A method and electronic device for predicting settlement of a filling area

Abstract

The present application provides a method and electronic device for predicting settlement in a filling area, the method comprising the following steps: obtaining historical settlement data, predicted engineering quantity data and predicted meteorological data of the filling area; performing polynomial fitting on the historical settlement data to obtain a settlement fitting polynomial; inputting the predicted engineering quantity data and predicted meteorological data into a support vector regression SVR model to obtain a prediction model fitting residual; and obtaining predicted settlement data of the filling area based on the settlement fitting polynomial and the prediction model fitting residual. The solution provided by the present application combines polynomial fitting with a support vector regression model, and can predict settlement problems in complex engineering environments. Polynomial fitting processes the overall trend of settlement data, and the SVR model compensates for the shortcomings of polynomial fitting in details, especially when dealing with complex engineering environments and changeable meteorological conditions.

Description

Technical Field

[0001] The invention relates to the fields of settlement monitoring, numerical modeling and engineering geology, and in particular to a method and electronic equipment for predicting settlement in a filling area. Background Art

[0002] With the rapid development of economy and urbanization in coastal areas, the shortage of land resources has become increasingly prominent, and land reclamation has become an important means to solve the problem of land resource shortage. However, in land reclamation projects, the settlement and consolidation of fillers often lead to the settlement of soft soil foundations, which makes the engineering structure unstable and damaged, and poses a severe challenge to the stability and safety of the project. Therefore, accurately predicting the settlement of soft soil foundations in the reclamation area is of great significance to ensure the quality and safety of the project and optimize the construction process.

[0003] In recent years, domestic and foreign scholars have conducted a lot of research on the prediction of soft soil foundation settlement. Traditional methods mainly include empirical formula method, finite element analysis method, grey system theory, etc. These methods can predict the settlement well under specific conditions, but due to the complexity and variability of soft soil foundation, they are limited by specific geological conditions and engineering background, and the calculation complexity is high. It is difficult to adapt to the actual needs of different projects, and often shows large errors and limitations in practical applications.

[0004] Therefore, finding a method that can accurately predict the settlement of the filling area has become an urgent problem to be solved. Summary of the invention

[0005] The embodiments of the present application provide a method and electronic equipment for predicting settlement in a filling area, which comprehensively considers the influence of factors such as engineering volume and meteorological factors on the settlement. By adopting this method in a large-area filling area, the settlement can be predicted quickly, accurately and comprehensively, and the settlement change trend can be grasped.

[0006] An embodiment of the present application provides a method for predicting settlement of a filling area, comprising the following steps: acquiring historical settlement data, predicted engineering quantity data, and predicted meteorological data of the filling area; performing polynomial fitting on the historical settlement data to obtain a settlement fitting polynomial; inputting the predicted engineering quantity data and predicted meteorological data into a support vector regression SVR model to obtain a prediction model fitting residual; and obtaining predicted settlement data of the filling area based on the settlement fitting polynomial and the prediction model fitting residual.

[0007] In one embodiment, the polynomial fitting of the historical settlement data to obtain the settlement fitting polynomial includes: using the least squares method to perform polynomial fitting on the historical settlement data to obtain an n-order fitting polynomial; determining the degree of the n-order polynomial based on the root mean square difference of polynomials of different degrees on the test data set, and using it as the settlement fitting polynomial.

[0008] In one embodiment, the method further includes: determining a historical settlement fitting result based on the historical settlement data and the settlement fitting polynomial; determining a historical polynomial fitting residual based on the historical settlement data and the historical settlement fitting result; and constructing an initial SVR model using SVR based on the historical polynomial fitting residual.

[0009] In one embodiment, the method further includes: acquiring historical engineering quantity data and historical meteorological data of the filling area; and training the initial SVR model through cross-validation based on the historical engineering quantity data and historical meteorological data to obtain the SVR model.

[0010] In one embodiment, the initial SVR model is trained by cross-validation based on the historical engineering quantity data and the historical meteorological data to obtain the SVR model, including: splitting the historical engineering quantity data and the historical meteorological data into a training set and a test set; using the training set to train the initial SVR model to obtain a trained SVR model; inputting the test set into the trained SVR model to obtain the historical model fitting residual; if the difference between the historical polynomial fitting residual and the historical model fitting residual meets a preset condition, the trained SVR model is used as the SVR model.

[0011] In one embodiment, if the difference between the historical polynomial fitting residual and the historical model fitting residual does not meet a preset condition, the parameters of the trained SVR model are adjusted and the training is continued.

[0012] In one embodiment, inputting the predicted engineering quantity data and predicted meteorological data into the SVR model includes: performing standardized conversion on the predicted engineering quantity data and predicted meteorological data, and inputting the standardized converted data into the SVR model.

[0013] In one embodiment, the engineering quantity data includes: the amount of blown sand for preloading and the amount of bagged sand weir construction; the meteorological data includes: weather data, wind direction data, wind force data and temperature data; wherein, the engineering quantity data is historical engineering quantity data or predicted engineering quantity data, and the meteorological data is historical meteorological data or predicted meteorological data.

