CN112711904A - Blasting vibration characteristic parameter prediction method based on SA-GA-BP - Google Patents
Blasting vibration characteristic parameter prediction method based on SA-GA-BP Download PDFInfo
- Publication number
- CN112711904A CN112711904A CN202011490918.0A CN202011490918A CN112711904A CN 112711904 A CN112711904 A CN 112711904A CN 202011490918 A CN202011490918 A CN 202011490918A CN 112711904 A CN112711904 A CN 112711904A
- Authority
- CN
- China
- Prior art keywords
- population
- neural network
- blasting vibration
- blasting
- vibration characteristic
- 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
- 238000005422 blasting Methods 0.000 title claims abstract description 80
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000013528 artificial neural network Methods 0.000 claims abstract description 70
- 238000002922 simulated annealing Methods 0.000 claims abstract description 45
- 230000002068 genetic effect Effects 0.000 claims abstract description 43
- 238000012549 training Methods 0.000 claims abstract description 39
- 239000002360 explosive Substances 0.000 claims abstract description 24
- 238000005457 optimization Methods 0.000 claims abstract description 9
- 230000008569 process Effects 0.000 claims abstract description 9
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 230000035772 mutation Effects 0.000 claims description 13
- 238000004880 explosion Methods 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 8
- 206010064571 Gene mutation Diseases 0.000 claims description 6
- 238000001816 cooling Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 5
- 239000011435 rock Substances 0.000 claims description 3
- 230000006870 function Effects 0.000 description 26
- 108090000623 proteins and genes Proteins 0.000 description 14
- 230000003044 adaptive effect Effects 0.000 description 9
- 238000010606 normalization Methods 0.000 description 6
- 238000010276 construction Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000012886 linear function Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000009412 basement excavation Methods 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 230000000977 initiatory effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Probability & Statistics with Applications (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting blasting vibration characteristic parameters based on an SA-GA-BP algorithm, belonging to the technical field of blasting vibration. According to the method, blasting vibration influence factors in a blasting engineering field are obtained, and then hole depth, filling, chassis resistance line, height difference, blasting source distance, total explosive loading and maximum explosive loading are determined according to a primary analytic hierarchy process and are used as training samples and prediction samples; determining a BP neural network topological structure, calculating an optimal weight and a threshold by using a genetic simulated annealing algorithm (SA-GA), decoding and assigning to a BP neural network system for training; initially constructing a blasting vibration characteristic parameter prediction model; carrying out error analysis on the prediction result; and finally, carrying out field prediction on the blasting vibration characteristic parameters in the blasting engineering field. According to the method, the optimal solution can be searched only by a small amount of samples through the optimization of the improved hybrid intelligent algorithm, the convergence speed is improved, and the situation of falling into the local optimal solution is avoided.
Description
Technical Field
The invention relates to a blasting vibration characteristic parameter prediction method based on SA-GA-BP, and belongs to the technical field of blasting vibration prediction.
Background
Blasting technology is being widely applied to the fields of mining, tunnel excavation, railway and highway cutting formation, construction of water conservancy and hydropower infrastructure, mountain moving and sea filling and other related earth and rockwork engineering. The blasting vibration effect brought by the method causes huge potential safety hazards to surrounding related personnel and facilities, for example, vibration generated during excavation of a subway tunnel disturbs people, and a building is damaged under the action of blasting seismic waves. Therefore, the explosion vibration strength is grasped in advance for the explosion engineering, so that a plurality of unnecessary accidents can be avoided.
The general blasting vibration intensity is a main criterion formed by a particle vibration speed peak value, a vibration frequency and a vibration duration; the traditional prediction method is mainly based on the linear and definite relationship between the blasting vibration influence factors and the characteristic parameters, and in fact, the relationship has obvious nonlinearity and uncertainty. In the prior art, an intelligent optimization algorithm is used for predicting certain nonlinear parameters, but the prediction effect of the intelligent optimization algorithm is often influenced by the initial weight and the randomness of a threshold value of a network system, so that local oscillation is easy to occur and a local optimal solution is easy to fall into.
Disclosure of Invention
The invention provides a blasting vibration characteristic parameter prediction method based on SA-GA-BP (SA-GA-BP), aiming at the technical problems in the prior art.
