Method for simulating and diagnosing deposition based on mechanism-data coupling drive

Abstract

The invention discloses a method for simulating and diagnosing siltation based on mechanism-data coupling driving, and belongs to the technical field of urban drainage system siltation condition simulation and diagnosis. According to the method, the liquid level and flow data of the key point location of the combined system pipeline are predicted through the LSTM neural network, and the sediment thickness is inversely calculated in combination with the Manning formula; rewriting and constructing a sediment mechanism model based on an SWMM model, and coupling a sediment control equation and a transportation equation; and calibrating key parameters of the mechanism model by using sediment thickness data output by the data driving model to form a hybrid simulation model. The mean square error of the hybrid model is less than 1%, the mean absolute error is 0.018 m, the correlation coefficient R2 is 0.68, and the time for simulating the change of a 10d deposition layer reaches a minute level. According to the method, the dependency of a mechanism model on field sampling data is reduced, the defect of insufficient interpretability of a data driving model is overcome, the precision and efficiency of combined system pipeline sediment simulation are improved, and the method is suitable for sediment management and control of an urban drainage system.

Description

Technical Field

[0001] This invention relates to the field of simulation and diagnosis technology for siltation conditions in urban drainage systems, and specifically to a mechanism-data coupling-driven simulation and diagnosis method for siltation. Background Technology

[0002] With the rapid urbanization in my country, the proportion of hardened urban surfaces has increased significantly, leading to a large amount of surface particles carried by rainwater runoff into combined sewer systems. Inadequate maintenance, poor sewage pumping station facilities, and in poor hydraulic conditions of the pipelines themselves are factors contributing to sedimentation, which can even develop into permanent sedimentary layers under consolidation. Sediments have a series of negative impacts on drainage systems; for example, deposited organic matter reacts within the pipes, consuming carbon sources and thus affecting the efficiency of sewage treatment plants. Combined sewer overflows can contribute 50% to 90% of conventional water pollution, with pipeline sediments accounting for a significant portion of this pollution. Severely silted areas may also experience sewage overflows or even urban flooding.

[0003] For predicting sediment thickness, the mechanism-data hybrid model can calculate hydrodynamic and SS concentration changes through mechanistic modeling, providing strong theoretical support and ensuring model accuracy. The data-driven model can predict sediment thickness development, thus replacing field data collection and providing a large amount of data for validating sediment transport capacity equations, improving model accuracy. The advantages of the mechanism-data hybrid model lie in the interpretability of the mechanism and the improvement of model accuracy.

[0004] Currently, simulation studies on sediment movement are relatively limited. Compared to single numerical simulations or mechanistic model simulations, using a hybrid mechanism-data-driven approach to simulate sediments in combined sewer systems reduces reliance on field data and facilitates the extension of the model to the scale of urban drainage systems, possessing significant theoretical guidance and engineering application prospects. Chinese invention patent CN119622974A discloses a "Numerical Simulation Method and System for Sediment Flushing and Deposition in Pipelines Based on Cellular Automata." This method constructs a one-dimensional cellular flushing and deposition scenario for pipelines, setting initial conditions for pipeline cell size, neighborhood relationships, states, and the mass of aqueous and sedimentary particles. It calculates the net vertical sedimentation of the cells and updates the sedimentary particle mass and deposition thickness of the cells at the next time step according to sedimentary particle update rules. It also calculates the horizontal water phase transport of the cells and updates the water phase particle mass and concentration of the cells at the next time step according to water phase particle update rules. Combining the updated pipeline cell states, the vertical sedimentation and horizontal transport calculations of the pipeline cells are cyclically performed to obtain the sediment flushing and deposition process inside the pipeline. The discrete calculation method for sediment scouring and deposition changes in pipelines can accurately simulate the migration and transformation process of sediments inside pipelines, providing technical support for the simulation of sediment scouring and deposition in pipe networks. The aforementioned existing patents did not use the mechanistic model disclosed in this application. Their models involve multiple parameters, such as cell size, neighborhood relationships, pipe diameter, slope, and roughness, the settings of which significantly affect the simulation results. However, in practical applications, accurately obtaining these parameters may be difficult, and the interactions between parameters may lead to uncertainties in the simulation results. Furthermore, the model calibration process may be complex, requiring substantial experimental or field observation data for support. Chinese invention patent CN112179278A, entitled "A Device and Method for Measuring Pipeline Deposit Thickness Based on Laser Ranging," discloses a device for measuring pipeline deposit thickness based on laser ranging. This device includes a tripod, a laser rangefinder fixed to the top of the tripod via a gimbal, and an auxiliary alignment device for aligning the center point of the laser rangefinder with the vertical diameter of the pipe opening. The patent describes a method for measuring pipeline deposit thickness when the pipeline is laid in an ideally horizontal state and when the rainwater pipeline has a slope. This method overcomes the limitations of traditional measurement methods in terms of pipe diameter and the inability to measure deep within the pipeline, while also improving measurement accuracy. The aforementioned prior art patents do not employ the model simulation method disclosed in this application. Laser ranging has high equipment costs and requires regular maintenance and calibration, increasing additional maintenance costs and manpower.

