Multi-energy system cluster fault intelligent diagnosis and analysis method

Abstract

The invention provides a multi-energy system cluster fault intelligent diagnosis and analysis method, which comprises the steps of collecting original vibration signal data corresponding to a bearing wear signal from a multi-energy system cluster equipment node, extracting specific frequency vibration and amplitude reduction process characteristics, and performing material type identification; according to the material type identification result, the propagation process of the bearing wear signal from the source device to the adjacent device is analyzed, and propagation path analysis data is obtained; according to the propagation path analysis data, identifying the signal contribution degree of each equipment node corresponding to the fault feature difference, and judging the origin point of the potential fault feature; obtaining a vibration frequency subset corresponding to the origin point of the fault feature, and extracting the fault feature after feature correction; and determining the position of fault source equipment in the multi-energy system cluster according to the extracted fault features.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and in particular to a multi-energy system cluster fault intelligent diagnosis analysis method. BACKGROUND

[0002] In the energy industry, the running stability of the multi-energy system cluster is crucial to ensure energy supply and industrial production. The vibration signals between devices propagate through pipes and connecting structures, which is an important means to monitor the health status of the devices. However, existing methods often ignore the influence of complex propagation paths on signals when analyzing vibration signal propagation, resulting in inaccurate fault source positioning. In particular, existing technologies focus on signal collection and analysis of a single device, and do not fully consider the comprehensive influence of physical connections and material differences between devices on vibration signals, making it difficult to accurately trace the fault in a dynamic running environment. In the multi-energy system cluster, the propagation of device vibration signals is significantly affected by the stiffness and density of pipe materials and the tightness of connecting methods. The stiffness and density differences of pipe materials will cause attenuation and deformation of vibration signals during propagation. For example, when the bearing of device A wears out, it will produce a specific frequency vibration of 2-4 kHz on the device casing. This vibration is transmitted to the steel pipe connected directly through the flange, and the steel pipe can well maintain the vibration amplitude and frequency characteristics due to its high stiffness characteristics. However, when the vibration signal continues to propagate to device B (a pump with an aluminum alloy shell) connected to the steel pipe through bolts, the original 2-4 kHz vibration signal will be significantly attenuated due to the lower density and different damping characteristics of aluminum alloy, with an amplitude reduction of about 60-70%, and the high-frequency components are filtered out. The main low-frequency vibration detected on device B is 1-2 kHz. This cascading propagation effect causes device B to also produce corresponding vibration responses, but the vibration characteristics are significantly different from those of the original fault source device A. This signal evolution makes the original fault characteristics significantly different on different devices, increasing the difficulty of fault source identification. SUMMARY

[0003] The present application provides a multi-energy system cluster fault intelligent diagnosis analysis method, comprising:

[0004] S101, collecting original vibration signal data corresponding to bearing wear signal from multi-energy system cluster device nodes, extracting specific frequency vibration and amplitude reduction process characteristics, and identifying material types;

[0005] S102, according to the material type identification result, analyzing the propagation process of the bearing wear signal from the source device to the adjacent device, and obtaining propagation path analysis data;

[0006] S103, according to the propagation path analysis data, identifying the signal contribution degree of each device node corresponding to the fault characteristic difference, and judging the origin point of the potential fault characteristic.

[0007] S104, obtain a vibration frequency subset corresponding to the origin point of the fault feature, and extract the fault feature after feature correction;

[0008] S105, confirming the fault source device position in the multi-energy system cluster according to the extracted fault feature;

[0009] S106, performing path optimization processing through the confirmed fault source device position, and generating an intelligent diagnosis scheme for the bearing wear signal in the multi-energy system cluster under the condition of different pipe materials, which increases the difficulty of fault source identification.

[0010] Preferably, step S101 comprises: obtaining original vibration signals from acceleration sensors of each device node in the multi-energy system cluster, performing fast Fourier transform on the original vibration signals to obtain frequency spectrum data, identifying vibration peak values and corresponding amplitude characteristics in a preset high frequency band, and calculating an initial attenuation characteristic by calculating the attenuation ratio of the amplitude characteristics between different collection points. According to the nodes with an amplitude reduction rate exceeding a first preset threshold and the nodes with a frequency shift exceeding a second preset threshold, a material damping coefficient mapping table is constructed, the material categories of the outer shells of the nodes and the connecting pipes are determined by comparing the low damping coefficient range of the steel pipes with the high damping coefficient range of the aluminum alloy, and the propagation characteristic identification result of the bearing wear signal in different material nodes is obtained by calculating the phase change amount of the vibration signal at the flange connection and the propagation time delay of the vibration signal at the bolt connection.

[0011] Preferably, step S102 comprises: constructing a vibration signal propagation relationship diagram between device nodes according to the difference of damping coefficients in the material type identification result, labeling the material property and connection mode of each node, calculating the acoustic impedance value by taking the square root of the product of the material density and the elastic modulus, and calculating the impedance mismatch coefficient by the ratio of the acoustic impedance between adjacent nodes. The impedance mismatch coefficient is used to calculate the reflection coefficient and the transmission ratio of the vibration signal at the flange connection according to the propagation theory of the vibration wave at the interface of different media. If the reflection coefficient exceeds the preset reflection threshold, it is determined that there is a partial reflection phenomenon of the vibration signal in the path, and the reflection coefficient and the transmission ratio are recorded as the interface propagation characteristics. The coupling relationship between the vibration signal and the natural frequency of the pipeline on different propagation paths is identified by the interface propagation characteristics and the frequency component variation of the vibration signal. When the vibration frequency is close to the natural frequency of the pipeline, resonance phenomenon occurs, and the resonance frequency point and the corresponding amplitude variation are recorded to construct a propagation path characteristic matrix containing the impedance mismatch coefficient, the interface propagation characteristics and the resonance frequency. According to the vibration energy variation of each path in the propagation path characteristic matrix, the least square method is used to fit the amplitude attenuation curve of the vibration signal along the pipeline of different materials to determine the main propagation path and the secondary propagation path from the source device to each adjacent device, and to obtain complete propagation path analysis data.

