A method for identifying the wear condition of a rolling linear guide pair

Abstract

This invention discloses a method for identifying the wear state of rolling linear guide pairs, applicable to the field of performance testing technology for rolling linear guide pairs. The invention acquires vibration signals of the rolling linear guide pair during its operation using a vibration signal acquisition system, while simultaneously measuring the change in preload drag force. It uses fully ensemble empirical mode decomposition based on adaptive noise to decompose and reconstruct the original vibration signals, reducing noise interference. A feature set is constructed by extracting multi-scale fuzzy entropy from the reconstructed signals. Kernel principal component analysis is used to reduce the dimensionality of the original feature space, constructing a new feature set. A random forest model based on the gray wolf optimization algorithm is established, and the model is trained using a mixed feature set of three wear states. The trained model is then used to identify the wear state of the rolling linear guide pair. The features extracted by this method accurately characterize the wear state of the guide pair and are suitable for identifying the wear state of rolling linear guide pairs.

Description

Technical Field

[0001] This invention relates to the field of performance testing technology for rolling linear guide pairs, and more specifically to a method for identifying the wear condition of rolling linear guide pairs. Background Technology

[0002] Rolling linear guide pairs, as precision transmission components, possess characteristics such as high positioning accuracy, good load-bearing capacity, low frictional resistance, and high reliability, and are widely used in the transmission systems of high-end CNC machine tools. In practical applications, to improve the rigidity and running accuracy of the guide pair, balls with different interference fits are typically used to generate preload within the guide pair to eliminate clearance. However, during the operation of the guide pair, wear between the balls and raceways gradually reduces the preload, leading to degradation in preload drag force, rigidity, and running accuracy. This affects the machining accuracy of the CNC machine tool and the quality of the machined parts. Therefore, accurately identifying the wear state of rolling linear guide pairs is crucial for ensuring product quality, production reliability, and safety.

[0003] Current research on wear condition identification mainly focuses on ball screw pairs, cutting tools, and gears, with a lack of research on identifying the wear condition of rolling linear guide pairs. Furthermore, because the vibration signals during the operation of rolling linear guide pairs contain a large amount of noise, extracting features characterizing the wear condition is difficult. Therefore, appropriate signal processing and feature extraction methods are crucial for improving identification accuracy.

[0004] Therefore, proposing a wear condition identification method for rolling linear guide pairs to solve the difficulties existing in the prior art is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method for identifying the wear condition of rolling linear guide pairs, which is applicable to the identification of wear condition under a single working condition, and solves the problem that there is a lack of current methods for identifying the wear condition of rolling linear guide pairs.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for identifying the wear condition of a rolling linear guide pair includes the following steps:

[0008] S1. Measure the change in preload and the corresponding vibration signal of the rolling linear guide pair using a preload drag force test bench and vibration signal acquisition system;

[0009] S2. Determine the wear condition of the rolling linear guide pair and use it as a label;

[0010] S3. The original vibration signal is decomposed and reconstructed using complete ensemble empirical mode decomposition based on adaptive noise.

[0011] S4. Perform multi-scale fuzzy entropy analysis on the reconstructed signal, extract fuzzy entropy at τ scales, and construct a feature set characterizing the wear state of the guide rail pair.

[0012] S5. Dimensionality reduction of the original feature space is performed using kernel principal component analysis.

[0013] S6. Establish a random forest model based on the gray wolf optimization algorithm. Standardize the labeled feature set and divide it into training and test sets according to the ratio. Import the training set into the model and use the gray wolf optimization algorithm to find the optimal values ​​of the number of decision trees and the minimum number of leaf nodes.

[0014] S7. Import the test set into the trained model to identify the wear state of the guide rail pair, and compare the identified wear state of the guide rail pair with the actual state to obtain the accuracy of the model.

[0015] S8. For the rolling linear guide pair to be identified, execute S1, S3-S5 to obtain its unlabeled feature set, and then use the trained model to identify the wear state of the rolling linear guide pair.

