Ground mechanics characteristic parameter interpretation method based on impact penetration sounding

Abstract

The invention provides a ground mechanical property parameter interpretation method based on impact penetration sounding, which comprises the following steps of: carrying an impact penetrometer by an unmanned aerial vehicle to carry out remote survey on a target ground, and carrying out data interpretation on dynamic response data acquired by the penetrometer by utilizing a radial basis function neural network model and a multi-target genetic algorithm; therefore, the mechanical characteristic parameters of the ground can be quickly and accurately obtained. The method is reasonable in conception, the ground mechanical property parameters of the soil can be rapidly and accurately interpreted, and the parameters can be used for equipment trafficability evaluation in a complex geological environment, so that the capability of coping with the complex geological environment is improved.

Description

Technical Field

[0001] The invention relates to the technical field of geological exploration, and in particular to a method for interpreting ground mechanical characteristic parameters based on impact penetration probing. Background Art

[0002] Frequent geological disasters are a major challenge facing my country. Especially under the influence of earthquakes, climate change and human activities, landslides, mudslides and other disasters are becoming more frequent, seriously threatening people's lives and property. These disasters are characterized by strong suddenness, high concealment and great destructiveness. They are extremely difficult to predict, forecast and prevent, often catching people off guard. When a disaster occurs, the geological and topographic information of the affected area is missing, and it is difficult for rescue personnel and equipment to arrive quickly, which greatly reduces the timeliness of rescue.

[0003] Existing geological disaster survey technologies mainly include remote sensing technology and in-situ soil penetration technology. Although remote sensing technology can realize the acquisition of geological information in a long-distance and wide area, its non-contact measurement method cannot accurately invert the mechanical information of the ground soil, and it is easily affected by meteorological conditions in complex disaster environments, and the credibility and resolution of the data are low. In-situ soil penetration technology includes static penetration technology and dynamic penetration technology. Static penetration technology, such as topographic survey vehicles, vehicle-mounted ground penetrators and cone penetrators, can accurately identify rock and soil types and mechanical parameters, but these methods all require manual close-up detection and cannot meet remote and real-time survey needs. Dynamic penetration technology, such as Figure 7 As shown, by using an impact penetration probe mounted on a drone to obtain the penetration velocity under the action of gravity, and interpreting the ground mechanical characteristic parameters based on the penetration resistance and acceleration, the limitations of traditional technologies in conducting remote, wide-area, and rapid soil mechanics surveys in areas that are difficult for personnel to access can be broken through.

[0004] The ground mechanical characteristic parameters cannot be directly measured by the impact penetration tester. At present, the research work on the interpretation of ground mechanical characteristic parameters by researchers at home and abroad mainly focuses on the research method of static penetration test, while the research on the interpretation of ground mechanical characteristic parameters based on impact penetration test is rarely mentioned. Summary of the invention

[0005] In response to the technical problems existing in the above-mentioned background technology, the present invention proposes a method for interpreting ground mechanical property parameters based on impact penetration probing. The method has a reasonable concept and can quickly and accurately interpret the mechanical parameters of the soil. These parameters can be used to evaluate the equipment's passability in complex geological environments, thereby improving the ability to cope with complex geological environments.

[0006] In order to solve the above technical problems, the present invention provides a method for interpreting ground mechanical characteristic parameters based on impact penetration probing, which uses an unmanned aerial vehicle carrying an impact penetrometer to conduct remote survey of the target ground, and uses a radial basis function neural network model and a multi-objective genetic algorithm to interpret the dynamic response data collected by the penetrometer, so as to quickly and accurately obtain the ground mechanical characteristic parameters.

[0007] The method for interpreting ground mechanical characteristic parameters based on impact penetration testing specifically comprises the following steps:

[0008] (1) Data Collection

[0009] First, an impact penetration test is carried out on the target ground. The drone carries the impact penetrometer and flies to the target ground. After reaching the predetermined height, the drone drops the impact penetrometer into the ground in a free-fall manner to measure the time-course data of the cone tip resistance and acceleration during the penetration of the impact penetrometer into the ground. Then, the data acquisition and processing module embedded in the impact penetrometer performs preliminary data processing to obtain high-quality time-course data of the cone tip resistance and acceleration.

