Rock breaking efficiency prediction method based on image recognition and machine learning

Abstract

A rock breaking efficiency prediction method based on image recognition and machine learning comprises the following steps: acquiring a rock surface image, performing graying processing by applying an image processing technology, and then performing pixel classification and image segmentation so as to clearly divide different minerals into a plurality of independent structure diagrams; modeling a rock material according to mineral components, importing boundary information of the plurality of structure diagrams into a model, generating a plurality of large-particle filled rock blocks with the same size as the plurality of structure diagrams, and calibrating mesoscopic parameter values according to rock mechanical characteristics; establishing a hob rock breaking discrete element calculation model, and performing multiple intrusion rock breaking simulation tests on rock samples with different mineral contents; collecting simulation results to construct a rock breaking efficiency data set; a prediction model is established based on a PSO-BPNN algorithm, and a rock breaking efficiency prediction model is obtained after training; and carrying out real-time prediction on the rock breaking efficiency by utilizing the established model. According to the method, the influence of rock mineral components on the rock breaking effect of the hob is considered, and efficient and accurate prediction of the rock breaking efficiency of the hob is realized by utilizing a method of combining rock microscopic components and macroscopic rock breaking behaviors.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent control, and specifically relates to a method for predicting rock-breaking efficiency based on image recognition and machine learning. Background Art

[0002] Tunnel boring machines (TBMs) have been widely used in urban rail transit, underground engineering, mining and other fields. Their high tunneling efficiency makes them the main equipment for modern tunnel construction. The cutter is one of the core components of the TBM and plays a crucial role in its tunneling efficiency and overall performance. The design, material selection and wear characteristics of the cutter directly affect the rock-breaking ability and construction safety of the TBM. However, when the TBM faces complex geological conditions, its rock-breaking efficiency is comprehensively affected by multiple factors such as rock composition, structure and cutter parameters. Accurately predicting the rock-breaking efficiency is crucial for improving construction efficiency, reducing costs and ensuring construction safety. Extracting rock composition based on image recognition technology, generating a large amount of rock-breaking data through numerical simulation, and combining machine learning methods to train and analyze these data to construct a rock-breaking efficiency prediction model is of great significance for improving tunneling efficiency.

[0003] Existing research methods usually analyze rock composition or tunneling parameters in isolation, lacking in-depth research on the interaction between the two. At present, there is no comprehensive rock-breaking efficiency prediction method that combines image recognition, numerical simulation and machine learning to support the full-process application from formation exploration and analysis to rock-breaking efficiency prediction. Therefore, there is an urgent need to provide a calculation method that integrates formation lithology recognition technology and rock-breaking effect evaluation, so as to more comprehensively realize the intelligent prediction of rock-breaking efficiency and provide a scientific basis for optimizing actual engineering construction parameters. Summary of the Invention

[0004] Aiming at the problems existing in the above-mentioned prior art, the present invention provides a method for predicting rock-breaking efficiency based on image recognition and machine learning. This method can effectively solve the problem that it is difficult for existing rock-breaking technologies to accurately design construction parameters for different rock types. By analyzing the influence of rock mineral composition on cutter rock-breaking, it can combine rock microscopic composition with macroscopic rock-breaking behavior for efficient and accurate prediction of rock-breaking efficiency.

[0005] To achieve the above object, the present invention provides a method for predicting rock-breaking efficiency based on image recognition and machine learning, including the following steps:

[0006] Step 1: Obtain the rock surface image, apply image processing technology to grayscale the obtained rock surface image, and then perform pixel classification and image segmentation to clearly divide different minerals into several independent structure diagrams.

[0007] Step 2: Model the rock material based on the Voronoi Grain - Based Model, import the boundary information of several structure diagrams into the model, generate several large - particle - filled rock blocks with the same size as the several structure diagrams, connect the centroids of the contact polygons to form mineral crystal boundaries, and at the same time, fill smaller particles in the polygon areas; in the numerical simulation software, group the large - particle - filled rock blocks according to the defined mineral layers, divide and adjust the meso - mechanical properties of different mineral types, and calibrate a set of the most suitable meso - parameter values to accurately reflect the composition and micro - structure characteristics of the rock;

