Geotechnical engineering catastrophe main control factor determination and safety evaluation method and system

Abstract

The invention discloses a geotechnical engineering catastrophe main control factor determination and safety evaluation method and system, and relates to the technical field of engineering safety evaluation, and the method comprises the steps: constructing a full convolutional network model for learning, combining with sparse drilling data, carrying out the interpolation prediction of an unknown region stratum image, and outputting a stratum profile prediction image; based on a BP neural network of a simulated annealing algorithm, inverting constitutive parameters through soil mass mechanical test data; integrating the stratum profile image, the soil constitutive parameters and the existing upper building information to carry out refined modeling; carrying out refined simulation under the influence of multiple factors on each potential factor, and analyzing and determining the weight of each factor from a simulation result by adopting an entropy weight method; and a multi-factor catastrophe evaluation model is established by using a Topsis method, quantitative safety evaluation is realized, and an optimal construction scheme is obtained. According to the scheme, catastrophe factor identification and construction scheme optimization are realized through an intelligent inversion and fine modeling method, and the accuracy of catastrophe prediction and the scientificity of safety decision are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engineering safety evaluation, and particularly relates to a method and system for determining a main control factor of geotechnical engineering disaster and safety evaluation. BACKGROUND

[0002] In the field of geotechnical engineering, artificial intelligence methods have been deeply integrated with geotechnical engineering design, which is profoundly changing the development trajectory of geotechnical engineering. With the continuous advancement of modern engineering construction, the challenges faced by geotechnical engineering are increasingly complex and diverse. On the one hand, the scale of engineering is constantly expanding, such as large-scale water conservancy projects, high-rise building deep foundation pits, highway slopes, and other projects, which have very high requirements for the stability and safety of rock-soil bodies. On the other hand, the properties of rock-soil bodies are complex and variable, and their mechanical behavior is influenced by multiple factors, including geological conditions (stratum lithology, rock-soil body structure, etc.), environmental factors (groundwater, rainfall, earthquakes, etc.), and engineering factors (excavation methods, support measures, etc.). The uncertainty of these factors makes geotechnical engineering problems highly nonlinear and dynamic, and traditional analysis methods and empirical design gradually show limitations in dealing with complex geotechnical engineering problems.

[0003] In the field of geotechnical engineering, although the integration of artificial intelligence methods with geotechnical engineering design and the development of fine numerical methods provide new ideas and methods for determining the main control factor of geotechnical disaster and safety evaluation, there are still many challenges and problems, such as: (1) coarse test analysis cannot provide fine analysis methods; (2) existing engineering design lacks quantitative evaluation based on multiple factors. SUMMARY

[0004] The present application provides a method for determining the main control factor of geotechnical engineering disaster and safety evaluation, which solves the problems of coarse test analysis and lack of quantitative evaluation based on multiple factors.

[0005] Technical solution: In a first aspect, the present application provides a method for determining the main controlling factor of geotechnical engineering disasters and safety evaluation, which comprises: constructing a full convolutional network model to learn the features of known stratum images, combining sparse borehole data to interpolate and predict the stratum images of unknown areas, and outputting the standardized stratum profile prediction images; using a BP neural network based on a simulated annealing algorithm to train the nonlinear mapping relationship between the constitutive parameters of the soil body and the stress-strain characteristics of the sample, and to invert the constitutive parameters of the soil body through soil mechanics test data; integrating the predicted stratum profile prediction images, the inverted soil body constitutive parameters, and the existing upper building information into numerical software for fine modeling to realize fine simulation of the response of geotechnical engineering construction; selecting each potential factor affecting disasters for specific geotechnical engineering, fine simulation under the influence of multiple factors, using the entropy weight method to analyze and determine the weight of each factor from the simulation results to obtain a disaster factor rating table matched therewith; using the Topsis method to establish a multi-factor disaster evaluation model to realize quantitative safety evaluation of geotechnical engineering and obtain the optimal construction scheme.

[0006] Further, the method outputs standardized stratum profile prediction images, including the following steps

[0007] The local sample image is mapped into a plurality of feature mapping images by performing convolution operation processing in the local sample image in the order of from left to right and from top to bottom through a preset step size.

[0008] The unknown area stratum image is interpolated using an extended normal equation simulation algorithm, and standardized to obtain the standardized stratum profile prediction image.

[0009] Further, the method uses an extended normal equation simulation algorithm to interpolate the unknown area stratum image, including the following steps,

[0010] The first round of interpolation is performed on the blank grid between the N original boreholes in the task area, and when all the blank grids matching the feature template are interpolated, the first round of interpolation is ended;

[0011] The second round of interpolation is performed on the blank grid between the K pixel grid strips with soil layer information in the task area, and the interpolation is completed until all the blank grids matching the feature template are interpolated.

