Electro-hydraulic actuator multi-channel heterogeneous signal small sample hierarchical fault diagnosis method

Abstract

The application relates to the technical field of general equipment fault diagnosis, and particularly relates to a multi-channel heterogeneous signal small sample hierarchical fault diagnosis method for an electro-hydraulic actuator, which comprises the following steps: acquiring multi-channel monitoring signals of the electro-hydraulic actuator at a target time sequence, wherein the monitoring signals are used for indicating the running state of internal components contained in the electro-hydraulic actuator, and the internal components at least include a driving motor, a hydraulic pump, a valve control unit and a hydraulic cylinder; dividing the monitoring signals into hydraulic side signals or motor side signals according to the sources of the monitoring signals; performing feature extraction on the hydraulic side signals to obtain hydraulic side features, performing feature extraction on the motor side signals to obtain motor side features, and fusing the hydraulic side features and the motor side features to obtain comprehensive features; and identifying the fault type of the electro-hydraulic actuator at the target time sequence based on the comprehensive features to obtain a fault diagnosis result of the electro-hydraulic actuator.

Description

Technical Field

[0001] This application relates to the field of equipment fault diagnosis technology, and in particular to a method for diagnosing faults in electro-hydraulic actuators using small-sample hierarchical signals from multi-channel heterogeneous signals. Background Technology

[0002] Electro-hydraulic actuators are key actuating components in complex equipment such as aerospace, naval vessels, and high-end electromechanical servo systems. Their main function is to combine motor drive capability with hydraulic transmission capability to achieve high-precision output of position, speed, or force. Any local performance degradation or component malfunction in an electro-hydraulic actuator can be gradually amplified through the energy transfer chain and control feedback chain, ultimately affecting the stability, responsiveness, and reliability of the entire actuation system. Therefore, conducting high-precision, early-stage, and fine-grained fault diagnosis research on electro-hydraulic actuators is of great significance for ensuring the safe operation of complex equipment and improving independent support capabilities.

[0003] Most existing fault diagnosis methods for electro-hydraulic actuators rely on operational monitoring data, using signal processing, feature extraction, and pattern recognition to achieve state identification and fault classification. With the development of intelligent diagnostic technology, data-driven methods based on deep learning are gradually being applied to fault identification in actuation systems, demonstrating certain advantages in automatic feature extraction and complex pattern discrimination.

[0004] However, existing methods still have significant limitations when applied to practical electro-hydraulic actuators. This is mainly because the monitoring data of electro-hydraulic actuators exhibits significant multi-channel heterogeneous characteristics, and most methods simply splice or uniformly model the multi-channel signals, easily obscuring key fault features and thus affecting the accuracy of the final fault diagnosis. Therefore, a better hierarchical fault diagnosis scheme for multi-channel heterogeneous signals of electro-hydraulic actuators is needed to overcome these difficulties. Summary of the Invention

[0005] In view of this, this application provides a method for stratified fault diagnosis of multi-channel heterogeneous signals of electro-hydraulic actuators with small sample sizes, in order to improve the accuracy of fault diagnosis of electro-hydraulic actuators.

[0006] In a first aspect, this application provides a method for diagnosing faults in electro-hydraulic actuators using small-sample hierarchical analysis of multi-channel heterogeneous signals. The method includes: Acquire multi-channel monitoring signals of the electro-hydraulic actuator at a target timing, the monitoring signals being used to indicate the operating status of the internal components contained in the electro-hydraulic actuator, the internal components including at least a drive motor, a hydraulic pump, a valve control unit, and a hydraulic cylinder; Based on the source of the monitoring signal, the monitoring signal is divided into hydraulic side signal or motor side signal; Feature extraction is performed on the hydraulic side signal to obtain hydraulic side features, feature extraction is performed on the motor side signal to obtain motor side features, and the hydraulic side features and the motor side features are fused to obtain comprehensive features; Based on the comprehensive characteristics, the fault type of the electro-hydraulic actuator in the target timing is identified, and the fault diagnosis result of the electro-hydraulic actuator in the target timing is obtained.

[0007] Furthermore, based on the comprehensive features, the fault type of the electro-hydraulic actuator in the target timing is identified to obtain the fault diagnosis result of the electro-hydraulic actuator in the target timing, including: The comprehensive features are input into a prototype network for fault diagnosis to obtain the target fault type of the electro-hydraulic actuator in the target time sequence and the target fault degree of the electro-hydraulic actuator in the target time sequence, wherein the target fault degree belongs to the target fault type. The method for stratified fault diagnosis of multi-channel heterogeneous signals in electro-hydraulic actuators also includes: Based on the target fault type and the target fault severity, a decision is made regarding the handling of the electro-hydraulic actuator.

[0008] Furthermore, the training process of the prototype network includes: The sample monitoring signals of the electro-hydraulic actuator at the sample timing are acquired through multiple channels. The sample monitoring signals have fault type labels and fault degree labels. Based on the source of the sample monitoring signals, the sample monitoring signals are divided into sample hydraulic side signals or sample motor side signals. Feature extraction is performed on the hydraulic side signal of the sample to obtain the hydraulic side feature, feature extraction is performed on the motor side signal of the sample to obtain the motor side feature, and the hydraulic side feature and the motor side feature are fused to obtain the comprehensive feature of the sample. The sample comprehensive features are input into the prototype network to obtain the sample fault type of the electro-hydraulic actuator in the sample time sequence and the sample fault degree of the electro-hydraulic actuator in the sample time sequence, wherein the sample fault degree belongs to the sample fault type. Based on the differences between the sample fault type and the fault type label, and the differences between the sample fault degree and the fault degree label, the prototype network is trained to obtain the trained prototype network.

[0009] Furthermore, the process of determining the sample fault type of the electro-hydraulic actuator in the sample timing includes: Based on the comprehensive features of samples with the same fault type label, a support set with the fault type label is constructed, and the comprehensive features of the samples in the support set are summed to obtain the sample features sum. Based on the number of samples and the sum of sample features in the support set, the prototype features of the fault types corresponding to the support set are obtained; Based on the distance between the comprehensive features of the sample and the prototype features corresponding to each of the fault types, the sample fault type of the electro-hydraulic actuator in the sample time sequence is determined.

