CN115391938B - A failure mode identification method for full-length bonded bolts under shear action - Google Patents

Abstract

This application proposes a method for identifying the failure mode of a full-length bonded anchor under shearing. Aiming at the problem that the failure mode of the anchor is not clearly defined when the anchor is sheared under the influence of different factors, based on the combination of material mechanics Deformation theory and strength theory, embed the judgment process of pure shear failure, tensile shear failure and tensile bending failure of anchor rods into the main program of UDEC calculation and obtain training data sets, and establish shearing effect through random forest classification algorithm and particle swarm optimization algorithm The mapping relationship between the failure mode of the lower bolt and each characteristic parameter can be used to effectively identify the failure mode of the bolt when it is sheared. The method proposed in this application can accurately determine the failure mode of the full-length bonded anchor according to the provided rock mass, structural surface and anchor parameters, and provide a reference for the evaluation of the shear performance of the anchor. The optimization of anchor support parameters is of great significance.

Description

A failure mode identification method for full-length bonded bolts under shear action

technical field

The application belongs to the technical field of slope and roadway support, and in particular relates to a failure mode identification method of a full-length bonded anchor rod under shearing action.

Background technique

The full-length bonded anchor has been widely used in slopes and underground chambers due to its simple structure, convenient construction, and economical advantages. Due to the existence of a large number of structural surfaces such as joints and fissures in the natural rock mass, under the action of artificial disturbance or in-situ stress, the local rock mass undergoes large shear deformation along the structural surfaces, which leads to the failure of the full-length bonded anchor. It is usually not pure tensile failure or mortar debonding failure in the traditional sense, but a combined failure under the combined action of shear force, bending moment and axial force.

The failure mode of the bolt is directly related to the rock, structural surface and anchoring parameters, but the current definition of the failure mode of the bolt under different characteristic parameter values is not clear. Since different failure modes have a significant impact on the resistance of the anchor when it yields or fails, establishing the mapping relationship between the failure mode of the anchor and each characteristic parameter can effectively predict the failure mode of the anchor, which is very important for evaluating the shear performance of the anchor. , And then guide the rock mass anchoring engineering practice has important practical significance.

Contents of the invention

The technical problem to be solved in this application is to provide a full-length bonded anchor failure mode under shearing in view of the unclear definition of the failure mode of the full-length bonded anchor under shear in the existing research identification method.

The technical solution adopted by this application to solve the technical problem is to propose a method for identifying the failure mode of a full-length bonded anchor under shear, which specifically includes the following steps:

Q1. Based on the UDEC software, the shear numerical model of the rock structural plane is established, and a full-length bonded anchor is implanted in the center of the numerical model;

Q2. Through the FISH programming language that comes with the UDEC software, the judgment process of pure shear failure, tensile shear failure and tensile bending failure of the anchor is embedded in the main program of numerical calculation;

Q3. Select the characteristic parameters that affect the failure mode of the bolt under shear action. The characteristic parameters include: bolt diameter D, bolt axial yield strength f _y , ultimate tensile strength f _u , bolt inclination angle α, rock unit The smaller value of the axial compressive strength and the uniaxial compressive strength of mortar σ _c and the dilatancy angle θ of the structural plane;

Q4. Select three groups of typical anchored structural surface shear tests, and substitute their relevant characteristic parameters into the numerical model. According to the shear force-shear displacement curve, the normal coupling spring and tangential coupling spring in the anchor unit Calibrate the spring parameters and ensure that the failure mode of the anchor in the numerical test is consistent with the failure mode of the anchor in the experiment;

/>

和

Q5. According to the common geological environment in engineering and the commonly used anchor bolt specifications, select the appropriate range and l _i (li is a positive integer, and the _{l i of} each parameter can be different) levels for the i-th characteristic parameter; The characteristic parameters are specifically:

/>

and

Q6. According to the scope and level of each characteristic parameter divided in step Q5, design n groups of numerical experiments through orthogonal experiments, and n should generally be greater than 20;

Q7. On the basis of steps Q3 and Q4, solve the n sets of numerical tests designed in step Q6 one by one, record the bolt failure mode under each set of numerical tests, and set up a data set E containing n samples, The form of each sample in the data set is E _i =(xi _, FM _i ), where xi ₌ (D _i ,f _yi ,f _ui ,α _i ,σ _ci ,θ _i ); i=1,2 ,3...,n, FM _i represents the failure mode of the bolt, where the pure shear failure is denoted as "PS" (pure-shear), the tension-shear failure is denoted as "TS" (tensile-shear), and the tension-bend Destruction is recorded as "TB"(tensile-bend);

