A mechanical and electrical system fault diagnosis system based on vibration law analysis

Abstract

The present invention relates to the technical field of electromechanical fault diagnosis, and provides an electromechanical system fault diagnosis system based on vibration law analysis. A multi-source data fusion module is used to obtain multidimensional data in the electromechanical system, and fuses the multi-source data with the time-frequency characteristics of the vibration signal as the core to generate a weighted fusion tensor. The diagnosis model module extracts working condition invariant features from the weighted fusion tensor generated by the multi-source data fusion module to solve the model generalization problem and outputs the fused feature vector. The diagnosis module combines the physical law of the vibration signal with the feature vector output by the diagnosis model module to perform fault diagnosis of the electromechanical system and provide visual feedback. The present invention realizes efficient integration and feature extraction of multi-source data by fusing multi-source data through the multi-source data fusion module to perform weighted tensor modeling, dual-channel feature learning of the diagnosis model module, and adversarial domain adaptation gradient inversion training, thereby solving the problems of low information utilization and insufficient model generalization ability.

Description

Technical Field

[0001] The present invention relates to the technical field of electromechanical fault diagnosis, in particular to an electromechanical system fault diagnosis system based on vibration law analysis. Background Art

[0002] Vibration pattern analysis of electromechanical equipment is one of the core means of fault diagnosis. Traditional methods in the past relied on manual feature extraction and engineering experience, such as wavelet packet transform and fast Fourier transform, but they are susceptible to noise interference and inefficient under complex working conditions. For example, the manual feature extraction process is easily affected by background noise, resulting in the loss of key fault information; second, the algorithm has poor adaptability to non-stationary signals and has difficulty capturing transient impact characteristics; third, the diagnostic process has a lag and cannot meet real-time requirements.

[0003] In recent years, advances in intelligent sensing technology have led to breakthroughs in fault monitoring methods based on vibration signal analysis. These new systems achieve global awareness of equipment status by building distributed sensor networks and, combined with deep learning methods, directly mine fault characterization information from raw vibration signals. Among these, intelligent monitoring systems based on high-frequency vibration sensors and big data analysis enable online assessment of equipment status, while end-to-end diagnostic methods based on deep neural networks significantly improve diagnostic accuracy through automatic feature extraction, demonstrating significant application value in industrial scenarios.

[0004] However, the prior art still has the following disadvantages:

[0005] 1. There are defects in the fusion of multi-source data. Different types of data are difficult to effectively integrate, resulting in low information utilization and inability to fully reflect the operating status of the electromechanical system.

[0006] 2. The model learning and generalization capabilities of feature learning are poor. The constructed models are often only applicable to specific scenarios or data sets. When faced with new working conditions, different equipment parameters or complex and changing operating environments, the model's diagnostic accuracy drops significantly, making it difficult to meet the diverse fault diagnosis needs in actual projects. Summary of the Invention

[0007] In order to solve the above technical problems, the present invention provides an electromechanical system fault diagnosis system based on vibration law analysis. Through the improved Hilbert-Huang transform and envelope spectrum entropy weighted tensor modeling of the multi-source data fusion module and the adversarial domain adaptive gradient inversion training of the diagnosis model module, efficient integration and feature extraction of multi-source data are achieved to solve the problems in the prior art.

[0008] A mechanical and electrical system fault diagnosis system based on vibration law analysis, comprising:

[0009] A multi-source data fusion module is used to obtain multi-dimensional data in the electromechanical system, fuse the multi-source data with the time-frequency characteristics of the vibration signal as the core, and generate a weighted fusion tensor;

[0010] A diagnostic model module, which is data-connected to the multi-source data fusion module, extracts working condition-invariant features from the weighted fusion tensor generated by the multi-source data fusion module, solves the model generalization problem, and outputs a fused feature vector;

[0011] The diagnosis module is connected to the diagnosis model module and the multi-source data fusion module, and combines the physical laws of the vibration signal with the characteristic vector output by the diagnosis model module to perform fault diagnosis of the electromechanical system and provide visual feedback.

