An engine fault diagnosis method, device, equipment and storage medium

Abstract

The application discloses an engine fault diagnosis method and device, equipment and storage medium, and relates to the technical field of fault diagnosis. The method comprises the following steps: synchronously collecting sound signals and vibration signals of an engine under a normal operation state, and performing feature extraction on the sound signals and the vibration signals to obtain corresponding target features; the target features comprise time domain features, frequency domain features and time-frequency domain features; statistical indexes of each target feature are determined through statistical characteristic analysis, all the target features are subjected to feature fusion based on the statistical indexes according to a target fusion mode to obtain fusion indexes; and a warning threshold is determined according to data statistical characteristics of the fusion indexes, so that the engine is subjected to fault monitoring by using the warning threshold. The warning threshold determined according to the sound and vibration data of the engine under the normal operation state is used for fault diagnosis, the high dependence on fault samples is avoided, and the comprehensiveness of fault analysis is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault diagnosis, and in particular relates to an engine fault diagnosis method, device, equipment and storage medium. BACKGROUND

[0002] Various unknown or not clearly defined fault types that may occur during the operation of a device system are referred to as uncertain fault modes. In an aero-engine, for example, although common typical faults include blade damage, bearing wear, aerodynamic instability, etc., due to the complexity of the operating conditions, limited data acquisition and the small probability of fault occurrence, many fault modes cannot be accurately modeled in advance.

[0003] In related technologies, mode-driven diagnosis is often used, such as the identification of bearing faults and blade cracks, which relies on prior knowledge. However, in a laboratory environment, certain high-risk faults cannot be artificially created, and it is more difficult to obtain data for fault states due to the small probability of fault occurrence. Therefore, how to achieve abnormal state recognition and analysis of uncertain fault modes in the absence of prior knowledge is a problem that needs to be solved at present. SUMMARY

[0004] Therefore, the purpose of the present application is to provide an engine fault diagnosis method, device, equipment and storage medium, which can avoid high dependence on fault samples and improve the comprehensiveness of fault analysis. The specific scheme is as follows:

[0005] In a first aspect, the present application discloses an engine fault diagnosis method, comprising:

[0006] Synchronously collecting sound signals and vibration signals of an engine in a normal operating state, and performing feature extraction on the sound signals and the vibration signals to obtain corresponding target features; the target features include time domain features, frequency domain features and time-frequency domain features;

[0007] Determining statistical indicators of each target feature through statistical characteristic analysis, and performing feature fusion on all target features based on the statistical indicators according to a target fusion mode to obtain a fusion indicator;

[0008] Determining a warning threshold according to the data statistical characteristics of the fusion indicator, so as to use the warning threshold to monitor the engine for faults.

[0009] Optionally, the determination of the statistical indicators of each target feature through statistical characteristic analysis and the feature fusion of all target features based on the statistical indicators according to the target fusion mode to obtain a fusion indicator comprise:

[0010] Establishing a corresponding probability distribution model for each target feature;

[0011] calculating an index for describing a distribution center, an index for describing a distribution dispersion, and a confidence interval according to the probability distribution model, to obtain a statistical index of the target feature;

[0012] According to the weight corresponding to each of the target features and the confidence interval corresponding to each of the target features, the fusion index is obtained by weighted summation.

[0013] Optionally, the statistical index of each of the target features is determined by statistical characteristic analysis, and the fusion index is obtained by feature fusion of all the target features according to a target fusion mode based on the statistical index, including:

[0014] The target feature is normalized to obtain a normalized feature corresponding to the target feature;

[0015] The probability distribution of the target feature is calculated by discretizing the normalized feature;

[0016] The entropy value of the target feature is calculated according to the probability distribution, and the information amount of the target feature is determined according to the entropy value;

[0017] The weight corresponding to each of the target features is determined according to the information amount;

[0018] The fusion index is obtained by weighted fusion based on the weight and the normalized feature corresponding to each of the target features.

[0019] Optionally, before the feature extraction of the sound signal and the vibration signal to obtain the corresponding target feature, the method further includes:

[0020] The sound signal and the vibration signal are segmented to obtain a plurality of sampling segments;

[0021] The pre-warning threshold is determined according to the data statistical characteristics of the fusion index, including:

[0022] The fusion index sequence is calculated according to the fusion index corresponding to each sampling segment;

[0023] The pre-warning threshold is determined based on the mean and the standard deviation of the fusion index sequence.

[0024] Optionally, the engine fault diagnosis method further includes:

[0025] The pre-warning threshold is dynamically adjusted according to the dynamic characteristics of the fusion index.

[0026] Optionally, the feature extraction of the sound signal and the vibration signal to obtain the corresponding target feature includes:

[0027] performing time domain analysis, frequency domain analysis and time-frequency domain analysis on the sound signal and the vibration signal respectively;

[0028] The time domain features, frequency domain features and time-frequency domain features of the sound signal and the vibration signal obtained through analysis are taken as the target features.

