Method and system for classifying rolling bearing condition based on motif features of horizontal visual graph

Abstract

To provide a method and a system for classifying a rolling bearing state based on a motif feature of a horizontal visual graph.SOLUTION: Converting the vibration signals of each type of rolling bearing into a horizontal visual graph, labeling nodes of the horizontal visual graph according to a time sequence to obtain n-point motifs, for the n-point motifs with m motif types, calculating an appearance frequency of each type of n-point motif, and obtaining a feature vector based on the appearance frequencies of all types of n-point motifs, extracting a feature element of a bearing vibration signal of a current fault type based on the feature vector, obtaining feature elements of bearing vibration signals of all known rolling bearing fault types, obtaining bearing vibration signals to be classified, and selecting a known rolling bearing fault type corresponding to a maximum value of similarities as a fault type of the bearing vibration signal to be classified.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] (CROSS-REFERENCE TO RELATED APPLICATIONS) This application claims priority to a Chinese patent application, application number 202411116610.8, entitled "Method and system for classifying rolling bearing conditions based on motif features of horizontal visibility graphs," filed with the State Intellectual Property Office of the People's Republic of China on August 15, 2024, the entire contents of which are incorporated herein by reference and constitute a part of the present invention for all purposes. The present invention relates to the technical field of equipment condition monitoring and fault diagnosis, and in particular to a method and system for classifying rolling bearing conditions based on motif features of horizontal visibility graphs. [Background technology]

[0002] Rolling bearings are fundamental components that play a vital role in machinery and equipment. Failures in rolling bearings have a significant impact on the normal operation of the machinery and equipment, making condition monitoring and fault diagnosis of rolling bearings extremely important. When a rolling bearing fault occurs, the resulting vibration signal exhibits strong non-stationary, nonlinear characteristics. Therefore, quickly extracting fault features from complex bearing vibration signals is a key challenge in condition monitoring and fault diagnosis of rolling bearings.

[0003] In the prior art, time-domain and frequency-domain methods are only suitable for analyzing stationary linear signals, and approximate entropy methods can only obtain a single feature, and their operating parameters must be manually set subjectively. Visual graphs can convert nonlinear time series into complex networks, and by analyzing the features of these networks, the dynamic characteristics of rolling bearing vibration signals can be obtained.

[0004] Chinese patent document, patent publication number CN109100142B, patent title "Graph theory-based semi-supervised bearing fault diagnosis method" provides a graph theory-based semi-supervised bearing fault diagnosis method, which first uses a visible graph algorithm to convert the bearing's original vibration acceleration signal obtained by a sensor into a complex network, then calculates the complex network's structural parameters, extracts the mean value and standard deviation of the degree distribution and the network complexity index, and finally uses graph-based semi-supervised learning to process unlabeled samples to realize bearing fault diagnosis.

[0005] Chinese patent document, patent publication number CN116977708B, patent title "Method and system for intelligent bearing diagnosis based on adaptive aggregated visible graph" conducts typical fault vibration tests on bearings, collects vibration acceleration signals under typical fault conditions, uses an adaptive aggregated visible graph algorithm to perform feature extraction and graph mapping on the collected time series signals, obtains graph data composed of nodes and edges that is used as input for a graph neural network, builds an end-to-end rolling bearing intelligent diagnosis framework based on an improved DiffPool graph classification algorithm, inputs the graph data into the diagnosis framework, optimizes a rolling bearing intelligent diagnosis model, and uses the optimized rolling bearing intelligent diagnosis framework to identify and classify different operating modes of the rolling bearing.