[0014] In one embodiment, the method further includes: preprocessing the original meteorology to obtain meteorological data; the preprocessing of the original meteorology to obtain meteorological data includes: segmenting the weather type and encoding the weather type to obtain weather data; segmenting the wind direction type, angle mapping and circular encoding to obtain wind direction data; using the daily average wind speed as wind speed data; and using the daily average temperature as temperature data.

[0015] An embodiment of the present application further provides an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the above-mentioned method for predicting settlement of a filling area.

[0016] The solution provided by the above embodiment of the present application combines polynomial fitting with the support vector regression model, which can predict settlement problems in complex engineering environments. Polynomial fitting processes the overall trend of settlement data, and the SVR model compensates for the shortcomings of polynomial fitting in details, especially when dealing with complex engineering environments and changeable meteorological conditions. The solution provided by the above embodiment comprehensively considers engineering quantity data and meteorological data, and the model has high applicability and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solution of the embodiments of the present application, the drawings required for use in the embodiments of the present application are briefly introduced below.

[0018] Figure 1 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application;

[0019] Figure 2 It is a flow chart of a method for predicting settlement of a filling area provided in an embodiment of the present application;

[0020] Figure 3 is a wind direction-angle mapping relationship diagram provided in an embodiment of the present application;

[0021] Figure 4 It is a polynomial fitting result diagram provided in an embodiment of the present application;

[0022] Figure 5 This is an output result diagram of an SVR model provided in an embodiment of the present application on a test set;

[0023] Figure 6 It is a predicted settlement data diagram of a comprehensive polynomial fitting and SVR model prediction results provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0025] Similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0026] With the development of data-driven technology, machine learning methods have gradually been applied to the field of soft soil foundation settlement prediction in filling areas. Compared with traditional methods, machine learning methods can process a large amount of complex data, capture nonlinear relationships, and improve the accuracy and reliability of predictions. Machine learning algorithms such as artificial neural networks (ANN), support vector machines (SVM), and random forests (RF) have been successfully applied to settlement prediction research. These methods can automatically learn and extract features from a large amount of historical data without complex theoretical derivation and calculation, and have good generalization ability and prediction accuracy. Although machine learning methods have achieved certain results in soft soil foundation settlement prediction, most existing studies are limited to a single algorithm or model, and have not fully considered the impact of multiple environmental factors on settlement, especially external conditions such as engineering volume and meteorological factors.

[0027] Figure 1 1 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device 100 can be used to execute the method for predicting the settlement of a filling area provided in an embodiment of the present application. Figure 1 As shown, the electronic device 100 includes: one or more processors 102 and one or more memories 104 storing instructions executable by the processors. The processor 102 is configured to execute the method for predicting settlement of a filling area provided in the following embodiment of the present application.

[0028] The processor 102 can be a gateway, a smart terminal, or a device including a central processing unit (CPU), a graphics processing unit (GPU), or other forms of processing units with data processing capabilities and / or instruction execution capabilities. It can process data of other components in the electronic device 100 and control other components in the electronic device 100 to perform desired functions.

[0029] The memory 104 may include one or more computer program products, and the computer program products may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 102 may run the program instructions to implement the method for predicting settlement of the filling area described below. Various applications and various data may also be stored in the computer-readable storage medium, such as various data used and / or generated by the application.

[0030] In one embodiment, Figure 1 The electronic device 100 shown may also include an input device 106, an output device 108, and a data acquisition device 110, which are interconnected via a bus system 112 and / or other forms of connection mechanisms (not shown). Figure 1 The components and structures of the electronic device 100 shown are merely exemplary and non-limiting. The electronic device 100 may also have other components and structures as required.

[0031] The input device 106 may be a device for a user to input instructions, and may include one or more of a keyboard, a mouse, a microphone, a touch screen, etc. The output device 108 may output various information (e.g., images or sounds) to the outside (e.g., a user), and may include one or more of a display, a speaker, etc. The data acquisition device 110 may collect relevant data of the filling area, such as historical settlement data, historical engineering quantity data, historical meteorological data, predicted engineering quantity data, and predicted meteorological data, and store the collected data in the memory 104 for use by other components. Exemplarily, the data acquisition device 110 may be a camera.

[0032] In one embodiment, the various components in the example electronic device 100 for implementing the method for predicting settlement of a fill area of ​​the embodiment of the present application may be integrated or dispersed, such as integrating the processor 102, memory 104, input device 106 and output device 108 into one, while separating the data acquisition device 110.

[0033] In one embodiment, the exemplary electronic device 100 for implementing the method for predicting settlement of a filling area according to the embodiment of the present application may be implemented as an intelligent terminal such as a smart phone, a tablet computer, a desktop computer, a server, a vehicle-mounted device, or the like.

[0034] Figure 2 FIG. 1 is a flow chart of a method for predicting settlement of a filling area according to an embodiment of the present application. Figure 2 As shown, the method includes the following steps 210-240.

[0035] Step 210: Obtain historical settlement data, predicted engineering quantity data and predicted meteorological data of the filling area.

[0036] Relevant data of the filling area may include settlement data, engineering quantity data and meteorological data.

