A batch rechecking method and device for soil test results based on MATLAB

Abstract

The application provides a batch rechecking method and device for soil test results based on MATLAB, which comprises the following steps: initializing a MATLAB running environment, reading experimental data including soil types in an.xls file and writing the data into a breast.mat program based on a BP neural network; loading the breast.mat program, dividing training data and rechecking data, and saving the data in traind_s and testd_s files after normalization processing; constructing a neural network, reading and training the neural network with the training data, and calculating a calculation variable obtained from the rechecking data by using the neural network; transforming the calculation result, comparing the calculation result with the soil types in the initialized rechecking data, and obtaining a group number that does not match, an accuracy rate of a rechecking group, and an error rate. The application rechecks batch soil test results based on MATLAB, simultaneously calculates data by using a neural network, and simultaneously improves the speed of rechecking soil test results and the accuracy of rechecking results.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of batch rechecking of geotechnical test results, in particular to a batch rechecking method of geotechnical test results based on MATLAB. The present application also relates to a batch rechecking device of geotechnical test results based on MATLAB. BACKGROUND

[0002] Geotechnical test is one of the important components of geotechnical engineering investigation, and the test results will directly affect the quality of the investigation report. Rechecking the test results after the test is completed is an important means to ensure the quality of the investigation report.

[0003] At present, the industry still generally adopts manual method in the rechecking link, and each test is discriminated one by one by comparing the apparent characteristics of the test soil sample and the test results of the soil parameters. The fine-grained soil includes silt, silty clay, silt clay, silt clay, clay, and the soil type is relatively complex, so the rechecking is relatively difficult. When the amount of geotechnical test is large, the rechecking efficiency is very low. SUMMARY

[0004] The present application aims to overcome the technical problems of the prior art, and provides a batch rechecking method of geotechnical test results based on MATLAB. The present application also relates to a batch rechecking device of geotechnical test results based on MATLAB.

[0005] The present application provides a batch rechecking method of geotechnical test results based on MATLAB, which comprises the following steps:

[0006] Initializing the MATLAB running environment, relying on the BP neural network, reading the experimental data including the soil type in the.xls file and writing into the breast.mat program;

[0007] Loading the breast.mat program, dividing the training data and the rechecking data, and saving them in the traind_s and testd_s files after normalization processing;

[0008] Constructing a neural network, reading and completing the neural network training with the training data, and calculating the calculation variables obtained from the rechecking data by using the neural network;

[0009] Converting the calculation results and comparing them with the soil type in the initialized rechecking data to obtain the inconsistent group number, the accuracy rate and the error rate of the rechecking group.

[0010] Optionally, the MATLAB (also known as.mat) running environment is initialized, which comprises the following steps:

[0011] Clearing the global variables, closing the running windows that are not closed, and deleting the unnecessary variables and files.

[0012] Optionally, the dividing training data and review data comprises:

[0013] The training data is saved in the breast.mat program; and the test data for review is saved in the.xls file.

[0014] Reading the test data in the.xls file, and writing the test data into the breast.mat program;

[0015] Loading the breast.mat program, and re-dividing the training data and review data.

[0016] Optionally, the test data comprises description items: soil type, water content, liquid limit and plastic limit.

[0017] Optionally, further comprising data preprocessing based on the description items, comprising:

[0018] a) soil type quantification, silt = 0, silty clay = 0.25, mucky silty clay = 0.5, mucky clay = 0.75, clay = 1, the preprocessed data being used as the result item of the training data and the comparison item of the review data;

[0019] b) calculating variables according to the water content, liquid limit and plastic limit, comprising: water content w, water content control variable a, plasticity index Ip, plastic limit Wp, wherein a(a>=0) = 1, a(a<0) = 0;

[0020] Optionally, comprising normalization processing:

[0021] Mean shift and variance standardization are performed on the calculated variables of the training data and the calculated variables of the review data.

[0022] Optionally, the neural network comprises:

[0023] The number of network hidden layers is 5, the number of neurons in each layer is [9, 7, 5, 3, 2] respectively, the transfer function between layers is tansig function, the maximum number of training times is 5000, and the network target error is 0.00001.

[0024] Optionally, the neural network is used to calculate the calculated variables obtained through the review data, comprising:

[0025] In a behavioral unit, the quantitative soil type calculation result of each test group is calculated.

[0026] In the calculation process, a priority assignment code based on the plasticity index Ip is set, and when the normalized Ip value is less than a certain threshold, the calculation result is assigned as 0.

