Method and system for predicting soil erosion and particle migration based on acoustic emission monitoring

Abstract

The present application relates to geotechnical engineering monitoring technical field, especially soil erosion and particle migration prediction method and system based on acoustic emission monitoring, the method comprises the following steps: obtaining soil quality ratio parameter, preparing test soil sample based on the soil quality ratio parameter; Constructing multi-field coupling test mechanism to carry out multi-field coupling soil test on the test soil sample; Establishing soil sample data acquisition device to obtain soil sample monitoring data in the process of multi-field coupling soil test; According to the soil sample monitoring data, the acoustic emission characteristic parameters of the test soil body are obtained, and a multivariate nonlinear regression model is established; Constructing soil prediction model to realize the prediction of soil erosion and particle migration. The present application breaks through the limitation of traditional experience model by fusing multi-field coupling monitoring and machine learning, realizes high-precision dynamic prediction of soil erosion and particle migration, provides scientific basis for roadbed engineering stability evaluation, and significantly improves prediction reliability and engineering applicability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geotechnical engineering monitoring, in particular to a soil erosion and particle migration prediction method and system based on acoustic emission monitoring. BACKGROUND

[0002] Traditional soil erosion monitoring methods have obvious limitations. Physical sampling method obtains samples through drilling or profile cutting, which can directly analyze particle composition, but the sampling process directly destroys the original structure of the soil, disturbs the pore water pressure field and stress state, resulting in data distortion, and the discrete monitoring mode is difficult to capture the instantaneous changes under sudden conditions such as heavy rain and earthquake, and the real-time performance is poor. The optical observation method is limited by the opacity of the soil, and can only monitor the surface layer of a few centimeters, and cannot capture the internal particle migration, and the image analysis relies on complex algorithms, with long processing period, which is difficult to meet the real-time early warning needs of engineering. The existing vibration-seepage coupling test device can simulate complex load, but it focuses on macroscopic parameter measurement and is insufficient in micro-particle motion monitoring. Traditional devices are difficult to capture micro-features such as fine particle migration path and pore channel evolution, and are dependent on invasive sensors, which will interfere with water flow and particle motion, resulting in data distortion. The lack of non-contact monitoring methods further hinders the in-depth analysis of the erosion mechanism. The above limitations highlight the need to develop new monitoring technologies and break through the traditional framework to provide more accurate basis for soil engineering stability evaluation. SUMMARY

[0003] In view of the defects in the prior art, the present application provides a soil erosion and particle migration prediction method and system based on acoustic emission monitoring.

[0004] To achieve the above object, in a first aspect, the application provides a soil erosion and particle migration prediction method based on acoustic emission monitoring, which comprises the following steps: obtaining soil quality proportioning parameters, preparing a test soil sample based on the soil quality proportioning parameters; constructing a multi-field coupling test mechanism, and performing a multi-field coupling soil test on the test soil sample according to the multi-field coupling test mechanism; establishing a soil sample data acquisition device, and acquiring soil sample monitoring data during the multi-field coupling soil test by using the soil sample data acquisition device; obtaining acoustic emission characteristic parameters of the test soil sample according to the soil sample monitoring data, and establishing a multi-element nonlinear regression model based on the acoustic emission characteristic parameters; constructing a soil prediction model in combination with the multi-element nonlinear regression model, and realizing prediction of soil erosion and particle migration according to the soil prediction model. The application can precisely prepare a test soil sample, simulate complex working conditions in combination with a multi-field coupling test, acquire comprehensive monitoring data by using a multi-source data acquisition device, extract acoustic emission characteristic parameters to construct a multi-element nonlinear regression model, and then combine machine learning to construct a prediction model. The application breaks through the limitations of traditional methods, realizes high-precision dynamic prediction of soil erosion and particle migration, provides a scientific basis for engineering stability evaluation, and improves prediction reliability and engineering applicability.

[0005] Optionally, the obtaining soil quality proportioning parameters and preparing a test soil sample based on the soil quality proportioning parameters comprises: obtaining mass proportioning ratios of sand, clay and silt, and taking the mass proportioning ratios as the soil quality proportioning parameters; obtaining soil compaction density, and filling the sand, the clay and the silt into a model box in layers in combination with the soil quality proportioning parameters and the soil compaction density to obtain the test soil sample. The application can precisely reproduce actual soil composition and structure by determining mass proportions of sand, clay and silt and compaction density, preparing a test soil sample in layers, reducing soil sample preparation errors, providing a standardized and consistent research object for subsequent multi-field coupling tests, and ensuring reliability of monitoring data and scientificity of test results.

