CN111259601A - Blasting blockiness prediction method, device and medium based on random GA-BP neural network group - Google Patents
Blasting blockiness prediction method, device and medium based on random GA-BP neural network group Download PDFInfo
- Publication number
- CN111259601A CN111259601A CN202010045195.7A CN202010045195A CN111259601A CN 111259601 A CN111259601 A CN 111259601A CN 202010045195 A CN202010045195 A CN 202010045195A CN 111259601 A CN111259601 A CN 111259601A
- Authority
- CN
- China
- Prior art keywords
- blasting
- neural network
- blockiness
- prediction
- random
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Physiology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a blasting blockiness prediction method, a blasting blockiness prediction device and a blasting blockiness prediction medium based on a random GA-BP neural network group, wherein the method comprises the following steps: acquiring blasting parameters of an area to be blasted; extracting blasting characteristic data based on the blasting parameters; predicting the blasting average block degree of the area to be blasted according to the blasting characteristic data and a preset random GA-BP neural network group blasting block degree prediction model; the preset random GA-BP neural network group blasting blockiness prediction model is obtained by training a random GA-BP neural network group through historical blasting data of a blasted area. The average blasting blockiness of the area to be blasted is predicted, so that blasting parameters can be reasonably adjusted and designed conveniently, and the blasting effect after blasting is ensured to meet the working condition requirement; the invention predicts the blasting average block degree by the random GA-BP neural network group blasting block degree prediction model, and effectively improves the precision and reliability of the blasting average block degree prediction value.
Description
Technical Field
The invention relates to the technical field of geotechnical engineering, in particular to a blasting blockiness prediction method, a blasting blockiness prediction device and a blasting blockiness prediction medium based on a random GA-BP neural network group.
Background
The size of the rock blasting block size is an important index for evaluating the blasting effect, subsequent procedures such as shoveling, transporting and the like are influenced, the rock blasting block size is closely related to the benefits of the whole related production process, and different blasting operations are carried out, and the target block size of the blasted rock is different. Therefore, many experts and scholars at home and abroad make many efforts to research the prediction model of the rock blasting block size.
The Kuz-Ram model is a first step blasting lump size prediction equation and is an equation for calculating the average fragment size and uniformity index n of a step blasting circle of which the screening curve follows the RR function. In addition, Split-Desktop software is used for carrying out image analysis to find out the block size distribution, the uniformity index and the average particle size of the broken rock, and then a multiple regression method and a decision test and evaluation laboratory (DEMATEL) technology are adopted to establish a new model of the uniformity index and the average block size. And a small blasting database is established by utilizing data of a plurality of blasting events carried out in different mines and rock stratums in the world, the data is divided into two groups by adopting a hierarchical clustering analysis method, and a prediction equation of blasting block size distribution of each group is established by utilizing multivariate regression analysis.
With the rapid development of machine learning, various methods for predicting the explosion blocking degree are developed. If the method is used for solving the problems existing in the traditional step blasting blockiness distribution evaluation method, the average particle size (X) generated by rock blasting of different mines is predicted by adopting a Support Vector Machine (SVM) regression method based on the statistical learning theory50). If a BP neural network is used for replacing multivariate regression analysis to establish a prediction equation of each group of blasting block size distribution, four training methods are adopted for training, and the Levenberg-Marquardt (LM) algorithm is determined to be the optimal training method. At present, BP neural network is widely used in the aspect of explosion block degree predictionThe preparation is used.
Although the BP neural network is widely applied to the prediction of the blasting block size, the BP neural network has the defects of low learning speed, high possibility of falling into a local minimum value and the like. In the process of actual application to blasting block degree prediction, the efficiency is low, the error of the predicted value is large, and the prediction reliability is not high.
Disclosure of Invention
The invention provides a random GA-BP neural network group-based blasting block size prediction method, a random GA-BP neural network group-based blasting block size prediction device and a random GA-BP neural network group-based blasting block size prediction medium, which are used for solving the problems of large error and insufficient reliability in the prior art based on BP neural network prediction blasting block size.
In a first aspect of the present invention, a method for predicting blasting blockiness based on a random GA-BP neural network group is provided, which includes:
acquiring blasting parameters of an area to be blasted;
extracting blasting characteristic data based on the blasting parameters;
predicting the blasting average block degree of the area to be blasted according to the blasting characteristic data and a preset random GA-BP neural network group blasting block degree prediction model; the preset random GA-BP neural network group blasting blockiness prediction model is obtained by training a random GA-BP neural network group through historical blasting data of a blasted area.
The average blasting block size after blasting in the area to be blasted can be predicted by the acquired blasting parameters of the area to be blasted and the trained random GA-BP neural network group blasting block size prediction model so as to reasonably adjust and design the blasting parameters and enable the predicted value of the average blasting block size to meet the requirement and ensure that the blasting effect after blasting is implemented meets the working condition requirement, wherein the blasted area and the area to be blasted are in the same area or the similar area. The invention predicts the blasting average block degree by a random GA-BP neural network group blasting block degree prediction model obtained by training a random GA-BP neural network group through the historical blasting data of the blasted area, the random GA-BP neural network group blasting block degree prediction model is established based on the thought of a random forest algorithm and by taking the GA-BP neural network as a core, the probability that the BP neural network is limited to a local minimum value can be reduced through a Genetic Algorithm (GA), and the output blasting average block degree prediction result is more accurate and stable by combining the thought of the random forest algorithm, so that the precision and the reliability of the blasting average block degree prediction value are effectively improved.
Further, the preset random GA-BP neural network group blasting blockiness prediction model is obtained by training a random GA-BP neural network group through historical blasting data of a blasted area, and specifically includes:
acquiring blasting parameters and blasting average blockiness in historical blasting data of a blasted area;
extracting blasting characteristic data of the blasted area based on the blasting parameters of the blasted area, performing standardized preprocessing on the blasting characteristic data, and then constructing a blasting sample set based on the preprocessed blasting characteristic data of the blasted area and the corresponding blasting average block size;
extracting N groups of sub-training sets from the blasting sample set by a random sampling method with a release function, wherein N is a preset value;
respectively training the GA-BP neural network by using N groups of sub-training sets by using blasting characteristic data as input and corresponding blasting average block degrees as output to obtain N GA-BP neural network blasting block degree prediction models; wherein, each training obtains a GA-BP neural network blasting block degree prediction model, and the GA-BP neural network blasting block degree prediction model is respectively checked by using a corresponding subtest set; if the prediction error of the GA-BP neural network blasting blockiness prediction model exceeds a preset value, randomly extracting a group of sub-training sets and corresponding sub-test sets from the blasting sample set again, establishing a GA-BP neural network blasting blockiness prediction model by utilizing the group of sub-training sets, and checking the GA-BP neural network blasting blockiness prediction model by using the corresponding sub-test sets until the prediction error of the obtained GA-BP neural network blasting blockiness prediction model does not exceed the preset value;
and calculating the weight corresponding to each of the N GA-BP neural network blasting blockiness prediction models according to the prediction error of each of the N GA-BP neural network blasting blockiness prediction models, and establishing a random GA-BP neural network group blasting blockiness prediction model based on the N GA-BP neural network blasting blockiness prediction models, wherein the blasting average blockiness prediction value output by the random GA-BP neural network group blasting blockiness prediction model is the sum of the products of the output of each of the N GA-BP neural network blasting blockiness prediction models and the weight corresponding to the output.
