Hydraulic loading system detection method and system based on multi-modal signals

Abstract

The invention discloses a hydraulic loading system detection method and system based on a multi-modal signal, and the method comprises the steps: firstly, collecting the multi-modal signal of a hydraulic loading system in real time, and carrying out the preprocessing of the multi-modal signal; then, multi-modal features corresponding to the hydraulic loading system are extracted from the multi-modal signals; and finally, extracting corresponding modal characteristic historical characteristic data from the multi-modal characteristic historical data, calculating to obtain a coupling factor corresponding to cross-modal nonlinear characteristic coupling of the modal characteristics of the multi-modal signal, combining with the extracted multi-modal characteristics, fusing and evaluating the comprehensive health index of the hydraulic loading system, and obtaining the comprehensive health index of the hydraulic loading system. The multi-modal signal efficient fusion analysis of the hydraulic loading system is realized, the accuracy and reliability of fault detection are improved, and the problems of missing detection and false alarm of a traditional single signal detection method are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of feature extraction analysis, and in particular to a hydraulic loading system detection method and system based on multi-modal signals. BACKGROUND

[0002] The hydraulic loading system, as an important experimental test equipment, has a wide application in the fields of material performance testing and structure loading test. The hydraulic loading system works in a wide flow regulation, high pressure and high temperature environment for a long time, resulting in frequent system failures, and it is difficult and time-consuming to troubleshoot. In order to monitor the running state of the hydraulic loading system, various sensors are used to collect various parameters in the running process of the hydraulic loading system, and the running state of the system is monitored in real time, so that the equipment works in a normal state or an optimal state. The traditional monitoring method usually only focuses on a single signal parameter, such as vibration signal or pressure signal, and this monitoring method is difficult to fully reflect the running state of the system. At present, the multi-modal signal fusion analysis method of the hydraulic loading system has technical bottlenecks in processing, and cannot realize efficient fusion analysis of multi-modal signals. SUMMARY

[0003] In view of the deficiencies of the prior art, the present application provides a hydraulic loading system detection method and system based on multi-modal signals, which can effectively fuse multi-modal signals to analyze the running state of the hydraulic loading system and improve the accuracy and reliability of fault detection. The specific technical solutions are as follows: In a first aspect, a hydraulic loading system detection method based on multi-modal signals is provided. In a first implementation manner of the first aspect, the method comprises: Obtaining multi-modal signals corresponding to the hydraulic loading system, and pre-processing the multi-modal signals; Extracting features of the pre-processed multi-modal signals to obtain multi-modal features corresponding to the hydraulic loading system; Determining a coupling factor corresponding to cross-modal nonlinear feature coupling according to historical data of the corresponding modal features, and combining the multi-modal features to evaluate a comprehensive health index of the hydraulic loading system.

[0004] In combination with the first implementation manner of the first aspect, in a second implementation manner of the first aspect, the multi-modal signals include sound signals, vibration signals, flow signals and / or pressure signals.

[0005] In combination with the first implementation manner of the first aspect, in a third implementation manner of the first aspect, the pre-processing of the multi-modal signals comprises timestamp synchronization processing of the multi-modal signals.

[0006] In conjunction with the first feasible method of the first aspect, in the fourth feasible method of the first aspect, the multimodal signal is preprocessed, including: performing noise reduction and filtering processing on various modal signals in the multimodal signal in parallel.

[0007] In conjunction with the first feasible method of the first aspect, in the fifth feasible method of the first aspect, the extracted features include time-domain features, frequency-domain features, and time-frequency-domain features of various modal signals; Time-domain characteristics include: root mean square, standard deviation, variance, mean, kurtosis, median, skewness, maximum and / or minimum values; Frequency domain features include: amplitude frequency information, phase frequency information, and / or power spectral density.

[0008] In conjunction with the first feasible method of the first aspect, in the sixth feasible method of the first aspect, the coupling factor corresponding to the cross-modal nonlinear characteristic coupling is determined based on the relevant modal characteristic historical data, including: Based on the historical data of the modal features, calibrate the empirical coefficients corresponding to the cross-modal nonlinear feature coupling; By combining the empirical coefficients and the corresponding historical modal feature data, the coupling factor corresponding to the cross-modal nonlinear feature coupling is calculated.