[0006] The inventors discovered that the prior art has the following deficiencies: (1) Mean values ​​and standard deviations are only suitable for processing linear data. When a rolling bearing fault occurs, the degree distribution of the complex network corresponding to the vibration signal is highly nonlinear. Therefore, the mean value and standard deviation of the degree distribution are difficult to accurately reflect the essential characteristics of the bearing vibration signal. This method can only distinguish between different types of bearing faults, such as faults between the inner ring, outer ring, and rolling element of the bearing. It is difficult to simultaneously distinguish between different types and severities of bearing faults, and it is also difficult to distinguish between single and multiple bearing faults. (2) Semi-supervised learning classification algorithms have data imbalance problems and require sample labeling, resulting in low efficiency. (3) The calculation speed is slower, making them unsuitable for online applications. Summary of the Invention

[0007] In order to solve the deficiencies of the prior art, the present invention provides a method and system for classifying rolling bearing conditions based on motif features of horizontal visibility graphs. The method proposed in the present invention can fully extract the dynamic information contained in rolling bearing vibration signals, comprehensively express the inherent features of rolling bearing vibration signals, effectively distinguish between different types and severities of rolling bearing faults, and accurately distinguish between single and multiple bearing faults. The condition classification results have high accuracy and precision, are robust to abnormal data, have fast calculation speed, and are suitable for online applications.

[0008] In a first aspect, a method for classifying rolling bearing conditions based on motif features of a horizontal visibility graph is provided, and includes the steps of: (1) obtaining rolling bearing vibration signals of a plurality of known fault types; (2) converting each type of rolling bearing vibration signal into a horizontal visibility graph, and labeling the nodes of the horizontal visibility graph according to time series to obtain n-point motifs (where n and m are both positive integers greater than 2); (3) calculating the occurrence frequency of each type of n-point motif among m types of motifs of the n-point motifs, and obtaining a motif feature vector based on the occurrence frequency of all types of n-point motifs; and (4) calculating the motif feature vector. (5) repeating steps (2) to (4) to obtain characteristic elements of the bearing vibration signals of all known rolling bearing fault types; and (6) obtaining the bearing vibration signal to be classified, and performing steps similar to those (2) to (4) to obtain characteristic elements of the bearing vibration signal to be classified, calculating the similarity between the characteristic elements of the bearing vibration signal to be classified and the characteristic elements of the bearing vibration signals of all known rolling bearing fault types, and selecting the known rolling bearing fault type that corresponds to the maximum similarity as the fault type of the bearing vibration signal to be currently classified.

[0009] In a second aspect, a system for classifying rolling bearing conditions based on motif features of a horizontal visibility graph is provided, the system including: an acquisition module configured to acquire rolling bearing vibration signals of a plurality of known fault types; a conversion module configured to convert each type of rolling bearing vibration signal into a horizontal visibility graph, label the nodes of the horizontal visibility graph according to a time series, and obtain n-point motifs (where n and m are both positive integers greater than 2); a calculation module configured to calculate the occurrence frequency of each type of n-point motif for m types of motif types of n-point motifs, and obtain a motif feature vector based on the occurrence frequency of all types of n-point motifs; and a calculation module configured to classify the current fault based on the motif feature vector. the classification module is configured to acquire the bearing vibration signal to be classified, to obtain the feature elements of the bearing vibration signal to be classified through the operation process from the conversion module to the extraction module, to perform a similarity calculation between the feature elements of the bearing vibration signal to be classified and the feature elements of the bearing vibration signal of all known rolling bearing fault types, and to select the known rolling bearing fault type corresponding to the maximum value of the similarity as the fault type of the bearing vibration signal to be currently classified.

[0010] The above technical solution has the following advantages or effects: (1) The present invention uses the appearance frequency of motifs in a visible graph to represent the dynamic characteristics of rolling bearing vibration signals, thereby solving the defect of the prior art that the dynamic characteristics of rolling bearing vibration signals cannot be accurately represented by calculating the mean value and standard deviation of the degree distribution. (2) The present invention can effectively distinguish between different types and severities of rolling bearing faults, accurately identify single and multiple bearing faults, and is robust to abnormal data. It has good fault identification and classification effects, while prior art can only identify different types of bearing faults. (3) The present invention realizes automatic classification of bearing conditions by extracting a minimum number of motif features, avoiding problems such as data imbalance, sample labeling, and reduced efficiency that arise in classification using semi-supervised learning classification algorithms. (4) The method of the present invention has linear computational complexity, fast computation speed, and is suitable for online or offline applications, while the prior art has high computational complexity, slow computation speed, and cannot be used for online applications. [Brief explanation of the drawings]

[0011] The drawings in the specification that form a part of this invention are provided for a further understanding of the invention, and the illustrative examples of the invention and their descriptions are for the purpose of interpreting the invention and are not intended to unduly limit the invention.