[0012] Preferably, the multi-source data fusion module includes a vibration signal processing unit and a weight calculation unit;

[0013] The signal processing unit collects multi-source data and real-time multi-source data from the history or laboratory failure of the electromechanical system, uses the improved Hilbert-Huang transform to generate a time-frequency spectrum matrix for the vibration signal data as a fusion benchmark feature, and constructs a multi-dimensional tensor in combination with the multi-source auxiliary data;

[0014] The weight calculation unit calculates the frequency band importance weights of the vibration time-frequency spectrum based on the constructed multidimensional tensor and adopts the envelope spectrum entropy as the weight index, and performs multi-source data tensor modeling based on the frequency band importance weights, constructing the time-frequency spectrum matrix of the vibration signal and the multi-source auxiliary parameters of temperature and current into a weighted fusion tensor.

[0015] Preferably, the signal processing unit receives the original vibration signal Generate a vibration time-frequency spectrum, where is the total number of t, Indicates the dimension The specific calculation process is as follows:

[0016] First, perform CEEMDAN decomposition: Decomposed into n IMF components and residuals:

[0017] ;

[0018] Among them, t represents the sampling time point of obtaining the sample, represents the i-th IMF component, is the residual component;

[0019] Hilbert transform: Calculate the analytical signal for each IMF component :

[0020] ;

[0021] in, is the Hilbert transform operator, j represents the imaginary unit, which is used to place the Hilbert transformed signal in the imaginary part of the complex plane;

[0022] After the Hilbert transform is completed, based on the analytical signal The instantaneous amplitude and instantaneous phase Calculate the time-frequency spectrum matrix :

[0023] ;

[0024] in, Represents frequency variables, that is, continuous frequency variables in spectrum analysis or the set of all discrete frequency components. When f has a subscript, that is, Represents the specific frequency value of the i-th discretization, which is determined by the Fourier transform resolution;

[0025] Finally, multi-source data alignment is performed, and linear interpolation is performed on the temperature and current low-frequency signals, which are aligned to the vibration signal timestamp to construct a multidimensional tensor. : Where D represents the dimension including the spectral frequency, amplitude, temperature and current during vibration.

[0026] Preferably, the weight calculation unit uses envelope spectrum entropy as a single weight indicator to simultaneously characterize the impact characteristics and frequency domain complexity of the vibration signal. The specific calculation process is as follows:

[0027] Envelope signal extraction for original vibration signal Perform Hilbert envelope demodulation to obtain the envelope signal :

[0028] ;

[0029] Then, the envelope spectrum entropy is calculated to obtain the envelope signal Fourier spectrum of ;

[0030] Define the probability of spectrum energy distribution :

[0031] ;

[0032] Calculate the spectral entropy H:

[0033] ;

[0034] calculate Band weight :

[0035] ;

[0036] in, Strengthen the high energy frequency band, corresponding to the fault characteristic frequency, Suppress broadband noise. A high entropy value indicates a dispersed spectrum, while a low entropy value indicates concentrated energy.

[0037] Then scale the vibration spectrum dimensions:

[0038] ;

[0039] Output weighted fusion tensor and the original time-frequency spectrum .

[0040] Preferably, the diagnostic model module includes a feature extraction unit and a generalization enhancement unit;

[0041] The feature extraction unit is used to extract time-frequency features based on the weighted data tensor generated by the multi-source data fusion module by establishing a dual-channel 1DCNN to obtain fault-related features, wherein the first channel extracts resonance frequency band features through frequency domain slices of the vibration time-frequency spectrum, and the second channel extracts impact waveform features through the time domain vibration amplitude sequence;

[0042] The generalization enhancement unit is used to process the feature data extracted by the feature extraction unit based on multi-source data of the electromechanical system history or laboratory failure, and to establish and enhance the generalization ability of the features to adapt to different working conditions and data distributions.