[0005] Sedimentation in combined sewer systems is a key target for urban water environment management. Sedimentation leads to reduced pipe flow capacity, sewage overflows, and increased risk of flooding, with sediment contributing significantly to combined sewer overflow pollution. Existing sediment simulation methods are mainly divided into two categories: mechanistic models and data-driven models. Mechanistic models, such as SWMM and SewerSedFoam, rely on complete basic and monitoring data, have long modeling cycles, and poor adaptability to local parameters. Data-driven models, such as artificial neural networks, while possessing a certain level of predictive accuracy, lack interpretability and struggle to handle unconventional operating conditions. Therefore, a hybrid simulation method that combines high accuracy, strong interpretability, and high computational efficiency is urgently needed. Summary of the Invention

[0006] The purpose of this invention is to propose a mechanism-data coupling-driven simulation-based method for diagnosing siltation to solve the problems mentioned in the background art. This invention can predict the thickness of deposits in combined sewer systems and is used to diagnose the siltation status of urban drainage system pipes.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solution: A mechanism-data coupling-driven simulation-based method for diagnosing siltation includes the following steps: S1: Data collection and preprocessing: By deploying one water quality flow meter and one level meter at each monitoring point, the basic data of the combined sewer system, monitoring data and operation and maintenance reports of the pipe network in the area are collected. The monitoring data includes time series reports of pipe network flow, level and rainfall. The monitoring interval is 30 min. The data is standardized to the range of (0,1). S2: Data-driven model construction and sediment thickness prediction based on LSTM: A neural network model containing LSTM, Dropout and Dense layers was constructed using TensorFlow Sequential. The training batches were set to 16-20 times and the training generations to 8-10 rounds. The adam optimization method was used. The model was trained by inputting preprocessed monitoring data to predict the flow rate and liquid level data at key points. The sediment thickness was then calculated by combining the Manning formula. S3: Construction of Sediment Mechanism Model for Combined Sediment Pipelines: Basic pipeline data and flow velocity, flow rate, and suspended solids (SS) concentration data were obtained through PySWMM and SWMM-API. Taking a single pipe segment as the control unit, the SWMM model was rewritten based on the sediment control equation and transport equation, and the initial update cycle of sediment thickness was set to 3 days. S4: Hybrid Model Construction and Parameter Calibration: Using the sediment thickness data predicted by the data-driven model in step S2, combined with the measured sediment thickness in the field, the key coefficients a, b, c, and d of the sediment transport equation in the mechanistic model were calibrated. After calibration, the coefficient values ​​were 0.37, 2.56, 1.88, and 2.59, respectively, and the sediment thickness update cycle was adjusted to 1.5 days. S5: Hybrid Model Simulation and Validation: Simulates the sediment thickness variation in combined sewer systems using a calibrated hybrid model. The simulation results are compared with measured data for validation, ensuring the model's mean square error is <1% and the correlation coefficient R0 is within acceptable limits. 2 ≥0.68.

[0008] Preferably, the foundation of the collection network in S1 includes parameters such as catchment area, overflow width, surface slope, percentage of impermeable surface, Manning coefficient, water storage depth in depressions, and infiltration rate.

[0009] Preferably, in step S2, the LSTM model has 4 sets of input data and 50 sets of output data. 80% of the output data is used during the Dropout layer iteration training. The model training time is about 45 minutes. The average absolute error of the predicted liquid level is 0.01m and the average relative error of the predicted flow rate is 13.4%.