[0012] Repeat the above steps for multiple rounds of interpolation until all the blank grids are filled.

[0013] Further, the BP neural network of the simulated annealing algorithm refers to inputting the soil sample mechanics test results into the BP neural network in the input layer; the hidden layer contains the nonlinear mapping relationship between the constitutive parameters and the stress-strain, strain body characteristics; and the output layer analyzes and gives the constitutive parameters.

[0014] Further, the method selects various potential factors affecting the disaster according to the specific geotechnical engineering, and carries out fine simulation under the influence of multiple factors, including the following steps,

[0015] According to the construction characteristics of the geotechnical engineering, various potential factors affecting the disaster are selected;

[0016] The predicted stratum profile image, the inverted soil mechanical parameters and the upper building information are integrated into the finite element software for calculation to simulate the stratum response law of the construction section.

[0017] Further, the method adopts the entropy weight method to analyze and determine the weight of each factor from the simulation results, including the following steps

[0018] Based on the data of various potential factors of the disaster, a judgment matrix of m schemes and n evaluation indexes is constructed;

[0019] The extreme value method is used to standardize the original data to obtain a standardized judgment matrix;

[0020] The normalized matrix is normalized, and the entropy value of each evaluation index is calculated.

[0021] Further, the method uses the Topsis method to establish a multi-factor disaster evaluation model, realizes quantitative safety evaluation of the geotechnical engineering, and obtains the optimal construction scheme, including the following steps,

[0022] According to the entropy value of each evaluation index, an initial matrix is formed, and normalized processing is carried out to obtain a normalized matrix;

[0023] The combination weight of each evaluation index is determined by using the entropy weight method and the analytic hierarchy process;

[0024] Based on the weight factor of each evaluation index, a weighted judgment matrix is constructed combined with the normalized matrix;

[0025] The positive and negative ideal state solutions are obtained, and the Euclidean distance between the entropy value of the evaluation index of each scheme disaster risk and the positive and negative ideal state solutions is calculated by using the Euclidean formula;

[0026] The Topsis evaluation value of the safety score of each scheme disaster risk is calculated, the schemes are sorted according to the size of the Topsis evaluation value of the safety score of each scheme disaster risk, and the optimal construction scheme is selected.

[0027] In a second aspect, the present application provides a geotechnical engineering disaster master control factor determination and safety evaluation system, comprising: an image generation module, configured to construct a full convolution network model to learn features of known stratum images, to combine sparse drilling data to interpolate and predict stratum images of unknown areas, and to output a standardized stratum profile prediction image; a parameter inversion module, configured to train a nonlinear mapping relationship between soil body constitutive parameters and sample stress-strain characteristics based on a BP neural network of a simulated annealing algorithm, and to invert the constitutive parameters of the soil body through soil mechanics test data; a refined modeling module, configured to integrate the predicted stratum profile prediction image, the inverted soil body constitutive parameters, and existing upper building information into numerical software to perform refined modeling, and to realize refined simulation of geotechnical engineering construction responses; a weight analysis module, configured to select various potential factors affecting disasters for specific geotechnical engineering, to perform refined simulation under the influence of multiple factors, to analyze and determine the weights of various factors from the simulation results using an entropy weight method, and to obtain a disaster factor rating table matched therewith; and a safety evaluation module, configured to establish a multi-factor disaster evaluation model using a Topsis method, to realize quantitative safety evaluation of geotechnical engineering, and to obtain an optimal construction scheme.

[0028] In a third aspect, the present application provides a computer device, comprising a memory and a processor, and the memory stores a computer program, wherein: the computer program is executed by the processor to realize any step of the geotechnical engineering disaster master control factor determination and safety evaluation method according to the first aspect of the present application.

[0029] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, wherein: the computer program is executed by the processor to realize any step of the geotechnical engineering disaster master control factor determination and safety evaluation method according to the first aspect of the present application.

[0030] Beneficial effects: Compared with the prior art, the significant advantages of the present application are: by constructing a BP neural network model based on simulated annealing algorithm optimization to invert the soil constitutive parameters, the rapid and high-precision inversion of key constitutive parameters is realized, the accuracy and efficiency of soil material parameter input are effectively improved, and accurate and reliable constitutive model support is provided for numerical simulation; by integrating stratum image, constitutive parameters and structure information for fine modeling and simulation, the image data is directly converted into model grid, the seamless connection from image intelligent recognition to finite element modeling is realized, the stratum response process under construction disturbance is truly reproduced through high-precision simulation, and the timeliness and spatial coverage of field monitoring are made up; by multi-factor simulation and entropy weight identification based on the fine simulation results, the differentiation logic of "main factor" and "secondary factor" under multi-factor conditions is clarified, which has strong interpretability, the subjective bias of artificial weighting is avoided by entropy weight method, the objectivity of the evaluation system is improved, and the weight result further participates in Topsis scheme evaluation and can also independently constitute a disaster risk grade early warning mechanism. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 Flow chart of the method for determining and safety evaluation of geotechnical engineering disasters in embodiment 1;