[0043] Preferably, the feature extraction unit first inputs a historical data set of multi-source data based on the history of electromechanical systems or laboratory faults, including vibration signals and fault labels of multiple working conditions;

[0044] Perform dual-channel convolution training calculations, learn resonance patterns in the frequency domain channel, and learn impact waveforms in the time domain channel. The specific process of feature extraction by establishing a dual-channel 1DCNN is as follows:

[0045] Channel 1, frequency domain characteristics:

[0046] enter , convolution kernel :

[0047] ;

[0048] Channel 2, time domain features: input vibration amplitude sequence , convolution kernel :

[0049] ;

[0050] Among them, amp represents the amplitude, h represents the output characteristics, Indicates the matching degree of the resonance feature of the tth at time t, and the output feature represents the impact feature matching degree at time t, Represents the convolution kernel sliding time offset, represents the bias term, Represents the bias term in the frequency domain feature calculation, That is, it represents the bias term in the time domain feature calculation, k represents the time width of the convolution kernel, is a nonlinear activation function;

[0051] Perform attention mechanism fusion to generate attention weights :

[0052] ;

[0053] Perform feature fusion:

[0054] ;

[0055] in, is the Sigmoid function, is the weight of the fully connected layer, Represents the bias term in the attention mechanism fusion calculation, and finally outputs the fused feature vector .

[0056] Preferably, the generalization enhancement unit performs adversarial domain adaptation on the feature extraction unit to solve the generalization problem;

[0057] Use equipment data from the laboratory or known working conditions as source domain data, and new equipment or variable working condition data from the field as target domain data;

[0058] Adversarial training: The feature extraction unit is trained through gradient reversal to extract fault features that are unrelated to the working conditions;

[0059] In the real-time inference stage, directly input the weighted fusion tensor based on real-time multi-source data, use the trained feature extraction unit to extract features, output the weighted fusion tensor based on real-time multi-source data, and calculate the fused feature vector .

[0060] Preferably, the diagnostic module generates a diagnostic result based on the characteristic vector and the physical law of vibration;

[0061] First, obtain the fault feature vector output by the diagnosis model module and the original vibration time-frequency spectrum generated by the multi-source data fusion module;

[0062] Fault classification and location: A normal state hypersphere is constructed through support vector data description, and the distance from the feature vector to the hypersphere boundary is calculated to achieve anomaly detection. Then, combined with the vibration signal characteristic frequency library or the bearing fault characteristic frequency calculation formula, the abnormal feature vector is frequency band deconvolved to locate the faulty component.

[0063] Preferably, the diagnostic module can also provide visual feedback, superimpose the energy heat map of the fault frequency band on the original time-frequency spectrum, mark the position deviation between the characteristic frequency and the actual resonance peak, and output the fault type, position confidence and time-frequency domain visualization map.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. The present invention uses an improved Hilbert-Huang transform in the signal processing unit of the multi-source data fusion module to generate a time-frequency spectrum matrix for the vibration signal as a fusion benchmark feature, combines multi-source auxiliary data to construct a multidimensional tensor, and uses envelope spectrum entropy as a weight indicator to calculate the frequency band importance weight of the vibration time-frequency spectrum and perform multi-source data tensor modeling, thereby achieving effective integration of multi-source data of different types, formats, sampling frequencies and dimensions, and solving the problem that it is difficult to effectively integrate multi-source data, resulting in low information utilization and inability to fully reflect the operating status of the electromechanical system.

[0066] 2. The present invention performs adversarial domain adaptation on the feature extraction unit through the generalization enhancement unit in the diagnostic model module, takes the equipment data from the laboratory or known working conditions as the source domain data, and the new equipment or variable working condition data from the field as the target domain data, and extracts fault features that are unrelated to the working conditions through gradient reversal training of the feature extraction unit, thereby achieving the generalization ability of the enhanced features and solving the problem that the model learning ability and generalization ability of feature learning are poor, the diagnostic accuracy decreases when facing new working conditions, different equipment parameters or complex and changeable operating environments, and it is difficult to meet the diverse fault diagnosis needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a schematic diagram of the data connection process between the modules and units of the present invention when working;

[0068] Figure 2 Schematic diagram of the process of calculating weighted fusion tensor by the multi-source data fusion module of the present invention;

[0069] Figure 3 It is a schematic diagram of the calculation flow of the diagnostic model module of the present invention. DETAILED DESCRIPTION

