Zero sample prototype extension method and system for relieving hubness problem and composite fault diagnosis

Abstract

The invention relates to a zero sample prototype extension method and system for relieving hubness and composite fault diagnosis, and belongs to the field of mechanical fault diagnosis. State signal samples of bearings of different types and different working conditions are collected through a bearing fault test bench, corresponding fault attribute prototypes are designed, prototype extension kernel function mapping prototypes are adopted for the designed fault attribute prototypes to a high-dimensional space, the distance between the designed attribute prototypes is enlarged, a neural network feature extractor is designed, and the fault attribute prototypes are extracted. And inputting a single fault signal sample and a corresponding fault prototype for training, inputting a composite fault signal into a trained feature extractor to obtain an attribute feature, deducing a corresponding composite fault type according to a distance, and evaluating a hubness problem by adopting a designed measurement method. According to the method, the hubness problem can be effectively relieved, the accuracy of composite fault diagnosis is improved, and the rolling bearing fault diagnosis of the composite fault can be realized only through single fault data.

Description

Technical Field

[0001] The invention belongs to the field of mechanical fault diagnosis, and relates to a zero-sample prototype expansion method and system for alleviating hubness problems and complex fault diagnosis. Background Art

[0002] Rotating machinery is widely used in wind power generation, aerospace, transportation and other fields, and plays a vital role in modern industry. Bearings, as one of the key components of rotating machinery, are crucial to the performance and reliability of the equipment. However, bearings are prone to failure due to long-term operation under complex working conditions. With the increasing complexity of rotating machinery, in addition to single faults, machinery will also experience compound faults, and the diagnosis of compound faults is more challenging than single faults. Therefore, it is of great significance to study high-precision diagnostic methods for bearings and other components.

[0003] Existing composite fault diagnosis methods can be divided into three categories: methods based on fault mechanism, methods based on signal processing technology, and methods based on artificial intelligence. The methods based on fault mechanism mainly use dynamics and modal analysis to understand the mechanism of composite faults and assist in formulating effective maintenance strategies. The methods based on signal processing mainly use signal processing technology to extract composite fault information and judge the fault type through expert experience. However, the fault mechanism and signal processing methods rely heavily on expert experience, which limits their application in actual industry. In recent years, artificial intelligence-based methods have attracted widespread attention in various fields due to their excellent performance. Artificial intelligence-based diagnosis methods are mainly divided into traditional machine learning methods and deep learning-based methods. Although traditional machine learning and deep learning-based methods achieve high diagnostic accuracy, they rely on a large amount of fault data. However, it is unrealistic to obtain a large amount of fault data, especially composite fault data, in actual industry. In recent years, Lampert et al. developed a zero-shot learning method for image recognition to overcome the difficulty of target data collection. The method based on zero-shot learning does not require target samples to participate in training. At present, zero-shot learning methods have also been extended to the field of fault diagnosis. Feng et al. first applied zero-shot learning to fault diagnosis, Qin et al. proposed a zero-shot attribute consistency model for fault diagnosis under unknown working conditions, and Hu et al. constructed a semantically consistent embedding model for industrial zero-shot fault diagnosis.

[0004] However, existing zero-shot diagnosis methods have the following problems:

[0005] 1) Current zero-shot diagnosis methods directly map the signal feature space to the attribute prototype space and use the nearest neighbor estimation as the metric. This approach is susceptible to the hubness problem, that is, samples of some prototypes appear in the nearest neighbor estimates of other prototypes, which has a negative impact on the accuracy of fault diagnosis.

[0006] 2) In previous work, the defined attributes only consider the fault type and frequency characteristics, lacking consideration of other typical fault characteristics. In addition, the existing hubness problem measurement methods target high-dimensional data and are not suitable for zero-shot learning in some cases.

[0007] The Hubness problem is a phenomenon in machine learning, which refers to the fact that in some data sets, some sample points are more likely to be misclassified as their neighbors by other sample points. These sample points are like "central points" or "hubs" in the data space, attracting a large number of other sample points, causing them to be easily misclassified in algorithms such as nearest neighbor classification.

