A method and device for monitoring the structural health of an electric fuel pump based on stress waves

Abstract

The present application provides a method and device for monitoring the structural health of an electric fuel pump based on stress waves, which belongs to the field of fault prediction and health management. The method provided by the present application includes: triggering a pulse electrical signal at the starting end of the electric fuel pump, and converting the pulse electrical signal into a stress wave signal incident medium; collecting the stress wave signal at the receiving end of the electric fuel pump, and converting the collected stress wave signal into an electrical signal; wherein the stress wave signal propagates along the direction of the external load of the electric fuel pump structure, and propagates from the starting end to the receiving end; receiving the electrical signal and the pulse electrical signal at the starting end, and performing health monitoring of the electric fuel pump structure according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end. The method and device for monitoring the structural health of an electric fuel pump based on stress waves provided by the present application can conveniently and comprehensively detect the health status of an electric fuel pump.

Description

Technical Field

[0001] The present application relates to the technical field of fault prediction and health management, and in particular to a method and device for monitoring the structural health of an electric fuel pump based on stress waves. Background Art

[0002] The electric fuel pump is the core accessory of the helicopter fuel system. It can provide the engine with fuel at a specified flow rate and pressure. Its working condition directly affects the performance of the helicopter engine. Therefore, early detection of electric fuel pump failures and maintenance can greatly improve the safety, reliability and economic benefits of the helicopter.

[0003] However, the existing helicopter HUMS (operation and monitoring system) system mainly monitors the helicopter's rotor system. The monitoring and diagnosis methods for the electromechanical system are difficult to meet the needs of the health monitoring of the electromechanical system, and there is even less support for the monitoring and diagnosis of the electric fuel pump. The current commonly used method for monitoring electric fuel pump faults is to use ultrasonic non-destructive testing (NDT) technology for detection, which requires the identification of the operating parameters of the electric fuel pump, and the identification process cannot identify many external defects of the electric fuel pump. The overall fault identification process is not comprehensive and convenient enough. Summary of the invention

[0004] In view of this, the present application provides a method and device for monitoring the structural health of an electric fuel pump based on stress waves, which can conveniently and comprehensively detect the health status of the electric fuel pump.

[0005] Specifically, the present application is implemented through the following technical solutions:

[0006] In a first aspect, the present application provides a method for monitoring the structural health of an electric fuel pump based on stress waves, the method comprising:

[0007] Triggering a pulse electrical signal at the starting end of the electric fuel pump, and converting the pulse electrical signal into a stress wave signal incident on a medium;

[0008] Collecting stress wave signals at the receiving end of the electric fuel pump and converting the collected stress wave signals into electrical signals; wherein the stress wave signals propagate along the direction of the external load of the electric fuel pump structure and propagate from the starting end to the receiving end;

[0009] The electrical signal and the pulse electrical signal at the starting end are received, and health monitoring of the electric fuel pump structure is performed according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end.

[0010] A second aspect of the present application provides a stress wave-based electric fuel pump structural health monitoring device, the device comprising a trigger module, a collection module and a receiving module;

[0011] Wherein, the trigger module is used to trigger a pulse electrical signal at the starting end of the electric fuel pump and convert the pulse electrical signal into a stress wave signal incident on the medium;

[0012] The acquisition module is used to acquire stress wave signals at the receiving end of the electric fuel pump and convert the acquired stress wave signals into electrical signals; wherein the stress wave signals propagate along the direction of the external load of the electric fuel pump structure and propagate from the starting end to the receiving end;

[0013] The receiving module is used to receive the electrical signal and the pulse electrical signal at the starting end, and perform health monitoring of the electric fuel pump structure according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end.

