A method for evaluating the potential of resource utilization of dredged spoil based on hierarchical characteristics

Abstract

The present invention provides a method for evaluating the potential of resource utilization of dredged spoil based on hierarchical characteristics, which relates to the technical field of dredged spoil analysis. The method includes: obtaining multiple groups of historical dredging analysis data on dredged spoil and constructing a sampling characteristic parameter matrix corresponding to the sampling data of each group of dredged spoil; constructing multiple combinations of physical and chemical characteristic parameters and conducting a coupling effect analysis to construct a first coupling effect matrix corresponding to the sampling data of each group of dredged spoil; performing feature sparsity analysis and feature stability analysis on each type of coupling characteristic parameter in the multiple first coupling effect matrices to generate a second coupling effect matrix corresponding to each first coupling effect matrix; constructing a sample data set and training to obtain a dredged spoil resource evaluation model, and then conducting a dredged spoil resource evaluation based on the dredged spoil resource evaluation model. The present invention realizes the improvement of the accuracy and comprehensiveness of the dredged spoil resource evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of dredged spoil analysis, and particularly to a method for evaluating the resource utilization potential of dredged spoil based on stratification characteristics. Background Art

[0002] With the continuous progress of projects such as water conservancy projects, port construction, and waterway dredging, a large amount of dredged spoil is generated during dredging operations. Dredged spoil is often regarded as waste when untreated, and random dumping may cause environmental pollution and affect the water ecosystem. However, dredged spoil contains rich minerals, organic matter, and other utilizable resources. Through scientific and reasonable evaluation and utilization, it can be transformed into road base materials, building fillers, soil conditioners, etc., thus realizing resource utilization and promoting the recycling of dredged spoil. How to scientifically and effectively evaluate the resource utilization potential of dredged spoil is an important topic in current engineering practice and the environmental protection field.

[0003] Some of the existing technologies relatively simply conduct resource evaluation and analysis based on the physical and chemical characteristics of dredged spoil, ignoring the coupling relationship between different characteristics. Due to the complex spatial distribution and stratification characteristics of the components of dredged spoil, there are complex mutual influences between these characteristics. Only conducting a superficial analysis of the characteristic parameters may lead to a large deviation in the evaluation results and cannot comprehensively reveal the resource potential of dredged spoil. Dredged spoil has obvious stratification characteristics, and the physical and chemical characteristics at different depths vary greatly. Simple methods such as overall sampling or independent characteristic analysis are difficult to reflect the distribution law of dredged spoil in the vertical direction, resulting in the inability to accurately evaluate the utilization value of dredged spoil at different depths, thereby reducing the accuracy of the evaluation. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a method for evaluating the resource utilization potential of dredged spoil based on stratification characteristics, which analyzes and mines the data in dredging operations through strategies such as multi-dimensional feature coupling relationship analysis, feature selection and optimization, overcomes the problems of insufficient multi-dimensional feature analysis ability and low model prediction accuracy in the existing technology, and realizes a high-precision evaluation of the resource utilization potential of dredged spoil.

[0005] A method for evaluating the resource utilization potential of dredged spoil based on stratification characteristics provided by the present invention includes:

[0006] Obtain multiple sets of historical dredging analysis data on dredged spoil, where the historical dredging analysis data includes dredged spoil sampling data and dredged resource evaluation data, and the dredged spoil sampling data includes stratified characteristic sampling data at multiple sampling points;

[0007] Extract multiple physical property parameter vectors and chemical parameter vectors from the sampling data of each group of dredged spoil, and construct a sampling characteristic parameter matrix corresponding to the sampling data of each group of dredged spoil;

[0008] Construct multiple physical and chemical characteristic parameter combinations, conduct coupling effect analysis on each physical and chemical characteristic parameter combination, and construct a first coupling effect matrix corresponding to the sampling data of each group of dredged spoil, including calculating the characteristic coupling parameter and coupling effect factor of each physical and chemical characteristic parameter combination according to multiple parameter vectors, determining the coupling characteristic parameter of each physical and chemical characteristic parameter combination according to multiple characteristic coupling parameters and coupling effect factors, and constructing a first coupling effect matrix containing multiple coupling characteristic parameters;

[0009] Conduct characteristic sparsity analysis and characteristic stability analysis on each type of coupling characteristic parameter in multiple first coupling effect matrices, calculate the characteristic sparsity parameter and characteristic stability parameter of each type of coupling characteristic parameter, and conduct characteristic screening on each first coupling effect matrix according to the characteristic sparsity parameter and characteristic stability parameter to generate a second coupling effect matrix corresponding to each first coupling effect matrix;

[0010] Based on the second coupling effect matrix, conduct clustering analysis on multiple groups of dredged spoil sampling data, generate a coupling classification label for each group of dredged spoil sampling data, add multiple coupling classification labels to multiple groups of historical dredging analysis data, construct a sample data set for training the dredged spoil resource evaluation model, and after training the dredged spoil resource evaluation model through the sample data set, conduct dredged spoil resource evaluation based on the dredged spoil resource evaluation model.