The technical scheme adopted by the invention for solving the technical problem is as follows:
a blasting vibration characteristic parameter prediction method based on SA-GA-BP comprises the following specific steps:
(1) acquiring blasting vibration influence factors in a blasting engineering site, determining hole depth, stuffing, chassis resistance line, elevation difference, blasting source distance, total explosive loading and maximum explosive loading as input parameters according to a main analytic hierarchy process, constructing a data set as a training sample and a prediction sample by taking blasting vibration peak speed and main frequency as output parameters, and normalizing original data of the data set constructed by the training sample and the prediction sample; wherein the blasting vibration influence factors comprise hole distance, row spacing, aperture, hole depth, stuffing, chassis resistance line, total explosive loading, maximum section explosive loading, unit consumption, elevation difference, blast source distance and rock physical and mechanical properties; the raw data comprises input parameters and output parameters;
(2) determining the number of input layers, output layers and hidden layers according to the input parameters and the output parameters respectively to determine a neural network topological structure, and establishing a BP neural network initial model;
(3) optimizing the BP neural network by the genetic simulated annealing algorithm (comprising the steps (3) to (5)): calculating a BP neural network connection weight and a threshold by using a genetic algorithm, decoding and assigning the weight and the threshold to a BP neural network system, and training the training sample in the step (1) by using the BP neural network system to obtain a new generation of population; in the genetic algorithm, each sample of a training sample is used as a random initial population, coding is carried out, and genetic operations of selection, crossing and variation are carried out to generate a new generation of population;
(4) judging whether the new generation population in the step (3) tends to a stable stage, and if so, optimizing a BP neural network by adopting a simulated annealing algorithm to obtain an optimal solution; if not, returning to the step (3) to train the BP neural network prediction model by taking the new generation population as a training sample, and optimizing the BP neural network connection weight and the threshold by using a Genetic Algorithm (GA);
(5) giving the optimal solution which tends to be stable in the step (4) to a BP neural network system to update the weight and the threshold value, and initially constructing a blasting vibration characteristic parameter prediction model;
(6) inputting the prediction sample obtained in the step (1) into the blasting vibration characteristic parameter prediction model constructed in the step (5), and performing error analysis on an output result; if the error exceeds the set threshold, returning to the step (3) to adjust the genetic iteration times of the Genetic Algorithm (GA) to continuously optimize the obtained connection weight and the threshold, and if the error does not exceed the set threshold, completing the blasting vibration characteristic parameter prediction model of the engineering site;
further, the threshold value initially set in the step (6) is not more than 10%;
(7) and (4) predicting the blasting vibration characteristic parameters of the engineering site by using the blasting vibration characteristic parameter prediction model completed in the step (6).
The blasting vibration influence factor of the step (1) comprises: hole distance, row spacing, aperture, hole depth, filling, chassis resistance line, total explosive loading, maximum section explosive loading, unit consumption, elevation difference, explosive source distance and rock physical and mechanical properties;
the step (2) of inputting parameters comprises the following steps: hole depth, filling, chassis resistance line, elevation difference, explosive source distance, total explosive loading, maximum explosive loading, and output parameters comprise: blasting vibration peak speed and main frequency of the blasting vibration characteristic parameter;
the preliminary construction of the BP neural network model optimized by the genetic simulated annealing algorithm comprises the following specific steps:
1) forming a seven-dimensional vector by using data of hole depth, stuffing, chassis resistance line, height difference, explosion source distance, total charge amount and maximum section charge amount of a new generation of population, randomly generating an initial population according to an individual vector and coding individuals of the initial population, wherein a fitness function of the initial population is constructed according to network errors because a simulated annealing algorithm is mainly used for optimizing a neural network system connection weight and a threshold;
2) calculating the fitness value of the population according to the fitness function relation, and arranging the individuals of the new generation of population according to the sequence of the fitness values from small to large; the smaller the fitness value, the better the individual gene;
3) averagely dividing the new generation population into three grades according to the size of the fitness value, namely dividing the new generation population into three grades of good grade, medium grade and poor grade, wherein the smaller the fitness value is, the better the grade is; randomly sampling and selecting individuals in three grades of good, medium and poor according to the ratio of 5:3:2 of good, medium and poor respectively to form a new population A;
4) carrying out self-adaptive cross and variation treatment on the population A: with probability P1Recombining the individuals to generate a new individual a so as to ensure good genes of the offspring; probability of re-passing P2Carrying out mutation processing on the new individual a to obtain a new individual b so as to continue the diversity of the population, and carrying out adaptive probability P1Eliminating gene mutation individuals so as to prevent the excellent individuals from generating gene mutation to influence offspring individuals to obtain a population B;
5) judging the stability of the population B, and if the population is stable, entering the optimization operation of the step 6); if the population is unstable, returning to the step 3) to re-sample and select individuals and execute genetic operation;
6) taking the optimal individual in the stable population as the initial solution S of a Simulated Annealing (SA) algorithm0And generating a new solution S using the state function1;
7) Taking the fitness function as a target evaluation function of the simulated annealing operation, and carrying out S) on the initial solution of the step 6)0And new solution S1Evaluating, and selecting whether to accept new solution according to Metropolis principle, if soIf yes, outputting an optimal solution; if not, cooling according to the temperature decay function, returning to the step 6) to regenerate the new solution S1;
8) Decoding the optimal solution obtained in the step 7), and updating the BP neural network connection weight and the threshold to obtain an optimal weight threshold prediction model;
the genetic algorithm mainly improves the aspects of coding mode, operator selection, crossover operator, mutation operator and the like, and the simulated annealing algorithm mainly improves the aspects of initial temperature selection, temperature reduction attenuation function, algorithm termination selection, target error function introduction and the like; a BP neural network connection weight and a threshold are optimized by improving a genetic simulated annealing algorithm, so that the BP neural network connection weight and the threshold are infinitely approximated to a target function to be obtained, and the purpose of reducing system errors is achieved.