[0008] Step 3: Establish a cutter rock - breaking model. Use the GBM model to randomly generate several divided areas for filling particles according to the particle size of rock minerals, fill the divided areas with spherical particles, group and name each divided area according to the proportion of the mineral composition content of the rock sample, and then, according to the grouping results, assign different mineral compositions and corresponding meso - parameters to different components; after the particles are generated, apply confining pressure to them, use the wall to simulate the cross - section shape of the cutter, conduct multiple intrusion simulation tests on rock specimens with different mineral contents, and change the rock mineral content, cutter spacing, and penetration depth, and conduct intrusion simulation tests by combining them in turn to achieve the failure within the mineral crystal;

[0009] Step 4: According to the numerical simulation process, collect the results of each simulation, record the mineral content, cutter spacing, penetration degree, and rock - breaking efficiency of the rock sample, and construct a data set for subsequent analysis and modeling use;

[0010] Step 5: Establish a prediction model based on the PSO - BPNN algorithm; divide the constructed data set into a training set and a test set according to a set ratio, use the training set to train the back - propagation neural network optimized by the particle swarm optimization, and at the same time, use the PSO algorithm to optimize the weights and biases of the back - propagation neural network so that it can more effectively learn the complex relationship between the input and output during the training process; by adjusting the velocity and position of the particles, make the PSO algorithm continuously search for the optimal fitness solution, thereby improving the performance of the model to obtain a rock - breaking efficiency prediction model; input the test data into the rock - breaking efficiency prediction model to obtain the corresponding rock - breaking efficiency prediction results, and use the set evaluation indicators to evaluate the prediction accuracy and reliability of the model;

[0011] Step 6: Prediction of rock - breaking efficiency; during the rock - breaking operation process, collect the rock surface images in real - time. After pre - processing the rock surface images, input them into the rock - breaking efficiency prediction model, use the rock - breaking efficiency prediction model to achieve the prediction of the rock - breaking efficiency, and output the prediction results.

[0012] As an optimization, in Step 5, the data set is randomly divided into a training set and a test set according to a ratio of 7:3.

[0013] As a preference, in step one, the Image J digital image processing and analysis software is used to perform grayscale processing on the rock surface image.

[0014] As a preference, in step two, the "trial and error calibration method" is adopted, and the mesoscopic parameter values are calibrated through continuous adjustment and trial calculation.

[0015] As a preference, in step three, the mineral content, hob spacing, and penetration degree of the rock in each simulation process are collected, and the specific energy SE is calculated according to formula (1) by using the work done by the hob and the rock-breaking area;

[0016]

[0017] In the formula, W is the work done by the tool in the rock-breaking process, that is, the external force work; V is the volume of the rock generated by the rock-breaking test; Wv is the work done by the penetration force of the hob; W r is the work done by the rolling force of the hob. When only considering the intrusion effect of the cutter head, W r = 0; S is the rock-breaking area;

[0018] As a preference, in step five, the process of obtaining the corresponding rock-breaking efficiency prediction result is as follows:

[0019] A1: Obtain the output h of the hidden layer according to formula (2);

[0020] h = f(W ih ·X + b h ) (2);

[0021] In the formula, W ih is the weight matrix from the input layer to the hidden layer, X is the input vector, b h is the bias of the hidden layer, and f(·) is the activation function tansing, whose output range is [-1, 1];

[0022] A2: Obtain the prediction result y of the output layer according to formula (3);

[0023] y = g(W ho ·h + b o ) (3);

[0024] In the formula, W ho is the weight matrix from the hidden layer to the output layer, b o is the bias of the output layer, and g(·) is the activation function purelin of the output layer, whose output range is (-∞, +∞).

[0025] As a preference, in step five, the process of optimizing the weights of the backpropagation neural network by using the PSO algorithm is as follows:

[0026] B1: Calculate the mean square error L according to formula (4);

[0027]

[0028] where N is the number of test samples, y i is the actual output, is the predicted output;

[0029] B2: Take the derivative of formula (4) and obtain the gradient of the loss with respect to the weights according to formula (5)

[0030]

[0031] where is the gradient of the loss with respect to the output, is the gradient of the output with respect to the weights;

[0032] B3: Obtain the updated weights W according to formula (6);

[0033]

[0034] where η is the learning rate, which controls the step size of weight update.