[0006] Optionally, the constructing a multi-field coupling test mechanism and performing a multi-field coupling soil test on the test soil sample according to the multi-field coupling test mechanism comprises: constructing the multi-field coupling test mechanism based on adjustment of a hydraulic gradient and a composite vibration load of the test soil sample; and causing soil erosion and particle migration of the test soil sample according to the multi-field coupling test mechanism to perform the multi-field coupling soil test. The application can simulate complex working conditions such as rainstorm and earthquake, precisely cause soil erosion and particle migration by adjusting and controlling a hydraulic gradient and a composite vibration load to construct a multi-field coupling mechanism, break through the limitations of traditional single factors, reproduce a multi-factor action scene in real engineering, and provide test data close to reality for subsequent monitoring and model construction.

[0007] Optionally, the soil sample data acquisition device is used to obtain soil sample monitoring data in the multi-field coupling soil body test process, including: constructing the soil sample data acquisition device based on an acoustic emission sensor array, an X-ray scanning device, a pore pressure sensor, and an acceleration vibration sensor; using the acoustic emission sensor array to collect acoustic emission signals of the test soil sample; obtaining three-dimensional image data of the test soil sample according to the X-ray scanning device; obtaining pore pressure of the test soil sample through the pore pressure sensor; obtaining vibration acceleration of the test soil sample according to the acceleration vibration sensor; and fusing the acoustic emission signals, the three-dimensional image data, the pore pressure, and the vibration acceleration to obtain the soil sample monitoring data. The data acquisition device is constructed by using multiple types of sensors, acoustic emission signals, three-dimensional images, pore pressure, and vibration acceleration are synchronously obtained, and multi-dimensional data is fused; both microcosmic characteristics of internal particle movement of the soil body and macroscopic mechanical parameters are captured, the traditional monitoring limitations are broken through, comprehensive and accurate data support is provided for subsequent model construction, and prediction reliability is improved.

[0008] Optionally, the three-dimensional image data of the test soil sample is obtained according to the X-ray scanning device, including: the X-ray scanning device adopts a dynamic interval scanning mode to scan the test soil sample to obtain an initial soil sample image; when the acoustic emission signal changes sharply, the dynamic interval scanning mode is switched to a high-frequency scanning mode; and the three-dimensional image data is obtained by three-dimensional reconstruction and data processing of the initial soil sample image. Through switching of the dynamic interval and high-frequency scanning modes and in combination with three-dimensional reconstruction processing, initial images are efficiently obtained, key changes can be accurately captured when the acoustic emission signal changes sharply, efficiency and accuracy are taken into account, internal structure evolution of the soil body is completely presented, and high-value three-dimensional data is provided for analysis of soil body erosion and particle migration.

[0009] Optionally, the method further comprises: extracting acoustic emission features from the soil sample monitoring data to obtain an event count rate, an energy integral, and a dominant frequency; using the event count rate, the energy integral, and the dominant frequency as the acoustic emission characteristic parameters; constructing a first multivariate nonlinear regression model based on the event count rate, the energy integral, and the dominant frequency; performing normalization processing on the acoustic emission characteristic parameters to obtain acoustic emission characteristic normalized parameters; performing single-factor correlation analysis on the acoustic emission characteristic parameters and the soil erosion amount of the sample soil to obtain a correlation analysis result; and combining the acoustic emission characteristic normalized parameters and the correlation analysis result to establish a second multivariate nonlinear regression model.

[0010] Optionally, the method further comprises:

[0011]

[0012] wherein, is a soil particle loss volume, is an event count rate, is an energy integral, is a dominant frequency, is a hydraulic gradient, is a vibration intensity factor, is an error term. The first multivariate nonlinear regression model integrates the effects of multiple factors on the particle loss volume, quantifies the correlation between acoustic emission features and vibration, improves the adaptability of the model to complex working conditions, and provides a reliable mathematical basis for prediction.

[0013] Optionally, the combining the multiple nonlinear regression model constructs a soil prediction model, and the soil erosion and particle migration are predicted according to the soil prediction model, including: obtaining multi-working condition data, dividing the multi-working condition data into a training set, a verification set and a test set; constructing a hybrid model structure based on machine learning, and establishing the soil prediction model according to the multi-working condition data and the hybrid model structure; and obtaining a soil particle loss volume prediction value according to the soil prediction model, and predicting the soil erosion and particle migration. The application realizes the high-precision prediction of the soil erosion and particle migration by the multi-working condition data division and the machine learning hybrid model construction, the combination of the multiple nonlinear regression model and the soil prediction model, the accurate output of the particle loss volume prediction value, the consideration of the data diversity and the model generalization ability, and the like.

[0014] Optionally, the obtaining a soil particle loss volume prediction value according to the soil prediction model, and predicting the soil erosion and particle migration, includes:

[0015]

[0016] wherein, is the soil particle loss volume prediction value, is a ReLU activation function, is a weight matrix, is a spatial feature, is a time feature, is a bias term. The application realizes the accurate output of the particle loss volume prediction value by the fusion of the spatial and time features through the activation function, the quantification of the influence of the weight matrix and the bias term, and the like, and improves the dynamic nature and precision of the soil erosion and migration prediction.