The invention has the beneficial effects that:
(1) the invention provides a blasting vibration characteristic parameter prediction method for optimizing a BP neural network based on an improved genetic simulated annealing algorithm, which is characterized in that on the basis of a BP neural network system, the improved genetic simulated annealing algorithm is utilized to optimize a connection weight and a threshold value in the BP neural network system, so that the connection weight and the threshold value are infinitely approximated to a highly nonlinear mapping relation between an input factor and an output factor;
(2) the invention provides a blasting vibration characteristic parameter prediction method for optimizing a BP neural network based on an improved genetic simulated annealing algorithm, which is characterized in that aiming at each factor in a blasting engineering field, a plurality of blasting parameters influencing the vibration speed are selected by a main analytic hierarchy process as influencing factors, and a blasting vibration characteristic parameter prediction model aiming at a specific engineering is constructed by combining blasting data samples of the engineering;
(3) based on the nonlinear characteristic of the BP neural network, the method can be well suitable for predicting the explosion vibration characteristic parameters with very complex action mechanisms, the prediction precision is higher than that of a Sagnac formula and various improved formulas thereof, and the simulation is closer to the real situation of the site;
(4) the prediction method of the improved hybrid intelligent algorithm overcomes the difficulties that the number of initial samples required by the traditional neural network prediction is more and the training precision is high, can search the optimal solution only by optimizing a small number of samples through the improved hybrid intelligent algorithm, improves the convergence speed and avoids the situation of trapping in the local optimal solution.
Drawings
FIG. 1 is a flow chart of a method for predicting blasting vibration characteristic parameters based on SA-GA-BP;
FIG. 2 is a process for optimizing BP neural network parameters by improving a genetic simulated annealing algorithm.
Detailed Description
The present invention will be further described with reference to the following embodiments.
Example 1: as shown in fig. 1, a method for predicting blasting vibration characteristic parameters based on SA-GA-BP specifically comprises the following steps:
(1) acquiring blasting vibration influence factors in a blasting engineering site: the method comprises the steps of determining hole depth, filling, chassis resistance line, elevation difference, total explosive loading and maximum explosive loading as input parameters according to a main level analysis method, and constructing a data set as a training sample and a prediction sample by taking blasting vibration peak speed and main frequency as output parameters; because the physical dimension is different, the original data of the data set constructed by the training sample and the prediction sample are unified to be normalized; wherein the raw data comprises input parameters and output parameters;
(2) determining the number of input layers, output layers and hidden layers according to the input parameters and the output parameters respectively to determine a neural network topological structure, and establishing a BP neural network initial model;
(3) optimizing the BP neural network by the genetic simulated annealing algorithm (comprising the steps (3) to (5)): training the BP neural network prediction model by using a training sample, calculating a BP neural network connection weight and a threshold by using a Genetic Algorithm (GA), decoding and assigning to a BP neural network system, and training the training sample in the step (1) by using the BP neural network system to obtain a new generation population; in the genetic algorithm, each sample of a training sample is used as a random initial population, coding is carried out, and genetic operations of selection, crossing and variation are carried out to generate a new generation of population;
(4) judging whether the new generation population in the step (3) tends to a stable stage, and if so, optimizing the BP neural network by adopting a simulated annealing algorithm to obtain an optimal solution; if not, returning to the step (3) to train the BP neural network prediction model by taking the new generation population as a training sample, and optimizing the BP neural network connection weight and the threshold by using a Genetic Algorithm (GA);
(5) giving the optimal solution which tends to be stable and optimized in the step (4) to a BP neural network system to update the weight and the threshold value, and initially constructing a blasting vibration characteristic parameter prediction model;
the optimization of BP neural network by the improved genetic simulated annealing algorithm comprises the following specific steps (see figure 2)
1) Forming a seven-dimensional vector by using data of hole depth, stuffing, chassis resistance line, height difference, explosion source distance, total charge amount and maximum section charge amount of a new generation of population, randomly generating an initial population according to an individual vector and coding individuals of the initial population, wherein a fitness function is constructed according to network errors as a simulated annealing algorithm is mainly used for optimizing a neural network system connection weight and a threshold value, and the fitness function is as shown in a formula (1);
where y (k) represents the k sample actual output value and y' (k) represents the k sample net output value;
2) calculating the fitness value of the population according to the fitness function relation (1), and arranging the individuals of the new generation of population according to the sequence of the fitness values from small to large; the smaller the fitness value is, the better the individual gene is, so the individual sequence of the new population is arranged from good to bad;
3) evenly dividing the new generation population into 3 levels of good and medium differences according to the size of the fitness value, randomly sampling and selecting individuals in each level, wherein the random sampling quantity is 5:3:2 according to the ratio of good to medium differences to form a new population A;
