Method and system for predicting wear life of overflow surface of release structure

Abstract

The invention provides a method and a system for predicting the wear life of an overflow surface of a release structure, and is suitable for the technical field of hydraulic engineering monitoring. The prediction method comprises the steps of data acquisition equipment arrangement, model training, actual measurement data acquisition, model application and result display and maintenance, and a prediction system is designed based on the method. A sensor is arranged to collect relevant abrasion data of the flow passing surface of the release structure, and an abrasion depth prediction model is designed based on a double-channel structure; static data features are extracted through full connection layer coding, long-term dependence features are extracted through LSTM and an attention mechanism and fused to obtain a comprehensive feature vector, and finally the wear depth is predicted; dynamic prediction of the wear life is achieved through model deployment and periodic actual measurement data input, the model is updated in combination with nondestructive testing data, and the prediction precision is improved; the method effectively solves the problem that the service life of the overflow face of the water drainage building is abraded in the complex water flow environment.

Description

Technical Field

[0001] The invention relates to the technical field of water conservancy project monitoring, and in particular to a method and system for predicting the wear life of a flow surface of a water discharge structure. Background Art

[0002] As an important structure in water conservancy projects, the wear performance of the flow surface of the discharge structure directly affects the safety and service life of the structure. Since the flow surface of the discharge structure is often subjected to multiple effects of high-speed water flow, cavitation bubbles and particles entrained by water flow during operation, its surface material often faces severe wear, scouring and cavitation damage problems.

[0003] In actual projects, existing technologies mainly predict and evaluate the wear of the flow surface of the spillway structure through laboratory simulation, and further use computational fluid dynamics (CFD) technology to simulate the flow field distribution, cavitation characteristics and shear stress distribution of high-speed water flow on the flow surface of the spillway structure, and combine material performance parameters to predict wear. However, laboratory simulation and computer simulation methods rely on preset environments or offline calculations, and cannot respond to the actual operating conditions of the flow surface of the spillway structure in real time, such as the dynamic changes of flow velocity and particle concentration, and are difficult to meet prediction needs in a rapidly changing water flow environment.

[0004] In addition, empirical formulas such as particle erosion rate and cavitation loss rate have been developed based on experiments and field observations to estimate the wear rate or life of the flow surface materials of the spillway structure. This method is usually based on simplified assumptions and is only applicable to predictions under specific conditions. It cannot effectively reflect the coupling effects of multiple factors such as water flow, particles, and temperature in complex flow fields. These formulas ignore the nonlinear effects of dynamic water flow characteristics, such as flow velocity fluctuations and shear stress changes, on wear behavior, resulting in large deviations in the prediction results.

[0005] Therefore, there is an urgent need for a method to predict the wear life of the flow surface of a spillway structure that can accurately predict and respond to dynamic changes in operating conditions. Summary of the invention

[0006] The purpose of the present invention is to solve the problem that the wear prediction of the flow surface of traditional water discharge structures cannot dynamically respond to changes in actual operating conditions and the prediction is inaccurate. A wear life prediction method and system for the flow surface of a water discharge structure are proposed, which makes the wear prediction of the flow surface of a water discharge structure more intelligent and accurate. At the same time, alarms can be provided in time to remind maintenance personnel to take maintenance measures. The method can be widely used in the field of water conservancy project monitoring technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The wear life prediction method of the flow surface of the discharge structure specifically includes the following steps:

[0009] S101. Arrangement of data acquisition equipment: Install the flow velocity sensor on the flow surface of the water discharge structure, and adjust the installation position and angle of the sensor according to the water depth and flow velocity range; determine the cavitation sensitive area and rapid flow area of ​​the flow surface of the water discharge structure model according to CFD numerical simulation, and arrange pressure sensors and wall shear stress sensors in the corresponding areas of the actual flow surface of the water discharge structure; set temperature sensors in the area where the water flow is relatively gentle near the flow surface of the water discharge structure;

[0010] S102, model training, performing data preprocessing based on an existing data set, wherein each data sample of the existing data set includes static data independent of time and time-related time series data, wherein the static data includes compressive strength, tensile strength, surface hardness, and impact and abrasion resistance of impact-resistant and wear-resistant concrete, and the time series data includes water flow velocity, shear stress, cavitation number, temperature, fluid pressure, particle concentration, particle size, and particle hardness;