[0012] Preferably, step S103 comprises: calculating the transmission efficiency of the vibration signal from the source point to each device node according to the propagation path analysis data, determining the energy distribution coefficient by the ratio of the received vibration energy of each node to the total input energy, and constructing an energy distribution matrix of each device node. The frequency offset and the amplitude variation rate of the vibration signal of each node are extracted using the energy distribution matrix. If the frequency offset of a certain node exceeds the preset offset threshold and the amplitude variation rate is lower than the preset attenuation threshold, the node is determined as a candidate point of the fault source, and the frequency characteristics and the amplitude characteristics of the candidate point are recorded. The time delay relationship between different node signals is obtained by calculating the cross-correlation coefficient of the frequency characteristics and the amplitude characteristics of the candidate point. The actual generation order of the vibration signals of each candidate point is determined by inversely calculating the signal generation time according to the vibration propagation speed and the time delay relationship. According to the actual generation order, the node that generates the vibration signal earliest is given the highest weight value, and the weight values of other nodes are distributed in descending order of time. The comprehensive contribution degree of each node is calculated by combining the energy distribution coefficient. When the comprehensive contribution degree of a certain device node exceeds the preset contribution threshold, the device is determined as the origin point of the potential fault characteristics.

[0013] Preferably, the step S104 comprises: extracting frequency components in a preset frequency band from the original vibration signal of the fault feature origin point by short-time Fourier transform, identifying the fundamental frequency and its multiple components, calculating the theoretical fault frequencies of the inner ring, outer ring and rolling elements according to the bearing rotating speed, the number of rolling elements and the contact angle, screening out the frequency components with a deviation within a preset tolerance range from the theoretical value, and constructing a vibration frequency subset. For each frequency component in the vibration frequency subset, the sideband frequency interval on both sides of the main frequency is identified, the ratio of the sideband amplitude to the main frequency amplitude is calculated as the modulation depth, and if the modulation depth exceeds a preset modulation threshold, the instantaneous phase is obtained by Hilbert transform, the phase delay is calculated and compensated according to the propagation path length and wave speed, and the corrected frequency feature set is obtained. According to the corrected frequency feature set, the peak amplitude of each frequency component is calculated as the peak frequency feature, the energy accumulation value in a specific frequency band is calculated as the frequency band energy feature, the spectral kurtosis feature is calculated by the ratio of the fourth-order cumulant to the square of the second-order cumulant, the peak frequency feature, the frequency band energy feature and the spectral kurtosis feature are combined to construct a fault feature vector, and the fault feature of the bearing wear is obtained.

[0014] Preferably, the step S105 comprises: according to the extracted fault feature vector, similarity matching is performed with the standard fault features of each device node in the multi-energy system cluster, the Euclidean distance value is obtained by calculating the square root of the sum of squares of the difference values of each dimension of the feature vector, and if the Euclidean distance value is less than a preset distance threshold, the node is marked as a suspected fault source. Through the device number of the suspected fault source, a pre-established device number and physical location correspondence table is queried to obtain the installation position coordinates of the device in the multi-energy system cluster, and the specific position of the fault source device is confirmed.

[0015] Preferably, the step S106 comprises: constructing a signal attenuation compensation function from the fault source to each monitoring point according to the confirmed fault source device position, in combination with vibration signal propagation characteristics, the compensation coefficient of the steel pipeline is set as the pipeline length multiplied by a low damping coefficient, the compensation coefficient of the aluminum alloy pipeline is set as the pipeline length multiplied by a high damping coefficient, and a material difference compensation parameter set is obtained. The material difference compensation parameter set is used to reconfigure the vibration monitoring network, the sensor arrangement density is increased at the material node with a damping coefficient exceeding a preset damping threshold, the signal amplification multiple of each monitoring point is adjusted according to the compensation parameter set, and if the cumulative attenuation of a certain path exceeds a preset attenuation threshold, a monitoring point is additionally arranged in the middle of the path, and an optimized monitoring layout is formed. Through the optimized monitoring layout, the fault feature recognition threshold corresponding to each propagation path is established according to the frequency offset law and amplitude attenuation characteristics under different material combinations, the corresponding recognition parameter set is automatically called when a specific material sequence is detected, and a fault judgment rule library is constructed. According to the fault judgment rule library, the diagnosis results of multiple monitoring points are fused by using a decision tree method, the monitoring points closer to the fault source are given higher weight values, the fault type and position are comprehensively judged, and an intelligent diagnosis scheme including the optimized configuration of monitoring points, material compensation parameters and recognition thresholds is formed.

[0016] The technical scheme provided by the embodiment of the present application can include the following beneficial effects:

[0017] The application discloses a multi-energy system cluster fault intelligent diagnosis analysis method, which solves the problem of a complex pipeline material difference and a large difficulty in fault source identification. The application collects device node vibration signals, extracts specific frequency and amplitude characteristics, accurately identifies material types, analyzes the propagation path of signals from the source device to the adjacent device, and reveals the propagation law of fault characteristics. Based on this, the application calculates the signal contribution degree of each node, locates the origin point of the fault characteristics, corrects the vibration frequency subset to extract accurate fault characteristics, and finally confirms the fault source device position. By optimizing the propagation path, the application generates an intelligent diagnosis scheme, effectively deals with signal attenuation and interference caused by pipeline material differences. The technical effect lies in significantly improving the accuracy and efficiency of fault source positioning in a multi-energy system cluster, reducing the difficulty of fault diagnosis in a complex environment, and providing a reliable basis for equipment maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The flowchart of the multi-energy system cluster fault intelligent diagnosis analysis method of the application. DETAILED DESCRIPTION

[0019] The technical scheme in the embodiments of the application will be described clearly and in detail below with reference to the drawings in the embodiments of the application. The described embodiments are only a part of the embodiments of the application.