[0029] Optionally, the fault monitoring of the engine by using the early warning threshold comprises:

[0030] synchronously collecting real-time sound signals and real-time vibration signals of the engine, and performing feature extraction on the real-time sound signals and the real-time vibration signals to obtain corresponding real-time target features;

[0031] performing feature fusion on all the real-time target features according to the target fusion mode to obtain a real-time fusion index;

[0032] comparing the real-time fusion index with the early warning threshold to determine whether the engine has a fault.

[0033] In a second aspect, the present application discloses an engine fault diagnosis device, comprising:

[0034] a feature extraction module configured to synchronously collect sound signals and vibration signals of the engine in a normal operating state, and perform feature extraction on the sound signals and the vibration signals to obtain corresponding target features; the target features include time domain features, frequency domain features and time-frequency domain features;

[0035] a feature fusion module configured to determine statistical indexes of each target feature through statistical characteristic analysis, and perform feature fusion on all the target features according to a target fusion mode based on the statistical indexes to obtain a fusion index;

[0036] a threshold determination module configured to determine an early warning threshold according to data statistical characteristics of the fusion index, so as to monitor the fault of the engine by using the early warning threshold.

[0037] In a third aspect, the present application discloses an electronic device, comprising:

[0038] a memory configured to save a computer program;

[0039] a processor configured to execute the computer program to implement the engine fault diagnosis method described above.

[0040] In a fourth aspect, the present application discloses a computer readable storage medium configured to store a computer program; wherein the computer program is executed by a processor to implement the engine fault diagnosis method described above.

[0041] In the application, sound signals and vibration signals of an engine in a normal running state are synchronously collected, and corresponding target features are obtained by performing feature extraction on the sound signals and the vibration signals. The target features include time domain features, frequency domain features and time-frequency domain features. Statistical indexes of each target feature are determined through statistical characteristic analysis, all target features are fused according to a target fusion mode based on the statistical indexes to obtain a fusion index, and a warning threshold is determined according to data statistical characteristics of the fusion index, so as to perform fault monitoring on the engine by using the warning threshold. As can be seen, according to the sound and vibration signals of the engine in the normal running state, the statistical indexes of the features are obtained through feature analysis and statistical characteristic analysis, the warning threshold is determined according to the fusion index after feature fusion, the normal running state of the engine is determined by constructing a baseline model, whether the engine has a fault is determined, and based on the fusion index, the health status of the whole system can be comprehensively analyzed, the abnormal state recognition and analysis of the uncertain fault mode are realized in the absence of prior knowledge, and the comprehensiveness of fault analysis is improved. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only a part of the embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor based on the provided drawings.

[0043] Figure 1 A flow chart of an engine fault diagnosis method provided by the present application is shown in the figure.

[0044] Figure 2 A specific engine fault diagnosis system framework provided by the present application is shown in the figure.

[0045] Figure 3 A structural schematic diagram of an engine fault diagnosis device provided by the present application is shown in the figure.

[0046] Figure 4 An electronic device structure provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0048] In the related art, mode-driven diagnosis is often used, such as identification of bearing failure and blade crack, which relies on prior knowledge. However, in a laboratory environment, some high-risk failures cannot be artificially manufactured, and some failures occur as a small probability event, so it is more difficult to obtain fault state data. Therefore, when facing unknown failure modes and lack of fault samples in complex working conditions, not only the operability in the research and development stage is limited, but also the engineering progress is delayed, and it is still difficult to meet the engineering requirements of real-time online abnormal monitoring and early warning. To overcome the above technical problems, the application provides an engine fault diagnosis method, which can avoid high dependence on fault samples and improve the comprehensiveness of fault analysis.

[0049] The engine fault diagnosis method disclosed by the embodiments of the application can avoid high dependence on fault samples and improve the comprehensiveness of fault analysis. Figure 1 The method can include the following steps:

[0050] Step S11: Synchronously collecting sound signals and vibration signals of the engine in a normal running state, and performing feature extraction on the sound signals and the vibration signals to obtain corresponding target features; the target features include time domain features, frequency domain features and time-frequency domain features.

[0051] First, the sound signals (xacoustic(n)) and the vibration signals (xvibration(n)) generated during the running process can be collected by the sensors installed at the key positions of the aero-engine, and the sampling frequency can be set to not less than 51200Hz. To ensure data accuracy and subsequent analysis reliability, the sound signals and the vibration signals need to be high-precision time-synchronized. Specifically, the synchronization can be performed by triggering a unified sampling clock through hardware, and the digital synchronization module is labeled with a time stamp to ensure that the sound signals and the vibration signals of each sampling point are aligned in time.

