Safety assessment method and safety assessment system for dike project

Abstract

The invention discloses a safety assessment method and a safety assessment system for an embankment project, and belongs to the technical field of embankment project safety. The method comprises the following steps: acquiring temperature change data of a dike area to be monitored; according to the temperature change data, temperature gradient distribution in the dike is calculated, and the delay influence of the temperature gradient distribution on the seepage field state is analyzed in combination with historical data of the seepage field state; the method comprises the following steps: deploying a distributed sensor network in a key area of a dike, monitoring the change of a seepage field state in real time, and obtaining high-precision seepage field state data; predicting a piping path through a random forest model according to a piping signal extracted from the high-precision seepage field state data in combination with the temperature gradient distribution and the seepage field state data; and on the basis of the prediction results of the temperature change data, the seepage field state data and the piping path, predicting the probability of piping occurrence and the early warning level. By integrating multi-source data, the piping risk can be evaluated more comprehensively, and misjudgment possibly caused by a single index is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of embankment engineering safety, and in particular relates to a safety assessment method and a safety assessment system for embankment engineering. Background Art

[0002] Levees are water-retaining structures built along rivers, canals, lakes, coasts, or along the edges of floodways, diversion areas, and reclaimed areas. They are primarily used to effectively resist flooding during flood seasons, protecting the lives and property of downstream residents and ensuring their normal lives and livelihoods. They are crucial for regional economic development, and therefore, dam safety monitoring has received significant attention and attention. Due to the complex and variable seepage paths along dams, traditional monitoring methods rely primarily on manual patrols, which are inefficient and lack comprehensive coverage.

[0003] Prior art CN115713187A discloses a dam monitoring and analysis system based on multi-sensor technology. This system uses soil moisture sensors, displacement sensors, crack sensors, and temperature sensors to acquire dam data. This system then calculates and analyzes the data to obtain dam status information. This system enables multi-dimensional monitoring of the dam's soil moisture content, seepage line, deformation, and internal leaks. This improves the efficiency of dam inspections and analysis, providing more comprehensive and rapid dam status feedback to management centers and maintenance personnel. Furthermore, this system uses dam status information to regularly and automatically detect dam seepage line anomalies, dangerous moisture levels, landslide, and piping risks, identifying potential dam safety hazards and triggering alerts that are simultaneously sent to management centers and dam maintenance personnel. This enables human-machine interaction, timely monitoring of the dam, and prevention of potential hazards. While this method enables intelligent and real-time monitoring of dams, it only predicts the severity of piping risks. It cannot predict the time window for piping occurrence or further predict future piping changes. Summary of the Invention

[0004] The main purpose of the present invention is to provide a safety assessment method and safety assessment system for embankment projects to solve the problem that the existing technology cannot predict the time window for the occurrence of piping and cannot make further predictions on future changes of piping.

[0005] To achieve the above object, the present invention provides a safety assessment method for embankment projects, comprising the following steps: Obtain temperature change data of the embankment area to be monitored; Calculating the temperature gradient distribution inside the dike based on the temperature change data, and analyzing the delayed effect of the temperature gradient distribution on the seepage field state in combination with historical data of the seepage field state; Deploy distributed sensor networks in key areas of the embankment to monitor changes in seepage fields in real time and obtain high-precision seepage field data; Predicting piping paths using a random forest model based on piping signals extracted from the high-precision seepage field data and in combination with the temperature gradient distribution and the seepage field data; Based on the temperature change data, the seepage field state data and the prediction result of the piping path, the probability of piping occurrence and the warning level are predicted.

[0006] Furthermore, the acquisition of temperature change data of the embankment area to be monitored includes: installing temperature sensors at different depths and positions of the embankment to form a temperature monitoring grid, monitoring the internal and surface temperatures of the embankment in real time, and obtaining temperature change data of the embankment area.

[0007] Furthermore, the historical data of the seepage field state includes seepage pressure, flow velocity and flow rate.

[0008] Furthermore, the delayed effect of temperature gradient distribution on the seepage state is expressed as follows: ; Among them, R(t) represents the cumulative effect, K(τ) is the kernel function describing the effect of temperature gradient, and t is time.

