Hypobaric analysis methods, equipment, and media that integrate model optimization and physical inversion

Abstract

This invention discloses a method, equipment, and medium for seepage pressure analysis that integrates model optimization and physical inversion. The method includes: calculating the correlation coefficient between a specified reservoir water level and seepage pressure under different time lags based on cross-correlation function analysis, and identifying the differentiated nonlinear characteristics corresponding to the maximum correlation coefficient; constructing multiple diagnostic prediction models in parallel for the differentiated nonlinear characteristics of different dam types, and establishing a model selection decision rule to select the optimal diagnostic model; using the lag parameter obtained from cross-correlation analysis to back-calculate the permeability coefficient using the heat transfer and diffusion theory formula of porous media; using the regression slope and theoretical value attenuation factor of the optimal model, combined with the modified diffusion equation, to back-calculate the permeability coefficient; performing credibility assessment and weighted fusion of the permeability coefficient to calculate the comprehensive permeability coefficient; and diagnosing the soil type and permeability characteristic level. This invention can accurately quantify the propagation delay characteristics of reservoir water level changes in porous media, improving the accuracy and engineering applicability of permeability coefficient inversion.

Description

Technical Field

[0001] This invention relates to the field of reservoir seepage pressure parameter data processing technology, specifically to a seepage pressure analysis method, equipment, and medium based on multi-model adaptive fusion and physical parameter inversion for fusion model optimization and physical inversion. Background Technology

[0002] Seepage pressure monitoring plays a crucial role in the safe operation of dam projects and reservoir management. The accuracy and diagnostic capabilities of its monitoring data directly affect the quality of dam safety assessment and emergency response. Existing seepage pressure monitoring and analysis systems mainly rely on traditional single data sources and static threshold judgment methods. Although machine learning techniques (such as support vector machines, random forests, and neural networks) have been applied to seepage pressure prediction in recent years, they mostly adopt general algorithm frameworks and fixed feature selections, lacking adaptive mechanisms for the differences in seepage characteristics of different dam types, and lacking deep integration of engineering physics knowledge. This makes it difficult to cope with the diagnostic challenges brought about by complex reservoir water level changes, nonlinear seepage pressure response, and differences in characteristics of different dam types in practical applications.

[0003] Especially under complex operating conditions, existing methods suffer from problems such as inaccurate identification of reservoir water level-seepage pressure relationship characteristics, poor model generalization ability, and lack of engineering physics guidance. These issues often lead to core technical challenges for seepage pressure monitoring systems in multi-source data collaborative processing and complex diagnostic scenarios, including the following:

[0004] (1) Single model is difficult to adapt to dam type differences: Existing seepage pressure diagnosis systems mostly adopt a single algorithm framework (such as linear regression, single neural network, etc.), but in actual engineering, the reservoir water level-seepage pressure relationship of different dam types shows significant nonlinear characteristics. At this time, a single model either overfits the characteristics of a certain type of dam and has poor generalization ability to other dam types, or adopts a general fitting, which makes it inaccurate for any dam type, making it difficult to meet the diverse needs of engineering diagnosis.

[0005] (2) Lack of objective evaluation criteria for model selection: Traditional methods often rely on human experience and a single indicator (such as R²) when selecting among multiple candidate models. They lack a multi-dimensional and multi-level model comparison and evaluation framework. Even if the goodness of fit of multiple models is calculated, there is no unified and quantitative adaptive selection mechanism, resulting in insufficient reliability and repeatability of diagnostic results.

[0006] (3) Disconnect between statistical models and physical parameters: Although existing machine learning methods can obtain the regression relationship between reservoir water level and seepage pressure (such as slope) Time delay However, these statistical parameters themselves do not have a clear engineering physical meaning and cannot be directly converted into decision parameters that engineers care about (such as permeability coefficient K, soil type), which can easily cause a conversion gap from data observation to engineering diagnosis.

[0007] (4) The physical reverse reasoning approach is singular and unreliable: Even if some studies attempt to use a single physical model (such as based solely on time delay), The diffusion equation can be used to back-calculate the permeability coefficient, but this single-path back-calculation is easily affected by factors such as measurement errors and model assumption biases. It lacks multi-angle verification mechanisms and credibility assessment, making it difficult to guarantee the credibility of the back-calculation results.

[0008] (5) Lack of quantitative transformation of engineering diagnostic conclusions: Even if the permeability coefficient K value is derived, it is difficult to further transform it into usable engineering diagnostic conclusions (such as soil type determination, permeability characteristic classification, etc.). Existing methods mostly rely on empirical range benchmarking and lack a scientific and automated diagnostic classification system.

[0009] To address these technical problems, this application proposes a osmotic pressure analysis method that integrates model optimization and physical inversion. Summary of the Invention

[0010] The main objective of this invention is to provide a fusion model optimization and physical inversion method for osmotic pressure analysis, in order to solve the technical problems mentioned in the background art, such as insufficient fusion of multi-source data, lack of systematic identification of reservoir water level-osmotic pressure relationship characteristics, difficulty of single diagnostic models to adapt to dam type differences, and faults in the conversion of statistical parameters and physical parameters.

[0011] The present invention solves the above-mentioned technical problems by adopting the following technical solutions:

[0012] A osmotic pressure analysis method that integrates model optimization and physical inversion, comprising the following steps performed using computer equipment:

[0013] Step S1. Cross-correlation delay analysis and engineering response feature identification: Based on the cross-correlation function (CCF) analysis, calculate the water level and seepage pressure of the specified reservoir at different time lags. The correlation coefficient is calculated, and the optimal delay time corresponding to the maximum correlation coefficient is identified. The causal relationship between reservoir water level changes and seepage pressure response, as well as the time characteristic parameters of engineering response, were quantified, and an engineering response characteristic parameter system with differentiated nonlinear characteristics was established.