[0032] Figure 2 Structural schematic diagram of the full convolution network in the present application in embodiment 1;

[0033] Figure 3 Stratum feature learning flow in embodiment 1;

[0034] Figure 4 Stratum profile information interpolation flow based on the full convolution model in embodiment 1;

[0035] Figure 5 Application program flow of stratum information AI reconstruction in embodiment 1;

[0036] Figure 6 Stratum conditions of a subway tunnel project in embodiment 1;

[0037] Figure 7 Schematic diagram of Topsis evaluation flow of disaster schemes in embodiment 1;

[0038] Figure 8 BP neural network algorithm flow based on simulated annealing in embodiment 1. DETAILED DESCRIPTION

[0039] The technical solutions of the present application will be further described in detail below with reference to the drawings.

[0040] Second, the term "one embodiment" or "an embodiment" as may appear herein is merely used to refer to a possible implementation, of various possible implementations, of the present application. The various features, structures, or characteristics of an embodiment described herein are not to be interpreted as meaning that all of these features, structures, or characteristics are required, or that these features, structures, or characteristics are the only possible features, structures, or characteristics, used in an implementation of the present application. Different implementations of the present application can have different features, structures, or characteristics.

[0041] Embodiment 1

[0042] With reference to Figures 1-8 The embodiment provides a geotechnical engineering disaster master control factor determination and safety evaluation method, including the following steps:

[0043] S1, a full convolutional network model is constructed to learn the known stratum image features, sparse borehole data is combined to interpolate and predict the stratum image of an unknown area, and a standardized stratum profile prediction image is output.

[0044] Specifically, in an open area near a construction section, intensive drilling is performed to obtain a learning image for stratum prediction, and the stratum information below is predicted based on sparse boreholes in the engineering section. The essence of stratum feature learning is to extract soil layer features of different scales based on a full convolutional model for subsequent interpolation of unknown strata. A full convolutional neural network (FCN) is built, the input is an image sample, and the output is a soil layer type prediction map. The main steps are as follows:

[0045] Image learning of the learning area: for the existing stratum (learning area) of known intensive borehole data, a conventional interpolation method (such as KNN) can be used to form a stratum profile image.

[0046] Grid learning of the learning area: for the learning area after the stratum profile image is formed, pixel grids are divided according to a self-determined ratio. Taking a 50m×50m stratum profile as an example, if the division is performed at a ratio of 1:2500, then a single pixel grid represents a 1m×1m soil layer.

[0047] Grid learning of the task area: for the engineering site (task area) with only a small amount of borehole data, pixel grids are divided according to a self-determined ratio (same as the learning area);

[0048] Task area template determination: for the unknown area in the task area after grid division, a plurality of soil layer feature templates of different specifications are determined;

[0049] Learning area feature extraction: using the determined soil layer feature templates of different specifications, all soil layer distribution features (template information) conforming to the template specifications are extracted and collected on the stratum profile map of the learning area;

[0050] Feature template classification: First, according to the total amount of soil layer feature template, the matching algorithm is selected, when the number of feature templates is less than 1000, XGBoost is used, when the number of feature templates is more than 1000, CatBoost or LightGBM is used (all three algorithms are variants of gradient boosting decision tree, among which XGBoost performs well in processing small and medium-sized feature classification tasks, CatBoost and LightGBM are suitable for large-scale feature processing, and CatBoost requires higher computing resources). Then, the soil layer feature templates are classified according to the spacing of different size grid columns in the pixel grid template.

[0051] It should be noted that the convolution kernel is convolved in the local sample image in the order from left to right and from top to bottom with a preset step size, so that the local sample image is mapped into a plurality of feature mapping images.

[0052] On the basis of completing the classification of soil layer feature template information, for the blank grid between any two drill holes in the task area, an extended normal equation simulation algorithm is used to interpolate and fill the random path.

[0053] Specifically, the first round of interpolation is performed on the blank grid between the N original drill holes (pixel grid strips with soil layer information) in the task area, and when all the blank grids matched with the feature templates are interpolated, the first round of interpolation is ended.

