A soil layer quantification method, system, device and medium based on a machine learning algorithm fusion prediction model

Abstract

The application provides a soil layer quantitative stratification method, system, device and medium based on a machine learning algorithm fusion prediction model, the method comprises the following steps: S1, collection and arrangement of hole pressure static sounding data and soil layer classification information; S2, establishing a basic model by using SVM, RF and XGBoost; S3, optimizing the hyperparameters of the basic model by using a Bayesian optimization algorithm; S4, establishing a fusion prediction model of three basic models; S5, training the fusion prediction model; S6, predicting the soil layer type of another site by using the trained fusion prediction model; S7, determining the division accuracy, processing the stratification results, and finally obtaining the soil layer quantitative stratification results. The application can quickly establish the quantitative relationship between the hole pressure static sounding data and the soil layer category, realize rapid and accurate classification, quantify the uncertainty of soil layer classification, and avoid the subjectivity of the traditional soil layer classification method and the limitations of a single algorithm.

Description

Technical Field

[0001] This invention relates to the field of geotechnical engineering, and in particular to a method, system, device, and medium for quantitative stratification of soil layers based on a machine learning algorithm-fused prediction model. Background Technology

[0002] Compared to other in-situ testing methods, static cone penetration testing (CPPT) is generally considered to be fast, continuous, and economical, and has become a representative method for determining soil stratification information in geotechnical engineering. Currently, surveyors mainly rely on experience and relevant formulas to classify soil layers in processing CPPT data, lacking research on the theory of CPPT data interpretation. This results in low efficiency and accuracy in layer classification, affecting the progress of actual engineering surveys and subsequent design calculations.

[0003] Chinese patent CN103898890A discloses a method for soil quantification and stratification based on double-bridge static cone penetration test (CPPT) data using a backpropagation (BP) neural network. After continuous preprocessing of the CPPT data, a BP neural network prediction model for soil quantification and stratification is established. This method interprets the CPPT data using a BP neural network, providing some guidance for the application of CPPT data in soil layer division. However, its data preprocessing method is too simplistic, offering almost no improvement in prediction accuracy compared to the original dataset. Furthermore, the BP neural network structure is simple, resulting in shallow data mining depth. In terms of output, the soil type labels provided by this method are too absolute, offering insufficient information and failing to provide adequate guidance for engineering designers.

[0004] A domestic paper with DOI 10.13349 / j.cnki.jdxbn.20170527.001 proposed a soil interface identification method based on probabilistic neural networks for pore pressure static cone penetration testing. It has a good effect on identifying soil interlayers, but there are still problems such as the input parameters not fully reflecting the characteristics of static cone penetration data and the model output results not being intuitive.

[0005] Therefore, there is an urgent need for a reliable, high-precision, and user-friendly method for quantifying and stratifying soil layers, so as to provide an effective basis for various calculations in geotechnical engineering investigation and design. Summary of the Invention

[0006] The first objective of this invention is to provide a method for quantitative stratification of soil layers based on a machine learning algorithm and a fusion prediction model, using existing pore pressure static cone penetration test data and corresponding soil layer classification information.

[0007] Therefore, the above-mentioned objective of the present invention is achieved through the following technical solution:

[0008] A method for quantitative soil stratification based on a prediction model fused with machine learning algorithms includes the following steps:

[0009] S1. Collection and organization of pore pressure static cone penetration data and soil layer classification information;

[0010] Collect and organize geotechnical engineering investigation reports and geotechnical test results, and organize the penetration data along the depth of the pore pressure static penetration test holes and soil layer classification information;

[0011] S2. Establish a basic model using machine learning algorithms SVM, RF, and XGBoost;

[0012] Machine learning algorithms SVM, RF, and XGBoost were used to establish basic prediction models with pore pressure static cone penetration data as input and soil layer type as output. The output results are the probability of each soil layer corresponding to each soil type.

