Vibrating screen cross-domain fault diagnosis method and system capable of enhancing domain self-adaption

Abstract

The invention discloses a vibrating screen cross-domain fault diagnosis method for enhancing domain self-adaption. The method comprises the following steps: dividing obtained multi-dimensional comprehensive fault data sets of vibrating screens in different fields into a source domain feature training set, a target domain feature training set and a target domain feature test set; mapping the source domain feature training set and the target domain feature training set to a random feature space for joint distribution difference processing so as to execute domain and class alignment of the features; performing domain feature discrimination processing on the source domain feature training set to enhance feature intra-class tightness and inter-class separability; and according to a joint distribution difference processing result and a domain feature discrimination processing result, a vibrating screen cross-domain fault diagnosis model is constructed based on an enhanced domain adaptive increment random vector function chain network, and is used for predicting a fault category of an unknown label of a target domain. According to the method, the challenge of domain offset is effectively overcome, and the performance and generalization ability of the vibrating screen fault diagnosis model are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of vibration screen fault diagnosis, and in particular relates to a vibration screen cross-domain fault diagnosis method and system with enhanced domain adaptation. Background Art

[0002] As a key equipment in the industrial sorting process, the stability of the operation of the vibrating screen is related to production efficiency and product quality. Since they often work continuously for a long time in harsh working environments and bear complex loads under variable working conditions, vibrating screens will inevitably have various faults. If these faults are not diagnosed and handled in a timely manner, they may lead to a decline in equipment performance and even cause safety accidents. Therefore, effective fault diagnosis of vibrating screens to ensure their normal operation and extend their service life is a vital and urgent problem in the industrial field.

[0003] At present, the fault diagnosis methods of vibrating screens are mainly divided into three aspects: the method based on mechanism model, the method based on expert knowledge and the data-driven method. However, these methods have a common premise assumption in practical application, that is, the training data and the test data must obey the same distribution. However, due to factors such as the diversity of equipment models, the interference of environmental noise and the natural aging of equipment, this assumption is often difficult to achieve, or even almost impossible to achieve. Therefore, it is particularly important to study the cross-domain fault diagnosis method of vibrating screens.

[0004] Cross-domain fault diagnosis methods can be divided into instance-based methods, model-based methods, and feature-based methods. The core idea of ​​instance-based methods is to effectively assign weights to labeled data instances in the source domain, so that the instance distribution in the source domain is closer to the instance distribution in the target domain, thereby building a learning model with high classification accuracy and reliability in the target domain. Model-based methods accelerate the training process of new tasks by reusing the parameters and structure of pre-trained models. This method assumes that there is common knowledge between the source and target tasks at the model level. The core idea of ​​feature-based methods is to align the source and target domain data in the feature space, and by minimizing the difference in the metric feature distribution between the two in the feature space, the feature knowledge learned in the source domain can better adapt to the fault mode of the target domain. Domain adaptation, as a typical representative of this type of method, has been widely studied.

[0005] However, most of the existing cross-domain fault diagnosis methods are performed in deep neural networks. For vibrating screens, the feature extraction process becomes extremely complicated and time-consuming because the data contained in the vibration signals collected are complex and diverse. In addition, the vibrating screen works in a complex and changeable environment, and the differences in the distribution of fault data under different working conditions will also lead to domain offset problems. Therefore, it is urgent to propose a cross-domain fault diagnosis method suitable for vibrating screens. Summary of the Invention

[0006] The present invention proposes a cross - domain fault diagnosis method and system for vibrating screens that enhance domain adaptation, effectively overcomes the challenges of domain shift, significantly improves the performance and generalization ability of the vibrating screen fault diagnosis model, and provides certain technical support for promoting the automation, informatization, and intelligent development of the coal washing and processing process.

