Hydraulic engineering blasting fragmentation prediction method and device based on physical information constraint and extreme learning machine, medium and equipment

Abstract

The invention discloses a hydraulic engineering blasting fragmentation prediction method and device based on physical information constraint and an extreme learning machine, a medium and equipment, and belongs to the technical field of artificial intelligence. The hydraulic engineering blasting lumpiness prediction method based on the physical information constraint and the extreme learning machine comprises the following steps: obtaining blasting characteristic parameters of a blasting scheme to be predicted; the blasting characteristic parameters of the blasting scheme to be predicted are input into a prediction model, and the prediction model is a trained hydraulic engineering blasting fragmentation prediction model based on physical information constraint and an extreme learning machine; and the prediction model carries out operation according to the blasting characteristic parameters of the to-be-predicted blasting scheme, and outputs a blasting fragmentation prediction result of the to-be-predicted blasting scheme. The hydraulic engineering blasting fragmentation prediction device based on the physical information constraint and the extreme learning machine, the medium and the equipment can be used for realizing the method. The blasting lumpiness prediction method can obtain the blasting lumpiness prediction result according to the blasting characteristic parameters of the to-be-predicted blasting scheme, and is simple, convenient and low in cost.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, medium and equipment for predicting the fragmentation of blasting in a water conservancy project based on physical information constraints and an extreme learning machine. Background Art

[0002] Blasting is a crucial process in the excavation of slopes or tunnels in water conservancy projects. Blasting fragment size is a key indicator of blasting effectiveness. The effectiveness of blasting impacts the efficiency and schedule of subsequent loading, transportation, and crushing operations. A good blasting fragment size distribution improves loading, transportation, and crushing efficiency while reducing energy consumption, thereby improving economic returns. However, if the blasting fragment size distribution is poor, with a high percentage of large fragments, secondary crushing may be necessary. Therefore, predicting the blasting fragment size distribution using existing blasting parameters is crucial.

[0003] Existing methods for predicting blasting fragmentation include theoretical models and machine learning. Theoretical models, which use a semi-empirical, semi-theoretical model proposed by previous researchers, predict blasting fragmentation. These methods primarily build machine learning models by learning the inherent connections between data, neglecting the theoretical foundations of blasting. This results in low prediction accuracy and poor adaptability to different rock masses. Summary of the Invention

[0004] In view of this, the present invention provides a method, device, medium and equipment for predicting the blasting fragmentation of water conservancy projects based on physical information constraints and extreme learning machines. It can obtain the blasting fragmentation prediction results according to the blasting characteristic parameters of the blasting scheme to be predicted. It is simple, convenient and low-cost, and thus more suitable for practical use.

[0005] In order to achieve the first objective above, the present invention provides a method for predicting the fragmentation of hydraulic engineering blasting based on physical information constraints and extreme learning machines. The method is as follows:

[0006] The method for predicting the blasting fragmentation of a hydraulic engineering project based on physical information constraints and an extreme learning machine provided by the present invention comprises the following steps:

[0007] Obtaining blasting characteristic parameters of the blasting plan to be predicted;

[0008] Inputting the blasting characteristic parameters of the blasting scheme to be predicted into a prediction model, wherein the prediction model is a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machines;

[0009] The prediction model performs calculations based on the blasting characteristic parameters of the blasting scheme to be predicted, and outputs a blasting fragmentation prediction result of the blasting scheme to be predicted.

[0010] The method for predicting the fragmentation of blasting in water conservancy projects based on physical information constraints and extreme learning machines provided by the present invention can also be further implemented by adopting the following technical measures.

[0011] Preferably, in the step of inputting the blasting characteristic parameters of the blasting scheme to be predicted into a prediction model, wherein the prediction model is a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine, the training method of the prediction model includes the following steps:

[0012] Obtaining historical blasting samples, wherein the historical blasting samples include blasting characteristic parameters and blasting characteristic images of the blasting plans of the historical blasting samples, and blasting fragmentation results after the blasting is completed, wherein the blasting fragmentation results after the blasting is completed include blasting characteristic parameters and characteristic images of the rock accumulation body after the blasting;

[0013] Inputting the blasting characteristic parameters of the blasting plan of the historical blasting sample and the blasting fragmentation result after the blasting is completed into the recognition software for data recognition to obtain a data set;

[0014] Training is performed based on the processed data set to obtain a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machines.

[0015] Preferably, the step of obtaining historical blasting samples, wherein the historical blasting samples include blasting characteristic parameters of the blasting plan of the historical blasting samples, and blasting fragmentation results after the blasting is completed, further includes the following steps:

[0016] Random classification is performed on the historical blasting samples, and a set proportion of historical samples are selected as the training sample set, and the remaining historical samples are used as the test sample set, where:

[0017] The training sample set is used for data training to obtain a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine, wherein the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine includes an input layer, an output layer and a hidden layer;

[0018] The test sample set is used to perform performance testing on a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine, and to adjust relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine.

[0019] Preferably, the blasting characteristic parameters of the blasting scheme include explosive consumption per unit, blasthole diameter, charge length, hole length, and explosive type.

[0020] Preferably, the test sample set is used to perform performance testing on a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine, and the process of adjusting relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine specifically includes the following steps:

[0021] The test sample set is input into the extreme learning machine prediction model with physical information constraints optimized by the particle swarm optimization algorithm to obtain the optimized blasting fragmentation test results.

[0022] Preferably, the relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine include:

[0023] The number of nodes in the input layer, output layer, and hidden layer, the regularization coefficient, and the activation function of the hidden layer;

[0024] The maximum number of iterations of particles, the number of swarms, the inertia weight, the learning factor, and the parameter search range in the particle swarm optimization algorithm, wherein the particle swarm optimization algorithm is used to search for the optimal parameter combination of the extreme learning machine.

[0025] Preferably, training is performed according to the processed data set to obtain a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine, specifically comprising the following steps:

[0026] Step S1: Record the blasting parameters under different blasting schemes. The blasting parameters should include explosive consumption, blasthole diameter, charge length, hole length, and explosive type, and record the degree of rock joint and fissure development. After the blasting test, take photos of the blasted ore. When taking photos, record the size of the blasted ore from different directions and angles.

