Rock slope reliability evaluation method based on hybrid multi-model probability inference and related equipment

Abstract

The application discloses a rock slope reliability evaluation method based on a hybrid multi-model probability inference and related equipment, which can be applied to the technical field of geotechnical engineering reliability analysis. The application obtains geotechnological parameters and field sample data of a target rock slope, constructs a corresponding candidate probability distribution model set, calculates the prior distribution of the geotechnological parameters corresponding to each candidate probability distribution model, estimates the posterior parameter distribution of each distribution parameter in each candidate probability distribution model by using Bayesian inference and Markov chain Monte Carlo sampling, calculates the posterior weight of each candidate probability distribution model, combines all the posterior parameter distributions to construct a hybrid multi-model probability density function, inputs the hybrid multi-model probability density function into a pre-constructed slope stability proxy model to calculate the slope failure probability, and finally determines the reliability level, so that the accuracy of the slope reliability evaluation result can be effectively improved under limited data.

Description

Technical Field

[0001] This application relates to the field of geotechnical engineering reliability analysis technology, and in particular to a method and related equipment for assessing the reliability of rock slopes based on hybrid multi-model probabilistic inference. Background Technology

[0002] In related technologies, rock slope reliability assessment methods typically assume that the input parameters follow a pre-defined probability distribution, such as a log-normal distribution. However, field data in geotechnical engineering is limited, making it impossible to determine which probability distribution best suits the current scenario. Existing rock slope reliability assessment methods often neglect the impact of this uncertainty, resulting in low accuracy. Furthermore, existing methods typically rely on a single likelihood function, i.e., a single distribution form, which also fails to address the impact of model uncertainty, thus hindering the improvement of slope reliability assessment accuracy.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0004] The main objective of this application is to propose a method and related equipment for assessing the reliability of rock slopes based on hybrid multi-model probabilistic inference, which can effectively improve the accuracy of slope reliability assessment results with limited data.

[0005] To achieve the above objectives, one aspect of this application proposes a method for assessing the reliability of rock slopes based on hybrid multi-model probabilistic inference, the method comprising the following steps: Obtain the geotechnical parameters and field sample data of the target rock slope; Construct a set of candidate probability distribution models corresponding to the soil and rock parameters, and calculate the prior parameter distribution range corresponding to each candidate probability distribution model in the set of candidate probability distribution models for the soil and rock parameters; the prior parameter distribution range includes the prior distribution of each distribution parameter corresponding to each soil and rock parameter; Based on the field sample data and the prior distribution, Bayesian inference and Markov chain Monte Carlo sampling are used to estimate the posterior parameter distribution of each distribution parameter in each candidate probability distribution model; Calculate the posterior weights for each of the candidate probability distribution models; Construct a hybrid multi-model probability density function based on all the said posterior parameter distributions and all the said posterior weights; Construct a slope stability proxy model; The hybrid multi-model probability density function is input into the slope stability surrogate model to calculate the failure probability of the target rock slope; The reliability level of the target rock slope is assessed based on the failure probability.

[0006] In some embodiments, the geotechnical parameters include zero confining pressure shear strength, nonlinear coefficient, and geotechnical strength index; the candidate probability distribution model includes a log-normal distribution model, a Gamma distribution model, an inverse Gaussian distribution model, or a log-logistic distribution model.

[0007] In some embodiments, estimating the posterior parameter distribution of each distribution parameter in each candidate probability distribution model based on the field sample data and the prior distribution, using Bayesian inference and Markov chain Monte Carlo sampling, includes: Based on the on-site sample data and the prior distribution, the first posterior distribution of each distribution parameter in each candidate probability distribution model is calculated using the Bayesian method. The first posterior distribution is sampled in parallel using multiple Markov chains; Calculate the convergence diagnostic index of the Markov chain based on the sampled data; If the convergence diagnostic index is determined to be less than a preset value, the posterior parameter distribution is calculated based on the current sampled data.

[0008] In some embodiments, calculating the posterior weights of each candidate probability distribution model includes: Calculate the modified Akaike Information Criterion value for each of the candidate probability distribution models; Select the smallest modified Akaike information criterion value from all the modified Akaike information criterion values ​​as the benchmark value; Calculate the difference between each of the modified Akaike information criterion values ​​and the baseline value; The posterior weight of each candidate probability distribution model is calculated based on the difference.

[0009] In some embodiments, constructing the slope stability proxy model includes: Several sets of input parameter samples are constructed based on the statistical characteristics of the soil and rock parameters; The input parameter samples are respectively input into a preset numerical simulation software, and the corresponding sample slope safety factor is calculated using the strength reduction method to form several sets of input-output data pairs; Based on the input-output data pairs, a proxy model to be processed is constructed using the regularized multivariate tensor spline method; Calculate the Nash-Sutcliffe efficiency coefficient of the agent model to be processed; Once the Nash-Sutcliffe efficiency coefficient is determined to be within the target range, the current surrogate model to be processed is used as the slope stability surrogate model.

[0010] In some embodiments, inputting the hybrid multi-model probability density function into the slope stability surrogate model to calculate the failure probability of the target rock slope includes: The hybrid multi-model probability density function is input into the slope stability surrogate model to calculate several target slope safety factors for the target rock slope. Calculate the failure probability of the target rock slope based on all the target slope safety factors.