[0013] S3. Use Bayesian optimization algorithm to optimize the hyperparameters of the basic model;

[0014] The hyperparameters of the SVM, RF, and XGBoost basic models used for the quantitative stratification of soil layers by pore pressure static cone penetration test were optimized using Bayesian optimization algorithms.

[0015] S4. Establish a fusion prediction model of the three basic models using a stacking approach;

[0016] A linear meta-model is established, and the outputs of the three basic models are fused together using a stacking method to form a machine learning algorithm fusion prediction model.

[0017] S5. Train the prediction model by fusing machine learning algorithms;

[0018] The collected static cone penetration test data and soil layer classification information are input into a machine learning algorithm to fuse and predict the model, and the model is trained.

[0019] S6. Use the trained machine learning algorithm to fuse the prediction model to predict the soil type of another site.

[0020] The trained machine learning algorithm is used to fuse the prediction model to predict the soil type of another site, output the probability of each soil type corresponding to the soil layer, and take the soil type corresponding to the highest probability value as the predicted soil type of the soil layer.

[0021] S7. Determine the stratification accuracy, perform stratification processing on the stratification results, and finally obtain the quantitative stratification results of the soil layers.

[0022] Based on the required precision for the project, the prediction results are processed by layering to obtain the final quantitative stratification results of the soil layers.

[0023] While adopting the above technical solutions, the present invention may also adopt or combine the following technical solutions:

[0024] As a preferred embodiment of the present invention: in step S1, the pore pressure static penetration data includes: cone tip resistance q c Side wall friction f s , Pore water pressure u2.

[0025] As a preferred technical solution of the present invention: in step S1, the soil layer classification information includes: silt and silty soil, silty clay, silt, and silty sand, and respectively assign corresponding soil type label values ​​of 0, 1, 2, and 3.

[0026] As a preferred technical solution of the present invention: In step S3, the three basic models SVM, RF, and XGBoost respectively use a set of optimal hyperparameters obtained by Bayesian optimization, as shown below:

[0027]

[0028] As a preferred technical solution of the present invention: In step S4, the specific process of establishing the linear meta-model is as follows: calculate the weighted average of the probability of each soil type corresponding to each soil layer in the results of each basic model, and the weight is determined by the 10-fold cross-validation results of each basic model on the training set of the basic model.

[0029] As a preferred technical solution of the present invention: In step S5, 80% of the dataset used to train the model is randomly divided for training the basic model, and the remaining 20% ​​is used for testing the meta-model. The meta-model is required to have a better prediction effect than all the basic models before it can be used for the prediction of soil layer types in another site.

[0030] As a preferred technical solution of the present invention: In step S7, the specific process of the layering process is as follows: merge the soil layers of the same type with a total thickness less than the division accuracy into the upper layer.

[0031] The second objective of this invention is to provide a soil layer quantification and stratification system.

[0032] Therefore, the above-mentioned objective of the present invention is achieved through the following technical solution:

[0033] A soil layer quantitative stratification system includes the following modules:

[0034] The data collection module is used to collect and organize geotechnical engineering investigation reports, geotechnical test results, and organize the penetration data and soil layer classification information along the depth of the pore pressure static penetration test borehole.

[0035] The basic model building module is used to build basic prediction models using machine learning algorithms SVM, RF, and XGBoost, respectively, with the pore pressure static cone penetration data obtained by the data collection module as input and the soil layer category obtained by the data collection module as output. The output result is the probability of each soil layer corresponding to each soil type.

[0036] The hyperparameter optimization module is used to optimize the hyperparameters of the SVM, RF, and XGBoost basic models obtained by the basic model building module using a Bayesian optimization algorithm.

[0037] The fusion prediction model building module is used to build a linear meta-model. It integrates the outputs of the three basic models obtained by the basic model building module in a stacking manner to form a machine learning algorithm fusion prediction model.