[0009] In conjunction with the first feasible method of the first aspect, in the seventh feasible method of the first aspect, the comprehensive health index of the hydraulic loading system is evaluated, including: Assign weight coefficients to each type of modality feature in the multimodal features; The comprehensive health index is calculated by combining the various modal features and their corresponding weight coefficients, as well as the coupling factor corresponding to the cross-modal nonlinear feature coupling. The specific calculation formula is as follows: ; in, For the first Normalized values ​​of modal features, For feature vectors, For the first The weight coefficients corresponding to the modal features of the item. This represents the total number of modal feature types. This is the scaling factor. It is the coupling factor for the nonlinear interaction between multimodal features.

[0010] In conjunction with the seventh feasible method of the first aspect, in the eighth feasible method of the first aspect, the entropy weight method is used to allocate the weight coefficients corresponding to the modal features.

[0011] In conjunction with the first implementable method of the first aspect, the ninth implementable method of the first aspect also includes: The comprehensive health index is compared with a preset threshold, and an alarm is triggered in response to the comprehensive health index being lower than the preset threshold.

[0012] In a second aspect, a hydraulic loading system detection system based on multi-modal signals is provided, comprising: A data acquisition module configured to acquire multi-modal signals corresponding to the hydraulic loading system and pre-process the multi-modal signals; A feature extraction module configured to extract features of the pre-processed multi-modal signals and acquire multi-modal features corresponding to the hydraulic loading system; A health assessment module configured to determine coupling factors corresponding to cross-modal nonlinear coupling based on historical data of the corresponding modal features, and combine the multi-modal features to assess a comprehensive health index of the hydraulic loading system.

[0013] Beneficial effects: The hydraulic loading system detection method and system based on multi-modal signals can acquire various modal signals corresponding to the hydraulic loading system, extract corresponding modal features from each type of modal signal, fuse the modal features corresponding to various modal signals, and determine coupling factors corresponding to cross-modal nonlinear coupling of different modal signals based on historical data, to assess the comprehensive health index of the hydraulic loading system, thereby realizing efficient fusion analysis of multi-modal signals of the hydraulic loading system, improving the accuracy and reliability of fault detection, and reducing the missed detection and false alarm problems of traditional single signal monitoring methods. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the specific embodiments of the present application, the drawings required to be used in the specific embodiments will be briefly introduced below. In all the drawings, the elements or parts are not necessarily drawn according to the actual proportions.

[0015] Figure 1 A flowchart of the hydraulic loading system detection method based on multi-modal signals provided by an embodiment of the present application; Figure 2 A system block diagram of the hydraulic loading system detection system based on multi-modal signals provided by an embodiment of the present application. DETAILED DESCRIPTION

[0016] The embodiments of the technical solutions of the present application will be described in detail below with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present application, and therefore only serve as examples, but cannot limit the protection scope of the present application.

[0017] As shown in the flowchart of the hydraulic loading system detection method based on multi-modal signals, the detection method comprises: Figure 1 ​Step 1, acquiring the multi-modal signals corresponding to the hydraulic loading system, and preprocessing the multi-modal signals; Step 2, extracting the features of the preprocessed multi-modal signals, and acquiring the multi-modal features corresponding to the hydraulic loading system; Step 3, determining the coupling factors corresponding to the cross-modal nonlinear feature coupling according to the historical data of the corresponding modal features, and combining the multi-modal features to evaluate the comprehensive health index of the hydraulic loading system.

[0018] Specifically, first, various modal signals of the hydraulic loading system, such as vibration signals, flow signals, etc., can be collected in real time through different signal collection devices, and the collected multi-modal signals can be preprocessed for subsequent analysis. At the same time, the past multi-modal feature historical data of the hydraulic loading system can be retrieved from the database, which includes the multi-modal feature data corresponding to the hydraulic loading system at different detection times in the past. Then, the modal feature combination corresponding to the hydraulic loading system can be extracted from the acquired multi-modal signals, such as the root mean square of the vibration signal, the label difference, etc., extracted from the vibration signal, and the power spectral density of the pressure signal, etc., extracted from the pressure signal. Finally, the historical feature data of the corresponding modal features can be extracted from the multi-modal feature historical data, so as to calculate the coupling factors corresponding to the cross-modal nonlinear feature coupling of the modal features of the multi-modal signals. In combination with the multi-modal features corresponding to the current multi-modal signals, the comprehensive health index of the hydraulic loading system is evaluated, realizing efficient fusion analysis of the multi-modal signals of the hydraulic loading system, improving the accuracy and reliability of fault detection, and reducing the problems of missed detection and false reporting of traditional single signal detection methods.