[0011] To achieve the above objectives, another aspect of this application proposes a rock slope reliability assessment device based on hybrid multi-model probabilistic inference, the device comprising: The acquisition module is used to acquire the geotechnical parameters and field sample data of the target rock slope; The first construction module is used to construct a set of candidate probability distribution models corresponding to the soil and rock parameters, and to calculate the prior parameter distribution range corresponding to each candidate probability distribution model in the set of candidate probability distribution models for the soil and rock parameters; the prior parameter distribution range includes the prior distribution of each distribution parameter corresponding to each soil and rock parameter. The estimation module is used to estimate the posterior parameter distribution of each distribution parameter in each candidate probability distribution model based on the field sample data and the prior distribution, using Bayesian inference and Markov chain Monte Carlo sampling. The first calculation module is used to calculate the posterior weights of each of the candidate probability distribution models; The second construction module is used to construct a hybrid multi-model probability density function based on the distribution of all the posterior parameters and all the posterior weights. The third building module is used to build a slope stability proxy model; The second calculation module is used to input the hybrid multi-model probability density function into the slope stability surrogate model to calculate the failure probability of the target rock slope. An evaluation module is used to assess the reliability level of the target rock slope based on the failure probability.

[0012] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0013] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described above.

[0014] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method.

[0015] The embodiments of this application include at least the following beneficial effects: This application provides a method and related equipment for assessing the reliability of rock slopes based on hybrid multi-model probabilistic inference. This scheme obtains the soil and rock parameters and field sample data of the target rock slope, constructs a set of candidate probability distribution models corresponding to the soil and rock parameters, calculates the prior distribution of the soil and rock parameters in each candidate probability distribution model in the set of candidate probability distribution models, and then, based on the field sample data and prior distribution, uses Bayesian inference and Markov chain Monte Carlo sampling to estimate the posterior parameter distribution of each distribution parameter in each candidate probability distribution model, calculates the posterior weight of each candidate probability distribution model, constructs a hybrid multi-model probability density function based on all posterior parameter distributions and all posterior weights, constructs a slope stability surrogate model, and then inputs the hybrid multi-model probability density function into the slope stability surrogate model to calculate the failure probability of the target rock slope. Based on the failure probability, the reliability level of the target rock slope is assessed, thereby effectively improving the accuracy of slope reliability assessment results with limited data. Attached Figure Description

[0016] Figure 1 This is a flowchart of the rock slope reliability assessment method based on hybrid multi-model probabilistic inference provided in the embodiments of this application; Figure 2 This is a complete implementation flowchart of the rock slope reliability assessment method based on hybrid multi-model probabilistic inference provided in the embodiments of this application; Figure 3 This is a schematic diagram of the engineering geological profile of a strongly weathered siltstone slope of a highway provided in an embodiment of this application; Figure 4 This is a schematic diagram of the Markov chain Monte Carlo (MCMC) sampling process provided in the embodiments of this application; Figure 5 This is a comparative diagram of the cumulative distribution function of slope failure probability obtained by different reliability assessment methods provided in the embodiments of this application; Figure 6 This is a schematic diagram comparing the slope failure probabilities obtained by three reliability assessment methods under different reinforcement schemes provided in the embodiments of this application. Figure 7 This is a schematic diagram of the structure of the rock slope reliability assessment device based on hybrid multi-model probabilistic inference provided in the embodiments of this application; Figure 8 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application.

[0018] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0019] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0021] In related technologies, rock slope reliability assessment methods typically assume that the input parameters follow a pre-defined probability distribution, such as a log-normal distribution. However, field data in geotechnical engineering is limited, making it impossible to determine which probability distribution best reflects the actual scenario. Existing rock slope reliability assessment methods often ignore the impact of this uncertainty, leading to overly optimistic assessment results—that is, underestimating the probability of failure—and reducing the accuracy of slope reliability assessments, potentially resulting in suboptimal engineering designs. Furthermore, existing methods often rely on a single likelihood function, i.e., a single distribution form, which also fails to address the impact of model uncertainty, thus hindering the improvement of slope reliability assessment accuracy.

[0022] In view of this, this application provides a method and related equipment for assessing the reliability of rock slopes based on hybrid multi-model probabilistic inference, which can effectively improve the accuracy of slope reliability assessment results with limited data.

[0023] The rock slope reliability assessment method based on hybrid multi-model probabilistic inference provided in this application relates to the field of geotechnical engineering reliability analysis technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the rock slope reliability assessment method based on hybrid multi-model probabilistic inference, but is not limited to the above forms.

[0024] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0025] The embodiments of this application will be described in detail below with reference to the accompanying drawings: Figure 1 This is an optional flowchart of the rock slope reliability assessment method based on hybrid multi-model probabilistic inference provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S110 to S180: Step S110: Obtain the soil and rock parameters and field sample data of the target rock slope; Step S120: Construct a set of candidate probability distribution models corresponding to the soil and rock parameters, and calculate the prior parameter distribution range corresponding to each candidate probability distribution model in the set of candidate probability distribution models for the soil and rock parameters; wherein, the prior parameter distribution range includes the prior distribution of each distribution parameter corresponding to each soil and rock parameter; Step S130: Based on the field sample data and the prior distribution, Bayesian inference and Markov chain Monte Carlo sampling are used to estimate the posterior parameter distribution of each distribution parameter in each candidate probability distribution model. Step S140: Calculate the posterior weights of each candidate probability distribution model; Step S150: Construct a mixture of multiple model probability density functions based on the distributions of all posterior parameters and all posterior weights; Step S160: Construct a slope stability proxy model; Step S170: Input the hybrid multi-model probability density function into the slope stability surrogate model to calculate the failure probability of the target rock slope; Step S180: Assess the reliability level of the target rock slope based on the failure probability.