[0012] [Figure 1] 1 is a flowchart of a method according to an embodiment of the present invention. [Figure 2(a)] 1 shows a bearing vibration signal under normal conditions. [Figure 2(b)] 1 shows a bearing vibration signal of an inner race premature failure. [Figure 2(c)] 1 shows a bearing vibration signal of an inner race fault. [Figure 2(d)] 1 shows a bearing vibration signal of an outer ring premature failure. [Figure 2(e)] 1 shows a bearing vibration signal of an outer ring fault. [Figure 2(f)] 1 shows a bearing vibration signal of a premature rolling element failure. [Figure 2(g)] 1 shows a bearing vibration signal of a rolling element fault. [Figure 2(h)] 1 shows a bearing vibration signal of premature cage failure. [Figure 2(i)] 1 shows a bearing vibration signal of a cage fault. [Figure 3] The results show the classification of nine types of bearing conditions using the relative frequencies of motif types labeled 1, 3, and 6, where NORM indicates a normal condition, DIF indicates premature inner ring failure, IF indicates inner ring failure, DOF indicates premature outer ring failure, OF indicates outer ring failure, DBF indicates premature rolling element failure, BF indicates rolling element failure, BCF indicates premature cage failure, and CF indicates cage failure. [Figure 4] The results show that the bearing conditions are classified into nine types using effective values, where NORM indicates a normal state, DIF indicates premature inner ring failure, IF indicates inner ring failure, DOF indicates premature outer ring failure, OF indicates outer ring failure, DBF indicates premature rolling element failure, BF indicates rolling element failure, BCF indicates premature cage failure, and CF indicates cage failure. [Figure 5] The results show that nine types of bearing conditions are classified using skewness, where NORM indicates a normal state, DIF indicates premature inner ring failure, IF indicates inner ring failure, DOF indicates premature outer ring failure, OF indicates outer ring failure, DBF indicates premature rolling element failure, BF indicates rolling element failure, BCF indicates premature cage failure, and CF indicates cage failure. [Figure 6] The results of classifying nine types of bearing conditions using approximate entropy, where NORM is normal, DIF is inner ring premature failure, IF is inner ring failure, DOF is outer ring premature failure, OF is outer ring failure, DBF is rolling element premature failure, BF is rolling element failure, BCF is cage premature failure, and CF is cage failure. [Figure 7(a)] 1 shows a bearing vibration signal under normal conditions. [Figure 7(b)] 1 shows a bearing vibration signal of an inner race fault. [Figure 7(c)] 1 shows a bearing vibration signal of a rolling element fault. [Figure 7(d)] 1 shows a bearing vibration signal of an outer ring fault. [Figure 7(e)] The bearing vibration signal of a combined fault (inner ring fault + outer ring fault) is shown. [Figure 8] The results show the classification of five types of bearing conditions using the relative frequencies of motif types labeled 1, 3, and 5. NORM indicates a normal condition, IF indicates an inner ring fault, BF indicates a rolling element fault, OF indicates an outer ring fault, and IF+OF indicates an inner ring fault plus an outer ring fault. [Figure 9] The results show that the bearing conditions are classified into five types using effective values, with NORM indicating a normal state, IF indicating an inner ring fault, BF indicating a rolling element fault, OF indicating an outer ring fault, and IF+OF indicating an inner ring fault plus an outer ring fault. [Figure 10] The results show that five types of bearing conditions are classified using skewness, where NORM indicates a normal state, IF indicates an inner ring fault, BF indicates a rolling element fault, OF indicates an outer ring fault, and IF+OF indicates an inner ring fault plus an outer ring fault. [Figure 11]The results show the classification of five types of bearing conditions using approximate entropy, where NORM indicates a normal state, IF indicates an inner ring fault, BF indicates a rolling element fault, OF indicates an outer ring fault, and IF+OF indicates an inner ring fault plus an outer ring fault. [Figure 12(a)] This is a motif figure corresponding to the first motif type. [Figure 12(b)] This is a motif figure corresponding to the second motif type. [Figure 12(c)] This is a motif figure corresponding to the third motif type. [Figure 12(d)] This is a motif figure corresponding to the fourth motif type. [Figure 12(e)] This is a motif figure corresponding to the fifth motif type. [Figure 12(f)] This is a motif figure corresponding to the sixth motif type. DETAILED DESCRIPTION OF THE INVENTION