[0014] The electric fuel pump structure health monitoring method and device based on stress waves provided in the present application monitor the health of the electric fuel pump structure based on stress waves. It does not need to identify the operating parameters of the product, but identifies damage by comparing the response signal of the structure being tested with that of the undamaged structure or certain characteristic parameters of the signal. Therefore, the detection process does not require the product to be in operation, and the detection is more convenient. In addition, the use of stress waves for the health monitoring of the electric fuel pump structure allows the placement of transmitting and receiving probes anywhere on the surface of the component, which facilitates comprehensive monitoring. In addition, the monitoring method based on stress waves propagates the stress wave signal along the direction of the external load of the electric fuel pump structure, and can detect defects perpendicular to the surface being tested that cannot be detected by ultrasonic non-destructive testing methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a monitoring device shown as an exemplary embodiment of the present application;

[0016] Figure 2 A flowchart of Embodiment 1 of the method for structural health monitoring of an electric fuel pump based on stress waves provided in the present application;

[0017] Figure 3 A flowchart of Embodiment 2 of the method for structural health monitoring of an electric fuel pump based on stress waves provided in this application;

[0018] Figure 4 This is a structural schematic diagram of Example 1 of the stress wave-based electric fuel pump structural health monitoring device provided in the present application. DETAILED DESCRIPTION

[0019] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0020] The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in this article refers to and includes any or all possible combinations of one or more associated listed items.

[0021] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0022] Specific embodiments are given below to introduce the technical solution of the present application in detail.

[0023] Before introducing this method, the device used in this method is introduced:

[0024] Figure 1 This is a schematic diagram of a monitoring device shown in an exemplary embodiment of the present application, please refer to Figure 1 , the device includes a waveform generator, a signal amplifier, a signal conversion device, a signal receiver and a signal processing device;

[0025] Wherein, the waveform generator is used to transmit a pulse electrical signal;

[0026] The signal amplifier is used to amplify the pulse electrical signal;

[0027] The signal conversion device is used to convert the pulse electrical signal after signal amplification into a stress wave signal, and convert the received stress wave signal into an electrical signal after receiving the transmitted stress wave signal;

[0028] The signal receiver is used to simultaneously receive the pulse electrical signal generated by the waveform generator and the electrical signal converted from the stress wave signal;

[0029] The signal processing device is used to analyze, process and store the pulse electrical signal and the electrical signal.

[0030] It should be noted that the waveform generator can be an arbitrary waveform generator, and the waveform generator generates a pulse electrical signal, which is amplified by a signal amplifier (which can be an arbitrary signal amplifier), and then converted into a stress wave signal incident in the medium by a signal conversion device at the starting end of the product (for example, a transmitting transducer), and the product receiving end receives the stress wave signal obtained after propagation (which may be different from the incident stress wave signal), and converts the stress wave signal obtained after propagation into an electrical signal through a signal conversion device (which can be a multi-channel receiving transducer), and inputs it into a signal receiver. At the same time, the signal receiver will also synchronously receive the pulse electrical signal generated by the waveform generator, and finally process and analyze the pulse electrical signal and the electrical signal through a signal processing device to obtain the monitoring result and store it.

[0031] Figure 2 This is a flowchart of the first embodiment of the electric fuel pump structural health monitoring method based on stress waves provided in this application. Please refer to Figure 2 , the method provided in this embodiment may include:

[0032] S201, triggering a pulse electrical signal at the starting end of the electric fuel pump, and converting the pulse electrical signal into a stress wave signal and injecting it into a medium.

[0033] It should be noted that a pulse electrical signal refers to a signal in which the voltage or current changes rapidly in a short period of time; a stress wave signal refers to a mechanical wave propagating in a material or structure, which is used to detect and evaluate defects inside the material and the health of the structure. Among them, the starting end corresponds to the receiving end, and both are set according to actual monitoring needs, and can be set anywhere on the surface of the component detected by the electric fuel pump. It should also be noted that the medium refers to the pump structure material of the electric fuel pump.

[0034] Specifically, an arbitrary waveform generator is used at the starting end to generate a pulse electrical signal, and then a signal amplifier is used to amplify the generated pulse electrical signal to increase the signal strength and enhance the signal penetration, thereby ensuring that the amplified signal is not easily distorted during the propagation process. Then, the amplified signal is converted into a stress wave signal through a signal conversion device at the starting end, wherein the signal conversion device at this location can be a transmitting transducer.

[0035] S202, collecting stress wave signals at the receiving end of the electric fuel pump, and converting the collected stress wave signals into electrical signals; wherein the stress wave signals propagate along the direction of the external load on the electric fuel pump structure, and propagate from the starting end to the receiving end.