[0011] A dual-pathway structure is adopted. The static path extracts features from static data, and the dynamic path extracts features from time series data. The extracted static features and dynamic features are fused to predict the wear depth. The static feature extraction module of the static path uses a fully connected layer to encode the static features, and the activation function uses ReLU to transform the static features into a high-dimensional vector representation to provide input for subsequent fusion. The dynamic feature extraction module of the dynamic path uses two layers of LSTM to extract long-term dependency features in the time series. The first layer outputs the hidden state of the entire time series, and the second layer only outputs the hidden state of the last time step. Then, an attention layer is added to further process the features, and the output is a high-dimensional representation of the dynamic features. The static features output by the static path and the dynamic features output by the dynamic path are fused into a comprehensive feature vector through a splicing operation. The fused comprehensive feature vector is further processed through two fully connected layers to extract nonlinear relationships, and then the wear depth is predicted at the output layer.

[0012] The preprocessed existing data set is divided into training set, validation set and test set in a ratio of 70%:15%:15%. Based on the dual-channel structure, the mean square error is used as the loss function, and the training set data is used to train the model; after the model training is completed, the test set is used for performance evaluation;

[0013] S103, measured data collection, static data are collected after the completion of the construction of the discharge structure, wherein the compressive strength and tensile strength of the impact-resistant and wear-resistant concrete on the flow surface of the discharge structure are measured by standard mechanical experiments based on on-site test blocks, and the surface hardness is measured by using a portable hardness tester on the impact-resistant and wear-resistant concrete on the flow surface of the discharge structure; an underwater steel ball method impact-resistant and wear-resistant test device is used to cause wear on the surface of the impact-resistant and wear-resistant concrete specimen by stirring the steel ball in the device, and the impact-resistant and wear strength is calculated based on the mass lost by the impact-resistant and wear-resistant concrete specimen per unit time; time series data are collected periodically, wherein the water flow velocity, shear stress, temperature and fluid pressure are collected based on the sensors arranged in step S101, and the cavitation number n is calculated by the local fluid pressure p, the saturated vapor pressure p of water v , water density ρ, and water flow velocity v,

[0014]

[0015] Particle concentration, particle size, and particle hardness are analyzed in the laboratory by periodic sampling of the water stream;

[0016] S104, model application, saving the trained and tested model, building a model deployment environment on the cloud server, completing model deployment, implementing model call through the REST API interface, using the wear life prediction system of the flow surface of the discharge structure to pre-process the collected measured data, calling the REST API interface to input the pre-processed measured data into the model, and obtaining the wear depth prediction value within the cycle;

[0017] S105. Result display and maintenance. The wear life prediction system displays the predicted wear depth of each cycle through a line graph, and displays the cumulative wear depth through a bar graph. When the cumulative wear depth exceeds the thickness of the impact-resistant and wear-resistant concrete on the flow surface of the spillway structure, a maintenance reminder is triggered. The maintenance personnel take maintenance measures on the flow surface of the spillway structure and adjust the model according to the actual wear conditions.

[0018] As a preferred technical solution of the present invention, in step S102, the data samples of the existing data set are obtained through actual existing project collection, CFD numerical simulation acquisition and scaled model experimental testing.

[0019] As a preferred technical solution of the present invention, in step S102, the time series data of the existing data sets have the same period.

[0020] As a preferred technical solution of the present invention, in step S102, the data preprocessing includes normalization of static data, and normalization of time series data after eliminating noise using smoothing filtering.

[0021] As a preferred technical solution of the present invention, in step S104, the data preprocessing is as follows: for missing values ​​and abnormal values ​​of static data, if it is the compressive strength and tensile strength of impact-resistant and wear-resistant concrete, the approximate value is obtained according to the on-site construction record. When there is no approximate value in the construction record, the theoretical value is obtained according to the design data. If it is the surface hardness and impact-resistant and wear-resistant strength, the test piece can be additionally made and then re-measured; for missing values ​​and abnormal values ​​of time series data, linear interpolation is used for filling. If a large amount of data is missing and abnormal during the entire acquisition period, the corresponding data samples are deleted; for samples with inconsistent periods in the time series data, the data are aligned by padding or truncation.