[0013] The following detailed description is intended to be illustrative and to provide further explanation of the present invention. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0014] Example 1: As shown in FIG. 1, this example provides a method for classifying rolling bearing conditions based on motif features of horizontal visibility graphs, and includes the steps of: obtaining rolling bearing vibration signals of multiple known fault types in step S101; converting each type of rolling bearing vibration signal into a horizontal visibility graph; labeling the nodes of the horizontal visibility graph according to time series; obtaining n-point motifs (where n and m are positive integers greater than 2) in step S102; calculating the occurrence frequency of each type of n-point motif for m types of n-point motifs; and obtaining a motif feature vector based on the occurrence frequency of all types of n-point motifs in step S103; and The method includes step S104 of extracting characteristic elements of the bearing vibration signal of the current fault type based on the vector; step S105 of repeating steps S102 to S104 to obtain characteristic elements of the bearing vibration signal of all known rolling bearing fault types; and step S106 of acquiring the bearing vibration signal to be classified, obtaining characteristic elements of the bearing vibration signal to be classified by steps similar to S102 to S104, calculating the similarity between the characteristic elements of the bearing vibration signal to be classified and the characteristic elements of the bearing vibration signal of all known rolling bearing fault types, and selecting the known rolling bearing fault type corresponding to the maximum value of the similarity as the fault type of the bearing vibration signal to be currently classified.

[0015] Furthermore, the step S101 of obtaining rolling bearing vibration signals of a plurality of known fault types includes collecting the rolling bearing vibration signals of the plurality of known fault types by a sensor.

[0016] Furthermore, the step S102 of converting each type of rolling bearing vibration signal into a horizontal visual graph includes: i Let be the i-th node in the horizontal visibility graph, and the k-th node between the i-th node and the j-th node is

number

number

[0017] As can be seen, rolling bearing vibration data x k Calculate the horizontal visibility graph of N-point horizontal visibility graph HVG N where k=1,2,...,N, and N is the length of the vibration data.

[0018] Furthermore, the step S102 of labeling the nodes of the horizontal visibility graph according to the time sequence and obtaining the n-point motif is performed by labeling the nodes of the horizontal visibility graph according to the time sequence. N It involves labeling the nodes of, and an n-point motif denotes all substructures consisting of n consecutive nodes {s, s+1, …, s+n-1} on the horizontal visibility graph, where s = 1, …, N-n+1.

[0019] Furthermore, in step S103, the m kinds of motif types present in the n-point motif include the following: The m kinds of motif types are defined as follows, and any four consecutive data points

number

[0020]

number

[0021]

number

[0022]

number

[0023]

number

[0024]

number

[0025]

number

[0026] Furthermore, in step S103, the occurrence frequency of each kind of n-point motif is calculated, and m kinds of motif types exist in the n-point motif, and the occurrence frequency of the i-th kind of motif type is P i n Then,

number

number

[0027] Furthermore, in step S103, a motif feature vector is obtained based on the frequency of occurrence of all types of n-point motifs, and an m-dimensional feature vector is obtained based on the frequency of occurrence of m types of motifs.