[0054] The second round of interpolation is performed on the blank grid between the K pixel grid strips with soil layer information (3 pixel grid strips of original drill holes and 2 pixel grid strips obtained by interpolation) in the task area, and the interpolation is repeated until all the blank grids matched with the feature templates are interpolated; the above steps are repeated for multiple rounds of interpolation until all the blank grids are filled; wherein, in this embodiment, N and K are valued as: N=3, K=5.

[0055] The stratigraphic profile prediction image is standardized to obtain a standardized stratigraphic profile prediction image, wherein the image format is output as a more intuitive png image, excel table and txt text format.

[0056] S2, a BP neural network model based on simulated annealing algorithm is trained to obtain a nonlinear mapping relationship between soil body constitutive parameters and sample stress-strain characteristics, and then the constitutive parameters are inversed through soil mechanics test data.

[0057] Specifically, soil sample is obtained by field sampling. The number of soil samples can be multiple, and the soil samples are numbered. Indoor mechanical tests are performed on each soil sample in the laboratory to obtain soil body mechanical test data and corresponding constitutive parameters of each soil sample, wherein the soil body mechanical test data includes water content, density, stress-strain data, and the soil body mechanical test data and corresponding constitutive parameters of each soil sample are summarized. At the same time, a BP neural network model is constructed, and a nonlinear mapping relationship between the constitutive parameters of the soil body and the stress-strain characteristics of the sample is trained, so that the constitutive parameters of the soil body can be quickly and accurately inverted from the soil body mechanical test data.

[0058] The trained BP neural network model includes an input layer, a hidden layer and an output layer. The specific operation method is that the input layer inputs the soil body mechanical test data into the BP neural network; the hidden layer includes a nonlinear mapping relationship between the constitutive parameters (M, eN, λ, m) and the stress-strain, strain body change characteristics; and the output layer inversely gives the constitutive parameters.

[0059] The network error function is set as a mean square error function (MSE) for measuring the deviation between the network prediction and the true test value, and the expression is,

[0060]

[0061] Wherein, E is the error, y j ' is the target output, y j is the prediction output of the BP neural network model, and l is the number of neurons in the output layer.

[0062] The global optimization of the weight adjustment process of the BP neural network is performed based on the simulated annealing (Simulated Annealing) mechanism to improve the convergence stability and global optimization ability of the BP neural network. The update weight and bias formula in the BP neural network model is,

[0063] Δw ij = -η·δ i ·a j

[0064] Δb i = -η·δ i

[0065] Wherein, w ij is the weight (the strength of the connection between neurons), η is the learning rate, δ i is the error term, a j is the output of the previous layer of neurons, and b i is the bias.

[0066] It should be noted that the simulated annealing mechanism allows partial "inferior solutions" to jump out of the local extremum region by controlling the "temperature" variable to decrease, and finally converges to the global optimal weight configuration.

[0067] S3, the predicted stratum profile prediction image, the inverted constitutive parameter and the existing upper building information are integrated into the numerical software for fine modeling, so as to realize fine simulation of the geotechnical engineering construction response.

[0068] According to the characteristics of geotechnical engineering construction, each potential factor affecting the disaster is selected; taking shield tunneling as an example, the disaster factors are selected from three aspects of building parameters, stratum parameters and excavation parameters, and twelve factors of stratum inclination, stratum thickness, groundwater level, internal friction angle, cohesion, Poisson's ratio, building quality, foundation width, foundation length, face pressure, tunneling speed and tunnel depth are selected.

[0069] The predicted stratum profile prediction image, the inverted constitutive parameter and the existing upper building information are integrated into the finite element software for calculation, so as to more accurately simulate the stratum response law of the construction section.

[0070] Further, according to the predicted stratum profile prediction image, the PNG image or Excel / TXT format is converted into a three-dimensional stratum grid model; each pixel block in the stratum profile prediction image is mapped to an element in the numerical model. The constitutive parameters output by the BP neural network are used to assign values to the properties of each stratum material; the BIM model or structural drawing is used to reconstruct the geometry information of the foundation and upper load; the upper structure load is directly applied to the surface element node of the soil layer. The model construction output is a complete initial finite element model integrating the stratum grid, physical property parameters and building structure load.

[0071] In order to simulate the details of stratum and non-homogeneous structure, a non-structural grid division strategy should be adopted to adapt to irregular geological interfaces and structural bottom contours.

[0072] Specifically, the grid division rule is: the element type is three-dimensional tetrahedral or hexahedral element, and the grid size is referenced to the resolution of the profile image; the element division density: stratum interface, structure edge area: dense grid, control the element size not more than 1m 3 ; stratum middle, load stable area: sparse grid, element volume up to 4m 3 . The obtained grid file is input into the simulation software (such as ABAQUS, FLAC3D) to generate a solution body.