[0027] Step S2: Input the obtained fragmentation picture of the blasted ore into Wipfrag software for identification, thereby obtaining a fragmentation distribution curve, and thus obtaining the average fragmentation size x, that is, the fragmentation size with an undersize accumulation rate of 50%, and obtaining a data set;

[0028] Step S3: Determine the characteristic variables of the model as explosive consumption per unit, blasthole diameter, and charge length, and the target variable as the average fragmentation size x. Normalize the characteristic variable data so that different characteristic variables have the same scale and eliminate the influence of dimension. The normalization formula is as follows:

[0029]

[0030] Where, X n is the dimensionless value after normalization; X is the original data; X min is the minimum value of the original data; X maxis the maximum value of the original data,

[0031] The obtained data is randomly divided into 80% of which is the training set for training the model and 20% is the test set for testing the performance of the model;

[0032] Step S4: Determine the loss function of the extreme learning machine that incorporates physical information constraints, wherein the loss function of the extreme learning machine that incorporates physical information constraints consists of two parts, one of which is the extreme learning machine model prediction loss F EML , and the other part is the Kuz-Ram theoretical model prediction loss F Theory , assign different weights to these two parts of loss, F EML The weight is η, F Theory The weight is 1-η, the value range of η is [0,1], and the loss function F of the extreme learning machine is as follows:

[0033] F=ηgF EML +(1-η)gF Theory

[0034]

[0035] Among them, i is the i-th training sample, yi is the actual value, is the prediction value of the extreme learning machine model, is the predicted value of the Kuz-Ram theoretical model, n is the number of all training samples, ξ is the regularization coefficient, and w is the weight between the input layer and the hidden layer of the extreme learning machine model;

[0036] The Kuz-Ram theoretical model is as follows:

[0037]

[0038] Where A is the rock coefficient, and its value is related to the degree of development of rock joints and fissures; q is the unit consumption of blasthole explosives, kg / m 3 ; Q is the charge amount per hole, kg; E is the explosive power; d is the borehole diameter, m; L is the actual charge length, m; ρ is the charge density in the hole, kg / m 3 ;

[0039] Step S5: Establish an extreme learning machine prediction model that incorporates physical information constraints, use the training set data for training, and use the particle swarm optimization algorithm to search for the optimal parameter combination of the extreme learning machine during training. The parameters include the weight η in the loss function, the weight matrix w from the input layer to the hidden layer of the extreme learning machine, and the bias b. The specific steps described in step 5 are as follows:

[0040] Step 501: Determine the parameters of the extreme learning machine, including the input layer, output layer, number of hidden layer nodes, regularization coefficient, activation function of the hidden layer, and determine the maximum number of iterations of particles in the particle swarm algorithm, the number of populations, inertia weight, learning factor, and parameter search range;

[0041] Step 502: Randomly initialize a group of particles. These particles represent the solution in the parameter search space. Train the model according to the parameters represented by different particles. Then calculate the loss value by using the prediction results of the training set and the loss function. The loss function is the loss function of the extreme learning machine with physical information constraints as described in step 4, which includes the prediction loss F of the extreme learning machine model. EML The loss F is predicted by the Kuz-Ram theoretical model Theory Then, the particle's velocity and coordinates are iteratively updated according to the velocity update formula and position update formula, and the hyperparameter combination is updated. The velocity update formula and position update formula are as follows:

[0042] ν ij (t+1)=wv ij (t)+c1r1(t)[p ij (t)-x ij (t)]+c2r2(t)[p gj (t)-x ij (t)]

[0043] x ij (t+1)=x ij (t)+ν ij (t+1)

[0044] Among them, v ij is the velocity of the i-th particle in the j-th dimension; x ij is the position of the i-th particle in the j-th dimension; p ij is the local optimal position of the i-th particle; p gj is the global optimal position of the particle swarm; w is the inertia weight; c1 and c2 are learning factors respectively; r1 and r2 are random numbers in the range of [0,1];

[0045] The training process of the extreme learning machine model is as follows:

[0046] The extreme learning machine model is an efficient single-hidden-layer feedforward neural network consisting of three layers: input layer, output layer, and hidden layer. First, the parameters are initialized. The parameters include the hidden layer weights and biases, which are determined by the weights and biases represented by the particles. Then, given a training set, the hidden layer output is calculated. The hidden layer output calculation formula is as follows:

[0047]

[0048] Where xi is the i-th input data, w i is the weight of the i-th input feature, b is the bias term, and F is the activation function;

[0049] Step S503: Determine the output weight β between the hidden layer and the output layer, wherein the principle for determining the output weight β between the hidden layer and the output layer is to minimize the loss function of the model;

[0050] Based on the loss value of each particle after it moves to a new position, its individual historical best position and the group historical best position are dynamically adjusted. When the loss value tends to be stable or the number of iterations reaches the maximum, the iteration loop is terminated, and the optimal parameter combination is obtained and output. The formula for updating the individual historical best position is as follows:

[0051]

[0052] The formula for updating the historical optimal position of the group is as follows:

[0053]

[0054] Where, is the historical optimal position of the i-th particle in the t-th generation; is the current particle position; f(*) is the objective function; G best t+1 is the historical optimal position of the group in generation t+1;

[0055] Step S6: Use the optimal parameters to reconstruct the extreme learning machine model, perform training set reconstruction test, input the test set into the extreme learning machine prediction model with physical information constraints optimized by the particle swarm optimization algorithm, and obtain the average block size Predicted value;

[0056] Step S7: Calculate the evaluation index of the model by using the actual value and the predicted value, wherein the model evaluation index includes accuracy (AR), mean square error (MSE), root mean square error (RMSE), determination coefficient (R 2 ), the specific calculation formula is as follows:

[0057]

[0058] Where i is the i-th test sample, y i is the actual value, is the predicted value, is the mean of the predicted values, and n is the number of all test samples.

[0059] In order to achieve the second objective, the present invention provides a device for predicting the fragmentation of blasting debris in water conservancy projects based on physical information constraints and extreme learning machines. The device is as follows:

[0060] The device for predicting the fragmentation of blasting in water conservancy projects based on physical information constraints and extreme learning machines provided by the present invention includes:

[0061] A characteristic parameter acquisition module is used to obtain blasting characteristic parameters of the blasting scheme to be predicted;

[0062] A prediction model, used for inputting blasting characteristic parameters of the blasting scheme to be predicted into the prediction model, wherein the prediction model is a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine;

[0063] The prediction result output module is used for outputting the prediction model, which performs data calculation based on the blasting characteristic parameters of the blasting scheme to be predicted, and outputs the blasting fragmentation prediction result of the blasting scheme to be predicted through the prediction result output module.

[0064] In order to achieve the third objective above, the technical solution of the computer-readable storage medium provided by the present invention is as follows:

[0065] The computer-readable storage medium provided by the present invention stores a water conservancy project blasting fragmentation prediction program based on physical information constraints and extreme learning machines. When the water conservancy project blasting fragmentation prediction program based on physical information constraints and extreme learning machines is executed by a processor, the steps of the water conservancy project blasting fragmentation prediction method based on physical information constraints and extreme learning machines provided by the present invention are implemented.