[0011] Preferably, conduct coupling effect analysis on each physical and chemical characteristic parameter combination, and construct a first coupling effect matrix corresponding to the sampling data of each group of dredged spoil, including:

[0012] Calculate the physical and chemical characteristic covariance according to the physical property parameter vector and chemical parameter vector corresponding to the physical and chemical characteristic parameter combination, calculate the standard deviations of the physical property parameter vector and chemical parameter vector respectively, and calculate the characteristic coupling parameter of the physical and chemical characteristic parameter combination according to the physical and chemical characteristic covariance, the standard deviations of the physical property parameter vector and chemical parameter vector;

[0013] Calculate the mutual information amount between the physical property parameter vector and chemical parameter vector corresponding to the physical and chemical characteristic parameter combination, calculate the information entropy of the physical property parameter vector and chemical parameter vector respectively, and calculate the coupling effect factor of the physical and chemical characteristic parameter combination according to the mutual information amount, the information entropy of the physical property parameter vector and chemical parameter vector;

[0014] The coupling characteristic parameters of the physicochemical characteristic parameter combination are calculated using the following formula:

[0015]

[0016] Wherein, is the coupling characteristic parameter of the physicochemical characteristic parameter combination composed of the th physical characteristic parameter and the th chemical characteristic parameter, is the coupling effect factor of the physicochemical characteristic parameter combination, is the physicochemical characteristic covariance of the physicochemical characteristic parameter combination, , are the standard deviations of the physical characteristic parameter vector and the chemical parameter characteristic vector corresponding to the physicochemical characteristic parameter combination respectively, is the adjustment coefficient;

[0017] The first coupling effect matrix of each group of dredged spoil sampling data is constructed according to the coupling characteristic parameters corresponding to multiple physicochemical characteristic parameter combinations.

[0018] Preferably, for the calculation of the characteristic sparse parameter and the characteristic stable parameter of the coupling characteristic parameter, it includes:

[0019] The characteristic sparse parameter of each type of coupling characteristic parameter is calculated according to multiple first coupling effect matrices. Among them, for any type of coupling characteristic parameter, the calculation of the characteristic sparse parameter is as follows:

[0020]

[0021] Wherein, is the characteristic sparse parameter of this type of coupling characteristic parameter, is the indicator function, , is the value of this type of coupling characteristic parameter in the th first coupling effect matrix, is the number of the first coupling effect matrices;

[0022] The coupling characteristic vectors of each type of coupling characteristic parameter are extracted according to multiple first coupling effect matrices. The Kendall correlation coefficient between any two types of coupling characteristic parameters is calculated, and the mean value of the Kendall correlation coefficients between each type of coupling characteristic parameter and the remaining multiple types of coupling characteristic parameters is calculated and denoted as the characteristic stable parameter of each type of coupling characteristic parameter.

[0023] Preferably, feature screening is performed on each first coupling effect matrix according to the characteristic sparse parameter and the characteristic stable parameter to generate a second coupling effect matrix corresponding to each first coupling effect matrix, including:

[0024] Calculate the feature screening parameters of each type of coupled feature parameter according to the feature sparsity parameter and the feature stability parameter, where:

[0025]

[0026] In the formula, is the feature screening parameter of the coupled feature parameter, is the feature stability parameter of the coupled feature parameter, and are the weight parameters of the feature sparsity parameter and the feature stability parameter respectively;

[0027] Remove multiple coupled feature parameters with feature screening parameters less than the feature screening threshold in each first coupled effect matrix, and generate a second coupled effect matrix corresponding to each first coupled effect matrix.

[0028] Preferably, perform clustering analysis on multiple groups of dredged spoil sampling data based on the second coupled effect matrix, and generate coupled classification labels for each group of dredged spoil sampling data, including:

[0029] Determine the frequency scores corresponding to multiple types of coupled feature parameters in multiple second coupled effect matrices, determine multiple target coupled feature parameters in multiple second coupled effect matrices based on the frequency scores and the feature screening parameters, and construct a feature reference template based on the multiple target coupled feature parameters;

[0030] Perform feature reconstruction processing on multiple second coupled effect matrices based on the feature reference template, and generate a third coupled effect matrix corresponding to each second coupled effect matrix;

[0031] Perform clustering analysis on multiple groups of dredged spoil sampling data based on multiple third coupled effect matrices, and generate coupled classification labels for each group of dredged spoil sampling data.

[0032] Preferably, it is characterized in that performing feature reconstruction processing on multiple second coupled effect matrices based on the feature reference template includes:

[0033] Match the coupled feature parameters existing in each second coupled effect matrix with the target coupled features in the feature parameter template, retain the coupled feature parameters that match the target coupled features in the feature parameter template, and remove the coupled feature parameters that do not match the target coupled features in the feature parameter template;

[0034] Missing markers are added to the coupling feature parameters that exist in the feature reference template but are missing in the second coupling effect matrix, and the coupling feature parameters with missing markers in each second coupling effect matrix are complemented based on the mean filling method to complete the feature reconstruction of each second coupling effect matrix, generating a third coupling effect matrix corresponding to each second coupling effect matrix.

[0035] Preferably, a sample data set for training the dredging waste resource evaluation model is constructed, and it further includes:

[0036] Based on the dredging resource evaluation data in the historical dredging analysis data, multiple resource evaluation labels are determined, the coupling classification labels of each group of historical dredging analysis data are added to the corresponding dredging waste sampling data, and each group of dredging waste sampling data is associated with the corresponding resource evaluation label and coupling classification label to construct a sample data set;

[0037] Using multiple groups of dredging waste sampling data containing coupling classification labels in the sample data set as the input of the model, and using multiple resource evaluation labels in the sample data set as the training target of the model, a dredging waste resource evaluation model is trained, where the dredging waste resource evaluation model is an LSTM model.

[0038] Preferably, the DBSCAN clustering algorithm is used to perform clustering analysis on multiple groups of dredging waste sampling data based on multiple third coupling effect matrices.