[0008] Hubness problem occurs because of:

[0009] Uneven data distribution: In the dataset, the number of samples in some categories is much larger than that in other categories, which makes the samples in these categories more likely to be misclassified as neighbors of other categories.

[0010] High dimension of feature space: In high-dimensional space, the distance between sample points becomes more difficult to calculate and Hubness problems are more likely to occur.

[0011] Inappropriate similarity measurement: Using an inappropriate similarity measurement method, such as Euclidean distance, may lead to the Hubness problem.

[0012] Impact of Hubness Problem:

[0013] Reduced classification accuracy: The Hubness problem will cause the accuracy of algorithms such as nearest neighbor classification to decrease, because it is easy to misclassify as a neighbor of the Hubness sample.

[0014] Affecting clustering results: The Hubness problem will affect the performance of the clustering algorithm because Hubness samples may be incorrectly clustered into clusters of other categories.

[0015] Increased computational complexity: In order to solve the Hubness problem, more complex algorithms or data preprocessing are required, which increases computational complexity.

[0016] In order to alleviate the hubness problem, the present invention proposes a prototype expansion mapping method. The method designs a prototype expansion kernel function to map linear and nonlinear prototype features to a relatively high-dimensional space to enhance separability. The prototype expansion kernel function can increase the distance between prototypes, thereby alleviating the hubness problem and improving diagnostic accuracy. In view of the problem that some important vibration features are ignored in previous attribute definitions, the present invention designs a new attribute prototype definition method to add more fault features to the attributes. In addition, the present invention also proposes a measurement method for the hubness problem in zero-sample learning. Summary of the invention

[0017] In view of this, an object of the present invention is to provide a zero-sample prototype expansion method and system for alleviating hubness problems and complex fault diagnosis.

[0018] In order to achieve the above object, the present invention provides the following technical solutions:

[0019] like Figure 1 As shown, a zero-shot prototype extension method for alleviating hubness problems and composite fault diagnosis includes the following steps:

[0020] S1: Collect state signal samples of bearings of different types and working conditions through the bearing fault test bench;

[0021] S2: Design corresponding fault attribute prototypes for various bearing fault signals based on the characteristics of bearing fault type, signal fault frequency and fault amplitude;

[0022] S3: For the designed fault attribute prototype, the prototype expansion kernel function is used to map the prototype to a high-dimensional space to expand the distance between the designed attribute prototypes;

[0023] S4: Design a neural network feature extractor and input single fault signal samples and corresponding fault prototypes for training;

[0024] S5: input the composite fault signal into the trained feature extractor to obtain attribute features;

[0025] S6: The nearest neighbor estimation method is used to calculate the distance between the extracted composite fault attribute features and the expanded fault attribute prototype features, and the corresponding composite fault type is inferred based on the distance;

[0026] S7: Use the designed metric method to evaluate the hubness problem.

[0027] Furthermore, in S2, the attribute definition method designed for rolling bearings includes nine attributes: fault type, impact frequency, and impact amplitude. The attributes designed for rolling bearings are shown in Table 1. Because the data of a single fault can be directly obtained, its corresponding attribute prototype P S can be defined directly. The attribute prototype P of the composite fault C It is necessary to define it through the logical operation of single fault attributes. The attributes related to fault type and impact frequency are obtained through logical OR operation:

[0028]

[0029] in 1 to n elements representing the ith prototype of a single fault.

[0030] The properties related to the fault impulse amplitude are obtained through logical OR operation:

[0031]

[0032] Among them, D A represents the dimension of the attribute, Indicates the n+1~Dth of the wth composite fault prototype A elements, Indicates the n+1~Dth of the i-th single fault prototype A elements;

[0033] Table 1

[0034]

[0035]

[0036] Furthermore, S3 proposes a prototype extension kernel function, which can effectively improve the distance between prototypes and alleviate the hubness problem. The specific definition of the function is as follows:

[0037]

[0038] Among them, γ1,γ2 are the weight parameters of the two parts of the kernel function, c is a constant term, d is the degree of the polynomial, γ1,γ2≥1,d∈N+,c≥0.