[0073] According to the specific construction environment and engineering characteristics, the bottom boundary is set as a fixed boundary condition (full displacement constraint), the horizontal side boundary is set as a free shear boundary, the vertical direction is allowed to displace, and the shield construction disturbance load and underground water level load are set; according to the construction conditions, static or dynamic analysis is selected, wherein the static analysis solves the ground deformation and stress under the shield thrust, and the dynamic analysis simulates the disturbance propagation in the advancing process. The final output variables are: stress field for judging stress concentration area; displacement field for evaluating ground surface settlement and structure displacement; pore pressure field for analyzing seepage and damage risk; ground deformation indicators such as deformation value and deformation direction. The output results are exported as profile graphs or table files for quantitative analysis of disaster factors such as the maximum settlement caused by shield interference and the underground water level fluctuation amplitude.

[0074] It should be noted that by combining the existing building three-dimensional model based on the stratum geometric image information and soil body constitutive parameter information, the fine response simulation of the geotechnical engineering construction process is realized. The core goal is to reproduce the response law of the real stratum stress field, deformation field and pore pressure field under construction disturbance, and to provide basic data support for subsequent disaster factor weight analysis.

[0075] S4, for specific geotechnical engineering, select each potential factor affecting disaster, and perform fine simulation under the influence of multiple factors, and use entropy weight method to analyze and determine the weight of each potential factor from the simulation results.

[0076] Specifically, based on the data of each potential factor of disaster, a judgment matrix of m schemes and n evaluation indexes is constructed, represented as,

[0077]

[0078] The normalized matrix is normalized, and the entropy value of each evaluation index is calculated.

[0079] Further, the extreme value method is used to standardize the original data (i.e. potential factor data). The following formula is used to process the evaluation index data (the larger the value, the more dangerous), and the standardized matrix that has eliminated the differences in the measurement units of each evaluation index is finally obtained:

[0080]

[0081] Where, r ij is the standardized data value, x ij is the original data, min(x ij ) is the minimum value of the jth evaluation index, and max(x ij ) is the maximum value of the jth evaluation index.

[0082] In order to avoid the situation that the data after standardization has low value or even negative value, the data is translated, the expression is,

[0083] r' ij = r ij + H;

[0084] wherein, r' ij is the original data after smoothing, H is the amplitude of translation of evaluation index, generally 0.01.

[0085] In order to facilitate the calculation of entropy, the standardization result r ij of each column (i.e. each factor) is normalized to obtain the proportion coefficient under each index, expressed as,

[0086]

[0087] wherein, p ij is the proportion value of the i-th scheme under the j-th evaluation index after normalization, which represents the summation of i from 1 to m.

[0088] In order to avoid the error of ln(0) in logarithmic term, when p ij = 0, it can be set as a very small value such as 10 -6 .

[0089] According to information theory, the more dispersed information entropy is, the greater the uncertainty is; the more concentrated the entropy value is, the more distinguishable the index is. The entropy value calculation formula is as follows:

[0090]

[0091] wherein, e j is the information entropy of the j-th evaluation index, ln(p ij ) is the natural logarithm of p ij , lnm is the standardization coefficient,

[0092] The information "effectiveness" of the index is measured by its redundancy, the greater the redundancy is, the more useful the index is. The redundancy of each index is calculated and the redundancy of all evaluation indexes is normalized to obtain the final weight, the expression is,

[0093] g j = 1 - e j ;

[0094]

[0095] wherein, g j is the redundancy of the j-th evaluation index, is the summation of redundancy of all evaluation indexes, h jThe weight of the jth evaluation index.

[0096] S5, using Topsis method to establish multi-factor catastrophe evaluation model, analyzing the influence degree of different factors on engineering construction process, realizing quantitative safety evaluation of geotechnical engineering and obtaining optimal construction scheme.

[0097] Specifically, according to the entropy value of each evaluation index, an initial evaluation matrix is formed, and normalization processing is performed to obtain a normalized matrix;

[0098] Further, first, an initial evaluation matrix is constructed, assuming that there are m tunneling schemes that need to be evaluated for construction risk, and the evaluation index is n, the jth evaluation index of the ith scheme is x ij Then, the initial evaluation matrix X composed of the evaluation index values of each scheme is:

[0099]

[0100] Since different evaluation indexes have different dimensions, in order to unify the processing, the following standard normalization formula is used to normalize each element x ij in the original matrix X to a vector, and the expression is

[0101]

[0102] Thus, the normalized matrix R is obtained as

[0103]

[0104] Using the analytic hierarchy process, the normalized eigenvector corresponding to the largest eigenvalue is used as the weight vector to solve the CR value of each matrix.