[0026] It is understood that the geotechnical parameters in this embodiment include, but are not limited to, zero confining pressure shear strength. Nonlinear coefficients Address strength index Elastic modulus Tensile strength intercept and Hoek-Brown material constants The field sample data in this embodiment comes from field surveys, laboratory tests, or in-situ tests. The candidate probability distribution model set in this embodiment... Including There are several candidate probability distribution models. These models can be lognormal, Gamma, inverse Gaussian, loglogic, Nakagami, or Weibull.

[0027] Specifically, in this embodiment, after constructing the set of candidate probability distribution models corresponding to the geotechnical parameters, for each candidate probability distribution model... Each geotechnical parameter is determined based on prior engineering knowledge. The prior parameter distribution range is defined. The prior engineering knowledge originates from regional geological data, similar engineering experience, or relevant literature reports. Each candidate probability distribution model includes several distribution parameters, and the range of distribution parameters formed for each geotechnical parameter is different in each candidate probability distribution model. Therefore, the prior parameter distribution range in this embodiment includes the prior distribution of each distribution parameter corresponding to each geotechnical parameter.

[0028] It is understood that, after obtaining the field sample data and prior distribution, this embodiment uses Bayesian inference and Markov chain Monte Carlo sampling to estimate the posterior parameter distribution of each distribution parameter in each candidate probability distribution model. Specifically, this embodiment can use Bayesian methods to calculate the first posterior distribution of each distribution parameter in each candidate probability distribution model based on the field sample data and prior distribution. Then, it performs parallel sampling of the first posterior distribution through multiple Markov chains, calculates the convergence diagnostic index of the Markov chains based on the sampled samples, and calculates the posterior parameter distribution based on the current sampled samples when the convergence diagnostic index is less than a preset value. For example, using a candidate probability distribution model... For example, using the on-site sample data D, the candidate probability distribution model is calculated using Bayes' theorem. Medium soil and rock parameters The probability of the posterior distribution : ; In the formula, Let be the likelihood function. For the prior distribution, This represents the marginal likelihood.

[0029] Then, the Markov Chain Monte Carlo (MCMC) method is used to calculate the probability of the above posterior distribution. Perform numerical sampling. Set up multiple Markov chains for parallel sampling, with each chain performing at least 10 samples. 4 Next, the Gelman-Rubin convergence diagnostic index was used. R stat Evaluate the convergence of the chain when R stat When the value is less than 1.2, the decision chain has reached a stable posterior distribution. The converged sample is used as the posterior distribution estimate of the parameters of each candidate model.

[0030] Simultaneously, this embodiment also calculates the posterior weights of each candidate probability distribution model. Specifically, this embodiment can calculate the modified Akaike Information Criterion value for each candidate probability distribution model; select the smallest modified Akaike Information Criterion value from all modified Akaike Information Criterion values ​​as the benchmark value; calculate the difference between each modified Akaike Information Criterion value and the benchmark value; and calculate the posterior weights of each candidate probability distribution model based on the differences. For example, using candidate probability distribution models... For example, candidate probability distribution model Corrected Akaike Information Criterion Value The calculation formula is as follows: ; This represents the maximum likelihood function value at the maximum likelihood function position. Candidate probability distribution model The number of parameters represented; This indicates the sample size of the on-site sample data.

[0031] In the model of all candidate probability distributions Minimum value As a baseline, calculate the candidate probability distribution model. correspond The difference between the value and the benchmark value : ; In the formula, Represents the candidate probability distribution model of value.

[0032] For candidate probability distribution model posterior weights The calculation process is as follows: ; In the formula, This indicates that the j-th candidate probability distribution model corresponds to... The difference between the value and the benchmark value.

[0033] Among them, for candidate probability distribution model Its corresponding posterior weights Used to characterize candidate probability distribution models given data conditions. This represents the probability of the optimal model.

[0034] Specifically, other candidate probability distribution models The process of calculating the value and the posterior weight, and the candidate probability distribution model. The calculation process and posterior weights The calculation process is the same.

[0035] It is understandable that this embodiment obtains the probability distribution model of all candidates. The probability and posterior weights of the posterior parameter distribution Then, a hybrid multi-model probability density function is constructed based on the distributions of all posterior parameters and all posterior weights. Specifically, in this embodiment, the probability density function is first constructed based on the posterior distribution of each candidate probability distribution model. After determining the posterior prediction density function, it is combined with the posterior weights. Calculate the probability density function of the mixed multi-model for soil parameter x by combining soil parameters: ; In the formula, This represents the probability density function of the hybrid multi-model corresponding to the geotechnical parameter x; The candidate probability distribution model is the posterior prediction density function corresponding to the m-th candidate probability distribution model.