[0019] In this embodiment, optionally, the multi-modal signals include sound signals, vibration signals, flow signals, and / or pressure signals.

[0020] Specifically, the sound signals, vibration signals, flow signals, and pressure signals have complementarity in physical dimensions. Different modal signals can reflect the system state from independent physical layers, and the vibration signals are sensitive to mechanical structure abnormalities; the pressure / flow signals directly represent the hydraulic circuit state; the sound signals capture fluid dynamics characteristics such as cavitation and cavitation. Different modal signals have information redundancy, can cross-verify and suppress false positives, and have unique fault characterization capabilities, which can break through the perception limitations of single signal monitoring. Therefore, in this embodiment, the acquired multi-modal signals can include sound signals, vibration signals, flow signals, and pressure signals.

[0021] It should be understood that this embodiment is only exemplified by sound signals, vibration signals, flow signals, and pressure signals, but the present application is not limited thereto, and can also include other signals having complementarity with the four modal signals.

[0022] In the embodiment, the multi-modal signals are optionally preprocessed, including: time stamp synchronization processing on the multi-modal signals.

[0023] Specifically, in order to avoid data distortion and increase fusion analysis error caused by time asynchronization, time stamp synchronization processing can be performed on all acquired modal signals to generate modal signals with synchronization marks. Specifically, the clock deviation of different modal signals is calibrated with 10 seconds as the reference, the time stamp is dynamically adjusted, and it is ensured that the cumulative synchronization error is lower than the set threshold. Then, synchronization marks are added to each type of modal signal.

[0024] In the embodiment, the multi-modal signals are optionally preprocessed, including: parallel noise reduction filtering processing on each type of modal signal in the multi-modal signals.

[0025] Specifically, in order to speed up the signal processing efficiency and improve the fusion analysis efficiency, noise reduction filtering processing can be performed in parallel on each type of modal signal after time stamp processing. During the noise reduction filtering processing, a high-pass filter with a cutoff frequency of 100 Hz can be used to filter out low-frequency mechanical interference in the modal signal, and then a band-stop filter with a center frequency of 1 kHz and a bandwidth of 200 Hz can be used to suppress background noise in the modal signal.

[0026] In the embodiment, the extracted features include time domain features, frequency domain features and time-frequency domain features of each type of modal signal. The time domain features include: root mean square, standard deviation, variance, average value, signal kurtosis, median, skewness, maximum value and / or minimum value. The frequency domain features include: amplitude-frequency information, phase-frequency information and / or power spectral density.

[0027] Specifically, in order to further speed up the signal processing efficiency and improve the fusion analysis efficiency, feature extraction can be performed in parallel on each type of preprocessed modal signal. The extracted modal features include but are not limited to: time domain features, frequency domain features and time-frequency domain features.

[0028] The time domain features can include but are not limited to: root mean square, standard deviation, variance, average value, signal kurtosis, median, skewness, maximum value and minimum value.

[0029] The frequency domain features can include but are not limited to: amplitude-frequency information, phase-frequency information and / or power spectral density.

[0030] The time-frequency domain features are obtained by performing short-time Fourier transform on the modal signals.

[0031] The hydraulic system failure often shows multi-signal nonlinear correlation. Therefore, the inherent defects of traditional linear weighted fusion can be solved by cross-modal nonlinear feature coupling analysis. Specifically, the coupling factor can be learned from historical data, the implicit fault transmission process is converted into a calculable parameter, and the environmental noise interference is identified, thereby improving the robustness of state evaluation and early fault detection rate.

[0032] In the embodiment, the coupling factor corresponding to the cross-modal nonlinear feature coupling is determined according to the historical data of the corresponding modal features, including: The empirical coefficient corresponding to the cross-modal nonlinear feature coupling is calibrated according to the historical data of the modal features. The coupling factor corresponding to the cross-modal nonlinear feature coupling is calculated in combination with the empirical coefficient and the historical data of the corresponding modal features.