[0105] Using the entropy weight method in objective weighting method, the weight β j of the n evaluation indexes is calculated; using the analytic hierarchy process in subjective weighting method, the weight α j of the n evaluation indexes is calculated. The subjective and objective weighting methods are combined for combined weighting:

[0106]

[0107] Based on the weight factor of each evaluation index, a weighted judgment matrix is constructed combined with the normalized matrix;

[0108] Specifically, using the calculated weight h j of each evaluation index, the normalized matrix R is weighted to construct a weighted judgment matrix V, represented as

[0109] v ij = h j · r ij ;

[0110] wherein v ij is the weighted value of the ith scheme on the jth evaluation index.

[0111] The obtained weighted judgment matrix is expressed as,

[0112]

[0113] The positive and negative ideal state solutions are obtained, and the Euclidean distance between the entropy value of the evaluation index of each scheme and the positive and negative ideal state solutions is calculated by the Euclidean formula.

[0114] Further, the positive ideal solution is,

[0115]

[0116] The negative ideal solution is,

[0117]

[0118] wherein I is the risk assessment attribute set; A + is the set of positive ideal solutions, A - is the set of negative ideal solutions; is the positive ideal solution, is the negative ideal solution,

[0119] The distance between the evaluation value of each scheme and the positive and negative ideal state is calculated by the Euclidean formula, and the expression is,

[0120]

[0121]

[0122] wherein, represents the Euclidean distance between the ith scheme and the positive ideal solution, represents the Euclidean distance between the ith scheme and the negative ideal solution, is the value of the jth attribute in the positive ideal solution, is the value of the jth attribute in the negative ideal solution,

[0123] The Topsis evaluation value of the safety score of each scheme is calculated. The expression is,

[0124]

[0125] wherein, The greater, the better the scheme.

[0126] According to the Topsis evaluation value of the disaster risk safety score of each scheme from large to small, the optimal construction scheme is selected.

[0127] The embodiment also provides a geotechnical engineering disaster master control factor determination and safety evaluation system, comprising:

[0128] The image generation module is configured to construct a full convolution network model to learn known stratum image features, to combine sparse borehole data to interpolate and predict stratum images of unknown areas, and to output standardized stratum profile prediction images; the parameter inversion module is configured to train a nonlinear mapping relationship between soil body constitutive parameters and sample stress-strain characteristics based on a BP neural network of a simulated annealing algorithm, and to invert the constitutive parameters of the soil body through soil body mechanics test data; the refined modeling module is configured to integrate the predicted stratum profile prediction images, the inverted soil body constitutive parameters, and existing upper building information into numerical software to perform refined modeling, and to realize refined simulation of geotechnical engineering construction responses; the weight analysis module is configured to select various potential factors influencing disasters for specific geotechnical engineering, to perform refined simulation under the influence of multiple factors, to analyze and determine the weights of various factors from simulation results by using an entropy weight method, and to obtain a disaster factor rating table matched therewith; and the safety evaluation module is configured to establish a multi-factor disaster evaluation model by using a Topsis method, to realize quantitative safety evaluation of geotechnical engineering, and to obtain an optimal construction scheme.

[0129] The embodiment also provides a computer device suitable for the geotechnical engineering disaster master control factor determination and safety evaluation method, comprising a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to realize the geotechnical engineering disaster master control factor determination and safety evaluation method proposed in the above embodiment.

[0130] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is configured to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.

[0131] The embodiment also provides a storage medium having a computer program stored thereon, the program being executed by a processor to implement the method for determining a main control factor of geotechnical engineering disaster and safety evaluation as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0132] To sum up, the present application realizes high-precision image restoration of the stratum profile in the unknown area by constructing a full convolution network model to predict the stratum image and output a standardized image, improves the efficiency and accuracy of the acquisition of stratum geometric information, and provides high-resolution and high-quality spatial data basis for subsequent modeling and simulation; the present application realizes rapid and high-precision inversion of key constitutive parameters by constructing a BP neural network model optimized based on the simulated annealing algorithm to invert the soil constitutive parameters, effectively improves the precision and efficiency of the input of soil material parameters, and provides accurate and reliable constitutive model support for numerical simulation; the present application realizes seamless connection from image intelligent recognition to finite element modeling by integrating the stratum image, constitutive parameters and structure information to perform fine modeling and simulation, directly converting image data into model grids, and realizing high-precision simulation to truly reproduce the stratum response process under construction disturbance, makes up for the timeliness and spatial coverage of field monitoring, and identifies the weight of the entropy weight method based on the fine simulation results, clearly defines the differentiation logic of the "main factor" and the "secondary factor" under the condition of multiple factors, has strong interpretability, avoids subjective bias of artificial weighting by the entropy weight method, improves the objectivity of the evaluation system, and further participates in Topsis scheme evaluation, and can also independently constitute a disaster risk grade early warning mechanism.