4) carrying out self-adaptive cross and variation treatment on the population A: with probability P1Recombining the individuals to generate a new individual A so as to ensure good genes of the offspring;
wherein, P10And P11For greater and lesser initial crossover probabilities, fmin、fmaxF is the minimum adaptive value of the current population, the average adaptive value and the smaller adaptive value in the crossed individuals;
the method for pairwise cross variation in the population comprises
Wherein x1,x2For individuals selected before crossover, x1′,x′2Is the individual generated after crossing, c is [0, 1 ]]A random number of;
adaptive mutation operation probability P2Determining a gene to be mutated of an individual by adopting a random sampling method when the individual is selected, and then carrying out mutation on the gene by adopting a non-uniform mutation operator in the following mutation mode;
wherein: x' is a gene xiGenes, G, produced after mutationmaxThe total iteration times in the current evolutionary algebra and algorithm; c. C1Is a random number on (0,1), ai,biThe maximum value and the minimum value in the value range of the variation gene are obtained;
probability of re-passing P2Carrying out mutation processing on the new individual A to obtain a new individual B so as to extend the diversity of the population through the self-adaptive probability P1Get rid of the mutated individuals of the geneSo as to prevent the excellent individuals from generating gene mutation to influence the offspring individuals, and obtain a population B;
5) judging the stability of the population B, and if the population is stable, entering the simulated annealing operation of the step 6); if the population is unstable, returning to the step 3) to re-sample and select individuals and execute genetic operation;
the stability determination method comprises
Wherein: f. ofmaxIs the mean fitness in the population, fminFor optimal individual fitness values in the population, when the average individual fitness value is approximately equal to the optimal fitness value (i.e., equation (5) is satisfied, whereA preset value), at which time, population evolution tends to a stable stage;
6) taking the optimal individual in the stable population as the initial solution S of a Simulated Annealing (SA) algorithm0And generating a new solution S using the state function1;
The state function is as follows:
S1=S0+α/20 (6)
wherein alpha is a random number distributed uniformly in [ -0.5,0.5]
7) Adopting the fitness function F to S in the step 1)0And S1Starting to evaluate, and judging new solution S according to Metropolis principle1If yes, the acceptance probability Q is defined as follows:
T0is the initial set temperature;
8) judging whether a Simulated Annealing (SA) algorithm end condition is met, and if so, outputting an optimal solution; if not, cooling according to the temperature attenuation function, and then turning to the step 6) to continue to execute the simulated annealing operation until the SA termination condition is met;
temperature decay function selection linear function
Tk+1=kt×Tk (8)
Wherein k istTaking 0.9 as the cooling rate;
9) decoding the optimal solution obtained in the step 8), and updating the BP neural network connection weight and the threshold to obtain an optimal weight threshold prediction model;
(6) inputting the prediction sample obtained in the step (1) into the optimal weight threshold prediction model (blasting vibration characteristic parameter prediction model) obtained in the step (5), and carrying out error analysis on an output result; if the error exceeds the initially set threshold (the initially set threshold is not more than 10%), returning to the step (3), adjusting the iteration times of the Genetic Algorithm (GA), and if the error does not exceed the initially set threshold, completing the construction process of the blasting vibration characteristic parameter prediction model of the engineering site, namely the optimal weight threshold prediction model of the engineering site;
(7) and (4) predicting the blasting vibration characteristic parameters of the engineering site by using the blasting vibration characteristic parameter prediction model completed in the step (6).
Example 2: the embodiment further describes the technical solution provided by the present invention with reference to specific examples;
in the example, a blasting vibration test of a slope treatment project of a certain strip mine is taken as a research object, and the project is located in the red river state of Yunnan province; the charge mode of the project adopts a coupling continuous charge structure, the aperture is 120mm, the row spacing is 3.5m, the hole spacing is 5.5m, the chassis resistance line is 3-5m, the step height is 8-15m, and the initiation network selects an electronic detonator-detonating tube mixed initiation network;
in order to predict the peak velocity and the dominant frequency of blasting vibration, a TC-4850 blasting vibration recorder and a matched velocity sensor are used for monitoring blasting events in the area, in addition, a GPS positioning system is also used for respectively measuring the positions of the instruments, and the horizontal distance and the vertical distance between an explosion point and a monitoring point, namely, the distance between an explosion source and the height difference, are respectively measured and calculated after blasting;
a blasting vibration characteristic parameter prediction method (shown in figure 1) based on SA-GA-BP comprises the following specific steps:
(1) acquiring blasting vibration influence factors in a blasting engineering site: the method comprises the steps of determining hole depth, filling, chassis resistance line, elevation difference, total explosive loading and maximum explosive loading as input parameters according to a main level analysis method, and constructing a data set as a training sample and a prediction sample by taking blasting vibration peak speed and main frequency as output parameters; because the physical dimension is different, the original data of the data set constructed by the training sample and the prediction sample are unified to be normalized; wherein the raw data comprises input parameters and output parameters;
in the embodiment, before the training operation of the neural network prediction model of the genetic simulated annealing algorithm is executed, in order to prevent errors of the results caused by dimensions of different input parameters and reduce the influence of the errors of the maximum value and the minimum value of the data on the whole training sample, the data is preprocessed by normalization;