[0010] (2) Database construction

[0011] The Latin hypercube sampling method is used to generate ground mechanical characteristic parameter samples in the soil parameter space. The CEL finite element method is used to simulate the parameters in the ground mechanical characteristic parameter samples to generate the dynamic response data of the penetrometer under different soil conditions, thereby constructing a database covering a wide range of soil conditions.

[0012] (3) Radial Basis Neural Network Model Training

[0013] The constructed database is used to train the radial basis function neural network model to establish the mapping relationship between the ground mechanical characteristic parameters and the penetrometer dynamic response data;

[0014] (4) Interpretation of ground mechanical characteristic parameters

[0015] The radial basis function neural network model is combined with a multi-objective genetic algorithm to achieve an accurate interpretation of the ground mechanical characteristic parameters.

[0016] The method for interpreting ground mechanical characteristic parameters based on impact penetration testing, wherein the specific process of generating samples of ground mechanical characteristic parameters by using the Latin hypercube sampling method in step (2) is as follows:

[0017] (2.1) Determination of parameter range

[0018] Determine the value range of each ground mechanical characteristic parameter, assuming that there are P parameters in total, forming a P-dimensional parameter space;

[0019] (2.2) Divide the parameter range into equal parts

[0020] The value range of each ground mechanical characteristic parameter is divided into N equally spaced intervals in its dimension, that is, the value range corresponding to each ground mechanical characteristic parameter is divided into N sub-intervals;

[0021] (2.3) Randomly select sample values ​​for each interval

[0022] For each dimension of the ground mechanical characteristic parameter, a value is randomly selected in each sub-interval to ensure that all intervals on this dimension can obtain a sample value;

[0023] (2.4) Sample arrangement

[0024] After the N samples of each ground mechanical characteristic parameter are randomly arranged, they form P-dimensional samples respectively. The sample points finally generated will cover the entire ground mechanical characteristic parameter space, and the samples in each dimension are evenly distributed.

[0025] The method for interpreting ground mechanical characteristic parameters based on impact penetration testing, wherein the specific process of training the radial basis function neural network model using the constructed database in step (3) is as follows:

[0026] (3.1) Input the sample points of the ground mechanical characteristic parameters generated by Latin hypercube sampling into the input layer of the radial basis neural network model;

[0027] (3.2) Using CEL finite element simulation, simulate each set of ground mechanical characteristic parameter sample points to obtain the dynamic response data of the penetrometer, and use these data as the output of the radial basis function neural network model;

[0028] (3.3) Activation function for constructing the structure of radial basis neural network model

[0029] The structure of the radial basis neural network model consists of an input layer, a hidden layer, and an output layer. The hidden layer uses the radial basis function as the activation function:

[0030]

[0031] In the above formula (1), r is the Euclidean distance from the input to the center of the hidden layer; σ is the width parameter of the Gaussian function.

[0032] The method for interpreting ground mechanical characteristic parameters based on impact penetration probing, wherein: the preliminary data processing performed in step (1) includes filtering and noise reduction.

[0033] The method for interpreting ground mechanical characteristic parameters based on impact penetration probing, wherein the process of training the radial basis function neural network model in step (3) is as follows: the radial basis function neural network model uses the ground mechanical characteristic parameters as input and the time history data of the cone tip resistance and acceleration as output. At the same time, in order to verify the accuracy of the radial basis function neural network model, a part of the samples are randomly selected from the database as the test set and the remaining part is used as the training set; the normalized root mean square error and the determination coefficient R are introduced. 2 To evaluate the prediction accuracy of the radial basis neural network model, the prediction accuracy of the radial basis neural network model is measured by the following two evaluation indicators:

[0034] One evaluation metric is the normalized root mean square error (RAAE):

[0035]

[0036] In the above formula (2), y pred,i is the model prediction value, y ture,i is the true value, n is the number of samples, y max and min are the maximum and minimum values ​​of the true value respectively;

[0037] Another evaluation indicator is the coefficient of determination R 2 :

[0038]

[0039] In the above formula (3), y is the mean of the true values;

[0040] If the prediction accuracy of the radial basis neural network model on the test set does not meet the requirements, then increase the training data and adjust the structure and model parameters of the radial basis neural network model until the prediction accuracy of the radial basis neural network model reaches the expected standard.