[0010] Preferably, the calculation method of Manning's formula in S2 is presented using a hydraulic performance diagram. The relationship between the liquid level and flow rate in the pipeline is derived using Manning's formula, achieving the goal of directly calculating the flow rate from a known liquid level. The calculation formula is as follows:

[0011] in, Q Predict traffic using LSTM; S f For the pipe friction slope ( S f =0.0007); n Manning coefficient ( n =0.013); A The cross-sectional area is the area of ​​the flow path. R The hydraulic radius; The difference between the liquid level data calculated by the Manning formula and the liquid level data predicted by LSTM is the thickness of the sediment in the pipe. For non-uniform steady flow in open channels, the relationship between upstream water level, downstream water level, and flow rate is non-linear, and the spatial variation of flow depth is described by the backwater differential equation:

[0012]

[0013]

[0014] in, h The water flows and the water is deep; x Distance along the river channel; S 0 represents the slope of the pipe bottom. V The velocity of the water flow; g It is the acceleration due to gravity; F For Froude number; K n This is the unit conversion factor for Manning's formula. K n =1; D m This represents the average water depth.

[0015] Preferably, the step of comparing the changes in suspended solids (SS) concentration in the simulated pipeline in S3 includes: S31: Construct a suspended solids (SS) model in the initially constructed SWMM model to simulate the concentration of suspended solids (SS) in the pipe network; S32: Determine the concentration pattern of suspended solids (SS) in wastewater discharge based on the sampling test values, and compare the simulated suspended solids (SS) values ​​with the probe monitoring values ​​and the actual sampling values; S33: Couple the suspended matter (SS) model with sediment motion equations to further improve sediment simulation capabilities.

[0016] Preferably, the suspended particulate matter (SS) model in S31 is mainly constructed based on the land use nature of the area to be measured and the survey situation. The land types involved in the catchment area are mainly divided into asphalt roads and green spaces. According to the average concentration of each pollutant in the initial rainwater monitoring sessions, the SWMM user manual, the environmental conditions around the monitoring points, and relevant literature of similar studies, relevant parameters such as surface pollutant accumulation and scouring are set for different land use types.

[0017] Preferably, the specific content of coupling the suspended matter (SS) model with the sediment motion equation in step S33 is as follows: Basic data such as pipe diameter, length, flow rate, velocity, and suspended solids (SS) were obtained using PySWMM and SWMM-API. Calculations were then performed using the sediment control equation and sediment transport capacity formula, with empirically chosen coefficients (a, b, c, and d = 0.21, 0.09, 0.53, and 0.21, respectively). A 3-day sediment thickness update cycle was selected in the SWMM model, and the mass of sediment change was calculated at each calculation step. The suspended solids (SS) concentration inside the pipe was correlated with the sediment transport capacity (SS). C vBy comparing the sediment thickness with the model, the effects of siltation and erosion can be determined. The change in sediment mass per unit time can be calculated, and the sediment thickness is updated every 3 days. This thickness is used as a new boundary condition input into the model, and the relevant results are stored in the hot start file. The SWMM model is called repeatedly for subsequent simulation calculations. This process continues until the simulation ends.

[0018] Preferably, the sediment governing equations and sediment transport capacity formulas are as follows:

[0019]

[0020] in, The density of suspended solids (SS). A s The cross-sectional area of ​​the sediment layer. C This refers to the concentration of suspended solids (SS). t For time, L For the length of the control unit, C v For the ability to transport sediments, g It is the acceleration due to gravity. d 50 The median particle size of particulate matter. s For specific gravity, v The average flow velocity in the pipe. D * Dimensionless particle size f c denoted as friction coefficient, with superscripts a, b, c, and d indicating coefficients.

[0021] Preferably, the S4 on-site measurement analyzes the particle size distribution of sediments using a standard sieve, and the measurement area is... L The main pipeline downstream of the Xinhe Pumping Station in the Chengbei sewage receiving area of ​​the city is 56.5 m long, 0.8 m in diameter, made of concrete, and has a slope of 0.43%.

[0022] A combined sewer system sediment simulation system includes a data acquisition module, a data preprocessing module, an LSTM prediction module, a mechanism model construction module, a parameter calibration module, and a simulation output module. The data acquisition module is used to collect basic pipeline data and monitoring data, and the simulation output module is used to output sediment thickness variation curves and liquid level and flow rate correlation data in real time.

