Slope condition assessment method and device, computer storage medium and computer program product

Abstract

This invention provides a slope condition assessment method and apparatus, a computer storage medium, and a computer program product. The method involves inputting multiple first analysis parameters of the slope to be assessed into a neural network model to obtain a first coefficient of the slope. A safety factor calculation module then calculates a second coefficient of the slope based on the first coefficient and either multiple first analysis parameters or multiple second analysis parameters. This invention allows for real-time slope assessment without requiring cumbersome manual calculations.

Description

[Technical Field]

[0001] This invention relates to the technical field of slope condition assessment, and more particularly to a slope condition assessment method and apparatus, computer storage medium and computer program product. [Previous Technology]

[0002] Two-thirds of Taiwan's area can be defined as hilly land. Earthquakes, torrential rains, groundwater, or other human activities often cause slope instability. This necessitates that the competent authorities for hilly land conduct regular slope inspections to prevent slope disasters. Due to limited engineering funds, slope inspections rely heavily on visual inspections, which are not only too subjective but also unable to provide quantitative assessments based on mechanical principles, thus drawing criticism.

[0003] Furthermore, existing research on predicting slope stability and its reliability analysis has yielded theories such as limit analysis, the Mohr-Coulomb failure criterion, and the Hoek-Brown failure criterion. The stability coefficient or safety factor of the slope obtained from these analytical theories must undergo a series of calculations and graphing before it can be obtained manually by consulting charts. The charts produced by these theories often use logarithmic representations for the stability coefficient axis, making manual chart consulting difficult. If there are gaps between the analytical parameters used to calculate the stability coefficient, various regression methods must be used to find their regularity. This not only causes problems such as complex calculations and difficult solutions, but also makes it impossible to apply immediately in the field, which is inconvenient for civil engineers. [Summary of the Invention]

[0004] To solve the above problems, the main objective of this invention is to provide a slope condition assessment method, comprising: inputting a complex first analysis parameter of the slope to be measured into a neural network model to obtain a first coefficient of the slope to be measured; and having a safety factor calculation module calculate based on the first coefficient and the complex first analysis parameter or complex second analysis parameter of the slope to be measured to obtain a second coefficient of the slope to be measured.

[0005] The slope condition assessment method described above further includes the following method: having a training set calculation module use the limit analysis method to calculate the complex first analysis parameters corresponding to the complex training slopes to obtain the first coefficient of each complex training slope; and having a deep neural network module train according to the complex first analysis parameters and the first coefficient of each complex training slope to obtain the neural network model, wherein the neural network model can predict the regression law between the complex first analysis parameters and the first coefficient.

[0006] As described above in the slope condition assessment method, the deep neural network module includes an input layer, at least two hidden layers and an output layer. The input layer is used to input the complex first analysis parameter, and the output layer is used to output the first coefficient of the slope to be measured.

[0007] As described above in the slope condition assessment method, the deep neural network module is constructed using the Keras and TensorFlow packages.

[0008] As described in the aforementioned slope condition assessment method, the complex first analysis parameter of the slope to be tested and the training slope is one of the following groups: a group consisting of disturbance factor, geological strength index, lithology coefficient, slope angle and horizontal seismic force coefficient; a group consisting of disturbance factor, geological strength index, lithology coefficient, slope angle and groundwater level; a group consisting of disturbance factor, geological strength index, lithology coefficient, slope angle, groundwater level and horizontal seismic force coefficient; a group consisting of soil cohesion, soil unit weight, slope height, soil friction angle, slope angle and groundwater level; a group consisting of soil cohesion, soil unit weight, slope height, soil friction angle, distance between the base and the slope, base load, slope angle and groundwater level; or a group consisting of effective soil cohesion, soil unit weight, slope height, effective soil friction angle, air intake suction value, particle size distribution, rainfall intensity, soil permeability coefficient, slope angle and groundwater level.