[0041] The method for interpreting the ground mechanical characteristic parameters based on impact penetration probing, wherein the specific process of step (4) is: taking the difference between the cone tip resistance and acceleration time history data predicted by the radial basis function neural network model and the experimental observation data as the objective function, minimizing the objective function through a multi-objective genetic algorithm, determining the ground mechanical characteristic parameters and interpreting them; in order to interpret the ground mechanical characteristic parameters, it is first necessary to establish an objective function to minimize the difference between the radial basis function neural network model prediction value and the experimental observation value, and its expression is:

[0042]

[0043] In formula (4), pred_a t and pred_F tare the cone tip resistance and acceleration time history values ​​predicted by the radial basis function neural network model; tar_a t and tar_F t They are the experimental observation data of cone tip resistance and acceleration respectively;

[0044] Through non-dominated sorting, crowding calculation and elite retention strategy, the global optimality of the Pareto optimal solution set formed after the iteration of the multi-objective genetic algorithm is ensured to accurately interpret the ground mechanical characteristic parameters. The specific interpretation process is:

[0045] (4.1) Initialize the population of the genetic algorithm

[0046] An initial population is randomly generated, each individual represents a set of ground mechanical characteristic parameters; the gene of each individual is represented by a parameter vector X = [φ, c, E, ν, ρ];

[0047] (4.2) Objective function calculation

[0048] For each individual, the radial basis function neural network model is used to predict the cone tip resistance and acceleration time history data, and the objective function value corresponding to the individual is calculated by the following formula (5):

[0049]

[0050] (4.3) Non-dominated sorting

[0051] Sort the individuals in the population according to Pareto superiority to form multiple non-dominated levels, where individuals in lower levels are superior to those in higher levels;

[0052] (4.4) Crowding calculation

[0053] Calculate the crowding degree of each individual to measure its density in the target space. Crowding is used to maintain the diversity of the population. The specific calculation formula is:

[0054]

[0055] in, and is the function value of the adjacent solution of the ith solution on the jth objective function, and are the maximum and minimum values ​​of the jth objective function respectively;

[0056] (4.5) Elite Retention Strategy

[0057] In each generation of the genetic algorithm population, non-dominated solutions are retained first, and individuals are selected in areas with high congestion to ensure that the population develops towards the Pareto frontier;

[0058] (4.6) Crossover and mutation

[0059] New individuals are generated through crossover and mutation operations in genetic algorithms, and these new individuals will replace inferior individuals in subsequent generations of iterations;

[0060] (4.7) Termination conditions

[0061] When the population reaches the maximum number of generations or the convergence change of the objective function value is less than the preset threshold, the genetic algorithm terminates and the optimal ground mechanical characteristic parameter vector is selected according to the decision criterion.

[0062] By adopting the above technical solution, the present invention has the following beneficial effects:

[0063] The method for interpreting the geomechanical characteristic parameters based on impact penetration probing of the present invention is well-conceived and can quickly and accurately interpret the geomechanical characteristic parameters of the soil, including cohesion c, internal friction angle φ, elastic modulus E, Poisson's ratio ν and density ρ, etc. The average interpretation error of multiple working conditions is less than 25%, and the time taken for conventional computer single working condition parameter interpretation is less than 1 minute. The present invention can be applied to remote and rapid survey of geomechanics in disaster areas and areas that are difficult for personnel to reach.