[0007] In a first aspect, the present invention discloses a cross - domain fault diagnosis method for vibrating screens that enhances domain adaptation, and the method includes the following steps:

[0008] Step A: Divide the obtained multi - dimensional comprehensive fault data sets of vibrating screens in different domains into a source - domain feature training set, a target - domain feature training set, and a target - domain feature test set;

[0009] Step B: Map the source - domain feature training set and the target - domain feature training set into a random feature space for joint distribution difference processing to perform domain and class alignment of features;

[0010] Step C: Perform domain - feature discrimination processing on the source - domain feature training set to enhance the intra - class compactness and inter - class separability of features;

[0011] Step D: Based on the joint distribution difference processing result and the domain - feature discrimination processing result, construct a cross - domain fault diagnosis model for vibrating screens based on an enhanced domain - adaptive incremental random vector functional link network to predict the fault categories of unknown labels in the target domain.

[0012] Further, in step A, the process of obtaining the multi - dimensional comprehensive fault data sets of vibrating screens in different domains includes the following steps:

[0013] Deeply analyze the operating mechanism of the vibrating screen under harsh working environments and severe loads to identify and predict its common fault types;

[0014] Based on the common fault types of the vibrating screen, install acceleration - temperature composite sensors at different key fault monitoring points of the vibrating screen to collect comprehensive fault data sets of different monitoring points of the vibrating screen in different domains, and generate multi - dimensional comprehensive fault data sets of vibrating screens in different domains.

[0015] Further, the common fault types of the vibrating screen at least include:

[0016] Faults of the screen box body including side plate cracking, lifting beam cracking, driving beam cracking, load - bearing beam cracking, small beam cracking, T - shaped steel cracking, small beam loosening, and T - shaped steel loosening; faults of the vibrator including bearing damage and gear damage; faults of the drive assembly including bearing damage and bearing seat loosening; faults of the screen plate assembly including screen plate loosening, side pressure plate loosening, inlet and outlet blind plate loosening, and rail seat loosening; and spring cracking faults.

[0017] Furthermore, the key fault monitoring points of the vibrating screen include at least the main motor, the transmission shaft, the vibrating screen motor side exciter, the vibrating screen non-motor side exciter, the middle part of the vibrating screen motor side side plate, the middle part of the vibrating screen motor side non-side plate, the vibrating screen non-motor side feed end and discharge end.

[0018] Step A further comprises:

[0019] The multi-domain combined statistical analysis method and principal component analysis method are used to perform multi-dimensional robustness processing in time domain, frequency domain, and time-frequency domain on the multi-dimensional comprehensive fault data sets of vibrating screens in different fields. The generated highly robust feature set is divided into source domain feature training set, target domain feature training set, and target domain feature test set.

[0020] Furthermore, the process of using the multi-domain combined statistical analysis method and principal component analysis method to perform multi-dimensional robustness processing of the multi-dimensional comprehensive fault data set of vibrating screens in different fields in the time domain, frequency domain, and time-frequency domain includes the following steps:

[0021] Step A1: Standardize the fault data using a multi-domain combined statistical analysis method to eliminate the dimensional influence of data in different fields, and extract the statistical characteristics of the vibrating screen fault data using a multi-domain combined method;

[0022] Step A2: Use the principal component analysis method to eliminate redundant features and generate a highly robust feature set, which specifically includes the following sub-steps:

[0023] Step A21: Calculate the covariance matrix of the feature set:

[0024]

[0025] In the formula, represents the original feature set obtained in step A1, b represents the total number of original features, represents the original feature set after standardization, express The transpose of

[0026] Step A22: perform eigendecomposition on the covariance matrix to obtain eigenvalues ​​λ and corresponding eigenvectors ν, arrange the eigenvalues ​​λ in descending order, and selectively retain the first k principal components, where k is determined based on the principle that the cumulative contribution rate is greater than 95%; then project the original eigenmatrix onto the selected principal components to obtain a highly robust feature set;

[0027] Step A3: Divide the highly robust feature set to generate the source domain feature training set Target domain feature training set and the target domain feature test set Among them, S represents the source domain, T represents the target domain, and N S is the number of samples in the source domain feature training set, N t is the number of samples in the target domain feature training set, N tu is the number of samples in the target domain feature test set, N t +N tu =N T and N tu is greater than N t ; is the i-th sample in the source domain feature training set, is the label of the i-th sample in the source domain feature training set; is the i-th sample in the target domain feature training set, is the label of the i-th sample in the target domain feature training set; is the i-th sample in the target domain feature test set; both the source domain and the target domain contain C fault categories, and each fault sample contains m features.