[0022] As a preferred technical solution of the present invention, in step S105, the maintenance measures include using non-destructive testing technology to measure the specific wear depth of the flow surface of the spillway structure, and using higher performance impact-resistant and wear-resistant concrete to repair areas with severe local wear. When the wear surface is very large and local repairs cannot meet the requirements for continued use, the entire flow surface of the spillway structure needs to be recast.

[0023] As a preferred technical solution of the present invention, in step S105, the adjustment of the model includes measuring the specific wear depth value of the flow surface of the discharge structure according to non-destructive testing after starting the maintenance of the flow surface of the discharge structure, proportionally adjusting the wear depth prediction value in the historical period so that the cumulative wear depth prediction value is consistent with the specific wear depth value, and then using the adjusted wear depth prediction value and the static data and time series data of the corresponding period as a new data set for updating and maintaining the model.

[0024] The wear life prediction system of the flow surface of the water discharge structure is used for the wear life prediction method of the flow surface of the water discharge structure, including a data acquisition module, a data processing module, a model prediction module, a result analysis and alarm module, a user interaction module and a model update module;

[0025] In the data acquisition module, static data is manually input, and time series data other than cavitation number, particle concentration, particle size, and particle hardness are collected by sending collection instructions to the sensor through a preset period. The cavitation number is obtained by automatic calculation, and the particle concentration, particle size, and particle hardness are obtained through experiments and then manually input;

[0026] The data processing module processes missing values ​​and abnormal values, deletes invalid data, and fills the collected time series data according to the curve change trend when it is less than one cycle, and cuts it to the length of one cycle when it exceeds one cycle;

[0027] The model prediction module inputs the pre-processed measured data into the model by calling the REST API interface to obtain the wear depth prediction value within the cycle;

[0028] The result analysis and alarm module accumulates the wear depth prediction values ​​in each cycle to obtain a cumulative wear depth prediction value, compares the cumulative wear depth prediction value with the preset impact-resistant and wear-resistant concrete thickness of the flow surface of the discharge structure, and triggers an alarm according to the comparison result;

[0029] The user interaction module provides a visual interface to display the wear depth prediction value and the cumulative wear depth prediction value in each cycle, and provides a historical data query page for static data and time series data;

[0030] The model updating module inputs the specific wear depth value of the flow surface of the discharge structure measured by non-destructive testing, adjusts the wear depth prediction value in the historical period so that the cumulative wear depth prediction value is consistent with the specific wear depth value, and inputs the adjusted wear depth prediction value and the static data and time series data of the corresponding period into the training model for continuous updating and maintenance.

[0031] The beneficial effects of the present invention are as follows: static data and time series data related to the wear of the flow surface of the discharge structure are collected by arranging flow rate sensors, pressure sensors, shear stress sensors and other equipment, and a wear depth prediction model is designed based on a dual-channel structure; static data is encoded and features are extracted through full-connection layer encoding, and time series data is extracted through LSTM and attention mechanism. Long-term dependent features are integrated into a comprehensive feature vector to finally predict the wear depth; through model deployment and periodic measured data input, dynamic prediction of wear life is achieved, maintenance reminders are triggered, and the model is adjusted and updated in combination with non-destructive testing data to improve prediction accuracy; this method effectively solves the problem that the existing technology cannot accurately predict the wear life of impact-resistant concrete surfaces in complex water flow environments, and realizes the automation of refined monitoring and dynamic maintenance decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of the wear life prediction method of the flow surface of the water discharge structure of the present invention;

[0033] Figure 2 It is a schematic diagram of the wear life prediction system of the flow surface of the water discharge structure of the present invention. DETAILED DESCRIPTION

[0034] The specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. It should be understood that the specific embodiments given here are only used to illustrate and explain the present invention and cannot be used to limit the present invention. It should be noted that many specific details are set forth in the following description to facilitate a full understanding of the present invention, but the present invention may also have other embodiments and variations thereof, and therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0035] Embodiment 1, the wear life prediction method of the flow surface of the discharge structure is as follows:

[0036] S101. Arrangement of data acquisition equipment. Install the flow velocity sensor on the flow surface of the water discharge structure. Adjust the installation position and angle of the sensor according to the water depth and flow velocity range. In order to ensure the accuracy of sensor measurement, select the area with large flow velocity changes in the water flow field, and the direction needs to be consistent with the water flow direction to reduce the measurement error caused by the installation angle deviation. According to the field test results, adjust the inclination angle of the sensor. Determine the cavitation sensitive area and rapid flow area of ​​the flow surface of the water discharge structure model according to CFD numerical simulation, and arrange pressure sensors and wall shear stress sensors in the corresponding area of ​​the actual water discharge structure flow surface. Set a temperature sensor in the area where the water flow is relatively gentle near the flow surface of the water discharge structure.