[0010] Furthermore, the process of determining the degree of sample failure of the electro-hydraulic actuator in the sample timing includes: Based on the comprehensive features of samples with the same fault severity label, a support set with the fault severity label is constructed, and the comprehensive features of the samples in the support set are summed to obtain the sample features sum. Based on the number of samples and the sum of sample features in the support set, the prototype features of the fault degree corresponding to the support set are obtained; The sample fault degree of the electro-hydraulic actuator in the sample time sequence is determined based on the distance between the comprehensive features of the sample and the prototype features corresponding to each fault degree.

[0011] Furthermore, prior to feature extraction, the method for stratified fault diagnosis of multi-channel heterogeneous signals in electro-hydraulic actuators, comprising: Obtain the sample mean and sample standard deviation of the sample monitoring signals of the same channel, and determine the channel sample mean and sample monitoring signal difference of the same channel; The normalized sample monitoring signal is determined based on the sample difference and sample standard deviation of the same channel.

[0012] Furthermore, the step of training the prototype network based on the difference between the sample fault type and the fault type label, and the difference between the sample fault degree and the fault degree label, to obtain the trained prototype network, includes: Based on the difference between the sample fault type and the fault type label, a fault type identification loss is constructed; Based on the difference between the sample fault degree and the fault degree label, a fault degree identification loss is constructed; Based on the fault type identification loss and the fault severity identification loss, the total loss of the prototype network is constructed, and the prototype network is trained based on the total loss to obtain the trained prototype network.

[0013] Secondly, this application provides a multi-channel heterogeneous signal small-sample hierarchical fault diagnosis device for electro-hydraulic actuators, the multi-channel heterogeneous signal small-sample hierarchical fault diagnosis device for electro-hydraulic actuators comprising: A signal module is used to acquire multi-channel monitoring signals of the electro-hydraulic actuator at a target timing. The monitoring signals are used to indicate the operating status of the internal components included in the electro-hydraulic actuator. The internal components include at least a drive motor, a hydraulic pump, a valve control unit, and a hydraulic cylinder. The segmentation module is used to segment the monitoring signal into hydraulic side signal or motor side signal according to the source of the monitoring signal; The extraction module is used to extract features from the hydraulic side signal to obtain hydraulic side features, extract features from the motor side signal to obtain motor side features, and fuse the hydraulic side features and the motor side features to obtain comprehensive features; The diagnostic module is used to identify the fault type of the electro-hydraulic actuator in the target timing based on the comprehensive characteristics, and to obtain the fault diagnosis result of the electro-hydraulic actuator in the target timing.

[0014] Thirdly, embodiments of this application also provide a multi-channel heterogeneous signal small-sample hierarchical fault diagnosis system for electro-hydraulic actuators. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the methods described above.

[0015] Fourthly, this application provides a computer-readable storage medium storing computer program instructions thereon, which, when executed by a processor, implement the above-described method. The computer-readable storage medium may be volatile or non-volatile.

[0016] Fifthly, this application provides an electronic device, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to implement the above-described method when executing instructions stored in the memory.

[0017] Sixthly, this application provides a computer program product including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is run in a processor of an electronic device, the processor in the electronic device performs the above-described method.

[0018] In the embodiments of this specification, the multi-channel monitoring signals are first divided into two groups according to their source: hydraulic side signals and motor side signals. Then, fault-related hydraulic side features and motor side features are extracted from each group of signals, and the two types of features are fused. Furthermore, based on the comprehensive features obtained from the fusion, the fault type of the electro-hydraulic actuator is identified. This process, by dividing the hydraulic side signals and motor side signals, avoids the feature overload and information interference problems that occur when directly mixing signals from different sources for modeling. This improves the characterization ability of multi-channel heterogeneous monitoring signals and the fault feature extraction effect, ultimately improving the accuracy of fault diagnosis for electro-hydraulic actuators. In addition, the above process is a fault diagnosis analysis performed on time-series data. It can extract abnormal change features of monitoring signals within a short time range and describe the continuous change relationship between different moments in the fault development process, improving the ability to identify complex time-series fault symptoms and giving the two sets of features good discriminative power and stability. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] Figure 1 The diagram above illustrates a flowchart of a method for diagnosing faults in electro-hydraulic actuators using small-sample hierarchical multi-channel heterogeneous signals.

[0021] Figure 2 The diagram above illustrates the architecture of a multi-channel heterogeneous signal small-sample hierarchical fault diagnosis model for electro-hydraulic actuators.

[0022] Figure 3 The diagram above illustrates a structural schematic of a multi-channel heterogeneous signal small-sample hierarchical fault diagnosis device for electro-hydraulic actuators.

[0023] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0024] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0025] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0026] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0027] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0028] Figure 1 A flowchart illustrating a method for diagnosing small-sample hierarchical faults in multi-channel heterogeneous signals of electro-hydraulic actuators according to an embodiment of this disclosure is shown. This method can be applied to a device for diagnosing small-sample hierarchical faults in multi-channel heterogeneous signals of electro-hydraulic actuators. The device can be a terminal device, a server, or other processing equipment. The terminal device can be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, vehicle-mounted device, wearable device, etc.

[0029] In some possible implementations, the method for diagnosing multi-channel heterogeneous signal small-sample hierarchical faults in electro-hydraulic actuators can be implemented by a processor calling computer-readable instructions stored in memory.

[0030] like Figure 1 As shown, the method for stratified fault diagnosis of multi-channel heterogeneous signals in electro-hydraulic actuators may include: In step S11, the multi-channel monitoring signals of the electro-hydraulic actuator at the target timing are acquired.

[0031] The monitoring signals are used to indicate the operating status of the internal components included in the electro-hydraulic actuator. These monitoring signals can be multi-channel signals such as pressure, flow rate, displacement, current, and rotational speed. The internal components include at least a drive motor, a hydraulic pump, a valve control unit, and a hydraulic cylinder. In addition to the synchronous acquisition of multi-channel monitoring signals from the sensors of the internal components during the operation of the electro-hydraulic actuator, data from the feedback measurement stage can also be used as monitoring signals.

[0032] Since fault identification relies on the continuous changes of signals within a certain time range, rather than the instantaneous value of a single sampling point, it is necessary to acquire the monitoring signal of the electro-hydraulic actuator within a certain time range (i.e., the target timing).

