute the inflow treatment group. The analysis of the data from each treatment group has enabled the identification of a discrepancy between the breach process and the breach evolution. As shown in Figure 4, the failure of each group occurred at different times.

As demonstrated in Figure 4, at the commencement of the test, the water level did not attain the height of the embankment, and the embankment itself remained stable. The seven test conditions of the test undergo three stages to complete the breach. The phenomenon under investigation can be summarised as follows: approximately 90 s after test commencement, the water level was observed to coincide with the embankment crest, with overtopping imminent; at around 120 s after the test began, the embankment gradually failed as a result of overtopping. At this time, a minor breach of the embankment was observed, and a narrow erosion trough gradually formed at the summit of the embankment. The flow velocity increased in comparison with the overtopping. The collapse of the soil on both sides of the embankment was due to the effect of water flow. At this time, the breach mainly developed in a downward direction, and the overall vertical erosion phenomenon was presented. At approximately 270 s, the vertical erosion of 90 s leads to a deeper and narrower breach shape. At this time, due to the erosion of the water flow, it can be observed that the soil on both sides of the breach begins to collapse downward. At this juncture, the flow rate reaches its maximum stage within the test. In light of the escalating phenomenon of water erosion, the morphology of the breach undergoes rapid and unpredictable changes. The breach as a whole demonstrates a rapid expansion phenomenon; however, when the test reaches 300 s, the breach gradually ceases to change. At this juncture, the breach has reached its culmination, and the final breach shape is now discernible. The flow velocity and the breach flow undergo a gradual decrease to zero, and the breach tends to be stable as a whole. It is evident from the experimental data that the test groups 2 and 3 demonstrate a higher velocity of response to elevated water levels in comparison to the treatment group 1. The occurrence of overtopping failure is observed to precede the initiation of the test by 120 s, and the rate of breach widening is also enhanced. In the treatment group 4 and 5, it was observed that differing embankment heights did not have a significant effect on the breach, and there was no significant change in the experimental phenomenon when compared with the treatment group 1. The phenomenon of 6 and 7 in the experimental group was similar to that in the treatment group 1, but the break widening was increased, and the widening time was also increased.

During the course of the experiment, the infringement of the seven groups of test conditions is exhibited in three stages. Initially, the infringement is minor, and the flow velocity and water depth undergo negligible changes. At this time, the phenomenon is characterised by vertical erosion and an inclination to spread horizontally. In the second stage of the collapse, the soil layer began to lose stability due to an increase in depth. The collapse mouth underwent a gradual expansion to both sides, and the slope surface away from the river channel exhibited a significantly faster rate of erosion than the river side, due to the action of the water flow. The flow rate and flow rate of the collapse mouth increased and reached maximum values, and the river water level decreased significantly. In the third stage, the discharge of a substantial volume of water into the submerged area results in a decline in the flow and velocity of the breach. This leads to a gradual cessation of breach widening and erosion, thereby stabilising the breach’s shape.

3.2. Influence of Key Factors on Breach Characteristics

3.2.1. Breach Morphology

The shape of the breach is described through the projection of images during, after and at the end of the test. The discrepancy analysis demonstrates the impact of varying operational conditions on the configuration of the breach. This study employed a DV camera to capture visual images of the entire process of model embankment failure. When processing breach width data, the scale placed during the experiment served as the reference point, enabling the determination of breach widening at different time points through simplified processing.

The shape of the breach in the water depth treatment group is shown in Figure 5.

As demonstrated in Figure 5, the water depth is identified as the primary variable, with the final depth of the breach from large to small being the secondary variable. The tertiary variable is the third condition, the secondary condition, and the primary condition. It has been observed that the rate of development of the breach is increasing gradually. The third test condition is the most rapidly expanding, followed by the second, and the first is the slowest.

As demonstrated in Figure 6, there is a lack of statistical significance between the three test conditions of the first, fourth and fifth variables of the test conditions, as well as the height of the embankment and the final depth of the breach. Furthermore, it was observed that the development speed of the breach was essentially unchanging. However, due to the occurrence of force majeure, even if the depth and width remain essentially unchanged, the final breach shape will vary considerably.

