A system and method for predicting foundation settlement risk of drainage gate

Abstract

The present invention provides a system and method for predicting the risk of foundation settlement of a drainage gate, which relates to the technical field of risk prediction. The system includes: an acquisition module for acquiring time series data; a first judgment module for judging whether the time series data is stable through the Hurt index; a difference module for determining the difference order and performing difference on the time series data to stabilize the time series data; a construction module for constructing an autoregressive integral moving average model in combination with extreme value theory; a second judgment module for judging whether the model residual of the autoregressive integral moving average model conforms to the random distribution; an adjustment module for adjusting the parameters of the autoregressive integral moving average model; an output module for outputting the predicted drainage gate foundation height within a preset time length using the autoregressive integral moving average model; and an early warning module for issuing a risk early warning when there is data lower than the preset drainage gate foundation height. The early warning of foundation settlement risk can be completed automatically with higher accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of risk prediction, and in particular to a system and method for predicting the risk of foundation settlement of a drainage gate. Background Art

[0002] A sluice gate is a facility used to control water flow, drain water or prevent flood backflow, usually located at the entrance or exit of water bodies such as rivers, lakes, and drainage channels. The sluice gate adjusts the water level by opening and closing the gate to ensure the normal passage or discharge of water and prevent the water level from overflowing or drying up due to excessive water level. The foundation settlement of a sluice gate refers to the phenomenon that the soil or bedrock where the sluice gate is located sinks under long-term load. Foundation settlement can be caused by a variety of reasons, such as soil compaction, changes in groundwater levels, differences in soil types, etc. Settlement can cause structural instability of the sluice gate and may even cause damage to the drainage function or the risk of flooding.

[0003] As an important part of water conservancy projects, the stability of drainage gates directly affects the water conservancy safety of surrounding areas. If the foundation settlement cannot be predicted and handled in time, it may lead to structural instability or functional damage of the drainage gate, causing serious flood risks. By predicting the risk of drainage gate foundation settlement, potential foundation problems can be identified in advance, and necessary repair or reinforcement measures can be taken, thereby effectively preventing safety accidents caused by settlement and ensuring the safety of life and property and the stability of the surrounding environment.

[0004] However, most of the existing drainage gate risk predictions are based on subjective judgment of foundation settlement values, which cannot accurately evaluate the foundation settlement process affected by multiple factors. This may lead to the problem that the foundation settlement exceeds the safety threshold and cannot be predicted in advance. The foundation settlement risk prediction has a large deviation, which is not conducive to the safe operation and maintenance of the drainage gate. Summary of the invention

[0005] In order to solve the technical problems that the risk prediction of drainage gates in the prior art is mostly based on subjective judgment of foundation settlement values, which cannot accurately evaluate the foundation settlement process affected by various factors, may cause the foundation settlement to exceed the safety threshold and cannot be predicted in advance, and the foundation settlement risk prediction has a large deviation, which is not conducive to the safe operation and maintenance of the drainage gate, the present invention provides a foundation settlement risk prediction system and method for a drainage gate.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] First aspect

[0008] An embodiment of the present invention provides a foundation settlement risk prediction system for a drainage gate, comprising:

[0009] An acquisition module, used to acquire time series data describing the historical foundation height of the drainage gate at different times;

[0010] The first judgment module is used to combine the adaptive time window describing the local volatility of the time series data and judge whether the time series data is stable through the Hurt index. If so, the construction module is called, otherwise, the difference module is called;

[0011] The difference module is used to determine the difference order in combination with the Hurt index, and to differentiate the time series data based on the difference order to stabilize the time series data;

[0012] A construction module is used to determine the parameters of the autoregressive integrated moving average model based on the stationary time series data in combination with the extreme value theory, and to construct the autoregressive integrated moving average model, wherein the parameters of the autoregressive integrated moving average model include the autoregressive order and the moving average order;

[0013] The second judgment module is used to judge whether the model residual of the autoregressive integrated moving average model conforms to the random distribution. If so, the output module is called, otherwise, the adjustment module is called;

[0014] An adjustment module, used for adjusting the parameters of the autoregressive integrated moving average model by a gradient descent method, and calling the second judgment module until the model residual of the autoregressive integrated moving average model conforms to a random distribution;

[0015] An output module, used for outputting the predicted drainage gate foundation height within a preset time period by using an autoregressive integrated moving average model;

[0016] The early warning module is used to issue a risk early warning when there is a predicted drainage gate foundation height that is lower than a preset drainage gate foundation height.

[0017] Second aspect

[0018] An embodiment of the present invention provides a method for predicting foundation settlement risk of a drainage gate, comprising:

[0019] S1: Obtain time series data describing the historical foundation height of the drainage gate at different times;

[0020] S2: Combined with the adaptive time window that describes the local volatility of the time series data, the Hurt index is used to determine whether the time series data is stable. If so, proceed to step S4; otherwise, proceed to step S3;

[0021] S3: Determine the difference order in combination with the Hurt index, and perform difference on the time series data based on the difference order to stabilize the time series data;

[0022] S4: Based on the stable time series data, the autoregressive integrated moving average model parameters are determined in combination with the extreme value theory, and the autoregressive integrated moving average model is constructed, wherein the autoregressive integrated moving average model parameters include the autoregressive order and the moving average order;

[0023] S5: Determine whether the model residual of the autoregressive integrated moving average model conforms to the random distribution. If so, proceed to step S7; otherwise, proceed to step S6;

[0024] S6: adjusting the parameters of the autoregressive integrated moving average model by the gradient descent method, and judging whether the model residual of the autoregressive integrated moving average model conforms to the random distribution, until the model residual of the autoregressive integrated moving average model conforms to the random distribution;

[0025] S7: using the autoregressive integrated moving average model to output the predicted drainage gate foundation height within a preset time period;

[0026] S8: Issue a risk warning when there is a predicted drain gate foundation height that exceeds the preset drain gate foundation height.