[0033] In step S12, the monitoring signal is divided into hydraulic side signal or motor side signal according to the source of the monitoring signal.

[0034] When the source of the monitoring signal is clear (such as directly from the drive motor or hydraulic pump), the monitoring signal can be directly classified according to its source. Specifically, the monitoring signal can be determined as either a hydraulic-side signal or a motor-side signal based on whether it comes from hydraulic-related equipment or motor-related equipment.

[0035] When the source of the monitoring signal is unknown, classification indicators can be used to determine the segmentation of the monitoring signal. Classification indicators are used to classify the monitoring signal into hydraulic-side signals or motor-side signals. Hydraulic-side signals typically manifest as pressure pulsations, flow changes, displacement response, and control errors; motor-side signals primarily reflect voltage, current, speed, and drive characteristics. Clearly, hydraulic-side signals and motor-side signals differ significantly in physical meaning, dynamic response characteristics, and statistical distribution. Therefore, one or more classification indicators can be preset based on one or more of the following: physical meaning, dimensional scale, spectral distribution, and dynamic response characteristics.

[0036] Because electro-hydraulic actuators operate in a highly dynamic, strongly coupled, and variable operating environment, there is a clear electromechanical-hydraulic coupling relationship between their internal structures such as motors and hydraulic pumps. Therefore, unlike existing methods that use unified modeling or simple splicing to process monitoring signals, step S12 divides the monitoring signals, fully considering the differences and coupling relationships between the hydraulic and motor side signals. This allows for the full discovery of key fault features that adequately express critical fault information, thereby improving the quality of subsequent feature representation and the final diagnostic accuracy.

[0037] In step S13, feature extraction is performed on the hydraulic side signal to obtain hydraulic side features, feature extraction is performed on the motor side signal to obtain motor side features, and the hydraulic side features and the motor side features are fused to obtain comprehensive features.

[0038] Since electro-hydraulic actuator failures usually do not manifest solely on the hydraulic or motor side, but rather cause simultaneous changes in signals on both sides, after classifying the monitoring signals, feature extraction can be performed on each type of monitoring signal to obtain hydraulic side features corresponding to the hydraulic side signals and motor side features corresponding to the motor side signals. The two sets of features are then fused to form a comprehensive feature that can reflect the overall state of the electro-hydraulic actuator.

[0039] Specifically, the fusion operation can be a direct splicing of hydraulic side features and motor side features, or a weighted splicing of hydraulic side features and motor side features. The weight can be determined by experts in complex equipment fields such as aerospace equipment, ship equipment, and high-end electromechanical servo systems to which electro-hydraulic actuators belong, or it can be determined based on historical monitoring data and corresponding fault diagnosis results.

[0040] By fusing the hydraulic side features and motor side features in step S13, information related to the same fault in the hydraulic side and motor side can be merged, so that information related to the same fault in signals from different sources can form a joint expression in the same feature space. This enables subsequent fault diagnosis to be based on the comprehensive state of the entire system, thereby improving the comprehensive characterization ability of compound faults, faults with ambiguous boundaries, and faults under complex operating conditions, and enhancing the adaptability of the fault diagnosis method under actual operating conditions.

[0041] In step S14, based on the comprehensive features, the fault type of the electro-hydraulic actuator in the target timing is identified, and the fault diagnosis result of the electro-hydraulic actuator in the target timing is obtained.

[0042] Specifically, a prototype network for fault diagnosis can be set up. After training, the comprehensive features are input into the prototype network to obtain the fault diagnosis results.

[0043] Existing fault diagnosis methods mostly focus on single-layer identification of fault categories. That is, they usually only output the result of "whether it is a fault" or "what type of fault it is", and lack the ability to further characterize the fault development stage, severity, and internal differences of similar faults.

[0044] In the process of health management and maintenance decision-making for electro-hydraulic actuators, fault type identification alone is often insufficient to support subsequent handling. Further assessment of the fault's stage (early, minor, or severe) is needed to provide a basis for predictive maintenance, residual performance evaluation, and maintenance prioritization. Therefore, how to simultaneously identify fault type and differentiate fault severity under small sample conditions has become a practical requirement in the existing technology. In one possible implementation, based on the comprehensive features, the fault type of the electro-hydraulic actuator at the target time sequence is identified to obtain the fault diagnosis result of the electro-hydraulic actuator at the target time sequence, including: The comprehensive features are input into a prototype network for fault diagnosis to obtain the target fault type of the electro-hydraulic actuator in the target time sequence and the target fault degree of the electro-hydraulic actuator in the target time sequence, wherein the target fault degree belongs to the target fault type. The method for stratified fault diagnosis of multi-channel heterogeneous signals in electro-hydraulic actuators also includes: The maintenance decision for the electro-hydraulic actuator is determined based on the target fault type and the target fault severity.

[0045] Specifically, the prototype network can determine the target fault type of the electro-hydraulic actuator in the target timing based on the comprehensive characteristics, and then determine the target fault degree of the electro-hydraulic actuator in the target timing from the multiple fault degrees included in the target fault type.

[0046] In one example, the fault types in the prototype network can include: normal state, leakage fault, clogging fault, and mixed fault; the fault severity under each fault type can include: normal, early fault, minor fault, and severe fault. The information represented by the fault severity differs for different fault types: for leakage faults, the severity level indicates the amount of leakage; for clogging faults, the severity level indicates the size of the effective flow area; for mixed faults, the severity level indicates the combination of leakage and effective flow area; for normal state, the severity level only indicates normal.

[0047] The above method aims at two-layer fault category identification (i.e., fault type identification and fault severity identification). The output results are not limited to the level of fault presence or fault type, but realize the joint diagnosis of fault "type-severity". It further describes the severity of the fault state, the development stage and the internal differences of similar faults. In this way, it can provide more hierarchical and fine-grained information support for the handling decisions of electro-hydraulic actuators (including health assessment, maintenance decision and condition management, etc.), thereby improving the engineering application value of fault diagnosis results.

[0048] In one example, the prototype network can be part of a fault diagnosis model. Multi-channel monitoring signals can be input into the trained diagnostic model, which then sequentially performs operations such as sample construction, feature extraction, fault type identification, and fault severity identification, ultimately outputting the fault diagnosis result of the electro-hydraulic actuator at the target time sequence, as shown in Equation 1. The architecture of the fault diagnosis model is described in subsequent embodiments and will not be repeated here.