[0039] The present invention has the following beneficial effects:

[0040] By introducing the coupling effect analysis between physical properties and chemical properties, the present invention constructs a coupling feature matrix to reveal the complex interaction relationships between different combinations of physical properties and chemical properties, and through the sparsity and stability analysis of coupling features, screens out the features that are stable and representative in each sample, combines the coupling features in multiple groups of historical data for clustering analysis, generates coupling classification labels and fuses them with the original features to construct a sample data set containing coupling feature labels, and uses the sample data set to train the dredging waste resource evaluation model, enabling the model to learn the coupling relationship between physical properties and chemical properties and its impact on the resource utilization potential, thereby improving the model's understanding ability of different dredging waste samples, enhancing the ability to capture complex feature relationships, realizing high-precision prediction of the resource utilization potential of dredging waste, and effectively improving the accuracy and comprehensiveness of dredging waste resource evaluation. Brief Description of the Drawings

[0041] Figure 1 It is a schematic flowchart of a method for evaluating the resource utilization potential of dredging waste based on hierarchical characteristics disclosed in an embodiment of the present invention. Detailed Embodiments

[0042] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0043] As Figure 1 shown, a method for evaluating the resource utilization potential of dredged spoil based on hierarchical characteristics disclosed in an embodiment of the present invention includes the following steps:

[0044] Step S10: Obtain multiple sets of historical dredging analysis data on dredged spoil. The historical dredging analysis data includes dredged spoil sampling data and dredging resource evaluation data. Extract multiple physical characteristic parameter vectors and chemical parameter characteristic vectors from each set of dredged spoil sampling data, and construct a sampling characteristic parameter matrix corresponding to each set of dredged spoil sampling data.

[0045] In this embodiment, the historical dredging analysis data may be data related to dredged spoil collected during dredging operations, including dredged spoil sampling data and resource evaluation data of dredged spoil. In order to evaluate the resource utilization potential of dredged spoil, hierarchical characteristic sampling data corresponding to multiple sampling points is often collected based on hierarchical sampling technology, which includes physical and chemical characteristic data of dredged spoil at different sampling depths. That is, for example, physical characteristics include particle size distribution, water content, density, etc., and chemical characteristics include organic matter content, heavy metal content, pH value, etc. The resource utilization potential of dredged spoil can be evaluated through dredged spoil sampling data. The dredging resource evaluation data may be the performance of each set of dredged spoil in actual utilization, such as strength, permeability, etc., or the resource utilization potential evaluated according to experimental results, such as the applicability as a building material, the efficacy of a soil conditioner, etc. The historical dredging analysis data contains rich characteristic information, providing a good data basis for the subsequent evaluation of the resource utilization potential of dredged spoil.

[0046] After obtaining multiple sets of historical dredging analysis data on dredged spoil, extract the physical characteristic parameter vectors and chemical characteristic parameter vectors at different sampling depths of each sampling point from each set of dredged spoil sampling data. Each parameter vector represents the value of a characteristic, such as density, water content, heavy metal content, etc., at different sampling points and depths. Arrange these physical and chemical characteristic parameter vectors according to the distribution characteristics of each sampling point and depth, and construct a sampling characteristic parameter matrix corresponding to each set of dredged spoil sampling data.

[0047] Step S20: Construct multiple physical and chemical characteristic parameter combinations, perform coupling effect analysis on each physical and chemical characteristic parameter combination, and construct a first coupling effect matrix corresponding to each set of dredged spoil sampling data.

[0048] In this embodiment, the combination of physicochemical characteristic parameters refers to the pairwise combination of different physical characteristics and chemical characteristics to form a number of characteristic pairs. For example, the combinations can be "particle size distribution - organic matter content", "density - heavy metal content", etc. The construction of the combination of physicochemical characteristic parameters helps to reveal the mutual influence between each characteristic pair, thereby providing a basis for subsequent coupling effect analysis. After determining multiple combinations of physicochemical characteristic parameters, a coupling effect analysis is performed on each combination of physicochemical characteristic parameters, and the degree of mutual influence between each combination of physicochemical characteristic parameters is quantified through data analysis techniques such as correlation analysis and interaction effect analysis. Specifically, the characteristic coupling parameter and the coupling effect factor of each combination of physicochemical characteristic parameters are calculated according to multiple parameter characteristic vectors, and the coupling characteristic parameter of each combination of physicochemical characteristic parameters is determined according to multiple characteristic coupling parameters and coupling effect factors. Then, a first coupling effect matrix corresponding to each group of dredged spoil sampling data is constructed through multiple coupling characteristic parameters representing the coupling relationship between different characteristic pairs. This matrix quantitatively describes the characteristic coupling effect of the dredged spoil, expresses the coupling relationship of each physicochemical characteristic combination in each group of dredged spoil sampling data, and provides a data basis for subsequent characteristic analysis and model training.

[0049] Step S30: Perform feature sparsity analysis and feature stability analysis on each type of coupling characteristic parameter in multiple first coupling effect matrices, and calculate the feature sparsity parameter and the feature stability parameter of each type of coupling characteristic parameter.