The weight corresponding to each of the N GA-BP neural network blasting blockiness prediction models is calculated by the following formula:
in the formula, eiFor the error, x, of the ith prediction model in the N GA-BP neural network blasting blockiness prediction modelsiAnd (4) the weight of the ith GA-BP neural network blasting blockiness prediction model.
The mathematical model of the output result of the random GA-BP neural network group blasting blockiness prediction model is as follows:
in the formula: x is the number ofiWeight, y, of the ith GA-BP neural network blasting blockiness prediction modeliAnd Y is the total output of the N GA-BP neural network blasting blockiness prediction models after weighted averaging, namely the output of the random GA-BP neural network blasting blockiness prediction model, namely the predicted value of the blasting average blockiness.
The random forest algorithm (RF) selects training data by adopting a random put-back sampling method, constructs a plurality of decision tree classifiers, constructs optimal segmentation by randomly selecting features, and finally combines the constructed weak classifiers to increase the overall effect. The randomness of the random forest algorithm (RF) is realized by the randomness of the selected data and characteristics, and even if the data or the characteristics are lost, the random forest algorithm (RF) has a more accurate result. In addition, since a plurality of classifiers are constructed, errors caused by the result of a certain classifier can be reduced or eliminated. The invention provides a random GA-BP neural network group blasting blockiness prediction model taking a GA-BP neural network blasting blockiness prediction model as a core by combining the thought of a random forest algorithm (RF) and replacing a decision tree classifier with the GA-BP neural network blasting blockiness prediction model. The random GA-BP neural network group blasting blockiness prediction model inherits the advantages of the idea of a random forest algorithm (RF), the accuracy and the stability of the output result of the random GA-BP neural network group blasting blockiness prediction model are guaranteed through balancing errors, and the defects of the fault tolerance capability and the generalization capability of the GA-BP neural network blasting blockiness prediction model are made up through strong generalization capability and fault tolerance capability.
Further, the GA-BP neural network is obtained by optimizing a BP neural network through a genetic algorithm, and specifically includes:
a1, randomly generating a plurality of groups of weight values and threshold values of the BP neural network, and coding the weight values and the threshold values;
a2, carrying out fitness evaluation on the multiple groups of weights and thresholds respectively;
a3, if one or more groups of evaluation results corresponding to the weight and the threshold value meet the requirements, namely the error of the blasting block degree predicted value output by the BP neural network output layer is within a preset range, selecting a group of weight and threshold value corresponding to the smallest error of the blasting block degree predicted value output by the BP neural network output layer as an optimal weight and an optimal threshold value, and taking the optimal weight and the optimal threshold value as the weight and the threshold value of the BP neural network to obtain the GA-BP neural network; if no evaluation result corresponding to one set of weight and threshold meets the requirement, selecting, crossing and mutating the multiple sets of weight and threshold to generate new multiple sets of weight and threshold, and returning to the step A2.
The fitness function adopted for respectively evaluating the fitness of the multiple groups of weights and thresholds is a BP neural network, the evaluation method adopts mean square error or root mean square error in the evaluation of the BP neural network, and if the mean square error or the root mean square error corresponding to the evaluation of the fitness of the group of weights and thresholds is smaller than a preset value, the evaluation result meets the requirement. Wherein the Mean Square Error (MSE) and the Root Mean Square Error (RMSE) are calculated as follows:
in the formula, yiFor the value of the average block size of the actual blasting,
and n is the total number of training samples.
The Genetic Algorithm (GA) is a randomized search method which is evolved by simulating the evolution rule of survival, superiority and inferiority of a suitable person in the natural world. The method continuously evolves a 'chromosome' group represented by problem solution codes by generations through a probabilistic optimization method of selection, intersection and variation, and finally converges to the most adaptive group, so that the optimal solution or the satisfactory solution of the problem is obtained. The method has the advantages of inherent hidden parallelism and better global optimization capability. A Genetic Algorithm (GA) is used for optimizing the BP neural network, the global search capability of the genetic algorithm is mainly utilized, the characteristic that fitness evaluation can be carried out on multiple groups of weights and thresholds is utilized to optimize the weights and the thresholds of the BP neural network, and therefore the possibility that the established GA-BP neural network blasting block degree prediction model falls into a minimum value is reduced.
Further, after the extracting of the blasting characteristic data of the blasted area based on the blasting parameters of the blasted area, the method further comprises: and carrying out standardized preprocessing on the obtained blasting characteristic data and the blasting average block size.
Further, the blasting parameters comprise a resistance line, inter-hole spacing, step height, blocking length, blast hole diameter, explosive unit consumption, rock elastic modulus and in-situ rock bulkiness; blasting characteristic data extracted based on the blasting parameters comprise step height to resisting line ratio, hole spacing to resisting line ratio, resisting line to blast hole diameter ratio, blocking length to resisting line ratio, explosive unit consumption, in-situ rock block size and rock elastic modulus.
In a second aspect of the present invention, a random GA-BP neural network group-based blasting blockiness prediction apparatus is provided, including:
the first data acquisition module is used for acquiring blasting parameters of an area to be blasted;
the first data extraction module is used for extracting blasting characteristic data based on the blasting parameters;
the blasting average block degree prediction module is used for predicting the blasting average block degree of the area to be blasted according to the blasting characteristic data and a preset random GA-BP neural network group blasting block degree prediction model; the preset random GA-BP neural network group blasting blockiness prediction model is obtained by training a random GA-BP neural network group through historical blasting data of a blasted area.
Further, after the extracting of the blasting characteristic data of the blasted area based on the blasting parameters of the blasted area, the method further comprises: and carrying out standardized preprocessing on the obtained blasting characteristic data and the blasting average block size.
The second data acquisition module is used for acquiring blasting parameters and blasting average blockiness in historical blasting data of a blasted area;
the second data extraction module is used for extracting blasting characteristic data of the blasted area based on the blasting parameters of the blasted area, carrying out standard pretreatment on the blasting characteristic data, and then constructing a blasting sample set based on the pretreated blasting characteristic data of the blasted area and the corresponding blasting average block size of the blasting characteristic data;
the random GA-BP neural network group blasting block degree prediction model generation module is used for generating a random GA-BP neural network group blasting block degree prediction model and comprises the following steps:
the sampling unit is used for extracting N groups of sub training sets and corresponding sub test sets from the blasting sample set by a random sampling method with a release function, wherein N is a preset value;
the GA-BP neural network blasting block degree prediction model generation unit is used for training the GA-BP neural network by using N groups of sub-training sets by taking blasting characteristic data as input and corresponding blasting average block degrees as output to obtain N GA-BP neural network blasting block degree prediction models; wherein, each training obtains a GA-BP neural network blasting block degree prediction model, and the GA-BP neural network blasting block degree prediction model is respectively checked by using a corresponding subtest set; if the prediction error of the GA-BP neural network blasting blockiness prediction model exceeds a preset value, randomly extracting a group of sub-training sets and corresponding sub-test sets from the blasting sample set again, establishing a GA-BP neural network blasting blockiness prediction model by utilizing the group of sub-training sets, and checking the GA-BP neural network blasting blockiness prediction model by using the corresponding sub-test sets until the prediction error of the obtained GA-BP neural network blasting blockiness prediction model does not exceed the preset value;
and the random GA-BP neural network group blasting blockiness prediction model generation unit is used for calculating respective corresponding weights according to respective prediction errors of the N GA-BP neural network blasting blockiness prediction models, establishing the random GA-BP neural network group blasting blockiness prediction model based on the N GA-BP neural network blasting blockiness prediction models, and outputting a blasting average blockiness prediction value which is the sum of products of respective outputs of the N GA-BP neural network blasting blockiness prediction models and the corresponding weights.