[0037] Settlement data refers to the data pair of time and settlement value of the filling area. (time, settlement value) can be used. The unit of time can be day, hour, minute, etc. This application does not limit it, but generally speaking, day is the most common time unit. Settlement data includes historical settlement data and predicted settlement data.

[0038] The historical settlement data refers to the actual settlement values ​​of the filling area every day in the past. Several settlement bars can be arranged in the filling area. The settlement data of these settlement bars can be summarized and averaged on a daily basis, and the average settlement value of the filling area every day can be used as the historical settlement data.

[0039] Regarding the predicted settlement data, it is the settlement data of the future time predicted by various methods. Traditional prediction methods can include empirical formula method, finite element analysis method, and gray system theory. The predicted settlement data of these methods have large errors with the actual situation. With the development of data-driven technology, machine learning methods are applied to the acquisition of predicted settlement data. Although the predicted settlement data obtained by machine learning methods have smaller errors than those obtained by traditional prediction methods, their algorithms or models are single, and the influence of various environmental factors on the predicted settlement data is not considered. Therefore, the accuracy of the predicted settlement data obtained by the machine learning method is still not guaranteed. The method of the present application is also used to obtain predicted settlement data. It is a settlement prediction method for filling areas based on polynomial fitting and support vector regression (SVR). It comprehensively considers the influence of engineering volume, meteorological factors and other factors on settlement. The method is used in large-scale filling areas to quickly, accurately and comprehensively predict settlement conditions and grasp the trend of settlement changes.

[0040] The engineering quantity data includes historical engineering quantity data and predicted engineering quantity data. Historical engineering quantity data refers to the actual engineering quantity of the filling area every day in the past, which can be accurately measured and recorded in the actual project. Predicted engineering quantity data refers to the predicted engineering quantity in the future. The predicted engineering quantity data can be predicted based on the historical engineering quantity data, or it can be determined according to the schedule of the engineering plan for the filling area. Whether it is historical engineering quantity data or predicted engineering quantity data, it can include the amount of filling sand for preloading and the amount of bagged sand weir construction. The amount of filling sand for preloading is considered because preloading accelerates the discharge of pore water in the foundation by applying an overload, thereby accelerating the consolidation of the foundation, improving the bearing capacity and stability of the foundation, and directly affecting the settlement of the foundation. The construction amount of bagged sand weir is considered because the construction of bagged sand weir directly affects the initial bearing capacity and stability of the foundation. Sufficient and uniform sand accumulation helps to form a stable foundation, which indirectly affects the settlement rate and settlement of the foundation.

[0041] Meteorological data includes historical meteorological data and forecast meteorological data. Historical meteorological data refers to the actual meteorological data of the filling area in the past and every day, which can be prepared for measurement and record in actual engineering. Forecast meteorological data refers to the predicted meteorological data of the future time, which is generally based on the weather forecast. Whether it is historical meteorological data or forecast meteorological data, it can include weather data, wind direction data, wind force data and temperature data. Weather data is considered because changes in precipitation and humidity under different weather conditions will directly affect the humidity and stability of the foundation. For example, rainfall will increase the moisture content in the foundation, which may cause the foundation to soften and settle. Wind direction data is considered because changes in wind direction may bring different air humidity and temperature, affect the dryness and wetness of the foundation surface, and cause uneven settlement of the foundation surface. Especially under windy weather conditions, changes in wind direction may affect the settlement process of the foundation. Wind force data is considered because strong winds will accelerate the evaporation of moisture on the foundation surface, affect the dryness of the foundation, and thus affect the settlement of the foundation. In addition, windy weather may affect the construction progress and quality, thereby indirectly affecting the settlement of the foundation. Temperature data is considered because temperature changes will affect the evaporation of water in the foundation and the consolidation rate of the soil, thereby affecting the settlement process of the foundation. In addition, temperature changes can reflect seasonal changes and have important reference value for long-term foundation settlement prediction.

[0042] Since meteorological data may not be numerical, there will be certain obstacles for the subsequent training and testing of SVR models. Therefore, it is necessary to preprocess the original meteorological data to obtain meteorological data. The original meteorological data includes weather type, wind direction type, wind speed and temperature.

[0043] For the preprocessing of weather types, weather types can be segmented and coded to obtain weather data. Regarding weather type segmentation, considering that weather types may change within a day, weather descriptions containing the word "turn" are segmented, and the weather before and after the turning point are coded separately. For example, "sunny turning cloudy" is segmented into "sunny" and "cloudy". If the weather description does not contain the word "turn", it is considered that the weather of the day has not changed, and the previous weather and the later weather are the same. Regarding weather type coding, in order to quantify the impact of weather on foundation settlement, the concept of precipitation coding level is introduced according to the difference in precipitation in different weather, and corresponding codes are assigned to different weather types. The weather type coding mapping is shown in Table 1. This mapping converts the weather description into numerical form for easy use in the model.