[0050] In this embodiment, considering that some coupling characteristics may not show significant coupling relationships in some samples, and the coupling characteristic relationships between different samples may be unstable. For example, in some samples, it shows a strong coupling relationship, while in other samples, it shows a weak coupling relationship. To avoid these factors having a greater impact on the generalization ability and prediction effect of the model, the first coupling effect matrix is further analyzed. Specifically, feature sparsity analysis and feature stability analysis are performed on the coupling characteristic parameters in multiple first coupling effect matrices. Feature sparsity and feature stability are important indicators for measuring feature parameters, used to evaluate the distribution of features in samples and the performance consistency. Among them, feature sparsity is used to measure whether a certain coupling feature widely exists in all samples, and feature stability is used to evaluate the performance consistency of the same coupling feature in different samples. By performing feature sparsity analysis and feature stability analysis on each type of coupling characteristic parameter in multiple first coupling effect matrices, the feature sparsity parameter and the feature stability parameter of each type of coupling characteristic parameter are calculated, which are important indicators for measuring the performance of each coupling feature in the samples and provide a quantitative basis for subsequent feature screening.

[0051] Step S40: Perform feature screening on each first coupling effect matrix according to the feature sparsity parameter and the feature stability parameter to generate a second coupling effect matrix corresponding to each first coupling effect matrix.

[0052] In this embodiment, the purpose of performing feature screening on each first coupling effect matrix is to remove features with excessively high sparsity, that is, features that are missing or invalid in most samples, and features with poor stability, that is, features that show inconsistent performance in different samples, so as to retain a more representative and stable feature set. After feature screening, each first coupling effect matrix will only retain the screened features to generate the corresponding second coupling effect matrix. The second coupling effect matrix is a feature matrix optimized through screening and can better reflect the key features of the resource utilization potential of dredged spoil.

[0053] Step S50: Perform clustering analysis on multiple groups of dredged spoil sampling data based on the second coupling effect matrix to generate a coupling classification label for each group of dredged spoil sampling data, and add multiple coupling classification labels to multiple groups of historical dredging analysis data to construct a sample data set for training the dredged spoil resource evaluation model.

[0054] In this embodiment, based on the previously generated second coupling effect matrix, clustering analysis is performed on all dredged spoil sampling data. The purpose of clustering analysis is to group dredged spoil samples with similar coupling feature parameters to reveal the differences in the resource utilization potential of different types of dredged spoil. According to the clustering results, the coupling classification label of each dredged spoil sample can be determined, such as "high resource utilization potential", "medium resource utilization potential", "low resource utilization potential", etc., or more detailed labels, such as more specific resource utilization potential levels of dredged spoil in different application scenarios. Add these coupling classification labels to multiple groups of historical dredging analysis data to form a complete sample data set containing coupling feature parameters and coupling classification labels.

[0055] Step S60: After training the dredged spoil resource evaluation model through the sample data set, perform dredged spoil resource evaluation based on the dredged spoil resource evaluation model.

[0056] In this embodiment, the sample data set containing historical dredging analysis data and coupling classification labels generated above is used to train the dredged spoil resource evaluation model. The trained dredged spoil resource evaluation model can predict the resource utilization potential of dredged spoil based on the dredged spoil sampling data that needs to be analyzed and input, and realize the automated evaluation of the resource utilization value of dredged spoil. Through the above method, the accuracy of the evaluation of the resource utilization of dredged spoil can be improved.

[0057] It is worth noting that some existing technologies extract the relationship between physical and chemical property data and its actual utilization effect through multiple actual application records in historical data, and construct preliminary resource utilization value criteria. For example, given the sampling and testing data of a pile of dredged waste and its resource utilization results in actual scenarios, a similar rule is constructed: "When the particle size is greater than a certain parameter and the heavy metal content is less than a certain parameter, the sample's utilization value is at a certain level."

[0058] However, this process fails to consider the complex interactions between characteristics. This single or few characteristic analysis method makes it difficult to reveal the coupling effects between different characteristics, resulting in the inability of the trained model for resource assessment to effectively capture the complex relationships between characteristics. The interaction between physical and chemical properties plays an important role in the resource utilization potential of dredged waste. The analysis and modeling of simple and primitive physical or chemical properties ignores the complex coupling relationships that may exist between characteristics. This makes it difficult for the model to explore the deep laws in the data during the learning process, resulting in poor model prediction results.

[0059] The embodiment of the present invention provides a method for evaluating the resource utilization potential of dredged waste based on layered characteristics, which introduces coupling characteristics, that is, constructs a coupling feature matrix based on the coupling effect between physical characteristics and chemical characteristics. These coupling feature matrices can make the complex relationship between characteristics explicit, and reveal the mutual influence between different combinations of physical and chemical characteristics through coupling effect analysis. The model can capture the nonlinear relationship between characteristics and learn the influence of these feature combinations on the resource utilization potential, thereby improving the learning ability of the model.

[0060] For example, by introducing the coupling features of particle size distribution and organic matter content, heavy metal content and pH value, the model can understand the changing trends of these combined characteristics at different depths and sampling points, and establish a stronger association with the actual resource utilization value label. The introduction of this coupling feature enables the model to learn complex relationships in the data at a higher dimension, thereby improving the performance of the model. By introducing feature coupling effects and feature selection strategies, the model can effectively capture the feature differences of dredged waste in different engineering scenarios and different application environments, improving the applicability and versatility of the model to multi-scenario and multi-type dredged waste samples, and can output higher-precision resource assessment results in more application scenarios.

[0061] In step S20, a coupling effect analysis is performed on each combination of physical and chemical characteristic parameters to construct a first coupling effect matrix corresponding to each set of dredged waste sampling data, specifically including:

[0062] Analyze the physical property parameter vector and chemical parameter characteristic vector corresponding to the physicochemical property parameter combination, and calculate the characteristic coupling parameter of the physicochemical property parameter combination.