[0009] Furthermore, a distributed sensor network is deployed in key areas of the embankment to monitor changes in the seepage field in real time and obtain high-precision seepage data, including: Install pressure sensors in key areas including the embankment slope, crest, and toe to monitor changes in pore water pressure and determine seepage path and intensity; Flow sensors are installed at different depths inside the embankment to measure water flow speed and direction and identify potential leakage points.

[0010] Furthermore, the piping signal includes an abnormal increase in soil pore water pressure, a sudden change in seepage velocity, or a sharp increase in seepage water volume.

[0011] Furthermore, the input of the random forest model is ; Where X represents the input feature matrix, T represents the temperature gradient distribution data, F represents the seepage field data, and S represents the piping signal data; The output of the random forest model is ; Among them, H(x) represents the prediction result of the piping path of the random forest model, which includes the potential area and time window of piping, M represents the number of decision trees, and h i (x) represents the prediction result of the i-th decision tree.

[0012] Furthermore, based on the temperature change data, the seepage field state data and the prediction result of the piping path, the probability of piping occurrence and the warning level are predicted, including: establishing a warning system based on time series analysis, the warning system including an autoregressive integrated moving average model, a principal component analysis model and a hidden Markov model; The temperature change data, the seepage field data and the prediction result of the piping path are used as inputs of the early warning system, and the piping occurrence probability and the early warning level are calculated through the designed fuzzy rules.

[0013] Furthermore, the autoregressive integrated moving average model is used to perform time series analysis of temperature change data. The specific analysis includes: predicting the predicted temperature within a predetermined future time based on the temperature change data, and determining whether the predicted temperature continues to be greater than a preset threshold.

[0014] Furthermore, the principal component analysis model is used to analyze the seepage field state data. The specific analysis includes: performing dimensionality reduction processing on the seepage field state data, extracting the most representative features as principal components to represent the seepage state at different positions of the dam, and judging the change of seepage pressure at the corresponding position according to the change of the principal component score.

[0015] Furthermore, the hidden Markov model is used to perform time series analysis of the prediction results of the piping path. The specific analysis includes: dividing the piping development process into three stages: latent period, development period and acceleration period according to its time window, and learning the characteristics and transition probabilities of these three stages based on historical data to predict the probability of its occurrence stage migration.

[0016] Another aspect of the present invention further provides a safety assessment system for implementing the aforementioned safety assessment method for embankment projects, comprising: A temperature change acquisition module is used to obtain temperature change data of the embankment area to be monitored; A time delay impact analysis module is used to calculate the temperature gradient distribution inside the dike based on the temperature change data, and analyze the time delay impact of the temperature gradient distribution on the seepage field state in combination with the historical data of the seepage field state; The seepage field data acquisition module is used to deploy a distributed sensor network in key areas of the embankment to monitor changes in the seepage field in real time and obtain high-precision seepage field data; a piping path prediction module, configured to predict the piping path using a random forest model based on piping signals extracted from the high-precision seepage field state data and in combination with the temperature gradient distribution and the seepage field state data; The piping early warning module is used to predict the probability of piping and the early warning level based on the temperature change data, the seepage field data and the prediction result of the piping path.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. By integrating multi-source data, including temperature changes, seepage field states, piping signals and other data, the present invention can more comprehensively assess piping risks, reduce misjudgments that may be caused by a single indicator, and improve prediction accuracy.

[0018] 2. The present invention combines historical data of seepage field states to analyze the delayed effect of temperature gradient on seepage field states, compares the time series of temperature gradient changes and seepage field state changes, and finds the correlation between the two and the time lag correlation relationship. This correlation can capture anomalies in the early stage of piping and provide early warning of the occurrence of piping.

[0019] 3. The present invention processes multiple input factors through a random forest model. The random forest model has the characteristics of strong noise resistance and low overfitting, and is suitable for processing complex data in engineering practice. Through the prediction of this random forest model, potential piping risk areas and time windows can be identified in a timely manner, providing a scientific basis for engineering decision-making.