carrying out normalization pretreatment on the data, wherein the normalization process formula is as follows:
in the formula: x is the data set before preprocessing, Xmax,XminThe maximum value and the minimum value of X are obtained, and Y is a normalized data set; in this example, x and y can be expressed as any one data set of hole depth, filling height, chassis resistance line, height difference, explosive source distance, total explosive loading, maximum explosive loading, and peak velocity and main frequency of explosive vibration;
the preprocessing program can be realized by MATLAB, wherein normalization can be converted by a MATLAB linear function; carrying out inverse normalization processing on the output data, wherein the inverse normalization processing is realized by adopting a command POSIMNMX;
table 1 gives 40 sets of test set data, which are randomly divided into training samples and prediction samples in a 1:1 ratio;
TABLE 1 blasting vibration monitoring data of 40 groups for side treatment of certain strip mine
In the embodiment, a matlab neural network tool box is used for prediction, and when training samples are completely trained and controlled within error precision, the optimal weight threshold model is established; (2) determining the number of input layers, output layers and hidden layers according to the input parameters and the output parameters respectively to determine a neural network topological structure, and establishing a BP neural network initial model;
because the input parameters adopt 7 indexes and the output parameters adopt 2 indexes, the input layer and the output layer of the BP neural network respectively adopt 7 layers and 2 layers, but the hidden layer is not determined by a proper formula at present, the number of the hidden layers is obtained according to an empirical formula in the common method, in order to reduce the iteration times and the operation time cost of the whole model in the operation process and increase the prediction precision, the hidden layer is optimized by a Genetic Algorithm (GA), the prediction precision is highest when the hidden layer is 13, and therefore 13 is selected as the number of the hidden layers, the topological structure of the BP neural network is determined to be 7-13-2;
and selecting a Sigmoid function for the excitation function of the hidden layer node, and selecting a linear transmission function for the excitation function of the output layer node. Meanwhile, the maximum training frequency is set to be 1000, the relative MSE target value is 0.001, and the learning rate is 0.1;
in the Genetic Algorithm (GA), because data are preprocessed, the value range of an initial population is set to be (-3,3), the size of the population is 100, the maximum algebra is 200, the cross probability is 0.5, and the variation probability is 0.2;
(3) optimizing the BP neural network by the genetic simulated annealing algorithm (comprising the steps (3) to (5)): training the BP neural network prediction model by using a training sample, calculating a BP neural network connection weight and a threshold by using a Genetic Algorithm (GA), decoding and assigning to a BP neural network system, and training the training sample in the step (1) by using the BP neural network system to obtain a new generation population; in the genetic algorithm, each sample of a training sample is used as a random initial population, coding is carried out, and genetic operations of selection, crossing and variation are carried out to generate a new generation of population;
(4) judging whether the new generation population in the step (3) tends to a stable stage, and if so, optimizing the BP neural network by adopting a simulated annealing algorithm to obtain an optimal solution; if not, returning to the step (3) to train the BP neural network prediction model by taking the new generation population as a training sample, and optimizing the BP neural network connection weight and the threshold by using a Genetic Algorithm (GA);
(5) giving the optimal solution which tends to be stable and optimized in the step (4) to a BP neural network system to update the weight and the threshold value, and initially constructing a blasting vibration characteristic parameter prediction model;
the optimization of the BP neural network by the improved genetic simulated annealing algorithm comprises the following specific steps (see figure 2),
1) forming a seven-dimensional vector by using data of hole depth, stuffing, chassis resistance line, height difference, explosion source distance, total charge amount and maximum section charge amount of a new generation of population, randomly generating an initial population according to an individual vector and coding individuals of the initial population, wherein a fitness function is constructed according to network errors as a simulated annealing algorithm is mainly used for optimizing a neural network system connection weight and a threshold value, and the fitness function is as shown in a formula (1);
where y (k) represents the k sample actual output value and y' (k) represents the k sample net output value;
2) calculating the fitness value of the population according to the fitness function relation (1), and arranging the individuals of the new generation of population according to the sequence of the fitness values from small to large; the smaller the fitness value is, the better the individual gene is, so the individual sequence of the new population is arranged from good to bad;
3) evenly dividing the new generation population into 3 levels of good and medium differences according to the size of the fitness value, randomly sampling and selecting individuals in each level, wherein the random sampling quantity is 5:3:2 according to the ratio of good to medium differences to form a new population A;
4) carrying out self-adaptive cross and variation treatment on the population A: with probability P1Recombining the individuals to generate a new individual A so as to ensure good genes of the offspring;
wherein, P10And P11For greater and lesser initial crossover probabilities, fmin、fmaxF is the minimum adaptive value of the current population, the average adaptive value and the smaller adaptive value in the crossed individuals;
the method for pairwise cross variation in the population comprises