Q8. Use the random forest classification algorithm to study the data set, use the x _i of each sample in the data set E in step Q7 as an input variable, and FM _i as an output variable to establish a random forest classification model;

Q9. Use the particle swarm optimization algorithm to intelligently optimize the hyperparameters in the random forest classification model established in step Q8, including the number G of decision trees, the maximum depth d of the random forest, and the number of attempts in a single tree when using the random attribute principle The maximum characteristic number k is used to establish an optimized rock bolt failure mode identification model;

Q10. For a full-length bonded bolt in a specific geological environment, when the rock masses on both sides of the structural surface slide or have a tendency to slide relative to each other, the bolt is subjected to strong shearing. At this time, determine the characteristic parameters [D ,f _y ,f _u ,α,σ _c ,θ], and bring it into the bolt failure mode identification model established in step Q9, the obtained failure mode is the shearing action of the bolt identified by the algorithm destruction mode.

Furthermore, in step Q1, the Rockbolt unit is used to simulate the full-length bonded anchor. Compared with the anchors simulated by other structural units, the Rockbolt unit has the ability to resist tension, shear, and bending at the same time, and is suitable for shearing of the anchor. Condition.

Further, in step Q2, the three failure modes of pure shear failure, tension shear failure and tension bending failure are embedded in the main calculation program, and the main flow is as follows:

a) Carry out the main program operation, when running to stepi (step i), find the address of the first node of the Rockbolt unit through the FISH built-in function, and extract the axial force, shear force and bending moment stored in the node;

b) Use the following formula (1) to judge whether the anchor rod enters the yield state:

A为锚杆截面积；In the formula: σ _e is the yield strength of the anchor, M ₀ and N ₀ are the bending moment and axial force of the anchor, W is the bending section coefficient,

A is the cross-sectional area of the bolt;

c) If the yield state judgment formula (1) is not satisfied, it is considered that the anchor has not yielded, and continue to use the Tresca failure criterion (2) to determine whether the pure shear failure of the anchor occurs:

In the formula: Q ₀ is the shear force at one point of the anchor rod, and _Nu is the axial ultimate force corresponding to the axial ultimate strength of the anchor rod.

If the above formula (2) is satisfied, it is considered that the pure shear failure of the anchor occurs, and the first string "PS" is output, and the yield value (yield-tension) and yield strain (tension-failure-strain) of the anchor are set as one The default value (1×10 ^-10 ), the program will determine that the anchor is damaged, and the calculation program will be terminated; if the above formula (2) is not satisfied, the anchor is still considered to be in an elastic state, and the next node address will be iterated repeatedly until The node traversal of the whole anchor rod is completed, let i=i+1, and proceed to the next main program operation;

The pure shear failure here is not strictly a pure shear failure with only shear stress, but because the axial force is small at this time and the shear force dominates, it is approximately considered to be a pure shear failure;

d) If the yield state judgment formula (1) is satisfied, it is considered that the anchor rod yields at this point, and the anchor rod enters a plastic state; in the main program, the bending moment M ₀ at this time is assigned to the plastic moment of the anchor rod (plastic- moment), the plastic hinge is formed, the bending moment will not increase after reaching the plastic moment, and the axial force will further increase with the increase of the shear displacement of the structural surface. At this time, it is necessary to judge the failure mode of the anchor rod after it enters the plastic state;

e) After the anchor rod enters the plastic state, if the axial force and shear force of a point meet the Mises failure criterion (3), it is considered that tensile shear failure occurs at this point:

If the above formula (3) is satisfied, then output the second character string "TS", and also set the yield value and yield strain of the anchor as the default value (1×10 ^-10 ), and the calculation program terminates;

f) If the above formula (3) is not satisfied, continue to judge: if the axial force and bending moment of a point satisfy the following relational formula (4), it is considered that tension bending failure occurs, and the third string "TB" is output, and the The yield value and yield strain of the anchor are set to the default value (1×10 ^-10 ), and the calculation program is terminated:

g) If the judgment formulas (3) and (4) above for tensile-shear failure and tensile-bending failure are not satisfied, it means that although the anchor rod has entered a plastic state, it still has not reached the failure limit. At this time, it is necessary to continue node traversal. If it is still unsatisfied after traversal, set i=i+1, and proceed to the next step of the main program operation until any one of the damage modes is successfully judged, and the main program terminates the operation.