[0027] The application also provides a batch re-inspection device for soil test results based on MATLAB, comprising:

[0028] A writing module is configured to read experimental data including soil types in an.xls file and write the data into a breast.mat program based on a BP neural network after initializing a MATLAB running environment;

[0029] A dividing module is configured to load the breast.mat program, divide training data and re-inspection data, and save the data in traind_s and testd_s files after normalization processing;

[0030] A training module is configured to build a neural network, read and complete the neural network training with the training data, and calculate a calculation variable obtained from the re-inspection data by using the neural network;

[0031] A comparison module is configured to compare the calculation result with the soil types in the re-inspection data after initialization, and obtain a group number, an accuracy rate and an error rate of a re-inspection group that are inconsistent.

[0032] The application has the following advantages and beneficial effects:

[0033] The application provides a batch re-inspection method for soil test results based on MATLAB, comprising: initializing a MATLAB running environment, reading experimental data including soil types in an.xls file and writing the data into a breast.mat program based on a BP neural network; loading the breast.mat program, dividing training data and re-inspection data, and saving the data in traind_s and testd_s files after normalization processing; building a neural network, reading and completing the neural network training with the training data, and calculating a calculation variable obtained from the re-inspection data by using the neural network; and comparing the calculation result with the soil types in the re-inspection data after initialization, and obtaining a group number, an accuracy rate and an error rate of a re-inspection group that are inconsistent. The application re-inspects soil test results in batches based on MATLAB, simultaneously calculates data by using a neural network, and simultaneously improves the speed of re-inspection of soil test results and the accuracy of re-inspection results. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 FIG. 1 is a batch re-inspection process diagram for soil test results based on MATLAB in the application.

[0035] Figure 2 FIG. 2 is an execution diagram for batch re-inspection of soil test results based on MATLAB in the application.

[0036] Figure 3is a result schematic diagram of data division, transformation and normalization processing in the application.

[0037] Figure 4 is a neural network training schematic diagram in the application.

[0038] Figure 5 is a batch rechecking device schematic diagram of soil test results based on MATLAB in the application. DETAILED DESCRIPTION

[0039] The application will be further described below in conjunction with the drawings and specific embodiments, so that those skilled in the art can better understand the application and implement it.

[0040] The following are examples of specific implementation processes provided for detailed description of the technical solutions claimed by the application, but the application can also be implemented in other ways different from the description, and those skilled in the art can implement the application by using different technical means under the guidance of the concept of the application, so the application is not limited by the following specific embodiments.

[0041] The application provides a batch rechecking method of soil test results based on MATLAB, comprising: initializing a MATLAB running environment, relying on a BP neural network, reading experimental data including soil types in an.xls file and writing into a breast.mat program; loading the breast.mat program, dividing training data and rechecking data, and saving in traind_s and testd_s files after normalization processing; constructing a neural network, reading and completing the neural network training with the training data, and calculating a calculation variable obtained from the rechecking data by using the neural network; transforming the calculation result and comparing with the soil types in the initialized rechecking data, obtaining inconsistent group numbers, accuracy and error rate of the rechecking group. The application rechecks the batch soil test results based on MATLAB, simultaneously calculates the data by using the neural network, and simultaneously improves the speed of soil test result rechecking and the accuracy of rechecking results.

[0042] MATLAB is a commercial mathematical software produced by MathWorks Company in the United States, which integrates many powerful functions such as numerical analysis, matrix calculation, scientific data visualization, modeling and simulation of nonlinear dynamic systems in an easy-to-use window environment, and provides a comprehensive solution for scientific research, engineering design and many scientific fields that must perform effective numerical calculation.

[0043] The application provides a batch rechecking method of soil test results based on MATLAB, which replaces manual calculation with mechanical calculation to efficiently complete soil test rechecking work.

[0044] Please refer to Figure 1 and Figure 2 As shown in S101, initialize the MATLAB running environment, rely on BP neural network, read the experimental data including soil type in.xls file and write into breast.mat program.

[0045] Specifically, first initialize the MATLAB running environment: clear global variables, close the running window that is not closed, delete unnecessary variables and files.

[0046] Secondly, read and divide the data:

[0047] In this application, the implementation of soil test result recheck function relies on BP neural network.

[0048] The training data of the neural network is saved in the breast.mat program, the experimental data for recheck is.xls file, read the content in.xls file, arrange the content by rows, and write the content into breast.mat program.

[0049] For example: the training data of the network is saved in the breast.mat program, and accumulates 996 rows; the experimental data for recheck is.xls file, and accumulates 89 rows; read the content in.xls file, and write the content into breast.mat program, then the data in breast.mat program can reach 1085 rows. Then, load breast.mat program, and redivide the training data and recheck data.

[0050] S102 loads breast.mat program, divides training data and recheck data, and saves them in traind_s and testd_s files after normalization,

[0051] Finally, load breast.mat program, and redivide the training data and recheck data.