[0063] In this embodiment, the Pearson correlation coefficient between the physical property parameter vector and the chemical parameter characteristic vector is used as the characteristic coupling parameter of the physicochemical property parameter combination. Specifically, calculate the physicochemical characteristic covariance of the physicochemical property parameter combination according to the physical property parameter vector and the chemical parameter characteristic vector, that is, the covariance between the physical property parameter vector and the chemical parameter characteristic vector, and calculate the standard deviations of the physical property parameter vector and the chemical parameter characteristic vector respectively. Then, calculate the Pearson correlation coefficient between the physical property parameter vector and the chemical parameter characteristic vector according to the physicochemical characteristic covariance, the standard deviations of the physical property parameter vector and the chemical parameter characteristic vector, and obtain the characteristic coupling parameter of the physicochemical property parameter combination.

[0064] Analyze the physical property parameter vector and the chemical parameter characteristic vector corresponding to the physicochemical property parameter combination, and calculate the coupling effect factor of the physicochemical property parameter combination.

[0065] In this embodiment, determine the coupling effect factor of the physicochemical property parameter combination based on the mutual information between the physical property parameter vector and the chemical parameter characteristic vector. Specifically, calculate the mutual information between the physical property parameter vector and the chemical parameter characteristic vector corresponding to the physicochemical property parameter combination, and calculate the information entropies of the physical property parameter vector and the chemical parameter characteristic vector respectively. Then, calculate the coupling effect factor of the physicochemical property parameter combination according to the mutual information, the information entropies of the physical property parameter vector and the chemical parameter characteristic vector. Among them, the coupling effect factor is the ratio of the mutual information to the sum of the information entropies of the physical property parameter vector and the chemical parameter characteristic vector, and the coupling effect factor can measure the interaction intensity and complexity between physical properties and chemical properties.

[0066] Finally, calculate the coupling characteristic parameter of the physicochemical property parameter combination through the characteristic coupling parameter and the coupling effect factor. The coupling characteristic parameter can comprehensively quantify the complex interaction relationship between physical properties and chemical properties. Through the coupling characteristic parameter, the distribution law of the property combination at different sampling points and depths can be better described, and the non-linear relationship between different property pairs and its impact on the potential of resource utilization can be revealed.

[0067] In this embodiment, the following formula is used to calculate the coupling characteristic parameter of the physicochemical property parameter combination:

[0068]

[0069] In the formula, is the The coupling characteristic parameters of the physicochemical characteristic parameter combination composed of the first physical characteristic parameter and the second chemical characteristic parameter, is the coupling effect factor of the physicochemical characteristic parameter combination, is the physicochemical characteristic covariance of the physicochemical characteristic parameter combination, , are the standard deviations of the physical characteristic parameter vector and the chemical parameter characteristic vector corresponding to the physicochemical characteristic parameter combination respectively, is the adjustment coefficient.

[0070] After obtaining the coupling characteristic parameters of each physicochemical characteristic parameter combination through the above calculations, a first coupling effect matrix corresponding to each group of dredged spoil sampling data is constructed based on multiple coupling characteristic parameters. Each element of the first coupling effect matrix represents the coupling effect intensity of one of the physicochemical characteristic combinations on the dredged spoil sampling data, thus comprehensively describing the interaction between the physical and chemical characteristics of each group of dredged spoil sampling data as a whole, revealing the complex relationship between the physical and chemical characteristics in the dredged spoil and its impact on the potential for resource utilization.

[0071] In step 30, the calculation of the characteristic sparse parameter and the characteristic stable parameter of the coupling characteristic parameter specifically includes:

[0072] The characteristic sparse parameter of each type of coupling characteristic parameter is calculated based on multiple first coupling effect matrices, which is used to measure the widespread existence of a certain type of coupling characteristic in all samples. The characteristic sparsity parameter can quantify whether a coupling characteristic exists in most samples. If a certain coupling characteristic is zero or missing in most samples, then this characteristic may not be representative or valuable.

[0073] In this embodiment, for any type of coupling characteristic parameter, the calculation of the characteristic sparse parameter is as follows:

[0074]

[0075] In the formula, is the characteristic sparse parameter of this type of coupling characteristic parameter, is the indicator function, , is the value of this type of coupling characteristic parameter in the th first coupling effect matrix, is the number of the first coupling effect matrices.

[0076] The value of the feature sparsity parameter ranges from 0 to 1. The closer the value is to 1, it indicates that the feature exists in most samples, that is, the sparsity is low, and it is a representative feature. The closer the value is to 0, it means that the feature does not exist in most samples, that is, the sparsity is high, and it may be an invalid or unstable feature. By calculating the feature sparsity parameter, features that are missing or invalid in most samples can be effectively filtered out, reducing the interference of invalid features on the model.

[0077] Further analyze the performance consistency of each type of coupled feature parameter in different samples, that is, different first coupled effect matrices. Measure whether the performance of a certain type of coupled feature in all samples is consistent by calculating the feature stability parameter. Specifically, extract the coupled feature vectors of each type of coupled feature parameter according to multiple first coupled effect matrices, calculate the Kendall correlation coefficient between any two types of coupled feature parameters according to the multiple coupled feature vectors, and calculate the mean of the Kendall correlation coefficients between each type of coupled feature parameter and the remaining multiple types of coupled feature parameters and denote it as the feature stability parameter of each type of coupled feature parameter.