[0016] Optionally, S1 specifically includes: placing the rolling linear guide pair on a reliability test bench for loading and running, stopping the test bench and removing the rolling linear guide pair after running for a specified mileage, collecting vibration signals under its operating state through a vibration signal acquisition system, and simultaneously measuring the preload of the rolling linear guide pair using a preload drag force test bench.

[0017] Optionally, S2 specifically involves: using the turning point of the trend of the preload driving force of the rolling linear guide pair as the dividing point, classifying the wear state of the rolling linear guide pair into running-in wear, stable wear, and rapid wear states in sequence; and determining its wear state based on the current trend of the preload driving force of the rolling linear guide pair.

[0018] Optionally, S3 specifically includes:

[0019] Add i (i = 1, 2, ..., n) sets of Gaussian white noise n to the original vibration signal x(t). i The signal to be decomposed is obtained:

[0020] x i (t)=x(t)+n i

[0021] Based on the signal to be decomposed x i The maximum and minimum points of (t) are constructed using cubic spline interpolation. i The upper and lower envelopes u of (t)i (t) and v i (t), the local mean of the signal is:

[0022] m i (t)=[u i (t)+v i (t)] / 2

[0023] Determine h i (t)=x i (t)-m i (t) In its signal, whether the number of extreme points and zero-crossing points are equal or differ by at most one, and whether the mean of the upper and lower envelopes at all points is zero, if these conditions are met, then the first intrinsic mode component (IMF) is obtained. 1i Otherwise for x i (t) Repeat the above steps until signal h i (t)=x i (t)-m i (t) satisfies the required conditions.

[0024] Optionally, S4 specifically includes:

[0025] For an original sequence of length N {x1, x2, ..., x...} N A scale factor τ = 1, 2, ... was introduced to establish a new coarse-grained vector, using the following formula:

[0026]

[0027] in, Given the length of the coarse-grained vector, the original sequence is divided into τ segments of length τ. coarse-grained sequence u j (τ).

[0028] Optionally, S5 specifically includes:

[0029] For the original feature set X i (i = 1, 2, ..., n) are standardized to obtain a new sample set x. i The formula used is:

[0030]

[0031] in, Let σ(X) be the mean of the original feature set, and σ(X) be the standard deviation of the original feature set.

[0032] A nonlinear mapping function φ(x) is introduced to map the sample set to a high-dimensional feature space F, and its covariance matrix W is calculated using the following formula:

[0033]

[0034] Where λ is the eigenvalue of matrix W, and A is the eigenvector of matrix W.

[0035] Optionally, S6 specifically includes the following steps:

[0036] S601: Randomly generate M individuals as the initial population, and set the maximum number of iterations to N;

[0037] S602: Input the training set into the random forest model to train the model. Use the accuracy of the wear state identification of the rolling linear guide pair as the fitness function. The formula used is:

[0038]

[0039] Where s is the number of correctly identified data in the dataset, and t is the total number of data in the dataset;

[0040] S603: Find the individual corresponding to the maximum value of the fitness function Fitness, start iterative optimization, and determine whether the current fitness is better than the maximum fitness of the previous generation or whether the maximum number of evolutionary iterations of the population has been reached. If the condition is met, retain the optimal fitness value and its corresponding coordinates, and assign them to the initial coordinates; otherwise, return to step S602.

[0041] S604: Outputs the optimal number of decision trees and the optimal minimum number of leaf nodes for the random forest algorithm, completing the construction of the wear condition identification model for rolling linear guide pairs.

[0042] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method for identifying the wear state of a rolling linear guide pair, which has the following beneficial effects:

[0043] (1) The present invention uses a complete set empirical mode decomposition based on adaptive noise to decompose and reconstruct the original vibration signal, which can not only eliminate the influence of noise, but also reconstruct the important information part of the original signal.