number

[0028] Furthermore, in step S104, the motif feature vector Z n Based on this, the characteristic elements of the bearing vibration signal of the current fault type are extracted and the three-dimensional matrix MAT m×a×b Suppose that m represents the feature vector matrix generated by all data segments in all rolling bearing states, m represents the number of motif types, a represents the number of data segments, b represents the number of rolling bearing states, 1≦j≦m, 1≦i≦a, and the feature vector MAT corresponding to the i-th data segment and the j-th motif type is j×i×b The standard deviation of

number

number

number

[0029] Furthermore, in step S106, the bearing vibration signal to be classified is acquired, and characteristic elements of the bearing vibration signal to be classified are obtained by steps similar to S102 to S104. A similarity calculation is performed between the characteristic elements of the bearing vibration signal to be classified and the characteristic elements of the bearing vibration signals of all known rolling bearing fault types, and the known rolling bearing fault type corresponding to the maximum similarity is selected as the fault type of the bearing vibration signal to be currently classified, and cosine similarity is used for the similarity calculation.

[0030] Experiment 1: Different Types and Severity Classifications of Bearing Failure. This experiment uses rolling bearing vibration data containing different bearing failure types and severities. These bearing vibration data cover nine different bearing conditions: normal, premature inner ring failure, inner ring failure, premature rolling element failure, rolling element failure, premature outer ring failure, outer ring failure, premature cage failure, and cage failure. Figure 2(a) shows the bearing vibration signal for a normal condition, Figure 2(b) shows the bearing vibration signal for a premature inner ring failure, Figure 2(c) shows the bearing vibration signal for an inner ring failure, Figure 2(d) shows the bearing vibration signal for a premature outer ring failure, Figure 2(e) shows the bearing vibration signal for an outer ring failure, Figure 2(f) shows the bearing vibration signal for a premature rolling element failure, Figure 2(g) shows the bearing vibration signal for a rolling element failure, Figure 2(h) shows the bearing vibration signal for a premature cage failure, and Figure 2(i) shows the bearing vibration signal for a cage failure.

[0031] First, the bearing vibration signal is converted into a horizontal visibility graph, and then the four-point motif features of this horizontal visibility graph are analyzed. The comparison results of the sensitive parameters are shown in Table 1. Table 1 shows the average values ​​of the standard deviations for the relative frequencies of certain motif types in all data segments under all bearing conditions (different fault types and severities).

[0032] Table 1: Comparison results of sensitive parameters in Experiment 1 [Table 1]

[0033] As can be seen from Table 1, motif types labeled 1, 3, and 6 are sensitive to changes in bearing condition. Figure 3 shows the results of extracting the relative frequencies of these three motif types and classifying nine bearing conditions. As can be seen from Figure 3, the method of the present invention can clearly separate the nine bearing conditions. To evaluate the performance of the method, classification was performed using effective value, skewness, and approximate entropy. The results are shown in Figures 4, 5, and 6, respectively. As can be seen from Figures 4 to 6, these three methods cannot effectively distinguish the nine bearing conditions. Table 2 compares the computational efficiency of these four methods. The time cost of the method of the present invention is comparable to that of effective value and skewness, demonstrating that the computational speed of the method of the present invention is faster. Table 2 compares the computational efficiency of nine bearing vibration signals using the method of the present invention, effective value, skewness, and approximate entropy.

[0034] Table 2. Comparison of computational efficiency of the four methods in Experiment 1 [Table 2]

[0035] Experiment 2: Identifying Single and Multiple Bearing Faults. In this example, verification was performed using rolling bearing vibration data containing different single and multiple faults. These bearing vibration data included five types of bearing conditions: normal, inner ring fault, rolling element fault, outer ring fault, and inner and outer ring fault. As shown in Figures 7(a) to 7(e), these five types of bearing vibration data show short abnormal signals at the beginning of the normal bearing signal and the multiple fault signal. Figure 7(a) shows the bearing vibration signal in a normal state, Figure 7(b) shows the bearing vibration signal with an inner ring fault, Figure 7(c) shows the bearing vibration signal with a rolling element fault, Figure 7(d) shows the bearing vibration signal with an outer ring fault, and Figure 7(e) shows the bearing vibration signal with a multiple fault (inner ring fault + outer ring fault).