[0027] The third aspect

[0028] An embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for predicting foundation settlement risk of a drainage gate as described in the first aspect is implemented.

[0029] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0030] In the present invention, based on the acquired time series data describing the historical foundation height of the drainage gate at different times, combined with the adaptive time window describing the local volatility of the time series data, the Hurt index is used to first determine whether the time series data is stable, and the non-stationary data is differentiated to stabilize the series data. The adaptive time window can better capture the local volatility of the time series data, improve the adaptability and prediction accuracy of the model, and the stationarity of the time series is determined by the Hurst index, and the non-stationary data is differentiated to ensure the stationarity of the model input data, thereby improving the stability of the model and the prediction accuracy. After that, the autoregressive integrated moving average model parameters are determined in combination with the extreme value theory, and the autoregressive integrated moving average model is constructed, so that the model can more accurately capture extreme values ​​and long-term trends, thereby significantly improving the accuracy and reliability of the prediction, and ensuring that the prediction results are more stable and reliable. For the constructed model, determine whether its model residual conforms to the random distribution. If not, adjust the model parameters in combination with the gradient descent method until its model residual conforms to the random distribution. By checking whether the model residual conforms to the random distribution, it can be determined whether the model fully fits the data. If the residual does not conform to the random distribution, it means that the model may have systematic deviations or fail to capture certain laws. By adjusting the model parameters in combination with the gradient descent method, the model can be automatically optimized so that the residual gradually approaches the random distribution. The gradient descent method gradually reduces the prediction error by iteratively adjusting the parameters, avoiding the tediousness and subjectivity of manual parameter adjustment, and ensuring that the model adaptively improves the prediction accuracy during the training process. Finally, the model whose model residual conforms to the random distribution is used to predict the foundation height of the drainage gate for risk warning. By combining the adaptive time window, Hurst index, extreme value theory and gradient descent method, the volatility of time series data can be more accurately processed, data stability can be ensured, and model parameters can be automatically optimized, which improves the adaptability, accuracy and stability of the model, reduces manual intervention, ensures that the prediction results are more reliable, and effectively improves the accuracy and reliability of the drainage gate foundation settlement risk warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 A schematic diagram of the structure of a foundation settlement risk prediction system for a drainage gate provided by an embodiment of the present invention;

[0033] Figure 2 A schematic flow chart of a method for predicting foundation settlement risk of a drainage gate provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0035] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0036] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0037] Reference Manual Attached Figure 1 , showing a schematic structural diagram of a foundation settlement risk prediction system for a drainage gate provided in an embodiment of the present invention.

[0038] An embodiment of the present invention provides a foundation settlement risk prediction system for a drainage gate, comprising:

[0039] The acquisition module 1 is used to obtain time series data describing the historical foundation height of the drainage gate at different times.

[0040] It should be noted that the time series data describing the historical foundation height of the drainage gate at different times are obtained to provide basic data for subsequent analysis. These data reflect the changing trend of the foundation height of the drainage gate at different time points, which is the core input for foundation settlement risk prediction and helps to reveal the laws and changing trends of foundation settlement.

[0041] The first judgment module 2 is used to combine the adaptive time window describing the local volatility of the time series data and judge whether the time series data is stable through the Hurt index. If so, the construction module is called, otherwise, the difference module is called.

[0042] Among them, the local volatility of data refers to the degree of volatility of the time series in a specific time period, which reflects the drastic degree of data change in this time period. Through local volatility, the short-term change trend of data can be captured. Adaptive time window is a method of dynamically adjusting the size of the analysis window. Its core idea is to adjust the size of the time window according to the local volatility of the data, so as to better capture the long-term dependency characteristics of different time periods, so as to make the analysis results more accurate and flexible. The core idea of ​​adaptive time window is to dynamically adjust the analysis window size of the time series so that the Hurst index can adaptively capture the long-term dependency characteristics of different time periods according to the local volatility of the data. This can solve the problem of neglecting the volatility change caused by the use of a fixed time window in the traditional Hurst index method. The Hurst index is an indicator to measure the long-term dependency of the time series. Its value is between 0 and 1. If the Hurst index is close to 0.5, it means that the sequence is random. If it is greater than 0.5, it means that the time series is persistent and tends to be stable. If it is less than 0.5, it means that the sequence is anti-persistent.

[0043] It should be noted that, combined with the adaptive time window that describes the local volatility of the time series, the Hurst index can better capture the long-term dependency characteristics of the time series by dynamically adjusting the window size. If the data is judged to be in a stable state by the Hurst index, it will directly proceed to the follow-up. If the data is in a non-stationary state, differential processing will be performed. This method can solve the problem of neglecting volatility changes caused by fixed windows in traditional methods and improve the accuracy and adaptability of analysis.

[0044] In a possible implementation, the first judgment module combines the adaptive time window describing the local volatility of the time series data and judges whether the time series data is stable by using the Hurt index, specifically including:

[0045] Determine the adaptive time window based on the local volatility of the time series data:

[0046]

[0047] in, Indicates time t The time window size, represents an adjustable constant that controls the relationship between window size and local volatility, Indicates the time series data in the current time window The standard deviation of the time series data at time t The local volatility of Represents the first i Observations, Indicates the time series data in the current time window The local mean within Represented on time series data The average value of .