[0037] S102, model training, data preprocessing based on existing data sets, including static data normalization, time series data normalization after eliminating noise using smoothing filtering; each data sample of the existing data set includes static data independent of time and time series data related to time, the static data includes compressive strength, tensile strength, surface hardness, and impact and wear strength of impact-resistant and wear-resistant concrete, and the time series data includes water flow rate, shear stress, cavitation number, temperature, fluid pressure, particle concentration, particle size, and particle hardness; the data samples of the existing data set are obtained through actual existing project collection, CFD numerical simulation acquisition, and scaled model experimental testing, and the time series data of the existing data set have the same period;

[0038] A dual-pathway structure is adopted. The static path extracts features from static data, and the dynamic path extracts features from time series data. The extracted static features and dynamic features are fused to predict the wear depth. The static feature extraction module of the static path uses a fully connected layer to encode the static features, and the activation function uses ReLU to transform the static features into a high-dimensional vector representation to provide input for subsequent fusion. The dynamic feature extraction module of the dynamic path uses two layers of LSTM to extract long-term dependency features in the time series. The first layer outputs the hidden state of the entire time series, and the second layer only outputs the hidden state of the last time step. Then, an attention layer is added to further process the features, and the output is a high-dimensional representation of the dynamic features. The static features output by the static path and the dynamic features output by the dynamic path are fused into a comprehensive feature vector through a splicing operation. The fused comprehensive feature vector is further processed through two fully connected layers to extract nonlinear relationships, and then the wear depth is predicted at the output layer.

[0039] The preprocessed existing data set is divided into training set, validation set and test set in a ratio of 70%:15%:15%. Based on the dual-channel structure, the mean square error is used as the loss function, and the training set data is used to train the model; after the model training is completed, the test set is used for performance evaluation;

[0040] S103, measured data collection, static data are collected after the completion of the construction of the discharge structure, wherein the compressive strength and tensile strength of the impact-resistant and wear-resistant concrete on the flow surface of the discharge structure are measured by standard mechanical experiments based on on-site test blocks, and the surface hardness is measured by using a portable hardness tester on the impact-resistant and wear-resistant concrete on the flow surface of the discharge structure; an underwater steel ball method impact-resistant and wear-resistant test device is used to cause wear on the surface of the impact-resistant and wear-resistant concrete specimen by stirring the steel ball in the device, and the impact-resistant and wear strength is calculated based on the mass lost by the impact-resistant and wear-resistant concrete specimen per unit time; time series data are collected periodically, wherein the water flow velocity, shear stress, temperature and fluid pressure are collected based on the sensors arranged in step S101, and the cavitation number n is calculated by the local fluid pressure p, the saturated vapor pressure p of water v , water density ρ, and water flow velocity v,

[0041]

[0042] The saturated vapor pressure of water is p v It is related to temperature. The saturated vapor pressure table of water at different temperatures has been entered into the wear life prediction system of the flow surface of the discharge structure. After collecting the temperature of the water flow, the value of the saturated vapor pressure is automatically obtained. Some data of the saturated vapor pressure table of water at different temperatures are shown in Table 1; particle concentration, particle size and particle hardness are obtained by periodic sampling in the water flow and analyzed in the laboratory;

[0043] Table 1 Saturated vapor pressure of water at different temperatures

[0044] Temperature / ℃ Saturated vapor pressure / kPa Temperature / ℃ Saturated vapor pressure / kPa Temperature / ℃ Saturated vapor pressure / kPa 0 0.61129 15 1.7056 30 4.2455 1 0.65716 16 1.8185 31 4.4953 2 0.70605 17 1.9380 32 4.7578 3 0.75813 18 2.0644 33 5.0335 ... ... ... ... ... ...