[0052] The feature extraction on the sound signals and the vibration signals to obtain the corresponding target features includes: performing time domain analysis, frequency domain analysis and time-frequency domain analysis on the sound signals and the vibration signals respectively; and taking the time domain features, the frequency domain features and the time-frequency domain features of the sound signals and the vibration signals obtained by the analysis as the target features.

[0053] The multi-dimensional characteristic features are obtained from the aero-engine operating sound and vibration signals by mathematical analysis method, which provides a quantitative basis for subsequent statistical modeling, feature fusion and abnormality determination. For example, in time domain analysis, the signal amplitude and waveform can be mathematically described to obtain features such as root mean square (RMS), peak factor (PF), kurtosis, etc. In frequency domain analysis, the signal can be mapped to frequency space by Fourier transform, which can be used to describe the main frequency, spectral centroid and energy distribution of the signal. In time-frequency domain analysis, the frequency characteristics of the signal changing with time can be obtained by short-time Fourier transform (STFT), wavelet transform or other time-frequency analysis methods. Through time-frequency domain features, the transient changes and multi-resolution energy distribution of the signal can be captured. The target features include but are not limited to: sound signal root mean square , sound signal peak factor , sound signal kurtosis , sound signal main frequency amplitude , sound signal spectral centroid , sound signal wideband energy ratio , vibration signal root mean square , vibration signal peak factor , vibration signal waveform index , vibration signal main frequency amplitude , vibration signal spectral centroid , vibration signal envelope demodulation amplitude Then, the 6-dimensional features of the acoustic and vibration signals are integrated to form a 12-dimensional feature vector F:

[0054] .

[0055] Step S12: determining the statistical index of each target feature by statistical characteristic analysis, and performing feature fusion on all target features according to the target fusion mode based on the statistical index to obtain a fusion index.

[0056] The statistical index can include an index describing the distribution center, an index describing the distribution dispersion, a confidence interval, etc., which can represent the variation range and distribution characteristics of the target feature under normal working conditions. It can also be the entropy value and information content of the target feature, etc. Then, based on the statistical index, all target features are fused according to the target fusion mode to obtain a fusion index. The target fusion mode is not limited in this embodiment, which can use weighted linear fusion (entropy weight method, principal component analysis), nonlinear fusion (geometric mean, neural network fusion, fuzzy fusion) or other mathematical transformation methods (Euclidean norm) to map the multi-dimensional features to a single fusion index, so as to take into account the information of each feature while suppressing redundant and noise interference.

[0057] In some embodiments, the statistical indicators of each target feature are determined by statistical property analysis, and a fusion indicator is obtained by fusing all the target features according to a target fusion manner based on the statistical indicators, including: establishing a corresponding probability distribution model for each target feature; calculating an indicator for describing a distribution center, an indicator for describing a distribution dispersion, and a confidence interval according to the probability distribution model to obtain the statistical indicators of the target features; and obtaining the fusion indicator by weighted summation according to the weight corresponding to each target feature and the confidence interval corresponding to each target feature. In the statistical property analysis link, the multi-dimensional features obtained in the feature extraction link are probabilistically modeled to quantify the statistical properties of the normal operating state, that is, a probability distribution model is established for each target feature Fi to describe the variation range and distribution characteristics thereof in the normal operating state, thereby providing a reliable basis for the diagnosis decision layer.

[0058] For example, the feature is subject to a normal distribution: Therefore, the probability distribution model is:

[0059] ; wherein, are the mean and standard deviation of the target feature, respectively, and exp[] represents an exponential function. Through the probability distribution model, the expected value, variance, and confidence interval of the feature can be calculated to describe the fluctuation range (confidence interval) of the feature in the normal operating state: The range can be used as a preliminary reference for abnormality judgment.

[0060] In addition to the normal distribution, other probability distribution models (such as lognormal, Gamma distribution, etc.) can also be applicable, and non-parametric methods such as kernel density estimation and empirical distribution function can also be used for statistical analysis; for example, the data distribution can be visualized by using a non-parametric method first, and then a suitable parametric model is selected for fitting. By modeling and quantifying the multi-dimensional features using probability and statistics theory, the fluctuation characteristics of the signal are converted into calculable statistical indicators, which provide a mathematical basis for subsequent feature fusion and abnormality judgment.

[0061] In the feature fusion link, the statistical indicators obtained by statistical property analysis are combined to fuse multiple target features to generate a single fusion indicator that can be used for abnormality judgment ; the multi-dimensional features are fused by mathematical methods to form a quantifiable fusion indicator, which provides a quantitative basis for diagnosis and decision making. This embodiment does not limit the specific fusion method, and weighted or other fusion methods can be used for feature combination.

[0062] For example, a weighted fusion method is used to generate a fusion indicator by linear combination:

[0063] ;

[0064] wherein m is the dimension of the target feature, i.e. there are m target features; is a weight, which can be pre-coordinated, or can be determined according to the feature information amount, importance or other evaluation indexes.