Further, the GA-BP neural network blasting blockiness prediction model generation unit includes:
the GA-BP neural network generation subunit is used for optimizing the BP neural network through a genetic algorithm to obtain the GA-BP neural network, and the optimization specific process comprises the following steps:
b1, randomly generating a plurality of groups of weight values and threshold values of the BP neural network, and coding the weight values and the threshold values;
b2, carrying out fitness evaluation on the multiple groups of weights and thresholds respectively;
b3, if one or more groups of evaluation results corresponding to the weight and the threshold value meet the requirements, namely the error of the blasting block degree predicted value output by the BP neural network output layer is within a preset range, selecting a group of weight and threshold value corresponding to the smallest error of the blasting block degree predicted value output by the BP neural network output layer as an optimal weight and an optimal threshold value, and taking the optimal weight and the optimal threshold value as the weight and the threshold value of the BP neural network to obtain the GA-BP neural network; if no evaluation result corresponding to one group of weight values and thresholds meets the requirements, selecting, crossing and mutating the multiple groups of weight values and thresholds to generate new multiple groups of weight values and thresholds, and returning to the step B2;
and the GA-BP neural network blasting block degree prediction model generation subunit is used for training the GA-BP neural network by using N groups of sub training sets and establishing N GA-BP neural network blasting block degree prediction models by using blasting characteristic data as input and corresponding blasting average block degrees as output.
Further, the blasting parameters comprise a resistance line, inter-hole spacing, step height, blocking length, blast hole diameter, explosive unit consumption, rock elastic modulus and in-situ rock bulkiness; blasting characteristic data extracted based on the blasting parameters comprise step height to resisting line ratio, hole spacing to resisting line ratio, resisting line to blast hole diameter ratio, blocking length to resisting line ratio, explosive unit consumption, in-situ rock block size and rock elastic modulus.
In a third aspect of the present invention, there is provided a computer readable storage medium storing program instructions adapted to be loaded by a processor and execute the method for predicting the blasting block size based on the random GA-BP neural network population according to the first aspect of the present invention.
Advantageous effects
The invention provides a blasting block size prediction method, a device and a medium based on a random GA-BP neural network group, which can predict the blasting average block size after blasting in a to-be-blasted area through the acquired blasting parameters of the to-be-blasted area and a trained random GA-BP neural network group blasting block size prediction model so as to reasonably adjust and design the blasting parameters, ensure that the predicted value of the blasting average block size meets the requirement and ensure that the blasting effect after blasting meets the working condition requirement. The invention predicts the blasting average block degree by a random GA-BP neural network group blasting block degree prediction model obtained by training a random GA-BP neural network group through the historical blasting data of the blasted area, the random GA-BP neural network group blasting block degree prediction model is established based on the thought of a random forest algorithm and by taking the GA-BP neural network as a core, the probability that the BP neural network is limited to a local minimum value can be reduced through a Genetic Algorithm (GA), and the output blasting average block degree prediction result is more accurate and stable by combining the thought of the random forest algorithm, so that the precision and the reliability of the blasting average block degree prediction value are effectively improved.
Drawings
Fig. 1 is a flowchart of a blasting blockiness prediction method based on a random GA-BP neural network group according to an embodiment of the present invention;
FIG. 2 is a flow chart of the blasting blockiness prediction model establishment for the random GA-BP neural network group provided by the embodiment of the present invention;
fig. 3 is a flowchart for optimizing each BP neural network by using a genetic algorithm according to an embodiment of the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
Example 1
As shown in fig. 1, the present embodiment provides a method for predicting a blasting block size based on a random GA-BP neural network group, including:
step S01: acquiring blasting parameters of an area to be blasted;
step S02: extracting blasting characteristic data based on the blasting parameters;
step S03: predicting the blasting average block degree of the area to be blasted according to the blasting characteristic data and a preset random GA-BP neural network group blasting block degree prediction model; the preset random GA-BP neural network group blasting blockiness prediction model is obtained by training a random GA-BP neural network group through historical blasting data of a blasted area.
The average blasting block size after blasting in the area to be blasted can be predicted by the acquired blasting parameters of the area to be blasted and the trained random GA-BP neural network group blasting block size prediction model so as to reasonably adjust and design the blasting parameters and enable the predicted value of the average blasting block size to meet the requirement and ensure that the blasting effect after blasting is implemented meets the working condition requirement, wherein the blasted area and the area to be blasted are in the same area or the similar area. The invention predicts the blasting average block degree by a random GA-BP neural network group blasting block degree prediction model obtained by training a random GA-BP neural network group through the historical blasting data of the blasted area, the random GA-BP neural network group blasting block degree prediction model is established based on the thought of a random forest algorithm and by taking the GA-BP neural network as a core, the probability that the BP neural network is limited to a local minimum value can be reduced through a Genetic Algorithm (GA), and the output blasting average block degree prediction result is more accurate and stable by combining the thought of the random forest algorithm, so that the precision and the reliability of the blasting average block degree prediction value are effectively improved.
Specifically, as shown in fig. 2, in this embodiment, the preset random GA-BP neural network group blasting blockiness prediction model is obtained by training a random GA-BP neural network group through historical blasting data of a blasted area, and specifically includes:
acquiring blasting parameters and blasting average blockiness in historical blasting data of a blasted area;
extracting blasting characteristic data of the blasted area based on the blasting parameters of the blasted area, performing standard pretreatment on the blasting characteristic data and the blasting average block size, and then constructing a blasting sample set based on the pretreated blasting characteristic data of the blasted area and the corresponding blasting average block size;
extracting N groups of sub-training sets and corresponding sub-test sets from the blasting sample set by a sampling method with random replacement, wherein N is a preset value, in the specific implementation, a bootstrap sampling method can be adopted, and the ratio of the number of samples in each group of sub-training sets to the number of samples in the sub-test sets is 3-10: 1;
respectively training the GA-BP neural network by using N groups of sub-training sets by using blasting characteristic data as input and corresponding blasting average block degrees as output to obtain N GA-BP neural network blasting block degree prediction models; wherein, each training obtains a GA-BP neural network blasting block degree prediction model, and the GA-BP neural network blasting block degree prediction model is respectively checked by using a corresponding subtest set; if the prediction error of the GA-BP neural network blasting blockiness prediction model exceeds a preset value, randomly extracting a group of sub-training sets and corresponding sub-test sets from the blasting sample set again, establishing a GA-BP neural network blasting blockiness prediction model by utilizing the group of sub-training sets, and checking the GA-BP neural network blasting blockiness prediction model by using the corresponding sub-test sets until the prediction error of the obtained GA-BP neural network blasting blockiness prediction model does not exceed the preset value;
and calculating the weight corresponding to each of the N GA-BP neural network blasting blockiness prediction models according to the prediction error of each of the N GA-BP neural network blasting blockiness prediction models, and establishing a random GA-BP neural network group blasting blockiness prediction model based on the N GA-BP neural network blasting blockiness prediction models, wherein the blasting average blockiness prediction value output by the random GA-BP neural network group blasting blockiness prediction model is the sum of the products of the output of each of the N GA-BP neural network blasting blockiness prediction models and the weight corresponding to the output.