[0133] Embodiment 2

[0134] Referring to Tables 1-3, the second embodiment of the present application is given to further verify the technical scheme of the present application, and experimental simulation data of the method for determining a main control factor of geotechnical engineering disaster and safety evaluation is given.

[0135] In a preferred embodiment (a subway tunnel project), a stratum with a range of 500m x 48m (span x depth) is selected as the research object, and there are 11 borehole data with a borehole spacing of about 47m. The vertical distribution of the stratum is clay layer, silty sand, silty clay and clayey silt in turn. First, the stratum profile of the two samples is divided into a pixel grid in proportion to 1:5000, and the corresponding horizontal and vertical resolutions are 2m and 0.4m respectively (i.e. the size of a single grid is 2m x 0.4m). Then, for the stratum profile of the task area, four equally spaced borehole data are given as sparse borehole information (i.e. four pixel grid strips with a spacing of about 67m on the image), forming the task area. Finally, the developed stratum profile information AI reconstruction program is executed to extract the soil layer feature template information of the learning area, and then the stratum profile information of the task area is predicted by multiple rounds of interpolation. The 2D images of the above size after interpolation prediction are assembled to obtain the 3D stratum information under the engineering section.

[0136] Further, in the subway tunnel project, 100 soil samples are obtained according to the soil type and engineering requirements. The soil samples are numbered one by one as No. 1-100, and the mechanical information of the soil samples is obtained by triaxial test. No. 1-90 soil samples are set as training samples, and the optimal BP neural network is obtained by BP neural network training and weight fine-tuning using error back propagation algorithm; No. 90-100 is set as a test sample to test and debug the BP neural network.

[0137] Preferably, the relative error control index in the BP neural network model is set to be less than 10%. Part of the samples are selected for testing, and the test results are shown in Table 1.

[0138] Table 1 Comparison of actual and predicted values of constitutive parameters

[0139]

[0140] As can be seen from Table 1, the relative error of each parameter is less than 10%, which meets the set requirement, indicating that the results obtained by the present application are more accurate and can be applied to engineering practical applications.

[0141] Further, on the basis of the above stratum geometry and soil mechanical information refinement, the construction section building is also modeled and integrated into the finite element software for further calculation to obtain the response law in the construction process.

[0142] Preferably, based on the characteristics of the engineering construction, various potential factors affecting disasters are selected. These factors are chosen from three aspects: building parameters, geological parameters, and excavation parameters. Twelve factors are selected: geological dip angle, geological thickness, groundwater level, internal friction angle, cohesion, Poisson's ratio, building quality, foundation width, foundation length, face pressure, tunneling speed, and tunnel depth. Then, the weight of each major disaster-controlling factor is determined using the entropy weight method, resulting in a matching disaster factor rating table, as shown in Table 2.

[0143] Table 2. Catastrophic Risk Factor Rating Table

[0144]

[0145] Furthermore, the Topsis method is used to establish an evaluation model for specific engineering schemes, analyze the degree of influence of different factors on the engineering construction process, and then optimize the existing parameters of the construction scheme. For specific implementation steps, please refer to step five. Examples of evaluation calculation results are shown in Table 3.

[0146] Table 3. Calculation results of Topsis evaluation

[0147]

[0148]

Claims

1. A method for determining the main controlling factors of geotechnical engineering disasters and for safety evaluation, characterized in that: include, A fully convolutional network model is constructed to learn the features of known stratigraphic images. Combined with sparse borehole data, stratigraphic images of unknown areas are interpolated and predicted, and standardized stratigraphic profile prediction images are output. A BP neural network based on the simulated annealing algorithm is trained to obtain the nonlinear mapping relationship between soil constitutive parameters and specimen stress-strain characteristics, and the constitutive parameters are inverted through soil mechanics test data. The predicted stratigraphic profile images, the inverted soil constitutive parameters, and the existing superstructure information are integrated into numerical software for refined modeling, thereby achieving refined simulation of geotechnical engineering construction response. For specific geotechnical engineering projects, various potential factors affecting disasters are selected, and a refined simulation under the influence of multiple factors is carried out. The entropy weight method is used to analyze and determine the weight of each factor from the simulation results, and a disaster factor rating table matching it is obtained. A multi-factor disaster assessment model was established using the Topsis method to achieve quantitative safety evaluation of geotechnical engineering and obtain the optimal construction plan.

2. The method for determining the main controlling factors of geotechnical engineering disasters and evaluating safety as described in claim 1, characterized in that: The output of the standardized stratigraphic profile prediction image includes the following steps: By performing convolution operations on local sample images in a left-to-right and top-to-bottom order with a preset stride, the local sample images are mapped into multiple feature image images. The extended normal equation simulation algorithm is used to interpolate the stratigraphic image of the unknown area and then perform standardization processing to obtain the standardized stratigraphic profile prediction image.