[0066] In order to achieve the fourth objective, the present invention provides an electronic device with the following technical solutions:

[0067] The electronic device provided by the present invention includes a memory and a processor, wherein the memory stores a water conservancy project blasting fragmentation prediction program based on physical information constraints and an extreme learning machine. When the water conservancy project blasting fragmentation prediction program based on physical information constraints and an extreme learning machine is executed by the processor, the steps of the water conservancy project blasting fragmentation prediction method based on physical information constraints and an extreme learning machine provided by the present invention are implemented.

[0068] The method, device, medium, and equipment for predicting the blasting fragmentation of a water conservancy project based on physical information constraints and an extreme learning machine provided by the embodiments of the present invention combine the Kuz-Ram theoretical model as physical prior knowledge with an extreme learning machine. By incorporating the Kuz-Ram theoretical model into the loss function of the extreme learning machine to constrain the training process of the neural network, a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is obtained. This model not only conforms to existing theoretical models but also mines inherent connections from data. On this basis, only the blasting parameters of the blasting scheme to be predicted need to be obtained, and the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine can be used to obtain the blasting fragmentation prediction result of the blasting scheme to be predicted. In particular, because the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is trained based on historical blasting samples, it can make the blasting fragmentation prediction result of the blasting scheme to be predicted simple, convenient, and low-cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0070] Figure 1 A flowchart of the overall steps of a method for predicting blasting fragmentation in water conservancy projects based on physical information constraints and extreme learning machines, provided in an embodiment of the present invention;

[0071] Figure 2 A schematic diagram of the data set construction process involved in the method for predicting the fragmentation of blasting in water conservancy projects based on physical information constraints and extreme learning machines provided in an embodiment of the present invention;

[0072] Figure 3 A flowchart of the operation of a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machines, which is involved in the water conservancy project blasting fragmentation prediction method based on physical information constraints and extreme learning machines provided in an embodiment of the present invention;

[0073] Figure 4 A schematic diagram of a blasting fragmentation prediction result of a blasting scheme to be predicted obtained by using a method for predicting blasting fragmentation of a water conservancy project based on physical information constraints and an extreme learning machine provided by an embodiment of the present invention;

[0074] Figure 5 A schematic diagram of the functional modules used in the device for predicting blasting fragmentation in water conservancy projects based on physical information constraints and extreme learning machines, and their association with blasting characteristic parameters and prediction results, provided by an embodiment of the present invention;

[0075] Figure 6 A schematic diagram of the structure of a device for predicting the fragmentation of blasting in water conservancy projects based on physical information constraints and extreme learning machines in the hardware operating environment provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0076] In order to solve the problems existing in the prior art, the present invention provides a method, device, medium and equipment for predicting the blasting fragmentation of water conservancy projects based on physical information constraints and extreme learning machines. The method can obtain the blasting fragmentation prediction result according to the blasting characteristic parameters of the blasting scheme to be predicted. The method is simple, convenient and low-cost, and thus more suitable for practical use.

[0077] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the method, apparatus, medium, and equipment for predicting blasting fragmentation in hydraulic engineering projects based on physical information constraints and an extreme learning machine, along with its specific implementation, structure, features, and effectiveness. In the following description, different references to "one embodiment" or "embodiment" do not necessarily refer to the same embodiment. Furthermore, features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0078] The term "and / or" in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B, specifically understood as: A and B may be included at the same time, A may exist alone, or B may exist alone, and any of the above three situations may exist.

[0079] A method for predicting the fragmentation of hydraulic engineering blasting based on physical information constraints and extreme learning machine

[0080] See attached Figure 1 -Attached Figure 4 The method for predicting the blasting fragmentation of a water conservancy project based on physical information constraints and an extreme learning machine provided by an embodiment of the present invention includes the following steps:

[0081] Obtain blasting characteristic parameters of the blasting plan to be predicted.

[0082] Specifically, the feature parameter acquisition module is used to acquire and store raw data, including physical parameters and image data, providing a data foundation for subsequent processing. A database or distributed storage system can be used to ensure data traceability and security.

[0083] The blasting characteristic parameters of the blasting scheme to be predicted are input into the prediction model, wherein the prediction model is a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine.

[0084] Specifically, the prediction model is the core module of the device for predicting the fragmentation of water conservancy project blasting based on physical information constraints and extreme learning machines provided by an embodiment of the present invention. The prediction model is an extreme learning machine model that integrates physical information constraints and is optimized by particle swarm optimization. First, the model structure parameters (number of nodes in the input layer / hidden layer / output layer, regularization coefficient, activation function) and particle swarm algorithm parameters (number of iterations, population size, inertia weight, learning factor, and search range) are determined, and the parameter combination space is characterized by randomly initializing the particle swarm. Each particle corresponds to the weight η of the ELM model, the input layer weight matrix w, and the bias b parameter. During the training process, a dual loss function (including ELM prediction loss and Kuz-Ram theoretical model loss) that integrates physical information constraints is used to evaluate performance. The particle swarm algorithm dynamically updates particle velocity and position, synchronously calculates hidden layer output, and optimizes output weight β to minimize the loss function. During the iterative process, the individual historical optimal position and the group global optimal position are continuously updated until the loss converges or the maximum number of iterations is reached, and the optimal parameter combination is obtained. The final reconstructed optimization model achieves accurate prediction of the average fragmentation size through the test set, combining the advantages of data-driven characteristics and physical mechanism constraints. In this example, WipFrag, a rock blasting fragmentation identification software used in the prediction model, is based on image analysis. It processes in-situ images of post-blasting debris piles and combines digital image processing techniques with statistical methods to quantify the fragmentation distribution of the crushed rock. This software collects high-definition images of the post-blasting debris pile and places calibration objects for geometric calibration. It then uses image preprocessing, edge detection, and deep learning segmentation algorithms to extract the fragmented rock contours. Using the calibration ratio, it calculates the equivalent fragmentation and generates a fragmentation distribution curve, ultimately outputting quantitative fragmentation data. This technology efficiently captures rock fragmentation characteristics in a non-contact manner.

[0085] The prediction model calculates the blasting characteristic parameters of the blasting scheme to be predicted and outputs the blasting fragmentation prediction result of the blasting scheme to be predicted.