The weight corresponding to each of the N GA-BP neural network blasting blockiness prediction models is calculated by the following formula:
in the formula, eiFor the error, x, of the ith prediction model in the N GA-BP neural network blasting blockiness prediction modelsiAnd (4) the weight of the ith GA-BP neural network blasting blockiness prediction model.
The mathematical model of the output result of the random GA-BP neural network group blasting blockiness prediction model is as follows:
in the formula: x is the number ofiWeight, y, of the ith GA-BP neural network blasting blockiness prediction modeliAnd Y is the total output of the N GA-BP neural network blasting blockiness prediction models after weighted averaging, namely the output of the random GA-BP neural network blasting blockiness prediction model, namely the predicted value of the blasting average blockiness.
The random forest algorithm (RF) selects training data by adopting a random put-back sampling method, constructs a plurality of decision tree classifiers, constructs optimal segmentation by randomly selecting features, and finally combines the constructed weak classifiers to increase the overall effect. The randomness of the random forest algorithm (RF) is realized by the randomness of the selected data and characteristics, and even if the data or the characteristics are lost, the random forest algorithm (RF) has a more accurate result. In addition, since a plurality of classifiers are constructed, errors caused by the result of a certain classifier can be reduced or eliminated. The invention provides a random GA-BP neural network group blasting blockiness prediction model taking a GA-BP neural network blasting blockiness prediction model as a core by combining the thought of a random forest algorithm (RF) and replacing a decision tree classifier with the GA-BP neural network blasting blockiness prediction model. The random GA-BP neural network group blasting blockiness prediction model inherits the advantages of a random forest algorithm (RF), the accuracy and the stability of the output result of the random GA-BP neural network group blasting blockiness prediction model are guaranteed through balancing errors, and the defects of the fault tolerance capability and the generalization capability of the GA-BP neural network blasting blockiness prediction model are made up through strong generalization capability and fault tolerance capability.
More specifically, as shown in fig. 3, the GA-BP neural network is obtained by optimizing a BP neural network through a genetic algorithm, and specifically includes:
a1, randomly generating a plurality of groups of weight values and threshold values of the BP neural network, and coding the weight values and the threshold values;
a2, carrying out fitness evaluation on the multiple groups of weights and thresholds respectively;
a3, if one or more groups of evaluation results corresponding to the weight and the threshold value meet the requirements, namely the error of the blasting block degree predicted value output by the BP neural network output layer is within a preset range, selecting a group of weight and threshold value corresponding to the smallest error of the blasting block degree predicted value output by the BP neural network output layer as an optimal weight and an optimal threshold value, and taking the optimal weight and the optimal threshold value as the weight and the threshold value of the BP neural network to obtain the GA-BP neural network; if no evaluation result corresponding to one set of weight and threshold meets the requirement, selecting, crossing and mutating the multiple sets of weight and threshold to generate new multiple sets of weight and threshold, and returning to the step A2.
The fitness function adopted for respectively evaluating the fitness of the multiple groups of weights and thresholds is a BP neural network, the evaluation method adopts mean square error or root mean square error in the evaluation of the BP neural network, and if the mean square error or the root mean square error corresponding to the evaluation of the fitness of the group of weights and thresholds is smaller than a preset value, the evaluation result meets the requirement. Wherein the Mean Square Error (MSE) and the Root Mean Square Error (RMSE) are calculated as follows:
in the formula, yiFor the value of the average block size of the actual blasting,
and n is the total number of training samples.
The Genetic Algorithm (GA) is a randomized search method which is evolved by simulating the evolution rule of survival, superiority and inferiority of a suitable person in the natural world. The method continuously evolves a 'chromosome' group represented by problem solution codes by generations through a probabilistic optimization method of selection, intersection and variation, and finally converges to the most adaptive group, so that the optimal solution or the satisfactory solution of the problem is obtained. The method has the advantages of inherent hidden parallelism and better global optimization capability. A Genetic Algorithm (GA) is used for optimizing the BP neural network, the global search capability of the genetic algorithm is mainly utilized, the characteristic that fitness evaluation can be carried out on multiple groups of weights and thresholds is utilized to optimize the weights and the thresholds of the BP neural network, and therefore the possibility that the established GA-BP neural network blasting block degree prediction model falls into a minimum value is reduced.
Of course, it should be understood that, in other embodiments, the Genetic Algorithm (GA) may be replaced by optimization algorithms such as Colonial Competition Algorithm (CCA), empire kingdom competition algorithm (ICA), artificial bee colony Algorithm (ABC), Artificial Immune System (AIS), Particle Swarm Optimization (PSO), and the like, and the replacement of the above algorithms also belongs to the protection scope of the present invention.
In this embodiment, the blasting parameters include a resistance line, an inter-hole distance, a step height, a blocking length, a blast hole diameter, a unit consumption of explosive, a rock elastic modulus and an in-situ rock bulk; blasting characteristic data extracted based on the blasting parameters comprise step height to resisting line ratio, hole spacing to resisting line ratio, resisting line to blast hole diameter ratio, blocking length to resisting line ratio, explosive unit consumption, in-situ rock block size and rock elastic modulus.
Example 2
The embodiment provides a blasting blockiness prediction device based on a random GA-BP neural network group, which includes:
the first data acquisition module is used for acquiring blasting parameters of an area to be blasted;
the first data extraction module is used for extracting blasting characteristic data based on the blasting parameters;
the blasting average block degree prediction module is used for predicting the blasting average block degree of the area to be blasted according to the blasting characteristic data and a preset random GA-BP neural network group blasting block degree prediction model; the preset random GA-BP neural network group blasting blockiness prediction model is obtained by training a random GA-BP neural network group through historical blasting data of a blasted area.
In this embodiment, the method further includes:
the second data acquisition module is used for acquiring blasting parameters and blasting average blockiness in historical blasting data of a blasted area;
the second data extraction module is used for extracting blasting characteristic data of the blasted area based on the blasting parameters of the blasted area, performing standard pretreatment on the blasting characteristic data and the blasting average block size, and then constructing a blasting sample set based on the pretreated blasting characteristic data of the blasted area and the corresponding blasting average block size;
the random GA-BP neural network group blasting block degree prediction model generation module is used for generating a random GA-BP neural network group blasting block degree prediction model and comprises the following steps:
the sampling unit is used for extracting N groups of sub training sets and corresponding sub test sets from the blasting sample set by a random sampling method with a release function, wherein N is a preset value;
the GA-BP neural network blasting block degree prediction model generation unit is used for training the GA-BP neural network by using N groups of sub-training sets by taking blasting characteristic data as input and corresponding blasting average block degrees as output to obtain N GA-BP neural network blasting block degree prediction models; wherein, each training obtains a GA-BP neural network blasting block degree prediction model, and the GA-BP neural network blasting block degree prediction model is respectively checked by using a corresponding subtest set; if the prediction error of the GA-BP neural network blasting blockiness prediction model exceeds a preset value, randomly extracting a group of sub-training sets and corresponding sub-test sets from the blasting sample set again, establishing a GA-BP neural network blasting blockiness prediction model by utilizing the group of sub-training sets, and checking the GA-BP neural network blasting blockiness prediction model by using the corresponding sub-test sets until the prediction error of the obtained GA-BP neural network blasting blockiness prediction model does not exceed the preset value;
and the random GA-BP neural network group blasting blockiness prediction model generation unit is used for calculating respective corresponding weights according to respective prediction errors of the N GA-BP neural network blasting blockiness prediction models, establishing the random GA-BP neural network group blasting blockiness prediction model based on the N GA-BP neural network blasting blockiness prediction models, and outputting a blasting average blockiness prediction value which is the sum of products of respective outputs of the N GA-BP neural network blasting blockiness prediction models and the corresponding weights.