[0036] It should be noted that a sensor can be arranged at the receiving end to collect stress wave signals, and the collected stress wave signals may be different from the incident stress wave signals. Furthermore, by setting a multi-channel signal conversion device, the collected stress wave signals are converted into electrical signals, and the multi-channel signal conversion device here can be a multi-channel receiving transducer. Among them, the multi-channel receiving transducer can simultaneously receive signals from multiple different positions or directions through multiple channels, improve the signal reception capacity, and reduce the signal loss or attenuation problem that may exist in a single channel. In this way, by integrating the signals of multiple channels, it is convenient to accurately locate the source of the signal, so that the defects and crack positions in the structure can be accurately detected.

[0037] In practical applications, when the stress wave signal propagates in the pump structure, due to the material properties of the structure and the direction of the external load, the signal will propagate along a specific path, and when encountering cracks, defects or other anomalies, the signal will be reflected, refracted or scattered.

[0038] S203, receiving the electrical signal and the pulse electrical signal at the starting end, and performing health monitoring of the electric fuel pump structure according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end.

[0039] It should be noted that a signal receiver is used to receive the electrical signal obtained after signal propagation and the pulse electrical signal at the starting end of synchronous reception. The signal receiver here may be a multi-channel ultrasonic receiver, and then the two received electrical signals are input into a signal processing device for processing. The signal processing device here may be a computer or other equipment.

[0040] It should be noted that before the health monitoring of the electric fuel pump structure is performed according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end, it includes: determining the signal frequency domain intervals under each health state according to the frequency domain characteristics of the historical fault signal of the electric fuel pump, and the various health states include healthy state, fault state and noise interference state; determining the constraints under each health state according to the signal frequency domain interval; determining the corresponding target signal frequency domain interval according to the constraints and the frequency domain characteristics of the received electrical signal; preprocessing the received electrical signal according to the preprocessing method corresponding to the target signal frequency domain interval, and the preprocessing means corresponding to different signal frequency domain intervals are different. By analyzing the frequency domain characteristics of the historical fault signal of the electric fuel pump to determine the respective signal frequency domain intervals under the health state (healthy, faulty, noise interference), the frequency domain feature models corresponding to different states can be established based on past experience. This allows for more accurate judgment of the health state to which the received electrical signal may belong in subsequent monitoring. For example, if the frequency domain characteristics of the electrical signal fall within a predetermined fault state frequency domain interval, the electric fuel pump can be more specifically suspected of having a potential fault, rather than simply relying on a general judgment based on real-time data, thereby improving the accuracy of the health status judgment.

[0041] In addition, based on the determined signal frequency domain interval, the constraints are further determined, and clear judgment boundaries are set for different health states. During real-time monitoring, the target signal frequency domain interval is determined based on the constraints and the frequency domain characteristics of the received electrical signal, which can more accurately evaluate the signal. For example, when judging whether an electric fuel pump is in a healthy state, it is not a broad comparison of real-time data with a universal standard, but under the determined health state constraints, the target signal frequency domain interval is compared to see if it meets the health state requirements, reducing the possibility of misjudgment.

[0042] Furthermore, different preprocessing methods are set to correspond to different signal frequency domain intervals, and the most appropriate preprocessing means can be selected according to the target signal frequency domain interval to which the electrical signal belongs. For example, for signal frequency domain intervals that are easily interfered by high-frequency noise, more effective preprocessing methods such as filtering can be used to remove noise and enhance the effectiveness of the signal. For signals in other frequency domain intervals, different preprocessing methods can be used, such as amplifying or attenuating signals in a specific frequency domain, so that the subsequent analysis of the electrical signal is more accurate and reliable, and avoids the masking or distortion of certain signal features due to a unified preprocessing method. It should also be noted that before the health monitoring of the electric fuel pump structure is performed based on the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end, it also includes: obtaining real-time monitoring performance data and stress wave data of the electric fuel pump in healthy and faulty states, and determining constraint standards based on the frequency domain characteristics of the acquired data; the real-time monitoring performance data at least includes the flow rate and boost value of the product; continuously collecting real-time monitoring performance data and stress wave data, and comparing the characteristic values ​​of the collected real-time monitoring performance data and the calculated collected stress wave data with the constraint standards. If it is within the healthy state of the constraint standards, it is a healthy state; otherwise, it is a faulty state.