[0014] Step S2. Parallel Construction and Objective Comparative Evaluation of Multiple Models: To address the potentially differentiated nonlinear characteristics of different dam types within the feature parameter system, multiple diagnostic prediction models, including linear regression, multinomial regression, and robust regression (Huber), are constructed in parallel. The aligned data is then fitted using the optimized data, and the coefficient of determination R is calculated for each model. 2The system employs a multi-dimensional evaluation framework, including Root Mean Square Error (RMSE) and Akaike Information Content (AIC), and verifies the statistical significance of regression relationships through residual diagnosis (normality test, heteroscedasticity detection, and autocorrelation analysis). Finally, based on the comprehensive comparison results of the multi-dimensional evaluation indicators, an objective model selection decision rule is established. The system automatically selects the optimal diagnostic model according to the priority of specified indicators such as RMSE or AIC, so as to achieve adaptive matching for the degree of nonlinearity of different dam types and ensure the optimal fit between the diagnostic model and the specific engineering characteristics.

[0015] Step S3. Dual-path reverse propagation mechanism for physical parameters:

[0016] (1) Design two independent physical reverse reasoning routes respectively:

[0017] Physical back-calculation route A based on the diffusion equation with time delay: Delay parameters obtained from cross-correlation analysis The permeability coefficient is calculated by using the heat transfer and diffusion theory formula of porous media.

[0018] Hydraulic physics back-calculation route B based on regression characteristics: based on the regression slope of the optimal model. The attenuation factor, combined with the theoretical value, is used to deduce the second set of estimated permeability coefficients using the modified diffusion equation.

[0019] (2) Credibility weighted fusion and parameter comprehensive estimation: The credibility of the two independent back-inferenced penetration coefficient estimates is then evaluated and weighted fusion is performed. The theoretical basis, parameter accuracy and physical rationality of each route are comprehensively considered to calculate the comprehensive penetration coefficient estimate and obtain a more robust and reliable back-inference result.

[0020] Step S4. Automatic Soil Type Diagnosis and Classification: The inversely derived permeability coefficient is compared with the standard reference range for classification to automatically diagnose soil type and permeability level. The diagnostic results are then visualized and output as an engineering report. An engineering information system integrating data management, analysis and processing, and diagnostic decision-making is developed to generate multi-dimensional diagnostic charts (reservoir water level-permeability fitting effect chart, model diagnosis four-in-one chart, etc.) and quantitative diagnostic reports, clearly showing the inversion process, parameter estimation, uncertainty assessment and final diagnostic conclusions, providing quantitative and interpretable decision support for engineers.

[0021] Preferably, the specific process for calculating the correlation coefficient based on the cross-correlation function analysis in step S1 includes:

[0022] Reservoir water level time series analysis using cross-correlation function (CCF) With osmotic pressure time series To perform delayed correlation analysis, the cross-correlation function is defined as follows:

[0023]

[0024] in, Indicated as time lag The correlation coefficient below, and These are the average reservoir water level and the average seepage pressure, respectively. Indicated as time lag The reservoir water level below Indicated as time lag The osmotic pressure below, The total number of data points. This represents the time lag steps, here. The data point number represents the current time. This represents how many data points have been shifted forward, and there exists a set of parameters. This represents the time difference between two data points;

[0025] Recognition Reaching the maximum value Optimal delay time Calculate the optimal delay time for:

[0026]

[0027] in, This delay parameter represents the sampling time interval (in hours). It is used to quantitatively reflect the time lag characteristics between reservoir water level changes and seepage pressure response, and is a key input for subsequent physical parameter inversion.

[0028] Preferably, the identification and extraction process of differential nonlinear features in step S1 includes:

[0029] Based on the optimal delay of identification Time alignment was performed on the original data sequence to align the reservoir water level sequence. Move backward Step, to obtain the aligned data pair sequence, we have:

[0030] Align reservoir water level :

[0031]

[0032] Alignment osmosis :

[0033]

[0034] Statistical features are extracted based on the aligned data, including the aligned reservoir water level. Sequence and Alignment Isobaric Pressure Maximum correlation coefficient of the series (Degree of correlation), initial correlation and optimal delay time This ultimately forms a set of engineering response characteristic parameters. .

[0035] Preferably, the parallel construction of the diagnostic prediction model in step S2 specifically includes:

[0036] a. Linear regression model, the calculation formula of which is:

[0037]

[0038] in, This is expressed as the slope of the linear regression. and These are represented by the linear regression intercept and the random error term of the model, respectively.

[0039] The model uses the least squares method to solve for the regression coefficients, and its 95% two-sided confidence intervals are... for:

[0040]

[0041] in, For standard error, For degrees of freedom of Distribution critical value, The significance level is used to determine the critical value.

[0042] b. The second-order polynomial regression model has the following calculation formula:

[0043]

[0044] in, , and These are respectively represented as the quadratic regression coefficient, the linear regression coefficient, the regression constant term, and the random error term of the model;

[0045] c. Robust regression model, the calculation formula of which is:

[0046]

[0047] in, Represented as the Huber loss function, Indicated as time lag The data residuals below.

[0048] Preferably, the specific operational procedures for calculating multi-dimensional evaluation indicators by fitting the diagnostic prediction model and verifying the statistical significance of the regression relationship through residual diagnosis in step S2 include:

[0049] For linear regression, second-order polynomial regression, and robust regression models, the following multidimensional evaluation indicators are calculated respectively:

[0050] Coefficient of determination : ;

[0051] Root mean square error : ;

[0052] Akaike Information Content (AIC): ;

[0053] in, For the sample size, The number of seepage pressure parameters in the model. Mean square error, For the first The actual observed osmotic pressure of each sample. This represents the average of the actual observed seepage pressure values. For the first The osmotic pressure model predicted values ​​for each sample;

[0054] Finally, the statistical significance of the regression relationship is verified using residual diagnostic indicators:

[0055] Normality test: The Shapiro-Wilk test was used to obtain standard values ​​for determining whether a distribution is normal.

[0056] Heteroscedasticity detection: Calculate the correlation coefficient between the residuals and the predicted values;

[0057] Autocorrelation detection: Calculate the Durbin-Watson statistic of the residuals;

[0058] Subsequently, in establishing the model selection decision rules, an optimal model was identified based on the RMSE metric, and cross-validation was performed using AIC to ensure that the selected model achieved an optimal balance between fitting accuracy and complexity. The optimal model identified based on the RMSE metric is as follows:

[0059]

[0060] in, Represented as the first The RMSE metric for each model.

