A Fault Diagnosis Operating Condition Domain Generalization Method and System Based on Operating Condition Independent Feature Representation
By embedding operating condition information into the model using a condition-independent feature representation method, the problem of cross-condition generalization in rotating machinery fault diagnosis is solved, achieving efficient diagnosis under unknown operating conditions and improving the applicability and robustness of the model.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG UNIV
- Filing Date
- 2026-03-16
- Publication Date
- 2026-06-30
AI Technical Summary
Existing deep learning methods suffer from a sharp decline in diagnostic performance under unknown operating conditions due to the domain differences between training and test data in rotating machinery fault diagnosis. Existing methods typically reduce model stability and computational cost, and have poor adaptability.
By using a condition-independent feature representation method, condition information is embedded into the model, and modulation parameters are generated using a learnable multilayer perceptron. This maps the vibration signal to a condition-independent feature representation space, achieving cross-condition generalization.
This improved the generalization ability and practical value of the rotating machinery fault diagnosis model under varying operating conditions, reduced data acquisition costs, and enhanced the robustness and applicability of the model.
Smart Images
Figure CN121834479B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of rotating machinery fault diagnosis technology, and in particular to a fault diagnosis operating condition domain generalization method and system based on operating condition-independent feature representation. Background Technology
[0002] Rotating machinery is a core component of critical equipment such as wind turbines, aero engines, and high-speed trains, and its operating status directly affects the safety and reliability of the equipment. Traditional fault diagnosis of rotating machinery mainly relies on human experience, judging by listening and touching, which is inefficient and highly subjective.
[0003] In recent years, deep learning-based fault diagnosis methods have gradually become a research hotspot, achieving fault identification by automatically extracting features from vibration signals. However, existing deep learning methods typically assume that training and test data follow the same distribution. In actual industrial scenarios, due to differences in equipment models, operating conditions, environmental conditions, and other factors, there are often significant domain differences between the training and test sets, leading to a sharp decline in the model's diagnostic performance under unknown operating conditions, thus limiting its engineering applications.
[0004] To address the problem of generalization across operating conditions, existing research often employs methods such as domain-invariant feature learning or data augmentation. However, these methods typically reduce model stability and interpretability while significantly increasing computational costs, resulting in high costs and poor adaptability in practical applications. Therefore, there is an urgent need for a lightweight method that can learn operating condition-independent feature representations using operating condition information to improve the generalization ability and practical value of rotating machinery fault diagnosis models under varying operating conditions. Summary of the Invention
[0005] To address the aforementioned issues, this invention proposes a fault diagnosis operating condition domain generalization method and system based on operating condition-independent feature representation. By embedding operating condition information into the model in a learnable manner, the vibration signal and intermediate features are guided to be mapped to a normalized feature space that is independent of operating condition changes, thereby achieving adaptation to unseen vibration signals from new operating conditions and solving the cross-operating condition generalization problem.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] In a first aspect, the present invention provides a fault diagnosis operating condition domain generalization method based on operating condition-independent feature representation, comprising the following steps:
[0008] Collect vibration signals and corresponding operating condition information of the target equipment under normal and fault conditions under multiple operating conditions;
[0009] The vibration signal is cut into segments of a preset length, and the working condition information is mapped into a working condition vector with the same length as the vibration signal segment.
[0010] Modulation parameters are generated by a learnable multilayer perceptron using the operating condition vector. These modulation parameters are used to adjust the vibration signal and its intermediate features step by step over time, mapping the vibration signal to an operating condition-independent feature representation space.
[0011] Fault classification is performed based on the mapped feature representation, and fault diagnosis results are output.
[0012] As an alternative implementation, the modulation parameters include two sets, wherein the first set of modulation parameters includes shift, scale, and gate, which are used to modulate the input vibration signal.
[0013] The second set of modulation parameters includes shift_mlp, scale_mlp, and gate_mlp, which are used to modulate the intermediate features of the vibration signal.