[0017] In a second aspect, the application provides a soil erosion and particle migration prediction system based on acoustic emission monitoring, which executes the soil erosion and particle migration prediction method based on acoustic emission monitoring provided by the application. The system comprises an input device, an output device, a processor and a memory, and is characterized in that the input device, the output device, the processor and the memory are connected with each other, wherein the memory is used for storing a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions. The application realizes the system monitoring and prediction of the soil erosion and particle migration by the connection of the input and output devices, the processor and the memory, the storage and calling of the program instructions for executing the prediction method, the guarantee of the data processing efficiency and the method execution stability, and the improvement of the convenience and reliability of the engineering application. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is a flow chart of the soil erosion and particle migration prediction method based on acoustic emission monitoring of the embodiment of the application;

[0019] Figure 2 A soil erosion and particle migration prediction system framework based on acoustic emission monitoring according to an embodiment of the present application;

[0020] Figure 3 A multi-field coupling soil test device according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] The specific embodiments of the present application will be described in detail below, it should be noted that the embodiments described herein are only used for illustration and do not limit the present application. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present application. However, it is obvious to those skilled in the art that the specific details are not necessarily used to implement the present application. In other instances, well-known circuits, software or methods have not been specifically described in order to avoid obscuring the present application.

[0022] Throughout the specification, the reference to "one embodiment", "an embodiment", "one example" or "an example" means that a particular feature, structure or characteristic described in connection with the embodiment or example is included in at least one embodiment of the present application. Therefore, the phrases "in one embodiment", "in an embodiment", "one example" or "an example" appearing throughout the specification are not necessarily all referring to the same embodiment or example. In addition, specific features, structures or characteristics can be combined in any appropriate combination and / or subcombination in one or more embodiments or examples. In addition, those skilled in the art should understand that the drawings provided herein are for illustrative purposes only and the drawings are not necessarily drawn to scale.

[0023] Please refer to Figure 1 An embodiment of the present application provides a soil erosion and particle migration prediction method based on acoustic emission monitoring, the method comprising the following steps:

[0024] S1, obtaining soil quality proportioning parameters, and preparing a test soil sample based on the soil quality proportioning parameters.

[0025] In the present embodiment, the mass proportioning ratios of sand, clay and silt are obtained, and the mass proportioning ratios are used as the soil quality proportioning parameters; the soil compaction density is obtained, and the sand, clay and silt are filled into a model box in layers in combination with the soil quality proportioning parameters and the soil compaction density to obtain the test soil sample.

[0026] Specifically, the soil sample is prepared according to the mass ratio, the mass proportioning ratio of sand is 40%-60%, the mass proportioning ratio of clay is 20%-30%, and the mass proportioning ratio of silt is 10%-20%; the soil compaction density is 1.6g / cm³-2.0g / cm³, and the sand, clay and silt are filled into the model box in layers to ensure that the density of each layer meets the range of the soil compaction density.

[0027] S2, construct a multi-field coupling test mechanism, and perform a multi-field coupling soil body test on the test soil sample according to the multi-field coupling test mechanism.

[0028] In this embodiment, the multi-field coupling test mechanism is constructed based on adjustment of the hydraulic gradient and the composite vibration load of the test soil sample; and the soil body erosion and particle migration of the test soil sample are caused according to the multi-field coupling test mechanism, so as to perform the multi-field coupling soil body test.

[0029] Specifically, the hydraulic gradient (denoted as i) of the test soil sample is regulated to a set value (0.5≤i≤4.0), and when the fluctuation rate of the pore water pressure is less than 5%, it is determined that the seepage is stable; the composite vibration load is applied to the test soil sample, the vibration frequency is 1Hz-20Hz, the acceleration amplitude is 0.2m / s 2 -1.5m / s 2 , and the duration is 10min-60min; and the multi-field coupling test mechanism is constructed.

[0030] Further, the hydraulic gradient and the vibration load are adjusted based on the multi-field coupling test mechanism, so that the soil body erosion and particle migration of the test soil sample occur, and the multi-field coupling soil body test is performed.

[0031] S3, establish a soil sample data acquisition device, and acquire soil sample monitoring data in the multi-field coupling soil body test process by using the soil sample data acquisition device.

[0032] In this embodiment, the soil sample data acquisition device is constructed based on an acoustic emission sensor array, an X-ray scanning device, a pore pressure sensor and an acceleration vibration sensor; the acoustic emission signal of the test soil sample is acquired by using the acoustic emission sensor array; the three-dimensional image data of the test soil sample is obtained according to the X-ray scanning device; the pore pressure of the test soil sample is obtained by the pore pressure sensor; the vibration acceleration of the test soil sample is obtained according to the acceleration vibration sensor; and the soil sample monitoring data is obtained by fusing the acoustic emission signal, the three-dimensional image data, the pore pressure and the vibration acceleration.