More specifically, the GA-BP neural network blasting blockiness prediction model generation unit includes:
the GA-BP neural network generation subunit is used for optimizing the BP neural network through a genetic algorithm to obtain the GA-BP neural network, and the optimization specific process comprises the following steps:
b1, randomly generating a plurality of groups of weight values and threshold values of the BP neural network, and coding the weight values and the threshold values;
b2, carrying out fitness evaluation on the multiple groups of weights and thresholds respectively;
b3, if one or more groups of evaluation results corresponding to the weight and the threshold value meet the requirements, namely the error of the blasting block degree predicted value output by the BP neural network output layer is within a preset range, selecting a group of weight and threshold value corresponding to the smallest error of the blasting block degree predicted value output by the BP neural network output layer as an optimal weight and an optimal threshold value, and taking the optimal weight and the optimal threshold value as the weight and the threshold value of the BP neural network to obtain the GA-BP neural network; if no evaluation result corresponding to one group of weight values and thresholds meets the requirements, selecting, crossing and mutating the multiple groups of weight values and thresholds to generate new multiple groups of weight values and thresholds, and returning to the step B2;
and the GA-BP neural network blasting block degree prediction model generation subunit is used for training the GA-BP neural network by using N groups of sub training sets and establishing N GA-BP neural network blasting block degree prediction models by using blasting characteristic data as input and corresponding blasting average block degrees as output.
In this embodiment, the blasting parameters include a resistance line, an inter-hole distance, a step height, a blocking length, a blast hole diameter, a unit consumption of explosive, a rock elastic modulus and an in-situ rock bulk; blasting characteristic data extracted based on the blasting parameters comprise step height to resisting line ratio, hole spacing to resisting line ratio, resisting line to blast hole diameter ratio, blocking length to resisting line ratio, explosive unit consumption, in-situ rock block size and rock elastic modulus.
For another specific implementation of the apparatus for predicting blasting block size based on a random GA-BP neural network group provided in this embodiment, reference may be made to embodiment 1 to provide a method for predicting blasting block size based on a random GA-BP neural network group, which is not described herein again.
Example 3
The present embodiment provides a computer-readable storage medium, which stores program instructions adapted to a processor to load and execute the method for predicting the blasting block size based on the random GA-BP neural network population according to embodiment 1.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The technical solution provided by the present invention is further described below with reference to a specific example.
In this example, the data in this text is derived from a database containing global multi-site blasting information established by hudavidiet al, which comprises 97 sets of blasting data, each set comprising 8 blasting parameters determining the size of the blasting block and the corresponding blasting average block (X) degree50) The method comprises the following steps of (1) respectively setting six main blasting design parameters, one rock mechanical property parameter and one rock structure parameter: resistance line (B, m), inter-hole spacing (S, m), step height (H, m), plugging length (T, m), blast hole diameter (D, m), specific charge (Pf, kg/m)3) Rock elastic modulus (E, GPa) and rock in situ bulk (X)BM). Since some blasting researchers have regarded the blasting design parameters as proportion, the blasting parameters of all the blasting data are extracted to obtain blasting characteristic data, i.e. input parameters of the model, including step height to resisting line ratio (H/B), inter-hole spacing to resisting line ratio (S/B), resisting line to blast hole diameter ratio (B/D), plugging length to resisting line ratio (T/B), specific charge (Pf), in-situ rock block size (X)B) And rock modulus of elasticity (E).
In the training process of the random GA-BP neural network group blasting blockiness prediction model (RGA-BPNNG model), in order to prevent the loss of precision caused by different dimensions and orders of training parameters and reduce the influence of errors of the maximum value and the minimum value in a data set on the whole data set, the data are subjected to standardization preprocessing. Table 1 shows the normalized pre-processed data range, which is kept at three decimal places.
Carrying out standardization preprocessing on the data, wherein the standardization process formula is as follows:
in the formula: x is the data set before normalization, xmeanIs the average value of x, y is the normalized data set, and n is the total number of samples; in this example, the data set represented by x and y may be the ratio of step height to resistance line, the ratio of inter-hole spacing to resistance line, and the ratio of resistance line to borehole diameterAny data set of the ratio, the plugging length to the resistance line ratio, the specific charge, the in-situ rock block degree, the rock elastic modulus and the blasting average block degree.
TABLE 1 normalized Pre-processed data Range
| Parameter(s) | H/B | S/B | B/D | T/B | Pf | XB | E | X50 |
| Categories | Input device | Input device | Input device | Input device | Input device | Input device | Input device | Output of |
| Minimum value | -1.618 | -1.234 | -1.938 | -1.132 | -1.313 | -2.042 | -1.113 | -1.500 |
| Maximum value | 4.805 | 2.127 | 2.504 | 5.056 | 3.104 | 2.336 | 1.708 | 3.490 |
Since there are seven input parameters, one output parameter, the input layer and the output layer of the BP neural network (BPNN) are 7 layers and 1 layer, respectively. The number of hidden layers has a significant impact on the performance of the BPNN. In general, the greater the number of hidden layers, the better the performance of the BPNN. But may result in too long training time or overfitting phenomena, so it is extremely important to select the appropriate number of hidden layers. At present, there is no suitable analytic formula to determine the number of hidden layer layers, and it is a common practice to select a suitable number of hidden layer layers according to an estimated value of the number of hidden layer layers obtained by an empirical formula or according to personal experience. In order to reduce the iteration times and the running time cost of the whole RGA-BPNNG model and increase the accuracy of the prediction result, a Genetic Algorithm (GA) is adopted to optimize the number of implicit layers of the BPNN, and the result shows that the prediction result is best when the number of implicit layers is 16, so that the structure of the BPNN is 7 multiplied by 16 multiplied by 1 by selecting 16 as the number of implicit layers.
In addition, the hidden layer selects a Tan-Sigmoid function as a transfer function, and the output layer selects a linear function as a transfer function. And selecting a Levenberg-Marquardt (trainlm) back propagation algorithm training function in the weight correction process. Meanwhile, the precision of the training model is set to be 0.001, the learning rate is set to be 0.1, and the maximum iteration number is set to be 1000.
In the Genetic Algorithm (GA), the initial population value range is set to (-3,3) because the data is normalized. The population size was set to 100, the maximum generation was 200, and the crossover and mutation probabilities were 0.8 and 0.15, respectively.