[0048] It should be noted that the advantage of determining the adaptive time window based on the local volatility of time series data is that it can dynamically adjust the size of the analysis window, so as to more accurately capture the changing trend of data in different time periods, avoid the problem that the fixed window method cannot adapt to changes in volatility, improve the ability to capture the long-term dependency characteristics of data, and enhance the accuracy and flexibility of the analysis results.

[0049] Calculate the ratio of the maximum cumulative deviation to the standard deviation within each adaptive time window:

[0050] ;

[0051] in, Indicates that time series data is t The cumulative deviation of time, R Indicates the maximum cumulative deviation within each adaptive time window Minimum cumulative deviation difference, S represents the standard deviation within the adaptive time window, i.e. .

[0052] The Hurt index is obtained by performing logarithmic fitting on the ratios in each adaptive time window:

[0053] ;

[0054] in, H represents the Hurt index, log represents the logarithmic function, C represents the intercept constant.

[0055] When the Hurt index value is 0.5, the output time series data is in a stable state, otherwise, the output time series data is in a non-stationary state.

[0056] Specifically, first, the size of the adaptive time window is determined according to the local volatility of the time series. The relationship between the window size and the local volatility is controlled by an adjustable constant, and the window size is optimized by minimizing the root mean square error so that the window can optimally fit the local volatility of the data. Next, the ratio of the maximum cumulative deviation to the standard deviation in each adaptive time window is calculated to reflect the degree of volatility of the data in the window. Then, the Hurst index is obtained by logarithmic fitting of the comparison values, which reflects the long-term dependence of the time series. Finally, the data is judged whether it is stable according to the value of the Hurst index. If the Hurst index is close to 0.5, the data is considered to be stable, otherwise it is non-stationary. Through this process, the stationarity of the time series can be effectively evaluated, and then it can be determined whether the data needs to be differentially processed.

[0057] The difference module 3 is used to determine the difference order in combination with the Hurt index, and to differentiate the time series data based on the difference order to stabilize the time series data.

[0058] Among them, the difference order refers to the number of times the time series data is differentiated. In time series analysis, differentiation is a process used to remove trend and seasonal fluctuations in the series, thereby making the series stable. The higher the difference order, the more times the difference is performed. Usually, the difference order is selected based on the Hurst index and the non-stationary characteristics of the series, with the aim of making the time series stable and facilitating subsequent modeling and prediction. Specifically, the difference order of the time series is first determined based on the Hurst index. The Hurst index provides information about the stationarity of the series, based on which it can be determined how many times the difference is needed to make the series stable. Once the difference order is determined, the original time series is differentiated, that is, the value at the current moment is subtracted from the value at the previous moment, and the operation is repeated until the data reaches a stable state. This process helps to remove trend changes in the time series, making the data more consistent with the assumption of a stationary series, thereby providing effective data support for subsequent modeling and prediction.

[0059] In a possible implementation manner, the formula for determining the difference order is specifically:

[0060] ;

[0061] in, d represents the difference order.

[0062] It is understandable that the difference order is adapted to the long-term dependency characteristics of the time series, which helps to stabilize the series and optimize the subsequent modeling and forecasting process.

[0063] Construction module 4 is used to determine the autoregressive integrated moving average model parameters based on the stationary time series data in combination with the extreme value theory, and to construct the autoregressive integrated moving average model, wherein the autoregressive integrated moving average model parameters include the autoregressive order and the moving average order.

[0064] Among them, extreme value theory is a statistical method used to study the distribution characteristics of extreme events that occur in certain random processes, especially the frequency and intensity of extreme values. In time series analysis, extreme value theory is often used to identify and model possible extreme fluctuations or outliers in data, so as to more accurately describe the tail behavior of the data. The autoregressive integrated moving average model (ARIMA) is a statistical model widely used for time series forecasting. The autoregressive order represents the relationship between the current value and its past p-time values, which determines how many historical observations need to be considered in the AR model. The moving average order represents the relationship between the current value and its past q-time error terms, which determines how many historical error terms need to be considered in the MA model.

[0065] It should be noted that extreme value theory can automatically identify extreme fluctuations in the data, and then more accurately determine the autoregressive order and moving average order. This method avoids the inefficiency and inaccuracy of traditional subjective selection, improves the accuracy and reliability of the model, and ensures the effectiveness of the prediction of the drainage gate foundation settlement.

[0066] In a possible implementation manner, the building block 4 is specifically used to:

[0067] Construct autocorrelation and partial autocorrelation functions for stationary time series data.

[0068] The function coefficients of the autocorrelation function and the partial autocorrelation function under different autoregressive integrated moving average model parameters are calculated, wherein the function coefficients include the autocorrelation coefficient and the partial autocorrelation coefficient.

[0069] Determine confidence intervals for function coefficients based on quantiles of the standard normal distribution:

[0070] ;

[0071] in, CI represents the confidence interval, Indicates the significance level corresponding to α The standard normal distribution quantile of n The sample size in the stationary time series data is the total number of observations in the stationary time series data.

[0072] Optionally, the significance level may be specifically set to 0.05.

[0073] The autoregressive integrated moving average model parameters corresponding to the function coefficients that exceed the confidence interval are determined as the extreme values ​​of the corresponding function.