[0061] Preferably, the specific operational procedure for calculating the permeability coefficient A in step S3 includes:

[0062] Using the heat transfer and diffusion theory of porous media, the propagation of reservoir water level changes in the soil follows a one-dimensional unsteady flow diffusion equation. The delay parameter is identified based on cross-correlation analysis. The inverse formula is established as follows:

[0063]

[0064] in, Porosity (a dimensionless constant). The hole depth is in meters (m). The delay time is in seconds. For the estimation of the permeability coefficient derived from this point, the longer the delay time, the slower the water spreads in the porous medium, and the smaller the corresponding permeability coefficient.

[0065] Preferably, the specific operational procedure for calculating the permeability coefficient B in step S3 includes:

[0066] A unit head change in reservoir water level, without attenuation, should produce a unit pressure response of the same magnitude at the seepage point. Therefore, an attenuation relationship is established between the theoretical slope and the actual slope. Let the theoretical slope be... for:

[0067]

[0068] Calculate the attenuation factor for:

[0069]

[0070] in, The actual regression slope, here is the attenuation factor. This reflects the degree of attenuation of the osmotic pressure response relative to the theoretical value;

[0071] Based on the attenuation factor, the diffusion equation is modified to obtain the following modified equation:

[0072]

[0073] in, The formula incorporates the influence of osmotic pressure attenuation characteristics as the attenuation coefficient. This approach quantifies the deviation of regression characteristics into an attenuation factor, which is then incorporated into the physical model through a correction coefficient, thus achieving a mapping from statistical observations to physical parameters.

[0074] Preferably, the specific calculation process for calculating the comprehensive penetration coefficient estimate in step S3 includes:

[0075] The two reverse-engineering routes, based on different physical foundations and data characteristics, yielded the following results: and Two independent valuations are used; to improve the robustness and reliability of the back-calculation results, a credibility-weighted fusion strategy is adopted as follows:

[0076] Based on correlation coefficient and attenuation factor To assess the rationality of each route and evaluate its credibility weight, the weight calculation formula is as follows:

[0077]

[0078]

[0079] in, For the first The maximum number of cross-correlation coefficients of the routes, This is the maximum correlation coefficient in cross-correlation analysis. The closer this index is to 1, the higher the credibility of the delayed route.

[0080] Calculate the overall penetration coefficient valuation for:

[0081]

[0082] Simultaneously, the degree of deviation between the two estimates is calculated, which serves as an indicator of the uncertainty of the inversion result. The calculation formula is as follows:

[0083]

[0084] when In this case, the results obtained by reverse engineering are highly reliable;

[0085] when At that time, the results were of moderate reliability;

[0086] when At this time, it is necessary to review the data quality and model assumptions.

[0087] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0088] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0089] As can be seen from the above technical solution, the present invention provides a osmotic pressure analysis method that integrates model optimization and physical inversion. Compared with the prior art, the present invention has the following advantages:

[0090] 1. This invention establishes a mechanism including delay parameters through cross-correlation function analysis. The engineering response characteristic parameter system of maximum correlation coefficient and initial correlation can accurately quantify the propagation delay characteristics of reservoir water level changes in porous media. It can not only obtain time delay information, but also objectively diagnose the linear / nonlinear characteristic intensity of different engineering processes by the strength of the correlation coefficient. Ultimately, it breaks the subjectivity of traditional experience judgment and realizes the systematic and quantitative transformation from observation data to engineering characteristics.

[0091] 2. This invention, by constructing a model spectrum covering different degrees of nonlinearity based on accurate feature recognition, can realize a scientific mapping mechanism from engineering features to model complexity. Specifically, through a three-model parallel framework, it forms a complete coverage from linear to nonlinear and from ideal to robust: the linear model corresponds to the ideal working condition with high correlation and weak interference, the second-order polynomial corresponds to the complex working condition with medium nonlinearity, and Huber regression corresponds to the actual working condition with strong interference and frequent anomalies.

[0092] 3. This invention addresses the deviation between two physical back-calculation routes: one based on the diffusion equation under time delay and the other on the hydraulic attenuation based on statistical response. It can establish a graded confidence assessment mechanism, so that the inversion results include not only parameter values, but also a clear confidence assessment.

[0093] 4. This invention significantly improves the accuracy and engineering practicality of permeability coefficient inversion through multi-dimensional feature interaction constraints, adaptive model selection, and reliable quantification, providing a scientific parameter basis and intelligent decision support for dam seepage monitoring, risk assessment, and long-term management.

[0094] It should be understood that the descriptions in this section are not intended to identify key or essential features of embodiments of the invention, nor are they intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Of course, implementing any product of the invention does not necessarily require achieving all of the advantages described above simultaneously. Attached Figure Description

[0095] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0096] Figure 1 This is a three-dimensional schematic diagram of the entire invention. Detailed Implementation

[0097] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0098] For details in the embodiments, please refer to Figure 1 .

[0099] The fusion model optimization and physical inversion seepage pressure analysis method proposed in this invention first achieves real-time acquisition of multi-source engineering parameters such as reservoir water level and seepage pressure values ​​through a distributed monitoring sensor network. Unified preprocessing methods, including missing value interpolation, data type standardization, and time series alignment, are employed to ensure the stability and accuracy of heterogeneous data. Second, cross-correlation function analysis is used to identify the time delay relationship between reservoir water level changes and seepage pressure response, establishing an engineering response time characteristic parameter system based on the delay time corresponding to the maximum correlation coefficient. Subsequently, to adapt to the diversity of seepage characteristics for different dam types, multiple prediction models, including linear, polynomial, and robust regression models, are innovatively constructed in parallel, providing differentiated fitting schemes for reservoir water level-seepage pressure relationships with different degrees of nonlinearity. Based on R... 2 Objective comparative evaluation was conducted using multiple dimensions such as RMSE and AIC. The statistical significance of the regression relationship was verified through residual diagnosis, achieving an adaptive selection mechanism for the model. A dual-path back-inference mechanism for physical parameters was proposed, based on time delay. and regression slope The attenuation factor is calculated, and the permeability coefficient K is inversely derived through the heat transfer and diffusion theory of porous media. The two estimates are then weighted and fused based on their credibility. Finally, by comparing the magnitude of K with the reference range, automatic diagnosis and classification early warning of soil type are achieved. An engineering information system integrating data management, analysis and processing, diagnostic decision-making, and visualization reports is developed.