[0044] (2) The multi-scale fuzzy entropy extracted by this invention can reflect the essence of the wear degree. The random forest model based on the gray wolf optimization algorithm can effectively identify the wear state of the rolling linear guide pair with high accuracy. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0046] Figure 1 A flowchart of a method for identifying the wear state of a rolling linear guide pair provided by the present invention;

[0047] Figure 2 This is a schematic diagram of the preload traction force measurement test bench for the rolling linear guide pair according to the present invention;

[0048] Figure 3 This is a schematic diagram of the vibration signal acquisition system for the rolling linear guide pair of the present invention;

[0049] Figure 4 This is a schematic diagram of the vibration signal acquisition system for the rolling linear guide pair of the present invention;

[0050] Figure 5 A detailed flowchart of a method for identifying the wear condition of a rolling linear guide pair provided by the present invention;

[0051] Figure 6 This is a schematic diagram of fatigue spalling on the raceway surface of the rolling linear guide pair according to the present invention.

[0052] Figure 7 This is a schematic diagram illustrating the wear state classification of the rolling linear guide pair according to the present invention;

[0053] Figure 8 The vibration signal time-domain waveforms of the rolling linear guide pair under three wear states of the present invention are shown.

[0054] Figure 9 This is a schematic diagram of the confusion matrix for identifying the wear state of a rolling linear guide pair based on the gray wolf optimization algorithm of the present invention;

[0055] Among them, 1-force gauge, 2-pad, 3-guide rail, 4-bed, 5-moving platform; 6-slider of the linear guide pair to be tested, 7-sensor, 8-push rod, 9-Prosig P8020 data acquisition system, 10-data cable, 11-PC; 12-bed, 13-guide rail pair responsible for bearing and guiding, 14-guide rail adapter plate, 15-servo motor, 16-gear, 17-rack, 18-pad, 19-force sensor, 20-handwheel. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Reference Figure 1 As shown, this invention discloses a method for identifying the wear condition of a rolling linear guide pair, comprising the following steps:

[0058] S1. Measure the change in preload and the corresponding vibration signal of the rolling linear guide pair using a preload drag force test bench and vibration signal acquisition system;

[0059] S2. Determine the wear condition of the rolling linear guide pair and use it as a label;

[0060] S3. The original vibration signal is decomposed and reconstructed using complete ensemble empirical mode decomposition based on adaptive noise.

[0061] S4. Perform multi-scale fuzzy entropy analysis on the reconstructed signal, extract fuzzy entropy at τ scales, and construct a feature set characterizing the wear state of the guide rail pair.

[0062] S5. Dimensionality reduction of the original feature space is performed using kernel principal component analysis.

[0063] S6. Establish a random forest model based on the gray wolf optimization algorithm. Standardize the labeled feature set and divide it into training and test sets according to the ratio. Import the training set into the model and use the gray wolf optimization algorithm to find the optimal values ​​of the number of decision trees and the minimum number of leaf nodes.

[0064] S7. Import the test set into the trained model to identify the wear state of the guide rail pair, and compare the identified wear state of the guide rail pair with the actual state to obtain the accuracy of the model.

[0065] S8. For the rolling linear guide pair to be identified, execute S1, S3-S5 to obtain its unlabeled feature set, and then use the trained model to identify the wear state of the rolling linear guide pair.

[0066] Furthermore, in one embodiment, such as Figure 2As shown, the preload drag force measurement test bench mainly consists of a force gauge 1, a pad 2, a bed 4, and a moving platform 5. The rolling linear guide pair 3 to be tested is fixed to the bed 4 using bolts and washers. The guide rail 3 should be parallel to the moving direction and measurement direction of the force gauge 1. The measurement stroke is set to 60% of the guide rail length, and the test speed is set to 11.6 mm / s (0.7 m / min). During data processing, 5% of the data before and after the measurement stroke is removed. The forward drag and reverse drag are measured 3 times each, and the average value is taken as the preload drag force of the guide rail pair 3 to be tested.

[0067] Combination Figure 3 ,based on Figure 2 The aforementioned pre-tensioning drag force measurement test bench is equipped with a vibration signal acquisition system. This system mainly includes a vibration sensor 7, push rods 8, a Prosig P8020 data acquisition system 9, a data cable 10, and a PC 11. The linear guide rail sub-slider 6 to be tested is placed between the two push rods 8. The sensor 7 is positioned on the top surface, side surface, and one end of the guide rail of the sub-slider. The sampling length is set to 60% of the guide rail length, the sampling frequency is set to 50Hz, and the slider running speed is set to 7m / min, 14m / min, and 21m / min, respectively. Vibration signals are collected from two reciprocating motions of the slider.