[0049] (Formula 1) in, This represents the fault diagnosis results output by the fault diagnosis model. This indicates the result of fault type identification. This indicates the result of fault degree identification.

[0050] In the embodiments of this specification, the multi-channel monitoring signals are first divided into two groups according to their source: hydraulic side signals and motor side signals. Then, fault-related hydraulic side features and motor side features are extracted from each group of signals, and the two types of features are fused. Furthermore, based on the comprehensive features obtained from the fusion, the fault type of the electro-hydraulic actuator is identified. This process, by dividing the hydraulic side signals and motor side signals, avoids the feature overload and information interference problems that occur when directly mixing signals from different sources for modeling. This improves the characterization ability of multi-channel heterogeneous monitoring signals and the fault feature extraction effect, ultimately improving the accuracy of fault diagnosis for electro-hydraulic actuators. In addition, the above process is a fault diagnosis analysis performed on time-series data. It can extract abnormal change features of monitoring signals within a short time range and describe the continuous change relationship between different moments in the fault development process, improving the ability to identify complex time-series fault symptoms and giving the two sets of features good discriminative power and stability.

[0051] In one possible implementation, the training process of the prototype network includes: The sample monitoring signals of the electro-hydraulic actuator at the sample timing are acquired through multiple channels. The sample monitoring signals have fault type labels and fault degree labels. Based on the source of the sample monitoring signals, the sample monitoring signals are divided into sample hydraulic side signals or sample motor side signals. Feature extraction is performed on the hydraulic side signal of the sample to obtain the hydraulic side feature, feature extraction is performed on the motor side signal of the sample to obtain the motor side feature, and the hydraulic side feature and the motor side feature are fused to obtain the comprehensive feature of the sample. The sample comprehensive features are input into the prototype network to obtain the sample fault type of the electro-hydraulic actuator in the sample time sequence and the sample fault degree of the electro-hydraulic actuator in the sample time sequence, wherein the sample fault degree belongs to the sample fault type. Based on the differences between the sample fault type and the fault type label, and the differences between the sample fault degree and the fault degree label, the prototype network is trained to obtain the trained prototype network.

[0052] The sample monitoring signal can be a monitoring signal of the electro-hydraulic actuator over a historical time period, or a monitoring signal over a simulated time period obtained by modeling and simulating the electric actuator. Similar to the aforementioned monitoring signals, the sample monitoring signal can also be a multi-channel heterogeneous signal.

[0053] In one example, a sliding time window approach can be used to construct samples for continuously monitored signals. Specifically, the sliding window length can be set to L, the window step size to S, and the starting time of the sliding window corresponding to the k-th sample can be denoted as t. kThen the k-th sample is represented as X. k =[x(t k ),x(t k+1 ), ,x(t k+L-1 The sample construction process involves extracting continuously acquired multi-channel monitoring signals into multiple local time segments with a fixed window length. Each time segment can be used as a sample input into the diagnostic model.

[0054] For any sampling time t, the instantaneous multi-channel observation vector of the electro-hydraulic actuator can be expressed as: x(t)=[x1(t),x2(t), ,x M (t)] T Where M represents the total number of monitoring channels, x m (t) represents the observation value of the m-th monitoring channel at time t. That is, x m (t) can represent the vector composed of the monitoring signals of each monitoring channel at the same sampling time.

[0055] During the sample labeling phase, each sample can be assigned a fault type label (ytype k) and a fault severity label (ydeg k) based on information such as the time and type of the fault. This results in a sample set as shown in Formula 2. (Formula 2) Where N represents the total number of samples in the sample set D, X k Let yk represent the k-th sample, ytype k represent the fault type label of the k-th sample, and ydeg k represent the fault degree label of the k-th sample.

[0056] Because different monitoring channels originate from different systems, such as hydraulic systems and motor drive systems, there are significant differences in physical meaning, dimensional range, noise components, and dynamic response speed among the channels. Directly inputting all raw signals into the model can easily lead to problems such as signals with large numerical ranges having an excessive impact on model training, while information with smaller amplitudes but capable of reflecting early faults being masked, resulting in training instability. Therefore, after constructing the sample set, each sample can be preprocessed and then divided into two groups according to its source: hydraulic side signals and motor side signals.

[0057] The preprocessing of the sample monitoring signals may include: performing time alignment on each monitoring channel to eliminate time offsets caused by different acquisition links; repairing missing values ​​or outliers to reduce the impact of sampling interruptions or abrupt changes; filtering and denoising each channel; and normalizing each monitoring channel to reduce the impact of different units on model training. In one possible implementation, before feature extraction, the method for hierarchical fault diagnosis of small samples of multi-channel heterogeneous signals from electro-hydraulic actuators includes: Obtain the sample mean and sample standard deviation of the sample monitoring signals of the same channel, and determine the channel sample mean and sample monitoring signal difference of the same channel; The normalized sample monitoring signal is determined based on the sample difference and sample standard deviation of the same channel.

[0058] The sample mean is the average value of the sample monitoring signals in the same channel, and the sample standard deviation is the standard deviation of the sample monitoring signals in the same channel.

[0059] In one example, the normalization formula shown in Equation 3 can be used to normalize the sample monitoring signal based on the sample mean and sample standard deviation.

[0060] (Formula 3) Where, μ m and σ m Let x represent the sample monitoring signal x of the m-th monitoring channel respectively. m The sample mean and sample standard deviation of (t) This represents the normalized sample monitoring signal.

[0061] After preprocessing, the samples can be divided into hydraulic-side and motor-side sample signal groups based on the source of the monitoring signals. To extract fault-related features from the two types of signals separately, independent feature extraction branches can be established for the hydraulic-side and motor-side sample signal groups. Each branch includes a one-dimensional convolutional layer, a nonlinear activation layer, a pooling layer, and a gated recurrent unit. The one-dimensional convolutional layer is used to extract local change features within a short time range, the pooling layer is used to compress the feature dimension while retaining the main information, and the gated recurrent unit is used to describe the continuous change relationship between the signal and previous time points.