[0070] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0071] The present invention provides a mechanical and electrical system fault diagnosis system based on vibration law analysis, comprising:

[0072] A multi-source data fusion module is used to obtain multi-dimensional data in the electromechanical system, fuse the multi-source data with the time-frequency characteristics of the vibration signal as the core, and generate a weighted fusion tensor;

[0073] A diagnostic model module, which is data-connected to the multi-source data fusion module, extracts working condition-invariant features from the weighted fusion tensor generated by the multi-source data fusion module, solves the model generalization problem, and outputs a fused feature vector;

[0074] The diagnosis module is connected to the diagnosis model module and the multi-source data fusion module, and combines the physical laws of the vibration signal with the characteristic vector output by the diagnosis model module to perform fault diagnosis of the electromechanical system and provide visual feedback. Example 1:

[0075] like Figure 1-Figure 3 As shown, in this embodiment, a large railway station building includes key electromechanical equipment such as passenger elevators, escalators, ventilation and air-conditioning units, distribution transformers, and automatic door systems. It faces daily challenges such as high-frequency start and stop of equipment, periodic fluctuations in passenger flow, and complex environmental noise.

[0076] Traditional monitoring systems have difficulty capturing the cumulative characteristics of equipment fatigue due to the scattered deployment of sensors and isolated data. They have even been closed to local areas due to drive chain breaks, seriously affecting passenger travel efficiency.

[0077] At the same time, the station faces extreme passenger flow during special periods such as the Spring Festival travel rush and Golden Week, requiring electromechanical equipment to operate continuously at full capacity and subject to temporary changes in operating mode. Traditional models are prone to misjudgment during sudden changes in operating conditions and are unable to adapt to the parameter differences of temporarily activated equipment, leading to misallocation of operation and maintenance resources. Therefore, the present invention introduces an electromechanical system fault diagnosis system based on vibration pattern analysis to facilitate fault diagnosis.

[0078] First, the signal processing unit collects multi-source data from historical or laboratory faults in railway station electromechanical systems, as well as real-time multi-source data. It uses an improved Hilbert-Huang transform to generate a time-frequency spectrum matrix for the vibration signal data, which serves as a fusion benchmark feature. This matrix is then combined with multi-source auxiliary data to construct a multidimensional tensor.

[0079] The signal processing unit receives the original vibration signal Generate a vibration time-frequency spectrum, where is the total number of t, Indicates the dimension The specific calculation process is as follows:

[0080] Perform CEEMDAN decomposition: Decomposed into n IMF components and residuals:

[0081] ;

[0082] Among them, t represents the sampling time point of obtaining the sample, represents the i-th IMF component, is the residual component;

[0083] Hilbert transform: Calculate the analytical signal for each IMF component :

[0084] ;

[0085] in, is the Hilbert transform operator, j represents the imaginary unit, which is used to place the Hilbert transformed signal in the imaginary part of the complex plane;

[0086] After the Hilbert transform is completed, based on the analytical signal The instantaneous amplitude and instantaneous phase Calculate the time-frequency spectrum matrix :

[0087] ;

[0088] in, Represents frequency variables, that is, continuous frequency variables in spectrum analysis or the set of all discrete frequency components. When f has a subscript, that is, Represents the specific frequency value of the i-th discretization, which is determined by the Fourier transform resolution;

[0089] Finally, multi-source data alignment is performed, and linear interpolation is performed on the temperature and current low-frequency signals, which are aligned to the vibration signal timestamp to construct a multidimensional tensor. : Where D represents the dimension including the spectral frequency, amplitude, temperature and current during vibration.

[0090] The weight calculation unit uses envelope spectrum entropy as a single weight indicator to simultaneously characterize the impact characteristics and frequency domain complexity of the vibration signal. The specific calculation process is as follows:

[0091] Envelope signal extraction for original vibration signal Perform Hilbert envelope demodulation to obtain the envelope signal :

[0092] ;

[0093] Then, the envelope spectrum entropy is calculated to obtain the envelope signal The Fourier spectrum of ;

[0094] Define the probability of spectrum energy distribution :

[0095] ;

[0096] Calculate the spectral entropy H:

[0097] ;

[0098] calculate Band weight :

[0099] ;

[0100] in, Strengthen the high energy frequency band, corresponding to the fault characteristic frequency, Suppress broadband noise. A high entropy value indicates a dispersed spectrum, while a low entropy value indicates concentrated energy.