To construct the RGA-BPNNG model and evaluate the performance of the model, 82 sets of data were used as the training set, and 15 sets of data were left as the test set. Only test set data is listed here, considering the amount of data, see table 2. Due to the process of constructing the GA-BP neural network blasting blockiness prediction model (GA-BPNN model), the iteration times are too many, the operation time is too long, and when the prediction result exceeds the limit, a new GA-BPNN model is reconstructed, so that the operation time cost is increased again. However, the more GA-BPNN models that are built, the better the performance of the RGA-BPNNG model. Therefore, in order to reduce the program runtime and guarantee the performance of the RGA-BPNNG model, 20 GA-BPNN models were built herein. In addition, the maximum relative error is less than 15 percent and is taken as the qualified basis of the GA-BPNN model. And after 20 GA-BPNN models are obtained, calculating the weight of the corresponding GA-BPNN model according to the average relative error. After the whole RGA-BPNNG model is built, the weight values and the threshold values of 20 groups of GA-BPNN models and the weight values corresponding to the GA-BPNN models are obtained.
TABLE 2 test set
In order to evaluate the stability and the accuracy of the prediction result of the random GA-BP neural network cluster blasting blockiness prediction model (RGA-BPNNG model), 3 RGA-BPNNG models are established by using the same training set and model parameters to carry out the test setPredicting, respectively establishing 3 GA-BP neural network blasting block degree prediction models (GA-BPNN models) as comparison, and adopting correlation coefficient (R)2) Three indexes of Root Mean Square Error (RMSE) and mean square error (MRE) are used as evaluation criteria. In the prediction program based on 6 models, the parameter settings of BPNN were matched to the parameters of the models. Table 3 and Table 4 show the predicted values and relative errors of the 3 RGA-BPNNG models and the 3 GA-BPNN models, respectively, and Table 5 shows the correlation coefficients (R) of the 6 models2) Three decimal places were kept in table 3, table 4, and table 5 for the Root Mean Square Error (RMSE) and average relative error (MRE) values.
Wherein the correlation coefficient (R)2) Calculated by the following formula:
in the formula, n represents that the RGA-BPNNG model predicts n test sets and obtains n prediction results,
for the ith prediction, YiThe actual value of the ith test sample in the n test sets. R2Larger values indicate better prediction by the model.
Root Mean Square Error (RMSE):
in the formula, n represents that the RGA-BPNNG model predicts n test sets and obtains n prediction results,
for the ith prediction, YiThe actual value of the ith test sample in the n test sets. Smaller RMSE values indicate better model prediction.
Mean Relative Error (MRE):
in the formula, n represents that the RGA-BPNNG model predicts n test sets and obtains n prediction results,
for the ith prediction, YiThe actual value of the ith test sample in the n test sets. Smaller MRE values indicate better model prediction.
TABLE 3 RGA-BPNNG model prediction values
TABLE 4 GA-BPNN model prediction values
TABLE 5 statistical indices of prediction models
As can be seen from Table 5, the R of the first and second RGA-BPNNG models2The same RMSE was 0.989 and 0.015, respectively, but the MRE of the first RGA-BPNNG model was 4.040, which is smaller than 4.936 of the second RGA-BPNNG model, so the predicted performance of the first RGA-BPNNG model was the best. The model with the worst prediction performance in the RGA-BPNNG model has better prediction effect than the model with the best prediction performance in the GA-BPNN model. As shown in tables 3 and 4, although the relative error of the prediction result of the RGA-BPNNG model is larger than that of the GA-BPNN model in a certain sample, the prediction effect of the RGA-BPNNG model is better than that of the GA-BPNN model in generalThe predicted effect of (2). In addition, the prediction result of the RGA-BPNNG model is relatively stable and accurate. Therefore, compared with the GA-BPNN model, the RGA-BPNNG model provided by the invention has higher prediction precision and more stable prediction effect, and shows that the RGA-BPNNG model provided by the invention is effective and can predict the explosion blocking degree more accurately.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A blasting blockiness prediction method based on a random GA-BP neural network group is characterized by comprising the following steps:
acquiring blasting parameters of an area to be blasted;
extracting blasting characteristic data based on the blasting parameters;
predicting the blasting average block degree of the area to be blasted according to the blasting characteristic data and a preset random GA-BP neural network group blasting block degree prediction model; the preset random GA-BP neural network group blasting blockiness prediction model is obtained by training a random GA-BP neural network group through historical blasting data of a blasted area.
2. The method of claim 1, wherein the predetermined random GA-BP neural network population blasting blockiness prediction model is obtained by training a random GA-BP neural network population according to historical blasting data of a blasted area, and specifically comprises:
acquiring blasting parameters and blasting average blockiness in historical blasting data of a blasted area;
extracting blasting characteristic data of the blasted area based on the blasting parameters of the blasted area, and then constructing a blasting sample set based on the blasting characteristic data of the blasted area and the corresponding blasting average block size;
extracting N groups of sub training sets and corresponding sub test sets from the blasting sample set by a random sampling method with a release function, wherein N is a preset value;
respectively training the GA-BP neural network by using N groups of sub-training sets by using blasting characteristic data as input and corresponding blasting average block degrees as output to obtain N GA-BP neural network blasting block degree prediction models; wherein, each training obtains a GA-BP neural network blasting block degree prediction model, and the GA-BP neural network blasting block degree prediction model is respectively checked by using a corresponding subtest set; if the prediction error of the GA-BP neural network blasting blockiness prediction model exceeds a preset value, randomly extracting a group of sub-training sets and corresponding sub-test sets from the blasting sample set again, establishing a GA-BP neural network blasting blockiness prediction model by utilizing the group of sub-training sets, and checking the GA-BP neural network blasting blockiness prediction model by using the corresponding sub-test sets until the prediction error of the obtained GA-BP neural network blasting blockiness prediction model does not exceed the preset value;
and calculating the weight corresponding to each of the N GA-BP neural network blasting blockiness prediction models according to the prediction error of each of the N GA-BP neural network blasting blockiness prediction models, and establishing a random GA-BP neural network group blasting blockiness prediction model based on the N GA-BP neural network blasting blockiness prediction models, wherein the blasting average blockiness prediction value output by the random GA-BP neural network group blasting blockiness prediction model is the sum of the products of the output of each of the N GA-BP neural network blasting blockiness prediction models and the weight corresponding to the output.
3. The random GA-BP neural network group-based blasting blockiness prediction method of claim 2, wherein the GA-BP neural network is obtained by optimizing a BP neural network through a genetic algorithm, and specifically comprises:
a1, randomly generating a plurality of groups of weight values and threshold values of the BP neural network, and coding the weight values and the threshold values;
a2, carrying out fitness evaluation on the multiple groups of weights and thresholds respectively;
a3, if one or more groups of evaluation results corresponding to the weight and the threshold value meet the requirements, namely the error of the blasting block degree predicted value output by the BP neural network output layer is within a preset range, selecting a group of weight and threshold value corresponding to the smallest error of the blasting block degree predicted value output by the BP neural network output layer as an optimal weight and an optimal threshold value, and taking the optimal weight and the optimal threshold value as the weight and the threshold value of the BP neural network to obtain the GA-BP neural network; if no evaluation result corresponding to one set of weight and threshold meets the requirement, selecting, crossing and mutating the multiple sets of weight and threshold to generate new multiple sets of weight and threshold, and returning to the step A2.