[0044] Table 1 Weather type code mapping table

[0045] Weather Type coding Weather Type coding clear 0 Thunderstorm 6 partly cloudy 1 light to moderate rain 7 Negative 2 moderate rain 8 fog 3 Moderate to heavy rain 9 light rain 4 heavy rain 10 shower 5 Sleet 11

[0046] For the preprocessing of wind direction types, the wind direction types can be segmented, angle-mapped, and circularly encoded to obtain wind direction data. Regarding wind direction type segmentation, similar to weather type segmentation, considering that the wind direction may change within a day, wind direction descriptions containing the word "to" (such as "southeast to south wind") are segmented, and the wind directions before and after the turn are extracted separately. If the wind direction description does not contain the word "to", it is considered that the wind direction does not change on that day, and the previous wind direction and the later wind direction are the same. Regarding angle mapping, to further quantify the wind direction data, the wind direction is mapped to an angle, with due north being 0°, and the angle is measured clockwise relative to the due north direction. The specific mapping relationship is as Figure 3 shown, with due north being 0°, northeast being 45°, due east being 90°, southeast being 135°, south being 180°, southwest being 225°, west being 270°, and northwest being 315°. Regarding circular encoding, since the wind direction is a periodic data, directly using the angle values may introduce discontinuity problems in data processing. To solve this problem, this paper adopts the method of circular encoding to convert the wind direction angle into two-dimensional sine and cosine values, and records the sine and cosine values of the wind direction angle. Through circular encoding, not only the information of wind direction changes is retained, but also the periodic characteristics of the wind direction data are maintained, which helps to quantify the impact of wind direction on foundation settlement.

[0047] For the preprocessing of wind force and temperature, since wind force and temperature are usually represented in numerical form, the daily average wind force can be used as the wind force data, and the daily average temperature can be used as the temperature data.

[0048] In step 210, the electronic device 100 can obtain the historical settlement data, predicted engineering data, and predicted meteorological data of the reclamation area. The historical settlement data, predicted engineering data, and predicted meteorological data of the reclamation area can be pre-stored in the electronic device 100, or stored outside the electronic device 100 and obtained through the data acquisition device 110.

[0049] Step 220: Perform polynomial fitting on the historical settlement data to obtain a settlement fitting polynomial.

[0050] As introduced above, the historical settlement data refers to the actual settlement values of the reclamation area every day in the past. A polynomial of degree n can be constructed to fit the actual settlement amount for each day. Specifically, the least squares method is used to perform polynomial fitting on the historical settlement data to obtain a polynomial of degree n. The historical settlement data can use data points ( X i , Y i ) to represent ( i = 0, 1,..., m ), where X i represents thei The time of sedimentation data, Y i Indicates i The settlement value of the settlement data. The n-th order fitting polynomial is: , where a k are polynomial coefficients. The least squares method makes the data points ( X i , Y i ) and the n-th degree fitting polynomial The error between them is minimized to capture the main trends of the settlement values ​​over time:

[0051] Formula (1)

[0052] The formula (1) can be used to adjust a k The value of is used to obtain the n-order fitting polynomial. n is generally greater than 2 and less than 6. If n is too large, overfitting may occur.

[0053] After obtaining the n-order fitting polynomial, the degree of the n-order polynomial is determined according to the root mean square error of different-order polynomials on the test data set, and used as the sedimentation fitting polynomial. The root mean square error is also known as RMSE, and its calculation formula is as follows:

[0054] Formula (2)

[0055] In formula (2), Y i Indicates i Settlement data (for example, i day’s sedimentation value), P n ( X i ) is the i The polynomial fitting result of the data (that is, the sedimentation value of the polynomial fitting), m is the number of sedimentation data, and n is the number of polynomial fitting.

[0056] By comparing the RMSE of polynomials of different degrees on independent test data sets through formula (2), the polynomial with the smallest RMSE is selected as the sedimentation fitting polynomial. That is, the degree of the n-th degree polynomial that minimizes the RMSE is determined, and the polynomial of this degree is used as the sedimentation fitting polynomial.

[0057] The test data set can be the historical settlement data of the filling area for any period of time. For example, it can be the historical settlement data of the filling area for any 7 days. By comparing the actual settlement value with the settlement value of the polynomial fitting, a suitable settlement fitting polynomial can be finally obtained.

[0058] For example, the test data set can be the actual settlement value of the filling area every day from January 1, 2024 to January 7, 2024, which can be determined by actual measurement. k The n-order fitting polynomial with values , take n=2, use the quadratic fitting polynomial to obtain the fitted settlement values ​​of the filling area every day from January 1, 2024 to January 7, 2024, and then calculate the RMSE when n=2 based on the actual settlement value and the fitted settlement value using formula (2). Similarly, the RMSE values ​​when n=3, n=4, n=5, etc. can also be obtained. When comparing the RMSE values ​​for different n values, assuming that the RMSE is the smallest when n=4, the quadratic fitting polynomial is used as the settlement fitting polynomial.

[0059] The appropriate settlement fitting polynomial can accurately predict the settlement value to a certain extent, but the settlement value determined by the polynomial fitting will be different from the actual settlement value, and this difference is called the polynomial fitting residual. Therefore, in order to obtain more accurate predicted settlement data, the polynomial fitting residual can be considered in advance, and the predicted settlement value obtained by the settlement fitting polynomial and the polynomial fitting residual considered in advance can be used together to determine the final predicted settlement data.