[0045] S104, model application, save the trained and tested model, build the model deployment environment on the cloud server, complete the model deployment, and call the model through the REST API interface. The wear life prediction system of the flow surface of the discharge structure performs data preprocessing on the collected measured data, including missing values ​​and abnormal values ​​of static data. If it is the compressive strength and tensile strength of impact-resistant and wear-resistant concrete, the approximate value is obtained according to the on-site construction record. When there is no approximate value in the construction record, the theoretical value is obtained according to the design data. If it is the surface hardness and impact-resistant strength, the test piece can be supplemented and then re-measured; For missing values ​​and abnormal values ​​of time series data, linear interpolation is used for filling. If a large amount of data is missing and abnormal during the entire collection cycle, the corresponding data sample is deleted; For samples with inconsistent periods in the time series data, the data is aligned by filling or truncation; Call the REST API interface to input the preprocessed measured data into the model to obtain the wear depth prediction value within the cycle;

[0046] S105. Result display and maintenance. The wear life prediction system displays the predicted wear depth of each cycle through a line graph, and displays the cumulative wear depth through a bar graph. When the cumulative wear depth exceeds the thickness of the impact-resistant and wear-resistant concrete on the flow surface of the spillway structure, a maintenance reminder is triggered. The maintenance personnel take maintenance measures on the flow surface of the spillway structure and adjust the model according to the actual wear conditions.

[0047] In this embodiment, the maintenance measures include using non-destructive testing technology to measure the specific wear depth of the flow surface of the spillway structure. For areas with severe local wear, higher performance impact-resistant and wear-resistant concrete is used for repair. When the wear surface is very large and local repairs cannot meet the requirements for continued use, the entire flow surface of the spillway structure is re-cast.

[0048] In this embodiment, the model is adjusted including starting the maintenance of the flow surface of the water discharge structure, measuring the specific wear depth value of a certain point on the flow surface of the water discharge structure according to non-destructive testing as A, and the wear depth prediction values ​​in the historical period are a1, a2, ..., a m , the wear depth prediction value in the historical period is adjusted proportionally to make the cumulative wear depth prediction value consistent with the specific wear depth value, and the adjusted wear depth prediction value a′1, a′2, ..., a′ is obtained m ,

[0049]

[0050] The adjusted wear depth prediction value and the static data and time series data of the corresponding cycle are then used as new data sets for model update and maintenance to further improve the accuracy of model prediction.

[0051] Embodiment 2, a wear life prediction system for the flow surface of a discharge structure, includes a data acquisition module, a data processing module, a model prediction module, a result analysis and alarm module, a user interaction module and a model update module.

[0052] In the data acquisition module, static data is manually entered. Except for the cavitation number, particle concentration, particle size, and particle hardness, the time series data is collected by sending collection instructions to the sensor through a preset period. The cavitation number is automatically calculated according to formula (1) based on the entered data in Table 1. The particle concentration, particle size, and particle hardness are obtained through experiments and then manually entered.

[0053] The data processing module handles missing values ​​and outliers, deletes invalid data, and fills in the collected time series data according to the curve change trend when it is less than one cycle. When it exceeds one cycle, it is truncated to the length of one cycle.

[0054] The model prediction module inputs the preprocessed measured data into the model by calling the REST API interface to obtain the predicted value of the wear depth within the cycle.

[0055] The result analysis and alarm module accumulates the wear depth prediction values ​​in each cycle to obtain the cumulative wear depth prediction value, compares the cumulative wear depth prediction value with the preset impact-resistant and wear-resistant concrete thickness of the flow surface of the discharge structure, and triggers an alarm based on the comparison result.

[0056] The user interaction module provides a visual interface to display the wear depth prediction value and the cumulative wear depth prediction value in each cycle, and provides a historical data query page for static data and time series data.

[0057] The model update module inputs the specific wear depth value of the flow surface of the discharge structure measured by non-destructive testing, adjusts the wear depth prediction value in the historical period to make the cumulative wear depth prediction value consistent with the specific wear depth value, and inputs the adjusted wear depth prediction value and the static data and time series data of the corresponding period into the training model for continuous updating and maintenance.