[0009] In the aforementioned slope condition assessment method, the safety factor calculation module calculates the second coefficient of the slope to be tested through a formula, wherein Nr is the first coefficient of the slope to be tested, σci is the uniaxial compressive strength in the complex second analysis parameter of the slope to be tested, γ is the unit weight in the complex second analysis parameter of the slope to be tested, H is the slope height in the complex second analysis parameter of the slope to be tested, and F is the second coefficient of the slope to be tested.

[0010] As described above in the slope condition assessment method, the safety factor calculation module calculates the second coefficient of the slope to be tested through a formula, where λ is the first coefficient of the slope to be tested, φ is the soil friction angle in the complex first analysis parameter of the slope to be tested, and F is the second coefficient of the slope to be tested.

[0011] As described above in the slope condition assessment method, the safety factor calculation module calculates the second coefficient of the slope to be tested through a formula, wherein N* is the first coefficient of the slope to be tested, is the soil effective friction angle in the complex first analysis parameter of the slope to be tested, and F is the second coefficient of the slope to be tested.

[0012] As described above, the slope condition assessment method further includes, after obtaining the second coefficient of the slope to be tested, having a failure probability calculation module calculate the failure probability of the slope to be tested using the Monte Carlo method.

[0013] Another object of the present invention is to provide a computer storage medium storing computer executable code, so that the aforementioned slope condition assessment method can be implemented after the computer executable code is executed.

[0014] Another object of the present invention is to provide a computer program product, including computer executable instructions, so that the aforementioned slope condition assessment method can be implemented after the computer executable instructions are executed.

[0015] Another object of the present invention is to provide a slope condition assessment device, including a memory and a processor, wherein the processor is used to implement the aforementioned slope condition assessment method by executing computer-executable instructions stored in the memory. [Simplified Explanation of the Diagram]

[0044] Figure 1 is a flowchart of the slope condition assessment method of the present invention.

[0045] Figure 2 is a flowchart of the steps for obtaining the neural network model in the slope condition assessment method of the present invention.

[0046] Figure 3 is a system architecture diagram of the first embodiment of the slope condition assessment device of the present invention.

[0047] Figure 4 is a system architecture diagram of the second embodiment of the slope condition assessment device of the present invention.

[0048] Figure 5 is a system architecture diagram of the third embodiment of the slope condition assessment device of the present invention.

[0049] Figure 6 is a system architecture diagram of the fourth embodiment of the slope condition assessment device of the present invention.

[0050] Figure 7 is a schematic diagram of the display screen of one embodiment of the slope condition assessment method of the present invention.

[0051] Figure 8 is a schematic diagram of another embodiment of the slope condition assessment method of the present invention.

[0052] Figure 9A shows the accuracy analysis results of the prior art.

[0053] Figure 9B shows the accuracy analysis results of the present invention.

Implementation Method

[0016] The implementation of the present invention will be described below by way of specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification, and it can also be implemented or applied by other different embodiments.

[0017] Please refer to Figures 3 to 6 first. The slope condition assessment method of the present invention can run under different architectures. As shown in Figure 3, the slope condition assessment method of the present invention can run in the slope condition assessment device 10. The slope condition assessment device 10 includes a memory 11 and a processor 12, and stores a neural network model 13, a safety factor calculation module 14, and a failure probability calculation module 15. The neural network model 13, the safety factor calculation module 14, and the failure probability calculation module 15 can be software, firmware, computer executable instructions, or data structures. After being read into the memory 11, they can be executed by the processor 12 to implement the slope condition assessment method of the present invention. In this embodiment, the slope condition assessment device 10 can be a mobile phone, tablet, or computer, etc., but is not limited thereto.

[0018] As shown in Figure 4, the slope condition assessment device 10 may only include a memory 11 and a processor 12. The neural network model 21, the safety factor calculation module 22, and the failure probability calculation module 23 are stored in the server 20. The slope condition assessment device 10 and the server 20 can communicate through a wired or wireless network. The slope condition assessment device 10 only needs to input instructions (or input the parameters required by the neural network model 21, the safety factor calculation module 22, and the failure probability calculation module 23) to allow the server 20 to execute the slope condition assessment method of the present invention, and then transmit the results back to the slope condition assessment device 10 for display. This architecture can save the operating resources of the slope condition assessment device 10.