[0100] At this point, by integrating various techniques such as cross-correlation analysis, multi-model adaptive fusion evaluation, and physical parameter inversion based on the characteristics of reservoir dam seepage monitoring data, and through the design of diverse models, the limitations of a single model in adapting to complex working conditions are overcome. This significantly improves the applicability and robustness of the method to different dam types, and realizes a complete mapping transformation from statistical observations to engineering decision parameters. It can significantly improve the accuracy of seepage pressure anomaly diagnosis and the reliability of early warning decisions, and can be applied to the field of intelligent monitoring and risk early warning of reservoir dam seepage pressure that requires high precision.

[0101] In further practical use, such as Figure 1 As shown, the specific steps include:

[0102] Step S1: Cross-correlation delay analysis and engineering response feature identification.

[0103] Based on the cross-correlation function (CCF) analysis method, the reservoir water level and seepage pressure were calculated at different time lags. The correlation coefficient is used to identify the optimal delay time corresponding to the maximum correlation coefficient. The causal relationship between reservoir water level changes and seepage pressure response, as well as the time characteristic parameters of engineering response, were quantified, and an engineering response characteristic parameter system was established.

[0104] In the specific implementation process, the specific operations include:

[0105] S11. Cross-correlation delay analysis and engineering response feature identification

[0106] Cross-correlation function (CCF) was used to analyze the reservoir water level time series. With osmotic pressure time series Perform delay correlation analysis. For a given time lag... Calculate the correlation coefficient between the two, where the correlation coefficient is defined as:

[0107]

[0108] in and These are the average values ​​of the reservoir water level and the seepage pressure, respectively. The total number of data points. This represents the time lag steps, here. The data point number represents the current time. This represents how many data points have been shifted forward, and there exists a set of parameters. This represents the time difference between two data points.

[0109] By iterating through all possible delays , From the maximum delay time and sampling interval Determine the maximum latency , recognition Optimal delay time to reach the maximum value Based on this, the optimal delay time is calculated:

[0110]

[0111] in This is the sampling time interval (in hours). This delay parameter... It quantitatively reflects the time lag characteristics between reservoir water level changes and seepage pressure response, and is a key input for subsequent physical parameter inversion.

[0112] S12, Data Alignment and Feature Parameter Extraction

[0113] Based on the optimal delay of identification Time alignment is performed on the original data sequence, shifting the reservoir water level sequence H backward. Step 1, to obtain aligned data pairs:

[0114] Align the reservoir water level:

[0115] Alignment of osmotic pressure:

[0116] Statistical features are extracted based on aligned data, including the correlation between the two sequences (maximum correlation coefficient). ), initial correlation and delay information This forms a set of engineering response characteristic parameters {τ, , }

[0117] In summary, cross-correlation analysis not only identified the delay parameter τ (time response characteristics), but more importantly, it obtained the correlation coefficient. This statistic implicitly reflects the quality of data and the degree of nonlinearity in the engineering process. That is, a high correlation coefficient indicates a clear physical coupling relationship between reservoir water level and seepage pressure and significant linear characteristics of the process, while a low correlation coefficient indicates large interference and strong nonlinear characteristics.

[0118] This diagnostic feature provides an objective basis for subsequent adaptive model selection: when When the correlation is close to 1, a simplified linear model is sufficient to describe the engineering process; when the correlation is weak, the system automatically tends to choose a second-order polynomial model or robust regression to capture the implicit nonlinear characteristics. Therefore, this step extracts not only the input parameters, but also a "diagnosis" of the characteristics of the entire system, which can be used to guide the complexity of subsequent models.

[0119] In summary, at this point, a cross-correlation function analysis is used to establish a system that includes delay parameters. The engineering response characteristic parameter system of maximum correlation coefficient and initial correlation can accurately quantify the propagation delay characteristics of reservoir water level changes in porous media. It can not only obtain time delay information, but also objectively diagnose the linear / nonlinear characteristic intensity of different engineering processes by the strength of the correlation coefficient. Ultimately, it breaks the subjectivity of traditional experience judgment and realizes the systematic and quantitative transformation from observation data to engineering characteristics.

[0120] Step S2: Parallel construction of multiple models and objective comparative evaluation.

[0121] To address the potentially differentiated nonlinear characteristics of different dam types, multiple diagnostic and predictive models, including linear regression, multinomial regression, and robust regression (Huber), were constructed in parallel. Aligned data was fitted using optimized data, and the R-squared value of each model was calculated. 2 The system employs multi-dimensional evaluation indicators such as coefficient of determination, root mean square error (RMSE), and AIC (Akaike Information Content). It verifies the statistical significance of regression relationships through residual diagnosis (normality test, heteroscedasticity detection, and autocorrelation analysis), establishing a comprehensive multi-indicator evaluation framework. Based on the comprehensive comparison results of these multi-dimensional evaluation indicators, it establishes objective model selection decision rules, automatically selecting the optimal diagnostic model according to priority indicators such as RMSE or AIC. Ultimately, it achieves adaptive matching for the degree of nonlinearity of different dam types, ensuring optimal adaptation of the diagnostic model to specific engineering characteristics.

[0122] In the specific implementation process, the specific operations include:

[0123] S21. Parallel Construction of Multiple Models and Regression Analysis

[0124] To address the differences in nonlinear characteristics among different dam types, three parallel model frameworks are constructed as follows:

[0125] (1) Linear regression model

[0126]

[0127] in, This is expressed as the slope of the linear regression. and These are represented as the linear regression intercept and the random error term of the model, respectively.

[0128] This linear regression model is the basic model for describing the relationship between reservoir water level and seepage pressure. Its physical meaning lies in assuming that the pore medium of the dam body is homogeneous and isotropic, and the seepage path is stable. At this time, a unit change in reservoir water level corresponds to a constant seepage pressure response amplitude.