Further, the random forest classification algorithm adopted in step Q8 is based on the C4.5 decision tree, and according to the basic theory of the C4.5 decision tree, the proportion of samples of the kth class in the current sample set R is pk (k=1, 2,...,|y|), the information entropy Ent(R) is used to measure the purity of the sample set, and the definition formula of information entropy is:

Assuming that the discrete attribute a has V possible values, the maximum information gain rate is used to select the partition attribute of the decision tree. The definition formula of the information gain rate is:

in:

In the formula: R ^v is all samples in R contained in the vth branch node that take the value of a ^v on attribute a.

作为候选划分点，然后像离散值一样来考察划分点，并选取最优划分点进行样本集合的划分，其中i为正整数。Since the input attributes of the present invention are all continuous values, the dichotomy method is adopted. For a certain continuous attribute a, the median point of the adjacent attribute value [a ⁱ , a ⁱ⁺¹ )

As a candidate division point, then examine the division point like a discrete value, and select the optimal division point to divide the sample set, where i is a positive integer.

Further, in step Q8, learning the data set E by using the random forest classification algorithm includes: according to the random forest classification theory, randomly sampling the data set E containing n training samples based on the bagging method, and obtaining G training samples containing m training samples The sampling set (G, n, and m are all positive integers) is used for the training of each decision tree respectively.

Further, in step Q8, the random forest classification algorithm uses the voting method to vote for the classification result of each decision tree, and the prediction result is the damage mode with the most votes, which is expressed as:

表示在第i个基学习器上在类别标记j(破坏模式)上的输出，G为决策树的数量，H(x)为最终输出。In the formula: X is the sample,

Denotes the output on the category label j (destruction mode) on the i-th base learner, G is the number of decision trees, and H(x) is the final output.

Further, in step Q9, it is necessary to select the number G of decision trees, the maximum depth d of the random forest, and the maximum number of features k tried in a single tree when the principle of random attributes is adopted. The present invention uses particle swarm optimization algorithm to select the The global intelligent optimization of parameters is carried out to obtain the optimal combination (G, d, k) to maximize the performance of the model, which specifically includes the following steps:

a) Initialize the particle and population speed, set the maximum number of iterations I _max =300, the number of particle populations is 25, and define the minimum error rate W _min ;

b) Determine the corresponding parameter combination (G, d, k) according to the particle number and population velocity, and substitute it into the model to obtain the prediction result;

c) Using the 10-fold cross-validation method, the error rate W is used to construct the fitness function in each cross-validation process, and the fitness value of the individual is calculated:

In the formula: K is the sample set, f represents the learner, f(xi ₎ is the prediction result of the random forest classification model, and y _i is the actual result;

d) Update particle velocity and particle position, perform next iteration, repeat step b and step c;

e) When the set maximum number of iterations is reached, the iteration ends, and the parameter combination (G _i , d _i , _ki ) corresponding to the minimum error rate W _min is output and substituted into the model to generate an optimized discrimination model.

A method for identifying the failure mode of a full-length bonded anchor under shear in this application is based on the combined deformation and strength theory of material mechanics, and embeds the three failure modes of the full-length bonded anchor under shear into the UDEC numerical calculation The program establishes data sets of different failure modes of bolts under different parameters through numerical calculation, and uses the algorithm combining random forest classification and particle swarm optimization to construct the mapping between the failure modes of bolts and rocks, structural surfaces and bolt parameters Based on the relationship, a model for identifying the failure mode of the full-length bonded anchor under shear action was established.

The implementation of a method for identifying the failure mode of a full-length bonded anchor under shear in the present application has the following beneficial effects: the mapping relationship between the failure mode of the anchor and the parameters of the rock, structural plane, and anchor is constructed, and the It is of great practical significance to identify the failure mode of the anchor under shear under different anchoring parameters in a specific geological environment, to accurately evaluate the shear resistance of the anchor, and to guide the engineering practice of rock mass anchorage.

Description of drawings

The application will be further described below in conjunction with the accompanying drawings and embodiments, in the accompanying drawings:

Fig. 1 is the general flowchart of bolt failure mode identification;

Fig. 2 is a flow chart of bolt failure mode determination in UDEC numerical software;

Fig. 3 is the flow chart of establishing the rock bolt failure mode identification model by adopting random forest classification algorithm and particle swarm optimization algorithm;

Figure 4 is a schematic diagram of the anchor rod being sheared.