[0019] Figure 5 shows the embodiment of Figure 3 with the addition of a deep neural network module 16 and a training set computing module 17, and Figure 6 shows the embodiment of Figure 4 with the addition of a deep neural network module 24 and a training set computing module 25 to the server 20. Similar to the above embodiments, the deep neural network modules 16, 24 and the training set computing modules 17, 25 can also be software, firmware, computer-executable instructions or data structures, which can be executed by the processor 12 or the server 20 of the slope condition assessment device 10.

[0020] The following describes the detailed technical content of the slope condition assessment method of the present invention and the above-mentioned modules.

[0021] Referring to Figure 1, the slope condition assessment method of the present invention includes: inputting the complex first analysis parameters of the slope to be measured into a neural network model 13, 21 to obtain the first coefficient of the slope to be measured (step S1). Then, the safety factor calculation modules 14, 22 are instructed to calculate based on the first coefficient and the complex first analysis parameters or complex second analysis parameters of the slope to be measured to obtain the second coefficient of the slope to be measured (step S2).

[0022] In this embodiment, depending on the slope conditions, the complex first analysis parameter, the first coefficient, and the steps for calculating the second coefficient are also different, which will be described in detail below.

[0023] In one embodiment, when considering seismic parameters for a rock slope, the complex first analysis parameters consist of a disturbance factor, a geological strength index, an intrinsic rock constant, a slope angle, and a seismic coefficient. The disturbance factor can be represented by the symbol D, the geological strength index by the symbol GSI, the lithology constant by the symbol mi, the slope angle by the symbol β, and the seismic coefficient by the symbol Kh. The first coefficient obtained from these complex first analysis parameters can be called a stability number, which can be represented by the symbol Nr. Next, the safety factor calculation modules 14 and 22 can calculate the second coefficient of the slope to be measured according to the formula, where Nr is the first coefficient, σci is the uniaxial compressive strength (unit kPa) in the complex second analysis parameter, γ is the unit weight (unit kN / m2) in the complex second analysis parameter, H is the slope height (unit meters (m)) in the complex second analysis parameter, and F is the second coefficient, which can be called the safety factor.

[0024] In one embodiment, when considering the influence of groundwater level parameters on rock slopes, the complex first analysis parameters consist of a disturbance factor, a geological strength index, a lithology coefficient, a slope angle, and a groundwater level. The groundwater level can be represented by the symbol hw / h. The remaining parameters are the same as in the aforementioned embodiments and will not be repeated here. The first coefficient obtained from these complex first analysis parameters can be called a stability coefficient. Furthermore, the calculation of the second coefficient of the slope under test is also the same as in the aforementioned embodiments and will not be repeated here.

[0025] In one embodiment, when considering the influence parameters of earthquake and groundwater level on rock slope, the complex first analysis parameters are composed of disturbance factor, geological strength index, lithology coefficient, slope angle, groundwater level and horizontal seismic force coefficient. The first coefficient obtained from these complex first analysis parameters can be called stability coefficient. The calculation of these complex first analysis parameters and the second coefficient of the slope to be measured are the same as those in the aforementioned embodiment, and will not be repeated here.

[0026] In one embodiment, when considering the influence of groundwater level on soil slope parameters, the complex first analytical parameters consist of soil cohesion, soil unit weight, slope height, soil friction angle, slope angle, and groundwater level. Soil cohesion can be represented by the symbol c (unit kPa), soil unit weight by the symbol γ (unit kN / m³), slope height by the symbol H (unit meters (m)), soil friction angle by the symbol β, slope angle by the symbol β°, and groundwater level by the symbol hw / h. The first coefficient obtained from these complex first analytical parameters can be called replacement parameters, and the replacement parameters can be represented by the symbol λ. Next, the safety factor calculation modules 14 and 22 can calculate the second coefficient of the slope to be measured according to the formula, where λ is the first coefficient, is the soil friction angle in the complex first analysis parameter, and F is the second coefficient, which can be called the safety factor.