[0052] The data file of fine-grained soil should include soil type, water content, liquid limit and plastic limit and other description items. When rechecking, the description items of the data should be preprocessed first:

[0053] a) Quantitative soil type, silt = 0, silty clay = 0.25, silt clay = 0.5, silt clay = 0.75, clay = 1. The preprocessed data is saved in trainl and testl files as the result item of training data and the comparison item of recheck data.

[0054] b) Obtain the calculation variables according to the water content, liquid limit and plastic limit, including the water content w, water content control variable a, plasticity index Ip and plastic limit Wp, wherein a(a>=0)=1, a(a<0)=0. The preprocessed data are saved in the traind and testd files as the calculation variables of the training data and the calculation variables of the review data, respectively.

[0055] As shown in Figure 3 , the calculation variables are subjected to mean translation and then variance standardization. For example, for the data in the traind and testd files, first mean translation is performed, and then variance standardization is performed, and the normalized data are saved in the traind_s and testd_s files, respectively.

[0056] S103 constructs a neural network, reads and completes the neural network training with the training data, and calculates the calculation variables obtained from the review data by using the neural network;

[0057] The neural network is constructed by calling the breast.mat program to divide the training data, complete grid training, set the number of network hidden layers, the transfer function between layers and the number of neurons in each layer that meet the training requirements, and the network training control indicators include the maximum number of training times and the network target error.

[0058] As shown in Figure 4 , in the present application, the number of network hidden layers is set to 5, the number of neurons in each layer is [9, 7, 5, 3, 2] respectively, the transfer function between layers is tansig function, the maximum number of training times is 5000, and the network target error is 0.00001.

[0059] The trained neural network is called;

[0060] The calculation variables obtained from the review data are calculated by using the neural network, and the quantitative soil type calculation results of each group of tests are obtained in units of behaviors. In the calculation process, the priority assignment code based on the plasticity index Ip is set, and when the normalized Ip value is less than a certain threshold, the calculation result is assigned as 0.

[0061] S104 converts the calculation results and compares them with the soil types in the initialized review data, obtains the group numbers that do not match, the accuracy rate of the review groups and the error rate.

[0062] In this application, the calculation results are converted and compared with the soil type in the initialized recheck data in units of groups, and the group number of the inconsistent results, the overall accuracy rate of all recheck groups and the error rate are displayed. In this embodiment, the overall accuracy rate of 89 groups of soil test is 89%, the error rate is 3.9%, there are 9 suspected test groups, and the results are as follows:

[0063] errorrate= 0.039.

[0065] Accuracy rate

[0066] 89.887640%

[0067] Error rate

[0068] 3.902866%

[0069] ans= 2 21 39 45 47 48 58 59 61

[0071] The plasticity index of the result is assigned as follows:

[0072]

[0073]

[0074] The calculation results are converted:

[0075] When the calculation result is greater than or equal to 0.875, the calculation result is converted to 1; when the calculation result is greater than or equal to 0.625 and less than 0.875, the calculation result is converted to 0.75; when the calculation result is greater than or equal to 0.375 and less than 0.625, the calculation result is converted to 0.5; when the calculation result is greater than or equal to 0.125 and less than 0.375, the calculation result is converted to 0.25; and when the calculation result is less than 0.125, the calculation result is converted to 0.

[0076] Finally, the converted calculation results are compared with the soil type in the initialized recheck data in units of lines, and the group number of the inconsistent results, the overall accuracy rate of all recheck groups and the error rate are displayed.

[0077] In this application, first, the recheck test data is read and imported, then the data is preprocessed to obtain quantitative soil type and calculation variable, and then neural network training and network calculation are performed, then the calculation results are converted in units of lines, and the soil type in the initialized recheck data is compared, and finally the group number of the inconsistent results, the overall accuracy rate of all recheck groups and the error rate are displayed. In this way, the recheck of soil test results is efficiently completed, and the efficiency is improved.

[0078] The present application also provides a MATLAB-based geotechnical test result batch re-inspection device for executing the above method process.

[0079] Please refer to the device Figure 5 As shown, it mainly includes:

[0080] The writing module 201 is used to initialize the MATLAB operating environment, and then read the experimental data including soil types in the .xls file based on the BP neural network and write the data into the breast.mat program;

[0081] The division module 202 is used to load the breast.mat program, divide the training data and the retest data, and save them in the traind_s and testd_s files respectively after normalization;

[0082] A training module 203 is used to construct a neural network, read and complete the training of the neural network using the training data, and use the neural network to calculate the calculated variables obtained from the re-inspection data;

[0083] The comparison module 204 is used to convert the calculation result and compare it with the soil type in the initialized re-inspection data to obtain the inconsistent group number, the accuracy rate and the error rate of the re-inspection group.