[0068] Next, a running-in test of the rolling linear guide pair under load was conducted, combined with Figure 4 The running-in test bench mainly consists of a bed 12, two load-bearing guide rail pairs 13, a guide rail adapter plate 14, a servo motor 15, gears 16, a rack 17, and a gantry loading device. The gantry loading device consists of pads 18, force sensors 19, and handwheels 20. The rolling linear guide rail pair under test is fixed on the guide rail adapter plate 14 on the side of the test bench, and pads 18 are installed on the slider. The handwheels 20 on both sides of the gantry loading mechanism apply load to the slider. The load magnitude is measured by the force sensor 19 on the loading block and transmitted to the data acquisition system. The test bench uses a servo motor 15 to drive the gear rack 16 and 17 to achieve the reciprocating motion of the slider under test. The test parameters are set as follows: load is 40% of the rated dynamic load, and running speed is 50 m / min. During the first 45 km of the guide rail pair's operation, the preload drag force is measured every 5 km and vibration signals are collected; after 45 km, the preload drag force is measured every 10 km and vibration signals are collected.

[0069] Furthermore, based on the changing trend of the preload driving force during the operation of the rolling linear guide pair, the wear state of the rolling linear guide pair is divided into running-in wear state, stable wear state, and severe wear state.

[0070] Furthermore, the original vibration signal is decomposed and reconstructed using complete ensemble empirical mode decomposition based on adaptive noise to eliminate the influence of noise. Specific steps include:

[0071] First, add i (i = 1, 2, ..., n) sets of Gaussian white noise n to the original vibration signal x(t). i The signal to be decomposed is obtained:

[0072] x i (t)=x(t)+n i

[0073] Based on the signal to be decomposed x i The maximum and minimum points of (t) are constructed using cubic spline interpolation. i The upper and lower envelopes u of (t) i (t) and v i (t), the local mean of the signal is:

[0074] m i (t)=[u i (t)+v i (t)] / 2

[0075] Determine h i (t)=x i (t)-m i (t) In its signal, whether the number of extreme points and zero-crossing points are equal or differ by at most one, and whether the mean of the upper and lower envelopes at all points is zero, if these conditions are met, then the first intrinsic mode component (IMF) is obtained. 1i Otherwise for x i (t) Repeat the above steps until signal h i (t)=x i (t)-m i (t) satisfies the required conditions;

[0076] The average value of the first eigenmode component is calculated to obtain the first-order eigenmode component of the complete ensemble empirical mode decomposition based on adaptive noise:

[0077]

[0078] The formula used to calculate the remainder r1(t) of the first-order modal component is:

[0079]

[0080] Add white noise to the remainder of the obtained modal components to construct a new signal to be decomposed, r1(t) + ε1E1(n). i ), calculate the second modal component

[0081]

[0082] The formula used to calculate the remainder r2(t) of the second-order modal component is:

[0083]

[0084] Repeat the above formula until the number of extreme points of the residual signal is less than 3, then stop the decomposition to obtain the K-order modal components and the residual term r. k (t), the final decomposition result can be expressed as:

[0085]

[0086] All intrinsic mode components are arranged in descending order of energy percentage, and the vibration signal is reconstructed based on their energy percentage using the following formula:

[0087]

[0088]

[0089] Where x(j) is the amplitude at each point in the time series, N is the length of the time series, and E i Let M be the energy of each eigenmode component, M be the number of eigenmode components obtained from the decomposition, and β be the energy of each eigenmode component. i This represents the energy percentage corresponding to this modal component;

[0090] The intrinsic mode components are filtered by setting a threshold. The energy ratio threshold is set to 70%. When the sum of the energy ratios of the current k intrinsic mode components exceeds 70%, the first k mode components are selected to reconstruct the original signal.