[0065] For example, the information entropy weight method is adopted, the Shannon entropy of each target feature is calculated to quantify its uncertainty; then the weight is calculated according to the entropy value: .

[0066] In some embodiments, before the feature extraction of the sound signal and the vibration signal obtains the corresponding target feature, the sound signal and the vibration signal are segmented to obtain a plurality of sampling segments; the determination of the early warning threshold according to the data statistical characteristics of the fusion index includes: calculating the fusion index corresponding to each sampling segment to obtain a fusion index sequence; and determining the early warning threshold based on the data statistical characteristics of the fusion index sequence. That is, the steps S11 and S12 are performed on each sampling segment to obtain the fusion index of each sampling segment, i.e. the fusion index sequence, and then the early warning threshold is determined according to the data statistical characteristics of the fusion index sequence. By converting the transient and irregular original signal into a smooth time sequence that can reflect the evolution trend of the device state, a stable and representative dynamic early warning threshold can be set based on the statistical characteristics (such as mean, variance, trend) of the sequence, effectively filtering transient noise interference and capturing the real performance degradation process.

[0067] In some embodiments, the determination of the statistical index of each target feature through statistical characteristic analysis, the feature fusion of all target features according to a target fusion mode based on the statistical index to obtain a fusion index includes: standardizing the target features to obtain the standardized features corresponding to the target features; discretizing the standardized features to statistically analyze the probability distribution of the target features; calculating the entropy value of the target features according to the probability distribution, determining the information amount of the target features according to the entropy value; determining the weight corresponding to each target feature according to the information amount; and performing weighted fusion based on the weight and the standardized feature corresponding to each target feature to obtain the fusion index. That is, the information entropy weight method is adopted to perform weighted fusion on the extracted acoustic and vibration features, and the information entropy weight method can automatically adjust the weight according to the information amount of each feature to ensure that the fused features can more truly reflect the running state of the system; the accuracy of feature fusion is improved, so that the system is more sensitive and efficient when processing complex signals.

[0068] Step S13: determining an early warning threshold according to the data statistical characteristics of the fusion index, so as to utilize the early warning threshold to monitor the engine fault.

[0069] By quantifying the statistical characteristics of the fusion indicators in the normal operating state and setting thresholds, the system can compare the real-time fusion indicators for deviation in the operation monitoring stage, realize abnormal diagnosis and early warning. In the absence of explicit prior information of fault mode, based on the multi-modal fusion analysis of acoustic signals and vibration signals in the operation process of aero-engine, an abnormal recognition process independent of priori is constructed, which sequentially completes signal collection, feature construction, abnormal recognition and real-time monitoring, realizes online monitoring and abnormal warning of engine operating state, and is suitable for bench test and in-service monitoring scene.

[0070] In the early warning threshold setting link, based on the statistical characteristics of the fusion indicators generated in the feature fusion link , the threshold value for abnormal judgment is determined, thereby providing reliable early warning basis for diagnosis decision. The data statistical characteristics can include mean or median representing the trend in the set, or standard deviation representing the dispersion degree, etc. The specific early warning threshold can be set according to confidence interval, quantile, probability density threshold or 3σ criterion, and the specific calculation method is not limited in the embodiment.

[0071] For example, the fusion feature indicator obeys the probability distribution (such as normal distribution) under normal working condition:

[0072] ; the early warning threshold T= , wherein k is the confidence coefficient.

[0073] In a specific embodiment, the early warning threshold determination includes the following process:

[0074] First, the continuously collected acoustic vibration data is cut into multiple sampling segments according to time window. For example, let the whole working condition set be For the cth working condition, if the total recording time is Tc (unit: second), the set analysis window length is Lc (unit: second), and the overlap rate is , then the step is:

[0075] ;

[0076] The number of effective sampling segments generated by this working condition is:

[0077] ;

[0078] Total sample segment number:

[0079] ;

[0080] Extract the multi-dimensional target feature vector from each sampling segment i:

[0081] ;

[0082] in, Let represent the j-th feature of the i-th sample segment, Ma represent the number of the a-th sound feature, and Mv represent the number of the v-th vibration feature. This indicates transpose; the total dimension of the target features (i.e., the total number of target features) is M = Ma + Mv.

[0083] Then, the features of each target are standardized:

[0084] ;

[0085] in, and Let be the mean and standard deviation of the j-th feature under normal conditions, respectively. The calculation formula is:

[0086] ;

[0087] Discretize each standardized feature, dividing the numerical range into K intervals, and calculate the probability distribution for each interval:

[0088] ;

[0089] Finally, calculate the feature entropy. :

[0090] ;

[0091] And further calculate the information content of each feature dimension:

[0092] ;

[0093] Through the above steps, statistical analysis and probabilistic modeling of all sample segment features are completed, generating indicator data for entropy-weighted fusion: the standardized value, probability distribution, entropy value, and information content of each feature. This data constitutes a normal operation benchmark model, providing a reliable basis for setting fusion indicators and early warning thresholds.