[0020] 4. The present invention constructs an early warning system based on time series analysis to capture the dynamic changes of various indicators. It not only focuses on the current status, but also considers historical trends and future forecasts. The output of the early warning level provides decision makers with intuitive and easy-to-understand risk assessment results, which helps to take corresponding prevention and control measures in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic flow chart showing a method for safety assessment of a levee project in a specific embodiment of the present invention is shown; Figure 2 A schematic diagram of the process of random forest model training in a specific embodiment of the present invention is shown; Figure 3 A schematic diagram of a process for predicting piping paths using a random forest model in a specific embodiment of the present invention is shown. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] In order to solve the shortcomings of the existing technology, the present invention discloses a safety assessment method for embankment engineering, such as Figure 1 As shown, the following steps are included: Step S101: Acquire temperature change data of the embankment area to be monitored.

[0024] Further, step S101 includes installing temperature sensors at different depths and locations of the embankment to form a temperature monitoring grid, and monitoring the internal and surface temperatures of the embankment in real time to obtain temperature change data of the embankment area.

[0025] The acquisition of embankment area temperature change data is an important part of flood control management. By deploying a network of temperature sensors, the internal and surface temperatures of the embankment can be monitored in real time. These sensors are usually installed at different depths and locations of the embankment to form a temperature monitoring grid. For example, on a 100-meter-long embankment, a set of sensors can be installed every 10 meters, with each set containing three measuring points: surface, middle, and deep. The data collected by the sensors is transmitted to the data center via a wireless network, and a temperature distribution map of the embankment is constructed.

[0026] This temperature-based piping warning method has the advantages of non-invasiveness and real-time. Compared with traditional regular patrols, an automated temperature monitoring system can work 24 hours a day without interruption, capturing subtle changes in embankment conditions in a timely manner. This not only improves monitoring efficiency, but also gives management personnel more emergency response time, effectively reducing the risk of flood disasters.

[0027] Step S102, according to the temperature change data, calculate the temperature gradient distribution of the internal of the embankment, combine the historical data of seepage field state, analyze the delay influence of temperature gradient distribution on seepage field state.

[0028] Specifically, the calculation of the temperature gradient distribution inside the embankment is the key to understanding the seepage dynamics. Temperature gradient reflects the direction and intensity of heat transfer, which can be calculated from the data obtained by the temperature sensor network. For example, in a 100-meter-long embankment, a set of temperature sensors is arranged every 10 meters, each set containing surface, middle, and deep measuring points. Through the temperature data of these measuring points, the vertical and horizontal temperature gradients can be calculated. The temperature gradient formula is: , ΔT represents the temperature gradient, T2 and T1 are the temperatures of the two measuring points, and d is the distance between the two points. Through this formula, the temperature gradient distribution inside the embankment can be obtained.

[0029] The vertical temperature gradient can reflect the seepage situation within the levee. Under normal circumstances, the surface temperature of the levee is greatly affected by the external air temperature, while the deep temperature is relatively stable. If an abnormal increase or decrease in the deep temperature is observed, it may mean the existence of a seepage channel. For example, under normal summer conditions, the surface temperature of a certain levee is 30°C, the deep temperature is 20°C, and the vertical temperature gradient is 1°C / meter. If the deep temperature suddenly drops to 15°C and the vertical temperature gradient increases to 1.5°C / meter, this may indicate the infiltration of cold water. The horizontal temperature gradient can be used to locate the location of seepage. If a local temperature anomaly is observed on a horizontal section of the levee, it may indicate the existence of a seepage channel there.

[0030] For example, if the temperature on a horizontal surface in the middle of a dike remains at 22°C for most of the area, but suddenly drops to 18°C ​​in a 10-meter section, creating a horizontal temperature gradient of 0.4°C / meter, this could indicate a seepage problem. Analyzing the delayed impact of the temperature gradient in conjunction with historical seepage data can better predict piping risks. Historical seepage data typically includes parameters such as seepage pressure, flow velocity, and flow rate.