[0064] The present invention constructs a physical model of impact penetration soil, and develops a method for interpreting ground mechanical characteristic parameters based on a radial basis function neural network model (RBF) and a multi-objective genetic algorithm (NSGA-II). Impact penetration experiments on sand and silty clay at different water contents show that, according to the method of the present invention, the ground mechanical characteristic parameters of the soil, including cohesion, internal friction angle, elastic modulus, Poisson's ratio and density, can be accurately interpreted through cone tip resistance and acceleration time history curves. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0066] Figure 1 It is a flow chart of the method for interpreting ground mechanical characteristic parameters based on impact penetration probing of the present invention;

[0067] Figure 2 It is a flow chart of the radial basis neural network model training involved in the method for interpreting ground mechanical characteristic parameters based on impact penetration probing of the present invention;

[0068] Figure 3This is a result diagram of the verification of the radial basis neural network model involved in the ground mechanical characteristic parameter interpretation method based on impact penetration probing of the present invention;

[0069] Figure 4 A flow chart of interpretation of ground mechanical characteristic parameters involved in the ground mechanical characteristic parameter interpretation method based on impact penetration probing of the present invention;

[0070] Figure 5 A flow chart of the NSGA-II algorithm involved in the method for interpreting ground mechanical characteristic parameters based on impact penetration probing of the present invention;

[0071] Figure 6 It is a schematic diagram of the CEL model involved in the method for interpreting ground mechanical characteristic parameters based on impact penetration probing of the present invention;

[0072] Figure 7 This is a site photo of the free-fall dynamic penetration technology. DETAILED DESCRIPTION

[0073] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] The present invention is further explained below in conjunction with specific implementation modes.

[0075] like Figure 1 As shown, this embodiment provides a method for interpreting ground mechanical characteristic parameters based on impact penetration probing, which uses an unmanned aerial vehicle carrying an impact penetrometer to conduct remote survey of the target ground (soil), and uses a radial basis function neural network model (i.e., RBF model) and a multi-objective genetic algorithm (NSGA-II) to interpret the dynamic response data collected by the penetrometer, so as to quickly and accurately obtain the ground mechanical characteristic parameters.

[0076] S100, Data Collection

[0077] First, an impact penetration test is carried out on the target ground (soil). The UAV carries the impact penetrometer and flies above the target ground. After reaching the predetermined height, the impact penetrometer penetrates into the ground in a free-fall manner to measure the cone tip resistance and acceleration time-history data during the impact penetrometer's penetration of the ground. Then, the data acquisition and processing module embedded in the impact penetrometer performs preliminary data processing, including filtering and noise reduction, to obtain high-quality cone tip resistance and acceleration time-history data.

[0078] S200, database construction

[0079] In order to establish the relationship between the dynamic response data of the penetrometer and the ground mechanical characteristic parameters, the present invention adopts the Latin hypercube sampling method to generate ground mechanical characteristic parameter samples in the soil parameter space (Latin hypercube sampling is a statistical method for generating sample points in a multidimensional parameter space, which can ensure that the sample values ​​of each dimension are evenly distributed, thereby avoiding the sample aggregation problem that may occur in traditional random sampling), and adopts the CEL finite element method to simulate the parameters in the ground mechanical characteristic parameter samples to generate the dynamic response data of the penetrometer under different soil conditions, thereby constructing a database covering a wide range of soil working conditions; the specific steps are as follows:

[0080] S201. Determine parameter range

[0081] For each ground mechanical characteristic parameter, such as cohesion c, internal friction angle φ, elastic modulus E, Poisson's ratio ν and density ρ, their value ranges are determined; assuming that the number of parameters is P, that is, a P-dimensional parameter space.

[0082] S202, dividing the parameter range into equal parts

[0083] The value range of each ground mechanics characteristic parameter is divided into N equally spaced intervals in its dimension, that is, the value range corresponding to each ground mechanics characteristic parameter is divided into N sub-intervals.