[0035] As an optimization, in step five, the process of adjusting the velocity and position of the particles is as follows:

[0036] C1: Update the velocity of the particle according to formula (7)

[0037]

[0038] where is the velocity of particle i at time t, w is the inertia weight, which controls the motion inertia of the particle, c1 is the individual learning factor, which controls the degree to which the particle moves towards its best position, c2 is the swarm learning factor, which controls the degree to which the particle moves towards the swarm best position, r1 and r2 are random numbers in the range [0, 1], p i is the best position of particle i, is the position of particle i at time t, g is the best position of the swarm;

[0039] C2: Update the position of the particle according to formula (8)

[0040]

[0041] In the present invention, by processing the acquired rock surface image based on image processing technology, particles of different mineral types can be grouped through preliminary analysis, so as to accurately obtain a number of structure diagrams, which is beneficial to the subsequent construction of an accurate rock sample model; by using numerical simulation to model the rock material and calibrate the mesoscopic parameter values, the constructed rock sample model can accurately reflect the composition and microscopic structure characteristics of the rock; by using the hob rock-breaking model to conduct hob penetration tests based on the calibrated mesoscopic parameters and changing the rock mineral content, hob spacing, and penetration depth for combined tests, a large amount of numerical simulation data can be accurately obtained, which is beneficial to constructing a high-precision rock-breaking efficiency prediction model based on this. By using machine learning technology to learn and train the obtained numerical simulation data, the complex relationship between parameters such as rock composition, cutter spacing, and penetration degree and rock-breaking efficiency can be effectively captured, so as to achieve more accurate rock-breaking efficiency prediction and provide a scientific basis for optimizing construction parameters such as penetration degree.

[0042] The present invention provides a comprehensive and efficient rock-breaking efficiency prediction method based on image recognition and machine learning, which effectively solves the problem that it is difficult for existing rock-breaking technologies to accurately design construction parameters for different rock strata types. By analyzing the influence of rock mineral composition on hob rock-breaking, it can combine the rock microscopic composition with the macroscopic rock-breaking behavior to efficiently and accurately predict the rock-breaking efficiency. This method can not only more accurately predict the rock-breaking efficiency, but also further optimize the cutter spacing and penetration degree according to the prediction results to achieve higher construction efficiency. Brief Description of the Drawings

[0043] Figure 1 is the flow chart of the present invention;

[0044] Figure 2 is the schematic diagram of the rock surface image obtained in the present invention;

[0045] Figure 3 is the structure diagram of the rock surface image in the present invention;

[0046] Figure 4 is the modeling diagram of the rock structure in the present invention;

[0047] Figure 5 is the schematic diagram of the structure before hob rock-breaking in the present invention;

[0048] Figure 6 is the schematic diagram of the structure after hob rock-breaking in the present invention;

[0049] Figure 7 is the fitting effect diagram of the BP neural network in the present invention. Detailed Description of the Invention

[0050] The present invention will be further described below in conjunction with the accompanying drawings.

[0051] As Figure 1 shown, the present invention provides a method for predicting rock-breaking efficiency based on image recognition and machine learning, including the following steps:

[0052] Step 1: As Figure 2 shown, obtain the rock surface image, apply image processing technology to grayscale the obtained rock surface image, and then perform pixel classification and image segmentation to clearly divide different minerals into several independent structure diagrams, as Figure 3 shown;

[0053] Specifically, use a high-definition camera to obtain the rock surface image by shooting. On this basis, introduce digital image processing technology, which can preliminarily analyze the composition of the rock sample without additional physical information, determine the mineral composition of the rock sample, and use the obtained different composition diagrams for the subsequent numerical simulation modeling process, which is beneficial to increasing the overall reference value of the subsequent model;

[0054] Step 2: As Figure 4 shown, build a model of the rock material based on the Voronoi Grain-Based Model (GBM), import the boundary information of several structure diagrams into the model, generate several large-particle-filled rock blocks with the same size as the several structure diagrams, connect the centroids of the contact polygons to form mineral crystal boundaries, and at the same time, fill smaller particles in the polygon area; group the large-particle-filled rock blocks according to the defined mineral layers in the numerical simulation software, divide and adjust the meso-mechanical properties of different mineral types, and calibrate a set of the most suitable meso-parameters to accurately reflect the composition and microstructure characteristics of the rock;

[0055] Specifically, import the boundary information of several structure diagrams into PFC2D. For the generated several large-particle-filled rock blocks, the light gray particles are feldspar, the gray particles are quartz, and the black particles are mica. As a further preference, the contents of feldspar, quartz, and mica are 60%, 35.3%, and 4.7% respectively;