[0023] The present invention further protects a computer device, the computer device including a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the instruction, program, code set or instruction set being loaded and executed by the processor to implement the above-described mechanism-data coupling driven simulation diagnosis of siltation method.

[0024] The present invention further protects a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the instruction, program, code set, or instruction set is loaded and executed by a processor to implement the above-described mechanism-data coupling driven simulation diagnosis of siltation method.

[0025] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention combines the strong interpretability of the mechanistic model with the high predictability of the data-driven model, reduces the dependence on field sampling data, eliminates the need for a large amount of complex field measurement work, and is suitable for promotion in urban-scale drainage systems; (2) This invention accurately predicts liquid level and flow data through LSTM model and, combined with local parameter calibration, significantly improves the accuracy of sediment simulation, with mean square error less than 1% and correlation coefficient reaching 0.68; (3) The present invention has high computational efficiency, requiring only minutes to simulate 10 days of sediment layer changes, and the sediment thickness update cycle is optimized to 15 days, which can quickly respond to the dynamic changes of pipeline deposition; (4) This invention can output the correlation change law between sediment thickness and liquid level and flow rate, providing a scientific basis for the formulation of combined sewer dredging schemes and the optimization of pump station operation, effectively reducing the risk of sewage overflow and waterlogging. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings involved in the embodiments are now briefly described. Obviously, the drawings in the following description are merely illustrative of some embodiments of the present invention. For those skilled in the art, other forms of drawings can be constructed based on these drawings without creative effort.

[0027] Figure 1 A flowchart of a mechanism-data coupling-driven simulation-based method for diagnosing siltation according to the present invention is shown below; Figure 2 This is a schematic diagram of a research area provided according to an embodiment of the present invention; Figure 3 A graph showing the changes in monitoring data such as liquid level, flow rate, and rainfall at key locations according to an embodiment of the present invention; Figure 4A schematic diagram illustrating the flow rate and liquid level monitoring values ​​and predicted values ​​at key locations according to an embodiment of the present invention; Figure 5 This is a schematic diagram of sediment thickness variation according to an embodiment of the present invention; Figure 6 This is a schematic diagram of a SWMM model pipeline network provided according to an embodiment of the present invention; Figure 7 This is a schematic diagram of simulated and monitored values ​​of SS concentration at a pipeline node according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the predicted pipeline deposit thickness according to an embodiment of the present invention; Figure 9 This is a schematic diagram illustrating the changes in pipe sediment thickness, liquid level, and flow rate calculated after refitting, according to an embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0029] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0030] This invention proposes a mechanism-data coupling-driven simulation-based method for diagnosing sludge accumulation, comprising the following steps: Step 1: Data Collection and Preprocessing Collect basic pipeline network data, monitoring data, and operation and maintenance reports for the combined sewer system area. Monitoring data includes time-series data on pipeline flow, liquid level, and rainfall, with a monitoring interval of 30 minutes to ensure data continuity and timeliness. Basic pipeline network data includes key parameters such as catchment area, overflow width, and surface slope, with the catchment area being 2.60 km². 2 The maximum infiltration rate is 75 mm·h -1 The minimum infiltration rate is 3.5 mm·h. -1 All collected data were standardized to the range of (0,1) to eliminate the impact of differences in data units on model training.

[0031] Step 2: LSTM-based data-driven model construction and sediment thickness prediction: A neural network model was constructed using the TensorFlow Sequential framework, comprising LSTM, Dropout, and Dense layers. The LSTM layer extracted local temporal variation features from the monitoring data, the Dropout layer enhanced the model's generalization ability, and the Dense layer extracted key information and output the prediction results. The training batch size was set to 16-20 epochs, with 8-10 training generations. The Adam optimization algorithm was used, inputting 4 sets of preprocessed monitoring data and outputting 50 sets of prediction results. The model training time was approximately 45 minutes. After training, it predicted flow rate and liquid level data at key locations, with an average absolute error of 0.01 m for liquid level prediction and an average relative error of 13.4% for flow rate prediction. Using the Manning formula, the pipe sediment thickness was calculated from the predicted flow rate and liquid level values.