[0028] Step B further includes:

[0029] Step B1: Map the source domain feature training set and the target domain feature training set into a random feature space to generate source domain random feature data and target domain random feature data;

[0030] Step B2: Use the source domain random feature data and the target domain random feature data to calculate the marginal distribution difference and the conditional distribution difference between the source domain and the target domain based on the maximum mean and covariance difference to generate a joint distribution difference.

[0031] Step C further includes:

[0032] Perform domain feature discrimination processing on the source domain random feature data generated in Step B, optimize the feature space by introducing a domain discrimination metric, and enhance the intra-class compactness and inter-class separability of the features in the random feature space.

[0033] Furthermore, in Step D, based on the enhanced domain adaptive incremental random vector functional link network, a vibration sieve fault diagnosis model is constructed to predict the fault categories with unknown labels in the target domain; the source domain feature training set and the target domain feature training set are used to train the vibration sieve fault diagnosis model to obtain the output weights of the vibration sieve fault diagnosis model, and fault diagnosis is performed on the target domain feature test set; specifically, it includes the following sub-steps:

[0034] Step D1: In the initialization stage of constructing the vibration sieve fault diagnosis model, set the upper limit of the maximum number of hidden layer nodes to L max , the current number of hidden layer nodes in the network is L, the initial error value of the model is set to e 0 , the tolerance error is ε, and a set of hyperparameters is defined and the random parameters are randomly assigned within the interval [-δ, δ];

[0035] Step D2: Calculate the maximum mean difference matrix M 0 , the maximum covariance difference matrix Q 0 , the within-class discriminant metric matrix L of the source domain tra-S and the between-class discriminant metric matrix L ter-S ;

[0036] Step D3: Randomly generate L hidden layer nodes in sequence within the range of the randomly assigned parameters. For each generated node, calculate the corresponding output H of the hidden layer nodes in the source domain S and the output H of the hidden layer nodes in the target domain t , and calculate H, H = [H S ; H t ;

[0037] Step D4: Calculate the class means in the output H of the hidden layer in the source domain to obtain the class centroid matrix of the source domain in the feature space S

[0038] Step D5: Calculate the joint distribution difference matrix M;

[0039] Step D6: Solve the output weights of the model in a globally optimal manner:

[0040]

[0041] In the formula, I represents the identity matrix, represents the transpose of H S , Y S represents the true output of the source domain feature training set, Y t represents the true output of the target domain feature training set;

[0042] Step D7: Calculate the trade-off factor in the joint distribution difference:

[0043]

[0044] In the formula, d mar is the marginal distribution difference, d con is the conditional distribution difference;

[0045] Step D8: Calculate the residual e L = H t β * - Y t , when the residual satisfies the tolerance error ε or the number of hidden layer nodes reaches L max ​When the model establishment is completed, go to step D9; otherwise, return to step D3 and add hidden layer nodes to the network, and continuously loop through steps D3 to D6 until the residual meets the tolerance error ε or the number of hidden layer nodes reaches L max and stop when it reaches;

[0046] Step D9: Use the vibrating screen fault diagnosis model to diagnose the samples of the target domain feature test set, and calculate the diagnostic accuracy rate, precision rate, and F1 score.

[0047] In a second aspect, the present invention discloses a vibrating screen cross-domain fault diagnosis system for enhancing domain adaptation, and the system includes:

[0048] A dataset construction module, configured to obtain multi-dimensional comprehensive fault datasets of vibrating screens in different domains, and divide them into a source domain feature training set, a target domain feature training set, and a target domain feature test set;

[0049] A joint distribution difference processing module, configured to map the source domain feature training set and the target domain feature training set into a random feature space for joint distribution difference processing to perform domain and class alignment of features;

[0050] A domain feature discrimination processing module, configured to perform domain feature discrimination processing on the source domain feature training set to enhance the intra-class compactness and inter-class separability of features;

[0051] A model construction and training module, configured to construct a vibrating screen cross-domain fault diagnosis model based on the enhanced domain adaptation incremental random vector functional link network according to the joint distribution difference processing result and the domain feature discrimination processing result, and predict the fault categories with unknown labels in the target domain.