[0084] S203, randomly select sample values ​​in each interval

[0085] For each dimension of the ground mechanical characteristic parameter, a value is randomly selected in each subinterval to ensure that all intervals on this dimension can obtain a sample value.

[0086] S204, sample arrangement

[0087] After N samples of each ground mechanical characteristic parameter are randomly arranged, they are respectively composed of P-dimensional samples; the sample points finally generated will cover the entire ground mechanical characteristic parameter space, and the samples in each dimension are evenly distributed.

[0088] The purpose of using Latin hypercube sampling is to ensure that the sample points are distributed as evenly as possible in the parameter space and avoid the problem of sample point clustering caused by random sampling. Subsequently, each set of soil parameters is simulated by using the CEL (Coupled Eulerian-Lagrangian) finite element method (the present invention adopts the coupled Eulerian-Lagrangian (CEL) finite element method, which is particularly suitable for dealing with large deformation problems and can accurately simulate the behavior of soil under dynamic loading conditions; the CEL finite element model used in the present invention is as follows Figure 6As shown in the figure, the cone in the model adopts Lagrangian unit with a diameter of 35.7 mm; the soil adopts Euler unit, and the size of the soil model is consistent with the size of the laboratory soil trough box, 1 meter wide, 1.5 meters long, and 1.2 meters high; in order to ensure numerical stability and calculation accuracy, the size of the Euler unit near the cone is selected as 2mm. ), the dynamic response data of the penetrometer under different soil conditions are generated, thus constructing a database covering a wide range of soil conditions for the subsequent training and verification of the radial basis function neural network model.

[0089] S300, radial basis neural network model training

[0090] The radial basis neural network model is trained using the constructed database; the specific training process is as follows (e.g. Figure 2 shown):

[0091] S301, input data: input the sample points of the ground mechanical characteristic parameters generated by Latin hypercube sampling into the input layer of the radial basis neural network model;

[0092] S302, output data: using CEL finite element simulation, simulate each set of ground mechanical characteristic parameter sample points to obtain the dynamic response data of the penetrometer, and use these data as the output of the radial basis function neural network model;

[0093] S303, constructing an activation function of the structure of the radial basis neural network model: The structure of the radial basis neural network model consists of an input layer, a hidden layer and an output layer, and the hidden layer uses a radial basis function (Gaussian function) as an activation function:

[0094]

[0095] In the above formula (1), r is the Euclidean distance from the input to the center of the hidden layer, and σ is the width parameter of the Gaussian function;

[0096] The mapping relationship between the ground mechanical characteristic parameters and the dynamic response data of the penetrometer is established. The radial basis function neural network model uses the ground mechanical characteristic parameters as input and the cone tip resistance and acceleration time history data as output. At the same time, in order to verify the accuracy of the radial basis function neural network model, a part of the samples are randomly selected from the database as the test set, and the rest are used as the training set. The normalized root mean square error (RAAE) and the determination coefficient R are introduced. 2 To evaluate the prediction accuracy of the radial basis neural network model; the prediction accuracy of the radial basis neural network model is measured by the following two evaluation indicators:

[0097] One evaluation metric is the normalized root mean square error (RAAE):

[0098]

[0099] In the above formula (2), y pred,i is the model prediction value, y ture,i is the true value, n is the number of samples, y max and min are the maximum and minimum values ​​of the true value respectively;

[0100] Another evaluation indicator is the coefficient of determination (R 2 ):

[0101]

[0102] In the above formula (3), is the mean of the true values;

[0103] If the prediction accuracy of the radial basis neural network model on the test set does not meet the requirements, then increase the training data and adjust the structure and model parameters of the radial basis neural network model until the prediction accuracy of the radial basis neural network model reaches the expected standard.

[0104] In this embodiment, the CEL finite element method is used to generate 4000 sets of penetrometer dynamic response time history data under different soil conditions through Latin hypercube sampling; based on these data, the radial basis function neural network model is trained. At the same time, in order to evaluate the prediction performance of the trained RBF, an additional 100 CEL simulations are performed, and these simulation results are compared with the prediction results of the RBF model. Figure 3 The prediction results of maximum acceleration, maximum penetration resistance and penetration depth are shown. The results show that the trained RBF model can effectively predict the dynamic response characteristics of the impact penetration process, thereby establishing a mapping relationship between the ground mechanical characteristic parameters and the dynamic response time history data of the penetrometer.