[0031] In this scheme, by comparing the time series of temperature gradient changes and seepage field changes, the correlation and time lag between the two can be found. For example, analyzing the data of the past five years, it is found that significant changes in temperature gradients usually precede changes in seepage field states by 12 to 24 hours. Specifically, when the vertical temperature gradient at a certain point on the embankment increases by 50% within 6 hours, the seepage pressure at that point increases by 30% within the subsequent 18 hours. This correlation provides a valuable time window for the early warning system. The delayed effect of temperature gradient distribution on seepage field states is expressed by the following formula: ; Where R(t) represents the cumulative effect, K(τ) is the kernel function that describes the effect of temperature gradient, and t is time.

[0032] The delayed effect of temperature gradients on seepage patterns may stem from the thermodynamic properties of soil. When seepage occurs, water flow gradually alters the soil's temperature distribution. However, due to the soil's heat capacity and thermal conductivity, these temperature changes take time to fully manifest. Therefore, changes in temperature gradients often precede noticeable changes in seepage patterns. Leveraging this delayed effect can help build more accurate early warning models.

[0033] Temperature gradient analysis can also help identify trends in seepage. If an abnormal temperature gradient is observed to gradually expand or increase in intensity, it may indicate that the seepage channel is expanding. For example, a 5-meter-wide temperature anomaly was detected on a dike with a temperature gradient of 0.5°C / meter. Over the next week, this anomaly expanded to 10 meters, and the temperature gradient increased to 0.8°C / meter. This trend indicates that the seepage problem is worsening and requires timely intervention.

[0034] Step S103, deploy a distributed sensor network in the key area of the embankment to monitor the changes in the seepage field in real time and obtain high-precision seepage field data.

[0035] Specifically, the deployment of a distributed sensor network in the key area of the embankment is an effective method for monitoring changes in the seepage field. This method can obtain high-precision seepage data in real time, providing key support for embankment safety management. In practical applications, various types of sensors can be selected, such as pressure sensors and flow sensors, and arranged at different positions and depths of the embankment. For example, pressure sensors are installed on the embankment slope, embankment top, and embankment foot to monitor changes in pore water pressure and determine the seepage path and intensity. Flow sensors are installed at different depths inside the embankment to measure water flow velocity and direction and identify potential leakage points.

[0036] These sensors form a network through wireless communication technology and transmit the collected data to the central control system in real time. Data transmission can use low-power wide-area network (LPWAN) technology such as LoRa or NB-IoT to ensure stable communication in harsh environments. The central control system uses high-performance computing devices and professional software to process and analyze the massive data received. During data analysis, various advanced algorithms can be applied. For example, time series analysis can be used to identify trends in seepage parameters, spatial interpolation techniques can be used to construct a three-dimensional seepage field model inside the embankment, and machine learning algorithms can be used to predict future changes in seepage conditions. These analysis results provide engineers with intuitive and comprehensive assessments of the embankment seepage conditions.

[0037] The acquisition of high-precision seepage field data is of great significance to embankment safety management. Through real-time monitoring, abnormal seepage phenomena such as piping or leakage can be detected in a timely manner, allowing appropriate protective measures to be taken. For example, when a pressure sensor in a certain area detects a sudden rise in pore water pressure exceeding the preset threshold, the system will immediately issue an alarm to alert staff to conduct on-site inspections and treatments. In addition, the long-term accumulation of high-precision seepage data can also be used to optimize embankment design and maintenance strategies. By analyzing seepage characteristics under different seasonal and water level conditions, engineers can better understand the hydraulic performance of the embankment, allowing for more reasonable design parameters and construction methods in future projects. For example, it may be found that a particular embankment structure is more prone to seepage problems under certain hydrological conditions, which will guide future embankment reconstruction work. In summary, the use of a distributed sensor network to monitor the seepage field of an embankment not only improves monitoring accuracy and real-time performance but also provides reliable data support for embankment safety management and engineering optimization, significantly improving the overall safety and operational efficiency of the embankment.

[0038] Step S104 , predicting a piping path using a random forest model based on the piping signal extracted from the high-precision seepage field data and in combination with the temperature gradient distribution and the seepage field data.

[0039] Specifically, this approach predicts the occurrence of piping by analyzing high-precision seepage data, temperature gradient distribution, and seepage field data. Piping is a major threat to levee safety, and its occurrence is often accompanied by significant changes in the seepage field.