Claims

1. A batch recheck method for geotechnical test results based on MATLAB, characterized in that: include: Initialize the MATLAB operating environment, rely on the BP neural network, read the experimental data including soil types in the .xls file and write it into the breast.mat program; Load the breast.mat program, divide the training data and retest data, and save them in the traind_s and testd_s files respectively after normalization; Constructing a neural network, reading and completing the training of the neural network using the training data, and using the neural network to calculate the calculated variables obtained from the re-inspection data; The calculation result is converted and compared with the soil type in the initialized re-inspection data to obtain the inconsistent group number, the accuracy rate and the error rate of the re-inspection group, including: when the calculation result is greater than or equal to 0.875, the calculation result is converted to 1; when the calculation result is greater than or equal to 0.625 and less than 0.875, the calculation result is converted to 0.75; when the calculation result is greater than or equal to 0.375 and less than 0.625, the calculation result is converted to 0.5; when the calculation result is greater than or equal to 0.125 and less than 0.375, the calculation result is converted to 0.25; when the calculation result is less than 0.125, the calculation result is converted to 0; The experimental data include the following descriptive items: soil type, moisture content, liquid limit and plastic limit; Data preprocessing is performed based on the description items, including: a) Quantification of soil type: silt = 0, silty clay = 0.25, silty silty clay = 0.5, silty clay = 0.75, and clay = 1. The preprocessed data is used as the result item of the training data and the comparison item of the retest data; b) obtaining calculation variables according to the moisture content, liquid limit and plastic limit, including: moisture content w, water content control variable a, plasticity index Ip, plastic limit Wp, where a (a>=0)=1, a (a<0)=0; The preprocessed data are used as calculation variables for training data and retest data.

2. The MATLAB-based batch re-examination method for geotechnical test results according to claim 1, characterized in that: Initialize the MATLAB operating environment, including: Clear global variables, close open run windows, and delete unnecessary variables and files.

3. The MATLAB-based batch re-examination method for geotechnical test results according to claim 1, characterized in that: The dividing of training data and recheck data includes: The training data are saved in the breast.mat program; the test data for retesting are saved in .xls files; Read the experimental data in the .xls file and write the experimental data into the breast.mat program; Load the breast.mat program and re-divide the training data and re-examination data.

4. The MATLAB-based batch re-examination method for geotechnical test results according to claim 1, characterized in that: Including normalization: The calculated variables of the training data and the calculated variables of the retest data are mean shifted and variance normalized.

5. The MATLAB-based batch re-examination method for geotechnical test results according to claim 1, characterized in that: The neural network comprises: The number of hidden layers in the network is 5, the number of neurons in each layer is [9, 7, 5, 3, 2], the transfer function between layers is the tansig function, the maximum number of training times is 5000, and the network target error is 0.00001.

6. The MATLAB-based batch rechecking method for geotechnical test results according to claim 1, characterized in that: The neural network is used to calculate the calculated variables obtained from the re-examination data, including: Calculate the quantitative soil type calculation results of each test group in rows; During the calculation process, a priority assignment code based on the plasticity index Ip is set. When the normalized Ip value is less than a certain threshold, the calculation result is assigned a value of 0.

7. A batch rechecking device for geotechnical test results based on MATLAB, characterized in that: include: The writing module is used to initialize the MATLAB operating environment, and then read the experimental data including soil types in the .xls file based on the BP neural network and write it into the breast.mat program; The partitioning module is used to load the breast.mat program, partition the training data and retest data, and save them in the traind_s and testd_s files after normalization; A training module, configured to construct a neural network, read and complete the training of the neural network using the training data, and calculate the calculated variables obtained from the retest data using the neural network; A comparison module is used to convert the calculation result and compare it with the soil type in the initialized re-inspection data to obtain the inconsistent group number, the accuracy rate and the error rate of the re-inspection group, including: when the calculation result is greater than or equal to 0.875, the calculation result is converted to 1; when the calculation result is greater than or equal to 0.625 and less than 0.875, the calculation result is converted to 0.75; when the calculation result is greater than or equal to 0.375 and less than 0.625, the calculation result is converted to 0.5; when the calculation result is greater than or equal to 0.125 and less than 0.375, the calculation result is converted to 0.25; when the calculation result is less than 0.125, the calculation result is converted to 0; The experimental data include the following descriptive items: soil type, moisture content, liquid limit and plastic limit; Data preprocessing is performed based on the description items, including: a) Quantification of soil type: silt = 0, silty clay = 0.25, silty silty clay = 0.5, silty clay = 0.75, and clay = 1. The preprocessed data is used as the result item of the training data and the comparison item of the retest data; b) obtaining calculation variables according to the moisture content, liquid limit and plastic limit, including: moisture content w, water content control variable a, plasticity index Ip, plastic limit Wp, where a (a>=0)=1, a (a<0)=0; The preprocessed data are used as calculation variables for training data and retest data.

Images