[0086] Specifically, the prediction result output module of the water conservancy project blasting fragmentation prediction device based on physical information constraints and extreme learning machines provided in an embodiment of the present invention intuitively displays the prediction results and measured results in the form of charts, heat maps, dynamic simulations, etc., which is convenient for analyzing error distribution, model performance and consistency of physical laws.

[0087] In this embodiment, the blasting characteristic parameters of each blasting scheme to be predicted, as well as the corresponding actual implementation results of the blasting block size can also be added to the training sample set and / or the test sample set. Therefore, the water conservancy project blasting block size prediction model based on physical information constraints and extreme learning machines provided by the embodiment of the present invention can have iteratively optimized application performance. Therefore, as the blasting blocks of the predicted blasting scheme increase, the water conservancy project blasting block size prediction model based on physical information constraints and extreme learning machines provided by the embodiment of the present invention tends to be more accurate.

[0088] The method for predicting the blasting fragmentation of a water conservancy project based on physical information constraints and an extreme learning machine, provided by an embodiment of the present invention, combines the Kuz-Ram theoretical model as physical prior knowledge with an extreme learning machine. By incorporating the Kuz-Ram theoretical model into the loss function of the extreme learning machine to constrain the training process of the neural network, a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is obtained. This model not only conforms to existing theoretical models but also mines inherent connections from the data. On this basis, only the blasting parameters of the blasting scheme to be predicted need to be obtained, and the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine can be used to obtain the blasting fragmentation prediction results of the blasting scheme to be predicted. In particular, because the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is trained based on historical blasting samples, it can make the blasting fragmentation prediction results of the blasting scheme to be predicted simple, convenient, and low-cost.

[0089] In the step of inputting the blasting characteristic parameters of the blasting scheme to be predicted into the prediction model, wherein the prediction model is a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine, the training method of the prediction model includes the following steps:

[0090] Obtaining historical blasting samples, wherein the historical blasting samples include blasting characteristic parameters and blasting characteristic images of the blasting plans of the historical blasting samples, and blasting fragmentation results after the blasting is completed, wherein the blasting fragmentation results after the blasting is completed include the blasting characteristic parameters and characteristic images of the rock accumulation body after the blasting;

[0091] Input the blasting characteristic parameters of the blasting plan of the historical blasting sample and the blasting fragmentation results after the blasting is completed into the recognition software for data recognition to obtain a data set;

[0092] According to the processed data set, training is performed to obtain a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine.

[0093] The process of obtaining historical blasting samples, wherein the historical blasting samples include blasting characteristic parameters of the blasting plan of the historical blasting samples, and the blasting fragmentation results after the blasting is completed, further includes the following steps:

[0094] For historical blasting samples, random classification is performed, and a set proportion of historical samples are selected as training sample sets, and the remaining historical samples are used as test sample sets, where:

[0095] The training sample set is used for data training to obtain a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine, wherein the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine includes an input layer, an output layer and a hidden layer;

[0096] The test sample set is used to perform performance testing on the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine, and to adjust relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine.

[0097] In this case, the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine obtained through training with the training sample set is further tested with the test sample set, which can make the relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine more accurate. Therefore, the blasting fragmentation prediction results of the predicted blasting scheme are more accurate.

[0098] Among them, the blasting characteristic parameters of the blasting plan include explosive consumption per unit, blasthole diameter, charging length, hole length, and explosive type.

[0099] The test sample set is used to perform performance testing on the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine, and the process of adjusting relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine specifically includes the following steps:

[0100] The test sample set is input into the extreme learning machine prediction model with physical information constraints optimized by the particle swarm optimization algorithm to obtain the optimized blasting fragmentation test results.

[0101] Among them, the relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine include:

[0102] The number of nodes in the input layer, output layer, and hidden layer, the regularization coefficient, and the activation function of the hidden layer;

[0103] The maximum number of iterations of particles, the size of the population, the inertia weight, the learning factor, and the parameter search range in the particle swarm optimization algorithm, where the particle swarm optimization algorithm is used to search for the optimal parameter combination of the extreme learning machine.

[0104] The method of training the processed data set to obtain a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine specifically includes the following steps:

[0105] Step S1: Record the blasting parameters under different blasting schemes. The blasting parameters should include explosive consumption, blasthole diameter, charge length, hole length, and explosive type, and record the degree of rock joint and fissure development. After the blasting test, take photos of the blasted ore. When taking photos, record the size of the blasted ore from different directions and angles.

[0106] Step S2: Input the obtained fragmentation picture of the blasted ore into Wipfrag software for identification, thereby obtaining a fragmentation distribution curve, and thus obtaining the average fragmentation size x, that is, the fragmentation size with an undersize accumulation rate of 50%, and obtaining a data set;

[0107] Step S3: Determine the characteristic variables of the model as explosive consumption per unit, blasthole diameter, and charge length, and the target variable as the average fragmentation size x. Normalize the characteristic variable data so that different characteristic variables have the same scale and eliminate the influence of dimension. The normalization formula is as follows:

[0108]

[0109] Where, X n is the dimensionless value after normalization; X is the original data; X min is the minimum value of the original data; X max is the maximum value of the original data,

[0110] The obtained data is randomly divided into 80% of which is the training set for training the model and 20% is the test set for testing the performance of the model;

[0111] Step S4: Determine the loss function of the extreme learning machine that incorporates physical information constraints, wherein the loss function of the extreme learning machine that incorporates physical information constraints consists of two parts, one of which is the extreme learning machine model prediction loss F EML , and the other part is the Kuz-Ram theoretical model prediction loss F Theory , assign different weights to these two parts of loss, F EML The weight is η, F Theory The weight is 1-η, the value range of η is [0,1], and the loss function F of the extreme learning machine is as follows:

[0112] F=ηgF EML +(1-η)gF Theory

[0113]

[0114] Among them, i is the i-th training sample, yi is the actual value, is the prediction value of the extreme learning machine model, is the predicted value of the Kuz-Ram theoretical model, n is the number of all training samples, ξ is the regularization coefficient, and w is the weight between the input layer and the hidden layer of the extreme learning machine model;

[0115] The Kuz-Ram theoretical model is as follows:

[0116]

[0117] Where A is the rock coefficient, and its value is related to the degree of development of rock joints and fissures; q is the unit consumption of blasthole explosives, kg / m 3 ; Q is the charge amount per hole, kg; E is the explosive power; d is the borehole diameter, m; L is the actual charge length, m; ρ is the charge density in the hole, kg / m 3 ;

[0118] Step S5: Establish an extreme learning machine prediction model that incorporates physical information constraints, use the training set data for training, and use the particle swarm optimization algorithm to search for the optimal parameter combination of the extreme learning machine during training. The parameters include the weight η in the loss function, the weight matrix w from the input layer to the hidden layer of the extreme learning machine, and the bias b. The specific steps in step 5 are as follows:

[0119] Step 501: Determine the parameters of the extreme learning machine, including the input layer, output layer, number of hidden layer nodes, regularization coefficient, activation function of the hidden layer, and determine the maximum number of iterations of particles in the particle swarm algorithm, the number of populations, inertia weight, learning factor, and parameter search range;

[0120] Step 502: Randomly initialize a group of particles. These particles represent the solution in the parameter search space. Train the model according to the parameters represented by different particles. Then calculate the loss value by using the prediction results of the training set and the loss function. The loss function is the loss function of the extreme learning machine with physical information constraints in step 4, which includes the prediction loss F of the extreme learning machine model. EML The loss F is predicted by the Kuz-Ram theoretical model Theory Then, the particle's velocity and coordinates are iteratively updated according to the velocity update formula and position update formula, and the hyperparameter combination is updated. The velocity update formula and position update formula are as follows:

[0121] ν ij (t+1)=wvij (t)+c1r1(t)[p ij (t)-x ij (t)]+c2r2(t)[p gj (t)-x ij (t)]

[0122] x ij (t+1)=x ij (t)+ν ij (t+1)

[0123] Among them, v ij is the velocity of the i-th particle in the j-th dimension; x ij is the position of the i-th particle in the j-th dimension; p ij is the local optimal position of the i-th particle; p gj is the global optimal position of the particle swarm; w is the inertia weight; c1 and c2 are learning factors respectively; r1 and r2 are random numbers in the range of [0,1];

[0124] The training process of the extreme learning machine model is as follows:

[0125] The extreme learning machine model is an efficient single-hidden-layer feedforward neural network consisting of three layers: input layer, output layer, and hidden layer. First, the parameters are initialized. The parameters include the hidden layer weights and biases, which are determined by the weights and biases represented by the particles. Then, given a training set, the hidden layer output is calculated. The hidden layer output calculation formula is as follows:

[0126]

[0127] Where x i is the i-th input data, w i is the weight of the i-th input feature, b is the bias term, and F is the activation function;

[0128] Step S503: Determine the output weight β between the hidden layer and the output layer, wherein the principle for determining the output weight β between the hidden layer and the output layer is to minimize the loss function of the model;

[0129] Based on the loss value of each particle after it moves to a new position, its individual historical best position and the group historical best position are dynamically adjusted. When the loss value tends to be stable or the number of iterations reaches the maximum, the iteration loop is terminated, and the optimal parameter combination is obtained and output. The formula for updating the individual historical best position is as follows:

[0130]

[0131] The formula for updating the historical optimal position of the group is as follows:

[0132]

[0133] Where, is the historical optimal position of the i-th particle in the t-th generation; is the current particle position; f(*) is the objective function; G best t+1 is the historical optimal position of the group in generation t+1;

[0134] Step S6: reconstructing the extreme learning machine model using the optimal parameters, performing a training set reconstruction test, and inputting the test set into the extreme learning machine prediction model with physical information constraints optimized by the particle swarm optimization algorithm to obtain the average block size x predicted value;

[0135] Step S7: Calculate the evaluation index of the model by using the actual value and the predicted value, wherein the model evaluation index includes accuracy (AR), mean square error (MSE), root mean square error (RMSE), determination coefficient (R 2 ), the specific calculation formula is as follows:

[0136]

[0137] Where i is the i-th test sample, y i is the actual value, is the predicted value, is the mean of the predicted values, and n is the number of all test samples.

[0138] A device for predicting the fragmentation of hydraulic engineering blasting based on physical information constraints and extreme learning machine

[0139] See attached Figure 5 The embodiment of the present invention provides a device for predicting the blasting fragmentation of a water conservancy project based on physical information constraints and an extreme learning machine, including:

[0140] The characteristic parameter acquisition module is used to obtain the blasting characteristic parameters of the blasting plan to be predicted.

[0141] Specifically, the feature parameter acquisition module is used to acquire and store raw data, including physical parameters and image data, providing a data foundation for subsequent processing. A database or distributed storage system can be used to ensure data traceability and security.

[0142] The prediction model is used to input the blasting characteristic parameters of the blasting scheme to be predicted into the prediction model, wherein the prediction model is a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine.

[0143] Specifically, the prediction model is the core module of the device for predicting the fragmentation of water conservancy project blasting based on physical information constraints and extreme learning machines provided by an embodiment of the present invention. The prediction model is an extreme learning machine model that integrates physical information constraints and is optimized by particle swarm optimization. First, the model structure parameters (number of nodes in the input layer / hidden layer / output layer, regularization coefficient, activation function) and particle swarm algorithm parameters (number of iterations, population size, inertia weight, learning factor, and search range) are determined, and the parameter combination space is characterized by randomly initializing the particle swarm. Each particle corresponds to the weight η of the ELM model, the input layer weight matrix w, and the bias b parameter. During the training process, a dual loss function (including ELM prediction loss and Kuz-Ram theoretical model loss) that integrates physical information constraints is used to evaluate performance. The particle swarm algorithm dynamically updates particle velocity and position, synchronously calculates hidden layer output, and optimizes output weight β to minimize the loss function. During the iterative process, the individual historical optimal position and the group global optimal position are continuously updated until the loss converges or the maximum number of iterations is reached, and the optimal parameter combination is obtained. The final reconstructed optimization model achieves accurate prediction of the average fragmentation size through the test set, combining the advantages of data-driven characteristics and physical mechanism constraints. In this example, WipFrag, a rock blasting fragmentation identification software used in the prediction model, is based on image analysis. It processes in-situ images of post-blasting debris piles and combines digital image processing techniques with statistical methods to quantify the fragmentation distribution of the crushed rock. This software collects high-definition images of the post-blasting debris pile and places calibration objects for geometric calibration. It then uses image preprocessing, edge detection, and deep learning segmentation algorithms to extract the fragmented rock contours. Using the calibration ratio, it calculates the equivalent fragmentation and generates a fragmentation distribution curve, ultimately outputting quantitative fragmentation data. This technology efficiently captures rock fragmentation characteristics in a non-contact manner.

[0144] The prediction result output module: the prediction model performs data calculation based on the blasting characteristic parameters of the blasting scheme to be predicted, and outputs the blasting fragmentation prediction result of the blasting scheme to be predicted through the prediction result output module.