[0038] The fusion prediction model training module is used to input the pore pressure static cone penetration data and soil layer classification information obtained by the data collection module into the machine learning algorithm fusion prediction model obtained by the fusion prediction model establishment module to train the model.

[0039] The prediction module is used to use the machine learning algorithm trained by the fusion prediction model training module to predict the soil type of another site, output the probability of each soil type corresponding to the soil layer, and take the soil type corresponding to the maximum probability value as the predicted soil type of the soil layer.

[0040] The stratification module is used to determine the stratification accuracy, perform stratification processing on the stratification results obtained by the prediction module, and finally obtain the soil layer quantitative stratification results.

[0041] A third objective of this invention is to provide an electronic device comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus, characterized in that:

[0042] The memory is used to store computer programs;

[0043] A processor is configured to execute a computer program stored in memory to implement the steps of the soil layer quantification and stratification method based on a machine learning algorithm fusion prediction model as described above.

[0044] Another objective of this invention is to provide a non-transitory readable storage medium storing a program that, when executed by a processor, implements the steps of the soil layer quantitative stratification method based on a machine learning algorithm fusion prediction model as described above.

[0045] This invention provides a method, system, device, and medium for soil layer quantitative stratification based on a machine learning algorithm-fused prediction model, which has the following beneficial effects:

[0046] 1) The soil layer quantitative stratification method based on a fusion prediction model using machine learning algorithms provided in this invention can leverage the advantages of multiple machine learning algorithms to quickly establish a quantitative relationship between pore pressure static cone penetration test data and soil layer categories, achieving rapid and accurate classification and quantifying the uncertainty of soil layer classification. Simultaneously, it avoids the subjectivity and limitations of single algorithms in traditional soil layer classification methods, improving the accuracy and reliability of classification. This invention also avoids the subjectivity and empirical nature of traditional soil layer classification methods, providing a scientific, objective, and intelligent solution to the soil layer quantitative stratification problem. This invention can, to a certain extent, reduce drilling work in engineering exploration, save exploration costs, and improve exploration efficiency.

[0047] 2) The soil layer quantification and stratification method based on machine learning algorithm fusion prediction model provided by this invention is not limited to soil layer classification. It is also applicable to the prediction and analysis of other geotechnical parameters (such as permeability coefficient and undrained shear strength) with spatial variability and correlation characteristics, demonstrating strong versatility. Furthermore, this invention allows for flexible selection and adjustment of machine learning algorithms and fusion strategies according to different geotechnical engineering problems and data characteristics, exhibiting strong adaptability and scalability. Attached Figure Description

[0048] Figure 1 The flowchart shows the soil layer quantitative stratification method based on machine learning algorithm fusion prediction model provided by the present invention.

[0049] Figure 2 This is a schematic diagram of the Bayesian optimization algorithm used for hyperparameter optimization of the basic model in machine learning algorithms.

[0050] Figure 3 A simplified diagram illustrating the structure of a prediction model that integrates machine learning algorithms.

[0051] Figure 4 The static cone penetration test curve and soil columnar section of JT33 borehole pressure test in Lianyungang, Jiangsu Province (with a division accuracy of 0.5m). Detailed Implementation

[0052] The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

[0053] A method for quantitative soil stratification based on a prediction model fused with machine learning algorithms includes the following steps:

[0054] S1. Collection and organization of pore pressure static cone penetration data and soil layer classification information;

[0055] We collected and organized geotechnical engineering investigation reports and geotechnical test results from Rudong, Jiangsu Province, and compiled the penetration data and soil layer classification information along the depth of the pore pressure static penetration test boreholes, totaling 800 sets. These included 200 sets each of silt and silty soil, silty clay, silt, and silty sand, corresponding to soil type label values ​​of 0, 1, 2, and 3, respectively. Examples of training samples are shown in Table 1 (only 80 sets of the 800 sets of data are displayed).

[0056] Table 1. Examples of XGBoost model training samples for quantified stratification of soil layers by pore pressure static cone penetration test.