[0062] For hydraulic side sample signals, the feature extraction process can be shown in Equation 4; for motor side sample signals, the feature extraction process can be shown in Equation 5. HH k=GRU H (Pool( (Conv1D H (XH k))))(Formula 4) HE k=GRUE (Pool( (Conv1D E (XE k)))) (Formula 5) Where XHk represents the hydraulic side sample signal, HHk represents the sample hydraulic side feature, XEk represents the motor side sample signal, HEk represents the sample motor side feature, and Conv1D H ( ) and Conv1D E ( The symbols ) represent the one-dimensional convolution operations of the hydraulic side branch and the motor side branch, respectively. ( ) represents a non-linear activation function, Pool( ) represents the pooling operation, GRU H ( ) and GRU E ( ) respectively represent the timing processing of the gated loop unit in the hydraulic side branch and the motor side branch.

[0063] After completing the time-series feature extraction, as shown in Formula 6, the intermediate features of the hydraulic side branch and the motor side branch outputs at each time step can be averaged to obtain two feature vectors: fH k=Avg(HH k), fE k=Avg(HE k) (Formula 6) Where fHk represents the average hydraulic side feature vector of the k-th sample, fEk represents the average motor side feature vector of the k-th sample, and Avg( ) represents the mean operation, HH k represents the hydraulic side feature of the sample, and HE k represents the motor side feature of the sample.

[0064] Formula 6 allows for the extraction of features from the hydraulic side signals reflecting fault information such as pressure fluctuations, abnormal flow rates, and displacement hysteresis, as well as features from the motor side signals reflecting fault information such as speed changes, voltage fluctuations, and abnormal currents. Furthermore, Formula 7 can be used to concatenate the hydraulic and motor side signals to obtain the comprehensive features of the samples.

[0065] (Formula 7) in, The hydraulic side characteristics of the sample after averaging the monitoring signal of the k-th sample are shown. W represents the average of the monitored signals of the k-th sample motor side characteristics. f Let b represent the characteristic transformation matrix. f This represents the bias vector. ( ) represents the nonlinear transformation function, zk This represents the comprehensive characteristics of the monitoring signal of the k-th sample. ; The symbol ] represents a vector concatenation operation.

[0066] Formula 7 allows for the merging of information related to the same fault from both the hydraulic and motor sides, enabling subsequent fault type and severity identification to be based on the overall state of the entire system.

[0067] After obtaining the comprehensive feature representation of the samples, a prototype network can be used to identify the fault type. The basic idea of ​​the prototype network is: in the comprehensive feature space, the mean of the comprehensive features of the supporting samples of each type of fault is used as the prototype of that type of fault, and then the fault type is classified according to the distance between the test sample and the prototypes of each type of fault.

[0068] Electro-hydraulic actuator fault data commonly suffers from small sample size issues. In practical engineering applications, acquiring typical faults, especially early faults, minor faults, and some complex faults, is costly. Some faults are even difficult to reproduce on a large scale in real equipment. Therefore, the number of labeled samples available for training is usually very limited. Traditional classification models that rely on a large number of labeled samples for training are prone to overfitting under these conditions, exhibiting insufficient ability to identify new operating condition samples, fault samples with ambiguous boundaries, and a small number of fault samples, making it difficult to meet the robustness and generalization requirements of engineering scenarios. In one possible implementation, the process of determining the sample fault type of the electro-hydraulic actuator in the sample time series includes: Based on the comprehensive features of samples with the same fault type label, a support set with the fault type label is constructed, and the comprehensive features of the samples in the support set are summed to obtain the sample features sum. Based on the number of samples and the sum of sample features in the support set, the prototype features of the fault types corresponding to the support set are obtained; Based on the distance between the comprehensive features of the sample and the prototype features corresponding to each of the fault types, the sample fault type of the electro-hydraulic actuator in the sample time sequence is determined.

[0069] Specifically, for the c-th type of fault, let its support set be S. c The support set can be extracted from the sample monitoring signals corresponding to the c-th type of fault (which have the fault type label corresponding to the c-th type of fault). The samples in the support set are processed through the aforementioned steps to obtain the sample comprehensive features.

[0070] Prototype features are used to represent the features corresponding to a fault type, and they can be determined based on the support set corresponding to that fault type. In one example, prototype features can be determined using Equation 8: (Formula 8) Where, p c Denotes the prototype features of the c-th type of fault in the comprehensive feature space, |S c | represents the number of samples in the support set for the c-th type of fault, z k This represents the comprehensive characteristics of the k-th sample (i.e., the sample monitoring signal) in the support set for the c-th type of fault.

[0071] For any sample monitoring signal, the comprehensive feature representation of the sample is denoted as z. Then, the distance between it and the feature of the c-th type of fault prototype can be expressed by Equation 9: (Formula 9) Where z represents the sample comprehensive characteristics of the electro-hydraulic actuator, p c Denotes the prototype features of the c-th type of fault in the comprehensive feature space, d c This represents the distance between the composite feature and the prototype feature of the c-th type of fault in the composite feature space.

[0072] The smaller the distance between the sample's overall features and the prototype features of a certain type of fault, the greater the probability that the sample's monitoring signal belongs to the fault type corresponding to that prototype feature. Based on the classification method of the prototype network, the probability that the sample's monitoring signal belongs to the c-th type of fault can be expressed by Formula 10: (Formula 10) Wherein, P(y type =c|z) represents the probability that the electro-hydraulic actuator belongs to the c-th type of fault, d c d represents the distance between the comprehensive features of a sample and the prototype features of the c-th type of fault in the comprehensive feature space. j This represents the distance between the comprehensive feature and the prototype feature of the j-th type of fault in the comprehensive feature space, where j∈[1,C] and C represents the total number of fault types.

[0073] Therefore, Formula 11 can be used to obtain the fault type identification result of the sample to be identified: (Formula 11) in, P(y) represents the fault type identification result. type =c|z) represents the probability that the electro-hydraulic actuator belongs to the c-th type of fault.

[0074] In engineering applications, simply identifying the type of a fault is often insufficient to meet state management requirements; further assessment of the fault's development stage or severity is also necessary. Therefore, after identifying the fault type, this specification further identifies different fault degrees within the identified fault types to achieve a joint output of both "type" and "degree" information. In one possible implementation, the process of determining the sample fault degree of the electro-hydraulic actuator in the sample timing sequence includes: Based on the comprehensive features of samples with the same fault severity label, a support set with the fault severity label is constructed, and the comprehensive features of the samples in the support set are summed to obtain the sample features sum. Based on the number of samples and the sum of sample features in the support set, the prototype features of the fault degree corresponding to the support set are obtained; The sample fault degree of the electro-hydraulic actuator in the sample time sequence is determined based on the distance between the comprehensive features of the sample and the prototype features corresponding to each fault degree.