[0033] Specifically, the soil sample data acquisition device is composed of a 32-channel acoustic emission sensor array, a pore pressure sensor, an acceleration vibration sensor and an X-ray Micro Computed Tomography (X-ray micro CT) system; the 32-channel acoustic emission sensor array has a sampling rate greater than or equal to 1 MHz, a 16-bit resolution, a 50 dB preamplifier and an adaptive threshold trigger mode; the pore pressure sensor is arranged according to each layer of soil sample; the specific configuration and working principle of the X-ray micro CT system adopts a high-resolution industrial micro-focus CT device, and the core components thereof include an X-ray source, a closed micro-focus X-ray tube, a focal point size adjustable to 5-50 microns, a tube voltage range of 30-160 kV, a tube current of 5-200 microamperes, penetration imaging of different density soil bodies (such as sand and clay) is realized by adjusting the voltage and current, and it is ensured that the particle size distribution (0.01-2 mm particle size) can be clearly identified in the erosion process, and the soil structure damage caused by excessive radiation can be avoided; a detector, a flat panel detector with 2048x2048 pixels, a pixel size of 50 microns, a frame rate of greater than or equal to 15 fps, support for dynamic tomography, and the ability to capture the instantaneous state of particle migration in the soil erosion process (such as piping channel formation and particle loss trajectory).

[0034] Further, the mechanical scanning system is equipped with a high-precision rotary table (positioning accuracy ±0.001°) and a three-dimensional translation table, which can realize 360° continuous rotary scanning, and the scanning range covers the entire model box, meeting the imaging needs of samples of different sizes.

[0035] In the process of the multi-field coupled soil body test, the acoustic emission sensor array is triggered to collect acoustic emission signals, and the acoustic emission signal preprocessing satisfies the following relationship:

[0036]

[0037] Wherein, is the filtered acoustic emission signal, is the original acoustic emission signal, represents a Hamming window function with a length of 512 points, is the background noise mean value.

[0038] In this embodiment, the X-ray scanning device adopts a dynamic interval scanning mode to scan the test soil sample to obtain an initial soil sample image; when the acoustic emission signal changes sharply, the dynamic interval scanning mode is switched to a high-frequency scanning mode; and three-dimensional reconstruction and data processing are performed on the initial soil sample image to obtain three-dimensional image data.

[0039] Specifically, for soil erosion monitoring, the "dynamic interval scanning" mode is adopted: in the initial stage of erosion (0h-2h), a full-range scan is performed every 30 minutes, and the single scan time is 10-15 minutes (adjusted according to the size of the model box), and the spatial resolution is set to 50μm (0.1mm or more particles can be identified); when the acoustic emission sensor detects intense signals (such as concentrated particle loss), it automatically switches to "high-frequency scanning mode", the scanning interval is shortened to 5 minutes, and the resolution is improved to 20μm, focusing on capturing the micro changes in the active erosion area. A CT scan marker point (containing a tantalum metal positioning ball) is reserved outside the model box for three-dimensional coordinate calibration of the scanning data, ensuring that the imaging data at different time points can be accurately superimposed and compared. After scanning, the CT reconstruction software is used to splice the tomographic images, generate a three-dimensional voxel model of the soil body, separate the soil particles and pore / water area using threshold segmentation algorithm, extract the particle volume, porosity, connectivity and other quantitative parameters before and after erosion, calculate the particle loss rate in different areas through time series analysis, and generate a dynamic evolution video of the erosion channel. The related parameters are input into the data acquisition system.

[0040] Further, the three-dimensional voxel model includes a three-dimensional voxel reconstruction algorithm, which satisfies the following relationship:

[0041]

[0042] wherein, is the voxel value at the coordinate in the three-dimensional space, is the three-dimensional coordinate point, is the total number of projections, is the index variable, is the projection intensity of the th time, is the natural logarithm base, is the linear attenuation coefficient of the soil body, is the penetration path length of the th projection, is the ray source coordinate of the th projection.

[0043] In an optional embodiment, the pore connectivity is quantified, which satisfies the following relationship:

[0044]

[0045] wherein, is the pore connectivity quantification index, is the number of connected pores, is the total number of pores.

[0046] In the embodiment, multi-modal data fusion is performed to associate the three-dimensional image data (such as pore distribution and particle size distribution) obtained by the X-ray micro-CT with the acoustic emission signal features (such as event count, energy integration and dominant frequency) in time and space, and a "microstructure-macro signal" mapping relationship is established; for example, when the CT identifies a sudden increase in porosity in a certain region, it is analyzed whether a high-frequency signal (characteristic of particle friction or collision) appears in the acoustic emission sensor corresponding to the region, and the consistency of the erosion mechanism is verified.

[0047] S4. Acoustic emission characteristic parameters of the test soil body are obtained according to the soil sample monitoring data, and a multivariate nonlinear regression model is established based on the acoustic emission characteristic parameters.