Among them, in Fig. 4, 1, anchor rod, 2, upper rock block, 3, lower rock block, 4, structural surface, 5, fixed surface, A, anchor rod shear area, B, movement direction.

Detailed ways

The implementation of the present invention will be described in detail with specific cases below in conjunction with the accompanying drawings.

Example 1

Please refer to the accompanying drawings 1-3, here is a specific description of the method proposed in the embodiment of the application; a method for identifying the failure mode of a full-length bonded anchor under shear action proposed in the embodiment of the application, specifically includes the following steps:

P1. Establish the shear numerical model of the anchored structural surface in UDEC software, and use the Rockbolt unit to simulate the anchor rod under the shear action;

P2. Express the bolt failure mode judgment process in accompanying drawing 2 with FISH language, and embed it in the main program of numerical simulation calculation;

P3. Select three sets of typical shear test data of anchored structural planes, and respectively substitute the parameters [D, f _y , f _u , α, σ _c , θ] into the numerical model established in P2. According to the shear force-shear Displacement curves, which calibrate the parameters of the normal coupling spring and tangential coupling spring in the Rockbolt unit, and ensure that the failure mode is consistent;

P4. According to the common geological environment in engineering and the experience of commonly used bolt specifications, select the appropriate range and level for each parameter, as shown in Table 1:

Table 1 Parameter value range and level

P5. A total of 25 sets of tests were designed through the L25 orthogonal table, as shown in Table 2, the 25 sets of test parameters were substituted into the numerical model established by P2 for calculation, and the corresponding bolt failure mode was obtained;

Table 2 Orthogonal test table

P6. Through the 25 sets of numerical experiments in P5, a data set E containing 25 sets of training data is obtained. The form of each sample in the data set E is E _i =( _xi , FM _i ), where xi ₌ (D _i ,f _yi ,f _ui ,α _i ,σ _ci ,θ _i ); i=1,2,3...,25;

P7. Using the random forest classification algorithm, set x _i = (D _i , f _yi , f _ui , α _i , σ _ci , θ _i ) in step P6 as input variables, and FM _i as output variables to establish bolt failure mode discrimination Model;

P8, use the particle swarm optimization algorithm in accompanying drawing 3 to the quantity G of the decision tree in the random forest classification model that step P7 establishes, the maximum depth d of random forest and the maximum characteristic that try in a single tree when adopting random attribute principle The number k is used for global optimization, and the optimized bolt failure mode identification model is established. Finally, when the number of trees is 90, the maximum depth is 4, and the maximum number of features is 3, the fitness is the highest (the error rate is the lowest), and the model reaches the optimum;

P9. For a specific case where the bolt is sheared, after determining the input parameters [D, f _y , _fu , α, σ _c , θ], substitute them into the discrimination model established in P8, and the corresponding Anchor failure mode.

Example 2

In order to illustrate better, an anchored structural surface shear test is taken as an example to describe in detail. Please refer to accompanying drawing 4, among Fig. 4, anchor rod 1 is inserted in the structural system that is made of upper rock block 2 and lower rock block 3, forms structure surface 4 between upper rock block 2 and lower rock block 3, and anchor rod 1 and The inclination angle between the structural planes 4 is α=90°, the diameter of the anchor rod 1 used in this test is D=4mm, the yield strength of the anchor rod 1 is f _y =475MPa, the ultimate strength f _u =580MPa, σ _c =51.44MPa, The dilation angle of the structural plane is θ=13.67°, the circular area A in Fig. 4 represents the shear area of the bolt 1, and the direction B indicated by the arrow represents the movement direction of the lower rock block. After the test, the anchor rod 1 was taken out for observation, and it was found that the anchor rod 1 had obvious tensile and shear failure characteristics.

All the parameters of embodiment 2 are within the value range of the data set feature parameters in embodiment 1, so the model established in embodiment 1 can be used for identification, and the model in embodiment 2 of the application can be identified by using the model of embodiment 1 The shear test is carried out to judge, and the final output bolt failure mode is "TS", which is consistent with the experimental observation results. Embodiment 2 further verifies that the method proposed in the embodiment of the present application has good recognition accuracy.