[0060] Step 230: Input the predicted engineering quantity data and predicted meteorological data into a support vector regression SVR model to obtain the prediction model fitting residual.

[0061] Before step 230, it is necessary to obtain an SVR model first, obtain the prediction model fitting residual through the SVR model that meets the preset conditions, and then determine the predicted settlement data together with the settlement fitting polynomial obtained in step 220. The specific process of obtaining the SVR model can be divided into two stages: the first stage is to obtain the initial SVR model, and the second stage is to train the initial SVR model using a cross-validation method to obtain an SVR model that meets the preset conditions.

[0062] The above-mentioned first stage specifically includes: determining the historical settlement fitting result based on the historical settlement data and the settlement fitting polynomial; determining the historical polynomial fitting residual based on the historical settlement data and the historical settlement fitting result; and constructing the initial SVR model using SVR based on the historical polynomial fitting residual.

[0063] In step 220, a suitable settlement fitting polynomial has been obtained. The time in the historical settlement data can be used to obtain the historical fitting settlement value on the settlement fitting polynomial (substituting the historical time into the settlement fitting polynomial), that is, the historical settlement fitting result. Then, the settlement value of the filling area in the historical settlement data (that is, the actual settlement value) and the historical fitting settlement value are used to determine the historical polynomial fitting residual. Specifically, the difference between the actual settlement value and the historical fitting settlement value can be calculated as the historical polynomial fitting residual. After obtaining the historical polynomial fitting residual, the residual can be further modeled using SVR based on the residual to obtain an initial SVR model.

[0064] Since the initial SVR model has not been trained and evaluated, its prediction accuracy and stability cannot be guaranteed. Therefore, it is necessary to carry out the second stage to obtain the SVR model. The second stage specifically includes: obtaining the historical engineering quantity data and historical meteorological data of the filling area; training the initial SVR model through cross-validation based on the historical engineering quantity data and historical meteorological data to obtain the SVR model.

[0065] The historical engineering quantity data and historical meteorological data have been introduced above and will not be repeated here. The electronic device 100 can obtain the historical engineering quantity data and historical meteorological data of the filling area. The historical engineering quantity data and historical meteorological data of the filling area can be pre-stored in the electronic device 100, or stored outside the electronic device 100 and collected by the data acquisition device 110.

[0066] Regarding training the initial SVR model through cross-validation based on the historical engineering quantity data and historical meteorological data to obtain the SVR model, it can specifically include: splitting the historical engineering quantity data and historical meteorological data into a training set and a test set; using the training set to train the initial SVR model to obtain a trained SVR model; inputting the test set into the trained SVR model to obtain the historical model fitting residual; if the difference between the historical polynomial fitting residual and the historical model fitting residual meets the preset condition, the trained SVR model is used as the SVR model; if the difference between the historical polynomial fitting residual and the historical model fitting residual does not meet the preset condition, adjusting the parameters of the trained SVR model and continuing the training.

[0067] Cross-validation means that when training a model, most of the data in the same data set is selected as the training set to train the model, and the remaining small part of the data is used as the test set to test the results of the model training. The model parameters are continuously adjusted according to the test results to obtain a model that meets the preset conditions.

[0068] Regarding the SVR model, whether it is the initial SVR model, the SVR model obtained during the training process, or the SVR model obtained by the final training, their input data are the engineering quantity data and meteorological data of the filling area, and the output data are the model fitting residuals, that is, the polynomial fitting residuals obtained through model prediction. Unlike the historical polynomial fitting residuals obtained by fitting polynomials and historical data, the model fitting residuals are a predicted value, while the historical polynomial fitting residuals are the actual residuals. Through the SVR model, not only can the historical model fitting residuals (model fitting residuals of the past time) be obtained, but also the predicted model fitting residuals (model fitting residuals of the future time) can be obtained.

[0069] Regarding splitting the historical engineering quantity data and historical meteorological data into a training set and a test set, in order to make the data volume in the training set and the test set adapt to the subsequent model training, for example, the data in the training set can be set to 90% of the total of the historical engineering quantity data and the historical meteorological data, and the data in the test set can be set to 10% of the total of the historical engineering quantity data and the historical meteorological data.

[0070] Regarding using the training set to train the initial SVR model to obtain the trained SVR model; inputting the test set into the trained SVR model to obtain the historical model fitting residual. Before the model training begins, the data in the training set and the test set can be standardized. The standardized transformation can be performed before the historical engineering quantity data and the historical meteorological data are split, or after the historical engineering quantity data and the historical meteorological data are split. For example, the standardized transformation is performed before the historical engineering quantity data and the historical meteorological data are split. Taking the standardized transformation of the historical engineering quantity data as an example, the mean and standard deviation of all the historical engineering quantity data can be calculated, and each historical engineering quantity data can be standardized according to formula (3):

[0071] Formula (3)

[0072] In formula (3), X is the original data, μ is the mean of the data, σ is the standard deviation of the data, X ’ is the standardized data.