3. The method for determining the main controlling factors of geotechnical engineering disasters and evaluating safety as described in claim 2, characterized in that: The method of interpolating stratigraphic images of unknown regions using the extended normal equation simulation algorithm includes the following steps. The first round of interpolation is performed on the blank grid between the N original boreholes in the task area. The first round of interpolation ends when all blank grids that match the feature template have been interpolated. The second round of interpolation is performed on the blank grids between the K pixel grids containing soil layer information in the task area until all blank grids matching the feature template are interpolated. Repeat the above steps for multiple rounds of interpolation until all blank grid cells are filled.

4. The method for determining the main controlling factors of geotechnical engineering disasters and evaluating safety as described in claim 3, characterized in that: The simulated annealing algorithm uses a BP neural network, where the input layer takes the mechanical test results of the soil sample and inputs them into the BP neural network; the hidden layer contains the nonlinear mapping relationship between constitutive parameters and stress-strain and strain deformation characteristics; and the output layer analyzes and provides the constitutive parameters.

5. The method for determining the main controlling factors of geotechnical engineering disasters and evaluating safety as described in claim 4, characterized in that: For specific geotechnical engineering projects, various potential factors affecting disasters are selected, and a refined simulation under the influence of multiple factors is conducted, including the following steps: Based on the characteristics of geotechnical engineering construction, select various potential factors that may affect disasters; The predicted stratigraphic profile images, inverted soil mechanical parameters, and superstructure information are integrated into finite element software for calculation to simulate the stratigraphic response characteristics of the construction section.

6. The method for determining the main controlling factors of geotechnical engineering disasters and evaluating safety as described in claim 5, characterized in that: The method of using entropy weight to analyze and determine the weights of each factor from the simulation results includes the following steps: Based on the data of various potential factors of disaster, construct a judgment matrix with m options and n evaluation indicators; The original data is standardized using the extreme value method to obtain a standardized judgment matrix; The standardized matrix is ​​normalized, and the entropy value of each evaluation index is calculated.

7. The method for determining the main controlling factors of geotechnical engineering disasters and evaluating safety as described in claim 6, characterized in that: The method of using the Topsis method to establish a multi-factor disaster assessment model to achieve quantitative safety evaluation of geotechnical engineering and obtain the optimal construction plan includes the following steps. An initial matrix is ​​formed based on the entropy values ​​of each evaluation index, and then normalized to obtain a normalized matrix. The combined weights of each evaluation index are determined using the entropy weight method and the analytic hierarchy process. Based on the weight factors of each evaluation indicator and combined with the normalized matrix, a weighted judgment matrix is ​​constructed. Obtain the positive and negative ideal state solutions, and calculate the Euclidean distance between the entropy value of the evaluation index of the disaster risk of each scheme and the positive and negative ideal state solutions using the Euclidean formula; Calculate the Topsis evaluation value of each scheme's disaster risk safety score, sort the schemes according to their Topsis evaluation values, and select the optimal construction scheme.

8. A system for determining the main controlling factors of geotechnical engineering disasters and evaluating safety, based on the method for determining the main controlling factors of geotechnical engineering disasters and evaluating safety as described in any one of claims 1 to 7, characterized in that: include, The image generation module is used to build a fully convolutional network model to learn the features of known stratigraphic images, combine sparse borehole data to interpolate and predict stratigraphic images in unknown areas, and output standardized stratigraphic profile prediction images. The parameter inversion module is used to train a BP neural network based on the simulated annealing algorithm to find the nonlinear mapping relationship between soil constitutive parameters and specimen stress-strain characteristics, and to invert its constitutive parameters through soil mechanics test data. The refined modeling module is used to integrate the predicted stratigraphic profile image, the inverted soil constitutive parameters, and the existing superstructure information into the numerical software for refined modeling, so as to realize the refined simulation of geotechnical engineering construction response. The weight analysis module is used to select various potential factors affecting disasters for specific geotechnical engineering projects, conduct refined simulations under the influence of multiple factors, and use the entropy weight method to analyze and determine the weight of each factor from the simulation results to obtain a disaster factor rating table that matches it. The safety evaluation module is used to establish a multi-factor disaster evaluation model using the Topsis method, so as to realize quantitative safety evaluation of geotechnical engineering and obtain the optimal construction plan.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method for determining the main controlling factors of geotechnical engineering disasters and evaluating safety as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the method for determining the main controlling factors of geotechnical engineering disasters and evaluating safety as described in any one of claims 1 to 7.

Images