[0036] Specifically, this embodiment employs a stochastic ensemble weighting mechanism to optimize the aforementioned hybrid multi-model probability density function. This embodiment defines a set of posterior weight vectors { w n} n =1 Each posterior weight vector corresponds to a combination of candidate probability distribution models. This embodiment is based on the original log-likelihood estimation. Normalization is performed: ; ; Then, based on the normalized log-likelihood estimate, the probability density function of the mixture multi-model is optimized to obtain the expression of the probability density function of the mixture multi-model in this embodiment: ; In the formula, N is the total number of weight vector schemes. For the first n The number of candidate probability distribution models included in each weighting scheme. For the first n In the weighting scheme, the first... m The weight coefficients of each candidate probability distribution model.

[0037] It is understandable that this embodiment will construct a slope stability surrogate model after obtaining the hybrid multi-model probability density function. In this embodiment, since slope stability analysis involves nonlinear failure criteria and complex stress-strain mechanisms, it is difficult to derive an analytical performance function. Therefore, this embodiment constructs an explicit surrogate model as the slope stability surrogate model. Specifically, this embodiment can construct several sets of input parameter samples based on the statistical characteristics of geotechnical parameters, input the input parameter samples into a preset numerical simulation software, and use the strength reduction method to calculate the corresponding sample slope safety factor to form several sets of input-output data pairs; then, based on the input-output data pairs, a surrogate model to be processed is constructed using the regularized multivariate tensor spline method; the Nash-Sutcliffe efficiency coefficient of the surrogate model to be processed is calculated, and it is determined that the Nash-Sutcliffe efficiency coefficient is within the target range, and the current surrogate model to be processed is used as the slope stability surrogate model.

[0038] In this embodiment, the statistical characteristics of the geotechnical parameters include the mean and standard deviation. H sets of input parameter samples (H≥500) are generated using Latin hypercube (LHS). The preset numerical simulation software for this embodiment can be ABAQUS. This embodiment uses the strength reduction method to calculate the corresponding slope safety factor FS. The rock constitutive model adopts the Hoek-Brown criterion, and the nonlinear Mohr-Coulomb failure surface is implemented using the UMAT subroutine.

[0039] This embodiment is based on the H-group input-output data pair { X , G The proxy model to be processed is constructed using the Regularized Multivariate Tensor Spline (RMTS) method. ; In the formula, For response field output; x k For the first k One input variable, Its corresponding basis functions of order; These are the multivariate spline coefficients.

[0040] The agent model to be processed is represented in matrix form G=F. , where F∈ × n To design the matrix, a regularization method based on the energy minimization principle is used to solve for the optimal coefficients: ; In the formula, ∈ n ×n This is the discretized energy matrix obtained from the second derivative of the response field.

[0041] The Nash-Sutcliffe efficiency coefficient (NSE) was used to evaluate the predictive accuracy of the surrogate model. Calculated on LHS validation set samples: ; In the formula, Let i be the true value of the i-th validation sample. Let be the predicted value of the RMTS surrogate model for the i-th validation sample. This is the average of the original numerical simulation results for all validation samples.

[0042] when When the target range [0.75, 1.0] is met, the current surrogate model is determined to have excellent predictive performance. Therefore, the current surrogate model is used as the slope stability surrogate model.

[0043] Understandably, in this embodiment, after calculating the slope stability surrogate model and the hybrid multi-model probability density function, the hybrid multi-model probability density function is input into the slope stability surrogate model to calculate several target slope safety factors for the target rock slope; then, based on all target slope safety factors, the failure probability of the target rock slope is calculated. Specifically, this embodiment uses the hybrid multi-model probability density function as the input distribution of geotechnical parameters, and calculates the target slope safety factor FS under a large number of random implementations through the slope stability surrogate model. The slope failure probability is then calculated based on the proportion of implementations where FS < 1 according to the limit state function Z = FS⁻¹. ; In the formula, Indicates the probability of slope failure; This represents the total number of times the safety factor of all target slopes occurs.

[0044] This embodiment assesses the reliability level of a target rock slope based on the calculated slope failure probability and with reference to the geotechnical engineering performance level grading standard. This effectively improves the accuracy of slope reliability assessment results even with limited data. Specifically, this embodiment can also output corresponding reinforcement design suggestions based on the calculated slope failure probability and its confidence interval to improve slope safety.

[0045] In some embodiments, a strongly weathered siltstone slope of a highway is taken as the research object, such as... Figure 3As shown, the slope is approximately 45m high with an angle of about 50°, and consists of a completely weathered layer, a strongly weathered layer, and a moderately weathered layer from top to bottom. The slope stability is primarily controlled by the mechanical properties of the strongly weathered siltstone, which has a fractured rock structure and well-developed joints and fissures, making it a typical rock slope controlled by nonlinear failure criteria. Figure 2 As shown, the complete implementation process of the method in this application embodiment in this scenario includes, but is not limited to, the following steps: Step S210: Obtain geotechnical parameters and field sample data.

[0046] Specifically, for this slope, this embodiment obtained 25 sets of field sample data corresponding to geotechnical parameters through on-site investigation, borehole sampling, and laboratory testing (as shown in Table 1). The obtained geotechnical parameters include elastic modulus. (GPa), tensile strength intercept (MPa), zero confining pressure shear strength (MPa), Hoek-Brown material constant Geological strength index GSI and nonlinear coefficient Each set of data includes the spatial location information of the sampling points (horizontal coordinates). and vertical coordinates The descriptive statistical characteristics of the 25 sets of field sample data are shown in Table 2. The mean values ​​of each parameter are as follows: GPa, MPa MPa GSI = 33 The standard deviation of each parameter reflects its degree of dispersion. The sample size of the above data is finite. This reflects a typical data sparsity problem in geotechnical engineering.