[0094] The standardized features of each sampling segment are weighted and fused using the information entropy weighting method. First, the weights are calculated using the calculated information content of each feature:

[0095] ;

[0096] Then, the M-dimensional feature values ​​are fused into a single feature index. The fusion index for the i-th sampling segment is:

[0097] ;

[0098] Calculate the fusion index sequence under normal conditions Mean and standard deviation:

[0099] ;

[0100] The early warning threshold is set as: , adopts criterion, k=3.

[0101] In some embodiments, the engine fault diagnosis method further comprises: dynamically adjusting the early warning threshold according to the dynamic characteristics of the fusion index. It can be understood that the dynamic adjustment of the early warning threshold is combined with the dynamic characteristics of the fusion index; for example, the engine will have progressive failures such as wear, fouling, performance degradation, etc. with the use time, and the early warning threshold should be dynamically adjusted following the change trend of the fusion index itself, and automatically adapt to the slow changes of the device working condition. Or the early warning threshold under different working modes (cruise mode, reasoning mode) of the engine should also be self-adaptive adjustment, the overall vibration level baseline will be raised when the load is increased, and the threshold will also be raised, avoiding false positives; by dynamically adjusting the threshold, the sensitivity and reliability of abnormal detection are considered. The dynamic early warning threshold setting based on normal state modeling realizes early identification of abnormalities; and the threshold can be adjusted in real time according to the characteristic distribution of the normal state, and the early warning standard is dynamically optimized for different operating conditions, so as to respond to the initial stage of failure faster.

[0102] In this application, in the face of fault mode uncertainty and lack of fault samples, the feature range and early warning threshold are determined using data in normal state, and effective detection of faults is realized through deviation analysis based on normal state modeling, which does not depend on a large number of fault samples, and can still realize efficient abnormal detection and early warning without fault samples.

[0103] In some embodiments, the fault monitoring of the engine using the early warning threshold comprises: synchronously collecting real-time sound signals and real-time vibration signals of the engine, and performing feature extraction on the real-time sound signals and the real-time vibration signals to obtain corresponding real-time target features; performing feature fusion on all the real-time target features according to the target fusion mode to obtain a real-time fusion index; and comparing the real-time fusion index with the early warning threshold to determine whether the engine has failed. That is, the real-time target features are fused according to the target fusion mode used in the early warning threshold determination process, and the previous weights are directly used without re-computation during fusion, such as weighted summation of the real-time target features and the weights corresponding to each target feature to obtain the real-time fusion index.

[0104] The traditional fault diagnosis method often needs a large number of fault samples for training and model construction, and the sound-vibration fusion diagnosis and early warning method based on normal operation state modeling provided in the application meets the needs of online monitoring, rapid response and fault early warning in engineering field, extracts and models sound-vibration multi-dimensional features from a large number of normal operation samples in the training stage through the fusion of engine sound signals and vibration signals, and constructs a high-dimensional fusion normal operation benchmark model. In the operation monitoring stage, the sound-vibration signals are collected in real time and the corresponding features are extracted, and the deviation is compared with the benchmark model, when the monitoring features deviate from the normal range and exceed the threshold, the system can realize abnormal diagnosis and early warning. Avoiding the high dependence on fault samples, even in the absence of fault samples, accurate abnormal detection can still be achieved through deviation analysis of normal state, greatly reducing the demand for fault samples, improving the universality and applicability of the system. Effectively solve the diagnosis problem under the condition of uncertain fault mode and scarce fault samples, with fast response, accurate identification and good engineering implementability.

[0105] As can be seen from the above, in the embodiment, the sound signal and the vibration signal of the engine in the normal operation state are synchronously collected, the corresponding target features are obtained by performing feature extraction on the sound signal and the vibration signal, the target features include time domain features, frequency domain features and time-frequency domain features, the statistical indexes of each target feature are determined through statistical characteristic analysis, the target features are fused according to a target fusion mode based on the statistical indexes to obtain fusion indexes, and the early warning threshold is determined according to the data statistical characteristics of the fusion indexes, so as to utilize the early warning threshold to perform fault monitoring on the engine. As can be seen, according to the sound-vibration signals of the engine in the normal operation state, the statistical indexes of the features are obtained through feature analysis and statistical characteristic analysis, the early warning threshold is determined according to the fusion indexes after feature fusion, the benchmark model of the engine in the normal operation state is constructed to determine whether the engine has a fault, and based on the fusion indexes, the health status of the whole system can be comprehensively analyzed, the abnormal state recognition and analysis of the uncertain fault mode can be realized in the absence of prior knowledge, and the comprehensiveness of fault analysis is improved.