[0129] In practical use, this model is applicable to the seepage process of homogeneous earth-rock dams under stable operating conditions, and has the advantages of fewer parameters, strong interpretability, and high computational efficiency.

[0130] Furthermore, the least squares method is used to solve for the regression coefficients, and the confidence intervals are calculated simultaneously.

[0131] At this point, regarding the regression slope Its 95% two-sided confidence interval is:

[0132]

[0133] in For standard error, For degrees of freedom of Distribution critical value, The significance level is used to determine the critical value.

[0134] (2) Second-order polynomial regression model

[0135]

[0136] This second-order polynomial regression model is used to capture the nonlinear characteristics in the reservoir water level-seepage pressure relationship.

[0137] The physical background is that in actual engineering, dam bodies often have heterogeneity (such as layered filling, local lens bodies), the saturation of the seepage path changes with the water level, and the geometric differences in the seepage field at high and low water levels. These factors cause the slope of the seepage pressure response to change with the reservoir water level.

[0138] In addition, the model employs multinomial feature transformation and linear regression for solution.

[0139] (3) Robust Regression Model - Huber Regression Model

[0140]

[0141] This robust regression model is used to address the problem of anomalous observations that are common in long-term monitoring data.

[0142] It's important to note that traditional least squares methods are highly sensitive to extreme values, and a single outlier can cause regression coefficients to deviate significantly from the true value. In actual engineering, factors such as instrument malfunctions, human disturbances, and sudden local leaks often lead to abnormal fluctuations in seepage pressure at specific moments. The Huber loss function combines the advantages of squared loss (sensitive to small errors) and absolute value loss (tolerant of large errors): it uses squared loss to maintain statistical efficiency when the residuals are within a reasonable range, and switches to linear growth when the residuals exceed a threshold to limit the influence weight of outliers. Therefore, this model ensures that even with extreme observations, the regression coefficients still reflect the true patterns of the main data, making it suitable for long-term monitoring sequences, scenarios with sudden disturbances, or inconsistent data quality. The Huber loss function is fault-tolerant to extreme observations.

[0143] At this point, using the Huber loss function to handle outliers makes the model more robust to extreme observations. By combining a linear regression model to capture the main linear features, a second-order polynomial regression model to capture nonlinear changes, and a robust regression model to ensure robustness against anomalies, a complete model cluster covering different engineering scenarios is formed.

[0144] In summary, the system can now automatically select the optimal model based on the data characteristics of a specific project.

[0145] S22. Calculation and Objective Comparison of Multi-Dimensional Evaluation Indicators

[0146] The following multidimensional evaluation indicators were calculated for the three models: (1) Coefficient of determination :

[0147]

[0148] (2) Root mean square error ( ):

[0149]

[0150] (3) Akaike Information Content (AIC):

[0151]

[0152] in, For the sample size, The number of seepage pressure parameters in the model. Mean square error, For the first The actual observed osmotic pressure of each sample. This represents the average of the actual observed seepage pressure values. For the first The osmotic pressure model prediction for each sample.

[0153] (4) Residual diagnostic indicators:

[0154] Normality test: The Shapiro-Wilk test was used to obtain the p-value;

[0155] Heteroscedasticity detection: Calculate the correlation coefficient between the residuals and the predicted values;

[0156] Autocorrelation detection: Calculate the Durbin-Watson statistic of the residuals.

[0157] S23. Calculation and Objective Comparison of Multi-Dimensional Evaluation Indicators

[0158] First, the optimal model is identified based on the RMSE metric:

[0159]

[0160] Simultaneously, cross-validation based on AIC is performed to ensure that the selected model achieves an optimal balance between fitting accuracy and complexity. Through automated decision-making logic, adaptive matching is achieved for different degrees of nonlinear characteristics of dam types, ensuring optimal adaptation of the model to specific engineering features.

[0161] At this point, by constructing a model family covering different degrees of nonlinearity based on accurate feature recognition, a scientific mapping mechanism from engineering features to model complexity can be achieved. Traditional methods often pre-fix a certain model (such as using only linear regression), and use the same modeling method regardless of the actual engineering features, resulting in insufficient fitting of strongly nonlinear processes and overcomplication of simple linear processes. Therefore, a three-model parallel framework is specifically used here to form a complete coverage from linear to nonlinear and from ideal to robust: the linear model corresponds to the ideal working condition with high correlation and weak interference, the second-order polynomial corresponds to the complex working condition with moderate nonlinearity, and Huber regression corresponds to the actual working condition with strong interference and frequent anomalies.

[0162] Simultaneously, this step also pre-diagnoses data quality and nonlinearity by combining the correlation coefficient and delay features extracted through cross-correlation function analysis. Then, it cross-constrains multiple specified models through multi-dimensional evaluation indicators, and finally automatically identifies the optimal model by objective decision rules, which can realize a closed-loop decision-making process of "feature diagnosis → model matching → complexity adaptation".

[0163] In summary, this step constructs a three-tiered, progressive model system: a linear model to capture the main process, a second-order polynomial model to capture nonlinear changes, and Huber robust regression to handle extreme disturbances. This allows for the cross-constraint of multi-dimensional evaluation indicators to address the nonlinear characteristics of different engineering scenarios—not just RMSE and R². 2 To assess the accuracy of the fit, a complexity penalty is introduced through AIC (to avoid the overfitting trap), and the validity of the model assumptions is verified through residual diagnostic tests (normality, heteroscedasticity, autocorrelation). Ultimately, this ensures that the selected model can accurately describe the data and has reasonable physical interpretation capabilities, rather than being a simple statistical fit.

[0164] Step S3: Dual-path reverse propagation mechanism for physical parameters.

[0165] Design two independent and undependent physical reverse inference routes, starting from time characteristics and regression characteristics respectively, to independently infer the penetration coefficient:

[0166] Route A (inverse calculation based on diffusion equations with cross-correlation time delay):

[0167] Delay parameters obtained from cross-correlation analysis Using the heat transfer and diffusion theory of porous media, the propagation of reservoir water level changes in the soil follows a one-dimensional unsteady flow diffusion equation, and the permeability coefficient is derived from the formula.