[0040] First, extracting piping signals from high-precision seepage data is a critical step. These signals may include an abnormal increase in soil pore water pressure, a sudden change in seepage velocity, or a sharp increase in infiltration volume. For example, during monitoring of a certain embankment, it was discovered that the pore water pressure in a local area increased from 20 kPa to 35 kPa in a short period of time. This abnormal change may indicate the early signs of piping.

[0041] In this program, if Figure 2 As shown in the figure, before using the random forest model for prediction, the random forest model needs to be trained for piping path identification. The training process includes the following steps: Select the random forest model; In this embodiment, the random forest model may include 100 decision trees, each using a randomly selected feature subset. Through cross-validation, the optimal parameter combination is selected, such as setting the tree depth to 8 and the minimum number of samples per node to 5.

[0042] Prepare historical temperature gradient distribution data, seepage field data, piping signal data and their corresponding piping results (including piping area and time window); Then, the historical temperature gradient distribution data, seepage field data, and piping signal data are used as input, and the corresponding piping results are used as output to train the random forest model to obtain a trained random forest model.

[0043] After training, the model can be used to predict new data. Consider monitoring data for a section of a dike: a temperature gradient of 0.05°C / m, a 15% increase in pore water pressure, and a detected acoustic signal with a frequency of 60 kHz. The model might output a 90% probability of a piping burst occurring in that area within the next 24 hours, along with the likely seepage path. This approach offers the advantage of comprehensively considering multiple factors and capturing their complex relationships. For example, a combination of temperature anomalies and pressure changes may be a more accurate predictor of piping bursts than a single indicator alone. Furthermore, the random forest model is robust to noise and prone to overfitting, making it suitable for processing complex data encountered in engineering practice. This approach allows for the timely identification of potential piping risk areas, providing a scientific basis for engineering decision-making. For example, after identifying high-risk areas, inspection frequency can be increased or temporary reinforcement measures implemented. Long-term, this predictive model can also guide the optimization of dike design and maintenance strategies, improving the overall safety and cost-effectiveness of projects.

[0044] Specifically, the input of the random forest model is Where X represents the input feature matrix, T represents the temperature gradient distribution data, F represents the seepage field data, and S represents the piping signal data. The output of the random forest model is ; Where H(x) represents the prediction result of the piping path of the random forest model, M represents the number of decision trees, and h i (x) represents the prediction result of the i-th decision tree. Random forest obtains the final result by averaging the predictions of multiple decision trees, that is, the final prediction result of the piping path, including the potential area and time window of piping. The piping path is predicted by the random forest model, such as Figure 3 As shown, the following steps are included: extracting piping signals; analyzing temperature gradients; studying seepage field states; determining potential areas; and determining time windows.

[0045] Step S105 , predicting the probability of occurrence and warning level of piping based on the temperature change data, the seepage field data and the prediction result of the piping path.

[0046] Furthermore, step S105 includes establishing an early warning system based on time series analysis, wherein the early warning system includes an autoregressive integrated moving average model, a principal component analysis model, and a hidden Markov model; The temperature change data, the seepage field data and the prediction result of the piping path are used as inputs of the early warning system, and the piping occurrence probability and the early warning level are calculated through the designed fuzzy rules.

[0047] Specifically, establishing an early warning system based on time series analysis is a key step in piping risk assessment. This system uses temperature change data, seepage field data, and piping path prediction results as input. Through comprehensive analysis, it outputs the probability of piping and the warning level, providing a scientific basis for flood control decision-making.

[0048] Time series analysis can employ the Autoregressive Integrated Moving Average (ARIMA) model. Using ARIMA, we can analyze temperature change data over time. This analysis involves predicting the predicted temperature for a predetermined period of time based on the temperature change data and determining whether the predicted temperature remains above a preset threshold. This model can capture trends, seasonality, and random fluctuations in the data, making it suitable for processing non-stationary time series data. For example, suppose that temperature data at a dam monitoring point shows a gradual upward trend from 20°C to 25°C for 30 consecutive days. The ARIMA model can predict temperature changes over the next seven days. If the predicted temperature continues to rise to 28°C, this may indicate increased seepage activity.