[0101] Then scale the vibration spectrum dimensions:

[0102] ;

[0103] Output weighted fusion tensor and the original time-frequency spectrum .

[0104] By performing CEEMDAN decomposition on the vibration signals of equipment such as escalator drive chains and elevator traction machines, a time-frequency spectrum matrix is generated as a fusion benchmark feature to solve the problem of weak early fault signals in strong noise environments.

[0105] Synchronously collect multi-source data such as vibration, current, speed, temperature and humidity, align low-frequency data to the vibration signal timestamp through linear interpolation, and construct a multi-dimensional tensor containing frequency, amplitude and operating parameters to achieve spatiotemporal unification of multi-dimensional data.

[0106] By using envelope spectrum entropy to calculate the importance weight of each frequency band of the vibration signal, the fault characteristic frequency band is enhanced, such as suppressing the ambient noise frequency band, thereby achieving the effect of improving the signal-to-noise ratio.

[0107] The diagnosis model module includes a feature extraction unit and a generalization enhancement unit;

[0108] The feature extraction unit is used to extract time-frequency features based on the weighted data tensor generated by the multi-source data fusion module by establishing a dual-channel 1DCNN to obtain fault-related features, wherein the first channel extracts resonance frequency band features through frequency domain slices of the vibration time-frequency spectrum, and the second channel extracts impact waveform features through the time domain vibration amplitude sequence;

[0109] The feature extraction unit first inputs a historical data set of multi-source data based on the history of electromechanical systems or laboratory failures, which contains vibration signals and fault labels under multiple working conditions;

[0110] Perform dual-channel convolution training calculations, learn resonance patterns in the frequency domain channel, and learn impact waveforms in the time domain channel. The specific process of feature extraction by establishing a dual-channel 1DCNN is as follows:

[0111] Channel 1, frequency domain characteristics:

[0112] enter , convolution kernel :

[0113] ;

[0114] Channel 2, time domain features: input vibration amplitude sequence , convolution kernel :

[0115] ;

[0116] Among them, amp represents the amplitude, h represents the output characteristics, Indicates the matching degree of the resonance feature of the tth at time t, and the output feature represents the impact feature matching degree at time t, Represents the convolution kernel sliding time offset, represents the bias term, Represents the bias term in the frequency domain feature calculation, That is, it represents the bias term in the time domain feature calculation, k represents the time width of the convolution kernel, is a nonlinear activation function;

[0117] Perform attention mechanism fusion to generate attention weights :

[0118] ;

[0119] Perform feature fusion:

[0120] ;

[0121] in, is the Sigmoid function, is the weight of the fully connected layer, Represents the bias term in the attention mechanism fusion calculation, and finally outputs the fused feature vector .

[0122] The generalization enhancement unit performs adversarial domain adaptation on the feature extraction unit to solve the generalization problem;

[0123] Use equipment data from the laboratory or known working conditions as source domain data, and new equipment or variable working condition data from the field as target domain data;

[0124] Adversarial training: The feature extraction unit is trained through gradient reversal to extract fault features that are unrelated to the working conditions;

[0125] In the real-time inference stage, directly input the weighted fusion tensor based on real-time multi-source data, use the trained feature extraction unit to extract features, output the weighted fusion tensor based on real-time multi-source data, and calculate the fused feature vector .

[0126] Through dual-channel 1DCNN feature extraction: construct frequency domain channels, extract resonance frequency band features, such as gearbox meshing frequency and time domain channels, extract impact waveform features, such as the instantaneous amplitude mutation of bearing spalling, and automatically learn the deep features of the equipment health status.