[0078] In this embodiment, the coupled feature vectors of each type of coupled feature extracted from multiple first coupled effect matrices contain the values of the feature in different samples. Then calculate the correlation between any two types of coupled feature parameters. In this implementation, the Kendall correlation coefficient is used to measure the correlation between any two types of coupled feature parameters in different samples. Then average the Kendall correlation coefficients between each type of coupled feature parameter and the remaining types of coupled feature parameters to obtain the feature stability parameter of each type of coupled feature parameter. The feature stability parameter can effectively evaluate the consistency of a certain coupled feature in multiple samples and screen out features that are stable and reliable in different dredged spoil samples. In this way, features that fluctuate greatly and are unstable in different samples can be filtered out.

[0079] In step S40, perform feature screening on each first coupled effect matrix according to the feature sparsity parameter and the feature stability parameter to generate a second coupled effect matrix corresponding to each first coupled effect matrix, specifically including:

[0080] Calculate the feature screening parameter of each type of coupled feature parameter according to the feature sparsity parameter and the feature stability parameter, which is used to measure the effectiveness and performance stability of the feature in all samples.

[0081] Specifically:

[0082]

[0083] In the formula, is the feature screening parameter of the coupled feature parameter, is the feature stability parameter of the coupled feature parameter, indicating the performance consistency of a certain coupled feature in different samples, , They are weight parameters for the feature sparsity parameter and the feature stability parameter respectively, and are used to control the influence weights of the two on the feature screening parameter.

[0084] The feature screening parameter synthesizes the information of the feature sparsity parameter and the feature stability parameter, and is a comprehensive indicator for measuring whether a coupled feature should be retained. The higher its value, the lower the sparsity of the feature (present in most samples) and the better the stability (consistent performance in different samples), indicating a high-quality coupled feature. By setting the threshold of the feature screening parameter, features with excessive sparsity or poor stability can be effectively filtered out, and only representative and stable coupled features are retained. Based on the feature screening threshold, multiple coupled feature parameters with feature screening parameters less than the feature screening threshold in each first coupled effect matrix are removed to generate a second coupled effect matrix corresponding to each first coupled effect matrix. The second coupled effect matrix is more refined than the first coupled effect matrix, only contains key coupled features after screening and optimization, reduces the noise interference caused by feature redundancy, and can be better used for model training.

[0085] In step S50, cluster analysis is performed on multiple groups of dredged spoil sampling data based on the second coupled effect matrix to generate coupled classification labels for each group of dredged spoil sampling data, including:

[0086] Determine the frequency scores corresponding to multiple types of coupled feature parameters in multiple second coupled effect matrices, determine multiple target coupled feature parameters in multiple second coupled effect matrices based on the frequency scores and the feature screening parameter, and construct a feature reference template based on the multiple target coupled feature parameters.

[0087] In this embodiment, considering that in the first coupled effect matrices corresponding to different dredged spoil sampling data, the coupling relationship strengths between different characteristic parameters are inconsistent, resulting in inconsistent retained characteristics in the multiple second coupled effect matrices generated after feature screening. For example, for the same coupled feature parameter, it may be retained in one matrix but missing in another matrix, resulting in feature misalignment and affecting subsequent data clustering.

[0088] In this case, a feature reconstruction strategy based on a feature reference template is designed. Specifically, the frequency of occurrence of each type of coupled feature parameter in all samples is statistically analyzed to determine the corresponding frequency score, which is used to measure the extent to which a certain coupled feature widely exists in all samples. Then, the frequency score and the feature screening parameter are combined to screen out multiple representative target coupled feature parameters. Exemplarily, a feature with a high frequency score and a feature screening parameter greater than a certain threshold is denoted as a target coupled feature parameter. By reasonably setting the corresponding thresholds for both, the selected target coupled feature parameters can not only stably exist in most samples but also exhibit good sparsity and stability. Then, a feature parameter template is constructed based on the selected target coupled feature parameters. This feature parameter template represents the set of high-quality coupled feature parameters in all second coupling effect matrices.

[0089] After performing feature reconstruction processing on multiple second coupling effect matrices based on the feature reference template, a third coupling effect matrix corresponding to each second coupling effect matrix is generated. Subsequently, clustering analysis is performed on multiple groups of dredging spoil sampling data based on the multiple third coupling effect matrices to generate a coupling classification label for each group of dredging spoil sampling data. The purpose of the feature reconstruction processing is to align the features of multiple second coupling effect matrices so that the data dimensions of all samples are consistent with the feature template, ensuring the consistency and integrity of the feature data and facilitating subsequent clustering analysis.

[0090] In this embodiment, the feature reconstruction processing of multiple second coupling effect matrices specifically includes:

[0091] Match the coupled feature parameters existing in each second coupling effect matrix with the target coupled features in the feature parameter template, retain the coupled feature parameters that successfully match the target coupled features in the feature parameter template, and remove the coupled feature parameters that do not successfully match the target coupled features in the feature parameter template to complete the preliminary feature screening step.

[0092] Then, missing marks are made for the coupled feature parameters that exist in the feature reference template but are missing in the second coupling effect matrix, and the coupled feature parameters with missing marks in each second coupling effect matrix are complemented based on the mean filling method. Specifically, for the coupled feature parameters with missing marks, calculate the mean of this coupled feature parameter in multiple second coupling effect matrices to complement the coupled feature parameters with missing marks. Through feature screening and feature complementation, the feature reconstruction of each second coupling effect matrix is completed, and finally, a third coupling effect matrix corresponding to each second coupling effect matrix is generated.