[0052] The beneficial effects of the present invention are as follows:

[0053] First, the enhanced domain adaptation vibrating screen cross-domain fault diagnosis method and system of the present invention, by deeply analyzing the operation mechanism of the vibrating screen, analyzing the vibration signals collected according to the characteristics of the vibrating screen, proposes a robust feature extraction process, and at the same time realizes cross-domain fault diagnosis by applying the enhanced domain adaptation method to the incremental stochastic neural network, effectively solving the domain shift problem caused by the distribution difference of fault data under different working conditions (such as factors such as equipment model changes, environmental noise interference, and natural aging of equipment in actual industry).

[0054] Second, the enhanced domain adaptation vibrating screen cross-domain fault diagnosis method and system of the present invention realize domain and class alignment by mapping the source domain and target domain features into a random feature space for joint distribution difference processing, and perform domain feature discrimination processing on the source domain to enhance intra-class compactness and inter-class separability, thereby realizing the domain adaptation ability and significantly improving the performance and generalization ability of the fault diagnosis model. Brief Description of the Drawings

[0055] Figure 1 It is a flowchart of the cross - domain fault diagnosis method for vibrating screens that enhances domain adaptability according to the present invention;

[0056] Figure 2 It is a schematic diagram of the data distribution difference of the present invention, where (a) is the group marginal distribution diagram of the source - domain vibrating screen, and (b) is the group marginal distribution diagram of the target - domain vibrating screen;

[0057] Figure 3 It is a schematic diagram of the structure of the enhanced domain - adaptive incremental random vector functional link network in the present invention. Detailed Embodiment

[0058] The following embodiments can enable those skilled in the art to understand the present invention more comprehensively, but do not limit the present invention in any way.

[0059] As Figure 1 shown, the present invention provides a cross - domain fault diagnosis method for vibrating screens that enhances domain adaptability, including the following steps:

[0060] S1: Obtain the vibration acceleration and temperature data of vibrating screens in different domains, construct a multi - dimensional fault data set, and generate a comprehensive fault data set by fusing the fault information of different monitoring points. The specific steps of S1 are as follows:

[0061] S11: The operating mechanism of the vibrating screen mainly depends on the periodic excitation force generated by the vibration motor. This force causes the screen body to generate reciprocating or rotational vibrations, resulting in the stratification and jumping of materials on the screen surface. During the vibration process, smaller particles pass through the sieve holes and fall to the lower layer, while larger particles remain on the screen surface and are finally discharged, achieving the classification screening of materials according to particle size. As shown in Table 1, in the actual use process, due to the harsh working environment of the vibrating screen, its main components are extremely prone to failure;

[0062] S12: According to common faults, acceleration - temperature composite sensors are respectively installed at 8 points, namely the main motor, the transmission shaft, the exciter on the motor side of the vibrating screen, the exciter on the non - motor side of the vibrating screen, the middle part of the side plate on the motor side of the vibrating screen, the middle part of the non - side plate on the motor side of the vibrating screen, the feeding end and the discharging end on the non - motor side of the vibrating screen. The acquisition frequency of the vibration signal is 51200Hz, and the acquisition time of each vibration signal is 320ms. The fault data of different measurement points of different vibrating screens are transmitted to the database through a wireless gateway. Use to represent the original fault data of the source domain collected, to represent the original fault data of the target domain collected.