[0145] Specifically, the prediction result output module of the water conservancy project blasting fragmentation prediction device based on physical information constraints and extreme learning machines provided in an embodiment of the present invention intuitively displays the prediction results and measured results in the form of charts, heat maps, dynamic simulations, etc., which is convenient for analyzing error distribution, model performance and consistency of physical laws.

[0146] The device for predicting the blasting fragmentation of a water conservancy project based on physical information constraints and an extreme learning machine, provided by an embodiment of the present invention, combines the Kuz-Ram theoretical model as physical prior knowledge with an extreme learning machine. By incorporating the Kuz-Ram theoretical model into the loss function of the extreme learning machine to constrain the training process of the neural network, a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is obtained. This model not only conforms to existing theoretical models but also mines inherent connections from data. On this basis, only the blasting parameters of the blasting scheme to be predicted need to be obtained, and the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine can be used to obtain the blasting fragmentation prediction result of the blasting scheme to be predicted. Since the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is trained based on historical blasting samples, the blasting fragmentation prediction result of the blasting scheme to be predicted can be simple, convenient, and low-cost.

[0147] Computer-readable storage medium

[0148] The computer-readable storage medium provided by an embodiment of the present invention stores a water conservancy project blasting fragmentation prediction program based on physical information constraints and extreme learning machines. When the water conservancy project blasting fragmentation prediction program based on physical information constraints and extreme learning machines is executed by a processor, the steps of the water conservancy project blasting fragmentation prediction method based on physical information constraints and extreme learning machines provided by the present invention are implemented.

[0149] The computer-readable storage medium provided by an embodiment of the present invention combines the Kuz-Ram theoretical model as physical prior knowledge with an extreme learning machine. By incorporating the Kuz-Ram theoretical model into the extreme learning machine's loss function to constrain the neural network training process, a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is obtained. This model not only conforms to existing theoretical models but also mines inherent connections from data. On this basis, only the blasting parameters of the blasting scheme to be predicted need to be obtained. The trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine can be used to obtain a blasting fragmentation prediction result for the blasting scheme to be predicted. Since the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is trained based on historical blasting samples, it can make the blasting fragmentation prediction result for the blasting scheme to be predicted simple, convenient, and low-cost.

[0150] electronic devices

[0151] An electronic device provided in an embodiment of the present invention includes a memory and a processor. The memory stores a water conservancy project blasting fragmentation prediction program based on physical information constraints and an extreme learning machine. When the water conservancy project blasting fragmentation prediction program based on physical information constraints and an extreme learning machine is executed by the processor, the steps of the water conservancy project blasting fragmentation prediction method based on physical information constraints and an extreme learning machine provided by the present invention are implemented.

[0152] The electronic device provided by an embodiment of the present invention combines the Kuz-Ram theoretical model as physical prior knowledge with an extreme learning machine. By incorporating the Kuz-Ram theoretical model into the extreme learning machine's loss function to constrain the neural network training process, a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is obtained. This model not only conforms to existing theoretical models but also mines inherent connections from data. On this basis, only the blasting parameters of the blasting scheme to be predicted need to be obtained. The trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine can be used to obtain the blasting fragmentation prediction results for the blasting scheme to be predicted. Since the trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine is trained based on historical blasting samples, it can make the blasting fragmentation prediction results for the blasting scheme to be predicted simple, convenient, and low-cost.

[0153] Reference Figure 6 , Figure 6 This is a schematic diagram of the structure of a device for predicting the fragmentation of blasting in water conservancy projects based on physical information constraints and extreme learning machines in the hardware operating environment of an embodiment of the present invention.

[0154] like Figure 6As shown, the device for predicting the fragmentation of blasting in water conservancy projects based on physical information constraints and an extreme learning machine may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and an input unit, such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a wireless fidelity (WI-FI) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk storage device. The memory 1005 may also be a storage device independent of the processor 1001.

[0155] Those skilled in the art will understand that Figure 6 The structure shown in does not constitute a limitation on the hydraulic engineering blasting fragmentation prediction device based on physical information constraints and extreme learning machines, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0156] like Figure 6 As shown, the memory 1005 as a storage medium may include an operating system, a data storage module, a network communication module, a user interface module, and a water conservancy project blasting fragmentation prediction program based on physical information constraints and extreme learning machines.

[0157] exist Figure 6 In the water conservancy project blasting fragmentation prediction device based on physical information constraints and extreme learning machine shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the water conservancy project blasting fragmentation prediction device based on physical information constraints and extreme learning machine of the present invention can be set in the water conservancy project blasting fragmentation prediction device based on physical information constraints and extreme learning machine. The water conservancy project blasting fragmentation prediction device based on physical information constraints and extreme learning machine calls the water conservancy project blasting fragmentation prediction program based on physical information constraints and extreme learning machine stored in the memory 1005 through the processor 1001, and executes the water conservancy project blasting fragmentation prediction method based on physical information constraints and extreme learning machine provided by the embodiment of the present invention.

[0158] Example

[0159] In this specific embodiment, taking a blasting project as an example, the average fragment size x is predicted by using the measured blasting parameters and fragment size distribution data. The method for predicting the fragment size of a hydraulic engineering blasting project based on physical information constraints and extreme learning machines provided in this embodiment is performed in the following steps:

[0160] Step 1: For each blasting operation, record the explosive consumption per unit, blasthole diameter, charge length, hole length, and explosive type before the blast. After the blast, use a camera to take a photo of the blasted ore deposit. Include a reference object when taking the photo, and determine the size of the reference object to facilitate subsequent fragment size identification using Wipfrag software.

[0161] Step 2: Use Wipfrag software to identify multiple photos taken during a blasting operation, obtaining multiple fragment size distribution curves. From these, we determine multiple average fragment sizes x. These average fragment sizes x are then averaged, and this average is used as the average fragment size x for that blasting operation. Furthermore, these parameters are compared with the parameters from the blasting operation described in Step 1 to form a data set. Repeating these two steps yielded a total of 85 data sets. Some of the resulting data is shown in the table below.

[0162] Table 1 Part of the data in the dataset

[0163]

[0164] Step 3: Normalize the explosive consumption per blasthole q, blasthole diameter d, and actual charge length L. All the data are randomly divided into 80% for training, totaling 68 data sets; and 20% for testing, totaling 17 data sets.

[0165] Step 4: Determine the form of the loss function of the extreme learning machine prediction model that incorporates physical information constraints. The loss function includes the extreme learning machine model prediction loss F EML (weighted by η) and the Kuz-Ram theoretical model prediction loss F Theory (The weight is 1-η). The value of η is obtained by particle swarm optimization algorithm. The Kuz-Ram theoretical model is shown in the following formula.