[0032] Step 3: Construction of a combined sewer system sedimentation mechanism model: Using PySWMM and the SWMM-API interface, basic data such as pipe diameter, length, flow velocity, flow rate, and suspended solids (SS) concentration are obtained. Taking a single pipe segment as a control unit, the system is based on sediment control equations. and transport equations The traditional SWMM model was rewritten to supplement the simulation function of sediment settling and scour processes, and the initial sediment thickness update cycle was set to 3 days.

[0033] Step 4: Hybrid Model Construction and Parameter Calibration Using sediment thickness data output from the data-driven model and combined with field-measured sediment thickness, the key coefficients a, b, c, and d of the sediment transport equation in the mechanistic model were locally calibrated. Before calibration, empirical default coefficients were used, resulting in significant prediction bias in the mechanistic model. After calibration, the coefficients were adjusted to 0.37, 2.56, 1.88, and 2.59, respectively. Simultaneously, the sediment thickness update cycle was optimized to 1.5 days to improve the model's adaptability to local sediment characteristics.

[0034] Step 5: Hybrid Model Simulation and Validation: The deposition thickness variation in combined sewer systems in the target area was simulated using a calibrated hybrid model. During the simulation, the SWMM model provided hydraulic parameters, while the data-driven model supplemented the thickness data; the two worked together to achieve high-precision simulation. Comparison of simulation results with field measured data showed that the model's mean square error was less than 1%, the mean absolute error was 0.018 m, and the correlation coefficient R0 was [missing value]. 2 The value is 0.68, and the simulation time for 10 days of sediment layer changes is in the minute range, which meets the application engineering requirements.

[0035] The following description, in conjunction with relevant accompanying drawings and specific examples, illustrates the mechanism-data coupling-driven simulation diagnostic method for siltation proposed in this invention.

[0036] Example 1: This invention proposes a mechanism-data coupling-driven simulation-based method for diagnosing sludge accumulation, which specifically includes the following: Please see Figure 1 , Figure 1 The flowchart of the mechanism-data coupling-driven simulation-based diagnosis of siltation method proposed in this invention is as follows: Figure 1 As shown, the proposed method includes: The process involves five steps: data collection and preprocessing, LSTM data-driven model building and prediction, mechanistic model building, hybrid model coupling and parameter calibration, and model validation.

[0037] To enable those skilled in the art to further understand the present invention, the following specific embodiment will be used to provide a detailed description of the mechanism-data coupling driven simulation diagnosis method for siltation proposed in this invention.

[0038] Please see Figure 2 , Figure 2 The typical drainage area selected for this invention is located in Lu'an City, with a total area of ​​2.6 km². 2 The area has a population of approximately 17,700. The Xinhe Pumping Station is located at the end of the area. The main water source for the area is sewage during the dry season, as well as water from a 0.5 km² section of... 2 Rainy season road runoff.

[0039] Please see Figure 3 , Figure 3 The monitoring data collected included pipeline flow, pipeline liquid level, and rain gauge time series data. The monitoring interval was 30 minutes, and the monitoring period was from January to June 2024. The above data was used to build an LSTM model.

[0040] Please see Figure 4 , Figure 4 The specific steps for using LSTM to predict flow rate and liquid level data at key locations are as follows: A total of 8825 sets of data were collected, with 80% used as training data and 20% as testing data. The model structure was built based on TensorFlow Sequential, mainly consisting of three layers: an LSTM layer, a Dropout layer, and a Dense layer. There were 4 sets of input data and 50 sets of output data. The Dropout layer enhances the model's generalization ability; each training iteration uses only 80% of the output data, i.e., 40 sets. The connection between the Dropout layer and the Dense layer is a fully connected layer. The Dense layer extracts the aforementioned key information and integrates it to output a single liquid level and flow rate prediction result. In terms of parameter settings, the training batch size was 16-20 times, the training generations were 8-10 epochs, and the parameters were optimized using the Adam optimization method. The model training process took approximately 45 minutes.

[0041] The accuracy of the model results is determined by the mean squared error (MSE), mean absolute error (MAE), and correlation coefficient (R²). 2 The evaluation was conducted using LSTM to predict liquid level changes. The MSE was <1%, the MAE was 0.01 m, and the R... 2 The value was 0.96; when LSTM predicted flow rate changes, the MSE was 13.4%, the MAE was 15.13 L / s, and the R... 2 The value is 0.73, indicating that the model's prediction performance is good.