Wherein x1,x2For individuals who are cross-forward selected, x'1,x′2Is the individual generated after crossing, c is [0, 1 ]]A random number of;
adaptive mutation operation probability P2Determining a gene to be mutated of an individual by adopting a random sampling method when the individual is selected, and then carrying out mutation on the gene by adopting a non-uniform mutation operator in the following mutation mode;
wherein: x' is a gene xiGenes, G, produced after mutationmaxFor summation in the current evolutionary algebra and algorithmGeneration times; c. C1Is a random number on (0,1), ai,biThe maximum value and the minimum value in the value range of the variation gene are obtained;
probability of re-passing P2Carrying out mutation processing on the new individual A to obtain a new individual B so as to extend the diversity of the population through the self-adaptive probability P1Eliminating gene mutation individuals so as to prevent gene mutation of individuals with excellent performance from affecting offspring individuals to obtain a population B;
5) judging the stability of the population B, and if the population is stable, entering the simulated annealing operation of the step 6); if the population is unstable, returning to the step 3) to re-sample and select individuals and execute genetic operation;
the stability determination method comprises
Wherein: f. ofmaxIs the mean fitness in the population, fminFor optimal individual fitness values in the population, when the average individual fitness value is approximately equal to the optimal fitness value (i.e., equation (5) is satisfied, whereA preset value), at which time, population evolution tends to a stable stage;
6) taking the optimal individual in the stable population as the initial solution S of a Simulated Annealing (SA) algorithm0And generating a new solution S using the state function1;
The state function is as follows:
S1=S0+α/20 (6)
wherein alpha is a random number distributed uniformly in [ -0.5,0.5]
7) Adopting the fitness function F to S in the step 1)0And S1Starting to evaluate, and judging new solution S according to Metropolis principle1If yes, the acceptance probability Q is defined as follows:
T0is the initial set temperature;
8) judging whether a Simulated Annealing (SA) algorithm end condition is met, and if so, outputting an optimal solution; if not, cooling according to the temperature attenuation function, and then turning to the step 6) to continue to execute the simulated annealing operation until the SA termination condition is met;
temperature decay function selection linear function
Tk+1=kt×Tk (8)
Wherein k istTaking 0.9 as the cooling rate;
9) decoding the optimal solution obtained in the step 8), and updating the BP neural network connection weight and the threshold to obtain an optimal weight threshold prediction model;
(6) inputting the prediction sample obtained in the step (1) into the optimal weight threshold prediction model (blasting vibration characteristic parameter prediction model) obtained in the step (5), and carrying out error analysis on an output result; if the error exceeds the initially set threshold (the initially set threshold is not more than 10%), returning to the step (3), adjusting the iteration times of the Genetic Algorithm (GA), and if the error does not exceed the initially set threshold, completing the construction process of the blasting vibration characteristic parameter prediction model of the engineering site, namely the optimal weight threshold prediction model of the engineering site;
in order to evaluate the accuracy of the prediction precision of the GA-SA-BP model, the GA-BP model is constructed by using the same training sample and the parameters of the prediction model to predict the prediction sample, the result is compared with the GA-SA-BP model, the relative error is used as the evaluation standard,
the predicted values and relative errors of the GA-BP model and the GA-SA-BP model are shown in Table 2,
TABLE 2 prediction values of GA-BP model and GA-SA-BP model
Evaluation indexes of the GA-BP model and the GA-SA-BP model are shown in Table 3
TABLE 3 evaluation indexes of GA-BP model and GA-SA-BP model
As can be seen from Table 3, the average relative errors of the peak vibration speed and the main frequency of the SA-GA-BP prediction model are smaller than those of the GA-BP prediction model, and the superiority of the prediction performance of the SA-GA-BP prediction model is reflected; from Table 2, it can be seen that although the relative error of each sample in the SA-GA-BP model is larger than that of the GA-BP model, the prediction effect of the SA-GA-BP model is not affected overall, and the prediction result is relatively stable and accurate. Compared with a GA-BP model, the SA-GA-BP model provided by the invention has higher reliability, higher prediction precision and more stable prediction result, and shows that the SA-GA-BP prediction model can more effectively and accurately predict the burst vibration characteristic parameters;
(7) and (4) predicting the blasting vibration characteristic parameters of the engineering site by using the blasting vibration characteristic parameter prediction model completed in the step (6).
The invention optimizes BP neural network parameters based on a genetic simulated annealing algorithm; and comprehensively considering the engineering site, and combining the blasting data sample of the engineering site to construct a blasting vibration characteristic parameter prediction model of the specific engineering. The invention solves the problem that the traditional algorithm needs a large number of training samples during prediction, can improve the prediction precision by only a small number of samples, has high convergence speed and can effectively avoid the problem of falling into the local optimal solution. The method is simple and convenient to operate and strong in overall search capability; and the prediction result can be expressed visually, and is simple and easy to understand.
Although the present invention has been described in detail with reference to the specific embodiments, the present invention is not limited to the embodiments described above, and various changes and modifications can be made without departing from the spirit and scope of the present invention by those skilled in the art. It should be understood that any modification, replacement, or improvement made within the spirit and scope of the present invention should be included in the following claims.