[0020] AsFigure 1 The embodiment of the multi-energy system cluster fault intelligent diagnosis analysis method can specifically include the following steps.

[0021] S101, collect the original vibration signal data corresponding to the bearing wear signal from the multi-energy system cluster device node, extract the specific frequency vibration and amplitude reduction process characteristics, and perform material type identification.

[0022] The original vibration signal is obtained from the acceleration sensor of each device node in the multi-energy system cluster, the frequency spectrum data is obtained by performing fast Fourier transform on the original vibration signal, the vibration peak value in the preset high frequency band and the corresponding amplitude characteristic are identified, and the initial attenuation characteristic is calculated by calculating the attenuation ratio between different collection points. According to the nodes with amplitude reduction rate exceeding the first preset threshold and the nodes with frequency shift exceeding the second preset threshold in the initial attenuation characteristic, a material damping coefficient mapping table is constructed, the material categories of the node device shell and the connecting pipe are judged by comparing the low damping coefficient range of the steel pipe with the high damping coefficient range of the aluminum alloy, and the material categories and the resonance frequency distribution in the frequency spectrum data are used to calculate the phase change amount of the vibration signal at the flange connection and the propagation time delay of the vibration signal at the bolt connection. The propagation characteristic identification result of the bearing wear signal at different material nodes is obtained.

[0023] In an embodiment, the multi-energy system cluster includes compressor units, delivery pump units, and heat exchanger units, and each device is connected to the pipe system through flanges or bolts. The acceleration sensor is installed on the surface of the device shell and the pipe connection node, and the sampling frequency is set to exceed the preset sampling threshold to ensure that the high-frequency vibration components generated by the bearing wear are captured.

[0024] Specifically, when performing fast Fourier transform on the collected original vibration signal, first, the time domain signal is processed with a window function to reduce the frequency spectrum leakage phenomenon. In the frequency spectrum data obtained after transformation, the vibration peak value in the preset high frequency band is focused on, and this frequency band usually corresponds to the fault characteristic frequency of the bearing inner ring or the rolling body. By comparing the amplitude difference of the source device node and the adjacent device node in the same frequency band, the attenuation ratio is calculated.

[0025] For example, when the compressor bearing wear generates a specific frequency vibration, the vibration is transmitted to the adjacent aluminum alloy pump shell through the steel pipe, and the amplitude attenuation ratio is obtained by calculating the logarithmic ratio of the amplitudes of the two nodes. At the same time, the shift amount of the frequency spectrum peak value from high frequency to low frequency is recorded, and this shift reflects the filtering effect of different materials on high frequency components.

[0026] The material damping coefficient mapping table is established based on the internal friction characteristics of the material. The low damping coefficient range of steel corresponds to smaller vibration attenuation, while the high damping coefficient range of aluminum alloy leads to more obvious vibration attenuation.

[0027] It should be noted that in the material category judgment process, the measured attenuation characteristics are matched with the standard values in the mapping table. When the amplitude reduction rate of a certain node exceeds the first preset threshold and the frequency offset exceeds the second preset threshold, it is determined that the node is a high damping material; otherwise, it is determined to be a low damping material. This judgment method avoids direct contact material detection and realizes material identification based on vibration response.

[0028] Propagation feature recognition is achieved by analyzing the changes of vibration signals under different connection modes. The rigid connection at the flange connection causes small phase changes in the vibration signal, while the bolt connection produces obvious propagation time delay due to the existence of contact gap. By calculating these characteristic parameters, the propagation feature recognition result of the bearing wear signal in the entire system is obtained.

[0029] S102, according to the material type identification result, the propagation process of the bearing wear signal from the source device to the adjacent device is analyzed to obtain the propagation path analysis data.

[0030] According to the difference of the damping coefficient in the material type identification result, the vibration signal propagation relationship diagram between the device nodes is constructed, the material properties and connection modes of each node are labeled, the acoustic impedance value is calculated by taking the square root of the product of the material density and the elastic modulus, and the impedance mismatch coefficient is calculated by calculating the ratio of the acoustic impedance between adjacent nodes. According to the propagation theory of vibration waves at different medium interfaces, the reflection coefficient and transmission ratio of the vibration signal at the flange connection are calculated using the impedance mismatch coefficient. If the reflection coefficient exceeds the preset reflection threshold, it is determined that there is a partial reflection phenomenon of the vibration signal in the path, and the reflection coefficient and transmission ratio are recorded as the interface propagation feature. By the frequency component change of the vibration signal and the interface propagation feature, the coupling relationship between the vibration signal and the pipe natural frequency on different propagation paths is identified. When the vibration frequency is close to the pipe natural frequency, resonance phenomenon occurs, and the resonance frequency point and the corresponding amplitude change are recorded to construct a propagation path feature matrix containing the impedance mismatch coefficient, the interface propagation feature and the resonance frequency. According to the vibration energy change of each path in the propagation path feature matrix, the least square method is used to fit the amplitude attenuation curve of the vibration signal along different material pipes to determine the main propagation path and the secondary propagation path from the source device to each adjacent device, and the complete propagation path analysis data is obtained.

[0031] In one embodiment, the vibration propagation path analysis of the multi-energy system cluster is based on the material type identification result, and by establishing the physical connection relationship and material property mapping between the device nodes, a complete vibration signal propagation network is formed to realize the quantitative analysis of the propagation characteristics from the source device to each adjacent device.