[0027] In one embodiment, considering the influence parameters of groundwater level and footing on the soil slope, the complex first analytical parameters consist of soil cohesion, soil unit weight, slope height, soil friction angle, distance from footing to slope, footing load, slope angle, and groundwater level. The distance from footing to slope can be represented by the symbol D (unit: meters (m)), and the footing load can be represented by the symbol q (unit: kN / m²). The remaining parameters are the same as in the aforementioned embodiment and will not be repeated here. The first coefficient obtained from these complex first analytical parameters can be called the substitution coefficient. Furthermore, the calculation of the second coefficient of the slope to be measured is also based on the formula, which is the same as in the aforementioned embodiment and will not be repeated here.

[0028] In one embodiment, considering the influence of groundwater level on unsaturated soil slopes, the complex first analytical parameters consist of effective soil cohesion, soil unit weight, slope height, effective friction angle, air-entry pressure, particle size distribution, rainfall intensity, soil permeability coefficient, slope angle, and groundwater level. Effective soil cohesion can be represented by the symbol c' (unit: kPa), effective friction angle by the symbol α, air-entry pressure by the symbol α (unit: 1 / kPa), particle size distribution by the symbol n, rainfall intensity by the symbol q (unit: mm / hr), and soil permeability coefficient by the symbol ks (unit: mm / hr). The remaining parameters are the same as in the aforementioned embodiment and will not be repeated. The first coefficient obtained from these complex first analytical parameters can be called the stability coefficient, which can be represented by the symbol N*. Next, the safety factor calculation modules 14 and 22 can calculate the second coefficient of the slope to be measured according to the formula, where N* is the first coefficient, which is the effective soil friction angle in the complex first analysis parameter, and F is the second coefficient, which can be called the safety factor.

[0029] In this embodiment, regardless of the slope condition mentioned above, after obtaining the second coefficient, the failure probability calculation modules 15 and 23 can use the Monte Carlo method to calculate the failure probability of the slope to be tested. For example, by inputting the coefficient of variation (COV) and standard deviation (Std) of the normal distribution, the failure probability Pf can be obtained so that civil engineers can evaluate whether the slope is stable.

[0030] In one embodiment, before applying the neural network models 13 and 21 in the slope state assessment method of the present invention, the neural network models 13 and 21 can be trained using artificial intelligence machine learning algorithms. Please refer to Figure 2 to understand the training process of the neural network model. In step S11, the complex first analysis parameters corresponding to the complex training slopes can be prepared as training data, and a training set calculation module 17 and 25 can be used to calculate them using the limit analysis method to obtain the first coefficients of each complex training slope. The so-called limit analysis method is one of the popular analysis methods in engineering. It combines the plastic limit theory and the upper and lower bound theory of finite elements to obtain the ultimate load of failure. It avoids the elastic-plastic analysis process and directly studies the limit state of the structure to solve for the ultimate load. In this embodiment, the training set calculation module 17 and 25 can use OptumCE software to calculate the first coefficients, but the present invention is not limited thereto.

[0031] Next, in step S12, a deep neural network module 16, 24 is trained according to the complex first analysis parameter and the first coefficient of each complex training slope to obtain neural network model 13, 21, wherein the neural network model 13, 21 can predict the regression law between the complex first analysis parameter and the first coefficient.

[0032] In one embodiment, the deep neural network modules 16 and 24 include an input layer, at least two hidden layers, and an output layer. Each layer may contain a different number of neurons. For example, the input layer may input 8 features, and there may be 3 hidden layers, each with 9 neurons. The output layer may output 4 neurons, etc., but the present invention is not limited thereto. In this embodiment, the input layer is used to input a complex first analysis parameter, and the output layer is used to output a first coefficient.