[0047] Table 1

[0048] Table 2

[0049] Step S220: Construct a set of candidate probability distribution models corresponding to geotechnical parameters.

[0050] Specifically, this embodiment focuses on three of the most critical geotechnical parameters (zero confining pressure shear strength). Nonlinear coefficients For both the geological intensity index (GSI) and the geological intensity index (GSI), a set of six candidate probability distribution models was constructed. The six candidate probability distribution models are the lognormal distribution, the Gamma distribution, the inverse Gaussian distribution, the loglogistic distribution, the Nakagami distribution, and the Weibull distribution.

[0051] Based on regional geological data, engineering experience with similar weathered rock masses, and relevant literature reports, the prior mean and standard deviation ranges of each soil and rock parameter were determined (as shown in Table 3). On this basis, the prior information was transformed into the prior parameter ranges of each candidate probability distribution model (as shown in Table 4).

[0052] Taking the log-normal distribution as an example, its two distribution parameters (Logarithmic mean) and The prior ranges for (logarithmic standard deviation) are as follows: for parameter, , ;for parameter, , For GSI parameters, , The parameter ranges for the remaining candidate models follow the same pattern (see Table 4 for details). The above prior ranges are set according to the principle of "broad and reasonable", ensuring that the prior information is broad enough to accommodate all physically possible values ​​and avoiding bias in posterior inference caused by an overly narrow prior range.

[0053] Table 3

[0054] Table 4

[0055] Step S230, Bayesian posterior inference and Markov chain Monte Carlo (MCMC) sampling.

[0056] Specifically, for each candidate probability distribution model Using the 25 sets of field data in Table 1 Bayes' theorem is used for posterior parameter inference. A prior distribution is defined. For a uniform distribution (i.e., a uniform distribution within the range defined in Table 4), the likelihood function... Constructed based on the probability density function of each candidate model.

[0057] The Markov Chain Monte Carlo (MCMC) method is used to numerically sample the posterior distribution of each candidate model. Four parallel Markov chains are set up, and each chain samples 10 times.4 The first 3,000 samples were discarded as a pre-burn-in period, and the remaining 7,000 samples were used as a numerical approximation of the posterior distribution.

[0058] The convergence of MCMC sampling was assessed using the Gelman-Rubin convergence diagnostic index. R stat Conduct an assessment. For example... Figure 4 As shown, the convergence diagnostic metrics for all candidate probability distribution models R stat The values ​​consistently decreased to below 1.2 after approximately 2000 iterations, indicating that each chain had converged to a stable posterior distribution. Finally, the mean and standard deviation of the posterior samples were used as Bayesian posterior estimates of the parameters for each candidate model.

[0059] Step S240: Calculate the posterior weights of each candidate probability distribution model; This embodiment calculates the posterior weights of each candidate probability distribution model based on the posterior inference results of each model using the Modified Akaike Information Criterion (AICc). Specifically, first, the AICc value of each candidate probability distribution model at the maximum likelihood estimation is calculated; then, using the minimum AICc value among all candidate probability distribution models as a benchmark, the AICc difference between each candidate probability distribution model is calculated. And based on this, calculate the posterior weights of each model. .

[0060] For example, with zero confining shear strength For example (as shown in Table 5), the AICc values ​​of the six candidate probability distribution models are as follows: log-normal distribution 78.93, Gamma distribution 79.45, inverse Gaussian distribution 80.28, log-logistic distribution 81.03, Nakagami distribution 79.68, and Weibull distribution 83.27. The log-normal distribution has the smallest AICc value (78.93), therefore its... The corresponding posterior weight is 0.265. The posterior weights for the Gamma and Nakagami distributions are 0.204 and 0.183, respectively; the posterior weight for the inverse Gaussian distribution is 0.136; the posterior weight for the log-logistic distribution is 0.092; and the posterior weight for the Weibull distribution is 0.030. These weights satisfy the normalization constraint. .

[0061] Similarly, for the GSI parameter, the inverse Gaussian distribution has the smallest AICc value (146.95) with a posterior weight of 0.228; for the nonlinear coefficient... The log-normal distribution has the smallest AICc value (42.18), with a weight of 0.278. Table 5 summarizes the complete weight calculation results for the three parameters under each candidate model.

[0062] Table 5

[0063] Step S250: Based on the calculated posterior weights The probability density function of the hybrid multi-model for geotechnical parameters was constructed according to the MMPI framework.

[0064] by For example, its MMPI mixture probability density function is expressed as: ; In the formula, For the first The posterior prediction density corresponding to each candidate model This represents the AICc weights for the model.

[0065] Specifically, this embodiment can employ a randomized ensemble weighting mechanism to optimize the aforementioned mixture density. This is based on the original log-likelihood estimation. After normalization, an optimized weight distribution is obtained. The final constructed hybrid probability density can simultaneously characterize the relative contribution of each candidate model and its combined performance.