[0168] In the specific implementation process, the specific operations include:

[0169] Delay parameters identified based on cross-correlation analysis Establish the inverse formula:

[0170]

[0171] in Porosity (dimensionless). The hole depth is in meters (m). The delay time is in seconds. The longer the delay time, the slower the water propagates in the porous medium, and the smaller the corresponding permeability coefficient.

[0172] Route B (Hydraulic backpropagation based on regression characteristics):

[0173] Since a unit change in reservoir water head should produce a unit pressure response of the same magnitude at the seepage point without attenuation, an attenuation relationship between the theoretical slope and the actual slope is established, based on the regression slope of the optimal model. The attenuation factor, combined with the theoretical value, is used to deduce the second set of estimated permeability coefficients using the modified diffusion equation.

[0174] In the specific implementation process, the specific operations include:

[0175] Let the theoretical slope be:

[0176]

[0177] Calculate the attenuation factor: ,in This represents the actual regression slope, where the attenuation factor reflects the degree of attenuation of the osmotic pressure response relative to the theoretical value.

[0178] The diffusion equation is modified based on the attenuation factor:

[0179]

[0180] in The formula incorporates the influence of osmotic pressure attenuation characteristics, where the attenuation coefficient is the attenuation coefficient.

[0181] This approach can quantify the bias of regression features into a decay factor, which is then incorporated into the physical model through a correction coefficient, thus achieving a mapping from statistical observations to physical parameters.

[0182] Then, credibility-weighted fusion and parameter comprehensive estimation are performed. By evaluating the credibility and weighting the penetration coefficient estimates from the two independent back-inference methods, and comprehensively considering the theoretical basis, parameter accuracy and physical rationality of each route, a comprehensive penetration coefficient estimate can be calculated to obtain a more robust and reliable back-inference result.

[0183] In the specific implementation process, the specific operations include:

[0184] Because the two reverse-engineering routes are based on different physical foundations and data characteristics, they respectively yield... and Two sets of independent estimates are used. To improve the robustness and reliability of the back-calculation results, a credibility-weighted fusion strategy is adopted. Specifically, the credibility weight of each route is first evaluated, based on the correlation coefficient and decay factor. To determine the rationality and calculate the weights, we have:

[0185]

[0186]

[0187] in This is the maximum correlation coefficient in cross-correlation analysis. The closer this index is to 1, the higher the credibility of the delayed route.

[0188] Next, calculate the overall penetration coefficient estimate:

[0189]

[0190] Simultaneously calculate the degree of deviation between the two estimates as an indicator of the uncertainty of the inversion result:

[0191]

[0192] when When the percentage is less than 20%, the reliability of the reverse-engineered results is high.

[0193] When 20%≤ When the percentage is less than 50%, the results are of moderate reliability.

[0194] when When the accuracy is ≥50%, the data quality and model assumptions need to be reviewed.

[0195] In summary, by setting up two physical back-inference routes—one based on time-delay diffusion equations and the other based on statistical response hydraulic attenuation—the two routes are completely independent and complementary in terms of physical foundation and data dependence. They can combine a correlation coefficient weighted fusion strategy to integrate multi-source estimates into comprehensive parameters, ultimately forming a complete technical closed loop of "feature diagnosis → adaptive modeling → dual-path back-inference → weighted fusion." Compared with traditional single models or manual experience selection, this fundamentally realizes the transformation from "experience-based judgment" to "data-driven scientific decision-making," making seepage pressure response modeling under different dam types and working conditions adaptive and objective.

[0196] Therefore, this step can overcome the limitations of using a single parameter to reverse the route:

[0197] Route A is based on the time information dimension (the delay parameter τ in step S2, following the physical mechanism of the diffusion equation), while Route B is based on the statistical information dimension (the regression slope decay factor in step S3, reflecting the response decay process of reservoir pressure and seepage pressure). The two routes are completely independent in terms of physical basis and data dependence. This design allows the two K-value estimations to essentially become "observing the same physical process from different perspectives." Specifically, this is achieved by setting the deviation between the two physical back-calculation routes: one based on the time delay diffusion equation and the other based on the statistical response hydraulic decay. This allows for the establishment of a tiered confidence assessment mechanism, ensuring that the inversion results include not only parameter values ​​but also explicit confidence assessments.

[0198] If the results of the two paths are consistent ( If the percentage is less than 20%, then they corroborate each other, have high credibility, and can be directly applied to engineering design;

[0199] When 20%≤ When the percentage is less than 50%, auxiliary verification is required.

[0200] If there is a large deviation ( If the percentage is ≥50%, it indicates a problem with the data quality or model assumptions, requiring reverse tracing to guide data and model improvement.

[0201] This mechanism enables engineering decision-makers to respond differently based on the credibility level, while supporting adaptive optimization of the model based on the inversion effect, forming a continuous improvement closed loop of "inversion-verification-optimization". Compared with the traditional single inversion method, through multi-dimensional feature interaction constraints, adaptive model selection and credibility quantification, it can significantly improve the accuracy and engineering practicality of permeability coefficient inversion, providing a parameter basis and intelligent decision support for dam seepage monitoring, risk assessment and long-term management.

[0202] Meanwhile, based on the weighted fusion strategy of the correlation coefficient in step S2, it is also possible to achieve a complete evaluation closed loop from multi-source features → weighted synthesis → uncertainty quantification.

[0203] The difference between the above reverse engineering process and the traditional single reverse engineering method lies in:

[0204] A. It possesses an inherent self-testing mechanism;

[0205] B. It can quantify the confidence level of the reverse-engineering results;

[0206] C. The robustness of parameter inversion is improved by leveraging the complementarity of multidimensional features.

[0207] Step S4: Automatic diagnosis and classification of soil type.

[0208] The system categorizes soil types and permeability levels by comparing the comprehensively back-calculated permeability coefficient with standard reference ranges. Diagnostic results are visualized and engineering reports are output.