[0033] In one embodiment, the deep neural network modules 16 and 24 can be constructed using tools such as Scikit-learn, Create ML, and Keras, or using other neural network training software. In a specific embodiment, the deep neural network modules 16 and 24 can be constructed using the Keras and TensorFlow tools to achieve the highest accuracy, but the present invention is not limited thereto. More specifically, the deep neural network modules 16 and 24 of the present invention can be of the type of recurrent neural network (RNN).

[0034] In this way, the neural network models 13 and 21 trained by the deep neural network modules 16 and 24 can predict the regression law between the complex first analysis parameter and the first coefficient. That is, when the complex first analysis parameter input to the neural network models 13 and 21 does not appear in the training data, the neural network models 13 and 21 can still predict the first coefficient corresponding to the complex first analysis parameter.

[0035] In one embodiment, neural network models 13 and 21 may be model files in h5 format, but the present invention is not limited thereto. In addition, depending on the operating system running in the slope condition assessment device 10, neural network models 13 and 21 may convert the h5 model files into different file formats, such as converting them into the iOS-compatible mlmodel format model files (e.g., using the Pvthon add-in Core ML Tools for conversion), and using the iOS-specific application programming software Xcode for APP development.

[0036] Please refer to Figures 7 and 8, which are schematic diagrams of different embodiments of the slope condition assessment method of the present invention (i.e., APP running screens). In the case of considering seismic parameters for rock slopes in Figure 7, the user can input parameters such as disturbance factor (D), geological strength index (GSI), lithology coefficient (mi), slope angle (β), and horizontal seismic force coefficient (kh) in input cells 31, 32, 33, 34, and 35 respectively, and then press button 41 to obtain the first coefficient (i.e., stability coefficient). Next, input parameters such as uniaxial compressive strength (σci), unit weight (γ), and slope height (H) in input cells 36, 37, and 38 respectively, and press button 42 to obtain the second coefficient (i.e., safety factor). The operation in Figure 8 (rock slope considering groundwater level influence parameters) is similar to that in Figure 7, and will not be described again here. In addition, input cells 31, 32, 33, 34, and 35 can be added or removed depending on the different complex first analysis parameters required for the slope conditions, and input cells 36, 37, and 38 can also be added or removed depending on the different formulas used to calculate the second coefficient. This invention is not limited to these.

[0037] In one embodiment, the slope condition assessment method of the present invention can be implemented by executing computer executable code through a computer storage medium storing computer executable code. Alternatively, the slope condition assessment method of the present invention can also be implemented by executing computer program products including computer executable instructions.

[0038] This invention uses regression analysis to verify the accuracy of the obtained neural network model. The regression analysis defines R as a coefficient that measures the correlation between two variables to determine the linear correlation between them. Figure 9A compares the neural network model trained using Extreme Learning Machine (ELM) in the prior art with the manually calculated Extreme Analysis method (OptumG2 software), achieving an R² value of 0.9949600102603778. Figure 9B compares the neural network model trained using Recurrent Neural Network (RNN) in this invention with the manually calculated Extreme Analysis method (OptumG2 software), achieving an R² value of 0.9991912520411319. Therefore, the R² value of this invention is slightly larger than that of the prior art, indicating that the accuracy of this invention is higher than that of the prior art.

[0039] Furthermore, ELM and the RNN of this invention were used to analyze the following actual cases (Table 1, the safety factor of which was obtained by limit analysis (OptumG2 software) of the actual cases). As can be seen from Table 2, the error between this invention and the actual case is 3.62%, which is much lower than the error of 7.63% between ELM and the actual case, thus sufficiently proving that this invention has the effect of high accuracy.

[0040] Table 1

[0041] Table 2

[0042] In summary, by means of the slope condition assessment method and device, computer storage medium and computer program product of the present invention, civil engineers can input the analysis parameters of the on-site slope into their mobile phones in real time, and easily obtain the stability coefficient and safety coefficient of the on-site slope. It has the effect of rapid prediction and convenience, and does not require complicated manual calculation.

[0043] The above embodiments are merely illustrative of the technical principles, features, and effects of the present invention, and are not intended to limit the scope of implementation of the present invention. Any person skilled in the art may modify and alter the above embodiments without departing from the spirit and scope of the present invention. However, any equivalent modifications and alterations made using the teachings of the present invention should still be covered by the following claims. The scope of protection of the present invention should be as listed in the following claims.