[0063] S2: Perform multi-dimensional robustness processing on the comprehensive fault data set in the time domain, frequency domain, and time-frequency domain to generate a highly robust feature set, and divide it into a source domain feature training set, a target domain feature training set, and a target domain feature test set. The specific steps of step S2 are as follows:

[0064] S21: Standardize the fault data to eliminate the influence of the dimension of data in different domains; extract the feature sets of the original fault data in the source domain and target domain in the time domain, frequency domain, and time-frequency domain The total number of features in both the source domain and the target domain is m. As shown in Table 2, this feature set consists of the mean value, root mean square value, frequency centroid, mean square frequency, wavelet transform coefficients, etc.;

[0065] S22: Use the principal component analysis method to eliminate redundant features and generate a highly robust feature set. The specific steps are as follows:

[0066] Calculate the covariance matrix of the feature set:

[0067]

[0068] In the formula, represents the source domain original feature set obtained in step S21, b represents the total number of source domain original features, represents the standardized source domain original feature set, represents transpose of;

[0069] Calculate the eigenvalues λ S and eigenvectors ν S , arrange the eigenvalues λ S in descending order, and selectively retain the first k S principal components, where k S is determined according to the principle that the cumulative contribution rate is greater than 95%; then project the source domain original feature set onto the selected principal components to obtain a highly robust feature set Obtain the target domain feature set using the same method

[0070] S23: Divide the highly robust feature set to generate a source domain feature training set containing N S feature samples N t a target domain feature training set with a small number of N feature samples tu and a target domain feature test set with N feature samples t where N tu +N T =N

[0071] Table 1 Vibration Screen Fault Type Table

[0072]

[0073] Table 2 Vibration Screen Fault Data Feature Table

[0074]

[0075]

[0076] S3: Map the source domain and target domain feature training sets to a random feature space for joint distribution difference processing to achieve domain and class alignment of features; the specific steps of step S3 are as follows:

[0077] As Figure 2 shown, due to the differences between cross-domain data, the data of different vibration screens are distributed differently on different coordinate axes; for this reason, align the features of the source domain and the target domain in the random feature space; the specific steps are as follows:

[0078] S31: Map the source domain and target domain feature training sets to a random feature space to generate multiple groups of random feature data;

[0079] S32: Use the source domain and target domain random feature data to calculate the marginal distribution difference and conditional distribution difference between the source domain and the target domain based on the maximum mean and covariance differences to generate a joint distribution difference; the joint distribution difference with a classifier is as follows:

[0080] L JMC = tr(β T HMH T β)

[0081] Let the joint distribution difference matrix where is the balance factor between the two differences; H = [H S , H T , H S represents the output matrix of the source domain hidden layer, and H T represents the output matrix of the target domain hidden layer;

[0082] S4: Perform domain feature discrimination processing on the source domain feature training set to enhance the within-class compactness and between-class separability of the features; the specific steps of step S4 are as follows:

[0083] To further effectively solve the domain shift problem and learn more domain discriminative features, enhance the within-class compactness and between-class separability by minimizing the within-class loss and maximizing the between-class loss, so as to improve the accuracy of the diagnosis results; the calculation of the within-class discrimination metric for the source domain is as follows:

[0084]

[0085] In the formula, L tra-S represents the within-class discriminant metric matrix L of the source domain tra-S ; represents H S transpose;

[0086] The between-class discriminant metric of the source domain is calculated as follows:

[0087]

[0088] In the formula, represents the class centroid matrix of the source domain in the feature space, represents transpose, L ter-S represents the between-class discriminant metric matrix;

[0089] S5: Generate an enhanced domain adaptation method according to the joint distribution difference and domain feature discrimination processing, and construct a cross-domain fault diagnosis model for vibrating screens; Step S5 is specifically as follows:

[0090] As Figure 3 shown, the present invention constructs a fault diagnosis model for vibrating screens based on enhanced domain adaptation, that is, an enhanced domain adaptation incremental random vector functional link network, which is used to predict the fault categories of unknown labels in the target domain; during the model training process, the source domain feature set and a small amount of target domain feature sets are used to construct a training set, and the remaining large amount of target domain feature sets are used to construct a test set; finally, the output weights of the model are obtained for fault diagnosis of the target domain test set, and the specific steps are as follows:

[0091] S51: Obtain a feature training set and a feature test set through S2, including a source domain feature training set a target domain feature training set a target domain feature test set Perform one-hot encoding on the training sample labels; Initialize various parameters for constructing the model: the upper limit of the maximum number of hidden layer nodes is L max , the current number of hidden layer nodes in the network is L, the initial error value of the model is set to e 0 , the tolerance error is ε, define a set of hyperparameters and the random parameters are randomly assigned within the interval [-v, δ];