[0091] Furthermore, multi-scale fuzzy entropy analysis is performed on the reconstructed signal to extract fuzzy entropy at τ scales (generally τ > 10, but this method uses τ = 16), constructing a feature set characterizing the wear state of the guide rail pair. The specific steps are as follows:

[0092] For an original sequence of length N {x1, x2, ..., x...} N A scale factor τ = 1, 2, ... was introduced to establish a new coarse-grained vector, using the following formula:

[0093]

[0094] in, Given the length of the coarse-grained vector, the original sequence is divided into τ segments of length τ. coarse-grained sequence u j (τ).

[0095] For a length of Given a coarse-grained sequence {u(i): 1≤i≤N / τ}, construct a vector. The formula used is:

[0096]

[0097] Where u0(i) is the mean of a time series of length m, expressed as:

[0098]

[0099] For two vectors and The distance between two vectors is represented by the maximum difference between corresponding elements, and the formula is:

[0100]

[0101] Through fuzzy function Calculate vectors and The similarity is determined by an exponential fuzzy function, and the formula used is:

[0102]

[0103] Where n and r are the gradient and width of the function boundary, respectively;

[0104] function φ m (n,r) is represented as:

[0105]

[0106] For vectors Repeated vectors The steps are to obtain the function φ m+1 (n,r), the formula used is:

[0107]

[0108] Based on function φ m (n,r) and φ m+1 (n,r), the fuzzy entropy of the coarse-grained sequence is calculated using the following formula:

[0109] FuzzyEn(m,n,r,N)=ln(φ m (n,r)-ln(φ m+1 (n,r))

[0110] Repeat the above steps for each coarse-grained sequence to obtain the fuzzy entropy of the time series under different scale factors, and construct a feature set representing the wear state.

[0111] Furthermore, the original feature space is reduced in dimensionality using kernel principal component analysis. The specific steps are as follows:

[0112] For the original feature set Xi (i = 1, 2, ..., n) are standardized to obtain a new sample set x. i The formula used is:

[0113]

[0114] in, Let σ(X) be the mean of the original feature set, and σ(X) be the standard deviation of the original feature set.

[0115] A nonlinear mapping function φ(x) is introduced to map the sample set to a high-dimensional feature space F, and its covariance matrix W is calculated using the following formula:

[0116]

[0117] Where λ is the eigenvalue of matrix W, A is the eigenvector of matrix W, and λ and A satisfy:

[0118] λA=WA

[0119]

[0120] Introducing a nonlinear function φ(x) at both ends i ) T The calculation formula is:

[0121]

[0122]

[0123] Since the form of the mapping function φ is unknown, a kernel function κ is introduced, and the formula used is:

[0124] κ(x i ,x j )=φ(x i ) T φ(x j )

[0125] The above formula can be expressed as:

[0126] KA=nλA

[0127] Where K is the kernel matrix corresponding to κ, its eigenvalues ​​are nλ, and the corresponding eigenvectors are A = (α1; α2; ...; α...). n For any sample, the data in the high-dimensional feature space F are reflected in the feature vector V. k The projection onto (k = 1, 2, ..., d) is the principal component feature of the original data x. A new feature set is constructed by selecting the first three principal component features, and the principal component features are represented as follows:

[0128]

[0129] Furthermore, a random forest model based on the Grey Wolf Optimization Algorithm is established. The labeled feature set is standardized and divided into training and test sets in an 8:2 ratio. The training set is imported into the model, and the Grey Wolf Optimization Algorithm is used to find the optimal values ​​for the number of decision trees and the minimum number of leaf nodes. The specific process is as follows:

[0130] S601: Randomly generate M individuals as the initial population, and set the maximum number of iterations to N;

[0131] S602: Input the training set into the random forest model to train the model. Use the accuracy of the wear state identification of the rolling linear guide pair as the fitness function. The formula used is:

[0132]

[0133] Where s is the number of correctly identified data in the dataset, and t is the total number of data in the dataset;

[0134] S603: Find the individual corresponding to the maximum value of the fitness function, Fitness, and start iterative optimization. Determine whether the current fitness is better than the maximum fitness of the previous generation or whether the maximum number of evolutionary iterations of the population has been reached. If the condition is met, retain the optimal fitness value and its corresponding coordinates and assign them to the initial coordinates; otherwise, return to the previous step.