[0074] Fit the extreme values ​​belonging to the same function through the generalized Pareto distribution to obtain the fitted values:

[0075] ;

[0076] in, x represents an extreme value, represents the function coefficient at the minimum extreme value, represents the width of the extreme value distribution, represents the shape parameter describing the distribution of extreme values, exp represents the exponential function, Indicates about The cumulative distribution function value of express The cumulative distribution function value when , log represents the logarithmic function, Indicates about The likelihood function of Indicates i extreme values, , m represents the total number of extreme values, Indicates that The smallest hour and as the fitted value.

[0077] The shape parameter determines the shape of the tail of the extreme value distribution. If the distribution has a heavy tail, the probability of an extreme value occurring is higher. If the distribution degenerates into an exponential distribution, that is, the tail is lighter, If the probability of extreme values ​​occurring is low, the distribution has a light tail. .in addition, The function coefficient representing the minimum extreme value refers to The function coefficients are the autocorrelation function and the partial autocorrelation function.

[0078] Compute the return period of each extreme value relative to the fitted value:

[0079] ;

[0080] in, Represents extreme values x The return cycle, Represents extreme values x The cumulative distribution function value of .

[0081] The extreme value corresponding to the maximum return period is taken as the optimal autoregressive integrated moving average model parameter.

[0082] Substitute the autoregressive integrated moving average model parameters into the autoregressive integrated moving average model to complete the construction of the autoregressive integrated moving average model:

[0083] ;

[0084] in, express t The observed value at the moment, c represents a constant term, Indicates i The autoregressive coefficients, , p represents the maximum autoregressive order, Indicates j The moving average coefficient, , q represents the maximum moving average order, Representation and t and j The associated lagged residuals, express t Time residual.

[0085] Specifically, the construction module describes in detail how to determine the optimal parameters of the autoregressive integrated moving average (ARIMA) model by combining the autocorrelation function and the partial autocorrelation function with the extreme value theory. First, the autocorrelation coefficient and partial autocorrelation coefficient of the stationary time series data are calculated to understand the intrinsic structure of the data. The confidence interval of the function coefficient is determined by the standard normal distribution, and the extreme values ​​with significant deviations are excluded. These extreme values ​​are fitted by the generalized Pareto distribution, and their return period is calculated. Finally, the extreme value with the largest return period is selected as the optimal model parameter. Finally, the autoregressive integrated moving average model is constructed by the determined parameters to ensure that the model can accurately reflect the changing trend of the data and optimize the prediction accuracy.

[0086] It should be noted that the precise statistical methods and extreme value theory are used to select the parameters of the autoregressive integrated moving average (ARIMA) model more efficiently and objectively. First, the intrinsic dependence and volatility characteristics of the data are captured by calculating the autocorrelation function and partial autocorrelation function of the time series data. Then, by analyzing the coefficients of these functions and determining their confidence intervals using the standard normal distribution, significant autoregressive coefficients and moving average coefficients are identified. Furthermore, combined with extreme value theory, extreme values ​​are identified and fitted, which reflect abnormal fluctuations in the data and are crucial for parameter selection. Finally, the optimal model parameters are determined by the return period analysis of extreme values. This method avoids the subjectivity and inefficiency of the information criterion in the traditional model parameter selection, and provides an automated, precise and effective solution to the volatility characteristics of time series data. This automatic selection process based on statistical laws can improve the model's fit and prediction accuracy, ensuring that the model has higher stability and reliability in complex practical applications.

[0087] The second judgment module 5 is used to judge whether the model residual of the autoregressive integrated moving average model conforms to the random distribution. If so, the output module is called, otherwise, the adjustment module is called.

[0088] Model residuals refer to the difference between actual observations and model predictions. In time series analysis, residuals reflect what the model cannot explain. Ideally, the residuals should have no obvious patterns or trends, indicating that the model has fitted the data well. Random distribution means that there is no regularity in the distribution of data, and each data point is independent of each other. Ideally, model residuals should follow a random distribution, that is, there is no systematic deviation or trend. If the model residuals show obvious structural patterns or cyclical fluctuations, it means that the model may not be adequately fitted or has not captured certain patterns.

[0089] The fitting effect of the model is determined by checking whether the residuals of the autoregressive integrated moving average model conform to the random distribution. If the residuals show a random distribution, it means that the model has fully captured the law of the data and the prediction effect is good, and the subsequent prediction steps can be entered. If the residuals do not conform to the random distribution, it means that the model may be biased or fail to fully fit the data and needs to be adjusted. Through this check, the reliability and prediction accuracy of the model are ensured.

[0090] In a possible implementation manner, the second judgment module 5 determines whether the model residual of the autoregressive integrated moving average model conforms to the random distribution, specifically including:

[0091] Compute the model residuals for an autoregressive integrated moving average model:

[0092] ;

[0093] in, express t The model residual at time t, They represent the actual observed value at time t and the model predicted value at time t respectively.

[0094] Calculate the model residual correlation coefficient:

[0095] ;

[0096] in, k represents the lag order, express of k The residual correlation coefficient of the order model is express The mean of The lag order is t - k The model residual at time t, n Represents the total number of model residuals.

[0097] Calculate the Q test statistic based on the model residual correlation coefficient:

[0098] ;

[0099] in, represents the Q test statistic.

[0100] Among them, the Q test statistic (usually called the Ljung-Box Q test statistic) is a method for testing whether there is autocorrelation in the residuals of a time series model. In time series modeling, the residuals should be random, have no regularity, and should obey white noise, that is, the residuals at each time point should be independent of each other and have the same distribution.