[0166]

[0167] Where A is the rock coefficient, and its value is related to the degree of development of rock joints and fissures; q is the unit consumption of blasthole explosives, kg / m 3 ; E is the explosive power; d is the borehole diameter, m; L is the actual charge length, m; ρ is the charge density in the hole, kg / m 3 .

[0168] Step 5: Build an extreme learning machine prediction model that incorporates physical information constraints. Training is performed using the training set data. During training, a particle swarm optimization algorithm is used to search for the optimal parameter combination for the extreme learning machine. The parameters include the weight η in the loss function, the weight matrix w from the input layer to the hidden layer of the extreme learning machine, and the bias b. The extreme learning machine has 3 input layer nodes, 1 output layer node, 100 hidden layer nodes, a regularization coefficient of 0.00001, and an activation function for the hidden layer. The maximum number of iterations for the particle swarm algorithm is set to 200, the population size is set to 100, the inertia weight and learning factor are set to the default values, the search range for the weight in the loss function is set to [0, 1], and the search range for the weight w and bias b is set to [-1, 1].

[0169] The weight η in the loss function of the extreme learning machine prediction model that incorporates physical information constraints is optimized by the particle swarm optimization algorithm and is 0.1031.

[0170] Step 6: Save the trained prediction model and input the test set into the model. The prediction results are shown in the attached Figure 3 Finally, the evaluation indicators of the model were calculated, and the accuracy (AR) was 80.88%, the mean square error (MSE) was 24.5947, the root mean square error (RMSE) was 4.9593, and the coefficient of determination (R 2 ) is 0.8898.

[0171] It can be seen that the method for predicting the fragmentation of hydraulic engineering blasting based on physical information constraints and extreme learning machines provided by the present invention has the following advantages over the prior art:

[0172] (1) By embedding the Kuz-Ram theoretical model as a physical constraint term into the loss function of the extreme learning machine, the theoretical deviation and data residual are optimized simultaneously during the parameter optimization process. Compared with traditional machine learning models, this method can simultaneously ensure that the prediction results conform to the basic theoretical model of blasting fragmentation, solving the theoretical inconsistency problem of pure data-driven models.

[0173] (2) The constraint mechanism based on physical priors enables the model to have the characteristic of self-explanatory parameters, which can significantly reduce the cost of cross-scenario application and break through the strong dependence of existing methods on specific geological conditions.

[0174] (3) The constraint framework integrating physical laws effectively suppresses data noise interference, reduces the risk of overfitting caused by data sparsity, and improves the practical value of industrial scenarios.

[0175] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0176] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A method for predicting blasting fragmentation in water conservancy projects based on physical information constraints and extreme learning machines, characterized in that: The following steps are involved: Obtaining blasting characteristic parameters of the blasting plan to be predicted; Inputting the blasting characteristic parameters of the blasting scheme to be predicted into a prediction model, wherein the prediction model is a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machines; The prediction model performs calculations based on the blasting characteristic parameters of the blasting scheme to be predicted, and outputs a blasting fragmentation prediction result of the blasting scheme to be predicted.

2. The method for predicting blasting fragmentation in water conservancy projects based on physical information constraints and extreme learning machines according to claim 1 is characterized in that: In the step of inputting the blasting characteristic parameters of the blasting scheme to be predicted into a prediction model, wherein the prediction model is a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine, the training method of the prediction model includes the following steps: Obtaining historical blasting samples, wherein the historical blasting samples include blasting characteristic parameters and blasting characteristic images of the blasting plans of the historical blasting samples, and blasting fragmentation results after the blasting is completed, wherein the blasting fragmentation results after the blasting is completed include blasting characteristic parameters and characteristic images of the rock accumulation body after the blasting; Inputting the blasting characteristic parameters of the blasting plan of the historical blasting sample and the blasting fragmentation result after the blasting is completed into the recognition software for data recognition to obtain a data set; Training is performed based on the processed data set to obtain a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machines.

3. The method for predicting blasting fragmentation in water conservancy projects based on physical information constraints and extreme learning machines according to claim 2 is characterized in that: The process of obtaining historical blasting samples, wherein the historical blasting samples include blasting characteristic parameters of the blasting plan of the historical blasting samples and blasting fragmentation results after the blasting is completed, further includes the following steps: Random classification is performed on the historical blasting samples, and a set proportion of historical samples are selected as the training sample set, and the remaining historical samples are used as the test sample set, where: The training sample set is used for data training to obtain a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine, wherein the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine includes an input layer, an output layer and a hidden layer; The test sample set is used to perform performance testing on a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine, and to adjust relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine.

4. The method for predicting blasting fragmentation in water conservancy projects based on physical information constraints and extreme learning machines according to claim 1 or 2, characterized in that: The blasting characteristic parameters of the blasting plan include explosive consumption per unit, blasthole diameter, charge length, hole length, and explosive type.

5. The method for predicting blasting fragmentation in water conservancy projects based on physical information constraints and extreme learning machines according to claim 3 is characterized in that: The test sample set is used to perform performance testing on a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine, and the process of adjusting relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and an extreme learning machine specifically includes the following steps: The test sample set is input into the extreme learning machine prediction model with physical information constraints optimized by the particle swarm optimization algorithm to obtain the optimized blasting fragmentation test results.

6. The method for predicting blasting fragmentation in water conservancy projects based on physical information constraints and extreme learning machines according to claim 3 is characterized in that: The relevant parameters in the water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine include: The number of nodes in the input layer, output layer, and hidden layer, the regularization coefficient, and the activation function of the hidden layer; The maximum number of iterations of particles, the number of swarms, the inertia weight, the learning factor, and the parameter search range in the particle swarm optimization algorithm, wherein the particle swarm optimization algorithm is used to search for the optimal parameter combination of the extreme learning machine.