[0048] In the embodiment, the event count rate, energy integration and dominant frequency are obtained by performing acoustic emission feature extraction on the soil sample monitoring data, and the event count rate, energy integration and dominant frequency are taken as the acoustic emission characteristic parameters; a first multivariate nonlinear regression model is constructed based on the event count rate, energy integration and dominant frequency; the acoustic emission characteristic parameters are normalized to obtain acoustic emission characteristic normalized parameters; a single-factor correlation analysis is performed on the acoustic emission characteristic parameters and the soil erosion amount of the sample soil to obtain a correlation analysis result; and a second multivariate nonlinear regression model is established in combination with the acoustic emission characteristic normalized parameters and the correlation analysis result.

[0049] Specifically, the event count rate, energy integration and dominant frequency are obtained by performing acoustic emission feature extraction on the acoustic emission signals in the soil sample monitoring data.

[0050] The above event count rate satisfies the following relationship:

[0051]

[0052] wherein, is the event count rate, is the total number of sampling points in the statistical time period, is an index variable of the sampling point, is an impulse function, is the total duration of the statistical time window.

[0053] It should be noted that when the signal amplitude exceeds the threshold value, .

[0054] The above energy integration satisfies the following relationship:

[0055]

[0056] wherein, is the energy integration of the acoustic emission signal, is the integration time interval, is the acoustic emission signal after filtering.

[0057] The main frequency satisfies the following relationship:

[0058]

[0059] wherein, is the main frequency of the acoustic emission signal, represents the independent variable corresponding to the maximum value, is the Fourier transform of the signal , and is the acoustic emission signal after filtering processing.

[0060] Further, a first multivariate nonlinear regression model is constructed based on the event count rate, energy integral and main frequency, satisfying the following relationship:

[0061]

[0062] wherein, is the soil particle loss volume, is the event count rate, is the energy integral, is the main frequency, is the hydraulic gradient, is the vibration intensity factor, is the error term.

[0063] Specifically, the vibration intensity factor satisfies the following relationship:

[0064]

[0065] wherein, is the vibration intensity factor, is the vibration duration, is the index variable of three directions, is the vibration acceleration time history of the th direction.

[0066] In the present embodiment, the acoustic emission characteristic parameters are normalized, satisfying the following relationship:

[0067]

[0068] wherein, is the acoustic emission characteristic normalized parameter, is the acoustic emission characteristic parameter, is the minimum value of the acoustic emission characteristic parameter, is the maximum value of the acoustic emission characteristic parameter.

[0069] Specifically, the Pearson correlation coefficient was used to conduct a single-factor correlation analysis on the acoustic emission characteristic parameters and soil erosion. The results of the correlation analysis showed that the Pearson correlation coefficient satisfies the following relationship:

[0070]

[0071] in, The Pearson correlation coefficient is used. For sample size, For the sample index variable, For the first Acoustic emission parameter values ​​for each sample, The sample mean of the acoustic emission parameters is denoted as . For the first The particle loss volume measurement value corresponding to each sample. This represents the sample mean of the volume of particles lost.

[0072] Furthermore, a second multivariate nonlinear regression model is established for the acoustic emission parameters and particle migration, satisfying the following relationship:

[0073]

[0074] in, This represents the amount of particle migration. For regression coefficients, For event count rate, The energy integral of the acoustic emission signal. The dominant frequency of the acoustic emission signal. For hydraulic gradient, Vibration intensity factor This is the error term.

[0075] It should be noted that the regression coefficients are solved using the least squares method.

[0076] S5. Construct a soil prediction model based on the multivariate nonlinear regression model, and use the soil prediction model to predict soil erosion and particle migration.

[0077] In this embodiment, multi-condition data is acquired and divided into training set, validation set and test set; a hybrid model structure is constructed based on machine learning, and a soil prediction model is established based on the multi-condition data and the hybrid model structure; the predicted value of soil particle loss volume is obtained according to the soil prediction model, so as to realize the prediction of soil erosion and particle migration.

[0078] Specifically, the data set is constructed by collecting 300 groups of multi-condition data, which is divided into training set, validation set and test set according to the ratio of 7:1.5:1.5; the input features include event count, energy integral, dominant frequency, hydraulic gradient, vibration intensity factor and CT porosity, and the output label is the predicted value of soil particle loss volume.

[0079] Further, a hybrid model structure (3D-CNN-LSTM hybrid model structure) is constructed based on a three-dimensional convolutional neural network (3D-CNN) and a long short-term memory network (LSTM); a 3D convolutional layer (kernel size 3×3×3, channel number 64→128→256); an LSTM layer (hidden unit 128, time step 100); and then a 3D-CNN-LSTM model is constructed as a soil prediction model by combining multi-condition data.

[0080] In this embodiment, the 3D-CNN-LSTM model is trained, the 3D-CNN module extracts spatial features through 3 convolutional layers, and the LSTM module inputs acoustic emission time sequence features (100×4 dimensions) and extracts time features through 2 LSTM layers.