[0066] like Figure 5 As shown by the CDF curve, the probability density distribution (represented by the cumulative distribution function) generated by the MMPI framework takes into account the goodness of fit of each candidate model, avoiding the bias caused by the selection of a single model. Compared with a single candidate model (such as using only a log-normal distribution) and a simple multi-model equal-weighted average, this embodiment constructs a more robust and flexible probability density representation through dual optimization of AICc weighting and random ensemble weights.

[0067] Step S260: Based on the calculated hybrid multi-model probability density function and slope stability surrogate model, calculate the slope safety factor FS.

[0068] Specifically, this embodiment uses the hybrid multi-model probability density function as the input distribution of geotechnical parameters and calculates the slope safety factor FS using a pre-constructed RMTS surrogate model (regularized multivariate tensor spline surrogate model). The RMTS surrogate model is constructed using 500 sets of Latin hypercube sampling (LHS) training samples, and its Nash-Sutcliffe efficiency coefficient (NSE) reaches 0.961, indicating that the surrogate model has excellent prediction accuracy.

[0069] For three reliability assessment methods (traditional Bayesian method (single model), multi-model quantization method (multi-model equal weight or discrete ensemble), and the MMPI method of this invention), the slope failure probability is calculated respectively. Each method uses candidate model densities of 1000, 3000, and 10000 respectively. J Simulations were performed to evaluate the convergence and stability of the results.

[0070] like Figure 5 As shown, the cumulative distribution functions (CDF) of failure probabilities obtained by the three methods exhibit significant differences: (1) Traditional Bayesian method: The CDF curve is the steepest and the distribution has the least dispersion. Its mean failure probability is about 1.67% (see Table 6), and the standard deviation is about 0.43%, indicating that the evaluation results given by this method are too concentrated and fail to fully reflect the uncertainty. The probability of failure exceeding the 2.3% performance threshold estimated by this method is only 7.62%.

[0071] (2) Multi-model quantization method: The CDF curve is flatter than that of the traditional method, and the dispersion of the distribution is increased. Its mean failure probability is about 1.90%, the standard deviation is about 0.65%, and the probability of exceeding the 2.3% threshold is about 27.81%. Although this method takes into account the existence of multiple candidate models, it lacks an effective probability weighted integration mechanism, resulting in high uncertainty of the evaluation results.

[0072] (3) The MMPI method of this application embodiment: the CDF curve lies between the two, but is generally shifted to the right. Its average failure probability is 2.28% ( J =10000), with a standard deviation of 0.548%, and a probability of exceeding the 2.3% performance threshold as high as 46.54% (as shown in Table 6). Compared with the above methods, the MMPI framework in this embodiment provides a more comprehensive and robust failure probability estimate based on a comprehensive consideration of parameter uncertainty and model form uncertainty, demonstrating significant methodological advancement.

[0073] Specifically, although the mean failure probability predicted by the MMPI method in this embodiment (2.28%) is only about 20% higher than that of the multi-model quantization method (1.90%), its probability of exceeding the 2.3% threshold (46.54%) is 1.67 times that of the multi-model quantization method (27.81%) and 6.1 times that of the traditional Bayesian method (7.62%). This difference has important decision-making significance in practical engineering—when the slope performance level crosses the boundary between "below average" and "poor", engineers will adopt drastically different reinforcement design strategies.

[0074] Table 6

[0075] Step S270: Decision guidance for slope reinforcement design based on the calculated failure probability.

[0076] Specifically, this embodiment illustrates the practical application value of the MMPI method in engineering reinforcement design decisions through three different anchor cable reinforcement schemes.

[0077] like Figure 6 As shown, the three reinforcement schemes are as follows: Option 1 (Single-stage anchor cable reinforcement): Only one row of prestressed anchor cables is arranged at the toe of the slope, with a design tension of 400kN; Option 2 (two-stage anchor cable reinforcement): A row of prestressed anchor cables is arranged at the toe of the slope and the middle of the slope, with a design tension of 400kN; Option 3 (Three-level anchor cable reinforcement): Three rows of prestressed anchor cables are arranged along the entire height of the slope, with a design tension of 400kN.

[0078] Using the three reliability assessment methods described above, the slope failure probability under each reinforcement scheme was calculated. The calculation results are summarized in Table 6. Based on Table 6 and Figure 6 As a result, the following conclusions can be drawn from this embodiment: (1) Traditional Bayesian method (single model): The average failure probability of the first scheme is about 0.04%, which has reached the "above average" performance level. The requirements of this method are as follows. If engineers make decisions solely based on this method, they will tend to adopt only the most economical single-stage reinforcement scheme. However, this conclusion is based on the implicit assumption that the model is "formally correct," ignoring a large number of uncertainties and constituting an overly optimistic assessment.

[0079] (2) Multi-model quantization method: The average failure probability of Scheme 1 is about 0.07%, which does not reach the threshold of 0.06%; the average failure probability of Scheme 2 is about 0.05%, which barely meets the performance requirements. This method suggests a two-level hardening scheme, which reflects a certain degree of conservatism. However, due to the failure to effectively integrate model weights, the evaluation results still have a large degree of dispersion.

[0080] (3) The MMPI method of this embodiment: the average failure probability of Scheme 1 is about 0.08%, which obviously does not meet the performance requirements; the average failure probability of Scheme 2 is about 0.06%, which is still close to or slightly higher than the threshold; the average failure probability of Scheme 3 is about 0.03%, which is significantly lower than the threshold of 0.06%, and meets the performance level of "above average".