As shown in Figure 7, the variables of test conditions one, six and seven correspond to the inflow, and the final depth of the breach is represented by c, b and a, from large to small. The development speed of the breach is increasing step by step; the width of the breach gradually increases from a, b, c, and develops violently from 2–4 min to form the final breach shape.

The projection of images during, after and at the conclusion of the test facilitates the depiction of the shape of the breach and the subsequent difference analysis. The influence of differing test conditions on the geometry of the breach is well documented. The test is recorded by the inflow, and the breach phenomenon commences gradually as the water flows over the dam. A comparison of the widening of the breach in the test group of Figure 5 reveals that as the water depth in the river increases, that is, as the water level difference across the river is amplified, the breach becomes deeper and wider at a faster rate. This finding suggests that the water level difference is a primary factor influencing the breach’s width. In circumstances where the water level difference between the two sides of the embankment is substantial, the potential energy generated by the water level is increased. In the event of a breach, the potential energy is converted into kinetic energy, resulting in the rapid development of breach widening. In circumstances where the water level is elevated, the seepage pressure exerted by the embankment soil is substantial. This has a detrimental effect on the stability and compactness of the embankment. Following the occurrence of a breach, the phenomenon of seepage deformation is exacerbated. As demonstrated in Figure 6, the relationship between breach widening and embankment height appears to be inconsequential. There is no significant difference in breach shape and breach widening rate observed in this group of experiments, and the experimental phenomenon is basically consistent with the three stages of breach. The analysis of the breach process and morphology of the three groups of test conditions depicted in Figure 7 reveals that as the inflow rate increases, the breach width also increases. It has been demonstrated that an increase in inflow results in an increase in breach duration and scouring time on both sides of the embankment.

3.2.2. Breach Flow Velocity

Once the test is complete, the flow rate for each test condition is extracted and processed by computer to eliminate invalid data. A comparison of the flow rate under various test conditions is then made, as shown in Figure 8.

As shown in Figure 8, the flow velocity reached its maximum approximately two minutes after the break was initiated, under the first, second and third test conditions. Thereafter, the flow velocity gradually declined alongside a reduction in the breach’s flow rate. The maximum flow velocities recorded for the three test conditions were 83.15 cm/s, 89.15 cm/s and 92.20 cm/s, respectively. There is a clear upward trend.

As shown in Figure 9, the variable under consideration is the embankment height in three test conditions (test conditions 1, 4 and 5). The flow rate gradually decreases in proportion to the breach flow rate, ultimately reaching zero. The maximum flow rates observed for the three distinct test conditions are 83.15 cm/s, 83.20 cm/s and 80.85 cm/s, respectively. A thorough examination of the data reveals no substantial increase or decrease, and no discernible trend is apparent across the three datasets.

As shown in Figure 10, the variable under consideration is the flow in three test conditions (test conditions 1, 6 and 7). The inflow rate decreases from 83.15 cm/s to 72.95 cm/s.

The analysis indicates that the velocity of the breach is influenced by three factors: the water level, the height of the embankment and the inflow. As Figure 8 shows, there is a direct correlation between the breach’s flow velocity and the water level. An increase in the water level has been shown to be accompanied by an increase in flow velocity. The elevated water level during the breach increases the difference in water level across the embankment, thereby increasing the potential energy of the water flow. In the event of a breach, more kinetic energy is converted from potential energy, and the flow velocity at the breach increases. As shown in Figure 9, embankment height has little effect on flow velocity at the breach. As shown in Figure 5, there is no discernible difference in the morphology of the breach, and a substantial decline in flow velocity occurs during its formation. Consequently, the flow velocity remains unaltered. As shown in Figure 10, there is a negative correlation between the inflow rate and the flow velocity at the breach. An increase in the inflow rate has been shown to result in an increase in the erosion time of the breach. Consequently, the embankment is unable to resist erosion from the water flow, resulting in rapid widening of the breach. This, in turn, widens the shape of the breach and gradually decreases the force of the water flow. Concurrently, a decline in flow velocity has been observed. It can therefore be concluded that the water level and inflow rate are significant factors in controlling the breach’s flow rate.