[0075] Specifically, for the r-th fault severity in the c-th type of fault, let its support set be S. c,r The support set can be extracted from the sample monitoring signal corresponding to the r-th fault level in the c-th fault category (which has the fault type label corresponding to the r-th fault level in the c-th fault category). The samples in the support set are processed by the aforementioned steps to obtain the sample comprehensive features.

[0076] The prototype features of the support set can be determined based on the support set. The sample monitoring signals in the support set are processed through the aforementioned steps to obtain the comprehensive sample feature representation z. k In one example, Formula 12 can be used to determine the prototype characteristics of the r-th fault severity in the c-th type of fault: (Formula 12) Where, p c,r Represents the prototype feature of fault severity r under fault type c in the comprehensive feature space, |S c,r | represents the number of samples in the support set for fault severity r under fault type c, z k This represents the comprehensive characteristics of samples that support centralized sample monitoring signals.

[0077] For any sample to be identified, the prototype network obtains its fault type identification result. Then, only in the set of fault severity corresponding to this fault type. The system further calculates the distance between its sample comprehensive feature representation z and the prototypes of each fault degree: (Formula 13) in, This represents the relationship between the comprehensive feature z of a sample and the fault type in the comprehensive feature space. Prototype characteristics of the lower fault level r The distance between them.

[0078] The smaller the distance between the comprehensive features of a sample and the prototype features of a certain fault degree, the greater the probability that the sample monitoring signal belongs to the fault degree corresponding to that prototype feature. Based on the minimum distance criterion, Formula 14 can be used to obtain the fault degree identification result of the sample to be identified: (Formula 14) in, This indicates the result of fault severity identification. This represents the relationship between the comprehensive feature z of a sample and the fault type in the comprehensive feature space. Prototype characteristics of the lower fault level r The distance between them.

[0079] Furthermore, Formula 15 can also be used to construct the probability prediction of fault severity based on each distance: (Formula 15) in, This indicates that the electro-hydraulic actuator belongs to a fault type. The probability of a failure level r. This represents the relationship between the comprehensive feature z of a sample and the fault type in the comprehensive feature space. Prototype characteristics of the lower fault level r The distance between them This represents the relationship between comprehensive features and fault types in the comprehensive feature space. The distance between the prototype features of the fault degree q, q∈ , Indicates the fault type The set of fault levels r.

[0080] In the above process, fault prototypes are constructed by utilizing the average features (i.e. prototype features) of various fault support samples (i.e. support set samples), and the fault type / degree is determined by the distance between the sample comprehensive features of the sample monitoring signal and the prototype features corresponding to each fault type / degree. This reduces the dependence on a large number of labeled samples and lowers the risk of overfitting that complex classifiers are prone to under small sample conditions, thereby improving the stability and generalization ability of fault type / degree recognition of sample monitoring signals.

[0081] During model training, fault type identification loss and fault type recognition loss can be combined to optimize the prototype network. In one possible implementation, training the prototype network based on the difference between the sample fault type and the fault type label, and the difference between the sample fault severity and the fault severity label, to obtain the trained prototype network includes: Based on the difference between the sample fault type and the fault type label, a fault type identification loss is constructed; Based on the difference between the sample fault degree and the fault degree label, a fault degree identification loss is constructed; Based on the fault type identification loss and the fault severity identification loss, the total loss of the prototype network is constructed, and the prototype network is trained based on the total loss to obtain the trained prototype network.

[0082] Specifically, the prototype network can be optimized using fault type identification loss and fault severity identification loss. The loss function corresponding to the former can be expressed by Equation 16, and the loss function corresponding to the latter can be expressed by Equation 17. (Formula 16) Among them, II ( ) is an indicator function, indicating the actual fault type y of the electro-hydraulic actuator. type Set the value to 1 if it matches the currently identified category c, otherwise set it to 0. type Loss for fault type identification, P(y) type =c|z) represents the probability that the electro-hydraulic actuator belongs to the c-th (∈C) type of fault, z represents the sample comprehensive feature of the sample monitoring signal, and C represents the total number of fault types. By optimizing this loss function, samples of the same type of fault can be clustered towards the corresponding fault prototype in the comprehensive feature space, while samples of different fault types can be separated from each other, thereby improving the accuracy and stability of fault type identification.

[0083] (Formula 17) Among them, II ( ) is an indicator function, representing the actual fault level y of the electro-hydraulic actuator. deg Set the value to 1 if it matches the current recognition level r, otherwise set it to 0. deg To identify the extent of the damage, This indicates that the sample monitoring signal belongs to a fault type. The probability of a fault level r, where z represents the sample comprehensive characteristics of the sample monitoring signal. Indicates the fault type The set of fault degrees r. By optimizing this loss function, samples of the same fault degree under the same fault can be clustered towards the corresponding fault prototype in the comprehensive feature space, while samples of different fault degrees are separated from each other, thereby improving the accuracy and stability of fault degree identification.

[0084] To balance the tasks of fault type identification and fault severity identification, the overall training objective can be represented using Equation 18: L=αL type +βL deg (Formula 18) Where α and β are non-negative weighting coefficients used to adjust the contribution ratio of the fault type identification task and the fault severity identification task to the total loss, and L is the total loss. type To identify losses based on fault type, L deg The loss is used for fault severity identification. By optimizing the total loss mentioned above, similar fault samples can be clustered in the comprehensive feature space towards the corresponding fault type prototype and fault severity prototype, thereby achieving joint optimization of fault type identification and fault severity identification.

[0085] Figure 2 A schematic diagram of the architecture of a multi-channel heterogeneous signal small-sample hierarchical fault diagnosis model for electro-hydraulic actuators is shown. Figure 2 As shown, the multi-channel heterogeneous signal small sample hierarchical fault diagnosis model of the electro-hydraulic actuator can include a data processing module, a feature extraction and fusion module, and a fault identification module.