[0081] Further, the spatial and time features are spliced through a fusion layer, and the soil particle loss volume prediction value is output through a fully connected layer, satisfying the following relationship:

[0082]

[0083] wherein, is the soil particle loss volume prediction value, is a ReLU activation function, is a weight matrix, is a spatial feature, is a time feature, is a bias term.

[0084] In this embodiment, the model is optimized by using an adaptive moment estimation optimizer (Adam) with an initial learning rate of 0.001, a batch size of 32, and an early stopping method (patience value=20). Cross-validation is used to evaluate the stability of the model, calculate the root mean square error and the determination coefficient. Prediction application: input new condition parameters, the model outputs the particle loss volume prediction curve within 60 minutes in the future, and gives the 95% confidence interval.

[0085] The above root mean square error satisfies the following relationship:

[0086]

[0087] wherein, is the root mean square error, is the number of samples, is the sample index variable, is the predicted value of the model for the th sample, is the true value of the th sample.

[0088] The above coefficient of determination satisfies the following relationship:

[0089]

[0090] wherein, is the coefficient of determination, is the number of samples, is the sample index variable, is the true value of the th sample, is the predicted value of the model for the th sample, is the mean of all sample true values.

[0091] In the present embodiment, the validation of the soil body prediction model adopts a dynamic weight adjustment strategy, and the loss function is defined to satisfy the following relationship:

[0092]

[0093] wherein, is the total loss function, is the dynamic weight coefficient, is the mean square error, is the mean absolute error, is the soil particle loss volume.

[0094] It should be noted that, , when the root mean square error of the validation set rises for three consecutive times, the learning rate decay is triggered (decay factor ).

[0095] In an optional embodiment, a random forest algorithm is introduced as a machine learning algorithm to construct a soil erosion prediction model; the soil erosion evolution data is divided into a training set, a validation set and a test set, and the soil erosion prediction model is trained, parameterized and evaluated; under multiple working conditions, the evolution law of soil erosion is predicted according to the soil erosion and particle migration prediction model.

[0096] It should be noted that the random forest model training, model parameters: the number of decision trees is 100, the maximum depth is 15, and the minimum leaf node sample number is 5; feature importance calculation: evaluate the weight of each feature through the Gini coefficient drop.

[0097] Please refer to Figure 2 In an optional embodiment, the present application provides a soil erosion and particle migration prediction system based on acoustic emission monitoring, which comprises an input device, an output device, a processor and a memory, and the hardware facilities are connected with each other, wherein the memory is used for storing a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions to execute the specific steps of the embodiments of the soil erosion and particle migration prediction method based on acoustic emission monitoring provided by the present application. The soil erosion and particle migration prediction system based on acoustic emission monitoring provided by the present application has complete structure, objective stability, and improves the overall applicability and practical application ability of the present application.

[0098] Please refer to Figure 3 , which is a schematic diagram of a multi-field coupling soil test device; an air compressor provides power to a water storage tank through a pressure stabilizing valve, and water is delivered to a model tank containing a test soil sample through a water pipe; the bottom of the model tank is equipped with a pore pressure sensor, a dynamic load control system and the like, and the side is provided with an acoustic emission sensor array, an eccentric vibrator, an acceleration vibration sensor and an X-ray micro-CT system for monitoring the mechanical, acoustic and other responses of the soil sample under different conditions; the water seepage of the soil sample flows into the water outlet tank through the model tank, and the collection of water and the monitoring of related parameters are completed by a thin film piezoelectric sensor, a filter screen and a water outlet; all collected data are transmitted to a data acquisition system and finally summarized to a machine learning platform, and the test data of the soil sample are analyzed in depth by means of machine learning technology. Specifically, it includes the following:

[0099] The sample loading system: the model tank is used for layer-by-layer filling of the test soil body; the pore pressure sensors are uniformly arranged in each soil layer inside the model tank for real-time monitoring of the pore water pressure distribution; the 32-channel acoustic emission sensor array arranged on six sides has a sampling rate of ≥1MHz, a 16bit resolution, a 50dB preamplifier and an adaptive threshold trigger mode; the acceleration vibration sensor is integrated at the bottom of the model tank for collecting vibration acceleration parameters. The sample loading system includes a soil storage device and a hose, the inside of the soil storage device is a storage cavity, a horizontal water inlet is formed on the left side of the storage cavity, four vertical water inlets are formed on the upper end of the storage cavity, and a water outlet is formed on the right side of the storage cavity, and the storage cavity is filled with a test soil sample.

[0100] Water supply system: air compressor and double pressure stabilizing valve cooperate to regulate the water flow of the inlet and outlet of the model box, to realize the dynamic control of hydraulic gradient; the water storage tank is connected to the model box through the air pressure pipe, and the hydraulic gradient control range is 0.1-5.0. The water supply system is composed of an air compressor, an air pressure pipe and a water storage tank; a precision pressure valve is used to realize precise air pressure control in cooperation with the air compressor; to ensure the stability of water pressure, a double pressure stabilizing valve is used to control the gas pressure of the water storage tank. The water flow of the model tank is adjusted by the precision control valve to realize the control of the hydraulic gradient, so that the test soil in the model tank reaches a stable seepage state; under the stable seepage state, the pore water pressure distribution of the test soil is obtained based on the data acquisition system.