[0209] Ultimately, an engineering information system integrating data management, analysis and processing, and diagnostic decision-making was developed. This system generates multi-dimensional diagnostic charts (such as reservoir water level-seepage pressure fitting effect charts and model diagnostic four-in-one charts) and quantitative diagnostic reports, clearly displaying the inversion process, parameter estimation, uncertainty assessment, and final diagnostic conclusions, providing quantitative and interpretable decision support for engineers.

[0210] In the specific implementation process, the specific operations include:

[0211] S41. Automatic Diagnosis and Classification of Soil Types

[0212] The permeability coefficient derived from the comprehensive back calculation By aligning with the standard reference range, a complete soil property diagnostic classification system was finally established, as shown in the table below:

[0213] Table 1: Soil Property Diagnostic Classification System

[0214]

[0215] S42. Construction of Engineering Information System Platform and Generation of Multi-Dimensional Diagnostic Charts

[0216] Develop an engineering information system platform that integrates data management, spatial visualization, and model calculation (which can be developed using Python + Django or a GIS platform).

[0217] The system includes a database module, a computing engine, and a visualization module. Based on these, the system can automatically generate multi-dimensional diagnostic charts, including: reservoir water level-seepage pressure time series comparison chart, model fitting effect evaluation chart, model diagnosis four-in-one chart (residual distribution, QQ plot, scale-location plot, residual time series plot), seepage pressure spatial distribution cloud map, key parameter sensitivity curves, etc., which can be used to intuitively display the characteristics of monitoring data, model fitting quality, and diagnostic process.

[0218] S43. Quantitative Diagnostic Report Preparation and Engineering Decision Support Output

[0219] Based on the inversion results and diagnostic indicators, a structured quantitative diagnostic report is automatically generated.

[0220] The report comprises five core sections:

[0221] (1) Inversion process documentation: used to show the initial parameter settings, iteration process, and convergence criteria;

[0222] (2) Parameter estimation results: used to output the point estimates of various types of inversion parameters;

[0223] (3) Uncertainty assessment: used to quantify the confidence interval of parameters, the correlation between parameters, and the forecast error range;

[0224] (4) Engineering diagnostic conclusions: used to give quantitative conclusions such as the dam body leakage risk level (low / medium / high) and abnormal seepage pressure location based on diagnostic index thresholds;

[0225] (5) Recommendations and warnings: Used to provide recommendations on follow-up monitoring priorities, maintenance measures or emergency response based on the diagnostic results.

[0226] The final report is presented in PDF or interactive web interface, providing engineering managers with quantitative, interpretable, and traceable decision support.

[0227] The system constructed by this method supports the acquisition and processing of multi-source heterogeneous data such as distributed sensor networks. It can be used to realize the multi-model fusion mechanism of engineering response feature identification and dam type feature adaptation based on cross-correlation delay analysis. At the same time, it proposes a dual-path back-inference system for physical parameters, which integrates statistical observations with the physical model of porous media, thereby constructing an integrated seepage pressure diagnosis system that can both ensure the accuracy of numerical diagnosis and realize the back-inference of dam type-specific engineering parameters.

[0228] In summary, from a technical perspective, this method enables real-time acquisition and unified preprocessing of multi-source heterogeneous engineering data. Through cross-correlation delay analysis, it accurately identifies the temporal relationship between reservoir water level and seepage pressure response. Multiple diagnostic models (linear, polynomial, robust regression, etc.) are constructed in parallel to adapt to the differences in seepage characteristics among different dam types. Based on multi-dimensional indicators (R... 2 Objective comparison and adaptive selection of parameters such as RMSE and AIC are performed, and a dual-path back-inference mechanism for physical parameters is proposed, based on time delay parameters respectively. and regression features (slope) The system uses the diffusion theory of porous media to invert the permeability coefficient K (including the attenuation factor) and performs a credibility-weighted fusion of the two estimates to achieve quantitative diagnosis and classification of soil types. This system not only overcomes the limitations of traditional diagnostic methods in single data processing and general algorithm adaptation, but also achieves a complete mapping transformation from statistical observation to engineering diagnosis through multi-model fusion and physical parameter inversion. This significantly improves the accuracy of permeability characteristic diagnosis under complex conditions and the scientific nature of engineering decisions, providing reliable technical support for the safety assessment and flood control management of reservoirs and dams.

[0229] Furthermore, from the perspective of engineering practicality and applicability, this method, by supporting the collaborative acquisition and unified preprocessing of multi-source heterogeneous data (such as reservoir water level, seepage pressure, time series, etc.) from distributed sensor networks, innovatively proposes a dual-path back-inference mechanism for physical parameters for the first time. This mechanism is based on the cross-correlation time delay parameter τ and the reservoir water level-seepage pressure regression characteristics (slope). The method independently inversely derives the permeability coefficient K using attenuation factors, and combines multi-model adaptive selection and confidence-weighted fusion to achieve a complete mapping transformation from statistical observations to engineering decision parameters. Through a scientific and automated diagnostic classification system, this method achieves accurate quantitative inversion of the permeability coefficient and intelligent diagnosis of soil types under complex working conditions, significantly improving the accuracy, reliability, and engineering application value of permeability pressure diagnosis.

[0230] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0231] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0232] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the fusion model preferred and physical inversion osmotic pressure analysis methods in the above embodiments.

[0233] It is understood that the system provided in the embodiments of the present invention corresponds to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

[0234] This application also provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, communication interface, and memory communicate with each other via the communication bus.

[0235] Memory, used to store computer programs;

[0236] The processor, when executing the program stored in the memory, implements the aforementioned osmotic pressure analysis method of fusion model optimization and physical inversion.

[0237] The communication bus mentioned in the above-mentioned electronic devices can be a standard bus for interconnecting peripheral components or an extended industrial standard structure bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc.

[0238] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0239] The memory may include random access memory or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0240] The processors mentioned above can be general-purpose processors, including central processing units, network processors, etc.; they can also be digital signal processors, application-specific integrated circuits, field-programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0241] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, an optical medium, or a semiconductor medium, etc.