[0042] Please see Figure 5 , Figure 5 The above is the calculation result of sediment thickness for predicting sediment development over 20 days in this embodiment.

[0043] Please see Figure 6 , Figure 6 The SWMM mechanism model constructed in this embodiment covers 6 sub-catchment areas, 391 nodes, 389 pipes, and 1 pumping station.

[0044] In this embodiment, the relevant parameters include: rainfall parameters, runoff parameters, infiltration parameters, simulation period, time step, simulation control equations, etc. The specific parameter settings are as follows:

[0045] Parameters related to surface pollutant accumulation and scouring were set for different land use types. Specifically, the maximum accumulation for roads was 200 kg / ha, with a accumulation constant of 0.5 d, a scouring coefficient of 0.006, and a scouring index of 1.70; the maximum accumulation for green spaces was 100 kg / ha, with a accumulation constant of 0.8 d, a scouring coefficient of 0.003, and a scouring index of 1.20. The simulated SS values ​​were compared with probe monitoring values ​​and actual sampled values. (See [link to simulation]). Figure 7The mean error of the three sets of data is less than 10%, the Nash efficiency coefficient (NSE) is greater than 0.57, and the model accuracy meets the requirements.

[0046] Uncalibrated pipe deposit thickness prediction results Figure 8 As shown, the mechanistic model predictions far exceed the data-driven predictions after the third period, requiring further calibration by combining the data-driven predictions with the measured values.

[0047] The fitted parameters a, b, c, and d, which conform to local characteristics, are 0.37, 2.56, 1.88, and 2.59, respectively. The refitted coefficients are used to rewrite the sedimentary mechanism model, with a sediment thickness update period of 1.5 days. Sediment thickness variations are shown below. Figure 9 As shown.

[0048] The sediment simulation method of this invention can reduce reliance on field data and improve simulation accuracy while maintaining the computational efficiency of the mechanistic model. After calibration and verification, the mechanistic model can be extended to the scale of urban drainage systems to diagnose pipe siltation. Of course, the sediment transport equation is currently mainly applied to circular pipes, and its applicability to simulating viscous sediments needs further optimization and verification.

[0049] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A mechanism-data coupling-driven simulation-based method for diagnosing siltation, characterized in that, Includes the following steps: S1. Data Collection and Preprocessing: Water quality flow meters and level gauges are installed at monitoring points in combined sewer systems to collect basic network data, monitoring data, and operation and maintenance reports for the area where the combined sewer system is located, and the collected data is processed in a standardized manner. S2. Neural Network Model Construction and Sediment Thickness Prediction: A neural network model based on LSTM was constructed. The preprocessed data in S1 was input into the constructed neural network model for training to predict the flow rate and liquid level data at key points, and the sediment thickness was calculated by combining the Manning formula. S3. Construction of a combined sewer system sedimentation mechanism model: Pipeline basic data, flow velocity, flow rate, and suspended solids concentration data are obtained through PySWMM and SWMM-API. Taking a single pipe segment as the control unit, the SWMM model is rewritten based on the sediment control equation and transport equation, and the initial update cycle of sediment thickness is set. S4. Hybrid Model Construction and Parameter Calibration: Using the sediment thickness data obtained in S2, combined with the field-measured sediment thickness, the key coefficients of the sediment transport equation in the mechanistic model were calibrated, and the sediment thickness update cycle was optimized based on the calibration results. S5. Hybrid Model Simulation and Validation: The calibrated hybrid model was used to fit the sediment thickness variation in the flow channel. The simulation results were compared with measured data for verification. When the model mean square error was <1% and the correlation coefficient R was <1%, the model was considered successful. 2 When the value is ≥0.68, the model is considered valid.

2. The method according to claim 1, characterized in that, The basic data of the pipeline network mentioned in S1 includes the catchment area, overflow width, surface slope, percentage of impermeable surface, Manning coefficient, water storage depth in depressions, and infiltration rate parameters. The monitoring data includes time-series reports of pipeline flow, liquid level, and rainfall. The data collection interval is 30 minutes, and the data is standardized to the range of (0,1).