[0101] Dynamic load control system: the system is arranged at the lower end of the model box, including an eccentric vibrator with adjustable amplitude and speed, and an acceleration vibration sensor for monitoring soil disturbance; the vibration table is designed to be retractable to adapt to different sizes of model boxes. The model box is vibrated based on the dynamic load control system, and the disturbance intensity of the test soil is monitored by the acceleration vibration sensor, including: establishing a vibration table surface, which is located below the model box and is designed as a retractable table to test different sizes of samples; the eccentric vibrator is used to apply a vibration load to the vibration table surface to vibrate the vibration table surface; the vibration load includes horizontal and vertical vibration loads; the vibration of the model box is realized based on the vibration of the vibration table surface, and the vibration parameters of the model box are obtained based on the data acquisition system to control the vibration parameters under various working conditions.

[0102] Data acquisition system: X-ray micro-CT system, including a micro-focus X-ray source, a 2048x2048 pixel flat panel detector, and a high-precision three-dimensional translation and rotation table; a synchronous acquisition module is used to fuse acoustic emission signals, pore pressure, vibration acceleration and CT three-dimensional image data.

[0103] Soil erosion collection system: conductive fiber filter screen, which monitors the thickness of particle accumulation by resistance change; thin film piezoelectric sensor, which is arranged in different aperture areas of the filter screen to capture particle size distribution in real time; multi-stage water outlet bucket, which is configured with multi-stage filter screens to separate different particle size erosion particles. The soil erosion collection system includes a filter screen, a thin film piezoelectric sensor and a water outlet bucket. The filter screen is made of conductive fiber composite structure, which monitors the thickness of particle accumulation by resistance change. The thin film piezoelectric sensor is distributed in different aperture areas of the filter screen to sense the pressure fluctuation when the particles are intercepted in real time, to judge the particle size and interception position, to calibrate the particle deposition amount, and to arrange different particle size filter screens on the water outlet bucket to obtain different particle size soil erosion amount. The soil erosion data is obtained based on the data acquisition system.

[0104] In an optional embodiment, the model box is tested under multiple working conditions through seepage of the soil body, data of soil body erosion and particle migration are obtained, and a multiple nonlinear regression model is established, including: the coupling includes seepage and vibration; the seepage in the horizontal and vertical directions causes the soil body to erode and migrate, thereby causing the soil body erosion; the hydraulic gradient of the seepage and the vibration frequency are changed respectively, thereby constructing multiple working conditions; under the multiple working conditions, the data of soil body erosion and particle migration are obtained based on a data acquisition system; and a multiple nonlinear regression model of the soil body seepage and vibration is established according to the data of soil body erosion and particle migration.

[0105] In summary, the method of the present application provides a soil body erosion and particle migration prediction method and system based on acoustic emission monitoring, the test soil body is layered and filled into the model box, and a pore pressure sensor, a 32-channel acoustic emission sensor array arranged on six surfaces, and an acceleration vibration sensor are preset inside the model box; the hydraulic gradient of the model box is controlled by an air compressor and a steady flow valve to obtain the pore water pressure distribution; the model box is vibrated based on a dynamic load control system, and the vibration information of the soil body is obtained according to the acceleration vibration sensor; the erosion process is three-dimensionally imaged and the particle migration inside the soil body is observed through X-ray micro-CT system assisted monitoring; the soil body erosion and particle migration are caused by seepage and vibration, the model box is tested under multiple working conditions, the evolution data of soil body erosion and particle migration are obtained, and a multiple nonlinear regression model is established; a soil body erosion and particle migration prediction model is constructed based on the evolution law of soil body erosion and particle migration combined with a machine learning algorithm, and the prediction of soil body erosion and particle migration is realized. The method of the present application is easy to understand, and provides a theoretical basis and technical support for further development of geotechnical engineering monitoring technology.

[0106] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should be covered in the scope of the claims and the description of the present application.