Claims (4)
1. A blasting vibration characteristic parameter prediction method based on SA-GA-BP is characterized by comprising the following specific steps:
(1) acquiring blasting vibration influence factors in a blasting engineering site, determining hole depth, stuffing, chassis resistance line, elevation difference, blasting source distance, total explosive loading and maximum explosive loading as input parameters according to a main analytic hierarchy process, constructing a data set as a training sample and a prediction sample by taking blasting vibration peak speed and main frequency as output parameters, and normalizing original data of the data set constructed by the training sample and the prediction sample; wherein the raw data comprises input parameters and output parameters;
(2) respectively determining the number of an input layer, an output layer and a hidden layer according to the input parameters and the output parameters, determining a topological structure of the BP neural network according to the number of the input layer, the output layer and the hidden layer, and establishing an initial model of the BP neural network;
(3) optimizing a BP neural network by a genetic simulated annealing algorithm: training the BP neural network prediction model by using a training sample, calculating a BP neural network connection weight and a threshold by using a Genetic Algorithm (GA), decoding and assigning to a BP neural network system, and training the training sample in the step (1) by using the BP neural network system to obtain a new generation population;
(4) judging whether the new generation population in the step (3) tends to a stable stage, and if so, optimizing a BP neural network by adopting a Simulated Annealing (SA) algorithm to obtain an optimal solution; if not, returning to the step (3) to train the BP neural network prediction model by taking the new generation population as a training sample, and optimizing the BP neural network connection weight and the threshold by using a Genetic Algorithm (GA);
(5) giving the optimal solution which tends to be stable and optimized in the step (4) to a BP neural network system to update the weight and the threshold value, and initially constructing a blasting vibration characteristic parameter prediction model;
(6) inputting the prediction sample obtained in the step (1) into the blasting vibration characteristic parameter prediction model constructed in the step (5), and performing error analysis on an output result; if the error exceeds the set threshold, returning to the step (3) to adjust the genetic iteration times of the Genetic Algorithm (GA) to continuously optimize the obtained connection weight and the threshold, and if the error does not exceed the set threshold, completing the blasting vibration characteristic parameter prediction model of the engineering site;
(7) and (4) predicting the blasting vibration characteristic parameters of the engineering site by using the blasting vibration characteristic parameter prediction model completed in the step (6).
2. The blasting vibration characteristic parameter prediction method based on the SA-GA-BP optimization BP neural network according to claim 1, characterized in that: the blasting vibration influence factors in the step (1) comprise hole distance, row spacing, aperture, hole depth, stuffing, chassis resistance wire, total explosive loading, maximum section explosive loading, unit consumption, height difference, blasting source distance and rock physical and mechanical properties.
3. The method for predicting the blasting vibration characteristic parameter based on the SA-GA-BP as claimed in claim 1, wherein: the optimization of the BP neural network by the genetic simulated annealing algorithm comprises the following specific steps:
1) forming a seven-dimensional vector by using data of hole depth, filling, chassis resistance line, height difference, explosion source distance, total charge amount and maximum charge amount of a new generation of population, randomly generating an initial population according to an individual vector, coding individuals of the initial population, and constructing a fitness function of the initial population according to network errors;
2) calculating the fitness value of the population according to the fitness function relation, and arranging the individuals of the new generation of population according to the sequence of the fitness values from small to large;
3) evenly dividing the new generation population into three levels according to the size of the fitness value, namely dividing the new generation population into three levels of good, medium and poor, and then randomly sampling and selecting individuals in the three levels of good, medium and poor according to the proportion of 5:3:2 to form a new population A;
4) carrying out self-adaptive cross and variation treatment on the population A: with probability P1Recombining the individuals to generate a new individual a by probabilityP2Carrying out mutation processing on the new individual a to obtain a new individual b so as to pass through the self-adaptive probability P1Eliminating the individual with gene mutation to obtain population B;
5) judging the stability of the population B, and if the population is stable, entering the optimization operation of the step 6); if the population is unstable, returning to the step 3) to re-sample and select individuals and execute genetic operation;
6) taking the optimal individual in the stable population as the initial solution S of a Simulated Annealing (SA) algorithm0And generating a new solution S using the state function1;
7) Taking the fitness function as a target evaluation function of the simulated annealing operation, and carrying out S) on the initial solution of the step 6)0And new solution S1Evaluating, selecting whether to accept a new solution according to a Metropolis principle, and if so, outputting an optimal solution; if not, cooling according to the temperature decay function, returning to the step 6) to regenerate the new solution S1;
8) Decoding the optimal solution obtained in the step 7), and updating the BP neural network connection weight and the threshold to obtain an optimal weight threshold prediction model.