[0039] Furthermore, the proof that the function mentioned in S3 is a kernel function is as follows. First, the symmetry of the function is proved:

[0040]

[0041] Next, we prove the positive definiteness of the function, that is, to prove that for any non-zero vector There are a TKa≥0, where K is the kernel matrix of the proposed function, as follows:

[0042]

[0043] According to the properties of linear kernel functions and polynomial kernel functions:

[0044]

[0045] Obviously, when γ1,γ2≥1:

[0046] a T Ka≥0(8)

[0047] In summary, the feasibility of the designed function as a kernel function is verified.

[0048] Furthermore, in S3, the relevant proof that the proposed kernel function can expand the distance between prototypes is as follows: Let P1 = [α1, α2, …, α m ] and P2=[β1,β2,…,β m ] represent the prototypes of any two faults respectively. After being mapped by the kernel function, the prototypes can be expressed as:

[0049]

[0050]

[0051] The Euclidean distance between two pre-mapping prototypes is:

[0052]

[0053] The Euclidean distance between two mapped prototypes is:

[0054]

[0055] The difference between the two types of distances can be calculated:

[0056]

[0057] When P1 and P2 are not zero vectors at the same time, we have:

[0058]

[0059] When α i ,β i ∈{0,1}:

[0060]

[0061] Therefore, according to the above analysis, the proposed kernel function can expand the distance between prototypes.

[0062] Furthermore, S4 designed a convolutional neural network as a feature extractor, which mainly consists of four convolutional modules, a flattening layer, and three fully connected layers. The first convolutional module includes a wide kernel convolutional layer, a batch normalization layer, and a maximum pooling layer. The second to fourth convolutional modules have the same configuration, including a normal convolutional layer, an expanded convolutional layer, a batch normalization (BN) layer, and a maximum pooling layer. The nonlinear rectified linear unit (ReLU) is used to activate all network layers. The specific parameters of the feature extractor are as follows: Figure 2 Take the fourth convolutional layer as an example, where (750,32) represents the input shape of the feature, 3-1-64-2 represents the convolution kernel size of 3, the stride of 1, the number of channels of 64, and the dilation rate of 2.

[0063] Furthermore, during the training of the model in S4, the mean square error loss is used to align the deep features and the mapping attribute prototypes:

[0064]

[0065] in, Represents the jth element of the attribute feature of the i-th sample, Represents the jth element of the attribute prototype of the i-th sample mapped by the kernel function, N S is the number of samples of a single fault, and m is the dimension of the design prototype.

[0066] Furthermore, the nearest neighbor estimation is used in S6 to infer the type of fault to be diagnosed:

[0067]

[0068] Among them, y c A label indicating a compound fault, Represents the attribute prototype matrix M mapped by the kernel function c The i-th element of the b-th row of .

[0069] Furthermore, the evaluation indicators for the hubness problem in zero-shot designed in S7 are as follows:

[0070]

[0071] Where N P is the number of prototypes tested, N k (i) indicates the number of test samples belonging to the i-th prototype that appear in the k nearest neighbors of other prototypes. The larger the Hub value, the more serious the Hubness problem.

[0072] The beneficial effects of the present invention are:

[0073] (1) The present invention can realize rolling bearing fault diagnosis of complex faults only by using single fault data.

[0074] (2) The model constructed by the present invention has higher diagnostic accuracy than existing classical and cutting-edge zero-sample models.

[0075] (3) The prototype expansion kernel function proposed in the present invention can effectively amplify the distance between prototypes, alleviate the hubness problem, and thus improve the diagnostic accuracy.

[0076] (4) The present invention proposes a measurement method for the hubness problem, which is more suitable for zero-sample learning than the original measurement method.

