CN110377980B - BP neural network-based rock joint surface peak shear strength prediction method - Google Patents

Abstract

A prediction method of peak shear strength of rock joint surface based on BP neural network comprises the following steps: (1) acquisition of training parameters: 1.1 Acquiring a three-dimensional morphology point cloud data set of the structural surface by a three-dimensional laser scanner, and acquiring the maximum possible contact area ratio and joint surface roughness parameters by a MATLAB program; 1.2 Obtaining the tensile strength of the rock and the normal stress applied to the rock through Brazilian split test, uniaxial test and shearing test; (2) Predicting peak shear strength through the trained neural network: 2.1 (1) the data obtained in the step (1) form neural network trainingIs a predictive dataset of (1); 2.2 Inputting training test data into BP neural network, and according to self-learning condition, the trained network can utilize logic and high-non-linear mapping relationship between data, and utilizes every neuron to already define connection weight value and threshold predictive peak value shear strength value tau _p . The method can be used for rapidly predicting the shear strength of the rock structure.

Description

BP neural network-based rock joint surface peak shear strength prediction method

Technical Field

The invention relates to a prediction method of peak shear strength of a rock joint surface based on a BP neural network, which is suitable for occasions of predicting the peak shear strength of a structural surface by analyzing factors affecting the mechanical performance of the structural surface.

Background

The structural surface is a geological interface which is continuously developed in the rock body in the long history process of rock body formation and geological action, so that weak links of the rock body are formed. For fractured rock mass, the shearing behavior of the structure is particularly important, because the structure controls the deformability and strength of the rock mass, and thus the stability of the rock mass, and the main failure mode is shearing failure along the structural surface. The peak shear strength of the structural plane is affected by a plurality of factors, the positive stress acting on the structural plane is an external factor, and the structural plane has mechanical characteristics such as anisotropy, size effect characteristics and the like and morphological characteristics which are internal factors. Grasselli found from a number of shear tests that the steepest raised areas of the upper and lower surfaces of the rock structure are critical locations for fracture, and believed to play a major role in fracture of the structural face as tensile rather than compressive fracture, taking into account sigma in the shear model _t And a maximum possible contact area ratio A ₀ As an intrinsic factor.

Disclosure of Invention

To overcome the defects of the prior artThe invention provides a prediction method for predicting the peak shear strength of a rock joint surface based on a BP neural network, which comprehensively considers various factors influencing the shear behavior of a structural surface: maximum possible contact area ratio A ₀ Roughness parameters of joint surface

Tensile strength sigma of rock _t Normal stress sigma exerted on rock _n And predicting peak shear strength through the trained BP neural network.

The technical scheme adopted for solving the technical problems is as follows:

the method for predicting the peak shear strength of the rock joint surface based on the BP neural network comprises the following steps:

(1) Acquisition of training parameters

1.1 Acquiring a three-dimensional morphology point cloud data set of the structural surface by a three-dimensional laser scanner, and obtaining the maximum possible contact area ratio A by a MATLAB program ₀ Roughness parameters of joint surface

1.2 Obtaining rock tensile Strength Sigma through Brazil split test, uniaxial test and shearing test _t Normal stress sigma exerted on rock _n ；

(2) Prediction of peak shear strength by trained neural networks

2.1 (1) the data obtained in step (1) form a predicted data set for neural network training

2.2 Inputting the obtained training test data into BP neural network, and according to self-learning condition, the trained network can utilize logic and high-non-linear mapping relationship between data, and utilizes every neuron to already define connection weight value and threshold predictive peak value shear strength value tau _p 。

The beneficial effects of the invention are mainly shown in the following steps: and the influence factors in the shearing process of the structural surface are fully considered, the weight proportion of each factor is analyzed through the highly nonlinear relation established by the neural network, and the shearing strength of the rock structure is rapidly predicted.

Drawings

Fig. 1 is a schematic diagram of a structural plane peak shear strength BP neural network prediction method flow.

FIG. 2 is a graphical representation of the comparison of the original and predicted values of peak shear strength.