[0058] In summary, the wear life prediction method and system of the flow surface of a discharge structure of the present invention has the characteristics of more intelligent and accurate wear prediction and automatic alarm when the wear life limit is reached in the field of water conservancy project monitoring technology.

[0059] It should be understood that the above-mentioned embodiments are one or more embodiments of the present invention, and there are many other embodiments and variations thereof based on the present invention; the variations and modifications made by ordinary technicians in this industry through the present invention without making pioneering innovations all fall within the scope of protection of the present invention.

Claims

1. The wear life prediction method of the flow surface of the discharge structure is characterized by The specific steps include: S101. Arrangement of data acquisition equipment: Install the flow velocity sensor on the flow surface of the water discharge structure, and adjust the installation position and angle of the sensor according to the water depth and flow velocity range; determine the cavitation sensitive area and rapid flow area of ​​the flow surface of the water discharge structure model according to CFD numerical simulation, and arrange pressure sensors and wall shear stress sensors in the corresponding areas of the actual flow surface of the water discharge structure; set temperature sensors in the area where the water flow is relatively gentle near the flow surface of the water discharge structure; S102, model training, performing data preprocessing based on an existing data set, wherein each data sample of the existing data set includes static data independent of time and time-related time series data, wherein the static data includes compressive strength, tensile strength, surface hardness, and impact and abrasion resistance of impact-resistant and wear-resistant concrete, and the time series data includes water flow velocity, shear stress, cavitation number, temperature, fluid pressure, particle concentration, particle size, and particle hardness; A dual-pathway structure is adopted. The static path extracts features from static data, and the dynamic path extracts features from time series data. The extracted static features and dynamic features are fused to predict the wear depth. The static feature extraction module of the static path uses a fully connected layer to encode the static features, and the activation function uses ReLU to transform the static features into a high-dimensional vector representation to provide input for subsequent fusion. The dynamic feature extraction module of the dynamic path uses two layers of LSTM to extract long-term dependency features in the time series. The first layer outputs the hidden state of the entire time series, and the second layer only outputs the hidden state of the last time step. Then, an attention layer is added to further process the features, and the output is a high-dimensional representation of the dynamic features. The static features output by the static path and the dynamic features output by the dynamic path are fused into a comprehensive feature vector through a splicing operation. The fused comprehensive feature vector is further processed through two fully connected layers to extract nonlinear relationships, and then the wear depth is predicted at the output layer. The preprocessed existing data set is divided into training set, validation set and test set in a ratio of 70%:15%:15%. Based on the dual-channel structure, the mean square error is used as the loss function, and the training set data is used to train the model; after the model training is completed, the test set is used for performance evaluation; S103, measured data collection, static data are collected after the completion of the construction of the discharge structure, wherein the compressive strength and tensile strength of the impact-resistant and wear-resistant concrete on the flow surface of the discharge structure are measured by standard mechanical experiments based on on-site test blocks, and the surface hardness is measured by using a portable hardness tester on the impact-resistant and wear-resistant concrete on the flow surface of the discharge structure; an underwater steel ball method impact-resistant and wear-resistant test device is used to cause wear on the surface of the impact-resistant and wear-resistant concrete specimen by stirring the steel ball in the device, and the impact-resistant and wear strength is calculated based on the mass lost by the impact-resistant and wear-resistant concrete specimen per unit time; time series data are collected periodically, wherein the water flow velocity, shear stress, temperature and fluid pressure are collected based on the sensors arranged in step S101, and the cavitation number n is calculated by the local fluid pressure p, the saturated vapor pressure p of water v , water density ρ, and water flow velocity v, Particle concentration, particle size, and particle hardness are analyzed in the laboratory by periodic sampling of the water stream; S104, model application, saving the trained and tested model, building a model deployment environment on the cloud server, completing model deployment, implementing model call through the REST API interface, using the wear life prediction system of the flow surface of the discharge structure to pre-process the collected measured data, calling the REST API interface to input the pre-processed measured data into the model, and obtaining the wear depth prediction value within the cycle; S105. Result display and maintenance. The wear life prediction system displays the predicted wear depth of each cycle through a line graph, and displays the cumulative wear depth through a bar graph. When the cumulative wear depth exceeds the thickness of the impact-resistant and wear-resistant concrete on the flow surface of the spillway structure, a maintenance reminder is triggered. The maintenance personnel take maintenance measures on the flow surface of the spillway structure and adjust the model according to the actual wear conditions.