[0036] First, the bearing vibration signal is converted into a horizontal visibility graph, and then the four-point motif features of this horizontal visibility graph are analyzed, and the comparison results of the sensitive parameters are shown in Table 3. Table 3 shows the average values ​​of the standard deviations for the relative frequencies of certain motif types of all data segments in all bearing conditions (different single faults and compound faults).

[0037] Table 3. Comparison results of sensitive parameters in Experiment 2 [Table 3]

[0038] As can be seen from Table 3, motif types labeled 1, 3, and 5 are sensitive to changes in bearing condition. Figure 8 shows the results of extracting the relative frequencies of these three motif types and classifying five bearing conditions. As can be seen from Figure 8, the method of the present invention can clearly separate these five bearing conditions, and the method of the present invention has strong robustness against abnormal data. To evaluate the performance of the method, the five bearing conditions were classified using effective value, skewness, and approximate entropy, respectively. The results are shown in Figures 9, 10, and 11, respectively. As can be seen from Figures 9 to 11, these three methods were unable to effectively distinguish these five bearing conditions and were significantly affected by abnormal data. Table 4 compares the computational efficiency of the four methods. The time cost of the method of the present invention is comparable to that of the effective value, indicating that the method of the present invention has a faster computational speed. Table 4 compares the computational efficiency of five bearing vibration signals using the method of the present invention, effective value, skewness, and approximate entropy.

[0039] Table 4. Comparison of computational efficiency of the four methods in Experiment 2 [Table 4]

[0040] Based on the test results and analysis, it was found that the method of the present invention can effectively distinguish between different types and severities of equipment faults, can effectively distinguish between single faults and multiple faults, has robustness against abnormal data, good state identification and classification effects, fast calculation speed, and has obvious advantages over effective value, skewness, and approximate entropy.

[0041] This invention uses the frequency of occurrence of motifs in a visible graph to represent the dynamic characteristics of rolling bearing vibration signals, and proposes a new method for extracting vibration signal features of complex equipment. The method of the present invention can effectively distinguish between different types and severities of rolling bearing faults, accurately identify single and multiple bearing faults, and is robust to abnormal data, has good fault identification and classification effects, and has clear advantages over RMS, skewness, and approximate entropy. The method of the present invention has linear-order computational complexity, is comparable to the RMS value in computational complexity, and has fast computation speed.

[0042] Example 2: This example provides a system for classifying rolling bearing conditions based on motif features of a horizontal visibility graph, including an acquisition module configured to acquire rolling bearing vibration signals of multiple known fault types; a conversion module configured to convert each type of rolling bearing vibration signal into a horizontal visibility graph, label the nodes of the horizontal visibility graph according to time series, and obtain n-point motifs (where n and m are positive integers greater than 2); a calculation module configured to calculate the occurrence frequency of each type of n-point motif for m types of n-point motifs, and obtain a motif feature vector based on the occurrence frequency of all types of n-point motifs; and a calculation module configured to calculate the occurrence frequency of each type of n-point motif for m types of n-point motifs based on the motif feature vector. the fault type of the bearing vibration signal to be classified; an extraction module configured to extract feature elements of the bearing vibration signal of the fault type; a repetition module configured to repeat the operation process from the conversion module to the extraction module to obtain feature elements of the bearing vibration signal of all known rolling bearing fault types; and a classification module configured to acquire the bearing vibration signal to be classified, obtain feature elements of the bearing vibration signal to be classified through the operation process from the conversion module to the extraction module, calculate similarities between the feature elements of the bearing vibration signal to be classified and the feature elements of the bearing vibration signal of all known rolling bearing fault types, and select the known rolling bearing fault type corresponding to the maximum similarity as the fault type of the bearing vibration signal to be currently classified.