[0077] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0079] Figure 1 Extending the diagnostic process of the method for zero-shot prototyping;

[0080] Figure 2 is the detailed parameters of the feature extractor;

[0081] Figure 3 The diagnostic accuracy of different methods under different loads in the NEEPU experiment

[0082] Figure 4 is the diagnostic accuracy of different methods under different loads in the BCF experiment

[0083] Figure 5 Diagnostic accuracy of the proposed model in different tasks; (a) 20Hz→25Hz; (b) 25Hz→20Hz; (c) 25Hz→30Hz; (d) 30Hz→25Hz;

[0084] Figure 6 Hub values ​​in NEEPU and BCF bearing experiments for different methods. DETAILED DESCRIPTION

[0085] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. In this specification, various details can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. In the absence of conflict, the features in the following embodiments and embodiments can be combined with each other.

[0086] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0087] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0088] The proposed zero-sample prototype expansion model is evaluated through three experiments on composite fault diagnosis of rolling bearings under different working conditions. The first experiment is based on the bearing dataset NEEPU collected by Northeast Electric Power University. The second experiment is verified using the self-made bearing composite fault experimental platform data (BCF). The third experiment is verified based on the dataset (HUST) of Huazhong University of Science and Technology.

[0089] The NEEPU dataset includes 7 different types of health states: normal state NC, outer ring fault OF, inner ring fault IF, rolling element fault BF, outer and inner ring fault OI, rolling element and outer ring fault BO, inner ring and ball fault IB. Various loads are applied to the bearing through a magnetic brake, including 0N·m (load 0), 0.1N·m (load 1), 0.2N·m (load 2) and 0.3N·m (load 3). The speed range is [1443,1478]rpm, and the sampling frequency is 12kHz. The number of samples for each state in the training set is 700. The number of samples for each state in the test set is 300. That is, the number of fault samples is 2800, and the number of compound fault samples in the test set is 900. In the proposed model, only single faults are involved in training, while compound faults are used for testing. The length of each sample is taken as 3072 to obtain more state information, and the sliding window sampling technique is used to obtain additional samples. Seven classic and cutting-edge zero-shot learning methods, including DAP, IAP, SJE, ESZSL, DEM, FDAT, and SCE, are compared to show the advantages of the proposed method. In addition, ablation experiments are performed to verify the proposed prototype expansion method, and the model without prototype expansion is recorded as ZSPE-WE.

[0090] Figure 3 The diagnostic accuracy of different zero-shot methods under four different loads is shown. Without the participation of compound faults in training, the average diagnostic accuracy of the ZSPE model for compound faults under four loads is 69.31%, 82.58%, 86.56% and 83.93%, respectively. The proposed model has higher accuracy than the other seven classic and advanced zero-shot learning models. Compared with the ZSPE-WE method, this method has a better improvement in accuracy, which verifies the effectiveness of the proposed prototype expansion method.

[0091] In order to further verify the generalization of the model, a zero-sample diagnosis experiment was carried out on a self-made bearing composite fault test bench. The platform is mainly composed of a load motor, a speed-torque meter, a planetary gear reducer, a bearing test module and a drive motor. The experimental data set includes 7 types of states of rolling bearings under 4 working conditions, which is the same as the NEEPU experiment. The type of the test bearing is SKF61908. Four loads are applied through the load motor, namely 0N·m (load 0), 2N·m (load 1), 4N·m (load 2), and 6N·m (load 3). The speed is set to 500rpm and the sampling frequency is 12kHz.

[0092] The average diagnostic accuracy of different zero-shot learning methods in the BCF experiment is as follows: Figure 4When only single fault data is involved in training, the average diagnostic accuracy of load 0 to load 3 composite faults is 65.87%, 75.64%, 96.71% and 87.33% respectively. Figure 4 It can be seen that the diagnostic accuracy of ZSPE is higher than that of other methods and ZSPE-WE. The accuracy of diagnosis further verifies the effectiveness of the proposed method and prototype extension module. In addition, the diagnostic accuracy of rolling bearing composite faults under different loads is acceptable, with the highest average accuracy reaching 96.71%, indicating that this method has great application potential in actual bearing zero-sample composite fault diagnosis.