[0057]

[0058]

[0059] S2. Establish a basic model using machine learning algorithms SVM, RF, and XGBoost;

[0060] Using Python, machine learning algorithms SVM, RF, and XGBoost were employed to establish data based on pore pressure static cone penetration test data (cone tip resistance q). c Side wall friction f s The basic prediction model uses pore water pressure (u2) as input and soil type label value as output. The SVM and RF models use the predict_proba() method for prediction, while the XGBoost model takes multi:softprob as the objective function. Finally, the output results of the three basic prediction models are the probabilities of each soil type corresponding to the soil layer.

[0061] S3. Use Bayesian optimization algorithm to optimize the hyperparameters of the basic model;

[0062] Using Python, the hyperparameters of the basic SVM, RF, and XGBoost models used for quantification of soil stratification by pore pressure static cone penetration testing were optimized using the Bayesian optimization algorithm. The cross-validation function in the Bayesian optimization used accuracy as the evaluation metric, with a 10-fold transition, and the Bayesian optimization iterations were performed 500 times. The optimal hyperparameters obtained by Bayesian optimization for each of the three basic models (SVM, RF, and XGBoost) are shown in Table 2.

[0063] Table 2

[0064]

[0065] S4. Establish a fusion prediction model of the three basic models using a stacking approach;

[0066] A linear meta-model is established, and the outputs of the three basic models are fused using a stacking method to form a machine learning algorithm fusion prediction model. In the dataset used for training the model (as shown in Table 1), 80% is randomly allocated for training the basic models, and the remaining 20% ​​is used for testing the meta-model. The random state of `train_test_split` is set to 0. The specific operation of the linear meta-model is to calculate the weighted average of the probabilities of each soil layer corresponding to each soil type in the results of each basic model. The weights are determined by the 10-fold cross-validation results of each basic model on the basic model training set. Accuracy is used as the evaluation metric. The 10-fold cross-validation results and model weights are shown in Table 3. An example of the meta-model implementation is shown in Table 4.

[0067] Table 3

[0068]

[0069] Table 4

[0070]

[0071] S5. Train and test the machine learning algorithm fusion prediction model;

[0072] The collected static cone penetration test data and soil layer classification information were input into a machine learning algorithm to fuse a prediction model. The model was then trained and tested. Table 1 shows the training set for the basic model and the test set for the meta-model. Table 2 shows the hyperparameters of the basic model, Table 3 shows the weights of each model, and Table 4 shows an example of the meta-model implementation. For the meta-model test set, the meta-model's prediction performance must outperform all basic models before it can be used to predict soil types at another site. Table 5 shows the prediction accuracy of each basic model and the fused model on the meta-model test set.

[0073] Table 5

[0074]

[0075] S6. Use the trained machine learning algorithm to fuse the prediction model to predict the soil type of another site.

[0076] The static pressure penetration test data of the JT33 static penetration test borehole in the exploration project of the scientific research project on key technologies for reinforcement of deep soft soil foundation of heavy-load ore stockpile in Lianyungang area, Jiangsu Province, were used to predict the soil type using a trained prediction model. The output of the prediction model is the probability value of each soil type corresponding to the soil layer. The soil type corresponding to the highest probability value is taken as the predicted soil type of the soil layer. The total test depth of the penetration test borehole is 35.0m. The static pressure penetration test data and prediction results are shown in Table 6.

[0077] Table 6. JT33 Pore Pressure Static Probe Data and Prediction Results for Lianyungang Ore Stockpile

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084]

[0085] S6. Determine the stratification accuracy, perform stratification processing on the stratification results, and finally obtain the quantitative stratification results of the soil layers.

[0086] Based on engineering requirements, the stratification accuracy was determined to be 0.5m. The predicted results were then processed by layering to obtain the final quantitative stratification results of the soil layers in the JT33 static penetration test borehole at the Lianyungang ore stockpile, as shown in Table 7. Table 7 also includes the JT33 pore pressure static penetration test curve and soil layer columnar section (0.5m stratification accuracy). Figure 3 As shown.