[0056] Step 3: Establish a hob rock-breaking model. Use the GBM model to randomly generate several divided regions for filling particles according to the particle size of rock minerals, and fill the divided regions with spherical particles. Group and name each divided region according to the content ratio of rock sample mineral components. Then, according to the grouping results, assign different mineral components and corresponding mesoscopic parameters to each component. After the particles are generated, apply confining pressure to them. Use a wall to simulate the cross-sectional shape of the hob. Conduct multiple intrusion simulation tests on rock specimens with different mineral contents, and change the rock mineral content, hob cutter spacing, and penetration depth, and conduct intrusion simulation tests in sequence to achieve the destruction within the mineral crystal, such as Figure 5 and Figure 6 as shown;

[0057] Step 4: According to the numerical simulation process, collect the results of each simulation, record the mineral content, hob cutter spacing, penetration degree, and rock-breaking efficiency of the rock sample, and construct a data set for subsequent analysis and modeling. Among them, the mineral content includes the proportion of the content of corresponding feldspar, quartz, mica, etc.;

[0058] As a further optimization, use a normalization function to normalize the data set, normalize the data to the range of 0-1 to improve the training effect, and then save the normalized parameters for subsequent denormalization processing;

[0059] Step 5: Establish a prediction model based on the PSO-BPNN algorithm. PSO-BPNN is a hybrid algorithm that combines the particle swarm optimization algorithm (PSO, Particle Swarm Optimization) and the backpropagation neural network (BPNN, Backpropagation Neural Network). PSO is an optimization algorithm based on swarm intelligence, which imitates the foraging behavior of bird flocks. Particles find the global optimal solution through mutual cooperation and information sharing. BPNN is a commonly used artificial neural network. It adjusts parameters through the error backpropagation algorithm to optimize the weights and biases of the network, thereby minimizing the output error. Combining PSO and BPNN results in the PSO-BPNN algorithm, which can effectively improve the prediction accuracy of the model. In PSO-BPNN, the PSO algorithm is used to optimize the initial weights and biases of BPNN, avoiding the local optimal problem that may be caused by random initialization. Then, through BPNN for training and prediction, it can ensure that the obtained model has excellent prediction accuracy.

[0060] Regarding the optimization problem of the hob rock-breaking process, the number of neurons in the input layer corresponds to variables such as the rock mineral content, hob spacing, and penetration depth. There is one neuron in the output layer, and the output is the optimization objective of rock-breaking efficiency. In a three-layer neural network, the number of neurons in the hidden layer is initially determined based on an empirical formula, and the performance of models corresponding to different numbers of hidden-layer neurons is judged by the mean square error (MSE). Based on this, by cycling through different numbers of hidden-layer nodes and comparing the training error (MSE, mean square error), the optimal number of hidden-layer nodes is determined. The optimal number of hidden-layer neurons is a specific value (which can be determined by tuning). Finally, the final BP neural network is constructed according to the optimal number of hidden-layer nodes.

[0061] The specific formula for the number of neurons in the hidden layer is: where hiddennum is the number of neurons in the hidden layer, n is the number of neurons in the input layer, preferably taking the value of 1, m is the number of neurons in the output layer, preferably taking the value of 5, and a is a constant with a value range between 1 and 10.

[0062] The constructed dataset is divided into a training set and a test set according to a set ratio. The role of the training set is to provide data for the model to learn, so as to obtain the required machine learning prediction model, that is, the rock-breaking efficiency prediction model. The test set is used to judge the accuracy of the generated prediction model.

[0063] The training set is used to train the backpropagation neural network (BPNN, Backpropagation Neural Network) based on particle swarm optimization (PSO, Particle Swarm Optimization). At the same time, the PSO algorithm is used to optimize the weights and biases of the backpropagation neural network, enabling it to more effectively learn the complex relationship between the input and output during the training process; by adjusting the velocity and position of the particles, the PSO algorithm continuously searches for the optimal fitness solution, thereby improving the performance of the model and obtaining the rock-breaking efficiency prediction model; the test data is input into the rock-breaking efficiency prediction model to obtain the corresponding rock-breaking efficiency prediction results, and the set evaluation indicators are used to evaluate the prediction accuracy and reliability of the model.

[0064] Step 6: Prediction of rock-breaking efficiency; during the rock-breaking operation, the rock surface image is collected in real time. After preprocessing the rock surface image, it is input into the rock-breaking efficiency prediction model, and the prediction model is used to realize the prediction of the hob rock-breaking efficiency and output the prediction results.