[0073] Each historical meteorological data can also be standardized in the same way as above, which will not be repeated here. After the standardized transformation is performed in the above way, the data in the training set and the test set are standardized, and then the model is trained using the standardized data.

[0074] If the difference between the historical polynomial fitting residual and the historical model fitting residual meets the preset condition, the trained SVR model is used as the SVR model; if the difference between the historical polynomial fitting residual and the historical model fitting residual does not meet the preset condition, the parameters of the trained SVR model are adjusted and the training is continued. Among them, the historical polynomial fitting residual is determined according to the historical settlement data and the settlement fitting result as described above, that is, the actual residual. The historical model fitting residual is the data output by the trained SVR model as described above, that is, the predicted residual. If the difference between the actual residual and the predicted residual meets the preset condition, for example, the preset condition can be that the actual residual minus the predicted residual is less than or equal to the first preset threshold, and for another example, the preset condition can also be that the RMSE between the actual residual and the preset residual is less than or equal to the second preset threshold, and the first preset threshold and the second preset threshold can both be set according to the empirical value. At this time, it means that the gap between the predicted residual and the actual residual is acceptable, so the training of the model can be stopped, and the trained SVR model can be used as the final SVR model. If the difference between the actual residual and the predicted residual does not meet the preset condition, for example, the preset condition may be that the actual residual minus the predicted residual is greater than the first preset threshold, and another example of the preset condition may be that the RMSE between the actual residual and the preset residual is greater than the second preset threshold. At this time, it means that the gap between the predicted residual and the actual residual is unacceptable. Therefore, it is still necessary to adjust the parameters of the trained SVR model, and then continue training until the difference between the actual residual and the predicted residual meets the preset condition, and then use the SVR model trained at this time as the final SVR model.

[0075] The parameters of the SVR model can be the type of kernel function, penalty parameter, gamma parameter, and epsilon parameter. The SVR model is trained using the training set, and the model parameters are gradually adjusted to achieve the best fit by minimizing the difference between the predicted residual and the actual residual. During the training process, the model is continuously iteratively optimized, and the optimal decision boundary is found by adjusting the parameters to minimize the residual prediction error. Then, an independent test set is used to verify the performance of the SVR model and evaluate the prediction accuracy and stability of the model. The effect of the model in residual prediction is evaluated by calculating the RMSE between the actual residual and the predicted residual.

[0076] Step 240: Obtain predicted settlement data of the filling area according to the settlement fitting polynomial and the prediction model fitting residual.

[0077] As mentioned above, settlement data is a data pair of time and settlement value, and the same is true for predicted settlement data. Therefore, settlement prediction for the filling area is actually a prediction of the settlement amount of the filling area in a certain period of time in the future. The specific time period to be predicted can be set manually before the prediction. For example, the settlement value of the filling area from January 1, 2025 to January 7, 2025 can be manually set. After determining the predicted time period, the predicted fitting settlement value can be obtained on the settlement fitting polynomial (substituting the predicted time into the settlement polynomial). According to historical experience, there will be a difference between the predicted fitting settlement value obtained by the settlement fitting polynomial and the settlement value caused by the actual settlement in the future. This difference is unavoidable for prediction by polynomial fitting, which brings a considerable deviation to the prediction result. Therefore, we can consider this difference in advance and add it in the prediction. Specifically, the predicted fitting settlement value can be added to the prediction model fitting residual to obtain the final predicted settlement value of the filling area.

[0078] The above embodiment combines polynomial fitting with support vector regression model, which can predict settlement problems in complex engineering environments. Polynomial fitting processes the overall trend of settlement data, and the SVR model compensates for the shortcomings of polynomial fitting in details, especially when dealing with complex engineering environments and changeable meteorological conditions. This method comprehensively considers engineering quantity data and meteorological data, and the model has high applicability and robustness.

[0079] Another embodiment of the present application is based on the land formation and deep foundation treatment project in a port area in East China. The loading height of the reclamation area is about 11 meters above the natural beach surface. About 80 settlement rods are deployed in the area. The settlement data of these settlement rods are summarized and averaged and regarded as the average settlement data of the reclamation area. The date range of the research data is from March 20, 2023 to May 31, 2024, of which the data from March 20, 2023 to April 30, 2024 is used as a training set, and the data from May 1 to 31, 2024 is used as a test set for prediction and verification of the accuracy of the model. The settlement data is monitored once a week, and the total settlement during the study period is about 2.1 meters. The linear interpolation method is used to interpolate to daily to ensure consistency with the time intervals of construction volume and meteorological data. The monitoring work uses the Sina Navigation T300 GNSS receiver, and its working mode is network RTK mode to ensure the accuracy and reliability of the data. The specific steps of this embodiment are as follows:

[0080] Step S1, data preparation: obtaining the historical settlement data, historical engineering quantity data (historical amount of blown fill sand for preloading and historical amount of bagged sand weir construction), predicted engineering quantity data (predicted amount of blown fill sand for preloading and predicted amount of bagged sand weir construction), historical weather data (historical weather, historical wind direction, historical wind force and historical temperature) and predicted meteorological data (predicted weather, predicted wind direction, predicted wind force and predicted temperature) of the filling area in this embodiment.