[0014] As an alternative implementation, the method maps vibration signals to a condition-independent feature representation space through a covariate knowledge embedding module, which performs the following operations:
[0015] The input vibration signal is subjected to layer normalization processing;
[0016] The normalized vibration signal is modulated using shift and scale from the first set of modulation parameters;
[0017] The modulated vibration signal is input into the feature extraction layer to extract features;
[0018] The extracted features are modulated using the gate from the first set of modulation parameters and then residually connected to the original input vibration signal.
[0019] As an alternative implementation, the feature extraction layer can be any one of a multilayer perceptron, a convolutional neural network, or a temporal Transformer network, and the covariate knowledge embedding module can be stacked and applied to multiple feature extraction layers.
[0020] As an alternative implementation, the operating condition information includes at least one of speed, load, and torque, and multiple operating condition variables are mapped to vectors of the same length as the vibration signal segment through independent fully connected layers and then summed and fused.
[0021] The parameters of the fully connected layer are updated through backpropagation.
[0022] As an alternative implementation, the method uses only data from a single or a few operating conditions during the training phase, and the method still has the ability to generalize fault diagnosis even when no operating conditions are encountered.
[0023] Secondly, the present invention provides a fault diagnosis operating condition domain generalization system based on operating condition-independent feature representation, comprising:
[0024] The data acquisition module is configured to: acquire vibration signals and corresponding operating condition information of the target equipment under normal and fault conditions under multiple operating conditions;
[0025] The data processing module is configured to: cut the vibration signal into segments of a preset length and map the working condition information into a working condition vector with the same length as the vibration signal segment;
[0026] The normalization module is configured to generate modulation parameters using a learnable multilayer perceptron via a working condition vector. The modulation parameters are used to adjust the vibration signal and its intermediate features step by step over time, mapping the vibration signal to a working condition-independent feature representation space.
[0027] The fault diagnosis module is configured to classify faults based on the mapped feature representation and output fault diagnosis results.
[0028] Thirdly, the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.
[0029] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in the first aspect.
[0030] Fifthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the method described in the first aspect.
[0031] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0032] This invention proposes a fault diagnosis operating condition domain generalization method based on operating condition-independent feature representation. By constructing an operating condition-independent feature representation space, vibration signals under different operating conditions are actively mapped to the same normalized domain, significantly improving the model's diagnostic performance under unknown operating conditions. Moreover, it only requires collecting vibration signals and operating condition information of the target equipment under a few typical operating conditions (or even a single operating condition), without needing to cover all possible operating conditions, greatly reducing data acquisition costs and annotation burden.
[0033] The CKE module proposed in this invention features plug-and-play functionality, allowing for flexible embedding into various feature extraction networks (such as MLP, CNN, and TST) to enhance the domain generalization ability of existing diagnostic models. The CKE module modulates both the input signal and intermediate features, enhancing the model's ability to model changes in operating conditions and improving the robustness of feature representation. By generating time-step modulation parameters, it achieves fine-grained adjustment of the vibration signal in the time dimension, better adapting to the local effects of operating condition fluctuations.
[0034] In summary, the method of the present invention significantly improves the applicability and engineering deployability of the rotating machinery fault diagnosis system under complex working conditions while ensuring diagnostic accuracy.
[0035] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0036] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0037] Figure 1 This is a flowchart of the fault diagnosis operating condition domain generalization method based on operating condition-independent feature representation of the present invention;
[0038] Figure 2 This is an architecture diagram of the fault diagnosis operating condition domain generalization model based on operating condition-independent feature representation of the present invention. Detailed Implementation
[0039] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0040] It should be noted that the following detailed description is exemplary and intended to provide further illustration of the invention. Unless otherwise specified, 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 pertains.
[0041] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form as well. Furthermore, it should be understood that the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but includes other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0042] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.