[0049] Seepage data can be analyzed using a principal component analysis (PCA) model. This analysis involves performing dimensionality reduction on the data, extracting the most representative features as principal components to represent the seepage state at different locations along the dam, and then determining changes in the seepage pressure at those locations based on changes in the principal component scores. For example, a PCA analysis of pressure data from 100 measuring points within the dam revealed that the first three principal components explained 85% of the data variance. These principal components may represent the seepage state upstream, midstream, and downstream of the dam, respectively. A continued increase in the score of the first principal component may indicate increasing seepage pressure upstream.

[0050] The Hidden Markov Model (HMM) can be used to analyze the time series of piping path prediction results. This model can describe the transition process of system states. Assume that the piping development process is divided into three phases based on its time window: the incubation phase, the development phase, and the acceleration phase. The HMM can learn the characteristics and transition probabilities of these three phases from historical data to predict the probability of phase transition. If the model predicts that the system will transition from the incubation phase to the development phase with a probability exceeding 60%, this will be a significant early warning signal. Based on these analysis results, a fuzzy logic-based early warning level assessment system can be constructed.

[0051] The system uses the outputs of ARIMA, PCA and HMM as input variables and calculates the probability of piping and the warning level through the set fuzzy rules. For example: 1. If the temperature forecast shows a stable trend, the seepage pressure main component score remains relatively stable, and the piping state is in the latent period, the system outputs a 20% probability of piping and gives a green warning level, which means it is relatively safe. 2. If the temperature forecast shows an upward trend or the seepage pressure main component score increases, and the piping state is in the latent period, the system outputs a 50% probability of piping and issues a blue warning level, indicating that there is a certain hidden danger; 3. If the temperature forecast shows an upward trend, the seepage pressure main component score increases, and the piping state is in the development stage, the system may output an 80% probability of piping and issue an orange warning level; 4. If the temperature forecast shows an upward trend, the seepage pressure main component score increases, and the piping state is in an accelerated period, the system may output a 100% probability of piping and issue a red warning level.

[0052] It should be noted that the above-mentioned piping probability and the corresponding warning level are defined by fuzzy rules. Therefore, fuzzy rules can be manually set according to actual needs to achieve different warning effects, which are all within the scope of protection of the present invention.

[0053] An embodiment of the present invention further provides a safety assessment system for implementing the aforementioned safety assessment method for embankment projects, comprising: A temperature change acquisition module is used to obtain temperature change data of the embankment area to be monitored; A time delay impact analysis module is used to calculate the temperature gradient distribution inside the dike based on the temperature change data, and analyze the time delay impact of the temperature gradient distribution on the seepage field state in combination with the historical data of the seepage field state; The seepage field data acquisition module is used to deploy a distributed sensor network in key areas of the embankment to monitor changes in the seepage field in real time and obtain high-precision seepage field data; a piping path prediction module, configured to predict the piping path using a random forest model based on piping signals extracted from the high-precision seepage field state data and in combination with the temperature gradient distribution and the seepage field state data; The piping early warning module is used to predict the probability of piping and the early warning level based on the temperature change data, the seepage field data and the prediction result of the piping path.

[0054] The embodiments described above are merely illustrative of embodiments of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. The present invention may also be implemented in other specific ways or in other specific forms without departing from the gist or essential features of the present invention. Therefore, the embodiments described should be considered in all respects as illustrative and not restrictive. The scope of the present invention should be described by the appended claims, and any variations equivalent to the intent and scope of the claims should also be included within the scope of the present invention.

Claims

1. A safety assessment method for embankment engineering, characterized in that: The following steps are involved: Obtain temperature change data of the embankment area to be monitored; Calculating the temperature gradient distribution inside the dike based on the temperature change data, and analyzing the delayed effect of the temperature gradient distribution on the seepage field state in combination with historical data of the seepage field state; Deploy distributed sensor networks in key areas of the embankment to monitor changes in seepage fields in real time and obtain high-precision seepage field data; Predicting piping paths using a random forest model based on piping signals extracted from the high-precision seepage field data and in combination with the temperature gradient distribution and the seepage field data; Based on the temperature change data, the seepage field state data and the prediction result of the piping path, the probability of piping occurrence and the warning level are predicted.