[0101] Calculate the probability of the Q test statistic under the prior assumption that the model residuals conform to a random distribution:

[0102] ;

[0103] in, The degrees of freedom are df The chi-square distribution of represents the probability value of the Q test statistic, Indicates that the degrees of freedom are df The probability value of the Q test statistic under the chi-square distribution.

[0104] When the occurrence probability value is greater than the preset occurrence probability value, the model residuals of the output autoregressive integrated moving average model conform to the random distribution, otherwise, the model residuals of the output autoregressive integrated moving average model do not conform to the random distribution.

[0105] Optionally, the preset occurrence probability value may be set to 0.05.

[0106] In the actual application process, first, the prediction error of the model is obtained by calculating the residuals. Then, the correlation between the residuals is analyzed to detect whether there are systematic deviations. Based on these residual correlation coefficients, the Q test statistic is used to evaluate the independence of the residuals, and then determine whether the model adequately fits the data. The probability value of the Q test can quantify whether the residuals conform to the random distribution, avoid human intervention, and objectively judge the effectiveness of the model. If the model residuals conform to the random distribution, it means that the model has fully captured the data characteristics and the prediction is reliable. If not, it is promptly prompted that the model may be biased and needs further optimization. This process ensures that there is no systematic regularity in the model residuals, improves the accuracy and reliability of the prediction, and reduces the errors caused by human selection or parameter adjustment.

[0107] The adjustment module 6 is used to adjust the parameters of the autoregressive integrated moving average model by the gradient descent method, and call the second judgment module until the model residual of the autoregressive integrated moving average model conforms to the random distribution.

[0108] Among them, the gradient descent method is an optimization algorithm used to minimize or maximize the objective function (for example, the loss function). It calculates the gradient (i.e., partial derivative) of the objective function with respect to the parameters, and then adjusts the parameters in the negative direction of the gradient to reduce the prediction error. In time series modeling, the gradient descent method can be used to adjust the parameters of the model to minimize the prediction error (residual) of the model. It is understandable that if the model residual does not conform to the random distribution, the gradient descent method is used to adjust the parameters of the autoregressive integrated moving average model. By iteratively adjusting the model parameters, the gradient descent method continuously optimizes the prediction effect until the residual conforms to the random distribution, indicating that the model fits sufficiently and can accurately reflect the data characteristics. This process improves the accuracy and reliability of the prediction by reducing the model bias.

[0109] In a possible implementation manner, the adjustment module 6 is specifically configured to:

[0110] Calculate the prediction loss value of the autoregressive integrated moving average model under the autoregressive integrated moving average model parameters:

[0111] ;

[0112] in, represents the autoregressive integrated moving average model parameters The predicted loss value under and Respectively t The actual observed value at the moment and t The model prediction value at that moment, L Indicates the length of time series data.

[0113] Calculate the gradient of the predicted loss value with respect to each autoregressive integrated moving average model parameter:

[0114] ;

[0115] in, represents partial derivative, Represents the prediction loss value Relative to the autoregressive integrated moving average model parameters gradient.

[0116] Based on the obtained gradient update, the corresponding autoregressive integrated moving average model parameters are:

[0117] ;

[0118] in, represents the updated autoregressive integrated moving average model parameters, Indicates the update step size.

[0119] It should be noted that by calculating the prediction loss value, the current prediction error of the model can be quantified to ensure the degree of fit of the model to the actual data. Next, the gradient of the loss function is solved so that the model can gradually adjust the parameters according to the direction of error change and optimize in the direction of minimizing the prediction error. Finally, the model parameters are updated based on the gradient and adjusted using an appropriate step size to ensure that each update can effectively improve the model performance. This process automatically optimizes the model, avoiding the tediousness and subjectivity of manual adjustments, while ensuring that the model gradually converges to the optimal prediction state, improving the stability and reliability of the prediction.

[0120] The output module 7 is used to output the predicted drainage gate foundation height within a preset time period using the autoregressive integrated moving average model.

[0121] It should be noted that those skilled in the art can set the preset duration according to actual needs, and the present invention is not limited thereto.

[0122] It is understandable that the constructed autoregressive integrated moving average model can be used to predict the drainage gate foundation height within a preset time period. As time goes by, the predicted foundation height will gradually stabilize, indicating that the foundation settlement rate slows down and eventually stops, maintaining a constant value. This process helps predict possible future settlement trends and provide a basis for risk warning.

[0123] In a possible implementation, the calculation method for predicting the height of the drainage gate foundation is specifically as follows:

[0124] ;

[0125] in, and Respectively represent the T Time lag h Step and h - i Step 1: Predict the drainage gate foundation height. Represents relative to T Time lag h - i The residual prediction value after step , It represents the value calculated based on the historical foundation height of the drainage gate at time T To time T + h The predicted mean within Indicates i The autoregressive coefficients, , p represents the maximum autoregressive order, Indicates j The moving average coefficient, , q Indicates the maximum moving average order.

[0126] It should be noted that by combining the predicted drainage gate foundation height with the residual prediction value of the lag step number, the dynamic change trend of foundation settlement can be more accurately reflected. By comprehensively predicting the foundation height and residuals with different time lags, the short-term fluctuations and long-term trends of foundation settlement can be effectively captured, the deviation of a single factor can be reduced, and the reliability and accuracy of the prediction results can be improved. In addition, the mean calculation based on the historical foundation height further improves the stability and robustness of the model, ensuring more accurate risk warnings.