[0086] The data processing module includes a signal acquisition and sample construction unit and a signal processing and grouping input unit. The former is used to acquire multi-channel signals from the hydraulic and motor sides, and construct sample X using a sliding time window. k Sample labeling (labels include fault type and fault severity); the latter is used for: time alignment, outlier repair, filtering, normalization, and dividing the input into hydraulic side or motor side.

[0087] The feature extraction and fusion module includes a hydraulic side and a motor side feature extraction unit, and a feature fusion and comprehensive representation construction unit. The former is used for extracting independent features (including hydraulic side features and motor side features); the latter is used for concatenating hydraulic side features and motor side features, and performing feature transformation to obtain a comprehensive feature representation.

[0088] The fault identification module includes a fault type identification unit, a fault severity identification unit, and a loss unit. The former is used to: construct prototype vectors for fault types (normal, blockage, leakage, mixed), calculate the distance between the comprehensive feature and each prototype vector, and the probability that the comprehensive feature belongs to each fault type, determine the fault type identification result, and calculate the fault similarity identification loss. The latter is used to: construct prototype vectors for fault severity (early, minor, severe) under the current fault type, calculate the distance between the comprehensive feature and each prototype vector, and the probability that the comprehensive feature belongs to each fault severity (these fault severity levels all belong to the aforementioned determined fault types), determine the fault severity identification result, and calculate the fault severity identification loss. The loss unit is used to calculate the total loss of the fault identification module.

[0089] After model training is completed, the multi-channel monitoring signals during the operation of the electro-hydraulic actuator under test are input into the model. The model is then used to complete sample construction, feature extraction, fault type identification, and fault degree identification in sequence. Finally, the fault diagnosis results of the sample under test can be output.

[0090] Existing electro-hydraulic actuator fault diagnosis technologies have shortcomings in terms of effective characterization of multi-channel monitoring signals, stable identification under small sample conditions, and joint output of fault type and fault degree. There is a lack of a fault diagnosis method that can be used for multi-source monitoring data from both the hydraulic and motor sides of electro-hydraulic actuators, while taking into account both small sample learning capabilities and hierarchical diagnosis requirements.

[0091] The aforementioned multi-channel heterogeneous signal small-sample hierarchical fault diagnosis model for electro-hydraulic actuators employs a novel method for multi-channel heterogeneous signal small-sample hierarchical fault diagnosis of electro-hydraulic actuators. This method can address multi-channel monitoring signals of electro-hydraulic actuators and, under small-sample conditions, achieve sample construction, signal preprocessing, hydraulic side and motor side group feature extraction, and comprehensive feature fusion. Furthermore, it can complete hierarchical fault diagnosis by identifying fault type and fault degree, thereby improving the accuracy, stability, and engineering applicability of electro-hydraulic actuator fault diagnosis.

[0092] The present invention also provides a fault diagnosis device for multi-channel heterogeneous signals with small sample hierarchical structure in electro-hydraulic actuators. Figure 3 This diagram illustrates a block diagram of a multi-channel heterogeneous signal small-sample hierarchical fault diagnosis device for electro-hydraulic actuators according to an embodiment of this specification. This multi-channel heterogeneous signal small-sample hierarchical fault diagnosis device for electro-hydraulic actuators can be a terminal device, a server, or other processing equipment. The terminal device can be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, vehicle-mounted device, wearable device, etc.

[0093] In some possible implementations, the multi-channel heterogeneous signal small-sample hierarchical fault diagnosis device for electro-hydraulic actuators can be implemented by a processor calling computer-readable instructions stored in memory.

[0094] like Figure 3 As shown, the electro-hydraulic actuator multi-channel heterogeneous signal small sample hierarchical fault diagnosis device 30 may include: The signal module 31 is used to acquire multi-channel monitoring signals of the electro-hydraulic actuator at the target timing. The monitoring signals are used to indicate the operating status of the internal components included in the electro-hydraulic actuator. The internal components include at least a drive motor, a hydraulic pump, a valve control unit, and a hydraulic cylinder. The segmentation module 32 is used to segment the monitoring signal into hydraulic side signal or motor side signal according to the source of the monitoring signal; The extraction module 33 is used to extract features from the hydraulic side signal to obtain hydraulic side features, extract features from the motor side signal to obtain motor side features, and fuse the hydraulic side features and the motor side features to obtain comprehensive features; The diagnostic module 34 is used to identify the fault type of the electro-hydraulic actuator in the target timing based on the comprehensive characteristics, and obtain the fault diagnosis result of the electro-hydraulic actuator in the target timing.

[0095] Thirdly, embodiments of this application also provide a multi-channel heterogeneous signal small-sample hierarchical fault diagnosis system for electro-hydraulic actuators. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the methods described above.

[0096] Fourthly, embodiments of this application also provide a computer-readable storage medium storing computer program instructions thereon, which, when executed by a processor, implement the above-described method. The computer-readable storage medium may be volatile or non-volatile.

[0097] Fifthly, embodiments of this application also provide an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to implement the above method when executing instructions stored in the memory.

[0098] Sixthly, this application provides a computer program product including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is run in a processor of an electronic device, the processor in the electronic device performs the above-described method.

[0099] This invention is now complete.

[0100] In summary, in the embodiments of this specification, the multi-channel monitoring signals are first divided into two groups according to their source: hydraulic side signals and motor side signals. Then, fault-related hydraulic side features and motor side features are extracted from each group of signals, and the two types of features are fused. Finally, based on the fused comprehensive features, the fault type of the electro-hydraulic actuator is identified. This process, by dividing the hydraulic side signals and motor side signals, avoids the feature overload and information interference problems that occur when directly mixing signals from different sources for modeling. This improves the characterization ability of multi-channel heterogeneous monitoring signals and the fault feature extraction effect, ultimately improving the accuracy of fault diagnosis for electro-hydraulic actuators. Furthermore, the above process is a fault diagnosis analysis performed on time-series data. It can extract abnormal change features of monitoring signals within a short time range and describe the continuous change relationship between different moments in the fault development process, improving the ability to identify complex time-series fault symptoms and giving the two sets of features good discriminative power and stability.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for stratified fault diagnosis of multi-channel heterogeneous signals in electro-hydraulic actuators using small samples, characterized in that, The method for stratified fault diagnosis of multi-channel heterogeneous signals in electro-hydraulic actuators includes: Acquire multi-channel monitoring signals of the electro-hydraulic actuator at a target timing, the monitoring signals being used to indicate the operating status of the internal components contained in the electro-hydraulic actuator, the internal components including at least a drive motor, a hydraulic pump, a valve control unit, and a hydraulic cylinder; Based on the source of the monitoring signal, the monitoring signal is divided into hydraulic side signal or motor side signal; Feature extraction is performed on the hydraulic side signal to obtain hydraulic side features, feature extraction is performed on the motor side signal to obtain motor side features, and the hydraulic side features and the motor side features are fused to obtain comprehensive features; Based on the comprehensive characteristics, the fault type of the electro-hydraulic actuator in the target timing is identified, and the fault diagnosis result of the electro-hydraulic actuator in the target timing is obtained.