[0043] Among them, by determining the constraint standards, a clear basis for judgment can be provided to distinguish between normal and faulty states, and the abnormal fluctuations in monitoring data that may be caused by noise and other interference signals can be reduced, and the probability of false alarms (i.e., misjudgment of normal conditions as faults) and missed alarms (i.e., misjudgment of fault conditions as normal conditions) can be reduced, thereby improving the reliability and accuracy of diagnosis. In addition, the constraint standards can also be used for warning and guiding repair and maintenance.

[0044] Specifically, in one embodiment, the product flow rate (15000kg / h) and the boost value (>200kpa) are used to determine whether the product is in a normal operating state, and then in a healthy state, that is, when the electric fuel pump is operating normally, a large amount of real-time monitoring performance data and stress wave data are collected; in a faulty state, that is, when the electric fuel pump fails or is artificially faulted, the corresponding monitoring performance data and stress wave data are collected. The collected data is preprocessed, including denoising, normalization, feature extraction, etc., and the time domain features and frequency domain features are obtained through feature extraction. The data in the healthy state and the faulty state are analyzed to find out their differences in various features and determine the feature threshold. Since the time domain signal of the high-frequency stress wave signal is easily affected by noise, it is difficult to judge directly by the time domain amplitude. Therefore, a fault judgment criterion based on frequency characteristics (such as peak SWF, energy integral SWF) is formulated as a constraint standard. According to the analysis results, a threshold is set for each feature to form a constraint standard for judging the health state of the electric fuel pump. For example, the range of peak SWF in a healthy state is AB, and the range in a fault state is CD; the range of energy integral SWF in a healthy state is EF, and the range in a fault state is GH.

[0045] After determining the constraint standards, continuously collect real-time monitoring performance data and stress wave data and perform the above-mentioned processing, perform feature extraction and calculate the eigenvalues ​​of the current data, compare the real-time eigenvalues ​​with the constraint standards, and if all eigenvalues ​​are within the range of the healthy state, the electric fuel pump is judged to be in a healthy state; if some eigenvalues ​​exceed the range of the healthy state, enter the fault detection process. Specifically, a continuous monitoring mechanism can be set up, and after an abnormality is detected within a continuous preset time, it is confirmed as a fault state. In this embodiment, there is no limitation on the continuous preset time, for example, it can be 5 consecutive seconds.

[0046] Combining the operations of the above two parts before performing health monitoring of the electric fuel pump structure according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end, the preprocessing method and constraints under different health states can be determined through the operations of the first part, and then the operations of the second part can be used to collect data in real time and judge the health state according to the constraint standards, thus forming a complete stress wave-based electric fuel pump structure health monitoring process, which ensures the effectiveness and accuracy of the monitoring method from different aspects.

[0047] It should also be noted that the health monitoring of the electric fuel pump structure is performed according to the time domain and frequency domain characteristics of the received electric signal and the pulse electric signal at the starting end, including: determining the number of different time scales of the received electric signal; determining the characteristic scale of the received electric signal according to the number of different time scales, and the number of characteristic scales corresponds to the number of different time scales; respectively extracting the time domain characteristics and frequency domain characteristics of the received electric signal at each characteristic scale. By determining the number of different time scales of the electric signal and determining the characteristic scale accordingly, the electric signal can be observed from multiple time scale levels. Signals at different time scales may contain different types of information. For example, a short time scale may reflect local, rapidly changing signal characteristics, which are related to small defects or local stress changes in the electric fuel pump structure; a long time scale may reflect overall, trending signal characteristics, which are related to the overall structural state of the electric fuel pump or long-term operating changes. Such a comprehensive analysis helps to avoid missing important information and more completely understand the structural health of the electric fuel pump.