[0043] Example 3 This embodiment further provides an electronic device including one or more processors, one or more memories, and one or more computer programs, wherein the processor is connected to the memory, and the one or more computer programs are stored in the memory, and when the electronic device is operated, the processor executes the one or more computer programs stored in the memory, thereby causing the electronic device to perform the method described in the above embodiment 1.

[0044] Example 4 This embodiment further provides a computer-readable storage medium for storing computer instructions that, when executed by a processor, complete the method of embodiment 1.

[0045] The above description is only a preferred embodiment of the present invention, and does not limit the present invention. Those skilled in the art can make various modifications and changes to the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. (1) acquiring rolling bearing vibration signals of a plurality of known fault types; (2) converting each type of rolling bearing vibration signal into a horizontal visibility graph, labeling the nodes of the horizontal visibility graph according to time series, and obtaining n-point motifs (where n and m are both positive integers greater than 2); (3) calculating the frequency of occurrence of each type of n-point motif for m types of motif types, and obtaining a motif feature vector based on the frequency of occurrence of all types of n-point motifs; (4) extracting feature elements of bearing vibration signals of the current fault type based on the motif feature vector; (5) repeating steps (2) to (4) to obtain characteristic elements of bearing vibration signals for all known rolling bearing fault types; (6) acquiring a bearing vibration signal to be classified, obtaining characteristic elements of the bearing vibration signal to be classified by steps similar to those of (2) to (4), calculating similarities between the characteristic elements of the bearing vibration signal to be classified and characteristic elements of bearing vibration signals of all known rolling bearing fault types, and selecting the known rolling bearing fault type corresponding to the maximum similarity as the fault type of the bearing vibration signal to be currently classified; The step of extracting feature elements of bearing vibration signals of a current fault type based on the motif feature vector includes: three-dimensional matrix MAT ｍ×ａ×ｂ Suppose j represents the feature vector matrix generated by all data segments in all rolling bearing states, m represents the number of motif types, a represents the number of data segments, b represents the number of rolling bearing states, 1≦j≦m, 1≦i≦a; The feature vector MAT corresponding to the jth motif type in the i-th data segment ｊ×ｉ×ｂ The standard deviation of [Equation 1] [Equation 2] Calculate by The average standard deviation of all a-segment data corresponding to the j-th motif type is [Equation 3] Calculate by Select the c motif types with the highest average frequency as the minimum number of feature elements. A computer-implemented method for classifying rolling bearing conditions based on motif features of horizontal visibility graphs, comprising:

2. The step of converting each type of rolling bearing vibration signal into a horizontal visual graph includes: The i-th data x of the rolling bearing vibration signal S ｉ Let be the i-th node in the horizontal visibility graph, and the k-th node located between the i-th node and the j-th node is [Equation 4] The i-th node and the j-th node are connected if ｊ represents the j-th data of the rolling bearing vibration signal S, [Equation 5] 2. The method for classifying rolling bearing conditions based on motif features of horizontal visibility graphs according to claim 1, wherein S is real-valued data of length N, and the horizontal visibility graph is a plan view containing N nodes.

3. The step of labeling the nodes of the horizontal visibility graph according to the time sequence and obtaining n-point motifs is as follows: HVG in chronological order Ｎ , N-n+1}, and an n-point motif refers to any substructure consisting of n consecutive nodes {s, s+1, ..., s+n-1} on the horizontal visibility graph, where s=1, ..., N-n+1.