[0032] Specifically, the calculation of acoustic impedance involves the extraction of basic physical parameters of materials. For steel pipes, its density is about 7850 kg / m 3, the acoustic impedance value is obtained by taking the square root of the product of density and elastic modulus, and the calculated result is about 4.07*10^7 kg / (m 2 ·s). While the density of aluminum alloy pipe is about 2700 kg / m 3 , the elastic modulus is about 70 GPa, and the corresponding acoustic impedance value is about 1.37*10^7 kg / (m 2 ·s). The impedance mismatch coefficient is determined by the ratio of the acoustic impedance of adjacent nodes. When the steel pipe connects the aluminum alloy equipment, the impedance mismatch coefficient is about 2.97. This significant impedance difference leads to a significant energy redistribution of the vibration signal at the material interface. The greater the impedance mismatch, the more obvious the reflection of the vibration signal at the interface, and the lower the proportion of energy transmitted to the downstream equipment.

[0033] It should be noted that the calculation of the reflection coefficient and the transmission ratio is based on the propagation theory of vibration waves at the interface of different media. The reflection coefficient R is calculated by the formula R=(Z2-Z1) / (Z2+Z1), where Z1 and Z2 are the acoustic impedance values of the two materials, respectively. When the absolute value of the reflection coefficient exceeds the preset reflection threshold of 0.3, it indicates that there is significant vibration reflection at the interface. The transmission ratio T is calculated by T=2Z2 / (Z2+Z1), which reflects the efficiency of vibration energy transmission to the downstream.

[0034] The recording of the interface propagation characteristics includes multiple dimensional parameters, in addition to the reflection coefficient and the transmission ratio, including the phase change and time delay of the vibration signal when passing through the interface, which together constitute the complete propagation characteristic description of the interface.

[0035] The coupling relationship between vibration and pipe natural frequency is identified by spectrum matching. The natural frequency of the pipe depends on its length, diameter, wall thickness and material properties. When the vibration frequency component generated by bearing wear approaches the natural frequency of a certain order of the pipe, it will excite the resonance response of the pipe.

[0036] For example, a steel pipe with a length of 3 meters and a diameter of 100 mm has a first-order bending natural frequency of about 150 Hz and a second-order natural frequency of about 600 Hz. When the bearing fault frequency approaches 600 Hz, the pipe produces a second-order resonance, and the vibration amplitude can be amplified by 3-5 times. The accurate identification of the resonance frequency point is achieved by peak detection of the frequency-sweep excitation or running vibration spectrum, and the amplification multiple corresponding to each resonance frequency is recorded to form a mapping relationship between frequency and amplification multiple.

[0037] Exemplarily, the propagation path characteristic matrix stores the propagation parameters of each path in a multi-dimensional array structure. The rows of the matrix correspond to different propagation paths, and the columns contain parameters such as impedance mismatch coefficient, reflection coefficient, transmission ratio, resonance frequency, and amplification factor. This matrix provides a quantitative basis for subsequent path optimization and fault tracing. Further, the amplitude decay curve of the vibration signal along the pipeline of different materials is obtained by least squares fitting. The vibration amplitude data of multiple measurement points is collected, with the propagation distance as the independent variable and the amplitude as the dependent variable, to fit the exponential decay function A(x) = A0 x exp(-αx), where α is the attenuation coefficient, which is related to the material damping characteristics.

[0038] It can be understood that by comparing the attenuation coefficients and energy transmission efficiencies of various propagation paths, the main propagation path and the secondary propagation path can be distinguished. The main path usually has a lower impedance mismatch coefficient and a smaller attenuation coefficient, and carries most of the vibration energy; the secondary path has a lower propagation efficiency due to material differences or loose connections and other factors. Complete propagation path analysis data includes path topological structure, propagation characteristic parameters, and energy distribution ratio.

[0039] S103, according to the propagation path analysis data, identify the signal contribution degree of each device node corresponding to the fault feature difference, and judge the origin point of the potential fault feature.

[0040] According to the propagation path analysis data, the transmission efficiency of the vibration signal from the source point to each device node is calculated, the energy distribution matrix of each device node is constructed by determining the energy distribution coefficient through the ratio of the vibration energy received by each node to the total input energy. Using the energy distribution matrix, the frequency offset and amplitude change rate of the vibration signal of each node are extracted. If the frequency offset of a certain node exceeds the preset offset threshold and the amplitude change rate is lower than the preset attenuation threshold, the node is determined as a candidate point of the fault source, and the frequency feature and amplitude feature of the candidate point are recorded. The mutual correlation coefficient of the frequency feature and the amplitude feature of the candidate point is calculated to obtain the time delay relationship between different node signals. According to the vibration propagation speed and the time delay relationship, the signal occurrence time is calculated in reverse, and the actual generation order of the vibration signal of each candidate point is determined. According to the actual generation order, the node that generates the vibration signal earliest is given the highest weight value, and the weights of other nodes are distributed in descending order of time, and the comprehensive contribution degree of each node is calculated according to the energy distribution coefficient. When the comprehensive contribution degree of a certain device node exceeds the preset contribution threshold, the device is judged as the origin point of the potential fault feature.

[0041] In one embodiment, the fault source positioning of the multi-energy system cluster is based on the comprehensive analysis of energy propagation characteristics and signal timing relationship, and the precise positioning of the single fault origin point is realized by quantifying the signal contribution degree of each device node from multiple vibration response points in reverse.

[0042] Specifically, the construction of the energy distribution matrix involves the step-by-step attenuation process of vibration energy on the propagation path. When the vibration energy generated by bearing wear propagates out from the source equipment, the initial energy is set as E0. After passing through the first steel pipe, the energy reaching the first connection node is E1 = E0 × exp(-α1 × L1) according to the attenuation coefficient α1 and the propagation distance L1. At the flange connection interface, the energy transmitted to the downstream is E1 × (1-R1) according to the reflection ratio R1. When propagating to the second equipment node, the attenuation coefficient α2 and the distance L2 of the aluminum alloy pipe also need to be considered, and the energy finally received by the node is E2 = E1 × (1-R1) × exp(-α2 × L2). By calculating the ratio of the energy received by each node to the total input energy of the system, the energy distribution coefficient is obtained.