[0105] S400, interpretation of ground mechanical characteristic parameters

[0106] The present invention combines the radial basis neural network model (RBF) with Figure 5 The multi-objective genetic algorithm (NSGA-II) shown in Figure 1 is combined with the ground mechanical characteristic parameters to achieve accurate interpretation of the ground mechanical characteristic parameters. Specifically, the difference between the cone tip resistance and acceleration time history data predicted by the RBF model and the experimental observation data is used as the objective function, and the ground mechanical characteristic parameters are determined by minimizing the objective function through the NSGA-II algorithm. In order to interpret the ground mechanical characteristic parameters, it is first necessary to establish an objective function to minimize the difference between the radial basis function neural network model prediction value and the experimental observation value. Its mathematical expression is:

[0107]

[0108] In formula (4), pred_a tand pred_F t are the cone tip resistance and acceleration time history values ​​predicted by the radial basis function neural network model; tar_a t and tar_F t They are the experimental observation data of cone tip resistance and acceleration respectively;

[0109] The NSGA-II algorithm ensures the global optimality of the Pareto optimal solution set (i.e., the set containing all non-dominated individuals) formed after iteration of the multi-objective genetic algorithm through non-dominated sorting, crowding calculation, and elite retention strategy, thereby accurately interpreting the ground mechanical characteristic parameters. The specific process is as follows (e.g. Figure 4 shown):

[0110] S401. Initialize the population of genetic algorithm

[0111] An initial population is randomly generated, each individual represents a set of soil geomechanical property parameters (for example, internal friction angle φ, cohesion c, elastic modulus E, Poisson's ratio ν and density ρ, etc.); the gene of each individual is represented by a parameter vector X = [φ, c, E, ν, ρ].

[0112] S402, objective function calculation

[0113] For each individual, the radial basis function neural network model is used to predict the cone tip resistance and acceleration time history data, and the objective function value corresponding to the individual is calculated by the following formula (5):

[0114]

[0115] S403, non-dominated sorting

[0116] The individuals in the population (here the population includes the initial population and the subsequent generation population) are sorted according to Pareto superiority to form multiple non-dominated levels. Individuals in lower levels are superior to individuals in higher levels.

[0117] S404. Congestion calculation

[0118] In order to maintain the diversity of the population, the crowding degree of each individual is calculated to measure its density in the target space; the crowding degree is used to maintain the diversity of the population, and the specific calculation formula is:

[0119]

[0120] Among them, in the above formula (6) and is the function value of the adjacent solution of the ith solution on the jth objective function, and are the maximum and minimum values ​​of the j-th objective function respectively.

[0121] S405, Elite Retention Strategy

[0122] In each generation of the genetic algorithm population, non-dominated solutions are retained first, and individuals are selected in areas with greater crowding to ensure that the population develops towards the Pareto frontier.

[0123] S406, Crossover and Mutation

[0124] New individuals are generated through crossover and mutation operations in the genetic algorithm, and these new individuals will replace inferior individuals in subsequent generations of iterations.

[0125] S407, Termination Condition

[0126] When the population reaches the maximum generation, or the convergence change of the objective function value is less than the preset threshold, the algorithm terminates; the optimal soil geomechanical property parameter vector is selected according to the decision criteria.

[0127] The interpretation effect of ground mechanical characteristic parameters is as follows:

[0128] The soil internal friction angle, cohesion and other parameters were calibrated through indoor undrained triaxial tests and compared with the interpretation results to verify the reliability of the interpretation method. The results show that the average interpretation error of the method under multiple working conditions is less than 25%, and the conventional computer single working condition parameter interpretation takes less than 1 minute.