[0092] S52: Calculate the maximum mean discrepancy matrix M 0 , the maximum covariance discrepancy matrix Q 0 , the within-class discriminant metric matrix L of the source domain tra-S and the between-class discriminant metric matrix L ter-S ;

[0093] S53: Randomly generate L hidden layer nodes in sequence within the allocation range of random parameters. For each generated node, calculate the output H of the corresponding source domain hidden layer node S and the output H of the target domain hidden layer node t , and calculate H;

[0094] S54: Calculate the class mean in the source domain hidden layer output H S to obtain the class centroid matrix of the source domain in the feature space

[0095] S55: Calculate the joint distribution difference matrix M;

[0096] S56: Solve the output weights of the model in a globally optimal manner:

[0097]

[0098] In the formula, I represents the identity matrix, represents the transpose of H S , Y S represents the true output of the source domain training set, Y t represents the true output of the target domain training set;

[0099] S57: Calculate the trade-off factor in the joint distribution difference:

[0100]

[0101] In the formula, d mar is the marginal distribution difference, d con is the conditional distribution difference;

[0102] S58: Calculate the residual e L = H t β * - Y t , when the residual satisfies the tolerance error ε or the number of hidden layer nodes reaches L max , the model establishment ends. Otherwise, return to step S53 and add hidden layer nodes to the network, continuously loop through steps S53 - S56 until the residual satisfies the tolerance error ε or the number of hidden layer nodes reaches L max and stop;

[0103] S59: Use the network model to diagnose the samples of the target domain test set Calculate the hidden layer output weight H tu through the random parameters generated in step S53 to obtain the output matrix T of the model tu = H tu β * ; Convert the output matrix and compare it with the true label Y of the test sample tu to calculate the accuracy, precision, and F1 score.

[0104] To illustrate the performance of the present invention in cross-domain fault diagnosis, a practical vibrating screen example is used for illustration. The same faults of three vibrating screens are selected for experiments, and the specific fault descriptions are shown in Table 3 below.

[0105] Four methods, namely RVFL, IRVFL, UD-RVFL, and DAMR-RWNN, are selected as the comparison methods for fault diagnosis in this paper. RVFL and IRVFL are basic models without domain adaptation, which are trained only using the labeled source domain feature set and then immediately evaluated on the target domain test set; UD-RVFL and DAMR-RWNN are cross-domain models considering domain shift. The method of the present invention is compared with the above four methods in terms of accuracy, precision, and F1-score. The diagnostic accuracy results of different methods are shown in Table 4 below. Tables 5 and 6 show the diagnostic precision and F1-score of the five methods; it can be seen from the tables that in the experiments of six cross-domain tasks, the method of the present invention can achieve the optimal values in terms of accuracy, precision, and F1-score. Moreover, compared with RVFL and IRVFL, the performance indicators of the present invention have a relatively high improvement, indicating that the cross-domain model can better solve the problem of domain shift, thereby improving the accuracy of fault diagnosis. By comparing with other methods, the method of the present invention performs best in the three evaluation indicators, which proves the effectiveness of this method.

[0106] Table 3 Faults selected for experiments

[0107]

[0108] Table 4 Diagnostic accuracy of different methods

[0109]

[0110] Table 5 Diagnostic precision of different methods

[0111]

[0112] Table 6 F1-score of different methods

[0113]

[0114] The above is only the preferred implementation manner of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. A method for cross-domain fault diagnosis of vibrating screens with enhanced domain adaptation, characterized in that: The method comprises the following steps: Step A, dividing the acquired multi-dimensional comprehensive fault data sets of vibrating screens in different fields into a source domain feature training set, a target domain feature training set and a target domain feature test set; Step B, mapping the source domain feature training set and the target domain feature training set into the random feature space for joint distribution difference processing to perform domain and class alignment of the features; Step C: Perform domain feature discrimination processing on the source domain feature training set to enhance the intra-class compactness and inter-class separation of features; Step D, according to the joint distribution difference processing results and the domain feature discrimination processing results, a vibrating screen cross-domain fault diagnosis model is constructed based on the enhanced domain adaptive incremental random vector function chain network to predict the fault category of the unknown label in the target domain.