[0043] Example 1
[0044] like Figure 1 As shown, this embodiment provides a fault diagnosis operating condition domain generalization method based on operating condition-independent feature representation, including the following steps:
[0045] S1. Collect vibration signals and corresponding operating condition information of the target equipment under normal and fault conditions under multiple operating conditions.
[0046] S2. Cut the vibration signal into segments of a preset length and map the working condition information into a working condition vector with the same length as the vibration signal segment;
[0047] S3. Using the working condition vector, a learnable multilayer perceptron is used to generate modulation parameters. The modulation parameters are used to adjust the vibration signal and its intermediate features step by step over time, and to map the vibration signal to a working condition-independent feature representation space.
[0048] S4. Based on the mapped feature representation, classify the faults and output the fault diagnosis results.
[0049] The specific solution of the present invention is as follows:
[0050] The method of this invention does not require modeling all possible operating conditions. It learns a condition normalization domain by collecting data from only a small number of operating conditions (or even just a single condition in extreme cases). This maps vibration signals to a unified representation space that is insensitive to operating conditions, thereby significantly improving cross-condition generalization capability. First, vibration signals and operating condition information that may affect signal changes are collected under several operating conditions of the target equipment, including normal and various fault states. Second, a vibration signal training model (rotating machinery fault diagnosis operating condition generalization model) is used, embedding the operating condition information into the model in a learnable manner. Finally, vibration signals and operating condition information are collected on the target equipment operating under any actual condition using the same method, input into the model, and the model's output is used to determine whether the rotating machinery has a fault and its possible fault states.
[0051] The specific steps for S1 are as follows:
[0052] Vibration signals x of the target equipment under normal and fault conditions under several typical operating conditions, as well as operating condition variables c that can significantly affect the statistical characteristics of the vibration signals, such as speed, load, and torque, are collected to construct a training set. This disclosure specifically emphasizes that it is not necessary to cover all operating conditions; training can be completed with only a small amount of data or data from a single operating condition.
[0053] The specific steps for S2 are as follows:
[0054] The vibration signal is cut into segments of length T.
[0055] ;
[0056] The operating condition covariates are extended into an operating condition sequence with the same length as the vibration signal segment through a learnable mapping function.
[0057] Suppose there are m covariates related to operating conditions:
[0058] ;
[0059] Each condition is mapped to a working condition vector of length T through a learnable fully connected layer:
[0060] ;
[0061] in It is a fully connected layer with independent covariates for each operating condition, and its parameters are updated by backpropagation.
[0062] The multiple load case vectors are then fused using addition. Since the load case vectors are generated through independent fully connected layers, this essentially achieves weighted fusion.
[0063] ;
[0064] Modulation parameters are generated using a learnable multilayer perceptron through the weighted fused operating condition vector.
[0065] The specific steps for S3 are as follows:
[0066] The model is trained using vibration signals and operating condition information (i.e., weighted and fused operating condition vectors) from the training set. The model learns the operating condition corresponding to the current vibration signal through embedding and guides the vibration signal to be normalized to a normalized domain independent of operating condition changes, thus achieving operating condition domain generalization. Specifically:
[0067] Reference Figure 2 The weighted and fused operating condition vectors are all used as inputs to the CKE module of the fault diagnosis operating condition domain generalization model based on operating condition-independent feature representation. Two sets of learnable modulation parameter sets are generated by modulating the MLP layer:
[0068] ;
[0069] Each parameter is a vector of length T, which provides the ability to modulate the operating conditions point by point in the time dimension.
[0070] ( ) is the first set of modulation parameters, used to adjust the input signal (the sum of the vibration signal and the position vector);
[0071] ( ), which is the second set of modulation parameters used to adjust intermediate features.
[0072] Global normalization is applied to the dataset before use. The vibration signal and position vector are summed and then input into the CKE module. The input signal is modulated using the first set of modulation parameters: after passing through Layer Norm in the CKE module, it is adjusted using shift and scale. Features are then extracted through the Model layer. The features are adjusted through the gate and then residually connected to the input.