2. The safety assessment method for embankment engineering according to claim 1, characterized in that: The step of obtaining temperature change data of the embankment area to be monitored includes: Temperature sensors are installed at different depths and locations on the embankment to form a temperature monitoring grid, which monitors the internal and surface temperatures of the embankment in real time and obtains temperature change data of the embankment area.

3. The safety assessment method for embankment engineering according to claim 1, characterized in that: The historical data of the seepage field state include seepage pressure, flow velocity and flow rate.

4. The safety assessment method for embankment engineering according to claim 1, characterized in that: The delayed effect of temperature gradient distribution on the seepage state is expressed as follows: ； Among them, R(t) represents the cumulative effect, K(τ) is the kernel function describing the effect of temperature gradient, and t is time.

5. The safety assessment method for embankment engineering according to claim 1, characterized in that: Deploy a distributed sensor network in key areas of the embankment to monitor changes in the seepage field in real time and obtain high-precision seepage data, including: Install pressure sensors in key areas including the embankment slope, crest, and toe to monitor changes in pore water pressure and determine seepage path and intensity; Flow sensors are installed at different depths inside the embankment to measure water flow speed and direction and identify potential leakage points.

6. The safety assessment method for embankment engineering according to claim 1, characterized in that: The piping signal includes an abnormal increase in soil pore water pressure, a sudden change in seepage velocity, or a sharp increase in seepage water volume.

7. The safety assessment method for embankment engineering according to claim 1, characterized in that: The input of the random forest model is ; Where X represents the input feature matrix, T represents the temperature gradient distribution data, F represents the seepage field data, and S represents the piping signal data; The output of the random forest model is ; Among them, H(x) represents the prediction result of the piping path of the random forest model, which includes the potential area and time window of piping, M represents the number of decision trees, and h i (x) represents the prediction result of the i-th decision tree.

8. The safety assessment method for embankment engineering according to claim 1, characterized in that: Predicting the probability and warning level of piping occurrence based on the temperature change data, the seepage field state data, and the prediction result of the piping path includes: Establishing an early warning system based on time series analysis, the early warning system includes an autoregressive integrated moving average model, a principal component analysis model, and a hidden Markov model; The temperature change data, the seepage field data and the prediction result of the piping path are used as inputs of the early warning system, and the piping occurrence probability and the early warning level are calculated through the designed fuzzy rules.

9. The safety assessment method for embankment engineering according to claim 8, characterized in that: The autoregressive integrated moving average model is used to perform time series analysis of temperature change data. The specific analysis includes: predicting the predicted temperature within a predetermined time in the future based on the temperature change data, and determining whether the predicted temperature is continuously greater than a preset threshold; And / or, the principal component analysis model is used to analyze the seepage field state data, specifically including: performing dimensionality reduction processing on the seepage field state data, extracting the most representative features as principal components to represent the seepage state at different locations of the dam, and judging the change of the seepage pressure at the corresponding location based on the change of the principal component score; And / or, the hidden Markov model is used to perform time series analysis of the prediction results of the piping path, and the specific analysis includes: dividing the piping development process into three stages: incubation period, development period and acceleration period according to its time window, and learning the characteristics and transition probabilities of these three stages based on historical data to predict the probability of its occurrence stage migration.

10. A safety assessment system for implementing the safety assessment method for embankment engineering according to any one of claims 1 to 9, characterized in that: include: A temperature change acquisition module is used to obtain temperature change data of the embankment area to be monitored; A time delay impact analysis module is used to calculate the temperature gradient distribution inside the dike based on the temperature change data, and analyze the time delay impact of the temperature gradient distribution on the seepage field state in combination with the historical data of the seepage field state; The seepage field data acquisition module is used to deploy a distributed sensor network in key areas of the embankment to monitor changes in the seepage field in real time and obtain high-precision seepage field data; a piping path prediction module, configured to predict the piping path using a random forest model based on piping signals extracted from the high-precision seepage field state data and in combination with the temperature gradient distribution and the seepage field state data; The piping early warning module is used to predict the probability of piping and the early warning level based on the temperature change data, the seepage field data and the prediction result of the piping path.

Images