[0087] Table 7. Quantitative stratification results of soil layers in static borehole JT33 at Lianyungang ore stockpile (0.5m stratification accuracy)

[0088]

[0089] The present invention also provides a soil layer quantitative stratification system, comprising the following modules:

[0090] The data collection module is used to collect and organize geotechnical engineering investigation reports, geotechnical test results, and organize the penetration data and soil layer classification information along the depth of the pore pressure static penetration test borehole.

[0091] The basic model building module is used to build basic prediction models using machine learning algorithms SVM, RF, and XGBoost, respectively, with the pore pressure static cone penetration data obtained by the data collection module as input and the soil layer category obtained by the data collection module as output. The output result is the probability of each soil layer corresponding to each soil type.

[0092] The hyperparameter optimization module is used to optimize the hyperparameters of the SVM, RF, and XGBoost basic models obtained by the basic model building module using a Bayesian optimization algorithm.

[0093] The fusion prediction model building module is used to build a linear meta-model. It integrates the outputs of the three basic models obtained by the basic model building module in a stacking manner to form a machine learning algorithm fusion prediction model.

[0094] The fusion prediction model training module is used to input the pore pressure static cone penetration data and soil layer classification information obtained by the data collection module into the machine learning algorithm fusion prediction model obtained by the fusion prediction model establishment module to train the model.

[0095] The prediction module is used to use the machine learning algorithm trained by the fusion prediction model training module to predict the soil type of another site, output the probability of each soil type corresponding to the soil layer, and take the soil type corresponding to the maximum probability value as the predicted soil type of the soil layer.

[0096] The stratification module is used to determine the stratification accuracy, perform stratification processing on the stratification results obtained by the prediction module, and finally obtain the soil layer quantitative stratification results.

[0097] The present invention also provides an electronic device, which includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus, characterized in that:

[0098] The memory is used to store computer programs;

[0099] A processor is configured to execute a computer program stored in memory to implement the steps of the soil layer quantification and stratification method based on a machine learning algorithm fusion prediction model as described above.

[0100] The present invention also provides a non-transitory readable storage medium storing a program, which, when executed by a processor, implements the steps of the soil layer quantitative stratification method based on a machine learning algorithm fusion prediction model as described above.

[0101] The aforementioned non-transitory readable storage medium can be any available medium or data storage device that can be accessed by the processor in the electronic device, including but not limited to magnetic storage such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), optical storage such as CDs, DVDs, BDs, HVDs, etc., and semiconductor storage such as ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs), etc.

[0102] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0103] These program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0104] These program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0105] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0106] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for quantitative soil stratification based on a prediction model fused with machine learning algorithms, characterized in that: The method includes the following steps: S1. Collection and organization of pore pressure static cone penetration data and soil layer classification information; Collect and organize geotechnical engineering investigation reports and geotechnical test results, and organize the penetration data along the depth of the pore pressure static penetration test holes and soil layer classification information; S2. Establish a basic model using machine learning algorithms SVM, RF, and XGBoost; Machine learning algorithms SVM, RF, and XGBoost were used to establish basic prediction models with pore pressure static cone penetration data as input and soil layer type as output. The output results are the probability of each soil layer corresponding to each soil type. S3. Use Bayesian optimization algorithm to optimize the hyperparameters of the basic model; The hyperparameters of the SVM, RF, and XGBoost basic models used for the quantitative stratification of soil layers by pore pressure static cone penetration test were optimized using Bayesian optimization algorithms. S4. Establish a fusion prediction model of the three basic models using a stacking approach; A linear meta-model is established, and the outputs of the three basic models are fused together using a stacking method to form a machine learning algorithm fusion prediction model. S5. Train the prediction model by fusing machine learning algorithms; The collected static cone penetration test data and soil layer classification information are input into a machine learning algorithm to fuse and predict the model, and the model is trained. S6. Use the trained machine learning algorithm to fuse the prediction model to predict the soil type of another site. The trained machine learning algorithm is used to fuse the prediction model to predict the soil type of another site, output the probability of each soil type corresponding to the soil layer, and take the soil type corresponding to the highest probability value as the predicted soil type of the soil layer. S7. Determine the stratification accuracy, perform stratification processing on the stratification results, and finally obtain the quantitative stratification results of the soil layers. Based on the required precision for the project, the prediction results are processed by layering to obtain the final quantitative stratification results of the soil layers.