2. The electro-hydraulic actuator multi-channel heterogeneous signal small sample layered fault diagnosis method according to claim 1, characterized in that, Based on the comprehensive features, the fault type of the electro-hydraulic actuator in the target timing is identified to obtain the fault diagnosis result of the electro-hydraulic actuator in the target timing, including: The comprehensive features are input into a prototype network for fault diagnosis to obtain the target fault type of the electro-hydraulic actuator in the target time sequence and the target fault degree of the electro-hydraulic actuator in the target time sequence, wherein the target fault degree belongs to the target fault type. The method for stratified fault diagnosis of multi-channel heterogeneous signals in electro-hydraulic actuators also includes: Based on the target fault type and the target fault severity, a decision is made regarding the handling of the electro-hydraulic actuator.

3. The electro-hydraulic actuator multi-channel heterogeneous signal small sample layered fault diagnosis method according to claim 2, characterized in that, The training process of the prototype network includes: The sample monitoring signals of the electro-hydraulic actuator at the sample timing are acquired through multiple channels. The sample monitoring signals have fault type labels and fault degree labels. Based on the source of the sample monitoring signals, the sample monitoring signals are divided into sample hydraulic side signals or sample motor side signals. Feature extraction is performed on the hydraulic side signal of the sample to obtain the hydraulic side feature, feature extraction is performed on the motor side signal of the sample to obtain the motor side feature, and the hydraulic side feature and the motor side feature are fused to obtain the comprehensive feature of the sample. The sample comprehensive features are input into the prototype network to obtain the sample fault type of the electro-hydraulic actuator in the sample time sequence and the sample fault degree of the electro-hydraulic actuator in the sample time sequence, wherein the sample fault degree belongs to the sample fault type. Based on the differences between the sample fault type and the fault type label, and the differences between the sample fault degree and the fault degree label, the prototype network is trained to obtain the trained prototype network.

4. The electro-hydraulic actuator multi-channel heterogeneous signal small sample layered fault diagnosis method according to claim 3, characterized in that, The process of determining the sample fault type of the electro-hydraulic actuator in the sample timing includes: Based on the comprehensive features of samples with the same fault type label, a support set with the fault type label is constructed, and the comprehensive features of the samples in the support set are summed to obtain the sample features sum. Based on the number of samples and the sum of sample features in the support set, the prototype features of the fault types corresponding to the support set are obtained; Based on the distance between the comprehensive features of the sample and the prototype features corresponding to each of the fault types, the sample fault type of the electro-hydraulic actuator in the sample time sequence is determined.

5. The electro-hydraulic actuator multi-channel heterogeneous signal small sample layered fault diagnosis method according to claim 3, characterized in that, The process of determining the degree of sample failure of the electro-hydraulic actuator in the sample timing includes: Based on the comprehensive features of samples with the same fault severity label, a support set with the fault severity label is constructed, and the comprehensive features of the samples in the support set are summed to obtain the sample features sum. Based on the number of samples and the sum of sample features in the support set, the prototype features of the fault degree corresponding to the support set are obtained; The sample fault degree of the electro-hydraulic actuator in the sample time sequence is determined based on the distance between the comprehensive features of the sample and the prototype features corresponding to each fault degree.

6. The method for stratified fault diagnosis of multi-channel heterogeneous signals in electro-hydraulic actuators according to claim 3, characterized in that, Before feature extraction, the method for stratified fault diagnosis of multi-channel heterogeneous signals of electro-hydraulic actuators using small samples includes: Obtain the sample mean and sample standard deviation of the sample monitoring signals of the same channel, and determine the channel sample mean and sample monitoring signal difference of the same channel; The normalized sample monitoring signal is determined based on the sample difference and sample standard deviation of the same channel.

7. The electro-hydraulic actuator multi-channel heterogeneous signal small sample layered fault diagnosis method according to claim 3, characterized in that, The prototype network is trained based on the difference between the sample fault type and the fault type label, and the difference between the sample fault severity and the fault severity label, to obtain the trained prototype network, including: Based on the difference between the sample fault type and the fault type label, a fault type identification loss is constructed; Based on the difference between the sample fault degree and the fault degree label, a fault degree identification loss is constructed; Based on the fault type identification loss and the fault severity identification loss, the total loss of the prototype network is constructed, and the prototype network is trained based on the total loss to obtain the trained prototype network.

8. An electro-hydraulic actuator multi-channel heterogeneous signal small sample hierarchical fault diagnosis device, characterized in that, The electro-hydraulic actuator multi-channel heterogeneous signal small sample hierarchical fault diagnosis device includes: A signal module is used to acquire multi-channel monitoring signals of the electro-hydraulic actuator at a target timing. The monitoring signals are used to indicate the operating status of the internal components included in the electro-hydraulic actuator. The internal components include at least a drive motor, a hydraulic pump, a valve control unit, and a hydraulic cylinder. The segmentation module is used to segment the monitoring signal into hydraulic side signal or motor side signal according to the source of the monitoring signal; The extraction module is used to extract features from the hydraulic side signal to obtain hydraulic side features, extract features from the motor side signal to obtain motor side features, and fuse the hydraulic side features and the motor side features to obtain comprehensive features; The diagnostic module is used to identify the fault type of the electro-hydraulic actuator in the target timing based on the comprehensive characteristics, and to obtain the fault diagnosis result of the electro-hydraulic actuator in the target timing.

9. An electronic device, comprising: include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A storage medium, characterized by The storage medium is configured to store computer-executable instructions that cause a computer to perform the method of any one of claims 1 to 7.

Images