[0043] For example, the compressor outputs 100 units of energy as a source point, after propagation, the pump group receives 35 units, the heat exchanger receives 20 units, and other nodes receive a total of 15 units, and the remaining 30 units are dissipated in the propagation process, so the energy distribution coefficients of each node are 0.35, 0.20 and 0.15 respectively.

[0044] It should be noted that the extraction of the frequency offset is achieved by comparing the main frequency components of the vibration signals of each node. Under normal circumstances, the characteristic frequency of the bearing inner ring fault remains constant, but the performance on different material equipment is different. The amplitude change rate is obtained by calculating the change slope of the vibration amplitude in the adjacent time window, reflecting the time-varying characteristics of the vibration intensity.

[0045] The determination of the fault source candidate point adopts a double screening mechanism. The first screening is based on the frequency offset. When the main frequency of a node deviates from the original fault frequency by less than 5% of the preset offset threshold, it indicates that the node well maintains the fault characteristics. The second screening is based on the amplitude change rate. When the amplitude attenuation rate is lower than the preset attenuation threshold of 30%, it indicates that the vibration energy loss of the node is small, which may be close to the fault source. The nodes that meet both conditions are marked as candidate points.

[0046] The calculation of the cross-correlation coefficient uses the normalized cross-correlation function to eliminate the influence of signal amplitude difference, and focuses on the similarity of signal waveform and time delay relationship.

[0047] Exemplarily, the acquisition and reverse calculation process of the time delay relationship plays a key role. When the compressor bearing fault vibration signal propagates in the steel pipe at a speed of 3000 m / s, it takes 1 ms to reach the first monitoring point 3 meters away, and 2 ms to reach the second monitoring point 6 meters away. Through cross-correlation analysis, it is found that the signal of the second monitoring point is delayed by 1 ms relative to the first monitoring point, which is consistent with the theoretical propagation time. In reverse calculation, according to the arrival time of the vibration signal recorded by each monitoring point, the corresponding propagation delay is subtracted, and the actual generation time of the vibration signal of each candidate point is restored.

[0048] For example, if the signal arrival time of three candidate points are t1 = 10 ms, t2 = 11 ms, t3 = 12 ms, and the actual generation time considering the propagation delay is t1' = 9 ms, t2' = 10.5 ms, t3' = 11.8 ms, the first candidate point generates vibration earliest, and is likely to be the fault source. Further, the weight distribution mechanism is set according to the time sequence of signal generation. The node that generates the vibration signal earliest is assigned a weight of 1.0, and the weights of subsequent nodes in time sequence are sequentially decreased, such as 0.7 for the second node and 0.4 for the third node, to ensure that time priority plays a dominant role in fault location.

[0049] It can be understood that the calculation of the comprehensive contribution degree combines the energy distribution coefficient and the time sequence weight in two dimensions. The comprehensive contribution degree of a certain node is equal to the product of its energy distribution coefficient and time sequence weight.

[0050] For example, the energy distribution coefficient of a certain candidate point is 0.35, and the time sequence weight is 1.0, so the comprehensive contribution degree is 0.35; the energy distribution coefficient of another candidate point is 0.40, but the time sequence weight is only 0.4, so the comprehensive contribution degree is 0.16. When the comprehensive contribution degree exceeds the preset contribution threshold of 0.3, the node is determined to be the origin point of the potential fault feature, and accurate fault location based on multi-dimensional information fusion is achieved.

[0051] S104, a vibration frequency subset corresponding to the origin point of the fault feature is obtained, and the fault feature is extracted after feature correction.

[0052] From the original vibration signal of the fault feature origin point, the frequency components in the preset frequency band are extracted through short-time Fourier transform, the fundamental frequency and its multiple frequency components are identified, the theoretical fault frequencies of the inner ring, the outer ring and the rolling elements are calculated according to the bearing rotating speed, the number of rolling elements and the contact angle, the frequency components with a deviation within a preset tolerance range from the theoretical value are screened out, and a vibration frequency subset is constructed. For each frequency component in the vibration frequency subset, the sideband frequency interval on both sides of the main frequency is identified, the ratio of the sideband amplitude to the main frequency amplitude is calculated as the modulation depth, and if the modulation depth exceeds a preset modulation threshold, the instantaneous phase is obtained through Hilbert transform, the phase delay is calculated and compensated according to the propagation path length and wave speed, and the corrected frequency feature set is obtained. According to the corrected frequency feature set, the peak amplitude of each frequency component is calculated as a peak frequency feature, the energy accumulation value in a specific frequency band is calculated as a frequency band energy feature, the spectral kurtosis feature is calculated through the ratio of the fourth-order cumulant to the square of the second-order cumulant, and the peak frequency feature, the frequency band energy feature and the spectral kurtosis feature are combined to construct a fault feature vector, and the fault feature of bearing wear is obtained.

[0053] In one embodiment, the fault feature extraction is based on multi-dimensional analysis of the original vibration signal from the origin point, and the accurate identification of bearing wear fault is realized through frequency domain transformation, feature correction and statistical quantity calculation.

[0054] Specifically, the construction of the vibration frequency subset needs to combine the structural parameters and operating conditions of the bearing. For a deep groove ball bearing with an inner diameter of 40 mm, an outer diameter of 90 mm, and containing 8 rolling bodies, when the rotating speed is 1500 r / min, the inner ring fault frequency is calculated as fi = 0.5 x n x z x (1 + d / D x cos a), where n is the rotating speed, z is the number of rolling bodies, d is the diameter of the rolling body, D is the pitch diameter, and a is the contact angle. Through short-time Fourier transform, a 256-point Hanning window is used with a 50% overlap rate to obtain a time-frequency spectrum. In the time-frequency spectrum, 120 Hz, 240 Hz, 360 Hz and other multiple frequency components are identified, which are consistent with the theoretically calculated inner ring fault frequency 119.7 Hz and its multiples, with a deviation within 0.3%, and these frequency components constitute the vibration frequency subset.