[0033] Specifically, the obtained multi-modal feature historical data includes the feature values of the multi-modal features corresponding to the hydraulic loading system at different detection times in the past. The feature values of the multi-modal features corresponding to each detection time are taken as samples to construct a sample data set. The historical feature values of the corresponding modal features are extracted from each sample to calibrate the empirical coefficient corresponding to the nonlinear combination of the modal features of various types of modal signals.

[0034] Specifically, the empirical coefficient can be determined by taking the absolute value of the correlation coefficient of the root mean square of the vibration signal and the power spectral density of the pressure signal in the multi-modal feature historical data The empirical coefficient can be determined by taking the ratio of the mean value and the standard deviation of the flow signal in the multi-modal feature historical data The empirical coefficient can be determined by taking the reciprocal of the short-time Fourier transform amplitude of the sound signal under the healthy state in the multi-modal feature historical data The empirical coefficient can be determined by setting the system sampling frequency as a fixed value .

[0035] In combination with the empirical coefficient and the historical feature values of the corresponding modal features in the sample, the coupling value of the multi-modal feature combination of various types of model signals in the sample is calculated. It should be understood that, in order to eliminate the dimensional difference, all the feature values of the modal features are subjected to standardization processing in the calculation process. Finally, in combination with the coupling values corresponding to all the samples in the sample data set, the coupling factor corresponding to the cross-modal nonlinear feature coupling of various types of model signals can be calculated.

[0036] For sound, vibration, flow, and pressure signals, the extracted modal features include: the root mean square and peak value of the vibration signal, the power spectral density of the pressure signal, the short-time Fourier transform amplitude of the sound signal, and the mean and labeled difference of the flow signal. Based on the historical feature values ​​corresponding to these modal features in the sample, the coupling factor characterizing the nonlinear interaction between multimodal features can be calculated, as shown in the following formula: ; in, , , , This is an empirical coefficient. The root mean square of the vibration signal. The power spectral density of the pressure signal. The amplitude of the short-time Fourier transform of the sound signal. The mean of the flow signal. The difference in the label of the flow signal. The peak value of the vibration signal. Coupling factor Represents the eigenvector. This represents the total number of modal feature types.

[0037] In this embodiment, optionally, the comprehensive health index of the hydraulic loading system is evaluated, including: Assign weight coefficients to each type of modality feature in the multimodal features; The comprehensive health index is calculated by combining the various modal features and their corresponding weight coefficients, as well as the coupling factor corresponding to the cross-modal nonlinear feature coupling. The specific calculation formula is as follows: ; in, For the first The normalized value of the modal feature is calculated using the following formula: ; For the extracted first Eigenvalues ​​of modal features , These are the first in the historical data of multimodal features. The maximum and minimum values ​​corresponding to the modal features of the item.

[0038] For the first The weight coefficients corresponding to the modal features of the item. This represents the total number of modal feature types in multimodal features. This is the scaling factor. It is the coupling factor for the nonlinear interaction between multimodal features.

[0039] In this embodiment, optionally, the entropy weight method is used to assign the weight coefficients corresponding to the modal features.

[0040] Specifically, based on the correlation between features and the state of the hydraulic loading system, the entropy weight method can be used to assign weight coefficients to various modal features according to the sample data in the sample dataset. The specific calculation formula is as follows: ; ; in, For the first Information entropy of modal features The total number of samples in the sample dataset. For the first Modal features of the term in the first The probability distribution of each sample is calculated using the following formula: ; For the first Modal features of the term in the first The original data from each sample.

[0041] In this embodiment, optionally, it also includes: The comprehensive health index is compared with a preset threshold, and an alarm is triggered in response to the comprehensive health index falling below the preset threshold.

[0042] like Figure 2 The diagram shown is a system block diagram of a hydraulic loading system detection system based on multimodal signals. The detection system includes: The data acquisition module is configured to acquire the multimodal signals corresponding to the hydraulic loading system and preprocess the multimodal signals. The feature extraction module is configured to extract features from the preprocessed multimodal signals to obtain the multimodal features corresponding to the hydraulic loading system. The health assessment module is configured to determine the coupling factor corresponding to the cross-modal nonlinear characteristic coupling based on the corresponding modal characteristic historical data, and to evaluate the comprehensive health index of the hydraulic loading system in combination with the multimodal characteristics.