[0081] Step S2, data preprocessing: average the wind force and temperature respectively, encode the weather and wind direction respectively, and the preprocessing example results are shown in Tables 2 to 5 respectively.

[0082] Table 2 Temperature pretreatment example table

[0083] date Minimum and maximum temperatures Average temperature 2023 / 3 / 20 11-12℃ 11.5℃ 2023 / 3 / 21 12-15℃ 13.5℃ 2023 / 3 / 22 13-15℃ 14℃ 2023 / 3 / 23 10-13℃ 11.5℃

[0084] Table 3 Wind preprocessing example table

[0085] date Wind range Average wind speed 2023 / 3 / 20 Level 5-6 Level 5.5 2023 / 3 / 21 Level 6-7 6.5 2023 / 3 / 22 Level 5-6 Level 5.5 2023 / 3 / 23 Level 7-8 7.5

[0086] Table 4 Weather coding example table

[0087] date Raw weather data Previous Weather Late Weather Previous Weather Code Late Weather Coding 2023 / 3 / 20 Cloudy to foggy Negative fog 2 3 2023 / 3 / 21 Moderate rain turning to light rain moderate rain light rain 8 4 2023 / 3 / 22 light rain light rain light rain 4 4 2023 / 3 / 23 Cloudy to light rain Negative light rain 2 4

[0088] Table 5 Wind force coding example table

[0089] date wind direction Early wind direction Later wind direction Early wind direction angle Later wind direction angle Early wind direction sine Previous wind direction cosine Later wind direction sine Late wind direction cosine 2023 / 3 / 20 Southeast to South Wind southeast South 135 180 0.707 -0.707 0 -1 2023 / 3 / 21 North Wind north north 0 0 0 1 0 1 2023 / 3 / 22 Northwest to north wind northwest north 315 0 -0.707 0.707 0 1 2023 / 3 / 23 East to northeast wind East northeast 90 45 1 0 0.707 0.707

[0090] Step S3, polynomial fitting of historical settlement data: When determining the number of polynomial fittings, it is necessary to strike a balance between fitting accuracy and overfitting risk. Table 6 lists the accuracy of polynomial fittings of different orders on test data:

[0091] Table 6 Polynomial fitting times and accuracy

[0092] Polynomial degree 1 2 3 4 5 6 RMSE / mm 150.7 49.7 19.2 24.9 24.7 46.4

[0093] As can be seen from Table 6, as the number of polynomials increases, RMSE first decreases rapidly, especially when the number increases from 1 to 3, RMSE decreases significantly, indicating that higher-order polynomials can better capture the trend of changes in settlement. However, when the number of polynomials exceeds 3, the RMSE changes no longer significantly, and even at 6, RMSE increases, indicating that the model may be overfitting. Therefore, a model with a polynomial order of 3 is selected, at which time RMSE reaches the minimum value (19.2 mm), and there is no obvious sign of overfitting compared to higher-order polynomials. The polynomial fitting results are shown in Figure 6. Figure 4 .

[0094] Depend on Figure 4It can be seen that the polynomial fitting curve can track the changing trend of the original settlement value well. In the training data from March 20, 2023 to April 30, 2024, the average residual is only 0.04mm, indicating that the fitting effect is good. However, at certain time points, there are still large residuals. On September 18, 2023, the residual reached a maximum value of 51.3mm. Combined with the engineering volume, the amount of sand blowing on that day was about 28733m³. Large-scale blowing and filling activities will significantly affect the settlement speed and amplitude of the foundation, especially when a large amount of blowing and filling is carried out in a short period of time, which is more likely to cause drastic fluctuations in the settlement value. Polynomial fitting may be insufficient in dealing with such sudden changes, resulting in large residuals.

[0095] Figure 4 The middle right figure shows the fitting results on the test data set in May 2024. It can be seen that in the test data stage, the polynomial fitting curve still follows the trend of the actual settlement value well. This shows that the polynomial fitting model has a strong generalization ability. During the entire time period of the test data set, the average value of the residual is 19.2mm, which shows that the polynomial fitting model can provide relatively accurate settlement predictions in most cases.

[0096] Step S4, calculate the polynomial fitting residual: calculate the residual between the actual daily settlement value during the study period and the settlement value fitted by the third-order polynomial in S3.

[0097] Step S5, support vector regression prediction residual: According to the residual obtained in step S4, an initial SVR model is constructed, and the initial SVR model is trained to obtain an SVR model that can be used to predict the residual. The input data of the SVR model include the predicted engineering quantity data (the predicted amount of blown sand for preloading and the predicted amount of bagged sand weir construction), the pre-processed predicted meteorological data (the predicted average wind force, the predicted average temperature, the predicted weather coding data, the predicted wind direction coding data), and the output data is the predicted residual data for the future time.

[0098] Step S5.1: Standardize the input data respectively.