[0127] The early warning module 8 is used to issue a risk early warning when there is a predicted drainage gate foundation height that is lower than a preset drainage gate foundation height.

[0128] It should be noted that those skilled in the art can set the size of the preset drainage gate foundation height according to actual needs, and the present invention is not limited here.

[0129] In a possible implementation manner, the early warning module 8 is specifically used for:

[0130] When there is a predicted drainage gate foundation height that is lower than the preset drainage gate foundation height, a risk warning is issued through a warning light, a speaker or information push.

[0131] In a possible implementation, it further includes:

[0132] The conversion module is used to convert the predicted drainage gate foundation height into the predicted drainage gate foundation height interval output:

[0133] ;

[0134] in, PI It indicates the predicted drainage gate foundation height interval. Represents the difference between the historical foundation height of the drainage gate and h The variance of the forecast errors.

[0135] Issue early warning for predicted drainage gate foundation height intervals that are lower than the preset drainage gate foundation height.

[0136] It is understandable that with the lag step length h That is, as the prediction step increases, the standard deviation of the prediction error will also increase, because future predictions are usually more uncertain than recent predictions. By converting the predicted drainage gate foundation height into a height interval and introducing the prediction error variance, the uncertainty of the prediction results can be more fully displayed. As the lag step increases, the standard deviation of the prediction error will also increase, reflecting the high uncertainty of future predictions. In this way, not only a specific prediction value can be provided, but also an interval range can be given to help more accurately assess potential risks, ensure that risk warnings are issued in a timely manner when the height is lower than the preset height, and improve the robustness of the prediction and the effectiveness of risk management.

[0137] In the actual application process, first, by obtaining historical foundation height data at different times, basic data is provided for subsequent analysis. Then, the stationarity of the data is judged by adaptive time window and Hurst index. If it is not stationary, differential processing is performed to stabilize the sequence. Then, the parameters of the autoregressive integral moving average model are automatically selected in combination with extreme value theory to improve the prediction accuracy. In this process, the model residuals need to conform to the random distribution. If not, the gradient descent method is used to adjust the model parameters until the residuals conform to the random distribution. After that, the optimized model is used to predict the foundation height of the drainage gate, and finally a risk warning is issued based on the prediction results. This method improves the accuracy and reliability of the prediction through precise model selection and optimization, ensuring timely warning of foundation settlement risks.

[0138] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0139] In the present invention, based on the acquired time series data describing the historical foundation height of the drainage gate at different times, combined with the adaptive time window describing the local volatility of the time series data, the Hurt index is used to first determine whether the time series data is stable, and the non-stationary data is differentiated to stabilize the series data. The adaptive time window can better capture the local volatility of the time series data, improve the adaptability and prediction accuracy of the model, and the stationarity of the time series is determined by the Hurst index, and the non-stationary data is differentiated to ensure the stationarity of the model input data, thereby improving the stability of the model and the prediction accuracy. After that, the autoregressive integrated moving average model parameters are determined in combination with the extreme value theory, and the autoregressive integrated moving average model is constructed, so that the model can more accurately capture extreme values ​​and long-term trends, thereby significantly improving the accuracy and reliability of the prediction, and ensuring that the prediction results are more stable and reliable. For the constructed model, determine whether its model residual conforms to the random distribution. If not, adjust the model parameters in combination with the gradient descent method until its model residual conforms to the random distribution. By checking whether the model residual conforms to the random distribution, it can be determined whether the model fully fits the data. If the residual does not conform to the random distribution, it means that the model may have systematic deviations or fail to capture certain laws. By adjusting the model parameters in combination with the gradient descent method, the model can be automatically optimized so that the residual gradually approaches the random distribution. The gradient descent method gradually reduces the prediction error by iteratively adjusting the parameters, avoiding the tediousness and subjectivity of manual parameter adjustment, and ensuring that the model adaptively improves the prediction accuracy during the training process. Finally, the model whose model residual conforms to the random distribution is used to predict the foundation height of the drainage gate for risk warning. By combining the adaptive time window, Hurst index, extreme value theory and gradient descent method, the volatility of time series data can be more accurately processed, data stability can be ensured, and model parameters can be automatically optimized, which improves the adaptability, accuracy and stability of the model, reduces manual intervention, ensures that the prediction results are more reliable, and effectively improves the accuracy and reliability of the drainage gate foundation settlement risk warning.

[0140] Reference Manual Attached Figure 2 , showing a schematic flow chart of a method for predicting foundation settlement risk of a drainage gate provided by the present invention.

[0141] The present invention also provides a method for predicting the risk of foundation settlement of a drainage gate, comprising:

[0142] S1: Obtain time series data describing the historical foundation height of the drainage gate at different times.

[0143] S2: Combined with the adaptive time window that describes the local volatility of the time series data, the Hurt index is used to determine whether the time series data is stationary. If so, the construction module is called, otherwise, the difference module is called.

[0144] S3: Determine the difference order in combination with the Hurt index, and differentiate the time series data based on the difference order to stabilize the time series data.

[0145] S4: Based on the stationary time series data, the autoregressive integrated moving average model parameters are determined in combination with the extreme value theory, and the autoregressive integrated moving average model is constructed, where the autoregressive integrated moving average model parameters include the autoregressive order and the moving average order.

[0146] S5: Determine whether the model residual of the autoregressive integrated moving average model conforms to the random distribution. If so, call the output module; otherwise, call the adjustment module.