2. The vibrating screen cross-domain fault diagnosis method with enhanced domain adaptation according to claim 1 is characterized in that: In step A, the process of acquiring the multi-dimensional comprehensive fault data set of vibrating screens in different fields includes the following steps: Conduct in-depth analysis of the operating mechanism of the vibrating screen under harsh working environment and severe loads to identify and predict its common fault types; Based on the common fault types of vibrating screens, acceleration-temperature composite sensors are installed at different key fault monitoring points of vibrating screens to collect comprehensive fault data sets of different monitoring points of vibrating screens in different fields and generate multi-dimensional comprehensive fault data sets of vibrating screens in different fields.

3. The vibrating screen cross-domain fault diagnosis method with enhanced domain adaptation according to claim 2 is characterized in that: The common fault types of the vibrating screen include at least: Screen box body failures including side panel cracking, lifting beam cracking, drive beam cracking, load-bearing beam cracking, small beam cracking, T-steel cracking, small beam loosening, T-steel loosening, vibrator failures including bearing damage and gear damage, drive component failures including bearing damage and bearing seat loosening, screen plate component failures including screen plate loosening, side pressure plate loosening, feed and discharge blind plate loosening, rail seat loosening, and spring cracking failures.

4. The vibrating screen cross-domain fault diagnosis method with enhanced domain adaptation according to claim 2 is characterized in that: The key fault monitoring points of the vibrating screen include at least the main motor, the transmission shaft, the vibrating screen motor side exciter, the vibrating screen non-motor side exciter, the middle of the vibrating screen motor side side plate, the middle of the vibrating screen motor side non-side plate, the vibrating screen non-motor side feed end and the vibrating screen non-motor side discharge end.

5. The vibrating screen cross-domain fault diagnosis method with enhanced domain adaptation according to claim 1 is characterized in that: Step A further comprises: The multi-domain combined statistical analysis method and principal component analysis method are used to perform multi-dimensional robustness processing in time domain, frequency domain, and time-frequency domain on the multi-dimensional comprehensive fault data sets of vibrating screens in different fields. The generated highly robust feature set is divided into source domain feature training set, target domain feature training set, and target domain feature test set.

6. The vibrating screen cross-domain fault diagnosis method with enhanced domain adaptation according to claim 5 is characterized in that: The process of using multi-domain combined statistical analysis method and principal component analysis method to perform multi-dimensional robustness processing of multi-dimensional comprehensive fault data sets of vibrating screens in different fields in time domain, frequency domain, and time-frequency domain includes the following steps: Step A1: Standardize the fault data using a multi-domain combined statistical analysis method to eliminate the dimensional influence of data in different fields, and extract the statistical characteristics of the vibrating screen fault data using a multi-domain combined method; Step A2: Use the principal component analysis method to eliminate redundant features and generate a highly robust feature set, which specifically includes the following sub-steps: Step A21: Calculate the covariance matrix of the feature set: In the formula, represents the original feature set obtained in step A1, b represents the total number of original features, represents the original feature set after standardization, express The transpose of Step A22: perform eigendecomposition on the covariance matrix to obtain eigenvalues ​​λ and corresponding eigenvectors v, arrange the eigenvalues ​​λ in descending order, and selectively retain the first k principal components, where k is determined based on the principle that the cumulative contribution rate is greater than 95%; then project the original eigenmatrix onto the selected principal components to obtain a highly robust feature set; Step A3: Divide the highly robust feature set to generate the source domain feature training set Target domain feature training set and the target domain feature test set Where S represents the source domain, T represents the target domain, and N S is the number of samples in the source domain feature training set, N t is the number of samples in the target domain feature training set, N tu is the number of samples in the target domain feature test set, N t +N tu =N T , and N tu Greater than N t ; is the i-th sample of the source domain feature training set, is the label of the i-th sample in the source domain feature training set; is the i-th sample of the target domain feature training set, is the label of the i-th sample in the target domain feature training set; is the i-th sample of the target domain feature test set; both the source domain and the target domain contain C fault categories, and each fault sample contains m features.