[0048] The stress wave signal generated by the electric fuel pump during operation may have complex characteristics, and its signal characteristics may show different patterns at different time scales. The multi-scale analysis method can better adapt to this complexity. And extracting time domain features and frequency domain features at each characteristic scale respectively can extract features according to the characteristics of the signal at different time scales. For example, at a short time scale characteristic scale, the extracted time domain features may highlight the instantaneous peak value, pulse width and other features of the signal related to local impact or micro cracks; while in the corresponding frequency domain features, the changes in high-frequency components may be highlighted, which is also related to local structural changes. At a long time scale characteristic scale, the time domain features may pay more attention to the mean and variance of the signal and other parameters reflecting the overall stability, while the frequency domain features may focus on the distribution of low-frequency components, which are related to the vibration characteristics of the overall structure or the state of large components. Such targeted feature extraction can more accurately reflect the health status of the electric fuel pump at different structural levels. The electric fuel pump structure health monitoring method based on stress wave provided in this embodiment triggers a pulse electric signal at the starting end of the electric fuel pump, and converts the pulse electric signal into a stress wave signal incident medium, then collects the stress wave signal at the receiving end of the electric fuel pump, and converts the collected stress wave signal into an electric signal; wherein the stress wave signal propagates along the direction of the external load of the electric fuel pump structure, and propagates from the starting end to the receiving end, and finally receives the electric signal and the pulse electric signal at the starting end, and performs health monitoring of the electric fuel pump structure according to the received electric signal and the pulse electric signal at the starting end. In this way, the health monitoring of the electric fuel pump structure based on stress wave does not need to identify the operating parameters of the product, but identifies the damage by comparing the response signal or certain characteristic parameters of the signal of the detected structure with that of the undamaged structure, so the detection process does not require the product to be in operation, and the detection is more convenient, and the use of stress wave for the health monitoring of the electric fuel pump structure can place the transmitting and receiving probes anywhere on the surface of the component, which is convenient for comprehensive monitoring. In addition, based on the monitoring method of stress waves, the stress wave signal propagates along the direction of the external load on the electric fuel pump structure, and can detect defects perpendicular to the inspected surface that cannot be detected by ultrasonic non-destructive testing methods.

[0049] Figure 3 This is a flow chart of Embodiment 2 of the electric fuel pump structural health monitoring method based on stress waves provided in this application. Figure 3 Based on the above embodiment, the health monitoring of the electric fuel pump structure is performed according to the received electrical signal and the pulse electrical signal at the starting end, and then includes:

[0050] S301, extracting data features from the stress wave signal represented by the electrical signal; the data feature extraction at least includes time domain feature extraction and frequency domain feature extraction.

[0051] Specifically, extracting data features from the stress wave signal represented by the electrical signal includes:

[0052] Preprocessing the stress wave signal represented by the electrical signal and extracting the stress wave factor;

[0053] Get the peak value based on the frequency domain features and time domain features:

[0054] ;

[0055] Among them, A(t) is the amplitude of the collected stress wave signal, t1 and t2 are the analysis periods;

[0056] Extract energy integral based on time domain features:

[0057] ;

[0058] Among them, A(t) is the amplitude of the collected stress wave signal, and t1 and t2 are the analysis periods.

[0059] In addition, empirical mode decomposition can also be used to extract intrinsic mode functions (IMFs) and residual components r(t). Empirical mode decomposition is used for adaptive decomposition of nonlinear and non-stationary data. This method does not require any basis functions to be set in advance. It decomposes the signal according to the different time scale characteristics of the data itself to obtain a series of intrinsic mode functions (IMFs) and residual components r(t). Different components represent the fluctuation characteristics of the original data at different time scales, as shown in the following formula:

[0060] ;

[0061] Among them, x(t) is the original signal, is the i-th intrinsic mode function component obtained by empirical mode decomposition, and r(t) is the residual component. As the feature parameter of the classifier.

[0062] S302: Input the results of data feature extraction into a preset number of neural network models for analysis.

[0063] It should be noted that BP network (back propagation neural network), XGBoost (extreme gradient boosting tree) and LSTM (long short-term memory network) can be used for analysis. Specifically, the defined condition monitoring diagnosis verification method is used to evaluate the diagnostic accuracy of each model. For example, the performance (including accuracy) of each model can be evaluated through cross-validation, accuracy, recall rate, F1 score and other methods.

[0064] Taking LSTM as an example, the following is a brief description of the neural network:

[0065] LSTM uses two gates to control the cell state The first is the forget gate, which determines the state of the unit at the previous moment. How many are retained to the present moment? ; The other is the input gate, which determines the input of the network at the current moment How much is saved to the cell state LSTM uses output gate to control the cell state How many outputs are there to the current output value of the LSTM Therefore, the special structure of the LSTM network enables it to solve time series problems. The special gate structure of LSTM can store input information of longer time series and prevent the gradient disappearance problem.