[0129] Table 1 Interpretation verification

[0130]

[0131] The present invention can quickly and accurately interpret ground mechanical characteristic parameters, including density, internal friction angle, cohesion, elastic modulus and Poisson's ratio, etc. These parameters can be used to evaluate the equipment trafficability in complex geological environments, thereby improving the ability to cope with complex geological environments.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for interpreting ground mechanical characteristic parameters based on impact penetration probing, characterized in that: The target ground is remotely surveyed by carrying an impact penetrometer on an unmanned aerial vehicle, and the dynamic response data collected by the penetrometer are interpreted using a radial basis function neural network model and a multi-objective genetic algorithm to quickly and accurately obtain the ground mechanical characteristic parameters.

2. The method for interpreting ground mechanical characteristic parameters based on impact penetration probing according to claim 1, characterized in that: The specific steps include: (1) Data Collection First, an impact penetration test is carried out on the target ground. The drone carries the impact penetrometer and flies to the target ground. After reaching the predetermined height, the drone drops the impact penetrometer into the ground in a free-fall manner to measure the time-course data of the cone tip resistance and acceleration during the penetration of the impact penetrometer into the ground. Then, the data acquisition and processing module embedded in the impact penetrometer performs preliminary data processing to obtain high-quality time-course data of the cone tip resistance and acceleration. (2) Database construction The Latin hypercube sampling method is used to generate ground mechanical characteristic parameter samples in the soil mechanical parameter space. The CEL finite element method is used to simulate the parameters in the ground mechanical characteristic parameter samples to generate the dynamic response data of the penetrometer under different soil conditions, thereby constructing a database covering a wide range of soil conditions. (3) Radial Basis Neural Network Model Training The constructed database is used to train the radial basis function neural network model to establish the mapping relationship between the ground mechanical characteristic parameters and the penetrometer dynamic response data; (4) Interpretation of ground mechanical characteristic parameters The radial basis function neural network model is combined with a multi-objective genetic algorithm to achieve accurate interpretation of the ground mechanical characteristic parameters; the ground mechanical characteristic parameters include cohesion c, internal friction angle φ, elastic modulus E, Poisson's ratio ν and density ρ.

3. The method for interpreting ground mechanical characteristic parameters based on impact penetration probing according to claim 2, characterized in that: The specific process of using the Latin hypercube sampling method to generate samples of the ground mechanical characteristic parameters in step (2) is as follows: (2.1) Determination of parameter range Determine the value range of each ground mechanical characteristic parameter, assuming that there are P parameters in total, forming a P-dimensional parameter space; (2.2) Divide the parameter range into equal parts The value range of each ground mechanical characteristic parameter is divided into N equally spaced intervals in its dimension, that is, the value range corresponding to each ground mechanical characteristic parameter is divided into N sub-intervals; (2.3) Randomly select sample values ​​for each interval For each dimension of the ground mechanical characteristic parameter, a value is randomly selected in each sub-interval to ensure that all intervals on this dimension can obtain a sample value; (2.4) Sample arrangement After the N samples of each ground mechanical characteristic parameter are randomly arranged, they form P-dimensional samples respectively. The sample points finally generated will cover the entire ground mechanical characteristic parameter space, and the samples in each dimension are evenly distributed.

4. The method for interpreting ground mechanical characteristic parameters based on impact penetration probing according to claim 2, characterized in that: The specific process of training the radial basis neural network model using the constructed database in step (3) is as follows: (3.1) Input the sample points of the ground mechanical characteristic parameters generated by Latin hypercube sampling into the input layer of the radial basis neural network model; (3.2) Using CEL finite element simulation, simulate each set of ground mechanical characteristic parameter sample points to obtain the dynamic response data of the penetrometer, and use these data as the output of the radial basis function neural network model; (3.3) Activation function for constructing the structure of radial basis neural network model The structure of the radial basis neural network model consists of an input layer, a hidden layer, and an output layer. The hidden layer uses the radial basis function as the activation function: In the above formula (1), r is the Euclidean distance from the input to the center of the hidden layer; σ is the width parameter of the Gaussian function.