[0147] S6: Adjust the parameters of the autoregressive integrated moving average model by the gradient descent method to determine whether the model residuals of the autoregressive integrated moving average model conform to the random distribution, until the model residuals of the autoregressive integrated moving average model conform to the random distribution.

[0148] S7: Use the autoregressive integrated moving average model to output the predicted drainage gate foundation height within a preset time period.

[0149] S8: Issue a risk warning when there is a predicted drainage gate foundation height that is lower than the preset drainage gate foundation height.

[0150] The method for predicting the risk of foundation settlement of a drainage gate provided by the present invention is applied to the above-mentioned system for predicting the risk of foundation settlement of a drainage gate, and achieves the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0151] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for predicting foundation settlement risk of a drainage gate as described in the method embodiment is implemented.

[0152] A computer-readable storage medium provided by the present invention can implement the steps and effects of the method for predicting foundation settlement risk of a drainage gate in the above-mentioned method embodiment. To avoid repetition, the present invention will not go into details.

[0153] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0154] There are a few points to note:

[0155] (1) The drawings of the embodiments of the present invention only involve structures related to the embodiments of the present invention. Other structures may refer to conventional designs.

[0156] (2) For the sake of clarity, in the drawings used to describe the embodiments of the present invention, the thickness of layers or regions is exaggerated or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0157] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0158] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A foundation settlement risk prediction system for a drainage gate, characterized in that: include: An acquisition module, used to acquire time series data describing the historical foundation height of the drainage gate at different times; The first judgment module is used to combine the adaptive time window describing the local volatility of the time series data and judge whether the time series data is stable through the Hurt index. If so, the construction module is called, otherwise, the difference module is called; A difference module, used for determining a difference order in combination with the Hurt index, and performing a difference on the time series data based on the difference order to stabilize the time series data; A construction module is used to determine the parameters of the autoregressive integrated moving average model based on the stationary time series data in combination with the extreme value theory, and to construct the autoregressive integrated moving average model, wherein the parameters of the autoregressive integrated moving average model include the autoregressive order and the moving average order; The second judgment module is used to judge whether the model residual of the autoregressive integrated moving average model conforms to the random distribution, and if so, call the output module, otherwise, call the adjustment module; An adjustment module, used for adjusting the parameters of the autoregressive integrated moving average model by a gradient descent method, and calling the second judgment module until the model residual of the autoregressive integrated moving average model conforms to a random distribution; An output module, used to output the predicted drainage gate foundation height within a preset time period using the autoregressive integrated moving average model; An early warning module, for issuing a risk early warning when there is a predicted drainage gate foundation height that is lower than a preset drainage gate foundation height; The calculation method of the predicted drainage gate foundation height is specifically as follows: ； in, and Respectively represent the T Time lag h Step and h - i Step 1: Predict the drainage gate foundation height. Represents relative to T Time lag h - i The residual prediction value after step , It represents the value calculated based on the historical foundation height of the drainage gate at time T To time T + h The predicted mean within Indicates i The autoregressive coefficients, , p represents the maximum autoregressive order, Indicates j The moving average coefficient, , q Indicates the maximum moving average order.

2. The foundation settlement risk prediction system for a drainage gate according to claim 1, characterized in that: The first judgment module combines the adaptive time window describing the local volatility of the time series data and judges whether the time series data is stable by using the Hurt index, specifically including: Determine the adaptive time window based on the local volatility of the time series data: ； in, Indicates time t The time window size, represents an adjustable constant that controls the relationship between window size and local volatility, Indicates the time series data in the current time window The standard deviation of the time series data at time t The local volatility of Represents the first i Observations, Indicates the time series data in the current time window The local mean within Represented on time series data The average value of Calculate the ratio of the maximum cumulative deviation to the standard deviation within each adaptive time window: ； in, Indicates that time series data is t The cumulative deviation of time, R Indicates the maximum cumulative deviation within each adaptive time window Minimum cumulative deviation difference, S represents the standard deviation within the adaptive time window, i.e. ; The ratios in each adaptive time window are logarithmically fitted to obtain the Hurt index: ； in, H represents the Hurt index, log represents the logarithmic function, C represents the intercept constant; When the Hurt index takes a value of 0.5, the time series data is output as a steady state; otherwise, the time series data is output as a non-steady state.

3. The foundation settlement risk prediction system for a drainage gate according to claim 2, characterized in that: The formula for determining the difference order is specifically: ； in, d represents the difference order.