[0065] As an optimization, in Step 5, the dataset is randomly divided into a training set and a test set according to a ratio of 7:3.

[0066] As an optimization, in Step 1, the Image J digital image processing and analysis software is used to perform grayscale processing on the rock surface image.

[0067] The particle contact model within and between minerals is given as PBM (Parallel - bonded model), and the meso - mechanical properties of different mineral types are classified. As an optimization, in Step 2, the "trial - and - error calibration method" is adopted, and continuous adjustment and trial calculations are carried out for the calibration of meso - parameter values.

[0068] As an optimization, in Step 3, the mineral content, hob spacing, and penetration degree of the rock during each simulation process are collected, and the specific energy SE is calculated according to formula (1) using the work done by the hob and the rock - breaking area; the definition of specific energy is the power required to cut a unit volume of rock, which can be used to evaluate the rock - breaking efficiency of the tool. The smaller the specific energy of the hob for breaking a unit volume of rock, the higher its rock - breaking efficiency.

[0069]

[0070] In the formula, W is the work done by the tool during rock - breaking, that is, the external force work; V is the volume of the rock generated by the rock - breaking test; Wv is the work done by the penetration force of the hob; W r is the work done by the rolling force of the hob. When only considering the intrusion effect of the cutter head, W r = 0; S is the rock - breaking area;

[0071] Forward propagation is the way adopted by the neural network during the calculation of the output. It gradually transforms the input data through each layer of the network into the final output, and through the introduction of activation functions, the neural network can model complex non - linear relationships. As an optimization, in Step 5, the process of obtaining the corresponding rock - breaking efficiency prediction result is as follows:

[0072] A1: Obtain the output h of the hidden layer according to formula (2);

[0073] h = f(W ih ·X + b h ) (2);

[0074] In the formula, W ih is the weight matrix from the input layer to the hidden layer, X is the input vector, b h is the bias of the hidden layer, and f(·) is the activation function tansing, whose output range is [-1, 1]; tansig helps the model learn non - linear relationships and is suitable for the hidden layer.

[0075] A2: Obtain the prediction result y of the output layer according to formula (3);

[0076] y = g(W ho ·h + bo ) (3);

[0077] Wherein, W ho is the weight matrix from the hidden layer to the output layer, b o is the bias of the output layer, and g(·) is the activation function purelin of the output layer, whose output range is (-∞, +∞). Purelin directly outputs the predicted value and is suitable for the output layer of regression tasks. The activation function not only introduces non-linearity to the neural network but also affects the training and performance of the network. Reasonable selection and application of the activation function are important parts of constructing an efficient neural network.

[0078] Backpropagation is the process of updating the network weights by calculating the gradient of the loss function. The process of calculating the gradient through the mean square error is one of the core steps of the backpropagation algorithm. Specifically, in the neural network, first, the predicted output of the network is obtained through forward propagation and compared with the actual target output. The mean square error, as the loss function, can be used to quantify the difference between the predicted result and the actual result. As a preference, in step five, the process of optimizing the weights of the backpropagation neural network using the PSO algorithm is as follows:

[0079] B1: Calculate the mean square error L according to formula (4);

[0080]

[0081] Wherein, N is the number of test samples, y i is the actual output, is the predicted output;

[0082] B2: By taking the derivative of the mean square error, the gradient of the loss function with respect to each weight can be obtained. The gradient represents the sensitivity of the loss function to the change of the weight, that is, under the current weight setting, if the weight increases or decreases, what impact will it have on the loss function. Take the derivative of formula (4) and obtain the gradient of the loss with respect to the weight according to formula (5)

[0083]

[0084] Wherein, is the gradient of the loss with respect to the output, is the gradient of the output with respect to the weight;

[0085] B3: After obtaining the gradient, the weights can be updated according to the gradient descent method. Specifically, obtain the updated weight W according to formula (6). The weight W is the parameter connecting neurons and determines the strength of signal transmission;

[0086]

[0087] Where η is the learning rate, which controls the step size of weight update. Through this process, the model gradually adjusts the weights in each iteration to reduce the prediction error, thereby effectively improving the accuracy and generalization ability of the model. By repeatedly executing this process, the neural network can continuously optimize its parameters and finally achieve better performance.