[0043] Specifically, the detection system comprises a data acquisition module, a feature extraction module and a health assessment module. Among them, the data acquisition module can acquire real-time multi-modal signals of the hydraulic loading system, and multi-modal feature historical data. The data acquisition module can preprocess the multi-modal signals to improve the signal quality of the multi-modal signals. The feature extraction module can extract features from the multi-modal signals to obtain the current multi-modal features of the hydraulic loading system. The health assessment module can extract the historical feature data of the corresponding modal features from the multi-modal feature historical data, and calculate the coupling factor corresponding to the cross-modal nonlinear feature coupling of the modal features of the multi-modal signals. Then, combined with the current multi-modal features, the comprehensive health index of the hydraulic loading system is fused to realize efficient fusion analysis of the multi-modal signals of the hydraulic loading system, improve the accuracy and reliability of fault detection, and reduce the missed detection and false alarm problems of the traditional single signal detection method.

[0044] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should be covered in the scope of the claims and the specification of the present application.

Claims

1. A detection method for a hydraulic loading system based on multimodal signals, characterized in that, include: Acquire the multimodal signals corresponding to the hydraulic loading system and preprocess the multimodal signals; Extract the features of the preprocessed multimodal signals to obtain the corresponding multimodal features of the hydraulic loading system; Based on the relevant historical data of modal features, the coupling factor corresponding to the cross-modal nonlinear feature coupling is determined, and the comprehensive health index of the hydraulic loading system is evaluated in combination with the extracted multimodal features.

2. The hydraulic loading system testing method according to claim 1, characterized in that, The multimodal signals include sound signals, vibration signals, flow signals, and / or pressure signals.

3. The method for detecting a hydraulic loading system according to claim 1, characterized in that, Preprocessing the multimodal signal includes: performing timestamp synchronization processing on the multimodal signal.

4. The method for detecting a hydraulic loading system according to claim 1, characterized in that, Preprocessing of multimodal signals includes parallel noise reduction and filtering of various modal signals in the multimodal signal.

5. The method for detecting a hydraulic loading system according to claim 1, characterized in that, The extracted features include time-domain features, frequency-domain features, and time-frequency-domain features of various modal signals; Time-domain characteristics include: root mean square, standard deviation, variance, mean, kurtosis, median, skewness, maximum and / or minimum values; Frequency domain features include: amplitude frequency information, phase frequency information, and / or power spectral density.

6. The method for detecting a hydraulic loading system according to claim 1, characterized in that, The coupling factor corresponding to the cross-modal nonlinear feature coupling is determined based on the relevant historical modal feature data, including: Based on the historical data of the modal features, calibrate the empirical coefficients corresponding to the cross-modal nonlinear feature coupling; By combining the empirical coefficients and the corresponding historical modal feature data, the coupling factor corresponding to the cross-modal nonlinear feature coupling is calculated.

7. The method for detecting a hydraulic loading system according to claim 1, characterized in that, Assess the overall health index of the hydraulic loading system, including: Assign weight coefficients to each type of modality feature in the multimodal features; The comprehensive health index is calculated by combining the various modal features and their corresponding weight coefficients, as well as the coupling factor corresponding to the cross-modal nonlinear feature coupling. The specific calculation formula is as follows: ； in, For the first Normalized values ​​of modal features, For feature vectors, For the first The weight coefficients corresponding to the modal features of the item. This represents the total number of modal feature types. This is the scaling factor. It is the coupling factor for the nonlinear interaction between multimodal features.

8. The method for detecting a hydraulic loading system according to claim 7, characterized in that, The entropy weighting method is used to assign weight coefficients to the modal features.

9. The method for detecting a hydraulic loading system according to claim 1, characterized in that, Also includes: The comprehensive health index is compared with a preset threshold, and an alarm is triggered in response to the comprehensive health index falling below the preset threshold.

10. A detection system for a hydraulic loading system based on multimodal signals, characterized in that, include: The data acquisition module is configured to acquire the multimodal signals corresponding to the hydraulic loading system and preprocess the multimodal signals. The feature extraction module is configured to extract features from the preprocessed multimodal signals to obtain the multimodal features corresponding to the hydraulic loading system. The health assessment module is configured to determine the coupling factor corresponding to the cross-modal nonlinear characteristic coupling based on the corresponding modal characteristic historical data, and to evaluate the comprehensive health index of the hydraulic loading system in combination with the multimodal characteristics.

Images