7. The vibrating screen cross-domain fault diagnosis method with enhanced domain adaptation according to claim 1 is characterized in that: Step B further comprises: Step B1: Map the source domain feature training set and the target domain feature training set into the random feature space to generate source domain random feature data and target domain random feature data; Step B2: Using the source domain random feature data and the target domain random feature data, the marginal distribution difference and conditional distribution difference between the source domain and the target domain based on the maximum mean and covariance differences are calculated to generate the joint distribution difference.

8. The vibrating screen cross-domain fault diagnosis method with enhanced domain adaptation according to claim 1 is characterized in that: Step C further comprises: The source domain random feature data generated in step B is processed by domain feature discrimination, and the feature space is optimized by introducing domain discrimination metrics to enhance the intra-class compactness and inter-class separability of features in the random feature space.

9. The vibrating screen cross-domain fault diagnosis method with enhanced domain adaptation according to claim 1 is characterized in that: In step D, a vibration screen fault diagnosis model is constructed based on an enhanced domain adaptive incremental random vector function chain network to predict the fault category of an unknown label in the target domain; the vibration screen fault diagnosis model is trained using a source domain feature training set and a target domain feature training set to obtain the output weights of the vibration screen fault diagnosis model, and fault diagnosis is performed on the target domain feature test set; specifically, the following sub-steps are included: Step D1: In the initialization stage of building the vibration screen fault diagnosis model, the upper limit of the maximum number of hidden layer nodes is set to L max , the current number of hidden layer nodes in the network is L, the initial error value of the model is set to e0, the tolerance error is ε, and a set of hyperparameters are defined And the random parameters are randomly assigned in the interval [-δ,δ]; Step D2: Calculate the maximum mean difference matrix M0, the maximum covariance difference matrix Q0, and the source domain intra-class discriminant metric matrix L tra-S And the inter-class discriminant metric matrix L ter-S ; Step D3: Randomly generate L hidden layer nodes in sequence within the distribution range of random parameters. For each generated node, calculate the corresponding source domain hidden layer node output H S and the target domain hidden layer node output H t , and calculate H, H = [H S ;H t ]; Step D4: Calculate the source domain hidden layer output H S The class mean in the feature space is obtained by taking the class centroid matrix of the source domain Step D5: Calculate the joint distribution difference matrix M; Step D6: Solve the output weights of the model through global optimization: In the formula, I represents the unit matrix, Indicates H S The transpose of S represents the true output of the source domain feature training set, Y t Represents the true output of the target domain feature training set; Step D7: Calculate the trade-off factor in the difference of joint distributions: Where, d mar is the marginal distribution difference, d con is the conditional distribution difference; Step D8: Calculate the residual e L =H t β * -Y t , when the residual meets the tolerance error ε or the number of hidden layer nodes reaches L max When , the model building is completed and the process goes to step D9; Otherwise, return to step D3 and add hidden nodes to the network, and continue to cycle through steps D3 to D6 until the residual meets the tolerance error ε or the number of hidden nodes reaches L max Stop when Step D9: Use the vibration screen fault diagnosis model to diagnose the target domain feature test set samples and calculate the diagnosis accuracy, precision, and F1 score.

10. A vibrating screen cross-domain fault diagnosis system with enhanced domain adaptation based on the method described in any one of claims 1 to 9, characterized in that: The system comprises: The dataset construction module is used to obtain multi-dimensional comprehensive fault datasets of vibrating screens in different fields and divide them into source domain feature training set, target domain feature training set and target domain feature test set; A joint distribution difference processing module is used to map the source domain feature training set and the target domain feature training set into a random feature space for joint distribution difference processing to perform domain and class alignment of features; The domain feature discrimination processing module is used to perform domain feature discrimination processing on the source domain feature training set to enhance the intra-class compactness and inter-class separation of features; The model building and training module is used to build a vibrating screen cross-domain fault diagnosis model based on the enhanced domain adaptive incremental random vector function chain network according to the joint distribution difference processing results and domain feature discrimination processing results, and predict the fault category of the unknown label in the target domain.

Images