[0127] The diagnosis module generates diagnosis results based on the characteristic vector and the physical laws of vibration;

[0128] First, obtain the fault feature vector output by the diagnosis model module and the original vibration time-frequency spectrum generated by the multi-source data fusion module;

[0129] Fault classification and location: A normal state hypersphere is constructed through support vector data description, and the distance from the feature vector to the hypersphere boundary is calculated to achieve anomaly detection. Then, combined with the vibration signal characteristic frequency library or the bearing fault characteristic frequency calculation formula, the abnormal feature vector is frequency band deconvolved to locate the faulty component.

[0130] By constructing an SVDD hypersphere based on historical normal data and calculating the distance from the current eigenvector to the hypersphere boundary in real time, quantitative detection of equipment anomalies is achieved. Combined with the vibration characteristic frequency library, faulty components are located through frequency band deconvolution.

[0131] At the same time, the diagnostic module can also provide visual feedback, superimpose the energy heat map of the fault frequency band on the original time-frequency spectrum, mark the position deviation between the characteristic frequency and the actual resonance peak, and output the fault type, location confidence and time-frequency domain visualization map.

[0132] The embodiments of the present invention are provided for the purpose of illustration and description. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A mechanical and electrical system fault diagnosis system based on vibration law analysis, characterized in that: include: A multi-source data fusion module is used to obtain multi-dimensional data in the electromechanical system, fuse the multi-source data with the time-frequency characteristics of the vibration signal as the core, and generate a weighted fusion tensor; A diagnostic model module, which is data-connected to the multi-source data fusion module, extracts working condition-invariant features from the weighted fusion tensor generated by the multi-source data fusion module, solves the model generalization problem, and outputs a fused feature vector; A diagnostic module, which is data-connected to the diagnostic model module and the multi-source data fusion module, combines the physical laws of the vibration signal with the characteristic vector output by the diagnostic model module to perform fault diagnosis of the electromechanical system and provide visual feedback; The multi-source data fusion module includes a vibration signal processing unit and a weight calculation unit; The vibration signal processing unit collects multi-source data and real-time multi-source data from the history or laboratory failure of the electromechanical system, uses the improved Hilbert-Huang transform to generate a time-frequency spectrum matrix for the vibration signal data as a fusion benchmark feature, and combines the multi-source auxiliary data to construct a multi-dimensional tensor; The weight calculation unit calculates the frequency band importance weight of the vibration time-frequency spectrum based on the constructed multidimensional tensor and adopts the envelope spectrum entropy as the weight index, and performs multi-source data tensor modeling based on the frequency band importance weight, constructing the time-frequency spectrum matrix of the vibration signal and the temperature and current multi-source auxiliary parameters into a weighted fusion tensor; The diagnostic model module includes a feature extraction unit and a generalization enhancement unit; The feature extraction unit is used to extract time-frequency features based on the weighted fusion tensor generated by the multi-source data fusion module by establishing a dual-channel 1DCNN to obtain fault-related features, wherein the first channel extracts resonance frequency band features through frequency domain slices of the vibration time-frequency spectrum, and the second channel extracts impact waveform features through the time domain vibration amplitude sequence; The generalization enhancement unit is used to process the feature data extracted by the feature extraction unit based on multi-source data of the electromechanical system history or laboratory failure, and to establish and enhance the generalization ability of the features to adapt to different working conditions and data distributions.

2. The electromechanical system fault diagnosis system based on vibration law analysis according to claim 1, characterized in that: The vibration signal processing unit receives the original vibration signal Generate a vibration time-frequency spectrum, where is the total number of t, Indicates the dimension The specific calculation process is as follows: First, perform CEEMDAN decomposition: Decomposed into n IMF components and residuals: ； Among them, t represents the sampling time point of obtaining the sample, represents the i-th IMF component, is the residual component; Hilbert transform: Calculate the analytical signal for each IMF component : ； in, is the Hilbert transform operator, j represents the imaginary unit, which is used to place the Hilbert transformed signal in the imaginary part of the complex plane; After the Hilbert transform is completed, based on the analytical signal The instantaneous amplitude and instantaneous phase Calculate the time-frequency spectrum matrix : ； in, Represents frequency variables, that is, continuous frequency variables in spectrum analysis or the set of all discrete frequency components. When f has a subscript, that is, Represents the specific frequency value of the i-th discretization, which is determined by the Fourier transform resolution; Finally, multi-source data alignment is performed, and linear interpolation is performed on the temperature and current low-frequency signals, which are aligned to the vibration signal timestamp to construct a multidimensional tensor. : Where D represents the dimension including the spectral frequency, amplitude, temperature and current during vibration.