4. The method of claim 2, wherein the method for predicting the blasting bulkiness based on the random GA-BP neural network population further comprises, after extracting the blasting characteristic data of the blasted area based on the blasting parameters of the blasted area: and carrying out standardized preprocessing on the obtained blasting characteristic data and the blasting average block size.
5. The method of any of claims 1 to 4, wherein the blasting parameters comprise a resistance line, an inter-hole distance, a step height, a plugging length, a blast hole diameter, a specific charge, a rock elastic modulus, and an in-situ rock bulkiness; blasting characteristic data extracted based on the blasting parameters comprise step height to resisting line ratio, hole spacing to resisting line ratio, resisting line to blast hole diameter ratio, blocking length to resisting line ratio, explosive unit consumption, in-situ rock block size and rock elastic modulus.
6. A blasting blockiness prediction device based on a random GA-BP neural network group is characterized by comprising the following components:
the first data acquisition module is used for acquiring blasting parameters of an area to be blasted;
the first data extraction module is used for extracting blasting characteristic data based on the blasting parameters;
the blasting average block degree prediction module is used for predicting the blasting average block degree of the area to be blasted according to the blasting characteristic data and a preset random GA-BP neural network group blasting block degree prediction model; the preset random GA-BP neural network group blasting blockiness prediction model is obtained by training a random GA-BP neural network group through historical blasting data of a blasted area.
7. The apparatus of claim 6, further comprising:
the second data acquisition module is used for acquiring blasting parameters and blasting average blockiness in historical blasting data of a blasted area;
the second data extraction module is used for extracting blasting characteristic data of the blasted area based on the blasting parameters of the blasted area, and then constructing a blasting sample set based on the blasting characteristic data of the blasted area and the corresponding blasting average block size;
the random GA-BP neural network group blasting block degree prediction model generation module is used for generating a random GA-BP neural network group blasting block degree prediction model and comprises the following steps:
the sampling unit is used for extracting N groups of sub training sets and corresponding sub test sets from the blasting sample set by a random sampling method with a release function, wherein N is a preset value;
the GA-BP neural network blasting block degree prediction model generation unit is used for training the GA-BP neural network by using N groups of sub-training sets by taking blasting characteristic data as input and corresponding blasting average block degrees as output to obtain N GA-BP neural network blasting block degree prediction models; wherein, each training obtains a GA-BP neural network blasting block degree prediction model, and the GA-BP neural network blasting block degree prediction model is respectively checked by using a corresponding subtest set; if the prediction error of the GA-BP neural network blasting blockiness prediction model exceeds a preset value, randomly extracting a group of sub-training sets and corresponding sub-test sets from the blasting sample set again, establishing a GA-BP neural network blasting blockiness prediction model by utilizing the group of sub-training sets, and checking the GA-BP neural network blasting blockiness prediction model by using the corresponding sub-test sets until the prediction error of the obtained GA-BP neural network blasting blockiness prediction model does not exceed the preset value;
and the random GA-BP neural network group blasting blockiness prediction model generation unit is used for calculating respective corresponding weights according to respective prediction errors of the N GA-BP neural network blasting blockiness prediction models, establishing the random GA-BP neural network group blasting blockiness prediction model based on the N GA-BP neural network blasting blockiness prediction models, and outputting a blasting average blockiness prediction value which is the sum of products of respective outputs of the N GA-BP neural network blasting blockiness prediction models and the corresponding weights.
8. The apparatus of claim 7, wherein the GA-BP neural network burst block size prediction model generation unit comprises:
the GA-BP neural network generation subunit is used for optimizing the BP neural network through a genetic algorithm to obtain the GA-BP neural network, and the optimization specific process comprises the following steps:
b1, randomly generating a plurality of groups of weight values and threshold values of the BP neural network, and coding the weight values and the threshold values;
b2, carrying out fitness evaluation on the multiple groups of weights and thresholds respectively;
b3, if one or more groups of evaluation results corresponding to the weight and the threshold value meet the requirements, namely the error of the blasting block degree predicted value output by the BP neural network output layer is within a preset range, selecting a group of weight and threshold value corresponding to the smallest error of the blasting block degree predicted value output by the BP neural network output layer as an optimal weight and an optimal threshold value, and taking the optimal weight and the optimal threshold value as the weight and the threshold value of the BP neural network to obtain the GA-BP neural network; if no evaluation result corresponding to one group of weight values and thresholds meets the requirements, selecting, crossing and mutating the multiple groups of weight values and thresholds to generate new multiple groups of weight values and thresholds, and returning to the step B2;
and the GA-BP neural network blasting block degree prediction model generation subunit is used for training the GA-BP neural network by using N groups of sub training sets and establishing N GA-BP neural network blasting block degree prediction models by using blasting characteristic data as input and corresponding blasting average block degrees as output.
9. The random GA-BP neural network population-based blasting bulkiness prediction device of any one of claims 6 to 8, wherein the blasting parameters comprise a resistance line, inter-hole spacing, step height, plug length, hole diameter, specific charge, rock elastic modulus, and in-situ rock bulkiness; blasting characteristic data extracted based on the blasting parameters comprise step height to resisting line ratio, hole spacing to resisting line ratio, resisting line to blast hole diameter ratio, blocking length to resisting line ratio, explosive unit consumption, in-situ rock block size and rock elastic modulus.