Detailed Description

The invention is further described below with reference to the accompanying drawings.

Referring to fig. 1 and 2, a method for constructing a structural plane peak shear strength prediction model based on a BP neural network includes the following steps:

(1) Acquisition of training parameters

1.1 Acquiring a three-dimensional morphology point cloud data set of the structural surface by a three-dimensional laser scanner, and obtaining the maximum possible contact area ratio A by a MATLAB program ₀ Roughness parameters of joint surface

1.2 Obtaining rock tensile Strength Sigma through Brazil split test, uniaxial test and shearing test _t Normal stress sigma exerted on rock _n ；

(2) Prediction of peak shear strength by trained neural networks

2.1 (1) the data obtained in step (1) form a predicted data set for neural network training

2.2 Inputting the obtained training test data into BP neural network, and according to self-learning condition, the trained network can utilize logic and high-non-linear mapping relationship between data, and utilizes every neuron to already define connection weight value and threshold predictive peak value shear strength value tau _p 。

The invention will be specifically described by way of examples:

(1) Constructing a prediction data set for BP neural network training according to 37 groups of test results of Grasselli on 7 rocks

The specific parameters are shown in Table 1./>

TABLE 1

Fig. 2 is a graph comparing the original and predicted values of peak shear strength. The original value of the test peak shear strength of Grasselli et al has better matching degree with the predicted value of the BP neural network. And the prediction result of the BP neural network is higher in reliability.

(2) Peak shear strength criteria set forth by Grasselli et al and Tang Zhicheng et al were chosen for comparison to verify the rationality of BP neural network model predictions.

Shear strength formula proposed by Grasselli et al:

/>

wherein: τ _p For peak shear strength, A ₀ For the maximum possible contact area ratio,

for maximum effective tilt angle in shear direction, C is a dimensionless roughness parameter, and is calculated by fitting, σ _n Is normal stress, sigma _t For rock tensile strength>

Is the basic friction angle and α is the shear direction angle.

Tang Zhicheng et al.

Wherein: τ _p For peak shear strength, sigma _n As a result of the normal stress,

at a basic friction angle, A ₀ For the maximum possible contact area ratio, +.>

For maximum effective tilt angle in shear direction, C is a dimensionless roughness parameter, and is calculated by fitting, σ _t Is the tensile strength of the rock.

The calculation results using the Grasselli model and the Tang Zhicheng model, the prediction results of the BP neural network model and the experimental measured peak shear strength are shown in Table 2.

TABLE 2

Average estimation errors of the Grasselli model, the Tang Zhicheng model and the BP neural network model were 24.1%, 14.8% and 8.2%, respectively. The three models have effectiveness in predicting the shear strength of the natural rock structure, and the BP neural network constructed by the invention has higher precision than Grasselli models and Tang Zhicheng models.

The embodiments described in this specification are merely illustrative of the manner in which the inventive concepts may be implemented. The scope of the present invention should not be construed as being limited to the specific forms set forth in the embodiments, but the scope of the present invention and the equivalents thereof as would occur to one skilled in the art based on the inventive concept.

Claims (1)

1. The method for predicting the peak shear strength of the rock joint surface based on the BP neural network is characterized by comprising the following steps of:

(1) Acquisition of training parameters

1.1 Acquiring a three-dimensional morphology point cloud data set of the structural surface by a three-dimensional laser scanner, and obtaining the maximum possible contact area ratio A by a MATLAB program ₀ Roughness parameters of joint surface

C is a dimensionless roughness parameter for maximum effective tilt along the shear direction;

1.2 Obtaining rock tensile Strength Sigma through Brazil split test, uniaxial test and shearing test _t Normal stress sigma exerted on rock _n ；

(2) Prediction of peak shear strength by trained neural networks

2.1 (1) the data obtained in step (1) form a predicted data set for neural network training

2.2 Inputting the obtained training test data into BP neural network, and according to self-learning condition, the trained network can utilize logic and high-non-linear mapping relationship between data, and utilizes every neuron to already define connection weight value and threshold predictive peak value shear strength value tau _p 。

Images