[0073] ;
[0074] The model layer here can be a common feature extractor (such as MLP, CNN, TST), which can easily replace the corresponding structure in a general fault diagnosis model to enhance its domain generalization ability. Intermediate features are obtained through the above steps.
[0075] Then, the intermediate features are adjusted using the same structure with the MLP regulator at its core. The intermediate features are modulated using a second set of parameters, and the modulation method is the same as that of the first set of parameters:
[0076] ;
[0077] The model learns simultaneously through backpropagation:
[0078] The model obtains a feature representation that is invariant across operating conditions by using the operating condition vector mapping function, modulation parameters, modulation strategy, and the high-dimensional features of the operating condition normalization domain itself.
[0079] The CKE module disclosed herein has the following technical features:
[0080] (1) Normalized operating condition domain:
[0081] The CKE module does not simply adjust features based on operating conditions. Instead, it adjusts the input signal and intermediate features by generating a parameter set based on the operating condition vector, learns an "operating condition-independent representation space", namely the operating condition normalization domain, and actively maps signals under different operating conditions to this domain, thereby significantly reducing statistical differences between operating conditions.
[0082] (2) Time step domain alignment:
[0083] The modulation parameter set is generated based on the operating condition sequence and is a vector of length T, enabling the model to normalize according to the local operating condition effects at each time step. Existing methods typically only use global operating condition information to modulate the signal and only support a few operating condition variables (such as rotational speed), and cannot normalize for local fluctuations in the time series.
[0084] (3) Dual-channel modulation:
[0085] Single operating condition modulation is equivalent to:
[0086] ;
[0087] It can only adjust the amplitude and phase effects caused by changes in operating conditions, but cannot completely eliminate the influence of operating conditions on the signal. It is proposed to use one set of modulation parameters for the original input signal and another set of modulation parameters (MLP regulator) for the intermediate characteristics.
[0088] (4) Plug-and-play operating condition normalization module:
[0089] The input and output dimensions of the CKE module are the same. Each CKE module can embed feature extraction networks of different layers (such as MLP, CNN, TST) to achieve "plug and play" generalization capability.
[0090] Vibration signals x and corresponding operating condition variables c are collected from actual operating equipment. A normalized domain mapping based on the operating condition is then performed using a trained CKE module to obtain an operating condition-independent feature representation. This representation is then input into a classifier to determine whether the equipment is abnormal and its fault type.
[0091] This method eliminates the need to collect training data for each operating condition, making it highly efficient to deploy and valuable in industrial settings.
[0092] This invention only requires collecting vibration signals and operating information under several operating conditions of the target equipment. In extreme cases, it only requires collecting vibration signals and operating information under one operating condition, rather than collecting vibration signals and operating information under every single operating condition, thus greatly saving manpower and resources. The collected vibration signals can effectively reflect the operating status of the target equipment, and the collected operating information can effectively guide the model operation, which is beneficial for accurately obtaining the operating status of the target equipment.
[0093] The model, by embedding operating condition signals to bridge the differences in operating condition domains, exhibits stronger embeddability compared to other methods. It is compatible with general fault diagnosis methods, enhancing its cross-operating condition generalization ability. This method effectively helps vibration signal-based fault diagnosis models achieve domain generalization, expanding their applicability.
[0094] Example 2
[0095] This embodiment provides a fault diagnosis operating condition domain generalization system based on operating condition-independent feature representation, including:
[0096] The data acquisition module is configured to: acquire vibration signals and corresponding operating condition information of the target equipment under normal and fault conditions under multiple operating conditions;
[0097] The data processing module is configured to: cut the vibration signal into segments of a preset length and map the working condition information into a working condition vector with the same length as the vibration signal segment;
[0098] The normalization module is configured to generate modulation parameters using a learnable multilayer perceptron via a working condition vector. The modulation parameters are used to adjust the vibration signal and its intermediate features step by step over time, mapping the vibration signal to a working condition-independent feature representation space.