[0106] For example Figure 2 A specific engine fault diagnosis system framework diagram is shown, which includes five parts of a signal acquisition and synchronization module, a normal operation modeling module, a real-time feature extraction and fusion module, a deviation identification and early warning module, and a system interface and display module. The modules are cooperatively operated to realize a closed loop from acquisition, processing of original sound-vibration signals to abnormal identification and early warning output.

[0107] The data layer is used to complete real-time collection of sound signals and vibration signals in the operation process of the aero-engine and ensure high-precision time synchronization of the two types of signals. The feature processing layer is used to construct a high-dimensional feature benchmark model of sound-vibration fusion under the fault-free operation state of the engine, that is, to determine feature indexes and early warning thresholds, and to extract current target features in real time and perform multi-physical quantity fusion processing during the operation of the engine. The diagnosis and decision layer is used to realize real-time monitoring and fault early warning of the operation state deviation. The interface layer is used for visualization and alarm linkage of the monitoring results, and realizes rapid response and display. The system is flexible and modular, and can adapt to detection requirements of different equipment and different fault types. Whether in actual aero-engines or in fault diagnosis of other equipment, the implementation method can be adjusted according to specific requirements.

[0108] The normal operation state modeling in the diagnosis and decision layer is the core, and the technical principle is to provide a reliable reference for abnormal diagnosis and early warning. In the modeling stage, a large number of normal operation samples of the aero-engine are analyzed, multi-dimensional features of sound and vibration signals are extracted, and probability distribution models of each feature are established based on statistical methods, so as to construct a high-dimensional fusion normal operation benchmark model, and at the same time, early warning thresholds are set according to the normal operation features, which provides a reliable basis for subsequent abnormal judgment and early warning in operation monitoring. That is, the system first collects the sound and vibration signals of the engine in operation, and through the links of feature extraction, statistical analysis, feature fusion and early warning threshold setting, a closed-loop diagnosis and early warning process is formed. In the monitoring stage, the system collects the sound and vibration signals of the engine in real time, extracts and fuses multi-dimensional features according to the data processing process established in the modeling stage, and compares the real-time fusion indexes obtained with the normal operation benchmark model. When the monitoring features exceed the normal range and exceed the early warning threshold, the system can realize abnormal diagnosis and early warning. Through normal operation state modeling, not only the sound and vibration feature space of the normal working condition is defined, but also a reliable basis for abnormal judgment and alarm of the fusion features is provided to realize online monitoring, rapid response and fault early warning.

[0109] Correspondingly, the embodiment of the application also discloses an engine fault diagnosis device, which refers to Figure 3 The device comprises:

[0110] The feature extraction module 11 is used for synchronously collecting sound signals and vibration signals of the engine in a normal operation state, and performing feature extraction on the sound signals and the vibration signals to obtain corresponding target features; the target features include time domain features, frequency domain features and time-frequency domain features;

[0111] The feature fusion module 12 is used for determining statistical indexes of each target feature by statistical characteristic analysis, and performing feature fusion on all target features according to a target fusion mode to obtain a fusion index based on the statistical indexes;

[0112] A threshold determination module 13 is configured to determine a pre-warning threshold according to a data statistical characteristic of the fusion index, so as to utilize the pre-warning threshold to monitor the engine failure.

[0113] As can be seen, in the embodiment, the sound signal and the vibration signal of the engine in the normal operation state are synchronously collected, and the corresponding target features are extracted from the sound signal and the vibration signal; the target features include time domain features, frequency domain features and time-frequency domain features; statistical indexes of each target feature are determined through statistical characteristic analysis, all target features are fused according to a target fusion mode based on the statistical indexes to obtain a fusion index; and a pre-warning threshold is determined according to a data statistical characteristic of the fusion index, so as to utilize the pre-warning threshold to monitor the engine failure. It can be seen that, according to the sound and vibration signals of the engine in the normal operation state, the statistical indexes of the features are obtained through feature analysis and statistical characteristic analysis, the pre-warning threshold is determined according to the fusion index after the feature fusion, the normal operation state of the engine is determined by constructing a baseline model, whether the engine fails is determined, and based on the fusion index, the health status of the whole system can be comprehensively analyzed, the abnormal state recognition and analysis of the uncertain failure mode are realized in the absence of prior knowledge, and the comprehensiveness of the failure analysis is improved.

[0114] In some specific embodiments, the feature fusion module 12 can specifically include:

[0115] A probability model construction unit is configured to establish a corresponding probability distribution model for each target feature;

[0116] A statistical index calculation unit is configured to calculate an index for describing a distribution center, an index for describing a distribution dispersion degree and a confidence interval according to the probability distribution model, so as to obtain the statistical index of the target feature;

[0117] A fusion index determination unit is configured to obtain the fusion index through weighted summation according to the corresponding weight of each target feature and the corresponding confidence interval of each target feature.