Claims

1. A slope condition assessment method, comprising: The complex first analysis parameters of the slope to be measured are input into a neural network model to obtain the first coefficient of the slope to be measured. The neural network model can predict the regression law between the complex first analysis parameters and the first coefficient. A safety factor calculation module is then set to calculate the second coefficient of the slope to be measured based on the first coefficient and the complex first analysis parameters or complex second analysis parameters of the slope to be measured.

2. The slope condition assessment method as described in Request 1 further includes: A training set computation module uses the limit analysis method to calculate the complex first analysis parameters corresponding to the complex training slopes to obtain the first coefficients of each complex training slope; and a deep neural network module is trained based on the complex first analysis parameters and first coefficients of each complex training slope to obtain the neural network model.

3. The slope condition assessment method as described in claim 2, wherein, The deep neural network module includes an input layer, at least two hidden layers, and an output layer. The input layer is used to input the complex first analysis parameter, and the output layer is used to output the first coefficient of the slope to be measured.

4. The slope condition assessment method as described in claim 2, wherein, This deep neural network module was built using the Keras and TensorFlow suites.

5. The slope condition assessment method as described in request item 1 or 2, wherein, The complex first analytical parameter of the slope under test and the training slope is one of the following groups: a group consisting of disturbance factor, geological strength index, lithology coefficient, slope angle, and horizontal seismic force coefficient; a group consisting of the disturbance factor, geological strength index, lithology coefficient, slope angle, and groundwater level; a group consisting of the disturbance factor, geological strength index, lithology coefficient, slope angle, groundwater level, and horizontal seismic force coefficient; a group consisting of soil coagulation... A group consisting of soil cohesion, soil unit weight, slope height, soil friction angle, slope angle, and groundwater level; a group consisting of soil cohesion, soil unit weight, slope height, soil friction angle, distance between the foundation and the slope, foundation load, slope angle, and groundwater level; or a group consisting of effective soil cohesion, soil unit weight, slope height, effective soil friction angle, air intake suction value, particle size distribution, rainfall intensity, soil permeability coefficient, slope angle, and groundwater level.

6. The slope condition assessment method as described in Request 1, wherein, The safety factor calculation module calculates the second coefficient of the slope to be tested using a formula, where Nr is the first coefficient of the slope to be tested, σci is the uniaxial compressive strength in the complex second analysis parameter of the slope to be tested, γ is the unit weight in the complex second analysis parameter of the slope to be tested, H is the slope height in the complex second analysis parameter of the slope to be tested, and F is the second coefficient of the slope to be tested.

7. The slope condition assessment method as described in Request 1, wherein, The safety factor calculation module calculates the second coefficient of the slope under test using a formula, where λ is the first coefficient of the slope under test, φ is the soil friction angle in the complex first analysis parameter of the slope under test, and F is the second coefficient of the slope under test.

8. The slope condition assessment method as described in Request 1, wherein, The safety factor calculation module calculates the second coefficient of the slope under test using a formula, where N* is the first coefficient of the slope under test, the effective soil friction angle is the first complex analysis parameter of the slope under test, and F is the second coefficient of the slope under test.

9. The slope condition assessment method as described in claim 1 further includes, after obtaining the second coefficient of the slope to be measured, having a failure probability calculation module calculate the failure probability of the slope to be measured using the Monte Carlo method.

10. A computer storage medium storing computer-executable code, wherein, upon execution, the computer-executable code implements the slope condition assessment method of any one of claims 1 to 9.

11. A computer program product, comprising computer-executable instructions, wherein, upon execution of the computer-executable instructions, the slope condition assessment method of any one of requests 1 to 9 is implemented.

12. A slope condition assessment device, comprising a memory and a processor, wherein, The processor is used to implement the slope condition assessment method of any one of requests 1 to 9 by executing computer-executable instructions stored in the memory.

Images