[0099] The fault diagnosis module is configured to classify faults based on the mapped feature representation and output fault diagnosis results.
[0100] It should be noted that the above modules correspond to the steps in Embodiment 1, and the examples and application scenarios implemented by the above modules and their corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules can be executed in a computer system as part of the system.
[0101] In further embodiments, the following is also provided:
[0102] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method described in Embodiment 1. For brevity, further details are omitted here.
[0103] It should be understood that in this embodiment, the processor can be a central processing unit (CPU) and a graphics processing unit (GPU). The processor can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.
[0104] A computer-readable storage medium for storing computer instructions that, when executed by a processor, perform the method of Embodiment 1.
[0105] The method in Example 1 can be directly executed by a hardware processor, or it can be executed by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.
[0106] A computer program product includes a computer program that, when executed by a processor, implements the method in Embodiment 1.
[0107] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which execute in a device on a target real or virtual processor to perform the processes / methods described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided among program modules as needed. The machine-executable instructions for the program modules can execute within a local or distributed device. In a distributed device, the program modules can reside in both local and remote storage media.
[0108] The computer program code used to implement the methods of the present invention may be written in one or more programming languages. This computer program code may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the computer or other programmable data processing device, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a computer, partially on a computer, as a stand-alone software package, partially on a computer and partially on a remote computer, or entirely on a remote computer or server.
[0109] In the context of this invention, computer program code or related data may be carried by any suitable carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals may include electrical, optical, radio, sound, or other forms of propagation signals, such as carrier waves, infrared signals, etc.
[0110] Those skilled in the art will recognize that the units and algorithm steps described in conjunction with the embodiments herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0111] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.
Claims
1. A fault diagnosis operating condition domain generalization method based on operating condition-independent feature representation, characterized in that, Includes the following steps: Collect vibration signals and corresponding operating condition information of the target equipment under normal and fault conditions under multiple operating conditions; The vibration signal is cut into segments of a preset length, and the working condition information is mapped into a working condition vector with the same length as the vibration signal segment. The operating condition information includes at least one of speed, load, and torque, and multiple operating condition variables are mapped to vectors of the same length as the vibration signal segment through independent fully connected layers and then summed and fused. The parameters of the fully connected layer are updated via backpropagation. Modulation parameters are generated by a learnable multilayer perceptron using the operating condition vector. These modulation parameters are used to adjust the vibration signal and its intermediate features step by step over time, mapping the vibration signal to an operating condition-independent feature representation space. The modulation parameters include two sets. The first set of modulation parameters includes shift, scale, and gate, which are used to modulate the input vibration signal. The second set of modulation parameters includes shift_mlp, scale_mlp, and gate_mlp, which are used to modulate the intermediate features of the vibration signal. The vibration signal is mapped to a condition-independent feature representation space through a covariate knowledge embedding module, which performs the following operations: The input vibration signal is subjected to layer normalization processing; The normalized vibration signal is modulated using shift and scale from the first set of modulation parameters; The modulated vibration signal is input into the feature extraction layer to extract features; The extracted features are adjusted using the gate in the first set of modulation parameters and residually connected to the original input vibration signal. The CKE module has the following technical features: (1) Normalized operating condition domain: The CKE module does not simply adjust features based on operating conditions. Instead, it adjusts the input signal and intermediate features by generating a parameter set based on the operating condition vector, learns an "operating condition-independent representation space", namely the operating condition normalization domain, and actively maps signals under different operating conditions to this domain, thereby significantly reducing statistical differences between operating conditions. (2) Time step domain alignment: The modulation parameter set is generated based on the operating condition sequence and is a vector of length T, which enables the model to be normalized at each time step according to the local operating condition influence. (3) Dual-channel modulation: Single operating condition modulation is equivalent to: ； It can only adjust the amplitude and phase effects caused by changes in operating conditions, but cannot completely eliminate the influence of operating conditions on the signal. It uses one set of modulation parameters for the original input signal and another set of modulation parameters for the intermediate features. (4) Plug-and-play operating condition normalization module: The input and output dimensions of the CKE module are the same, and each CKE module can be embedded with different layers of feature extraction networks to achieve "plug and play" generalization capability. Fault classification is performed based on the mapped feature representation, and fault diagnosis results are output.