4. The m motif types present in the n-point motif include: The m motif types are defined as follows, and any four consecutive data points [Equation 6] In this case, x ｌ represents the l-th data point, and x ｌ＋１ represents the l+1th data point, and x ｌ＋２ represents the l+2th data point, and x ｌ＋３ represents the l+3th data point, l represents a positive integer and 1≦l≦N−3, and N represents the data length; The inequality corresponding to the first kind of motif type is [Equation 7] and The inequality corresponding to the second motif type is [Equation 8] and The inequality corresponding to the third motif type is [Equation 9] and The inequality corresponding to the fourth motif type is [Equation 10] and The inequality corresponding to the fifth motif type is [0011] and The inequality corresponding to the sixth motif type is [0012] The method for classifying rolling bearing conditions based on motif features of horizontal visibility graphs according to claim 3, characterized in that:

5. Calculate the frequency of occurrence of each type of n-point motif, There are m kinds of motif types in the n-point motif, and the occurrence frequency of the i-th motif type is P ｉ ｎ Then, [0013] The occurrence frequency of the i-th motif type P ｉ ｎ is as follows: [0014] Here, q ｉ The method for classifying rolling bearing conditions based on motif features of horizontal visibility graphs according to claim 4, characterized in that: represents the number of occurrences of the i-th motif type.

6. Based on the frequency of occurrence of all kinds of n-point motifs, a motif feature vector is obtained, and based on the frequency of occurrence of m kinds of motif types, an m-dimensional feature vector is obtained. [Equation 15] The method for classifying rolling bearing conditions based on motif features of horizontal visibility graphs according to claim 5, characterized in that:

7. 2. The method for classifying rolling bearing conditions based on motif features of horizontal visibility graphs according to claim 1, characterized in that: a bearing vibration signal to be classified is acquired; feature elements of the bearing vibration signal to be classified are obtained by steps similar to those of (2) to (4); a similarity calculation is performed between the feature elements of the bearing vibration signal to be classified and the feature elements of the bearing vibration signals of all known rolling bearing fault types; the known rolling bearing fault type corresponding to the maximum similarity is selected as the fault type of the bearing vibration signal to be currently classified; and a cosine similarity is adopted for the similarity calculation.

8. 2. The method for classifying rolling bearing conditions based on motif features of horizontal visibility graphs according to claim 1, wherein the step of acquiring rolling bearing vibration signals of a plurality of known fault types includes collecting the rolling bearing vibration signals of the plurality of known fault types by a sensor.

9. an acquisition module configured to acquire rolling bearing vibration signals of a plurality of known fault types; a transformation module configured to transform each type of rolling bearing vibration signal into a horizontal visibility graph, label the nodes of the horizontal visibility graph according to time series, and obtain n-point motifs, where n and m are both positive integers greater than 2; A calculation module configured to calculate the frequency of occurrence of each type of n-point motif for m types of motif types, and obtain a motif feature vector based on the frequency of occurrence of all types of n-point motifs; an extraction module configured to extract feature elements of a bearing vibration signal of a current fault type based on the motif feature vector; a repeating module configured to repeat the process from the conversion module to the extraction module to obtain characteristic elements of bearing vibration signals of all known rolling bearing fault types; a classification module configured to acquire a bearing vibration signal to be classified, obtain feature elements of the bearing vibration signal to be classified through an operation process from the conversion module to the extraction module, perform a similarity calculation between the feature elements of the bearing vibration signal to be classified and feature elements of bearing vibration signals of all known rolling bearing fault types, and select the known rolling bearing fault type corresponding to the maximum similarity as the fault type of the bearing vibration signal to be currently classified; Extracting feature elements of bearing vibration signals of current fault types based on the motif feature vector includes: three-dimensional matrix MAT ｍ×ａ×ｂ Suppose j represents the feature vector matrix generated by all data segments in all rolling bearing states, m represents the number of motif types, a represents the number of data segments, b represents the number of rolling bearing states, 1≦j≦m, 1≦i≦a; The feature vector MAT corresponding to the jth motif type in the i-th data segment ｊ×ｉ×ｂ The standard deviation of [0016] [Equation 17] Calculate by The average standard deviation of all a-segment data corresponding to the j-th motif type is [Equation 18] Calculate by The occurrence frequency of c motif types with the maximum average value is selected as the minimum number of feature elements. A classification system for rolling bearing conditions based on motif features of horizontal visibility graphs, characterized by:

Images