3.3. Energy-Driven Mechanism of Embankment Breach

The observed effects of water level and inflow velocity on embankment erosion in this study fundamentally involve the conversion of potential energy into kinetic energy, which drives the erosion process. This potential energy is primarily gravitational, stored as head pressure determined by both water level height and water volume—higher water levels yield greater gravitational potential energy. Concurrently, inflow velocity continuously replenishes the water body, sustaining the supply of potential energy and providing the fundamental energy basis for ongoing embankment erosion.

When water overflows the crest to form an initial breach channel, gravitational potential energy is converted into kinetic energy through gravitational work. According to the definition of the Froude number: $F_r={\displaystyle \frac{v}{\sqrt{gH}}}$, Flow velocity v is positively correlated with hydraulic head H. When this kinetic energy impacts the embankment breach and initiates scouring, it ultimately leads to the stripping and transport of embankment soil through erosion. Increased inflow not only amplifies hydraulic energy by expanding the breached area but also continuously replenishes water volume, allowing erosion effects to accumulate. This manifests as a positive correlation between breach scouring rate and inflow velocity. Consequently, water level and inflow velocity are key factors governing breach scouring rates.

4. Discussion

The breach process mechanism is explored through the analysis of multiple breach test sets, with particular reference to breach morphology characteristics and flow velocity data. The seven groups of test conditions have been categorised into three distinct stages: vertical erosion (stage I), breach expansion (stage II) and breach stability (stage III). Stage II is the primary stage of breach development and sees significant changes in velocity and widening velocity. As shown in Figure 11.

4.1. Vertical Erosion Stage (Stage I)

After the test commenced, the inflow rate into the reservoir reached the river water level, causing water to overflow the crest of the embankment. As the inflow rate continued to increase, the water flowed toward the downstream side. The overflowing water eroded the sediment at the weak points of the embankment, forming cracks that became the initial breach. As inflow volume into the channel continued to rise, the water eroded a narrow scour channel on the downstream face of the embankment. At this stage, the flow velocity was too low to cause extensive scouring and rapid breach expansion. Increased moisture content in the sediment enhanced inter-particle friction, making it easier for water to dislodge particles—laying the groundwork for Phase II. At this stage, flow velocity and sediment friction mutually constrain each other, each increasing as the other decreases. As particle friction diminishes while inflow volume continuously rises, flow velocity gradually increases, enhancing the traction effect on sediment at the contact surface. Consequently, the breach develops downward. During this phase, the breach progresses slowly, with sediment being entrained by the flow and forming fan-shaped deposits on the leeward side, primarily manifesting as vertical erosion. Field observations indicate that during the breach of the Changkai embankment, a funnel-shaped breach first formed on the outer side, accompanied by multiple cracks on the inner side. Subsequently, the upstream slope collapsed, causing the entire embankment to slide downward by 4–5 m. The slid embankment was rapidly overtopped by the flow. This period represents the optimal window for breach closure. At this stage, a stable flow channel has not yet formed, the breach exhibits a “V”-shaped cross-section, erosion transitions from particle detachment to channelized cutting, and the breach rate gradually accelerates. Protective measures for this stage include constructing a temporary water-retaining embankment upstream of the breach. Simultaneously, filling the breach channel with reinforced cages containing embedded boulders, reinforced with sandbags, can slow channel erosion rates within 1–2 h. This buys time for subsequent permanent closure. Missing this window increases closure difficulty once the breach enters the expansion stage.