4. The SA-GA-BP based blasting vibration characteristic parameter prediction method according to claim 1, wherein: and (4) the threshold initially set in the step (6) is not more than 10%.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011490918.0A CN112711904A (en) | 2020-12-17 | 2020-12-17 | Blasting vibration characteristic parameter prediction method based on SA-GA-BP |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011490918.0A CN112711904A (en) | 2020-12-17 | 2020-12-17 | Blasting vibration characteristic parameter prediction method based on SA-GA-BP |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112711904A true CN112711904A (en) | 2021-04-27 |
Family
ID=75545103
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011490918.0A Pending CN112711904A (en) | 2020-12-17 | 2020-12-17 | Blasting vibration characteristic parameter prediction method based on SA-GA-BP |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112711904A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113591321A (en) * | 2021-08-09 | 2021-11-02 | 中钢集团马鞍山矿山研究总院股份有限公司 | Method for generating three-dimensional high-simulation structural model of mine soil-rock mixture |
CN113987033A (en) * | 2021-12-28 | 2022-01-28 | 国网江西省电力有限公司电力科学研究院 | Main transformer online monitoring data group deviation identification and calibration method |
CN114781040A (en) * | 2022-05-10 | 2022-07-22 | 青岛理工大学 | Building blasting vibration safety evaluation method, device, equipment and medium |
CN115630257A (en) * | 2022-12-19 | 2023-01-20 | 中南大学 | Blasting funnel volume prediction method |
CN116992770A (en) * | 2023-08-07 | 2023-11-03 | 中国铁建大桥工程局集团有限公司 | Wall protection control blasting method based on GOA-DBN neural network |
CN117669393A (en) * | 2024-02-01 | 2024-03-08 | 昆明理工大学 | Blasting block uncertainty prediction method and system |
CN117949798A (en) * | 2024-01-31 | 2024-04-30 | 中山大学 | Multi-parameter characterization device and method for dynamic characteristics of power field effect transistor |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111275252A (en) * | 2020-01-16 | 2020-06-12 | 南华大学 | Blasting peak value speed prediction method, device and medium based on RGA-BPNNG |
-
2020
- 2020-12-17 CN CN202011490918.0A patent/CN112711904A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111275252A (en) * | 2020-01-16 | 2020-06-12 | 南华大学 | Blasting peak value speed prediction method, device and medium based on RGA-BPNNG |
Non-Patent Citations (2)
Title |
---|
李明洲 等: "基于SA-GA-BP的某***扫雷器电液伺服***建模", 《兵工自动化》 * |
高富强 等: "***震动速度峰值预测模型的比较和讨论", 《***》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113591321A (en) * | 2021-08-09 | 2021-11-02 | 中钢集团马鞍山矿山研究总院股份有限公司 | Method for generating three-dimensional high-simulation structural model of mine soil-rock mixture |
CN113591321B (en) * | 2021-08-09 | 2023-10-03 | 中钢集团马鞍山矿山研究总院股份有限公司 | Method for generating three-dimensional high-simulation structural model of mine earth-rock mixture |
CN113987033A (en) * | 2021-12-28 | 2022-01-28 | 国网江西省电力有限公司电力科学研究院 | Main transformer online monitoring data group deviation identification and calibration method |
CN114781040A (en) * | 2022-05-10 | 2022-07-22 | 青岛理工大学 | Building blasting vibration safety evaluation method, device, equipment and medium |
CN115630257A (en) * | 2022-12-19 | 2023-01-20 | 中南大学 | Blasting funnel volume prediction method |
CN116992770A (en) * | 2023-08-07 | 2023-11-03 | 中国铁建大桥工程局集团有限公司 | Wall protection control blasting method based on GOA-DBN neural network |
CN116992770B (en) * | 2023-08-07 | 2024-03-22 | 中国铁建大桥工程局集团有限公司 | Wall protection control blasting method based on GOA-DBN neural network |
CN117949798A (en) * | 2024-01-31 | 2024-04-30 | 中山大学 | Multi-parameter characterization device and method for dynamic characteristics of power field effect transistor |
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 |
---|---|---|
CN112711904A (en) | Blasting vibration characteristic parameter prediction method based on SA-GA-BP | |
CN110298503B (en) | Tunnel rock burst early warning method based on microseismic information and deep convolutional neural network | |
CN105260599A (en) | Rockburst dynamic prediction method based on BP neural network modeling | |
CN109934398A (en) | A kind of drill bursting construction tunnel gas danger classes prediction technique and device | |
CN108985340A (en) | Shield construction ground predicting subsidence method based on dual model fusion | |
CN110348046B (en) | Drilling track multi-target optimization method based on fuzzy comprehensive evaluation | |
CN103711523A (en) | Method for predicating gas concentration in real time based on local decomposition-evolution neural network | |
Ning et al. | GA-BP air quality evaluation method based on fuzzy theory. | |
CN112100927A (en) | Prediction method for slope deformation and soft soil foundation settlement based on GA-BP neural network | |
CN114723095A (en) | Missing well logging curve prediction method and device | |
CN115130375A (en) | Rock burst intensity prediction method | |
Raji et al. | A new approach in well placement optimization using metaheuristic algorithms | |
CN112800673A (en) | Method for predicting blasting block degree based on SA-GA-BP algorithm | |
CN115980826A (en) | Rock burst intensity prediction method based on weighted meta-heuristic combined model | |
CN104732067A (en) | Industrial process modeling forecasting method oriented at flow object | |
CN110441815B (en) | Simulated annealing Rayleigh wave inversion method based on differential evolution and block coordinate descent | |
CN111950203A (en) | Blasting vibration speed prediction method based on adaptive neural fuzzy inference system | |
Xu et al. | Stability Analysis of Geotechnical Landslide Based on GA‐BP Neural Network Model | |
CN114819178A (en) | Railway construction progress index prediction and online updating method | |
CN117517061A (en) | Fine intelligent sensing method and system for strength characteristics of all-domain rock of tunnel face | |
Zhu | Quantitative evaluation of coal-mining geological condition | |
CN106503793A (en) | A kind of neural network short-term wind speed forecasting method based on improvement difference algorithm | |
CN115962692A (en) | Blasting parameter optimization method applied to open blasting engineering | |
Chatterjee et al. | Genetic algorithm-based neural network learning parameter selection for ore grade evaluation of limestone deposit | |
Yuan et al. | Peak particle velocity and principal frequency prediction based on RS-FNN comprehension method for blasting vibration |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210427 |