[0093] In order to further verify the generalization ability of the model, the present invention conducted cross-domain zero-sample diagnosis experiments for multiple tasks based on the HUST bearing dataset. In these experiments, only data under known working conditions were used for training, and data under unknown working conditions were directly tested. A mechanical fault simulator was used to collect bearing fault data sets. The simulator mainly consists of a speed controller, a motor, a shaft, an acceleration sensor, a test bearing, and a data acquisition board. In this experiment, five rolling bearings in different states were used, including normal state, outer ring fault, inner ring fault, ball fault, and outer ring and inner ring combined fault. This experiment uses three different speeds of 20Hz, 25Hz, and 30Hz, and the sampling frequency is set to 25.6kHz.

[0094] In four different cross-domain zero-shot diagnosis experiments, the average diagnostic accuracy of the model was 92.53% ± 1.14, 92.87% ± 2.15, 97.69% ± 1.06, and 97.68% ± 0.88, respectively. It can be seen that in all four cases, the diagnostic accuracy exceeded 90%. Among them, the diagnostic accuracy in two cases exceeded 97%, which further proved the generalization performance of the proposed model. In addition, the confusion matrix obtained from the four types of fault diagnosis experiments is shown in Figure 2. Figure 5 As shown, (a) is 20Hz→25Hz; (b) is 25Hz→20Hz; (c) is 25Hz→30Hz; (d) is 30Hz→25Hz; which further verifies the performance of the proposed model.

[0095] In addition, in the two experiments of NEEPU and BCF, the proposed metric (Equation (18)) is used to measure the hubness problem to illustrate the effectiveness of the proposed method in alleviating the hubness problem. Figure 6 As shown in the figure, the Hub value of ZSPE-WE is much higher than that of ZSPE in all eight working conditions of the two experiments. The results show that the proposed method reduces the number of occurrences of the corresponding sample features in the nearest neighbor estimation of other prototypes. It can be concluded that the proposed method can identify the features corresponding to the correct prototype, thereby improving the accuracy of diagnosis.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A zero-shot prototype extension method for alleviating hubness problems and complex fault diagnosis, characterized by: The following steps are involved: S1: Collect state signal samples of bearings of different types and working conditions through the bearing fault test bench; S2: Design corresponding fault attribute prototypes for various bearing fault signals based on the characteristics of bearing fault type, signal fault frequency and fault amplitude; S3: For the designed fault attribute prototype, the prototype expansion kernel function is used to map the prototype to a high-dimensional space to expand the distance between the designed attribute prototypes; S4: Design a neural network feature extractor and input single fault signal samples and corresponding fault prototypes for training; S5: input the composite fault signal into the trained feature extractor to obtain attribute features; S6: The nearest neighbor estimation method is used to calculate the distance between the extracted composite fault attribute features and the expanded fault attribute prototype features, and the corresponding composite fault type is inferred based on the distance; S7: Use the designed metric method to evaluate the hubness problem.

2. The zero-sample prototype extension method for alleviating hubness problems and complex fault diagnosis according to claim 1, characterized in that: In said S2, the attribute prototype includes attributes of fault type, impact frequency and impact amplitude; Prototype P of compound fault C The prototype P of a single fault S The logical operation is: The properties related to the fault type and impact frequency are obtained through logical OR operation: in, represents the 1st to nth elements of the wth composite fault prototype, Represents the 1st to nth elements of the i-th single fault prototype; The properties related to the fault impulse amplitude are obtained through logical OR operation: Among them, D A represents the dimension of the attribute, Indicates the n+1~Dth of the wth composite fault prototype A elements, Indicates the n+1~Dth of the i-th single fault prototype A elements.

3. The zero-sample prototype extension method for alleviating hubness problems and complex fault diagnosis according to claim 1, characterized in that: In S3, the prototype expansion kernel function is defined as: Among them, γ1, γ2 are the weight parameters of the two parts of the kernel function, c is a constant term, d is the degree of the polynomial, γ1, γ2 ≥ 1, d∈N+, c ≥ 0, and N+ represents a positive integer.