[0118] In some specific embodiments, the feature fusion module 12 can specifically include:

[0119] A standardization processing unit is configured to perform standardization processing on the target features to obtain corresponding standardized features of the target features;

[0120] A discretization unit is configured to statistically analyze the probability distribution of the target features by discretizing the standardized features;

[0121] An information amount calculation unit is configured to calculate an entropy value of the target features according to the probability distribution, and determine the information amount of the target features according to the entropy value.

[0122] a weight determination unit configured to determine a weight corresponding to each target feature according to the information amount;

[0123] a fusion index determination unit configured to perform weighted fusion on the normalized features based on the weight corresponding to each target feature to obtain the fusion index.

[0124] In some specific embodiments, the engine fault diagnosis apparatus can specifically include:

[0125] a segmentation unit configured to segment the sound signal and the vibration signal to obtain a plurality of sampling segments before performing feature extraction on the sound signal and the vibration signal to obtain corresponding target features;

[0126] The threshold determination module 13 can specifically include:

[0127] a sequence acquisition unit configured to calculate the fusion index corresponding to each sampling segment to obtain a fusion index sequence;

[0128] a threshold determination unit configured to determine the early warning threshold based on the mean and the standard deviation of the fusion index sequence.

[0129] In some specific embodiments, the engine fault diagnosis apparatus can specifically be configured to dynamically adjust the early warning threshold according to the dynamic characteristics of the fusion index.

[0130] In some specific embodiments, the feature extraction module 11 can specifically include:

[0131] an analysis unit configured to perform time domain analysis, frequency domain analysis, and time-frequency domain analysis on the sound signal and the vibration signal, respectively;

[0132] a feature extraction unit configured to take the time domain features, the frequency domain features, and the time-frequency domain features corresponding to the sound signal and the vibration signal obtained by analysis as the target features.

[0133] In some specific embodiments, the threshold determination module 13 can specifically include:

[0134] a real-time target feature extraction unit configured to synchronously acquire real-time sound signals and real-time vibration signals of the engine, and perform feature extraction on the real-time sound signals and the real-time vibration signals to obtain corresponding real-time target features;

[0135] a real-time fusion index determination unit configured to perform feature fusion on all the real-time target features according to the target fusion mode to obtain a real-time fusion index;

[0136] The fault analysis unit is configured to determine whether the engine has a fault by comparing the real-time fusion index with the early warning threshold.

[0137] Further, the embodiment of the present application further discloses an electronic device, referring to Figure 4 The contents in the drawings cannot be considered as any limitation on the use range of the present application.

[0138] Figure 4 A structural schematic diagram of an electronic device 20 is provided in the embodiment of the present application. The electronic device 20 specifically can include at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25 and a communication bus 26. The memory 22 is configured to store a computer program, and the computer program is loaded and executed by the processor 21 to implement the related steps in the engine fault diagnosis method disclosed in any of the foregoing embodiments.

[0139] In the embodiment, the power supply 23 is configured to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol followed by the communication interface 24 can be any communication protocol applicable to the technical solution of the present application, which is not limited specifically herein; the input / output interface 25 is configured to obtain external input data or output data to the outside, and the specific interface type can be selected according to the specific application needs, which is not limited specifically herein.

[0140] In addition, the memory 22 as a carrier for resource storage can be a read-only memory, a random access memory, a magnetic disk or an optical disk, etc., and the resources stored thereon include an operating system 221, a computer program 222 and data 223 including statistical indexes, etc., and the storage mode can be temporary storage or permanent storage.

[0141] The operating system 221 is configured to manage and control each hardware device on the electronic device 20 and the computer program 222, so as to implement the operation and processing of the processor 21 on the mass data 223 in the memory 22, and the operating system 221 can be Windows Server, Netware, Unix, Linux, etc. The computer program 222 can further include computer programs for completing other specific work in addition to the computer programs for completing the engine fault diagnosis method executed by the electronic device 20 disclosed in any of the foregoing embodiments.

[0142] Further, the embodiment of the present application further discloses a computer storage medium, and the computer storage medium stores computer executable instructions. When the computer executable instructions are loaded and executed by a processor, the steps of the engine fault diagnosis method disclosed in any of the foregoing embodiments are implemented.

[0143] Further, the embodiment of the present application further discloses a computer program product comprising a computer program which, when executed by a processor, implements the steps of the engine fault diagnosis method disclosed in any of the foregoing embodiments.

[0144] The embodiments in the specification are described in progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the embodiments can be mutually referred to. For the device disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part.

[0145] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0146] Finally, it should be noted that in this document, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0147] The above describes in detail the engine fault diagnosis method, device, equipment and storage medium provided by the present application. The principles and implementation manners of the present application are described by applying specific examples in this document. The above embodiment description is only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range will be changed; and the above description should not be understood as limiting the present application.