3. The electromechanical system fault diagnosis system based on vibration law analysis according to claim 2, characterized in that: The weight calculation unit uses envelope spectrum entropy as a single weight indicator to simultaneously characterize the impact characteristics and frequency domain complexity of the vibration signal. The specific calculation process is as follows: Envelope signal extraction for original vibration signal Perform Hilbert envelope demodulation to obtain the envelope signal : ； Then, the envelope spectrum entropy is calculated to obtain the envelope signal Fourier spectrum of ; Define the probability of spectrum energy distribution : ； Calculate the spectral entropy H: ； calculate Band weight : ； in, Strengthen the high energy frequency band, corresponding to the fault characteristic frequency, Suppress broadband noise. A high entropy value indicates a dispersed spectrum, while a low entropy value indicates concentrated energy. Then scale the vibration spectrum dimensions: ； Output weighted fusion tensor and the original time-frequency spectrum .

4. The electromechanical system fault diagnosis system based on vibration law analysis as claimed in claim 3, characterized in that: The feature extraction unit first inputs a historical data set of multi-source data based on the history of electromechanical systems or laboratory faults, including vibration signals and fault labels of multiple working conditions; Perform dual-channel convolution training calculations, learn resonance patterns in the frequency domain channel, and learn impact waveforms in the time domain channel. The specific process of feature extraction by establishing a dual-channel 1DCNN is as follows: Channel 1, frequency domain characteristics: enter , convolution kernel : ； Channel 2, time domain features: input vibration amplitude sequence , convolution kernel : ； Among them, amp represents the amplitude, h represents the output characteristics, Represents the resonance feature matching degree at time t, and the output feature represents the impact feature matching degree at time t, Represents the convolution kernel sliding time offset, represents the bias term, Represents the bias term in the frequency domain feature calculation, That is, it represents the bias term in the time domain feature calculation, k represents the time width of the convolution kernel, is a nonlinear activation function; Perform attention mechanism fusion to generate attention weights : ； Perform feature fusion: ； in, is the Sigmoid function, is the weight of the fully connected layer, Represents the bias term in the attention mechanism fusion calculation, and finally outputs the fused feature vector .

5. The electromechanical system fault diagnosis system based on vibration law analysis according to claim 4, characterized in that: The generalization enhancement unit performs adversarial domain adaptation on the feature extraction unit to solve the generalization problem; Use equipment data from the laboratory or known working conditions as source domain data, and new equipment or variable working condition data from the field as target domain data; Adversarial training: The feature extraction unit is trained through gradient reversal to extract fault features that are unrelated to the working conditions; In the real-time inference stage, directly input the weighted fusion tensor based on real-time multi-source data, use the trained feature extraction unit to extract features, output the weighted fusion tensor based on real-time multi-source data, and calculate the fused feature vector .

6. The electromechanical system fault diagnosis system based on vibration law analysis according to claim 1, characterized in that: The diagnostic module generates a diagnostic result based on the characteristic vector and the physical law of vibration; First, obtain the fault feature vector output by the diagnosis model module and the original vibration time-frequency spectrum generated by the multi-source data fusion module; Fault classification and location: A normal state hypersphere is constructed through support vector data description, and the distance from the feature vector to the hypersphere boundary is calculated to achieve anomaly detection. Then, combined with the vibration signal characteristic frequency library or the bearing fault characteristic frequency calculation formula, the abnormal feature vector is frequency band deconvolved to locate the faulty component.

7. The electromechanical system fault diagnosis system based on vibration law analysis according to claim 6, characterized in that: The diagnostic module can also provide visual feedback, superimpose the energy heat map of the fault frequency band on the original time-frequency spectrum, mark the position deviation between the characteristic frequency and the actual resonance peak, and output the fault type, location confidence and time-frequency domain visualization map.

Images