2. The fault diagnosis operating condition domain generalization method based on operating condition-independent feature representation as described in claim 1, characterized in that, The feature extraction layer can be any of the following: multilayer perceptron, convolutional neural network, or temporal Transformer network, and the covariate knowledge embedding module can be stacked and applied to multiple feature extraction layers.
3. The fault diagnosis operating condition domain generalization method based on operating condition-independent feature representation as described in claim 1, characterized in that, The method uses data from only a single or a few operating conditions during the training phase, and still has the ability to generalize fault diagnosis even when no operating conditions are encountered.
4. A fault diagnosis operating condition domain generalization system based on operating condition-independent feature representation, characterized in that, include: The data acquisition module is configured to: acquire vibration signals and corresponding operating condition information of the target equipment under normal and fault conditions under multiple operating conditions; The operating condition information includes at least one of speed, load, and torque, and multiple operating condition variables are mapped to vectors of the same length as the vibration signal segment through independent fully connected layers and then summed and fused. The parameters of the fully connected layer are updated via backpropagation. The data processing module is configured to: cut the vibration signal into segments of a preset length and map the working condition information into a working condition vector with the same length as the vibration signal segment; The normalization module is configured to generate modulation parameters using a learnable multilayer perceptron via a working condition vector. The modulation parameters are used to adjust the vibration signal and its intermediate features step by step over time, mapping the vibration signal to a working condition-independent feature representation space. The modulation parameters include two sets. The first set of modulation parameters includes shift, scale, and gate, which are used to modulate the input vibration signal. The second set of modulation parameters includes shift_mlp, scale_mlp, and gate_mlp, which are used to modulate the intermediate features of the vibration signal. The vibration signal is mapped to a condition-independent feature representation space through a covariate knowledge embedding module, which performs the following operations: The input vibration signal is subjected to layer normalization processing; The normalized vibration signal is modulated using shift and scale from the first set of modulation parameters; The modulated vibration signal is input into the feature extraction layer to extract features; The extracted features are adjusted using the gate in the first set of modulation parameters and residually connected to the original input vibration signal. The CKE module has the following technical features: (1) Normalized operating condition domain: The CKE module does not simply adjust features based on operating conditions. Instead, it adjusts the input signal and intermediate features by generating a parameter set based on the operating condition vector, learns an "operating condition-independent representation space", namely the operating condition normalization domain, and actively maps signals under different operating conditions to this domain, thereby significantly reducing statistical differences between operating conditions. (2) Time step domain alignment: The modulation parameter set is generated based on the operating condition sequence and is a vector of length T, which enables the model to be normalized at each time step according to the local operating condition influence. (3) Dual-channel modulation: Single operating condition modulation is equivalent to: ； It can only adjust the amplitude and phase effects caused by changes in operating conditions, but cannot completely eliminate the influence of operating conditions on the signal. It uses one set of modulation parameters for the original input signal and another set of modulation parameters for the intermediate features. (4) Plug-and-play operating condition normalization module: The input and output dimensions of the CKE module are the same, and each CKE module can be embedded with different layers of feature extraction networks to achieve "plug and play" generalization capability. The fault diagnosis module is configured to classify faults based on the mapped feature representation and output fault diagnosis results.
5. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method according to any one of claims 1-3.
6. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, perform the method described in any one of claims 1-3.
7. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the method described in any one of claims 1-3.