[0088] Use the current velocity and historical optimal position of the particle; As an optimization, in step five, the process of adjusting the velocity and position of the particle is as follows:

[0089] C1: Update the velocity of the particle according to formula (7)

[0090]

[0091] In the formula, is the velocity of particle i at time t, w is the inertia weight, which controls the motion inertia of the particle, c1 is the individual learning factor, which controls the degree of the particle moving towards its best position, c2 is the swarm learning factor, which controls the degree of the particle moving towards the swarm best position, r1 and r2 are random numbers in the range of [0, 1], p i is the best position of particle i, is the position of particle i at time t, and g is the best position of the swarm;

[0092] C2: Update the position of the particle according to formula (8)

[0093]

[0094] The present invention continuously seeks the globally optimal particle position and records the convergence curve by cyclically updating the velocity and position of the particle. Calculate the error of the PSO-BP neural network, perform simulation prediction on the test set, and anti-normalize the normalized data to the actual value. The fitting effect is as Figure 7 shown, including regression analysis diagrams of training, validation, testing, and overall fitting effects. The horizontal axis is "target value" in all cases, representing the true output of the data, and the vertical axis is "output" in all cases, representing the output value predicted by the model. The solid line represents the regression line of the prediction result and the true result, and the dashed line represents the ideal Y = T line, that is, the case where the prediction result is exactly the same as the true result. The correlation coefficient R in the figure represents the fitting degree on the training set, and being close to 1 indicates that the model fits well on the training set.

[0095] In the present invention, by processing the acquired rock surface image based on image processing technology, particles of different mineral types can be grouped through preliminary analysis, so as to accurately obtain a number of structure diagrams, which is conducive to constructing an accurate rock sample model subsequently; by using numerical simulation to model the rock material and calibrate the mesoscopic parameter values, the constructed rock sample model can accurately reflect the composition and microscopic structure characteristics of the rock; by using a cutter rock breaking model to conduct cutter penetration tests based on the calibrated mesoscopic parameters and changing the rock mineral content, cutter spacing, and penetration depth for combined tests, a large amount of numerical simulation data can be accurately obtained, which is conducive to constructing a high-precision rock breaking efficiency prediction model based on this. By using machine learning technology to learn and train the obtained numerical simulation data, the complex relationship between parameters such as rock composition, cutter spacing, and penetration degree and rock breaking efficiency can be effectively captured, so as to achieve more accurate rock breaking efficiency prediction and provide a scientific basis for optimizing construction parameters such as penetration degree.

[0096] The present invention provides a comprehensive and efficient rock breaking efficiency prediction method based on image recognition and machine learning, which effectively solves the problem that it is difficult to accurately design construction parameters for different rock layer types in the existing rock breaking technology. By analyzing the influence of rock mineral composition on cutter rock breaking, it can combine the rock microscopic composition with the macroscopic rock breaking behavior to efficiently and accurately predict the rock breaking efficiency. This method can not only more accurately predict the rock breaking efficiency, but also further optimize the cutter spacing and penetration degree according to the prediction results to achieve higher construction rock breaking efficiency.

Claims

1. A method for predicting the rock-breaking efficiency based on image recognition and machine learning, characterized in that, It includes the following steps: Step 1: Obtain the rock surface image, apply image processing technology to grayscale the obtained rock surface image, and then perform pixel classification and image segmentation to clearly divide different minerals into several independent structure diagrams; Step 2: Based on the Voronoi Grain-Based Model, model the rock material, import the boundary information of several structure diagrams into the model, generate several large-particle filled rock blocks with the same size as the several structure diagrams, connect the centroids of the contact polygons to form mineral crystal boundaries, and at the same time, fill smaller particles in the polygon area; in the numerical simulation software, group the large-particle filled rock blocks according to the defined mineral layers, divide and adjust the meso-mechanical properties of different mineral types, and calibrate a set of most suitable meso-parameters to accurately reflect the composition and microstructure characteristics of the rock; Step 3: Establish a cutter rock-breaking model, use the GBM model to randomly generate several division regions for filling particles according to the particle size of the rock minerals, fill the division regions with spherical particles, group and name each division region according to the proportion of the mineral composition content of the rock sample, and then, according to the grouping results, assign different mineral compositions and corresponding meso-parameters to different components; apply confining pressure to the particles after generation, use the wall to simulate the cross-sectional shape of the cutter, conduct multiple intrusion simulation tests on rock samples with different mineral contents, and change the rock mineral content, cutter spacing, and penetration depth, and conduct intrusion simulation tests in combination in turn to achieve the failure within the mineral crystal; Step 4: According to the numerical simulation process, collect the results of each simulation, record the mineral content, cutter spacing, penetration degree, and rock-breaking efficiency of the rock sample, and construct a data set for subsequent analysis and modeling; Step 5: Establish a prediction model based on the PSO-BPNN algorithm; divide the constructed data set into a training set and a test set according to a set ratio, use the training set to train the backpropagation neural network based on particle swarm optimization, and at the same time, use the PSO algorithm to optimize the weights and biases of the backpropagation neural network to enable it to more effectively learn the complex relationship between the input and output during the training process; by adjusting the velocity and position of the particles, make the PSO algorithm continuously search for the optimal fitness solution, thereby improving the performance of the model and obtaining a rock-breaking efficiency prediction model; input the test data into the rock-breaking efficiency prediction model to obtain the corresponding rock-breaking efficiency prediction result, and use the set evaluation index to evaluate the prediction accuracy and reliability of the model; Step 6: Prediction of rock-breaking efficiency; During the rock-breaking operation process, collect the rock surface image in real time, after preprocessing the rock surface image, input it into the rock-breaking efficiency prediction model, use the rock-breaking efficiency prediction model to achieve the prediction of the rock-breaking efficiency, and output the prediction result.