Claims

1. A method for predicting soil erosion and particle migration based on acoustic emission monitoring, characterized in that, Includes the following steps: Obtain soil mass mix proportion parameters, and prepare soil samples based on the soil mass mix proportion parameters to obtain test soil samples; A multi-field coupling test mechanism is constructed, and multi-field coupling soil tests are conducted on the test soil sample based on the multi-field coupling test mechanism. A soil sample data acquisition device is established, and the soil sample monitoring data during the multi-field coupled soil test is acquired using the soil sample data acquisition device. The acoustic emission characteristic parameters of the test soil sample are obtained based on the soil sample monitoring data, and a multivariate nonlinear regression model is established based on the acoustic emission characteristic parameters. A soil prediction model is constructed by combining the aforementioned multivariate nonlinear regression model, and the prediction of soil erosion and particle migration is realized based on the aforementioned soil prediction model. The construction of the multi-field coupling test mechanism, and the conduct of multi-field coupling soil tests on the test soil samples based on the multi-field coupling test mechanism, includes: The multi-field coupling test mechanism is constructed by adjusting the hydraulic gradient and composite vibration load of the test soil sample. Based on the aforementioned multi-field coupling test mechanism, soil erosion and particle migration are induced in the test soil sample to conduct the multi-field coupling soil test; The process of obtaining acoustic emission characteristic parameters of the test soil sample based on the soil sample monitoring data, and establishing a multivariate nonlinear regression model based on the acoustic emission characteristic parameters, includes: Acoustic emission feature extraction is performed on the soil sample monitoring data to obtain event count rate, energy integral, and dominant frequency, and the event count rate, energy integral, and dominant frequency are used as the acoustic emission feature parameters; A first multivariate nonlinear regression model is constructed based on the event count rate, the energy integral, and the dominant frequency. The acoustic emission feature parameters are normalized to obtain the acoustic emission feature normalized parameters; The correlation analysis results were obtained by performing a single-factor correlation analysis between the acoustic emission characteristic parameters and the soil erosion amount of the test soil sample; A second multivariate nonlinear regression model is established by combining the normalized parameters of the acoustic emission characteristics and the correlation analysis results; The construction of the first multivariate nonlinear regression model based on the event count rate, the energy integral, and the dominant frequency includes: in, This represents the volume of soil particles lost. For event count rate, For energy integral, Main frequency, For hydraulic gradient, Vibration intensity factor This is the error term.

2. The method for predicting soil erosion and particle migration based on acoustic emission monitoring according to claim 1, characterized in that, The process of obtaining soil mass mix design parameters and preparing test soil samples based on these parameters includes: Obtain the mass proportions of sand, clay and silt, and use the mass proportions as the soil mass proportion parameters; The soil compaction density is obtained, and the sand, clay and silt are layered and filled into the model box in combination with the soil mass ratio parameters and the soil compaction density to obtain the test soil sample.

3. The method for predicting soil erosion and particle migration based on acoustic emission monitoring according to claim 1, characterized in that, The establishment of the soil sample data acquisition device, and the use of the soil sample data acquisition device to acquire soil sample monitoring data during the multi-field coupled soil test, includes: The soil sample data acquisition device is constructed based on an acoustic emission sensor array, an X-ray scanning device, a pore pressure sensor, and an acceleration vibration sensor. The acoustic emission signal of the test soil sample was acquired using the acoustic emission sensor array. The three-dimensional image data of the test soil sample were obtained using the X-ray scanning device. The pore pressure of the test soil sample is obtained through the pore pressure sensor. The vibration acceleration of the test soil sample is obtained from the acceleration vibration sensor. The soil sample monitoring data is obtained by fusing the acoustic emission signal, the three-dimensional image data, the pore pressure, and the vibration acceleration.

4. The method for predicting soil erosion and particle migration based on acoustic emission monitoring according to claim 3, characterized in that, The acquisition of three-dimensional image data of the test soil sample using the X-ray scanning device includes: The X-ray scanning device uses a dynamic interval scanning mode to scan the test soil sample to obtain an initial soil sample image; When the acoustic emission signal changes drastically, the dynamic interval scanning mode switches to a high-frequency scanning mode; The initial soil sample image is reconstructed in three dimensions and the data is processed to obtain the three-dimensional image data.

5. The method for predicting soil erosion and particle migration based on acoustic emission monitoring according to claim 1, characterized in that, The construction of a soil prediction model based on the aforementioned multivariate nonlinear regression model, and the prediction of soil erosion and particle migration based on the soil prediction model, includes: Acquire multi-condition data and divide the multi-condition data into a training set, a validation set, and a test set; A hybrid model structure is constructed based on machine learning, and the soil prediction model is established based on the multi-condition data and the hybrid model structure. Based on the soil prediction model, the predicted volume of soil particles lost is obtained, thereby enabling the prediction of soil erosion and particle migration.

6. The method for predicting soil erosion and particle migration based on acoustic emission monitoring according to claim 5, characterized in that, The step of obtaining the predicted volume of soil particle loss based on the soil prediction model to predict soil erosion and particle migration includes: in, This represents the predicted volume of soil particle loss. It is the ReLU activation function. This is the weight matrix. For spatial features, As a time feature, This is a bias term.

7. A soil erosion and particle migration prediction system based on acoustic emission monitoring, characterized in that, The system includes an input device, an output device, a processor, and a memory, which are interconnected. The memory stores a computer program, which includes program instructions. The processor is configured to invoke the program instructions to execute the method for predicting soil erosion and particle migration based on acoustic emission monitoring as described in any one of claims 1-6.

Images