7. The method for predicting blasting fragmentation in water conservancy projects based on physical information constraints and extreme learning machines according to claim 2, characterized in that: According to the processed data set, training is performed to obtain a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine, which specifically includes the following steps: Step S1: Record the blasting parameters under different blasting schemes. The blasting parameters should include explosive consumption, blasthole diameter, charge length, hole length, and explosive type, and record the degree of rock joint and fissure development. After the blasting test, take photos of the blasted ore. When taking photos, record the size of the blasted ore from different directions and angles. Step S2: Input the obtained fragmentation picture of the blasted ore into Wipfrag software for identification, thereby obtaining a fragmentation distribution curve and obtaining the average fragmentation size. That is, the block size with an undersize accumulation rate of 50% is obtained, and the data set is obtained; Step S3: Determine the characteristic variables of the model as explosive consumption per unit, blasthole diameter, and charge length, and the target variable as average block size Normalize the feature variable data so that different feature variables have the same scale and eliminate the influence of dimension. The normalization formula is as follows: Where, X n is the dimensionless value after normalization; X is the original data; X min is the minimum value of the original data; X max is the maximum value of the original data, The obtained data is randomly divided into 80% of which is the training set for training the model and 20% is the test set for testing the performance of the model; Step S4: Determine the loss function of the extreme learning machine that incorporates physical information constraints, wherein the loss function of the extreme learning machine that incorporates physical information constraints consists of two parts, one of which is the extreme learning machine model prediction loss F EML , and the other part is the Kuz-Ram theoretical model prediction loss F Theory , assign different weights to these two parts of loss, F EML The weight is η, F Theory The weight is 1-η, the value range of η is [0,1], and the loss function F of the extreme learning machine is as follows: F=ηgF EML +(1-η)gF Theory Among them, i is the i-th training sample, yi is the actual value, y( i is the predicted value of the extreme learning machine model, y( i is the predicted value of the Kuz-Ram theoretical model, n is the number of all training samples, ξ is the regularization coefficient, and w is the weight between the input layer and the hidden layer of the extreme learning machine model; The Kuz-Ram theoretical model is as follows: Where A is the rock coefficient, and its value is related to the degree of development of rock joints and fissures; q is the unit consumption of blasthole explosives, kg / m 3 ; Q is the charge amount per hole, kg; E is the explosive power; d is the borehole diameter, m; L is the actual charge length, m; ρ is the charge density in the hole, kg / m 3 ; Step S5: Establish an extreme learning machine prediction model that incorporates physical information constraints, use the training set data for training, and use the particle swarm optimization algorithm to search for the optimal parameter combination of the extreme learning machine during training. The parameters include the weight η in the loss function, the weight matrix w from the input layer to the hidden layer of the extreme learning machine, and the bias b. The specific steps described in step 5 are as follows: Step 501: Determine the parameters of the extreme learning machine, including the input layer, output layer, number of hidden layer nodes, regularization coefficient, activation function of the hidden layer, and determine the maximum number of iterations of particles in the particle swarm algorithm, the number of populations, inertia weight, learning factor, and parameter search range; Step 502: Randomly initialize a group of particles. These particles represent the solution in the parameter search space. Train the model according to the parameters represented by different particles. Then calculate the loss value by using the prediction results of the training set and the loss function. The loss function is the loss function of the extreme learning machine with physical information constraints as described in step 4, which includes the prediction loss F of the extreme learning machine model. EML The loss F is predicted by the Kuz-Ram theoretical model Theory Then, the particle's velocity and coordinates are iteratively updated according to the velocity update formula and position update formula, and the hyperparameter combination is updated. The velocity update formula and position update formula are as follows: ν ij (t+1)=wv ij (t)+c1r1(t)[p ij (t)-x ij (t)]+c2r2(t)[p gj (t)-x ij (t)] x ij (t+1)=x ij (t)+ν ij (t+1) Among them, v ij is the velocity of the i-th particle in the j-th dimension; x ij is the position of the i-th particle in the j-th dimension; p ij is the local optimal position of the i-th particle; p gj is the global optimal position of the particle swarm; w is the inertia weight; c1 and c2 are learning factors respectively; r1 and r2 are random numbers in the range of [0,1]; The training process of the extreme learning machine model is as follows: The extreme learning machine model is an efficient single-hidden-layer feedforward neural network consisting of three layers: input layer, output layer, and hidden layer. First, the parameters are initialized. The parameters include the hidden layer weights and biases, which are determined by the weights and biases represented by the particles. Then, given a training set, the hidden layer output is calculated. The hidden layer output calculation formula is as follows: Where x i is the i-th input data, w i is the weight of the i-th input feature, b is the bias term, and F is the activation function; Step S503: Determine the output weight β between the hidden layer and the output layer, wherein the principle for determining the output weight β between the hidden layer and the output layer is to minimize the loss function of the model; Based on the loss value of each particle after it moves to a new position, its individual historical best position and the group historical best position are dynamically adjusted. When the loss value tends to be stable or the number of iterations reaches the maximum, the iteration loop is terminated, and the optimal parameter combination is obtained and output. The formula for updating the individual historical best position is as follows: The formula for updating the historical optimal position of the group is as follows: Where, is the historical optimal position of the i-th particle in the t-th generation; is the current particle position; f(*) is the objective function; G best t+1 is the historical optimal position of the group in generation t+1; Step S6: Use the optimal parameters to reconstruct the extreme learning machine model, perform training set reconstruction test, input the test set into the extreme learning machine prediction model with physical information constraints optimized by the particle swarm optimization algorithm, and obtain the average block size Predicted value; Step S7: Calculate the evaluation index of the model by using the actual value and the predicted value, wherein the model evaluation index includes accuracy (AR), mean square error (MSE), root mean square error (RMSE), determination coefficient (R 2 ), the specific calculation formula is as follows: Where i is the i-th test sample, yi is the actual value, is the predicted value, is the mean of the predicted values, and n is the number of all test samples.

8. A device for predicting the fragmentation of blasting in water conservancy projects based on physical information constraints and extreme learning machines, characterized in that: include, A characteristic parameter acquisition module is used to obtain blasting characteristic parameters of the blasting scheme to be predicted; A prediction model, used for inputting blasting characteristic parameters of the blasting scheme to be predicted into the prediction model, wherein the prediction model is a trained water conservancy project blasting fragmentation prediction model based on physical information constraints and extreme learning machine; The prediction result output module is used for outputting the prediction model, which performs data calculation based on the blasting characteristic parameters of the blasting scheme to be predicted, and outputs the blasting fragmentation prediction result of the blasting scheme to be predicted through the prediction result output module.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a water conservancy project blasting fragmentation prediction program based on physical information constraints and extreme learning machines. When the water conservancy project blasting fragmentation prediction program based on physical information constraints and extreme learning machines is executed by the processor, it implements the steps of the water conservancy project blasting fragmentation prediction method based on physical information constraints and extreme learning machines described in any one of claims 1-7.

10. An electronic device, characterized in that: The invention comprises a memory and a processor, wherein the memory stores a water conservancy project blasting fragmentation prediction program based on physical information constraints and an extreme learning machine, and when the water conservancy project blasting fragmentation prediction program based on physical information constraints and an extreme learning machine is executed by the processor, the steps of the water conservancy project blasting fragmentation prediction method based on physical information constraints and an extreme learning machine described in any one of claims 1 to 7 are implemented.

Images