[0055] It should be noted that the appearance of the sideband frequency reflects the modulation phenomenon caused by the fault. When there is local peeling on the inner ring of the bearing, an impact will be generated every time the rolling body passes through the damaged position, and this periodic impact will modulate the natural vibration of the bearing. The sideband frequency interval is equal to the shaft rotating frequency 25 Hz, and the modulation depth is determined by the ratio of the sideband amplitude to the main frequency amplitude. When this ratio exceeds the preset threshold value of 0.15, it indicates that there is a significant modulation phenomenon.

[0056] Phase compensation is realized through Hilbert transform. The vibration signal propagates from the fault source to the sensor position, which will produce a phase delay, and the delay amount is φ = 2πf x L / c, where f is the frequency, L is the propagation distance, and c is the wave speed. For a propagation distance of 3 meters and a wave speed of 3000 m / s, the phase delay of the 120 Hz frequency component is 0.72 radians. By inversely compensating the phase delay, the original phase characteristics of the signal are restored.

[0057] The fault feature vector contains three key indicators. The peak frequency feature reflects the dominant frequency component of the fault; the frequency band energy feature is calculated by integrating the total energy in the 100-500 Hz frequency band, which reflects the severity of the fault; the spectral kurtosis is calculated by K = E[(x-μ)^4] / σ^4-3, which is particularly sensitive to impact faults and is close to 0 under normal conditions and significantly increases when there is a fault.

[0058] For example, after feature extraction, the feature vector of the bearing inner ring with slight wear shows a main frequency of 120 Hz, a frequency band energy of 0.8, and a spectral kurtosis of 3.5, while the feature vector of the bearing inner ring with severe wear shows a main frequency of 120 Hz accompanied by rich harmonics, a frequency band energy of 2.5, and a spectral kurtosis of more than 8, realizing the quantitative evaluation of the fault degree.

[0059] S105, confirming the fault source device position in the multi-energy system cluster according to the extracted fault feature.

[0060] According to the extracted fault feature vector, the similarity of each device node in the multi-energy system cluster is matched with the standard fault feature, the Euclidean distance value is obtained by calculating the square sum of the difference of each dimension of the feature vector, and if the Euclidean distance value is less than the preset distance threshold, the node is marked as a suspected fault source. Through the device number of the suspected fault source, the correspondence table of the device number and the physical position is queried to obtain the installation position coordinates of the device in the multi-energy system cluster, and the specific position of the fault source device is confirmed.

[0061] In an embodiment, the confirmation of the fault source device position is realized based on two links of feature matching and position mapping. Each device node in the multi-energy system cluster has a unique number and a corresponding physical installation position, and the accurate positioning is completed through feature similarity determination and position query.

[0062] Specifically, the similarity matching of the fault feature vector and the standard fault feature adopts Euclidean distance measurement. Let the extracted fault feature vector be V1=[f1,e1,k1], wherein f1 is the peak frequency 120Hz, e1 is the frequency band energy 2.5, and k1 is the spectral kurtosis 8; the standard fault feature vector is V2=[f2,e2,k2], and the corresponding values are

[0063] [119Hz,2.3,7.8]. The Euclidean distance is calculated as d=sqrt[(120-119) 2 +(2.5-2.3) 2 +(8-7.8) 2 ]. When the distance value is less than the preset distance threshold 2.0, it is determined that the matching is successful, and the corresponding device node is marked as a suspected fault source.

[0064] It should be noted that the correspondence table of the device number and the physical position is established when the system is initialized.

[0065] For example, the number CP-001 corresponds to the compressor unit No. 1 located in the A area, the 3rd row, and the 2nd column, and the number PM-005 corresponds to the delivery pump group No. 5 located in the B area, the 1st row, and the 4th column. By querying the correspondence table, the accurate position of the fault source device is directly obtained. The confirmation process realizes the mapping from the abstract feature space to the specific physical space, so that the maintenance personnel can quickly locate the fault device for repair.

[0066] S106, through the confirmed fault source device position, a path optimization process is performed to generate an intelligent diagnosis scheme for the bearing wear signal in the multi-energy system cluster under the condition of different pipe materials, which increases the difficulty of fault source identification.

[0067] According to the confirmed fault source device position, a signal attenuation compensation function from the fault source to each monitoring point is constructed in combination with the vibration signal propagation characteristics, the compensation coefficient of the steel pipeline is set as the pipeline length multiplied by a low damping coefficient, the compensation coefficient of the aluminum alloy pipeline is set as the pipeline length multiplied by a high damping coefficient, and a material difference compensation parameter set is obtained. The material difference compensation parameter set is used to reconfigure the vibration monitoring network, the sensor arrangement density is increased at the material node with a damping coefficient exceeding a preset damping threshold, the signal amplification multiple of each monitoring point is adjusted according to the compensation parameter set, if the cumulative attenuation of a certain path exceeds a preset attenuation threshold, a monitoring point is additionally arranged in the middle of the path, and an optimized monitoring layout is formed. Through the optimized monitoring layout, a fault feature recognition threshold corresponding to each propagation path is established according to the frequency offset law and the amplitude attenuation characteristics under different material combinations, when a specific material sequence is detected, the corresponding recognition parameter set is automatically called, and a fault judgment rule library is constructed. According to the fault judgment rule library, the diagnosis results of multiple monitoring points are fused by using a decision tree method, the monitoring points closer to the fault source are given higher weight values, the fault type and position are comprehensively judged, and an intelligent diagnosis scheme including the optimized configuration of monitoring points, material compensation parameters and recognition thresholds is formed.

[0068] In an embodiment, the intelligent diagnosis scheme compensates and corrects the influence of material differences through path optimization processing. The scheme comprehensively considers the influence of different material pipelines on vibration signal propagation in the multi-energy system cluster, and establishes a complete diagnosis system from signal compensation, monitoring optimization to intelligent judgment.