[0135] S604: Outputs the optimal number of decision trees and the optimal minimum number of leaf nodes for the random forest algorithm, completing the construction of the wear condition identification model for rolling linear guide pairs.

[0136] Finally, the test set data is imported into the trained random forest model to identify the wear state of the rolling linear guide pair. The specific process is as follows: Figure 5 As shown.

[0137] The method of the present invention will be illustrated below through an embodiment.

[0138] The experimental rolling linear guide pair was selected as the SHS35V type guide rail manufactured by THK Corporation of Japan. The main parameters are shown in Table 1.

[0139] Table 1

[0140]

[0141] During the running and testing process, the preload drag force of the rolling linear guide pair was continuously measured, and vibration data was collected simultaneously, until fatigue spalling appeared on the raceway surface of the rolling linear guide pair, such as... Figure 6 As shown, the total running distance and mileage of the rolling linear guide at this time is 205km.

[0142] Next, the wear state of the rolling linear guide pair is classified according to the changing trend of the preload traction force during the running process, such as... Figure 7 As shown in the figure, the curves exhibit a clear three-stage trend, which can be divided into the break-in wear stage, the stable wear stage, and the severe wear stage. The time-domain curves of the vibration signals in the three stages are as follows: Figure 8 As shown, the test data used are the vibration signals of the top surface of the slider of the guide rail pair at a speed of 7 m / min. Among them, 100 sets of data were selected for the running-in wear stage, 320 sets of data for the stable wear stage, and 60 sets of data for the severe wear stage. Each set of data contains 1024 data points, and the total data length is 480×1024.

[0143] Next, the original signal is decomposed using a complete set empirical mode decomposition method based on adaptive noise to obtain different intrinsic mode components. Suitable mode components are then selected for reconstruction by using the energy proportion threshold of the mode components, thereby reducing noise in the vibration signal.

[0144] The next step is to extract fuzzy entropy at 16 scales from the reconstructed signal and construct a feature set characterizing the wear state of the linear guide pair.

[0145] Next, the original feature space is reduced in dimensionality using the kernel principal component analysis method, with the number of principal components set to 3, that is, the feature space is reduced from 16 dimensions to 3 dimensions, forming a feature set with a length of 480×3.

[0146] After establishing the random forest model based on the gray wolf optimization algorithm, the training set data was imported into the model for training. The optimized random forest had 28 decision trees and a minimum number of leaf nodes of 4.

[0147] Finally, the test set is imported into the trained model for wear state identification. The confusion matrix for identification is as follows: Figure 9 As shown, the recognition accuracy can reach 97.917%.

[0148] To verify the recognition effect of this invention at different positions and speeds of the rolling linear guide pair, the wear state was identified using vibration signals from the top and side surfaces of the slider and the guide rail at speeds of 7 m / min, 14 m / min, and 21 m / min. The identification results are shown in Table 2. Observation reveals that for the vibration signal from the top surface of the slider, the recognition accuracy is highest at 7 m / min, while the recognition results are poorer at 14 m / min and 21 m / min. For the vibration signal from the side surface of the slider, the recognition accuracy is similar across the three speeds, all around 90%. For the vibration signal from the guide rail, the recognition accuracy can reach over 90% at all three speeds, indicating that this invention has certain applicability for identifying the wear state of rolling linear guide pairs under single operating conditions.