[0099] Step S5.2, support vector regression model training: After cross-validation, the model parameters are set as follows: the kernel function selects the radial basis function; the penalty parameter is set to 1000; the gamma value is set to 0.05; and the epsilon value is set to 0.1. The output results of the SVR model on the test set are as follows Figure 5 shown.

[0100] Step S5.3, model evaluation: The prediction results of the SVR model on the test set show that the RMSE between the original residual and the SVR model output is 3.4 mm, indicating that the SVR model has a high residual prediction accuracy for the test data. In addition, the change trend of the original residual and the SVR output results is consistent, and the correlation coefficient reaches 0.85, which further proves the effective modeling ability of the SVR model for residual values.

[0101] Step S6, comprehensive prediction: In this embodiment, the final settlement prediction result, that is, the predicted settlement data, is obtained by combining the polynomial fitting result with the residual prediction result of the SVR model, such as Figure 6 shown.

[0102] The comprehensive prediction results show that the average difference between the predicted value and the actual measured settlement value is 3.1 mm, which indicates that the comprehensive method has shown high accuracy in settlement prediction. According to the law of error propagation, since the final predicted value is the linear superposition of the polynomial fitting result and the SVR prediction result, the accuracy of the final predicted value should be the square root of the sum of the squares of the errors of the two, which is 24 mm. In addition, the correlation coefficient between the comprehensive prediction results and the measured settlement values ​​is 0.99, which further indicates that this method can accurately reflect the changes in settlement and proves the effectiveness and reliability of the method combining polynomial fitting with the SVR model.

[0103] In several embodiments provided in the present application, the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and a part of a module, a program segment or a code contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of boxes in the block diagram and / or the flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0104] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0105] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

Claims

1. A method for predicting settlement of a filling area, characterized in that: The following steps are involved: Obtaining historical settlement data, predicted engineering quantity data and predicted meteorological data of the filling area; wherein the predicted engineering quantity data refers to the predicted engineering quantity at a future time, and the predicted meteorological data refers to the predicted meteorological data at a future time; Performing polynomial fitting on the historical settlement data to obtain a settlement fitting polynomial; Determine the historical settlement fitting result according to the historical settlement data and the settlement fitting polynomial; determine the historical polynomial fitting residual according to the historical settlement data and the historical settlement fitting result; and construct an initial SVR model using support vector regression SVR according to the historical polynomial fitting residual; Acquire historical engineering quantity data and historical meteorological data of the filling area; train the initial SVR model through cross-validation based on the historical engineering quantity data and historical meteorological data to obtain an SVR model; input the predicted engineering quantity data and predicted meteorological data into the SVR model to obtain the prediction model fitting residual; Obtaining predicted settlement data of the filling area according to the settlement fitting polynomial and the prediction model fitting residual; The method of training the initial SVR model through cross-validation based on the historical engineering quantity data and historical meteorological data to obtain the SVR model includes: splitting the historical engineering quantity data and historical meteorological data into a training set and a test set; using the training set to train the initial SVR model to obtain a trained SVR model; inputting the test set into the trained SVR model to obtain the historical model fitting residual; if the difference between the historical polynomial fitting residual and the historical model fitting residual meets a preset condition, using the trained SVR model as the SVR model.

2. The method for predicting settlement of a filling area according to claim 1, characterized in that: The performing polynomial fitting on the historical settlement data to obtain a settlement fitting polynomial includes: Performing polynomial fitting on the historical settlement data using the least square method to obtain an n-order fitting polynomial; According to the root mean square deviation of polynomials of different degrees on the test data set, the degree of the n-order polynomial is determined and used as the sedimentation fitting polynomial.

3. The method for predicting settlement of a filling area according to claim 1, characterized in that: If the difference between the historical polynomial fitting residual and the historical model fitting residual does not meet the preset condition, the parameters of the trained SVR model are adjusted and the training is continued.

4. The method for predicting settlement of a filling area according to claim 1, characterized in that: The inputting the predicted engineering quantity data and the predicted meteorological data into the SVR model comprises: The predicted engineering quantity data and predicted meteorological data are standardized and converted, and the standardized and converted data are input into the SVR model.

5. The method for predicting settlement of a filling area according to claim 1, characterized in that: The engineering quantity data includes: the amount of blown fill sand for preloading and the amount of bagged sand weir construction; Meteorological data include: weather data, wind direction data, wind speed data and temperature data; The engineering quantity data are historical engineering quantity data or predicted engineering quantity data, and the meteorological data are historical meteorological data or predicted meteorological data.

6. The method for predicting settlement of a filling area according to claim 5, characterized in that: The method further comprises: preprocessing the original meteorology to obtain meteorological data; The preprocessing of the original meteorological data to obtain meteorological data includes: Segment and encode weather types to obtain weather data; The wind direction types are segmented, angle mapped and ring-coded to obtain wind direction data; The daily average wind speed is used as the wind speed data; The daily average temperature is used as the temperature data.

7. An electronic device, characterized in that: The electronic device comprises: processor; a memory for storing processor-executable instructions; The processor is configured to execute the method for predicting settlement of a filling area as described in any one of claims 1 to 6.

Images