5. The method for interpreting ground mechanical characteristic parameters based on impact penetration probing according to claim 2, characterized in that: The preliminary data processing performed in step (1) includes filtering and noise reduction.

6. The method for interpreting ground mechanical characteristic parameters based on impact penetration probing according to claim 2, characterized in that: The process of training the radial basis function neural network model in step (3) is as follows: the radial basis function neural network model uses the ground mechanical characteristic parameters as input and the time history data of the cone tip resistance and acceleration as output. At the same time, in order to verify the accuracy of the radial basis function neural network model, a part of the samples are randomly selected from the database as the test set and the remaining part is used as the training set; the normalized root mean square error and the determination coefficient R are introduced. 2 To evaluate the prediction accuracy of the radial basis neural network model, the prediction accuracy of the radial basis neural network model is measured by the following two evaluation indicators: One evaluation metric is the normalized root mean square error (RAAE): In the above formula (2), y pred,i is the model prediction value, y ture,i is the true value, n is the number of samples, y max and min are the maximum and minimum values ​​of the true value respectively; Another evaluation indicator is the coefficient of determination R 2 : In the above formula (3), y is the mean of the true values; If the prediction accuracy of the radial basis neural network model on the test set does not meet the requirements, then increase the training data and adjust the structure and model parameters of the radial basis neural network model until the prediction accuracy of the radial basis neural network model reaches the expected standard.

7. The method for interpreting ground mechanical characteristic parameters based on impact penetration probing according to claim 2, characterized in that: The specific process of step (4) is as follows: taking the difference between the cone tip resistance and acceleration time history data predicted by the radial basis function neural network model and the experimental observation data as the objective function, minimizing the objective function through a multi-objective genetic algorithm, determining the ground mechanical characteristic parameters and interpreting them; in order to interpret the ground mechanical characteristic parameters, it is first necessary to establish an objective function to minimize the difference between the radial basis function neural network model prediction value and the experimental observation value, and its expression is: In formula (4), pred_a t and pred_F t are the cone tip resistance and acceleration time history values ​​predicted by the radial basis function neural network model; tar_a t and tar_F t They are the experimental observation data of cone tip resistance and acceleration respectively; Through non-dominated sorting, crowding calculation and elite retention strategy, the global optimality of the Pareto optimal solution set formed after the iteration of the multi-objective genetic algorithm is ensured to accurately interpret the ground mechanical characteristic parameters. The specific interpretation process is: (4.1) Initialize the population of the genetic algorithm An initial population is randomly generated, each individual represents a set of ground mechanical characteristic parameters; the gene of each individual is represented by a parameter vector X = [φ, c, E, ν, ρ]; (4.2) Objective function calculation For each individual, the radial basis function neural network model is used to predict the cone tip resistance and acceleration time history data, and the objective function value corresponding to the individual is calculated by the following formula (5): (4.3) Non-dominated sorting Sort the individuals in the population according to Pareto superiority to form multiple non-dominated levels, where individuals in lower levels are superior to those in higher levels; (4.4) Crowding calculation Calculate the crowding degree of each individual to measure its density in the target space. Crowding is used to maintain the diversity of the population. The specific calculation formula is: in, and is the function value of the adjacent solution of the ith solution on the jth objective function, and are the maximum and minimum values ​​of the jth objective function respectively; (4.5) Elite Retention Strategy In each generation of the genetic algorithm population, non-dominated solutions are retained first, and individuals are selected in areas with high congestion to ensure that the population develops towards the Pareto frontier; (4.6) Crossover and mutation New individuals are generated through crossover and mutation operations in genetic algorithms, and these new individuals will replace inferior individuals in subsequent generations of iterations; (4.7) Termination conditions When the population reaches the maximum number of generations or the convergence change of the objective function value is less than the preset threshold, the genetic algorithm terminates and the optimal soil geomechanical property parameter vector is selected according to the decision criteria.

Images