[0149] Table 2

[0150]

[0151] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0152] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying the wear condition of a rolling linear guide pair, characterized in that, Includes the following steps: S1. Measure the change in preload and the corresponding vibration signal of the rolling linear guide pair using a preload drag force test bench and vibration signal acquisition system; S2. Determine the wear condition of the rolling linear guide pair and use it as a label; S3. The original vibration signal is decomposed and reconstructed using complete ensemble empirical mode decomposition based on adaptive noise. S4. Perform multi-scale fuzzy entropy analysis on the reconstructed signal to extract... Fuzzy entropy at various scales is used to construct a feature set characterizing the wear state of the guide rail pair; S5. Dimensionality reduction of the original feature space is performed using kernel principal component analysis. S6. Establish a random forest model based on the gray wolf optimization algorithm. Standardize the labeled feature set and divide it into training and test sets according to the ratio. Import the training set into the model and use the gray wolf optimization algorithm to find the optimal values ​​of the number of decision trees and the minimum number of leaf nodes. S7. Import the test set into the trained model to identify the wear state of the guide rail pair, and compare the identified wear state of the guide rail pair with the actual state to obtain the accuracy of the model. S8. For the rolling linear guide pair to be identified, execute S1, S3-S5 to obtain its unlabeled feature set, and then use the trained model to identify the wear state of the rolling linear guide pair.

2. The method for identifying the wear condition of a rolling linear guide pair according to claim 1, characterized in that, S1 specifically includes: placing the rolling linear guide pair on a reliability test bench for loading and running, stopping the test bench and removing the rolling linear guide pair after running for a specified mileage, collecting vibration signals under its operating state through a vibration signal acquisition system, and measuring the preload of the rolling linear guide pair using a preload drag force test bench.

3. The method for identifying the wear condition of a rolling linear guide pair according to claim 1, characterized in that, S2 specifically refers to: taking the turning point of the trend of the preload driving force of the rolling linear guide pair as the dividing point, dividing the wear state of the rolling linear guide pair into running-in wear, stable wear and rapid wear in sequence; and determining its wear state based on the current trend of the change of the preload driving force of the rolling linear guide pair.

4. The method for identifying the wear condition of a rolling linear guide pair according to claim 1, characterized in that, S3 specifically includes: In the original vibration signal Add to Group Gaussian white noise The signal to be decomposed is obtained: According to the signal to be decomposed The maximum and minimum points are constructed using cubic spline interpolation. upper and lower envelopes and The local mean of the signal is: judge In its signal, whether the number of extreme points and zero-crossing points are equal or differ by at most one, and whether the mean of the upper and lower envelopes at all points is zero, if these conditions are met, then the first intrinsic mode component is obtained. Otherwise Repeat the above steps until the signal... The required conditions must be met.

5. The method for identifying the wear condition of a rolling linear guide pair according to claim 1, characterized in that, S4 specifically includes: For length of The original sequence A scaling factor was introduced. To establish a new coarse-grained vector, the formula used is: in, Given the length of the coarse-grained vector, the original sequence is divided into A length of coarse-grained sequences .

6. The method for identifying the wear condition of a rolling linear guide pair according to claim 1, characterized in that, S5 specifically includes: For the original feature set Standardization is performed to obtain a new sample set. The formula used is: in, The average value of the original feature set. The standard deviation of the original feature set; Introducing nonlinear mapping functions Map the sample set to a high-dimensional feature space In the middle, calculate its covariance matrix. The formula used is: in, For matrix eigenvalues, For matrix eigenvectors.

7. The method for identifying the wear condition of a rolling linear guide pair according to claim 1, characterized in that, S6 specifically includes the following steps: S601: Randomly generate M individuals as the initial population, and set the maximum number of iterations to N; S602: Input the training set into the random forest model to train the model. Use the accuracy of the wear state identification of the rolling linear guide pair as the fitness function. The formula used is: in, s Identify the correct number for the dataset. t The total number of datasets; S603: Find the individual corresponding to the maximum value of the fitness function Fitness, start iterative optimization, and determine whether the current fitness is better than the maximum fitness of the previous generation or whether the maximum number of evolutionary iterations of the population has been reached. If the condition is met, retain the optimal fitness value and its corresponding coordinates, and assign them to the initial coordinates; otherwise, return to step S602. S604: Outputs the optimal number of decision trees and the optimal minimum number of leaf nodes for the random forest algorithm, completing the construction of the wear condition identification model for rolling linear guide pairs.

Images