[0242] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0243] Furthermore, it should be noted that if any directional indication (such as up, down, left, right, front, back, etc.) is involved in the embodiments of the present invention, the directional indication is only used to explain the relative positional relationship and movement of each component in a specific posture. If the specific posture changes, the directional indication will also change accordingly.

[0244] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the meaning of "and / or" throughout the text includes three parallel solutions; for example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously. Furthermore, in the embodiments of this invention, "multiple" refers to two or more. Moreover, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

Claims

1. A osmotic pressure analysis method integrating model optimization and physical inversion, characterized in that, include: Step S1. Calculate the correlation coefficient between the water level and seepage pressure of the specified reservoir under different time lags based on cross-correlation function analysis, and identify the differential nonlinear characteristics corresponding to the maximum correlation coefficient; Step S2. For the differentiated nonlinear characteristics of different dam types, construct multiple diagnostic prediction models in parallel, fit and calculate the multi-dimensional evaluation indicators of each model, and verify the statistical significance of the regression relationship through residual diagnosis. Based on the comprehensive comparison results of the multi-dimensional evaluation indicators, establish model selection decision rules and select the optimal diagnostic model according to the priority of the indicators. Step S3. Obtain the optimal delay time based on cross-correlation analysis. The permeability coefficient can be calculated by using the heat transfer and diffusion theory formula of porous media. The permeability coefficient is calculated by back-calculating the regression slope of the optimal model and the theoretical attenuation factor, combined with the modified diffusion equation. Regarding the permeability coefficient , Perform credibility assessment and weighted fusion to calculate the comprehensive penetration coefficient; Step S4. Compare the comprehensively back-calculated permeability coefficient with the standard reference range to classify and diagnose the soil type and permeability grade; In step S1, the process of identifying the factor that maximizes the cross-correlation function formula result is described. Optimal delay time Calculate the optimal delay time for: in, The sampling time interval; The process for identifying and extracting differential nonlinear features in step S1 includes: Based on the optimal delay of identification Time alignment was performed on the original data sequence to align the reservoir water level sequence. Move backward Step, to obtain the aligned data pair sequence, we have: Align reservoir water level : Alignment osmosis : Extract statistical features, including aligned reservoir water levels. Sequence and Alignment Isobaric Pressure Maximum correlation coefficient of the series Initial correlation and optimal delay time This ultimately forms a set of engineering response characteristic parameters. ; In step S3, the permeability coefficient is calculated backwards. The specific operating procedures include: Let the theoretical slope be for: Calculate the attenuation factor for: in, This represents the actual regression slope; Based on the attenuation factor, the diffusion equation is modified to obtain the following modified equation: in, This is the attenuation coefficient.

2. The osmotic pressure analysis method based on the fusion model optimization and physical inversion as described in claim 1, characterized in that, The specific process for calculating the correlation coefficient based on the cross-correlation function analysis in step S1 includes: The cross-correlation function is defined as follows: in, Indicated as time lag The correlation coefficient below, and These are the average reservoir water level and the average seepage pressure, respectively. Indicated as time lag The reservoir water level below Indicated as time lag The osmotic pressure below, The total number of data points. This represents the number of time lag steps.

3. The osmotic pressure analysis method based on the fusion model optimization and physical inversion as described in claim 2, characterized in that, The parallel construction of the diagnostic prediction model in step S2 specifically includes: a. Linear regression model, the calculation formula of which is: in, This is expressed as the slope of the linear regression. and These are represented as the linear regression intercept and the random error term of the model, respectively. The model uses the least squares method to solve for the regression coefficients; b. The second-order polynomial regression model has the following calculation formula: in, , and These are respectively represented as the quadratic regression coefficient, the linear regression coefficient, the regression constant term, and the random error term of the model; c. Robust regression model, the calculation formula of which is: in, Represented as the Huber loss function, Indicated as time lag The data residuals below.

4. The osmotic pressure analysis method based on the fusion model optimization and physical inversion as described in claim 3, characterized in that, The specific operational procedures for calculating multi-dimensional evaluation indicators by fitting the diagnostic prediction model and verifying the statistical significance of the regression relationship through residual diagnosis in step S2 include: For linear regression, second-order polynomial regression, and robust regression models, the following multidimensional evaluation indicators are calculated respectively: Root mean square error : ; Akaike Information Content (AIC): ; in, For the sample size, The number of seepage pressure parameters in the model. Mean square error, For the first The actual observed osmotic pressure of each sample. For the first The osmotic pressure model predicted values ​​for each sample; Subsequently, in establishing the model selection decision rules, an optimal model was identified based on the RMSE metric, and cross-validation was performed using AIC to ensure that the selected model achieved an optimal balance between fitting accuracy and complexity. The optimal model identified based on the RMSE metric is as follows: in, Represented as the first The RMSE metric for each model.

5. The osmotic pressure analysis method based on the fusion model optimization and physical inversion as described in claim 1, characterized in that, In step S3, the permeability coefficient is calculated backwards. The specific operating procedures include: The optimal delay time was identified using the heat transfer and diffusion theory of porous media and based on cross-correlation analysis. The inverse formula is established as follows: in, Porosity For the depth of the hole, For the optimal delay time, This is the estimated penetration coefficient derived from the reverse calculation.

6. The osmotic pressure analysis method based on the fusion model optimization and physical inversion as described in claim 1, characterized in that, The specific calculation process for calculating the comprehensive penetration coefficient estimate in step S3 includes: Based on correlation coefficient and attenuation factor To assess the rationality of each route and evaluate its credibility weight, the weight calculation formula is as follows: in, For the first The maximum number of cross-correlation coefficients of the routes, This is the maximum correlation coefficient in cross-correlation analysis. The closer this index is to 1, the higher the credibility of the delayed route. Calculate the overall penetration coefficient valuation for: Simultaneously, the degree of deviation between the two estimates is calculated, which serves as an indicator of the uncertainty of the inversion result. The calculation formula is as follows: when In this case, the results obtained by reverse engineering are highly reliable; when At that time, the results were of moderate reliability; when At this time, it is necessary to review the data quality and model assumptions.

7. A computer-readable storage medium, characterized in that, The system stores a computer program that, when executed by a processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 6.

8. A computer device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 6.

Images