4.2. Breach Expansion Stage (Stage II)

As the breach widens, both flow rate and velocity at the breach site increase, accelerating scouring and deepening the scour channel. At this stage, the breach transitions from primarily vertical erosion to lateral erosion. As water scours the soil on both sides of the breach and permeates through the soil, the initial phase resembles Stage I. However, as the depth increases and gravitational forces on the soil intensify, the soil on both sides becomes unstable and collapses downward. rapidly widening the breach. Once the breach widens, a large volume of water from the river channel surges into it, further widening and deepening the breach, accelerating the breaching process. As massive discharge flows through the breach, the river channel water level drops rapidly. The discharge volume far exceeds the inflow, causing peak flow velocity. The downstream face of the embankment sustains continuous impact, washing away all sediment deposited during Stage I and soil collapsed from the breach. Only when flow velocity and volume gradually decrease does breach development slow, forming a wide and deep breach morphology. This phase represents the breach expansion stage, the second phase of embankment failure. In the original embankment, this period constituted the breach outbreak phase, where the breach widened to 60 m within just 30 min. The flowing water scoured and entrained massive amounts of sediment, causing severe embankment damage, localized collapses, and pronounced, violent destruction. Given the characteristics of this stage, effective sealing of the breached dam body is not feasible. Engineering efforts should focus on personnel evacuation and disaster loss mitigation. Methods such as downstream channel diversion and reservoir peak-shaving operations can be employed to reduce inflow volume, slow the rate of breach expansion, and lower flood peak levels.

4.3. Breach Stabilization Stage (Stage III)

The third stage of breach development is the stabilization stage. During this phase, the breach morphology stabilizes, with both vertical and lateral expansion ceasing, representing the final breach configuration. At this point, the lowered water level within the channel reduces the water level difference between the upstream and downstream sides of the embankment. The hydrodynamic forces become insufficient to sustain flow velocity and widen the breach further. With flow velocity slowing below the critical velocity for the embankment material tested, extensive sediment deposition occurs behind the embankment. This deposition provides feedback support to the embankment structure. By the final stage of breach, flow rates at the breach nearly equalize with inflow rates until the breach process concludes. Flow velocity at the breach gradually approaches zero. The embankment reaches a state of dynamic equilibrium, marking the complete termination of the breach process and full stabilization of the breach morphology. In the prototype scenario, the breach at Changkai Embankment widened to 347 m and formed a scour pit over 4 m deep. The flow cross-section stabilized, the breach ceased expanding, and its shape approximated a trapezoid. The discharge rate gradually decreased, marking the tail-end phase of the breach where destructive forces weakened. This represents a critical window for post-breach remediation. Reinforcement should be applied based on the breach morphology, with structural optimization in vulnerable areas—such as installing drainage channels to reduce seepage pressure and deploying sufficient sensors for real-time monitoring.

4.4. Engineering Significance

The breached section of the Changkai Embankment was primarily filled with fine silt, exhibiting poor construction quality and inadequate seepage resistance. Prolonged rainfall saturated the embankment soil, reducing its shear strength. Consequently, the Embankment proved unable to withstand the assault of a flood exceeding historical records, leading to the breach. Conducting physical model tests based on this incident not only recreates the core mechanical mechanisms of embankment breaches and replicates the evolutionary patterns during critical stages but also provides robust, actionable scientific evidence for emergency assessments, risk classification, and response decision-making following similar embankment failures. Simultaneously, it offers targeted, safe, and implementable technical guidance for formulating on-site disaster response plans, selecting appropriate technologies, and optimizing operations. This constitutes a crucial technical foundation for enhancing emergency response capabilities and improving disaster prevention and mitigation effectiveness in the event of embankment breaches.

5. Restrictions

This study employed physical model experiments to reveal the influence patterns of water level, embankment height, and inflow discharge on the failure process of earth-rock dams. However, the applicability of its conclusions is constrained by experimental design and objective conditions, presenting the following key limitations.

5.1. Repeatability of the Test

Since the test apparatus simulates the dam breach process, each test condition can only be run once, making it difficult to precisely replicate the results of each test. This limitation prevents the quantification of data variability. The results of a single test can only reflect the breach process under that specific condition and lack statistical representativeness, thus cannot be quantitatively evaluated. Its core purpose is to reveal the qualitative relationship between conditions and breaches, rather than to provide quantitative scale parameters.