[0069] Specifically, the construction of the signal attenuation compensation function is based on the physical property difference of the material. For a steel pipeline, the low damping property of the steel makes the vibration signal attenuate slowly, and the compensation coefficient C_steel=L x a_steel, where L is the pipeline length and a_steel is the damping coefficient of the steel material, and the typical value is 0.02-0.05 / m. The aluminum alloy pipeline has a higher internal friction property, and the damping coefficient a_aluminum reaches 0.06-0.12 / m, and the corresponding compensation coefficient C_aluminum=L x a_aluminum. When the vibration signal propagates through a 3-meter steel pipe and then through a 2-meter aluminum pipe, the cumulative compensation coefficient is C_total=3 x 0.03+2 x 0.09. This means that the received signal needs to be multiplied by a compensation factor of exp(0.27) to restore the original signal strength. By establishing a compensation parameter lookup table of different material sequences, fast compensation calculation of complex propagation paths is realized. The material difference compensation parameter set contains all possible material combinations in the system and the corresponding compensation coefficients, forming a complete compensation system.

[0070] It should be noted that the reconfiguration of the monitoring network adopts a non-uniform arrangement strategy. In high-damping material nodes, such as aluminum alloy pump shells or rubber shock absorber joints, the sensor arrangement density is increased to 1.5-2 times the regular density. This dense arrangement can capture weak signals caused by material damping and avoid missing fault features.

[0071] The adjustment of sensor density follows the principle of signal attenuation degree. When the signal attenuation of a certain section of the pipeline exceeds 50%, an additional monitoring point is added at the midpoint of the section; when the attenuation exceeds 70%, two monitoring points are added according to the three-section principle. This dynamic adjustment ensures that the signal always maintains a detectable intensity level during propagation.

[0072] The signal amplification factor is automatically adjusted according to the attenuation coefficient in the compensation parameter set. The monitoring point of the steel pipe 5 meters away from the fault source is set to an amplification factor of 2, while the monitoring point of the aluminum alloy section 10 meters away is set to an amplification factor of 5, achieving signal intensity equalization.

[0073] Exemplarily, the construction process of the fault determination rule library involves multi-dimensional parameter mapping. The rule library stores the frequency offset laws under different material sequences, such as the "steel-aluminum-steel" sequence, which causes a 2-4 kHz frequency band to shift to 1.5-3 kHz, with an offset of about 25%; the "steel-rubber-aluminum" sequence has an offset of up to 40%. The amplitude attenuation characteristics are also recorded, with a pure steel pipe path having an amplitude attenuation rate of 10% / m and a mixed path containing an aluminum alloy section having an attenuation rate of 25% / m. When the system detects a specific material sequence, it automatically extracts the corresponding identification parameter set from the rule library, including the frequency offset compensation value, the amplitude attenuation compensation coefficient, and the fault feature threshold adjustment amount. These parameters are dynamically loaded according to the actual propagation path, enabling adaptive diagnosis under different working conditions. Further, the decision tree method uses a distance weighting strategy when integrating the results of multiple monitoring points. The weight of the monitoring point closest to the fault source is set to 1.0, and the weight decreases by 0.2 for every additional 3 meters of propagation distance. Through this weighting mechanism, the diagnosis results of the near-source monitoring points dominate in the final determination.

[0074] It can be understood that the intelligent diagnosis scheme formed contains three core components: the monitoring point optimization configuration table, the material compensation parameter library, and the adaptive identification threshold set. This scheme can automatically adjust the diagnosis parameters according to the actual material distribution and equipment layout, overcoming the interference of material differences on fault identification and achieving accurate fault location in complex material environments.

[0075] The above only describes the preferred embodiments of the present application. It should be noted that those skilled in the art can make several improvements and supplements without departing from the principles of the present application, and these improvements and supplements should also be considered within the scope of protection of the present application.

Claims

1. A multi-energy system cluster fault intelligent diagnosis analysis method, characterized in that, Comprise: S101, collect bearing wear signal corresponding original vibration signal data from multi-energy system cluster device node, extract specific frequency vibration and amplitude reduction process characteristics, and perform material type identification; S102, according to the material type identification result, analyze the propagation process of bearing wear signal from the source device to the adjacent device, and obtain the propagation path analysis data; S103, according to the propagation path analysis data, identify the signal contribution degree of each device node corresponding to the fault feature difference, and judge the origin point of the potential fault feature; S104, obtain the vibration frequency subset corresponding to the origin point of the fault feature, extract the fault feature after feature correction; S105, according to the extracted fault feature, confirm the fault source device position in the multi-energy system cluster; S106, through the confirmed fault source device position, carry out path optimization processing, and generate intelligent diagnosis scheme of bearing wear signal in multi-energy system cluster under the condition of pipe material difference leading to the increase of fault source identification difficulty. 2.The method of claim 1, wherein, Step S101 includes: Obtain the original vibration signal from the acceleration sensor of each device node in the multi-energy system cluster, perform fast Fourier transform on the original vibration signal to obtain frequency spectrum data, identify the vibration peak and its corresponding amplitude characteristic in the preset high frequency band, calculate the attenuation ratio of the amplitude characteristic between different collection points to obtain the initial attenuation characteristic. According to the nodes with amplitude reduction rate exceeding the first preset threshold and the nodes with frequency offset exceeding the second preset threshold in the initial attenuation characteristic, a material damping coefficient mapping table is constructed, the material damping coefficient mapping table is constructed, the low damping coefficient range of steel pipe and the high damping coefficient range of aluminum alloy are compared, and the material category of each node device shell and connecting pipe is judged. The material category and the resonance frequency distribution in the frequency spectrum data are used to calculate the phase change amount of the vibration signal at the flange connection and the propagation time delay of the vibration signal at the bolt connection, and the propagation characteristic identification result of the bearing wear signal at different material nodes is obtained. 3.The method of claim 1, wherein, Step S102 includes: According to the difference of the damping coefficient in the material type identification result, a vibration signal propagation relationship diagram between device nodes is constructed, the material property and the connection mode of each node are labeled, the acoustic impedance value is calculated by taking the square root of the product of the material density and the elastic modulus, and the impedance mismatch coefficient is obtained by calculating the ratio of the acoustic impedance between adjacent nodes. According to the propagation theory of vibration waves at different medium interfaces, the reflection coefficient and the transmission ratio of the vibration signal at the flange connection are calculated by using the impedance mismatch coefficient. If the reflection coefficient exceeds the preset reflection threshold, it is determined that there is a partial reflection phenomenon of the vibration signal in the path, and the reflection coefficient and the transmission ratio are recorded as the interface propagation characteristics. According to the interface propagation characteristics and the frequency component change of the vibration signal, the coupling relationship between the vibration signal and the natural frequency of the pipeline on different propagation paths is identified. When the vibration frequency is close to the natural frequency of the pipeline, resonance phenomenon occurs, and the resonance frequency point and the corresponding amplitude change are recorded to construct a propagation path characteristic matrix containing the impedance mismatch coefficient, the interface propagation characteristics and the resonance frequency. According to the vibration energy change of each path in the propagation path characteristic matrix, the least square method is used to fit the amplitude attenuation curve of the vibration signal along the pipeline of different materials to determine the main propagation path and the secondary propagation path from the source device to each adjacent device, and to obtain complete propagation path analysis data.