[0093] The process of performing clustering analysis on multiple groups of dredged spoil sampling data based on multiple third coupling effect matrices. In this embodiment, an algorithm based on density is used to perform clustering analysis on multiple groups of dredged spoil sampling data. Exemplarily, after flattening multiple third coupling effect matrices to obtain corresponding one-dimensional vectors, DBSCAN clustering algorithm is used to perform clustering on multiple groups of dredged spoil sampling data. DBSCAN clustering algorithm is suitable for situations where there may be density differences, outliers or abnormal points in the samples, and can automatically identify clustering regions with higher density and regard sample points with lower density as noise points. This method can better handle the complex distribution characteristics and abnormal samples that may exist in the dredged spoil sampling data. Finally, according to the clustering results generated by the DBSCAN clustering algorithm, the coupling classification labels corresponding to each group of dredged spoil sampling data regarding coupling characteristics can be obtained, which contain the coupling associations between physicochemical characteristic parameters. The sample data set constructed by combining the coupling classification labels reveals the potential characteristic relationships in the dredged spoil, fully excavates the complex internal characteristics of the generated dredged spoil samples, and enables the dredged spoil resource evaluation model to better understand and analyze the comprehensive characteristics of the dredged spoil, thereby improving the accuracy and comprehensiveness of the resource evaluation.

[0094] In step S50, for the construction of the sample data set used to train the dredged spoil resource evaluation model, a conventional data set can be first constructed based on multiple groups of historical dredging analysis data. For example, multiple resource evaluation labels can be determined according to the dredging resource evaluation data. Taking the application scenario of building materials as an example, relevant labels such as strength evaluation, durability evaluation, density and volume evaluation can be obtained. Specifically, they can be multiple evaluation levels or specific values set based on empirical knowledge, etc. First, a conventional data set is constructed with the dredged spoil sampling data and multiple resource evaluation labels in multiple groups of historical dredging analysis data, and on this basis, the corresponding coupling classification labels are added to the dredged spoil sampling data in each group of historical dredging analysis data. The coupling classification labels contain the coupling characteristics between physicochemical properties. After associating each group of dredged spoil sampling data with the corresponding resource evaluation labels and coupling classification labels, a sample data set containing rich characteristic information is constructed.

[0095] During the process of training the dredged spoil resource evaluation model with the sample data set, multiple groups of dredged spoil sampling data containing coupling classification labels are used as the input of the model, and multiple resource evaluation labels are used as the training targets of the model. The dredged spoil resource evaluation model that can be used to evaluate the potential of dredged spoil resource utilization is trained, so as to achieve an accurate evaluation of the potential of dredged spoil resource utilization.

[0096] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The parts not described in detail in this specification belong to the prior art well known to those skilled in the art.

Claims

1. A method for evaluating the resource utilization potential of dredged waste based on stratification characteristics, characterized in that: include: Acquire multiple groups of historical dredging analysis data on dredged waste, the historical dredging analysis data including dredged waste sampling data and dredging resource assessment data, the dredged waste sampling data including stratified characteristic sampling data of multiple sampling points; Extract multiple physical characteristic parameter vectors and chemical parameter characteristic vectors from each set of dredged slag sampling data, and construct a sampling characteristic parameter matrix corresponding to each set of dredged slag sampling data; Constructing multiple physical and chemical characteristic parameter combinations, performing coupling effect analysis on each physical and chemical characteristic parameter combination, and constructing a first coupling effect matrix corresponding to each group of dredged waste sampling data, including calculating characteristic coupling parameters and coupling effect factors of each physical and chemical characteristic parameter combination according to multiple parameter characteristic vectors, determining coupling characteristic parameters of each physical and chemical characteristic parameter combination according to the multiple characteristic coupling parameters and coupling effect factors, and constructing a first coupling effect matrix containing multiple coupling characteristic parameters; Performing feature sparsity analysis and feature stability analysis on each type of coupling feature parameters in a plurality of first coupling effect matrices, calculating feature sparsity parameters and feature stability parameters of each type of coupling feature parameters, performing feature screening on each first coupling effect matrix according to the feature sparsity parameters and feature stability parameters, and generating a second coupling effect matrix corresponding to each first coupling effect matrix; Based on the second coupling effect matrix, cluster analysis is performed on multiple groups of dredged waste sampling data to generate coupled classification labels for each group of dredged waste sampling data. Multiple coupled classification labels are added to multiple groups of historical dredging analysis data to construct a sample data set for training a dredged waste resource assessment model. After the dredged waste resource assessment model is obtained through training with the sample data set, dredged waste resource assessment is performed based on the dredged waste resource assessment model.

2. The method for evaluating the resource utilization potential of dredged waste based on stratification characteristics according to claim 1 is characterized in that: The coupling effect analysis is performed on each combination of physical and chemical characteristic parameters to construct the first coupling effect matrix corresponding to each set of dredged waste sampling data, including: The physicochemical characteristic covariance is calculated based on the physical characteristic parameter vector and the chemical parameter characteristic vector corresponding to the physicochemical characteristic parameter combination, the standard deviation of the physical characteristic parameter vector and the chemical parameter characteristic vector are calculated respectively, and the characteristic coupling parameter of the physicochemical characteristic parameter combination is calculated based on the physicochemical characteristic covariance, the standard deviation of the physical characteristic parameter vector and the chemical parameter characteristic vector; Calculate the mutual information between the physical property parameter vector and the chemical parameter characteristic vector corresponding to the physical and chemical property parameter combination, calculate the information entropy of the physical property parameter vector and the chemical parameter characteristic vector respectively, and calculate the coupling effect factor of the physical and chemical property parameter combination according to the mutual information, the information entropy of the physical property parameter vector and the chemical parameter characteristic vector; The coupling characteristic parameters of the physical and chemical characteristic parameter combination are calculated using the following formula: In the formula, For the The physical characteristic parameters and The coupled characteristic parameters of the physical and chemical characteristic parameter combination composed of chemical characteristic parameters, is the coupling effect factor of the physical and chemical characteristic parameter combination, is the physicochemical characteristic covariance of the physicochemical characteristic parameter combination, , are the standard deviations of the physical characteristic parameter vector and chemical parameter characteristic vector corresponding to the physical and chemical characteristic parameter combination, is the adjustment coefficient; The first coupling effect matrix of each group of dredged slag sampling data is constructed according to the coupling characteristic parameters corresponding to the respective combinations of multiple physical and chemical characteristic parameters.