[0066] Specifically, the use of LSTM networks can solve the long-term dependency problem of RNNs, making it possible to learn longer time series information. LSTM networks have extremely strong data fitting capabilities, and are more robust and versatile. In addition, the network parameters of the LSTM network's recurrent layer are shared, and their parameters will not increase with the length of the time series.

[0067] S303, adjusting the parameters of the models according to the comparison of the analysis accuracy of each model, and selecting the neural network model with the highest accuracy as the fault diagnosis model.

[0068] It should be noted that according to the evaluation results, the model parameters are adjusted to optimize the model performance. The model parameter adjustment methods may include grid search, random search, Bayesian optimization, etc. The diagnostic accuracy of each model is compared, and the model with the highest accuracy is selected as the final fault diagnosis model. In this way, the parameters are adjusted continuously during the use of the model in order to better meet the changes in the equipment operating environment and conditions, ensure the diagnostic accuracy of the model, and through such an online learning process, the latest status of the equipment is reflected in real time, the robustness and stability of the model are improved, and subsequent optimization and maintenance are facilitated.

[0069] In specific implementation, according to the analysis accuracy of each model, the model parameters are adjusted, and the neural network model with the highest accuracy is selected as the fault diagnosis model, including: determining the model characteristics according to the distribution characteristics of the received electrical signal; determining the model set to be compared according to the model characteristics; using the historical fault signals and recognition results of the electric fuel pump as samples to train each model to be compared in the model set to be compared; using the manual judgment results and real-time fault signals as screening samples to calculate the recognition accuracy of each model to be compared after training; and selecting the model to be compared with the highest recognition accuracy as the fault diagnosis model.

[0070] The model features are determined according to the distribution characteristics of the received electrical signals, so that the models selected and trained later can closely fit the signal characteristics generated by the actual operation of the electric fuel pump. When different electric fuel pumps are working, the distribution of their stress wave signals may vary due to factors such as structure, working conditions, and fault type. By determining the model features in a targeted manner, the model can better capture the signal information related to the fault, avoid the inadaptability that may be caused by using a general model, and thus improve the accuracy of fault diagnosis. The set of models to be compared is determined based on the determined model features to ensure that the selected model has the ability to process specific signal features in theory. Then, the historical fault signals and recognition results of the electric fuel pump are used as samples to train these models, so that the models can learn the signal patterns under real fault conditions. Since the historical fault signals contain various actual fault conditions, the model can gradually grasp the relationship between different faults and signal characteristics during the learning process, so that the fault type and degree can be more accurately judged when facing new signals.

[0071] Furthermore, the recognition accuracy of the trained model is calculated using manual judgment results and real-time fault signals as screening samples, realizing real-time feedback adjustment of the model. During the actual operation of the electric fuel pump, factors such as operating conditions and environment may change, resulting in new changes in the fault mode. By collecting fault signals in real time and combining them with manual judgment results to evaluate the model accuracy, it is possible to timely discover the performance of the model under the current actual operating conditions, and then optimize the model in a targeted manner, so that it always maintains accurate diagnosis capabilities for the latest status of the equipment. Continuously adjusting the model parameters according to actual conditions and selecting the model with the highest accuracy as the fault diagnosis model will help improve the model's adaptability to different operating conditions and fault conditions.

[0072] It should also be noted that the neural network model with the highest accuracy is selected as the fault diagnosis model, and then, it includes:

[0073] (1) After feature extraction, the newly acquired stress wave signal is input into the fault diagnosis model for diagnosis.

[0074] It should be noted that the input of the fault diagnosis model is the characteristic value, and the output is the diagnosis result.

[0075] (2) Based on the diagnosis results, the fault cause, fault location and fault severity of the electric fuel pump are obtained.

[0076] Specifically, for example, the cause of the fault may include mechanical wear, electrical fault, structural crack, etc. The fault location is the specific location of the fault in the electric fuel pump, and the degree of the fault is used to determine whether it is necessary to immediately stop operation for maintenance or whether it can be delayed to a planned maintenance cycle.

[0077] (3) Select the timing and method of repair and maintenance and the spare parts management plan based on the cause, location and severity of the fault.

[0078] Specifically, the timing and method of maintenance includes selecting the appropriate maintenance time based on the diagnosis results. For example, a serious fault requires immediate downtime for maintenance, while a minor fault can be scheduled for the next maintenance cycle. It can also include determining the maintenance method, including replacing damaged parts, repairing cracks, correcting electrical problems, etc.