2. The method according to claim 1, characterized in that: In step S1, the pore pressure static cone penetration test data includes: cone tip resistance, sidewall friction resistance, and pore water pressure.

3. The method according to claim 1, characterized in that: In step S1, the soil classification information includes: silt and silty soil, silty clay, silt, and silty sand, which are assigned corresponding soil type label values ​​of 0, 1, 2, and 3, respectively.

4. The method according to claim 1, characterized in that: In step S4, the specific process of establishing the linear meta-model is as follows: calculate the weighted average of the probabilities of each soil type corresponding to each soil layer in the results of each basic model. The weights are determined by the 10-fold cross-validation results of each basic model on the training set of the basic model.

5. The method according to claim 1, characterized in that: In step S5, the dataset used to train the model is randomly divided into two parts: 80% is used to train the base model, and the remaining 20% ​​is used to test the meta-model. The meta-model is required to outperform all the base models before it can be used to predict soil types in another site.

6. The method according to claim 1, characterized in that: In step S7, the specific process of layer merging is as follows: merge similar soil layers with a total thickness less than the division accuracy into the next layer.

7. A soil layer quantitative stratification system, characterized in that: The soil layer quantitative stratification system is based on the soil layer quantitative stratification method based on machine learning algorithm fusion prediction model as described in claim 1, and includes the following modules: The data collection module is used to collect and organize geotechnical engineering investigation reports, geotechnical test results, and organize the penetration data and soil layer classification information along the depth of the pore pressure static penetration test borehole. The basic model building module is used to build basic prediction models using machine learning algorithms SVM, RF, and XGBoost, respectively, with the pore pressure static cone penetration data obtained by the data collection module as input and the soil layer category obtained by the data collection module as output. The output result is the probability of each soil layer corresponding to each soil type. The hyperparameter optimization module is used to optimize the hyperparameters of the SVM, RF, and XGBoost basic models obtained by the basic model building module using a Bayesian optimization algorithm. The fusion prediction model building module is used to build a linear meta-model. It integrates the outputs of the three basic models obtained by the basic model building module in a stacking manner to form a machine learning algorithm fusion prediction model. The fusion prediction model training module is used to input the pore pressure static cone penetration data and soil layer classification information obtained by the data collection module into the machine learning algorithm fusion prediction model obtained by the fusion prediction model establishment module to train the model. The prediction module is used to use the machine learning algorithm trained by the fusion prediction model training module to predict the soil type of another site, output the probability of each soil type corresponding to the soil layer, and take the soil type corresponding to the maximum probability value as the predicted soil type of the soil layer. The stratification module is used to determine the stratification accuracy, perform stratification processing on the stratification results obtained by the prediction module, and finally obtain the soil layer quantitative stratification results.

8. An electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus, characterized in that: The memory is used to store computer programs; A processor, the processor being configured to execute a computer program stored in a memory to implement the steps of the soil layer quantitative stratification method based on a machine learning algorithm fusion prediction model as described in any one of claims 1-6.

9. A non-transitory readable storage medium, characterized in that: The non-transient readable storage medium stores a program, which, when executed by a processor, implements the steps of the soil layer quantitative stratification method based on a machine learning algorithm fusion prediction model as described in any one of claims 1-6.

Images