Embodiments of the present invention have been described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementations, and the above-mentioned specific implementations are only illustrative, rather than restrictive, and those of ordinary skill in the art will Under the enlightenment of the present invention, many forms can also be made without departing from the gist of the present invention and the protection scope of the claims, and these all belong to the protection of the present invention.

Claims (7)

1. The method for judging the damage mode of the full-length adhesive anchor rod under the shearing action is characterized by comprising the following steps:

q1, establishing a rock structural surface shearing numerical model based on UDEC software, and implanting a full-length adhesive anchor rod in the center of the numerical model;

q2, embedding a judging process of pure shearing damage, tensile shearing damage and stretch bending damage of the anchor rod into a numerical calculation main program through a self-contained FISH programming language of UDEC software;

q3, selecting characteristic parameters influencing the anchor rod damage mode under the shearing action, wherein the characteristic parameters comprise: diameter D of anchor rod and axial yield strength f of anchor rod _y Ultimate tensile strength f _u The anchor rod inclination angle alpha, the rock uniaxial compressive strength and the smaller value sigma of the mortar uniaxial compressive strength _c And a structural face shear angle θ;

q4, selecting three groups of typical anchor-added structural surface shear tests, substituting the relevant characteristic parameters into a numerical model, calibrating normal coupling springs and tangential coupling spring parameters in the anchor rod unit according to a shear force-shear displacement curve, and ensuring that the failure mode of the anchor rod in the numerical test is consistent with the failure mode of the anchor rod in the test;

q5, selecting a proper range and l according to the common geological environment in engineering and the common anchor rod specification _i A level; the characteristic parameters are specifically as follows:

and->

Q6, designing n groups of numerical tests through orthogonal tests according to the range and the level of each characteristic parameter divided in the step Q5;

q7, on the basis of the steps Q3 and Q4, solving n groups of numerical tests designed in the step Q6 one by one, recording anchor rod damage modes under each group of numerical tests, and establishing a data set E containing n samples, wherein each sample in the data set E is in the form of E _i ＝(x _i ,FM _i ) Wherein x is _i ＝(D _i ,f _yi ,f _ui ,α _i ,σ _ci ,θ _i ) The method comprises the steps of carrying out a first treatment on the surface of the i=1, 2,3., n, FM represents the failure mode of the bolt;

q8, learning the data set E by adopting a random forest classification algorithm, and obtaining x of each sample in the data set E in the step Q7 _i FM as an input variable _i As output variable, establishing a random forest classification model;

q9, intelligently optimizing the super parameters in the random forest classification model established in the step Q8 by adopting a particle swarm algorithm, wherein the super parameters comprise the number G of decision trees, the maximum depth d of the random forest and the maximum feature number k tried in a single tree by adopting a random attribute principle, and establishing an optimized anchor rod damage mode judgment model;

q10, for a full-length adhesive anchor rod in a specific geological environment, when rock masses at two sides of a structural surface slide relatively or have a tendency of sliding relatively, the anchor rod is subjected to strong shearing action, and at the moment, characteristic parameters [ D, f are determined _y ,f _u ,α,σ _c ,θ]And (3) introducing the failure mode of the anchor rod into the failure mode judging model of the anchor rod established in the step (Q9), wherein the obtained failure mode is the failure mode of the anchor rod under the shearing action judged by an algorithm.

2. The method of claim 1, wherein the step Q1 uses a Rockbolt unit to simulate a full length cohesive bolt.

3. The method for judging the failure mode of a full-length adhesive anchor rod under a shearing action according to claim 2, wherein the process of embedding the judging process of pure shearing failure, stretch shearing failure and stretch bending failure into the numerical calculation main program in the step Q2 is as follows:

a) When the main program operation is carried out and the operation is carried out to the stepi, the address of the first node of the Rockbolt unit is found through a FISH built-in function, and the axial force, the shearing force and the bending moment stored in the node are extracted;

b) Judging whether the anchor rod enters a yield state or not by using the following formula (1):

wherein: sigma (sigma) _e For the yield strength of the anchor rod, M ₀ And N ₀ Is the bending moment and the axial force of the anchor rod, W is the bending section coefficient,

a is the sectional area of the anchor rod;

c) If the yield state judging formula (1) is not satisfied, the anchor rod is considered to be unyielding, and the Tresca breaking criterion (2) is continuously used for judging whether the anchor rod is broken by pure shearing or not:

wherein: q (Q) ₀ Is the shearing force of one point of the anchor rod, N _u The axial limit force is corresponding to the axial limit strength of the anchor rod;

if the formula (2) is met, the anchor rod is considered to be damaged by pure shearing, a first character string is output, the yield value and the yield strain of the anchor rod are set as default values, the program can judge that the anchor rod is damaged, and the calculation program is terminated; if the formula (2) is not satisfied, considering that the anchor rod is still in an elastic state, repeating iteration on the next node address until node traversal of the whole anchor rod is completed, and performing next main program operation by making i=i+1;

d) If the yield state judging formula (1) is satisfied, the anchor rod is considered to yield at the point, and the anchor rod enters a plastic state; bending moment M at this time in the main routine ₀ The plastic moment assigned to the anchor rod is formed by plastic hinge, the bending moment is not increased after reaching the plastic moment, the axial force is further increased along with the increase of the shearing displacement of the structural surface, and the failure mode of the anchor rod after entering the plastic state is judged at the moment;

e) If the axial force and shear force of a point meet Mises failure criterion (3), it is considered that a pull shear failure occurs at that point:

if the formula (3) is satisfied, outputting a second character string, setting the yield value and the yield strain of the anchor rod as default values, and terminating the calculation program;

f) If the above formula (3) is not satisfied, continuing the determination: if the axial force and the bending moment at one point meet the following relational expression (4), the stretch bending damage is considered to occur, a third character string is output, the yield value and the yield strain of the anchor rod are set as default values, and the calculation program is terminated:

g) If the judgment formulas (3) and (4) of the pull-up shear damage and the stretch bending damage are not satisfied, the anchor rod is in a plastic state, but the damage limit is not reached, node traversal is needed to be continued at the moment, if the judgment formulas are still not satisfied after the node traversal, i=i+1 is needed to carry out next main program operation until any damage mode judgment is successful, and the main program terminates operation.

4. The method for judging failure mode of full-length adhesive anchor rod under shearing action as claimed in claim 1, wherein the random forest classification algorithm adopted in the step Q8 uses C4.5 decision tree as a base learner, and the proportion of the kth sample in the current sample set R is p according to the basic theory of C4.5 decision tree _k (k=1, 2, |y|) the information entropy is used to measure the purity of the sample set:

assuming that the discrete attribute a has V possible values, the maximum information gain rate is used for selecting the partition attribute of the decision tree:

wherein:

wherein: r is R _v All of the R's included for the v-th branch node have a value of a on attribute a ^v Is a sample of (a).

5. The method for judging the failure mode of the full-length adhesive anchor rod under the shearing action according to claim 1, wherein the step Q8 is characterized in that a random forest classification algorithm is adopted to learn the data set E, and the method comprises the steps of randomly sampling the data set E containing n training samples based on a bagging method according to a random forest classification theory to obtain G sampling sets containing m training samples, wherein the G sampling sets are respectively used for training each decision tree.

6. The method for judging the failure mode of a full-length adhesive anchor rod under the shearing action according to claim 1, wherein the random forest classification algorithm in the step Q8 adopts a voting method to vote on the classification result of each decision tree, and the predicted result is the failure mode with the most votes, and is expressed as follows by a formula:

wherein:

is shown in the firstOutput on class label j on i base learners.

7. The method for judging the failure mode of a full-length adhesive anchor rod under the shearing action according to claim 1, wherein the intelligent optimization of the super parameters in the random forest classification model established in the step Q8 by adopting a particle swarm algorithm in the step Q9 comprises the following steps:

a) Initializing particle and population speed, and setting maximum iteration number I _max =300, particle population number 25, defining minimum error rate W _min ；

b) Determining corresponding parameter combinations (G, d, k) according to the particle numbers and the population speeds, and substituting the parameter combinations into the random forest classification model to obtain a prediction result;

c) Adopting a 10-fold cross verification method, constructing an fitness function by adopting an error rate W in each fold cross verification process, and calculating the fitness value of an individual:

wherein: k is the sample set, f represents the learner, f (x _i ) Predicting the result, y, for the random forest classification model _i Is the actual result;

d) Updating the particle speed and the particle position, performing the next iteration, and repeating the step b and the step c;

e) Ending the iteration when the set maximum iteration number is reached, and outputting the minimum error rate W _min Corresponding parameter combinations (G _i ,d _i ,k _i ) Substituting the model into the random forest classification model to generate an optimized judgment model.

Images