[0079] Among them, the spare parts management plan is to determine the parts that need to be replaced based on the diagnosed cause and location of the fault, purchase and reserve key spare parts in advance, ensure that key spare parts can be supplied in time when needed, and reduce maintenance downtime.

[0080] The stress wave-based electric fuel pump structural health monitoring method provided in this embodiment extracts data features from stress wave signals represented by electrical signals, and inputs the extracted data features into a preset number of neural network models for analysis, and then adjusts the model parameters based on the analysis accuracy of each model, and selects the neural network model with the highest accuracy as the fault diagnosis model. In this way, the reliability and service life of the electric fuel pump can be effectively improved, unexpected downtime can be reduced, and maintenance costs can be optimized by using the neural network model with the highest accuracy for fault diagnosis and making maintenance management decisions based on the diagnosis results.

[0081] Corresponding to the aforementioned embodiment of a method for monitoring the structural health of an electric fuel pump based on stress waves, the present application also provides an embodiment of a device for monitoring the structural health of an electric fuel pump based on stress waves.

[0082] Figure 4 This is a schematic diagram of the structure of the first embodiment of the electric fuel pump structural health monitoring device based on stress waves provided in this application. Figure 4 , the device provided in this embodiment includes a trigger module 410, a collection module 420 and a receiving module 430;

[0083] Wherein, the trigger module 410 is used to trigger a pulse at the starting end of the electric fuel pump and convert the pulse into a stress wave signal incident on the medium;

[0084] The acquisition module 420 is used to acquire stress wave signals at the receiving end of the electric fuel pump and convert the acquired stress wave signals into electrical signals; wherein the stress wave signals propagate along the direction of the external load of the electric fuel pump structure and propagate from the starting end to the receiving end;

[0085] The receiving module 430 is used to receive the electrical signal and the pulse electrical signal at the starting end, and perform health monitoring of the electric fuel pump structure according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end.

[0086] The device of this embodiment can be used to perform Figure 2 The steps, specific implementation principles and implementation processes of the method embodiment shown are similar and will not be repeated here.

[0087] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0088] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0089] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for monitoring the structural health of an electric fuel pump based on stress waves, characterized in that: The method comprises: Triggering a pulse electrical signal at the starting end of the electric fuel pump, and converting the pulse electrical signal into a stress wave signal incident on a medium; Collecting stress wave signals at the receiving end of the electric fuel pump and converting the collected stress wave signals into electrical signals; wherein the stress wave signals propagate along the direction of the external load of the electric fuel pump structure and propagate from the starting end to the receiving end; Receiving the electrical signal and the pulse electrical signal from the starting end; Determine the signal frequency domain intervals in each health state according to the frequency domain characteristics of the historical fault signal of the electric fuel pump, wherein each health state includes a healthy state, a fault state and a noise interference state; Determine the constraint conditions under each health state according to the signal frequency domain interval; Determining a corresponding target signal frequency domain interval according to the constraint condition and the received frequency domain characteristics of the electrical signal; Preprocessing the received electrical signal according to the preprocessing method corresponding to the frequency domain interval of the target signal; Obtain real-time monitoring performance data and stress wave data of the electric fuel pump in healthy and faulty states, and determine constraint criteria based on the frequency domain characteristics of the acquired data; Continuously collect real-time monitoring performance data and stress wave data, and compare the characteristic values ​​of the collected real-time monitoring performance data and the calculated and collected stress wave data with the constraint standard; The health monitoring of the electric fuel pump structure is performed according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end.

2. The method according to claim 1, characterized in that The health monitoring of the electric fuel pump structure according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end includes: determining a number of different time scales of the received electrical signal; Determining a characteristic scale of the received electrical signal according to the number of different time scales, the number of the characteristic scales corresponding to the number of different time scales; The time domain features and frequency domain features of the electrical signal received at each characteristic scale are extracted respectively.

3. The method according to claim 1, characterized in that After the health monitoring of the electric fuel pump structure is performed according to the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end, the method further includes: Performing data feature extraction on the stress wave signal represented by the electrical signal; the data feature extraction at least includes time domain feature extraction and frequency domain feature extraction; The results of data feature extraction are input into a preset number of neural network models for analysis; By comparing the analysis accuracy of each neural network model, the parameters of the neural network model are adjusted, and the neural network model with the highest accuracy is selected as the fault diagnosis model.