4. The method of claim 1, wherein, Step S103 comprises: According to the propagation path analysis data, the transmission efficiency of the vibration signal from the source point to each device node is calculated, the energy distribution coefficient is determined by the ratio of the received vibration energy of each node to the total input energy, and the energy distribution matrix of each device node is constructed. According to the energy distribution matrix, the frequency offset and the amplitude change rate of the vibration signal of each node are extracted. If the frequency offset of a certain node exceeds the preset offset threshold and the amplitude change rate is lower than the preset attenuation threshold, it is determined that the node is a candidate point of the fault source, and the frequency characteristics and the amplitude characteristics of the candidate point are recorded. The time delay relationship between signals of different nodes is obtained by calculating the cross-correlation coefficient of the frequency characteristics and the amplitude characteristics of the candidate point. According to the vibration propagation speed and the time delay relationship, the occurrence time of the signal is inversely calculated to determine the actual generation order of the vibration signal of each candidate point. According to the actual generation order, the node that generates the vibration signal earliest is given the highest weight value, and the weights of other nodes are distributed in descending order of time sequence. The comprehensive contribution degree of each node is calculated by combining the energy distribution coefficient. When the comprehensive contribution degree of a certain device node exceeds the preset contribution threshold, it is determined that the device is the origin point of the potential fault characteristics.

5. The method of claim 1, wherein, Step S104 comprises: From the original vibration signal of the fault feature origin point, the frequency components in the preset frequency band are extracted by short-time Fourier transform, the fundamental frequency and its multiple components are identified, the theoretical fault frequencies of the inner ring, the outer ring and the rolling elements are calculated according to the bearing rotating speed, the number of rolling elements and the contact angle, the frequency components with a deviation within a preset tolerance range from the theoretical value are screened out, and a vibration frequency subset is constructed. For each frequency component in the vibration frequency subset, the sideband frequency interval on both sides of the main frequency is identified, the ratio of the sideband amplitude to the main frequency amplitude is calculated as the modulation depth, and if the modulation depth exceeds a preset modulation threshold, the instantaneous phase is obtained by Hilbert transform, the phase delay is calculated according to the propagation path length and wave speed and is compensated to obtain a corrected frequency feature set. According to the corrected frequency feature set, the peak amplitude of each frequency component is calculated as a peak frequency feature, the energy accumulation value in a specific frequency band is calculated as a frequency band energy feature, the spectral kurtosis feature is calculated by the ratio of the fourth-order cumulant to the square of the second-order cumulant, and the peak frequency feature, the frequency band energy feature and the spectral kurtosis feature are combined to construct a fault feature vector to obtain the fault feature of bearing wear.

6. The method of claim 1, wherein, Step S105 comprises: According to the extracted fault feature vector, similarity matching is performed with the standard fault features of each device node in the multi-energy system cluster, the Euclidean distance value is obtained by calculating the square root of the sum of squares of the difference values of each dimension of the feature vector, and if the Euclidean distance value is less than a preset distance threshold, the node is marked as a suspected fault source. According to the device number of the suspected fault source, the correspondence table of device number and physical location is queried to obtain the installation location coordinates of the device in the multi-energy system cluster, and the specific location of the fault source device is confirmed.

7. The method of claim 1, wherein, Step S106 comprises: According to the confirmed fault source device location, a signal attenuation compensation function from the fault source to each monitoring point is constructed in combination with the vibration signal propagation characteristics, the compensation coefficient of the steel pipeline is set as the pipeline length multiplied by a low damping coefficient, the compensation coefficient of the aluminum alloy pipeline is set as the pipeline length multiplied by a high damping coefficient, and a material difference compensation parameter set is obtained. The material difference compensation parameter set is used to reconfigure the vibration monitoring network, the sensor arrangement density is increased at the material nodes with a damping coefficient exceeding a preset damping threshold, the signal amplification multiple of each monitoring point is adjusted according to the compensation parameter set, and if the cumulative attenuation of a certain path exceeds a preset attenuation threshold, a monitoring point is added in the middle of the path to form an optimized monitoring layout. According to the frequency offset law and amplitude attenuation characteristics under different material combinations, the fault feature recognition threshold corresponding to each propagation path is established, the corresponding recognition parameter set is automatically called when a specific material sequence is detected, and a fault judgment rule library is constructed. According to the fault judgment rule library, the diagnosis results of multiple monitoring points are fused by using the decision tree method, the monitoring points closer to the fault source are given higher weight values, the fault type and location are comprehensively determined, and an intelligent diagnosis scheme including the optimized configuration of monitoring points, material compensation parameters and recognition threshold is formed.

Images