3. The method for evaluating the resource utilization potential of dredged waste based on stratification characteristics according to claim 2 is characterized in that: The calculation of characteristic sparse parameters and characteristic stable parameters of coupled characteristic parameters includes: The characteristic sparse parameters of each type of coupling characteristic parameters are calculated according to the multiple first coupling effect matrices, wherein for any type of coupling characteristic parameters, the characteristic sparse parameters are calculated as follows: In the formula, is the characteristic sparse parameter of this type of coupling characteristic parameter, is the indicator function, , The characteristic parameters of this type of coupling are The value of the first coupling effect matrix, is the number of the first coupling effect matrix; The coupling characteristic vectors of each type of coupling characteristic parameters are extracted according to multiple first coupling effect matrices, the Kendall correlation coefficients between any two types of coupling characteristic parameters are calculated according to the multiple coupling characteristic vectors, and the average value of the Kendall correlation coefficients between each type of coupling characteristic parameters and the remaining multiple types of coupling characteristic parameters is calculated and recorded as the characteristic stability parameter of each type of coupling characteristic parameter.

4. The method for evaluating the resource utilization potential of dredged waste based on stratification characteristics according to claim 3 is characterized in that: Performing feature screening on each first coupling effect matrix according to the feature sparse parameter and the feature stable parameter, generating a second coupling effect matrix corresponding to each first coupling effect matrix, including: The characteristic screening parameters of each type of coupling characteristic parameters are calculated based on the characteristic sparsity parameters and the characteristic stability parameters, where: In the formula, is the feature screening parameter for the coupled feature parameters, is the characteristic stability parameter of the coupled characteristic parameter, , are the weight parameters of feature sparsity parameter and feature stability parameter respectively; Based on the feature screening threshold, multiple coupling feature parameters whose feature screening parameters are smaller than the feature screening threshold in each first coupling effect matrix are removed to generate a second coupling effect matrix corresponding to each first coupling effect matrix.

5. The method for evaluating the resource utilization potential of dredged waste based on stratification characteristics according to claim 4 is characterized in that: Based on the second coupling effect matrix, cluster analysis is performed on multiple groups of dredged slag sampling data to generate coupling classification labels for each group of dredged slag sampling data, including: Determine frequency scores corresponding to multiple types of coupling feature parameters in multiple second coupling effect matrices, determine multiple target coupling feature parameters in the multiple second coupling effect matrices based on the frequency scores and feature screening parameters, and construct a feature reference template based on the multiple target coupling feature parameters; Performing feature reconstruction processing on a plurality of second coupling effect matrices based on a feature reference template to generate a third coupling effect matrix corresponding to each second coupling effect matrix; Based on multiple third coupling effect matrices, cluster analysis is performed on multiple groups of dredged waste sampling data to generate coupling classification labels for each group of dredged waste sampling data.

6. The method for evaluating the resource utilization potential of dredged waste based on stratification characteristics according to claim 5 is characterized in that: Performing feature reconstruction processing on multiple second coupling effect matrices based on the feature reference template includes: Matching the coupling characteristic parameters existing in each second coupling effect matrix with the target coupling characteristics in the characteristic parameter template, retaining the coupling characteristic parameters that successfully match the target coupling characteristics in the characteristic parameter template, and removing the coupling characteristic parameters that fail to match the target coupling characteristics in the characteristic parameter template; The coupling feature parameters that exist in the feature reference template but are missing in the second coupling effect matrix are marked as missing, and the coupling feature parameters with missing marks in each second coupling effect matrix are completed based on the mean filling method to complete the feature reconstruction of each second coupling effect matrix and generate a third coupling effect matrix corresponding to each second coupling effect matrix.

7. The method for evaluating the resource utilization potential of dredged waste based on stratification characteristics according to claim 6 is characterized in that: The sample data set constructed for training the dredging spoil resource assessment model also includes: Based on the dredging resource assessment data in the historical dredging analysis data, multiple resource assessment labels are determined, and the coupling classification labels of each group of historical dredging analysis data are added to the corresponding dredging spoil sampling data, and each group of dredging spoil sampling data is associated with the corresponding resource assessment label and the coupling classification label to construct a sample data set; Taking multiple groups of dredged slag sampling data containing coupled classification labels in the sample data set as the input of the model, and taking multiple resource evaluation labels in the sample data set as the training target of the model, a dredged slag resource evaluation model is trained, wherein the dredged slag resource evaluation model is an LSTM model.

8. The method for evaluating the potential for resource utilization of dredged waste based on stratification characteristics according to claim 7 is characterized in that: The DBSCAN clustering algorithm was used to perform cluster analysis on multiple groups of dredging waste sampling data based on multiple third coupling effect matrices.

Images