4. The zero-sample prototype extension method for alleviating hubness problems and complex fault diagnosis according to claim 1, characterized in that: In S4, the neural network feature extractor is a convolutional neural network; the convolutional neural network consists of four convolutional modules, a flattening layer and three fully connected layers; the first convolutional module includes a wide kernel convolutional layer, a batch normalization layer and a maximum pooling layer; the second to fourth convolutional modules have the same configuration, including a normal convolutional layer, an extended convolutional layer, a batch normalization BN layer and a maximum pooling layer; The nonlinear rectified linear unit ReLU is used to activate all network layers; during the training of the convolutional neural network, the mean square error loss is used to align the deep features and the mapping attribute prototypes: in, Represents the jth element of the attribute feature of the i-th sample, Represents the jth element of the attribute prototype of the i-th sample after being mapped by the kernel function, N S is the number of samples of a single fault, and m is the dimension of the design prototype.

5. The zero-sample prototype extension method for alleviating hubness problems and complex fault diagnosis according to claim 1, characterized in that: In S6, the nearest neighbor estimation is used to infer the type of fault to be diagnosed: Among them, y c A label indicating a compound fault, Represents the attribute prototype matrix M mapped by the kernel function c The i-th element of the b-th row of .

6. The zero-sample prototype extension method for alleviating hubness problems and complex fault diagnosis according to claim 1, characterized in that: In S7, the evaluation index of the hubness problem is defined as: Among them, N P is the number of prototypes tested, N k (i) represents the number of test samples belonging to the i-th prototype that appear in the k nearest neighbors of other prototypes. The larger the Hub value, the more serious the Hubness problem.

7. A system for alleviating hubness problems and diagnosing compound faults, characterized by: include: A signal acquisition module, used to collect state signal samples of the bearing; The attribute prototype design module is used to design the corresponding fault attribute prototype according to the characteristics of the bearing fault type, signal fault frequency and fault amplitude; The prototype expansion module is used to map the designed fault attribute prototype to a high-dimensional space using the prototype expansion kernel function to expand the distance between the designed attribute prototypes; The feature extraction module is used to design a neural network feature extractor and input a single fault signal sample and the corresponding fault prototype for training; The diagnosis module is used to input the composite fault signal into the trained feature extractor to obtain the attribute feature, and use the nearest neighbor estimation method to calculate the distance between the extracted composite fault attribute feature and the expanded fault attribute prototype feature, and infer the corresponding composite fault type according to the distance; The evaluation module is used to evaluate the hubness problem using the designed measurement method.

8. The system for alleviating hubness problems and diagnosing complex faults according to claim 7, characterized in that: The attribute prototype design module further includes a fault type identification unit, a fault frequency extraction unit and a fault amplitude extraction unit; All three modules receive bearing status signal samples from the signal acquisition module; The fault type identification unit is responsible for identifying the type of bearing fault, the fault frequency extraction unit is responsible for extracting the fault frequency characteristics of the bearing fault signal, and the fault amplitude extraction unit is responsible for extracting the fault amplitude characteristics of the bearing fault signal; These three modules pass their respective outputs to the attribute prototyping module; The attribute prototype design module designs the corresponding fault attribute prototype according to the output results of these three modules.

9. The system for alleviating hubness problems and diagnosing complex faults according to claim 7, characterized in that: The prototype extension module further includes a kernel function calculation unit and a mapping unit; Both modules receive fault attribute prototypes from the attribute prototyping module; The kernel function calculation unit is responsible for calculating the kernel function value between attribute prototypes according to the preset kernel function formula; The mapping unit receives the kernel function values ​​from the kernel function calculation unit and uses these values ​​to map the attribute prototype to a high-dimensional space; The mapping unit outputs the expanded attribute prototype to the feature extraction module.

10. The system for alleviating hubness problems and diagnosing complex faults according to claim 7, characterized in that: The evaluation module further includes a hubness problem calculation unit and an output unit; Both modules receive the fault type inference results from the diagnosis module; The hubness problem calculation unit is responsible for inferring the results based on the fault type and calculating the value of each test sample in the hubness problem metric; The output unit receives the calculation result from the Hubness problem calculation unit and outputs the evaluation result of the hubness problem.

Images