4. The foundation settlement risk prediction system for a drainage gate according to claim 1, characterized in that: The building blocks are specifically used for: Establishing the autocorrelation function and partial autocorrelation function of the stationary time series data; Calculating function coefficients of the autocorrelation function and the partial autocorrelation function under different autoregressive integrated moving average model parameters, wherein the function coefficients include the autocorrelation coefficient and the partial autocorrelation coefficient; Determine the confidence intervals for the coefficients of the function based on the quantiles of the standard normal distribution: ； in, CI represents the confidence interval, Indicates the significance level corresponding to α The standard normal distribution quantile of n It represents the sample size in the stationary time series data, that is, the total number of observation points in the stationary time series data; Determine the autoregressive integrated moving average model parameter corresponding to the function coefficient that exceeds the confidence interval as the extreme value of the corresponding function; Fit the extreme values ​​belonging to the same function through the generalized Pareto distribution to obtain the fitted values: ； in, x represents an extreme value, represents the function coefficient at the minimum extreme value, represents the width of the extreme value distribution, represents the shape parameter describing the distribution of extreme values, exp represents the exponential function, Indicates about The cumulative distribution function value of express The cumulative distribution function value when , log represents the logarithmic function, Indicates about The likelihood function of Indicates i extreme values, , m represents the total number of extreme values, Indicates that The smallest hour and as the fitted value; Compute the return period of each extreme value relative to the fitted value: ； in, Represents extreme values x The return cycle, Represents extreme values x The cumulative distribution function value of ; The extreme value corresponding to the maximum return period is used as the optimal autoregressive integrated moving average model parameter; Substitute the autoregressive integrated moving average model parameters into the autoregressive integrated moving average model to complete the construction of the autoregressive integrated moving average model: ； in, express t The observed value at time, c represents a constant term, Indicates i The autoregressive coefficients, , p represents the maximum autoregressive order, Indicates j The moving average coefficient, , q represents the maximum moving average order, Representation and t and j The associated lagged residuals, express t Time residual.

5. The foundation settlement risk prediction system for the drainage gate according to claim 4 is characterized in that: The second judgment module judges whether the model residual of the autoregressive integrated moving average model conforms to the random distribution, specifically including: Calculate the model residuals for the autoregressive integrated moving average model: ； in, express t The model residual at time t, Respectively represent the actual observed value at time t and the model predicted value at time t; Calculate the residual correlation coefficient of the model: ； in, k represents the lag order, express of k The residual correlation coefficient of the order model is express The mean of The lag order is t - k The model residual at time t, n Represents the total number of model residuals; The Q test statistic is calculated based on the residual correlation coefficient of the model: ； in, represents the Q test statistic; Calculate the probability of the Q test statistic under the prior assumption that the model residuals conform to a random distribution: ； in, The degrees of freedom are df The chi-square distribution of represents the probability value of the Q test statistic, Indicates that the degrees of freedom are df The probability value of the Q test statistic under the chi-square distribution; When the occurrence probability value is greater than a preset occurrence probability value, the model residuals output of the autoregressive integrated moving average model conform to the random distribution, otherwise, the model residuals output of the autoregressive integrated moving average model do not conform to the random distribution.

6. The foundation settlement risk prediction system for a drainage gate according to claim 1, characterized in that: The adjustment module is specifically used for: Calculate the predicted loss value of the autoregressive integrated moving average model under the autoregressive integrated moving average model parameters: ； in, represents the autoregressive integrated moving average model parameters The predicted loss value under and Respectively t The actual observed value at the moment and t The model prediction value at that moment, L Indicates the length of time series data; Calculate the gradient of the predicted loss value with respect to each autoregressive integrated moving average model parameter: ； in, represents partial derivative, Represents the prediction loss value Relative to the autoregressive integrated moving average model parameters The gradient of Based on the obtained gradient update, the corresponding autoregressive integrated moving average model parameters are: ； in, represents the updated autoregressive integrated moving average model parameters, Indicates the update step size.

7. The foundation settlement risk prediction system for a drainage gate according to claim 1, characterized in that: The early warning module is specifically used for: When there is a predicted drainage gate foundation height that is lower than the preset drainage gate foundation height, a risk warning is issued through a warning light, a speaker or information push.

8. The foundation settlement risk prediction system for a drainage gate according to claim 1, characterized in that: Also includes: A conversion module is used to convert the predicted drainage gate foundation height into a predicted drainage gate foundation height interval output: ； in, PI It indicates the predicted drainage gate foundation height interval. Represents the difference between the historical foundation height of the drainage gate and h The associated forecast error variance; A risk warning is issued for the predicted drainage gate foundation height interval that is lower than the preset drainage gate foundation height.

9. A method for determining the risk of foundation settlement of a drainage gate, characterized in that: include: S1: Obtain time series data describing the historical foundation height of the drainage gate at different times; S2: Combined with the adaptive time window describing the local volatility of the time series data, the Hurt index is used to determine whether the time series data is stable. If so, the construction module is called, otherwise, the difference module is called; S3: determining a difference order in combination with the Hurt index, and performing differentiation on the time series data based on the difference order to stabilize the time series data; S4: Based on the stable time series data, the autoregressive integrated moving average model parameters are determined in combination with the extreme value theory, and the autoregressive integrated moving average model is constructed, wherein the autoregressive integrated moving average model parameters include the autoregressive order and the moving average order; S5: Determine whether the model residual of the autoregressive integrated moving average model conforms to the random distribution, if so, call the output module, otherwise, call the adjustment module; S6: adjusting the parameters of the autoregressive integrated moving average model by a gradient descent method, and determining whether the model residual of the autoregressive integrated moving average model conforms to a random distribution, until the model residual of the autoregressive integrated moving average model conforms to a random distribution; S7: using the autoregressive integrated moving average model to output the predicted drainage gate foundation height within a preset time period; S8: Issue a risk warning when there is a predicted drainage gate foundation height that is lower than the preset drainage gate foundation height; The calculation method of the predicted drainage gate foundation height is specifically as follows: ； in, and Respectively represent the T Time lag h Step and h - i Step 1: Predict the drainage gate foundation height. Represents relative to T Time lag h - i The residual prediction value after step , It represents the value calculated based on the historical foundation height of the drainage gate at time T To time T + h The predicted mean within Indicates i The autoregressive coefficients, , p represents the maximum autoregressive order, Indicates j The moving average coefficient, Indicates the maximum moving average order.

Images