2. The method for predicting the rock-breaking efficiency based on image recognition and machine learning according to claim 1, wherein, In Step 5, randomly divide the data set into a training set and a test set according to a ratio of 7:

3.

3. A rock-breaking efficiency prediction method based on image recognition and machine learning according to claim 1, characterized in that, In Step 1, use the Image J digital image processing and analysis software to grayscale the rock surface image.

4. A rock-breaking efficiency prediction method based on image recognition and machine learning according to claim 1, characterized in that, In step two, the "trial and error calibration method" is adopted, and the mesoscopic parameter values are calibrated through continuous adjustment and trial calculation.

5. A rock-breaking efficiency prediction method based on image recognition and machine learning according to claim 1, characterized in that, In step three, the mineral content, cutter spacing, and penetration of the rock in each simulation process are collected, and the specific energy SE is calculated using the cutter work and the rock-breaking area according to formula (1). In the formula, W is the work done by the tool during rock breaking, i.e., the external force work; V is the volume of the rock generated by the rock breaking test; Wv is the work done by the cutter penetration force; W r is the work done by the cutter rolling force. When only considering the intrusion effect of the cutter head, W r = 0; S is the rock breaking area.

6. The rock-breaking efficiency prediction method based on image recognition and machine learning according to claim 1, wherein In step five, the process of obtaining the corresponding rock-breaking efficiency prediction results is as follows: A1: Obtain the output h of the hidden layer according to formula (2). h = f(W ih ·X + b h ) (2); where, W ih is the weight matrix from the input layer to the hidden layer, X is the input vector, and b h is the bias of the hidden layer, f(·) is the activation function tansing, and its output range is [-1, 1]; A2: Obtain the prediction result y of the output layer according to formula (3). y = g(W ho ·h + b o ) (3); where W ho is the weight matrix from the hidden layer to the output layer, and b o is the bias of the output layer. g(·) is the activation function purelin of the output layer, and its output range is (-∞, +∞).

7. A method for predicting the rock-breaking efficiency based on image recognition and machine learning according to claim 1, characterized in that In step five, the process of optimizing the weights of the backpropagation neural network using the PSO algorithm is as follows: B1: Calculate the mean square error L according to formula (4). where N is the number of test samples, and y i is the actual output, and is the predicted output; B2: Differentiate formula (4) and obtain the gradient of the loss with respect to the weights according to formula (5). wherein, is the gradient of the loss with respect to the output, is the gradient of the output with respect to the weights; B3: Obtain the updated weight W according to formula (6). In the formula, η is the learning rate, which controls the step size of weight update.

8. A method for predicting the rock-breaking efficiency based on image recognition and machine learning according to claim 1, characterized in that, In step five, the process of adjusting the velocity and position of the particles is as follows: C1: Update the velocity of the particle according to formula (7) Wherein, is the velocity of particle i at time t, w is the inertia weight to control the motion inertia of the particle, c1 is the individual learning factor to control the degree of the particle moving towards its best position, c2 is the swarm learning factor to control the degree of the particle moving towards the swarm best position, r1 and r2 are random numbers within the range of [0, 1], p i is the best position of particle i, is the position of particle i at time t, and g is the best position of the swarm; C2: Update the position of the particle according to formula (8)

Images