5.2. Scale Effect Analysis

This study investigates embankment failure imagery under varying water levels, embankment heights, and inflow rates. The physical model experiments were designed in strict adherence to the principle of geometric similarity. Since the embankment breach process is gravity-dominated, with both flow dynamics and breach propagation influenced by gravity, the Froude number was employed to correlate the head, velocity, and flow characteristics between the prototype embankment and the model. Consequently, the Froude similarity criterion served as the core design principle for the model, ensuring similarity in the breach process across different variable combinations. Three variable-scale ratios were established based on different embankment heights: 1:36, 1:28, and 1:22, with a crest width scale ratio of 1:100. Due to scale effects and model materials, absolute erosion rates cannot be directly scaled using similarity ratios. However, since the experiments focus on the macroscopic phenomena and patterns of embankment failure rather than precisely replicating erosion parameters, this does not conflict with the core research objectives of this study.

Due to inherent constraints of the physical model and material characteristics, this study exhibits Reynolds number distortion and permeability coefficient errors. The particle size distribution of the prototype soil and its hydraulic properties cannot be scaled proportionally within the model, resulting in a lower model Reynolds number compared to the prototype. However, since embankment failure is primarily influenced by gravity, the Reynolds number distortion affects only the local microscopic scale and does not compromise the analysis of macroscopic patterns. To minimize existing proportional distortions, this physical model experiment selected soil types and particle sizes similar to the embankment prototype and controlled soil moisture content. This ensured consistent core conditions when the model was subjected to hydraulic failure.

6. Conclusions

Based on the model of the Changkai Embankment, this study conducted indoor model experiments to record and analyze the evolution of overtopping breach morphology and the progression of water level and flow velocity during breaches in homogeneous earthen embankments under various operating conditions. It qualitatively examined the influence patterns of different water levels, embankment heights, and inflow rates on embankment breaches. The expansion rate of the breach width is positively correlated with water level and inflow volume: higher water levels store greater gravitational potential energy. When the embankment breaches, the conversion of this potential energy into kinetic energy increases in both magnitude and efficiency. Simultaneously, the continuous replenishment from inflow sustains the dynamic conversion process between potential and kinetic energy. This maintains the water flow’s strong scouring capacity, which persistently acts upon the breached embankment soil. Ultimately, this propels the continuous expansion of the breach width. The following conclusions are derived from this research:

(1)

The process of dike breaching can be categorised into three distinct stages. Firstly, there is the vertical erosion stage (Stage I, 90–120 s), followed by the breach expansion stage (Stage II, 120–270 s). The third and final stage is the breach stabilisation stage (Stage III, 270–300 s). The initial phase of breaching, Stage I, is characterised by gradual vertical erosion. Stage II is characterised by accelerated expansion of the breach and the attainment of peak flow rate. The breaching process culminates in Stage III. At this point, the flow rate has decreased and the breach morphology has stabilised, concluding the breaching process.

(2)

It is evident that the river water level and inflow are pivotal factors in determining the final morphology of the levee breach and the velocity at the breach. However, the height of the levee has been shown to have no significant influence on the morphology or velocity of the breach.

(3)

The results indicate that the width of the breach ranges from 6 cm to 12 cm. Specifically, the greater the water depth (i.e., the greater the difference in water level across the river), the deeper and wider the breach becomes and the faster it widens. In cases of large inflow, the breach is wider. Greater inflow results in a longer breach duration, extended scouring on both sides of the embankment and a larger breach cross-section. In summary, the water level difference and inflow are key factors influencing the breach width.

(4)

In sub-aquatic conditions, the maximum flow velocities observed in the control group were 83.15 cm/s, 89.15 cm/s and 92.20 cm/s, respectively. This suggests a direct correlation between water level and flow velocity: an increase in water level is associated with a corresponding increase in flow velocity. It has been demonstrated that an increase in water level results in a corresponding rise in the potential energy of the water flow. During breaching, a greater proportion of this potential energy is converted into kinetic energy, which increases the flow velocity at the breach. The inflow rate is found to be inversely proportional to the flow velocity at the breach. As the inflow rate increases, the breach widens and the force of the water flow gradually diminishes, resulting in a decrease in flow velocity.