3. The method according to claim 1, characterized in that, The neural network model described in S2 is built using the TensorFlow Sequential framework, and its network structure includes LSTM layers, Dropout layers, and Dense layers. The training parameters for the neural network model are: 16-20 training batches, 8-10 training generations, and the optimizer is adam; The neural network model has 4 sets of input data and 50 sets of output data; the Dropout layer uses 80% of the output data during iterative training, and the model training time is 40~50 minutes.

4. The method according to claim 3, characterized in that, The specific content of back-calculating sediment thickness using the Manning formula as described in S2 includes the following: By combining hydraulic performance diagrams, the relationship between liquid level and flow rate in the pipeline is derived using Manning's formula. The specific calculation formula is as follows: In the formula, Q Predict traffic using LSTM; S f For pipe friction slope; n This is the Manning coefficient; A The cross-sectional area is the area of ​​the flow path. R The hydraulic radius; The difference between the liquid level data calculated using the Manning formula and the liquid level data predicted by the neural network model is used to calculate the thickness of the sediment in the pipe. For non-uniform steady flow in open channels, the relationship between upstream and downstream water levels and flow rate is non-linear, and the spatial variation of the flow depth is described using the backwater differential equation. In the formula, h The water flows and the water is deep; x Distance along the river channel; S 0 represents the slope of the pipe bottom. V The velocity of the water flow; g It is the acceleration due to gravity; D m The average water depth; F For Froude number; K n This is the unit conversion factor for Manning's formula. K n =1.

5. The method according to claim 1, characterized in that, The method for obtaining suspended solids concentration data as described in S3 specifically includes the following: S3.

1. Further construct a suspended solids model in the initially constructed SWMM model to simulate the concentration of suspended solids in the pipe network; S3.2 Determine the pattern of suspended solids concentration in wastewater discharge based on the actual sampling values, and compare the simulated suspended solids concentration in S3.1 with the probe monitoring values ​​and the actual sampling values; S3.

3. Couple the suspended matter model with the sediment motion equation to further improve the sediment simulation function.

6. The method according to claim 5, characterized in that, The method for constructing the SS model described in S3.1 specifically includes the following: Based on the land use characteristics of the area to be tested and the survey results, the land types involved in the catchment area are divided into asphalt roads and green spaces. According to the average concentration of each pollutant in the initial rainwater monitoring sessions, the SWMM user manual, the environmental conditions around the monitoring points, and relevant literature from similar studies, surface pollutant accumulation and scouring parameters are set for different land use types.

7. The method according to claim 6, characterized in that, S3.3 specifically includes the following: Pipe diameter, length, flow rate, flow velocity, and suspended solids concentration were obtained using PySWMM and SWMM-API, and calculations were performed using the sediment governing equations and sediment transport capacity formulas. In the formula, Density of suspended matter; A s The cross-sectional area of ​​the sediment layer; C This refers to the concentration of suspended solids. t For time; L The length of the control unit; C v For the ability to transport sediments; g It is the acceleration due to gravity; d 50 The median particle size; s Specific gravity; v The average flow velocity in the pipe; D * Dimensionless particle size; f c is the friction coefficient; superscripts a, b, c, d are coefficients; In the SWMM model, a sediment thickness update cycle of 3 days is selected, and the mass of sediment change is calculated at each calculation step. The concentration of suspended solids inside the pipeline is compared with the sediment transport capacity to determine the siltation and scouring situation. The change in sediment mass per unit time is calculated. The sediment thickness is updated every 3 days and used as a new boundary condition to input into the model. The relevant results are stored in the hot start file. The SWMM model is called cyclically for subsequent simulation calculations. Continue the above process until the simulation ends.

8. A combined sewer sediment simulation system applying the method of any one of claims 1-7, characterized in that, It includes a data acquisition module, a data preprocessing module, an LSTM prediction module, a mechanism model construction module, a parameter calibration module, and a simulation output module; wherein, the data acquisition module is used to collect basic pipeline network data and monitoring data, and the simulation output module is used to output sediment thickness change curves and liquid level and flow rate correlation data in real time.

9. A computer device, characterized in that, The computer device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, and the instruction, program, code set, or instruction set is loaded and executed by the processor to implement the mechanism-data coupling driven simulation diagnostic siltation method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, at least one program, code set, or instruction set, which is loaded and executed by a processor to implement the mechanism-data coupling driven simulation diagnostic siltation method as described in any one of claims 1-7.

Images