[0081] The above comparison results show that if only the traditional Bayesian method is used, engineers may underestimate the actual risks and assume that a single-level reinforcement is sufficient to meet the safety requirements, thus leaving potential safety hazards. However, by using the MMPI method in this embodiment, engineers will clearly realize that only a three-level full-height reinforcement scheme can meet the design safety objectives with sufficient confidence.

[0082] Therefore, the MMPI framework in this embodiment provides a more conservative and safer decision-making basis for engineering design by comprehensively quantifying parameter uncertainty and model form uncertainty, and is especially suitable for reliability assessment of geotechnical engineering under sparse data conditions.

[0083] Furthermore, based on the values ​​in Table 6, it can be seen that the MMPI method in this embodiment, at J=1000, has statistical indicators (mean failure probability 2.211%, standard deviation 0.593%) that are very close to those at J=10000 (mean 2.284%, standard deviation 0.548%), and the difference in the probability exceeding the 2.3% threshold (41.77% vs. 46.54%) is within an acceptable range. This indicates that the method in this embodiment can achieve stable convergence with a relatively small computational ensemble size (requiring only 1000 candidate model densities), while traditional methods (such as multi-model quantization) still have a significantly higher standard deviation (0.685%) at J=10000 (0.649%), indicating a slower convergence speed.

[0084] By introducing the RMTS proxy model and the efficient integration mechanism of the MMPI framework, the computation time for a complete reliability assessment in this embodiment is about 20 times that of a set of slope numerical simulations, which is far lower than the tens of thousands of finite element calculations required by traditional direct Monte Carlo simulation, thus significantly improving the engineering applicability of the method.

[0085] Please see Figure 7 This application also provides a rock slope reliability assessment device based on hybrid multi-model probabilistic inference, the device comprising: The acquisition module is used to acquire the geotechnical parameters and field sample data of the target rock slope; The first construction module is used to construct a set of candidate probability distribution models corresponding to the soil and rock parameters, and to calculate the prior parameter distribution range corresponding to each candidate probability distribution model in the set of candidate probability distribution models for the soil and rock parameters; wherein, the prior parameter distribution range includes the prior distribution of each distribution parameter corresponding to each soil and rock parameter; The estimation module is used to estimate the posterior parameter distribution of each distribution parameter in each candidate probability distribution model based on the field sample data and the prior distribution, using Bayesian inference and Markov chain Monte Carlo sampling. The first calculation module is used to calculate the posterior weights of each candidate probability distribution model; The second building module is used to construct a hybrid multi-model probability density function based on the distribution of all posterior parameters and all posterior weights. The third building module is used to build a slope stability proxy model; The second calculation module is used to input the hybrid multi-model probability density function into the slope stability surrogate model and calculate the failure probability of the target rock slope. The assessment module is used to evaluate the reliability level of a target rock slope based on the probability of failure.

[0086] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0087] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0088] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0089] Please see Figure 8 , Figure 8 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 810 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 820 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 820 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 820 and is called and executed by the processor 810 using the methods described in the embodiments of this application. The input / output interface 830 is used to implement information input and output; The communication interface 840 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 850 transmits information between various components of the device (e.g., processor 810, memory 820, input / output interface 830, and communication interface 840); The processor 810, memory 820, input / output interface 830 and communication interface 840 are connected to each other within the device via bus 850.

[0090] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0091] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0092] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0093] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0094] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0095] The rock slope reliability assessment method and related equipment based on hybrid multi-model probabilistic inference provided in this application have the following beneficial effects: (1) Comprehensive quantification of dual uncertainties significantly improves assessment accuracy: This embodiment, by constructing a set of multiple candidate probability distribution models, Bayesian posterior inference, and AICc weighted ensemble, simultaneously quantifies parameter uncertainty (the specific parameters of the distribution are unknown) and model form uncertainty (the type of distribution is unknown) within the same framework for the first time. Compared with traditional Bayesian methods, the mean failure probability given by this invention (e.g., 2.28%) is significantly higher than that of traditional methods (1.67%), and is closer to the true risk level of the slope, avoiding overly optimistic assessment results caused by ignoring model form uncertainty.

[0096] (2) Effectively avoids subjectivity in model selection and improves the reliability of evaluation results: This embodiment does not force the selection of a single "best" distribution model, but instead combines multiple candidate models with reasonable fit under limited data through a weighted integration of multiple models. This mechanism fundamentally avoids the subjective bias introduced by human selection of a single model, making the evaluation results no longer sensitive to model selection.

[0097] (3) Significantly reduced computational costs and improved evaluation efficiency: This embodiment replaces traditional finite element numerical simulation with an RMTS proxy model, requiring only 500 sets of training samples to construct a high-precision (NSE>0.90) proxy model, reducing the computation time for subsequent tens of thousands of safety factor assessments from several days to minutes. Simultaneously, the MMPI framework achieves this with a smaller computational integration scale ( J Stable convergence can be achieved with a value of 1000, which is significantly better than traditional methods.

[0098] (4) Providing quantitative basis for engineering decision-making: This embodiment not only provides a single failure probability point estimate, but also provides a complete probability distribution of the failure probability (such as a 90% confidence interval), enabling engineers to intuitively understand the uncertainty range of the evaluation results. More importantly, by comparing the failure probabilities under different reinforcement schemes, this invention can clearly guide engineers to select the appropriate reinforcement level (such as single-level, two-level, or three-level anchor cable reinforcement), realizing the direct transformation of reliability analysis results into engineering decisions.