4. The method according to claim 3, characterized in that Extracting data features of the stress wave signal represented by the electrical signal includes: Preprocessing the stress wave signal represented by the electrical signal and extracting the stress wave factor; Obtain peak values ​​based on frequency domain features and time domain features; Extract energy integral based on time domain features; The intrinsic mode functions and residual components are extracted based on empirical mode decomposition.

5. The method according to claim 3, characterized in that: The method of comparing the analysis accuracy of each neural network model, adjusting the parameters of the neural network model, and selecting the neural network model with the highest accuracy as the fault diagnosis model includes: Determining model features according to distribution characteristics of the received electrical signal; Determine a model set to be compared according to the model characteristics; Using the historical fault signals and identification results of the electric fuel pump as samples, training each model to be compared in the set of models to be compared; Using manual judgment results and real-time fault signals as screening samples, calculate the recognition accuracy of each model to be compared after training; The model to be compared with the highest recognition accuracy is selected as the fault diagnosis model.

6. The method according to claim 3, characterized in that After selecting the neural network model with the highest accuracy as the fault diagnosis model, the method further includes: Extracting features of the newly acquired stress wave signal and inputting it into the fault diagnosis model for diagnosis; Obtain the fault cause, fault location and fault degree of the electric fuel pump based on the diagnosis result; Select the timing and method of repair and maintenance as well as the spare parts management plan based on the cause, location and severity of the fault.

7. The method according to claim 1, characterized in that The real-time monitoring performance data at least includes the flow rate and boost value of the product; The characteristic values ​​of the collected real-time monitoring performance data and the calculated collected stress wave data are compared with the health state of the constraint standard. If they are within the health state of the constraint standard, it is a healthy state; otherwise, it is a fault state.

8. The method according to claim 1, characterized in that Different signal frequency domain intervals correspond to different preprocessing methods.

9. The method according to claim 1, characterized in that: The method is applied to a stress wave-based electric fuel pump structural health monitoring device, which includes a waveform generator, a signal amplifier, a signal conversion device, a signal receiver and a signal processing device; wherein, The waveform generator is used to transmit a pulse electrical signal; The signal amplifier is used to amplify the pulse electrical signal; The signal conversion device is used to convert the pulse electrical signal after signal amplification into a stress wave signal, and convert the received stress wave signal into an electrical signal after receiving the transmitted stress wave signal; The signal receiver is used to simultaneously receive the pulse electrical signal generated by the waveform generator and the electrical signal converted from the stress wave signal; The signal processing device is used to analyze, process and store the pulse electrical signal and the electrical signal.

10. A stress wave-based electric fuel pump structural health monitoring device, characterized in that: The device comprises a trigger module, a collection module and a receiving module; Wherein, the trigger module is used to trigger a pulse electrical signal at the starting end of the electric fuel pump and convert the pulse electrical signal into a stress wave signal incident on the medium; The acquisition module is used to acquire stress wave signals at the receiving end of the electric fuel pump and convert the acquired stress wave signals into electrical signals; wherein the stress wave signals propagate along the direction of the external load of the electric fuel pump structure and propagate from the starting end to the receiving end; The receiving module is used to receive the electrical signal and the pulse electrical signal of the starting end; Determine the signal frequency domain intervals in each health state according to the frequency domain characteristics of the historical fault signal of the electric fuel pump, wherein each health state includes a healthy state, a fault state and a noise interference state; Determine the constraint conditions under each health state according to the signal frequency domain interval; Determining a corresponding target signal frequency domain interval according to the constraint condition and the received frequency domain characteristics of the electrical signal; Preprocessing the received electrical signal according to the preprocessing method corresponding to the frequency domain interval of the target signal; Obtain real-time monitoring performance data and stress wave data of the electric fuel pump in healthy and faulty states, and determine constraint criteria based on the frequency domain characteristics of the acquired data; Continuously collect real-time monitoring performance data and stress wave data, and compare the characteristic values ​​of the collected real-time monitoring performance data and the calculated and collected stress wave data with the constraint standard; The health monitoring of the electric fuel pump structure is performed based on the time domain and frequency domain characteristics of the received electrical signal and the pulse electrical signal at the starting end.

Images