2. The wear life prediction method of the flow surface of a discharge structure according to claim 1, characterized in that: In step S102, data samples of the existing data set are obtained through actual existing project collection, CFD numerical simulation acquisition, and scaled model experimental testing.

3. The wear life prediction method of the flow surface of a discharge structure according to claim 1, characterized in that: In step S102, the time series data of the existing data sets have the same period.

4. The wear life prediction method of the flow surface of a discharge structure according to claim 1 is characterized in that: In step S102, the data preprocessing includes normalizing static data and normalizing the time series data after eliminating noise using smoothing filtering.

5. The wear life prediction method of the flow surface of a discharge structure according to claim 1, characterized in that: In step S104, the data preprocessing is as follows: for missing values ​​and abnormal values ​​of static data, if it is the compressive strength and tensile strength of impact-resistant and wear-resistant concrete, the approximate value is obtained according to the on-site construction record. When there is no approximate value in the construction record, the theoretical value is obtained according to the design data. If it is surface hardness and impact-resistant and wear-resistant strength, additional test pieces can be made and then re-measured; for missing values ​​and abnormal values ​​of time series data, linear interpolation is used to fill them. If a large amount of data is missing and abnormal during the entire acquisition period, the corresponding data samples are deleted; for samples with inconsistent periods in the time series data, the data are aligned by padding or truncation.

6. The wear life prediction method of the flow surface of a discharge structure according to claim 1, characterized in that: In step S105, the maintenance measures include using non-destructive testing technology to measure the specific wear depth of the flow surface of the spillway structure. For areas with severe local wear, higher performance impact-resistant and wear-resistant concrete is used for repair. When the wear surface is very large and local repairs cannot meet the requirements for continued use, the entire flow surface of the spillway structure needs to be recast.

7. The wear life prediction method of the flow surface of a discharge structure according to claim 1 is characterized in that: In step S105, the adjustment of the model includes measuring the specific wear depth value of the flow surface of the discharge structure according to non-destructive testing after starting the maintenance of the flow surface of the discharge structure, proportionally adjusting the wear depth prediction value in the historical period so that the cumulative wear depth prediction value is consistent with the specific wear depth value, and then using the adjusted wear depth prediction value and the static data and time series data of the corresponding period as a new data set for updating and maintaining the model.

8. A system for predicting the wear life of a flow surface of a discharge structure, used in a method for predicting the wear life of a flow surface of a discharge structure as claimed in any one of claims 1 to 7, characterized in that: Specifically, it includes data acquisition module, data processing module, model prediction module, result analysis and alarm module, user interaction module and model update module; In the data acquisition module, static data is manually input, and time series data other than cavitation number, particle concentration, particle size, and particle hardness are collected by sending collection instructions to the sensor through a preset period. The cavitation number is obtained by automatic calculation, and the particle concentration, particle size, and particle hardness are obtained through experiments and then manually input; The data processing module processes missing values ​​and abnormal values, deletes invalid data, and fills the collected time series data according to the curve change trend when it is less than one cycle, and cuts it to the length of one cycle when it exceeds one cycle; The model prediction module inputs the pre-processed measured data into the model by calling the REST API interface to obtain the wear depth prediction value within the cycle; The result analysis and alarm module accumulates the wear depth prediction values ​​in each cycle to obtain a cumulative wear depth prediction value, compares the cumulative wear depth prediction value with the preset impact-resistant and wear-resistant concrete thickness of the flow surface of the discharge structure, and triggers an alarm according to the comparison result; The user interaction module provides a visual interface to display the wear depth prediction value and the cumulative wear depth prediction value in each cycle, and provides a historical data query page for static data and time series data; The model updating module inputs the specific wear depth value of the flow surface of the discharge structure measured by non-destructive testing, adjusts the wear depth prediction value in the historical period so that the cumulative wear depth prediction value is consistent with the specific wear depth value, and inputs the adjusted wear depth prediction value and the static data and time series data of the corresponding period into the training model for continuous updating and maintenance.

Images