[0099] (5) Applicable to real-world engineering scenarios with sparse data: This embodiment is specifically designed to address the common problem of data scarcity in geotechnical engineering. Guided by prior knowledge, a reliable assessment can be completed with only a small amount of field data (such as 20-30 sets), which has good engineering practicality and promotional value.

[0100] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0101] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0102] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0103] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0104] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0105] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0106] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0107] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0108] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0109] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0110] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for assessing the reliability of rock slopes based on hybrid multi-model probabilistic inference, characterized in that, The method includes the following steps: Obtain the geotechnical parameters and field sample data of the target rock slope; Construct a set of candidate probability distribution models corresponding to the soil and rock parameters, and calculate the prior parameter distribution range corresponding to each candidate probability distribution model in the set of candidate probability distribution models for the soil and rock parameters; the prior parameter distribution range includes the prior distribution of each distribution parameter corresponding to each soil and rock parameter; Based on the field sample data and the prior distribution, Bayesian inference and Markov chain Monte Carlo sampling are used to estimate the posterior parameter distribution of each distribution parameter in each candidate probability distribution model; Calculate the posterior weights for each of the candidate probability distribution models; Construct a hybrid multi-model probability density function based on all the said posterior parameter distributions and all the said posterior weights; Construct a slope stability proxy model; The hybrid multi-model probability density function is input into the slope stability surrogate model to calculate the failure probability of the target rock slope; The reliability level of the target rock slope is assessed based on the failure probability.

2. The method according to claim 1, characterized in that, The geotechnical parameters include zero confining pressure shear strength, nonlinear coefficient, and geotechnical strength index; the candidate probability distribution model includes a log-normal distribution model, a Gamma distribution model, an inverse Gaussian distribution model, or a log-logarithmic distribution model.

3. The method according to claim 1, characterized in that, The step of estimating the posterior parameter distribution of each distribution parameter in each candidate probability distribution model based on the on-site sample data and the prior distribution, using Bayesian inference and Markov chain Monte Carlo sampling, includes: Based on the on-site sample data and the prior distribution, the first posterior distribution of each distribution parameter in each candidate probability distribution model is calculated using the Bayesian method. The first posterior distribution is sampled in parallel using multiple Markov chains; Calculate the convergence diagnostic index of the Markov chain based on the sampled data; If the convergence diagnostic index is determined to be less than a preset value, the posterior parameter distribution is calculated based on the current sampled data.

4. The method according to claim 1, characterized in that, The calculation of the posterior weights for each candidate probability distribution model includes: Calculate the modified Akaike Information Criterion value for each of the candidate probability distribution models; Select the smallest modified Akaike information criterion value from all the modified Akaike information criterion values ​​as the benchmark value; Calculate the difference between each of the modified Akaike information criterion values ​​and the baseline value; The posterior weight of each candidate probability distribution model is calculated based on the difference.

5. The method according to claim 1, characterized in that, The construction of the slope stability proxy model includes: Several sets of input parameter samples are constructed based on the statistical characteristics of the soil and rock parameters; The input parameter samples are respectively input into a preset numerical simulation software, and the corresponding sample slope safety factor is calculated using the strength reduction method to form several sets of input-output data pairs; Based on the input-output data pairs, a proxy model to be processed is constructed using the regularized multivariate tensor spline method; Calculate the Nash-Sutcliffe efficiency coefficient of the agent model to be processed; Once the Nash-Sutcliffe efficiency coefficient is determined to be within the target range, the current surrogate model to be processed is used as the slope stability surrogate model.

6. The method according to claim 1, characterized in that, The step of inputting the hybrid multi-model probability density function into the slope stability surrogate model to calculate the failure probability of the target rock slope includes: The hybrid multi-model probability density function is input into the slope stability surrogate model to calculate several target slope safety factors for the target rock slope. Calculate the failure probability of the target rock slope based on all the target slope safety factors.

7. A rock slope reliability assessment device based on hybrid multi-model probabilistic inference, characterized in that, The device includes: The acquisition module is used to acquire the soil and rock parameters and field sample data of the target rock slope; The first construction module is used to construct a set of candidate probability distribution models corresponding to the soil and rock parameters, and to calculate the prior parameter distribution range corresponding to each candidate probability distribution model in the set of candidate probability distribution models for the soil and rock parameters; the prior parameter distribution range includes the prior distribution of each distribution parameter corresponding to each soil and rock parameter. The estimation module is used to estimate the posterior parameter distribution of each distribution parameter in each candidate probability distribution model based on the field sample data and the prior distribution, using Bayesian inference and Markov chain Monte Carlo sampling. The first calculation module is used to calculate the posterior weights of each of the candidate probability distribution models; The second construction module is used to construct a hybrid multi-model probability density function based on the distribution of all the posterior parameters and all the posterior weights. The third building module is used to build a slope stability proxy model; The second calculation module is used to input the hybrid multi-model probability density function into the slope stability surrogate model to calculate the failure probability of the target rock slope. An evaluation module is used to assess the reliability level of the target rock slope based on the failure probability.

8. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

Images