Claims

1. An engine failure diagnosis method characterized by comprising: The method comprises the following steps: Synchronously collecting sound signals and vibration signals of an engine in a normal running state, and performing feature extraction on the sound signals and the vibration signals to obtain corresponding target features; the target features include time domain features, frequency domain features, and time-frequency domain features; Determining statistical indexes of each target feature through statistical characteristic analysis, and performing feature fusion on all target features based on the statistical indexes according to a target fusion mode to obtain a fusion index; Determining a pre-warning threshold according to data statistical characteristics of the fusion index, so as to use the pre-warning threshold to monitor faults of the engine; The method for determining the statistical indexes of each target feature through statistical characteristic analysis, and performing feature fusion on all target features based on the statistical indexes according to a target fusion mode to obtain a fusion index comprises any one of the following modes: A corresponding probability distribution model is established for each target feature; indexes for describing distribution centers, indexes for describing distribution dispersion degrees, and confidence intervals are calculated according to the probability distribution model, so as to obtain the statistical indexes of the target features; the information entropy weight method is used to calculate the weight corresponding to each target feature according to the Shannon entropy of each target feature; the fusion index is obtained by weighted summation according to the weight corresponding to each target feature and the confidence interval corresponding to each target feature; Or, the target features are normalized to obtain standardized features corresponding to the target features; the probability distribution of the target features is obtained by discretizing the standardized features; the entropy value of the target features is calculated according to the probability distribution, and the information amount of the target features is determined according to the entropy value; the weight corresponding to each target feature is determined according to the information amount; the fusion index is obtained by weighted fusion based on the weight corresponding to each target feature and the standardized feature.

2. The engine failure diagnosis method according to claim 1, characterized by, Before the feature extraction on the sound signals and the vibration signals to obtain corresponding target features, the method further comprises the following steps: Segmenting the sound signals and the vibration signals to obtain a plurality of sampling segments; The method for determining a pre-warning threshold according to data statistical characteristics of the fusion index comprises the following steps: Calculating the fusion index corresponding to each sampling segment to obtain a fusion index sequence; The pre-warning threshold is determined based on the mean value and the standard deviation of the fusion index sequence.

3. The engine failure diagnostic method according to claim 1, characterized by, The method further comprises the following steps: The pre-warning threshold is dynamically adjusted according to the dynamic characteristics of the fusion index.

4. The engine failure diagnosis method according to claim 1, characterized by, The feature extraction on the sound signals and the vibration signals to obtain corresponding target features comprises the following steps: Performing time domain analysis, frequency domain analysis, and time-frequency domain analysis on the sound signals and the vibration signals respectively; The time domain features, the frequency domain features, and the time-frequency domain features corresponding to the sound signals and the vibration signals obtained by analysis are taken as the target features.

5. The engine failure diagnostic method according to any one of claims 1 to 4, characterized by, The method for monitoring faults of the engine using the pre-warning threshold comprises the following steps: Synchronously collecting real-time sound signals and real-time vibration signals of the engine, and performing feature extraction on the real-time sound signals and the real-time vibration signals to obtain corresponding real-time target features; Performing feature fusion on all real-time target features according to the target fusion mode to obtain a real-time fusion index; By comparing the real-time fusion index with the early warning threshold, it is determined whether the engine has a fault.

6. An engine failure diagnosis device characterized by comprising: The method comprises the steps of: a feature extraction module is configured to synchronously collect sound signals and vibration signals of the engine in a normal operating state, and extract features from the sound signals and the vibration signals to obtain corresponding target features; the target features include time domain features, frequency domain features, and time-frequency domain features; a feature fusion module is configured to determine statistical indexes of each target feature by statistical property analysis, and fuse all target features according to a target fusion mode based on the statistical indexes to obtain a fusion index; a threshold determination module is configured to determine an early warning threshold according to data statistical characteristics of the fusion index, so as to monitor the engine for faults by using the early warning threshold; wherein the feature fusion module is configured to establish a corresponding probability distribution model for each target feature, calculate indexes for describing distribution centers, indexes for describing distribution dispersion degrees, and confidence intervals according to the probability distribution model, so as to obtain the statistical indexes of the target features; calculate the weight of each target feature according to the Shannon entropy of each target feature by using an information entropy weight method; and obtain the fusion index by weighted summation according to the weight of each target feature and the confidence interval of each target feature. Alternatively, the feature fusion module is configured to perform standardization processing on the target features to obtain standardized features corresponding to the target features; discretize the standardized features to obtain the probability distribution of the target features; calculate the entropy value of the target features according to the probability distribution, determine the information amount of the target features according to the entropy value; determine the weight of each target feature according to the information amount; and perform weighted fusion on the weight and the standardized features of each target feature to obtain the fusion index.

7. An electronic device, comprising: The method comprises the steps of: a memory for storing a computer program; a processor for executing the computer program to implement the engine fault diagnosis method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, a memory for storing a computer program; wherein the computer program is executed by a processor to implement the engine fault diagnosis method according to any one of claims 1 to 5.

Images