10. A computer readable storage medium, characterized in that the storage medium stores program instructions adapted to be loaded by a processor and to execute the method of random GA-BP neural network population-based blasting blockiness prediction according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010045195.7A CN111259601A (en) | 2020-01-16 | 2020-01-16 | Blasting blockiness prediction method, device and medium based on random GA-BP neural network group |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010045195.7A CN111259601A (en) | 2020-01-16 | 2020-01-16 | Blasting blockiness prediction method, device and medium based on random GA-BP neural network group |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111259601A true CN111259601A (en) | 2020-06-09 |
Family
ID=70945167
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010045195.7A Pending CN111259601A (en) | 2020-01-16 | 2020-01-16 | Blasting blockiness prediction method, device and medium based on random GA-BP neural network group |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111259601A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112541392A (en) * | 2020-11-09 | 2021-03-23 | 北方***科技有限公司 | Open bench blasting prediction method based on deep neural network |
| CN112800673A (en) * | 2021-01-27 | 2021-05-14 | 昆明理工大学 | Method for predicting blasting block degree based on SA-GA-BP algorithm |
| CN113340164A (en) * | 2021-06-21 | 2021-09-03 | 中铁十九局集团矿业投资有限公司 | Method for predicting rock blasting block size |
| CN113515891A (en) * | 2021-06-04 | 2021-10-19 | 浙江永联民爆器材有限公司 | Method for predicting and optimizing quality of emulsion explosive |
| CN113553699A (en) * | 2021-06-28 | 2021-10-26 | 中国矿业大学(北京) | Prediction method for average lump degree of ore blasting |
| CN113569487A (en) * | 2021-08-02 | 2021-10-29 | 昆明理工大学 | Method for predicting step blasting throwing effect based on BP neural network |
| WO2022240049A1 (en) * | 2021-05-11 | 2022-11-17 | 주식회사 한화 | Blasting management system for analysis of vibration and fragmentation caused by blasting |
| CN113837440B (en) * | 2021-08-20 | 2023-09-12 | 中国矿业大学(北京) | Blasting effect prediction method and device, electronic equipment and medium |
| CN117113162A (en) * | 2023-05-23 | 2023-11-24 | 南华大学 | Eddar-rock structure background discrimination and graphic method integrating machine learning |
| CN117669393A (en) * | 2024-02-01 | 2024-03-08 | 昆明理工大学 | Blasting block uncertainty prediction method and system |
| CN117852717A (en) * | 2024-01-10 | 2024-04-09 | 中国气象局气象发展与规划院 | Typhoon disaster prediction method for optimizing back propagation neural network based on genetic algorithm |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030191728A1 (en) * | 2002-03-27 | 2003-10-09 | Kulkarni Bhaskar Dattatray | Performance of artificial neural network models in the presence of instrumental noise and measurement errors |
| CN103778469A (en) * | 2013-01-23 | 2014-05-07 | 辽宁工程技术大学 | Blasting scheme selection method based on neural network optimization genetic algorithm |
| CN109102109A (en) * | 2018-07-16 | 2018-12-28 | 四川大学 | A kind of piece of exploitation of stone explosion gradation prediction technique |
-
2020
- 2020-01-16 CN CN202010045195.7A patent/CN111259601A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030191728A1 (en) * | 2002-03-27 | 2003-10-09 | Kulkarni Bhaskar Dattatray | Performance of artificial neural network models in the presence of instrumental noise and measurement errors |
| CN103778469A (en) * | 2013-01-23 | 2014-05-07 | 辽宁工程技术大学 | Blasting scheme selection method based on neural network optimization genetic algorithm |
| CN109102109A (en) * | 2018-07-16 | 2018-12-28 | 四川大学 | A kind of piece of exploitation of stone explosion gradation prediction technique |
Non-Patent Citations (6)
| Title |
|---|
| 史秀志等: "基于PCA-BP神经网络的岩石***平均粒径预测" * |
| 常晓雪;于立涛;撖奥洋;安树怀;张智晟;: "基于Bagging算法和Elman-NN融合的短期负荷预测模型" * |
| 汪志国;唐学义;陈东方;严鹏;陈永长;冷智诚;: "基于神经网络的采场凿岩***参数优化及应用" * |
| 沈掌泉,孔繁胜: "基于广义回归网络的动态权重回归型神经网络集成方法研究" * |
| 沈掌泉等: "基于个体选择的动态权重神经网络集成方法研究" * |
| 王德永等: "基于GA-BP神经网络矿岩***参数优选" * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112541392A (en) * | 2020-11-09 | 2021-03-23 | 北方***科技有限公司 | Open bench blasting prediction method based on deep neural network |
| CN112800673A (en) * | 2021-01-27 | 2021-05-14 | 昆明理工大学 | Method for predicting blasting block degree based on SA-GA-BP algorithm |
| WO2022240049A1 (en) * | 2021-05-11 | 2022-11-17 | 주식회사 한화 | Blasting management system for analysis of vibration and fragmentation caused by blasting |
| CN113515891A (en) * | 2021-06-04 | 2021-10-19 | 浙江永联民爆器材有限公司 | Method for predicting and optimizing quality of emulsion explosive |
| CN113515891B (en) * | 2021-06-04 | 2024-02-20 | 浙江永联民爆器材有限公司 | Emulsion explosive quality prediction and optimization method |
| CN113340164A (en) * | 2021-06-21 | 2021-09-03 | 中铁十九局集团矿业投资有限公司 | Method for predicting rock blasting block size |
| CN113553699A (en) * | 2021-06-28 | 2021-10-26 | 中国矿业大学(北京) | Prediction method for average lump degree of ore blasting |
| CN113553699B (en) * | 2021-06-28 | 2023-06-06 | 中国矿业大学(北京) | Prediction method for average block size of ore blasting |
| CN113569487B (en) * | 2021-08-02 | 2023-08-08 | 昆明理工大学 | BP neural network-based method for predicting step blasting throwing effect |
| CN113569487A (en) * | 2021-08-02 | 2021-10-29 | 昆明理工大学 | Method for predicting step blasting throwing effect based on BP neural network |
| CN113837440B (en) * | 2021-08-20 | 2023-09-12 | 中国矿业大学(北京) | Blasting effect prediction method and device, electronic equipment and medium |
| CN117113162A (en) * | 2023-05-23 | 2023-11-24 | 南华大学 | Eddar-rock structure background discrimination and graphic method integrating machine learning |
| CN117113162B (en) * | 2023-05-23 | 2024-02-02 | 南华大学 | Eddar-rock structure background discrimination and graphic method integrating machine learning |
| CN117852717A (en) * | 2024-01-10 | 2024-04-09 | 中国气象局气象发展与规划院 | Typhoon disaster prediction method for optimizing back propagation neural network based on genetic algorithm |
| CN117669393A (en) * | 2024-02-01 | 2024-03-08 | 昆明理工大学 | Blasting block uncertainty prediction method and system |
| CN117669393B (en) * | 2024-02-01 | 2024-04-19 | 昆明理工大学 | Blasting block uncertainty prediction method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111259601A (en) | Blasting blockiness prediction method, device and medium based on random GA-BP neural network group | |
| US12014282B2 (en) | Data processing method and apparatus, electronic device, and storage medium | |
| KR20210040248A (en) | Generative structure-property inverse computational co-design of materials | |
| CN110298663B (en) | Fraud transaction detection method based on sequence wide and deep learning | |
| CN107169628B (en) | Power distribution network reliability assessment method based on big data mutual information attribute reduction | |
| CN111275252A (en) | Blasting peak value speed prediction method, device and medium based on RGA-BPNNG | |
| Zeng et al. | A GA-based feature selection and parameter optimization for support tucker machine | |
| CN110826618A (en) | Personal credit risk assessment method based on random forest | |
| CN109388565B (en) | Software system performance optimization method based on generating type countermeasure network | |
| CN111105035A (en) | Neural network pruning method based on combination of sparse learning and genetic algorithm | |
| CN112560985A (en) | Neural network searching method and device and electronic equipment | |
| CN115510042A (en) | Power system load data filling method and device based on generation countermeasure network | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| CN111400964B (en) | Fault occurrence time prediction method and device | |
| CN113642727A (en) | Training method of neural network model and processing method and device of multimedia information | |
| CN117407665A (en) | Retired battery time sequence data missing value filling method based on generation countermeasure network | |
| CN117233633A (en) | Battery health degree prediction method based on AGA-Elman algorithm | |
| Zhao et al. | A new feature selection algorithm in text categorization | |
| CN117291069A (en) | LSTM sewage water quality prediction method based on improved DE and attention mechanism | |
| CN116680969A (en) | Filler evaluation parameter prediction method and device for PSO-BP algorithm | |
| CN116628488A (en) | Training method of wind power prediction model, wind power prediction method and device | |
| CN114708639B (en) | FPGA chip for face recognition based on heterogeneous impulse neural network | |
| CN115600926A (en) | Post-project evaluation method and device, electronic device and storage medium | |
| CN116307139A (en) | Wind power ultra-short-term prediction method for optimizing and improving extreme learning machine | |
| US20230334360A1 (en) | Model-independent feature selection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |