A method and system for early warning of abnormality in anti-surge and pressure relief valve groups with axial flow structure

Abstract

This invention discloses a method and system for early warning of coordinated anomalies in anti-surge and pressure relief valve groups for axial flow structures, relating to the field of valve group monitoring technology. The method includes the following steps: constructing a unified time baseline based on the valve group opening / closing cycle signal and the axial flow aerodynamic response signal; extracting the nonlinear frequency superposition trajectory and identifying high-order resonance peaks; generating a frequency anomaly fingerprint matrix through energy concentration intervals and dynamic slope factors; matching modal characteristics to construct a modal response phase field; and locking in the neighborhood of dangerous frequencies; further calculating the energy amplification curve and coupling risk coefficient to generate a dynamic risk curve; and dynamically shifting the valve group opening / closing frequency domain window when the risk increases, so that the operating frequency is far from the natural frequency. This invention achieves deep coupling analysis of valve group signals and aerodynamic responses through a unified time baseline and multi-scale phase difference mapping, enabling accurate identification of high-order resonances and locking in the neighborhood of dangerous frequencies, thus achieving early warning of aerodynamic instability risks.

Description

Technical Field

[0001] This invention relates to the field of valve group monitoring technology, specifically to a method and system for early warning of abnormalities in anti-surge and pressure relief valve groups with axial flow structures. Background Technology

[0002] Coordinated anomaly early warning for anti-surge pressure relief valve groups in axial flow structures refers to a technical means of monitoring and providing anomaly warnings for multiple pressure relief valve groups in devices centered around axial compressors or fans to prevent surge, a dangerous operating condition characterized by reverse oscillations and drastic flow fluctuations caused by compression instability. Its core principle is that the action of a single pressure relief valve is often insufficient to effectively address the overall aerodynamic instability risk under complex operating conditions. Therefore, it is necessary to model and dynamically analyze the opening and closing states, response timing, pressure relief volume, and the matching relationship between these factors and dynamic changes in the flow field from the perspective of "valve group coordination." When certain valves are detected to exhibit sluggish action, unbalanced response, or coordinated failure, an anomaly warning is triggered, indicating a potential surge risk or aerodynamic instability trend in the system. This mechanism allows for the early identification of potential abnormal signs before stall, sudden efficiency drops, or even structural damage occur, providing a reliable basis for proactively adjusting valve strategies, optimizing control logic, and preventing accidents.

[0003] The existing technology has the following shortcomings:

[0004] In existing technologies, anti-surge and pressure relief valve groups in axial flow structures typically rely on periodic opening and closing to stabilize the flow field. However, during dynamic frequency scanning, the opening and closing cycles of some valves are prone to superposition effects under nonlinear conditions, resulting in high-order resonant peaks. When these resonant peaks couple with the natural frequency of the axial flow structure, the local vibration intensity of the valve body is amplified sharply, which in turn induces crack propagation in the valve seat material under high-cycle fatigue. In severe cases, it may even lead to resonance failure of the entire structure due to the collective coupling of the valve group.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for early warning of abnormalities in the anti-surge and pressure relief valve group of axial flow structure, so as to solve the problems in the background art mentioned above.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for early warning of abnormality in a group of anti-surge and pressure relief valves with an axial flow structure, comprising the following steps:

[0008] A unified time baseline is constructed based on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal. Multi-scale phase difference mapping is performed on the unified time baseline. Frequency superposition trajectory under nonlinear conditions is extracted from the multi-scale phase difference mapping to identify high-order resonance peaks within a unified time frame.

[0009] The energy concentration interval is extracted by frequency superposition trajectory, and the dynamic slope factor is calculated to characterize the energy transition intensity, thereby generating a fingerprint matrix that characterizes the frequency amplification trend, and the fingerprint matrix is ​​used as a constraint condition for intrinsic frequency matching analysis.

[0010] Based on the fingerprint matrix, the fingerprint features are matched one by one with the modal characteristics of the axial flow structure, and a modal response phase field is constructed during the matching process to achieve real-time coupling analysis under unified conditions, thereby locking the dangerous frequency neighborhood that overlaps with the natural frequency.

[0011] Based on the modal response phase field, a force distribution weight is introduced into the neighborhood of the dangerous frequency, and the amplification effect of the energy coupling path is dynamically calculated to form a continuously evolving energy amplification curve, which is used to characterize the local response risk of the structure.

[0012] The coupling risk coefficient between the high-order resonant peak and the natural frequency is calculated by using the energy amplification curve, and a dynamic risk curve is generated by combining the frequency evolution trend on a unified time baseline, so as to achieve quantitative identification and real-time early warning of coupling risk.

[0013] Based on the risk dynamic curve, when the coupling risk between the high-order resonance peak and the natural frequency is identified, the spectral characteristics of the valve group opening and closing cycle signal are dynamically monitored. Under the condition that the resonance factor continues to increase, the frequency domain window of the valve group opening and closing cycle signal is shifted, so that the overall operating frequency of the valve group is pushed away from the natural frequency neighborhood.

[0014] Preferably, extracting the frequency superposition trajectory under nonlinear conditions includes the following steps:

[0015] The valve group opening and closing cycle signal and axial flow aerodynamic response signal are acquired and synchronized under a unified reference clock through time normalization and cubic spline interpolation. Random noise is eliminated by wavelet threshold denoising algorithm to form a unified time baseline.

[0016] On a unified time baseline, continuous wavelet transforms were performed on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal to obtain the instantaneous phase at multiple scales, and the phase difference was calculated to construct a multi-scale phase difference mapping.

[0017] Multi-scale phase difference mapping is transformed into a three-dimensional surface model, and contour lines are drawn on the surface to extract the frequency superposition trajectory under nonlinear conditions. A nonlinear energy density calculation method is introduced on the frequency superposition trajectory, and the high-order resonance peak is identified by combining the phase difference convergence phenomenon.

[0018] Preferably, generating a fingerprint matrix characterizing the frequency amplification trend includes the following steps:

[0019] After obtaining the frequency superposition trajectory, instantaneous energy is calculated on the frequency superposition trajectory, and the energy distribution curve is obtained by smoothing through sliding window integration, and the energy concentration interval is extracted.

[0020] The first derivative of the energy distribution curve is calculated within the energy concentration interval, and the instantaneous growth rate is calculated using the central difference method. The dynamic slope factor is then standardized to characterize the energy transition intensity.

[0021] Using the time series of the energy concentration interval as the horizontal axis and the frequency series of the frequency superposition trajectory as the vertical axis, and mapping the dynamic slope factor value to the intensity identifier, a two-dimensional matrix is ​​established to generate a fingerprint matrix.

[0022] Preferably, constructing the modal response phase field includes the following steps:

[0023] Fingerprint features are extracted based on the obtained fingerprint matrix, and the center frequency, frequency bandwidth and energy intensity of the fingerprint features are matched with the natural frequency, mode shape and damping ratio of the axial flow structure. When the frequency difference is less than the preset frequency difference threshold and the energy intensity exceeds the background energy level, the corresponding fingerprint features are determined to be related to the modal characteristics.

[0024] After matching is completed, the dynamic slope factor and energy intensity are mapped to a three-dimensional distribution using time and space modal morphology as coordinates, and the modal response phase field is constructed. The instantaneous phase is extracted by performing Hilbert transform on the energy envelope signal, thus forming a time-space-phase mapping relationship.

[0025] In the modal response phase field, phase-locked regions are identified, and when the energy intensity continuously exceeds three times the background energy level and the phase difference remains within the phase difference threshold range, the continuous frequency range corresponding to the phase-locked region is determined as the danger frequency neighborhood and marked on the unified time baseline.

[0026] Preferably, the background energy level is the average energy intensity obtained from the valve group opening and closing cycle signal and the axial flow aerodynamic response signal on a unified time baseline under the stable operating conditions of the axial flow structure.

[0027] Preferably, based on the modal response phase field, a force distribution weight is introduced into the neighborhood of the dangerous frequency, and the amplification effect of the energy coupling path is dynamically calculated to form a continuously evolving energy amplification curve, including the following steps:

[0028] Based on the phase field of the modal response, a force distribution weight is introduced into the neighborhood of the dangerous frequency, and the relative displacement amplitude of each discrete point in the mode shape is converted into a weight value and weighted and superimposed on the time series.

[0029] After introducing the force distribution weights, the phase difference sequence and energy intensity sequence in the neighborhood of the dangerous frequency are weighted, and the propagation direction of the energy coupling path is obtained by calculating the phase gradient.

[0030] The ratio of weighted energy intensity to reference energy intensity is calculated at discrete points along the energy coupling path to obtain the instantaneous energy amplification coefficient and identify the energy amplification interval. The energy amplification interval is integrated and combined with the phase difference evolution to form a dynamic amplification effect curve.

[0031] The dynamic amplification effect curves are spliced ​​together in chronological order to form an energy amplification curve, and then labeled with intensity distribution weights to characterize the overall risk evolution trend and local response risk.

[0032] Preferably, generating a risk dynamic curve by combining the frequency evolution trend on a unified time baseline includes the following steps:

[0033] After constructing the energy amplification curve, the characteristic parameters of the high-order resonance peak are extracted and compared with the natural frequency. The coupling risk coefficient is obtained by weighted fusion of frequency proximity and energy amplification factor.

[0034] After obtaining the coupling risk coefficient, the frequency evolution trajectory under the unified time baseline is synchronized with the energy amplification curve, and the coupling risk coefficient is recalculated at each time point so that the coupling risk coefficient continuously evolves into a risk dynamic curve over time. The risk dynamic curve is then weighted and adjusted in combination with the slope of the frequency evolution trend.

[0035] After the risk dynamic curve is generated, the risk level is determined based on the preset multi-level thresholds. When the risk dynamic curve exceeds the danger threshold, an early warning is triggered, and the corresponding time interval and frequency interval are output to characterize the potential danger frequency neighborhood.

[0036] Preferably, when a coupling risk is identified between the higher-order resonant peak and the natural frequency, pushing the overall operating frequency of the valve group away from the natural frequency neighborhood includes the following steps:

[0037] Based on the risk dynamic curve, dangerous frequency neighborhoods are identified and fast Fourier transform is performed on the valve group opening and closing cycle signal to extract spectral features and compare them with dangerous frequency neighborhoods to determine whether there is overlap.

[0038] After identifying the overlap between the spectral features and the neighborhood of the dangerous frequency, the spectral features are dynamically monitored and the resonance factor is calculated. When the resonance factor shows a continuous increasing trend after smoothing, the risk is determined to change from a potential state to an actual state.

[0039] After the risk is determined to be in a real state, the frequency domain window of the valve group opening and closing cycle signal is shifted. By adjusting the opening and closing cycle parameters or phase difference distribution, the dominant frequency is gradually moved away from the natural frequency, and the risk dynamic curve and spectrum characteristics are monitored in real time during the shift process.

[0040] After the migration is completed, the spectral characteristics are recalculated and combined with the risk dynamic curve to confirm whether the dominant frequency has moved away from the inherent frequency and the coupling risk coefficient has decreased, thus forming a closed-loop risk avoidance mechanism.

[0041] The operating frequency refers to the set frequency of the valve group opening and closing cycle signal, which is usually adjusted through the valve group opening and closing cycle parameters.

[0042] The dominant frequency refers to the main frequency component in the system response obtained through spectrum analysis, representing the most influential frequency in the system.

[0043] The anti-surge and pressure relief valve group collaborative anomaly early warning system of axial flow structure includes a time baseline construction module, a fingerprint matrix generation module, a modal phase field construction module, an energy amplification curve formation module, a risk dynamic curve generation module, and a frequency domain window migration module;

[0044] The time baseline construction module constructs a unified time baseline based on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal, and performs multi-scale phase difference mapping on the unified time baseline. It extracts the frequency superposition trajectory under nonlinear conditions from the multi-scale phase difference mapping and identifies the high-order resonance peak.

[0045] The fingerprint matrix generation module extracts the energy concentration interval through frequency superposition trajectory and calculates the dynamic slope factor to characterize the energy transition intensity, thereby generating a fingerprint matrix that characterizes the frequency amplification trend.

[0046] The modal phase field construction module, based on the fingerprint matrix, matches the fingerprint features in it with the modal characteristics of the axial flow structure one by one, and constructs the modal response phase field during the matching process to achieve real-time coupling analysis under unified conditions and lock the dangerous frequency neighborhood that overlaps with the natural frequency.

[0047] The energy amplification curve forming module, based on the modal response phase field, introduces force distribution weights into the neighborhood of the danger frequency and dynamically calculates the amplification effect of the energy coupling path to form a continuously evolving energy amplification curve;

[0048] The risk dynamic curve generation module calculates the coupling risk coefficient between the higher-order resonant peak and the natural frequency through the energy amplification curve, and generates a risk dynamic curve by combining the frequency evolution trend on a unified time baseline.

[0049] The frequency domain window migration module, based on the risk dynamic curve, dynamically monitors the spectral characteristics of the valve group opening and closing cycle signal when it identifies a coupling risk between the high-order resonance peak and the natural frequency. Under the condition that the resonance factor continues to increase, it migrates the frequency domain window of the valve group opening and closing cycle signal, so that the overall operating frequency of the valve group is pushed away from the natural frequency neighborhood.

[0050] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0051] This invention performs multi-scale phase difference mapping on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal on a unified time baseline, and constructs a fingerprint matrix by combining the frequency superposition trajectory. Then, the fingerprint matrix is ​​matched with the modal characteristics one by one. This can accurately identify high-order resonance peaks and lock the neighborhood of dangerous frequencies, thereby clarifying the coupling relationship between aerodynamic disturbances and structural modes. This allows potential abnormal risks to be detected in the early stages and ensures the ability to provide early warning of aerodynamic instability trends under complex operating conditions.

[0052] This invention introduces force distribution weights into the modal response phase field, dynamically calculates the energy coupling path and generates an energy amplification curve, and further realizes the active migration of the valve group's opening and closing frequency domain window under the drive of the risk dynamic curve, so that the overall operating frequency of the valve group is far away from the natural frequency. This enables rapid avoidance measures to be taken after risk identification, effectively reducing the risk of local structural response and significantly improving the operational stability and safety of the device. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0054] Figure 1 This is a flowchart of the method for early warning of abnormality in the anti-surge pressure relief valve group of the axial flow structure of the present invention;

[0055] Figure 2 This is a flowchart of the method for generating a fingerprint matrix based on frequency superposition trajectory according to the present invention;

[0056] Figure 3 Here is a flowchart of the method for constructing the modal response phase field according to the present invention;

[0057] Figure 4 This is a schematic diagram of the module of the anti-surge and pressure relief valve group collaborative abnormal early warning system of the axial flow structure of the present invention. Detailed Implementation

[0058] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0059] This invention provides, for example Figures 1 to 3 The method for early warning of abnormality in the anti-surge and pressure relief valve group of the axial flow structure shown includes the following steps:

[0060] S101. Construct a unified time baseline based on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal, and perform multi-scale phase difference mapping on the unified time baseline. Extract the frequency superposition trajectory under nonlinear conditions from the multi-scale phase difference mapping in order to identify high-order resonance peaks within the unified time frame.

[0061] The specific steps for extracting the frequency superposition trajectory under nonlinear conditions are as follows:

[0062] First, the valve group opening / closing cycle signal and the axial flow aerodynamic response signal are acquired, and a unified time baseline is constructed based on these signals. In specific operation, the valve group opening / closing cycle signal is obtained through continuous sampling of the opening / closing status sensors of the anti-surge and pressure relief valves. These sensors include a position displacement sensor and a valve stem motion current sensor. The former records the specific opening angle of the valve disc during the opening and closing process, while the latter captures the driving force fluctuations of the valve during opening and closing. Through joint sampling by these two types of sensors, the time interval, duration, amplitude, and flow disturbance fluctuations caused by the valve group opening and closing can be obtained. The axial flow aerodynamic response signal is acquired in real time using high-frequency pressure sensors and hot-wire anemometers deployed at the compressor inlet, middle section, and outlet. The pressure sensors acquire dynamic pressure fluctuation signals within the flow channel, and the hot-wire anemometer acquires velocity pulsation signals within the flow channel, thus forming aerodynamic response data containing multiple positions and variables. Because the sampling frequencies of the valve group opening / closing cycle signal and the axial flow aerodynamic response signal are usually inconsistent—for example, the sampling frequency of the valve displacement sensor is 5kHz, while the sampling frequency of the compressor pressure sensor is 20kHz—direct comparison will result in time alignment errors. To solve this problem, the valve group opening / closing cycle signal and the axial flow aerodynamic response signal are first time-normalized separately. Specifically, the highest sampling frequency is selected as the reference time step, and cubic spline interpolation is performed on the lower sampling frequency signal within this reference time step to ensure that its time resolution is consistent with that of the higher sampling frequency signal. Subsequently, the two signals are synchronized under a unified reference clock to ensure that the valve opening / closing state and the flow field aerodynamic response at the same moment can accurately correspond. To ensure the validity of the signal, noise suppression processing is further performed after time alignment. Specifically, a wavelet threshold denoising algorithm is applied to each time sequence to remove random noise components unrelated to valve action at the wavelet decomposition scale, so that the finally constructed unified time baseline can reflect the true temporal relationship between the valve group opening / closing cycle signal and the axial flow aerodynamic response signal with high fidelity.

[0063] After establishing a unified time baseline, multi-scale phase difference mapping is performed on the valve group opening / closing cycle signal and the axial flow aerodynamic response signal on the unified time baseline. In the specific implementation, continuous wavelet transforms are first performed on both signals to decompose the time series into multiple scales to obtain the instantaneous frequency distribution. The scale range covers the entire frequency band from tens of hertz to thousands of hertz, ensuring coverage of the low-frequency disturbances of valve opening / closing and the high-frequency vortex of the compressor. After wavelet transform, each signal corresponds to the instantaneous phase at multiple scales at a given time point. Next, at the same time point, the instantaneous phase of the valve group opening / closing cycle signal is subtracted from the instantaneous phase of the axial flow aerodynamic response signal to obtain the phase difference value at that scale. The phase difference values ​​at all scales are continuously calculated in the time direction, forming a three-dimensional data set of time-scale-phase difference. Based on this, a multi-scale phase difference mapping map is constructed, which shows the variation of the phase difference between valve group opening / closing and aerodynamic response over time in different frequency ranges. In phase difference mapping, if the phase difference at a certain scale remains stable near a fixed value over a long period, it indicates a strong phase-locked relationship between valve group opening / closing and aerodynamic response within that frequency range. If the phase difference exhibits continuous drift or jumps, it indicates the presence of nonlinear coupling within that frequency range. To enhance the reliability of the mapping results, phase-locked values ​​and phase-difference entropy are calculated for the phase difference sequence at each scale. The former measures the degree of phase synchronization, while the latter measures the disorder of the phase difference. Through joint analysis of these two values, it is possible to clearly distinguish which frequency ranges are stable phase-synchronous regions and which are unstable phase-drift regions, thus laying a data foundation for the subsequent extraction of frequency superposition trajectories.

[0064] Finally, based on multi-scale phase difference mapping, frequency superposition trajectories under nonlinear conditions are extracted, and higher-order resonant peaks are identified within a unified temporal framework. In the specific implementation, the multi-scale phase difference mapping is first transformed into a three-dimensional surface model, with the three coordinate axes representing time, scale, and phase difference value, respectively. By plotting contour lines on the phase difference surface, trajectories exhibiting similar evolution trends in phase difference across different time scales can be obtained; these trajectories are the frequency superposition trajectories. Each frequency superposition trajectory corresponds to the coupling characteristics of the valve group opening / closing cycle signal and the axial flow aerodynamic response signal within that frequency range. Subsequently, a nonlinear energy density calculation method is introduced onto the frequency superposition trajectories. Specifically, a Hilbert transform is performed on the corresponding valve group opening / closing cycle signal and aerodynamic response signal on the frequency superposition trajectory to extract the instantaneous energy envelope, which is then integrated over the time scale to obtain the energy intensity value at each time point of the frequency superposition trajectory. By continuously measuring the energy intensity across the entire trajectory, frequency ranges with concentrated energy can be identified. For these energy ranges, the variation in the opening amplitude of the valve group opening / closing cycle signal and the pressure amplitude fluctuation of the aerodynamic response signal are further combined to determine whether an abnormal amplification effect exists within that frequency range. When the energy intensity on the frequency superposition trajectory exceeds three times the baseline energy level, and the corresponding phase difference curve exhibits continuous convergence, a higher-order resonance peak is identified within that frequency range. After identifying the higher-order resonance peak, this result is compared with the original valve group opening / closing cycle signal and aerodynamic response signal under a unified time baseline. By checking for significant abnormal fluctuations in the amplitude of the original signals at the same time points, it is verified whether the identified higher-order resonance peak truly reflects the nonlinear coupling between the valve group and the aerodynamic response, rather than being a false peak caused by noise or occasional fluctuations. Through this specific process, higher-order resonance peaks can be identified with high precision within a unified time-series framework, providing a reliable basis for subsequent energy concentration analysis and risk warning.

[0065] It should be noted that:

[0066] The high-order resonance peak refers to an abnormal energy peak, different from the system's fundamental frequency or low-order harmonics, formed by the superposition, doubling, or combination effect of multiple frequency components when the valve group opening and closing cycle signal and the axial flow aerodynamic response signal are nonlinearly coupled. This abnormal energy peak is characterized by a significant amplification of energy intensity and a continuously converging phase difference within a specific frequency range. It reflects the strong coupling characteristics between aerodynamic excitation and structural response at the high-order frequency level and is an important abnormal frequency indicator that may induce local vibration amplification and structural risks.

[0067] Through the above steps, the nonlinear frequency superposition trajectory extraction and high-order resonance peak identification of the valve group opening and closing cycle signal and the axial flow aerodynamic response signal under a unified time baseline were realized. This not only solved the timing deviation problem caused by inconsistent sampling frequencies and signal asynchrony, but also revealed the complex nonlinear coupling relationship through multi-scale phase difference mapping. Furthermore, the high accuracy of the identified high-order resonance peak was ensured through energy density calculation and phase difference verification.

[0068] S102. Based on the frequency superposition trajectory, extract the energy concentration interval and calculate the dynamic slope factor to characterize the energy transition intensity, thereby generating a fingerprint matrix that characterizes the frequency amplification trend, and using the fingerprint matrix as a constraint condition for the inherent frequency matching analysis.

[0069] The specific steps for generating a fingerprint matrix representing the amplification trend of frequency anomalies are as follows:

[0070] After obtaining the frequency superposition trajectory, the energy distribution characteristics of the trajectory are analyzed to extract energy concentration intervals. In the specific implementation, instantaneous energy is first calculated for each time point on the frequency superposition trajectory. Instantaneous energy is obtained by summing the squares of the instantaneous amplitudes of the valve group opening / closing cycle signal and the axial flow aerodynamic response signal at that time point. Subsequently, the instantaneous energy values ​​are arranged chronologically along the frequency superposition trajectory and smoothed using a sliding window integration method. The window length is set to at least ten times the sampling period based on the sampling frequency to ensure suppression of short-term fluctuations. This method yields a continuous energy distribution curve over time. If the energy value in a certain continuous interval is significantly higher than that of neighboring intervals, for example, exceeding three times the average energy level of neighboring intervals, then that interval is marked as an energy concentration interval. This method accurately identifies the time periods and frequency ranges where abnormal energy accumulation occurs on the frequency superposition trajectory, providing an initial range for subsequent energy transition intensity analysis.

[0071] After obtaining the energy concentration interval, the energy evolution trend within this interval is analyzed to calculate the dynamic slope factor and characterize the intensity of energy transitions. Specifically, within the energy concentration interval, the first derivative of the energy distribution curve is calculated to obtain the instantaneous energy growth rate over time. To avoid abrupt changes caused by occasional noise, a three-point central difference formula is used for numerical approximation when calculating the instantaneous growth rate, i.e., the energy values ​​of the previous, current, and next time moments are used for joint calculation to ensure the smoothness of the slope calculation. After obtaining the energy growth rate, it is standardized to distribute it within the range of 0 to 1, and defined as the dynamic slope factor. The larger the dynamic slope factor, the stronger the energy transition within this interval, and the more anomalous the frequency superposition phenomenon experienced by the system. By calculating the dynamic slope factor, the local energy peaks within the energy concentration interval can be distinguished from the overall evolution trend, thereby accurately characterizing the intensity features of energy transitions on a unified time baseline.

[0072] Subsequently, a fingerprint matrix characterizing the frequency amplification trend is generated based on the energy concentration interval and the dynamic slope factor. In practice, the time series of the energy concentration interval is first used as the x-axis, and the frequency series of the corresponding frequency superposition trajectory is used as the y-axis to establish a frequency-time coordinate framework on a two-dimensional plane. Then, the values ​​corresponding to the dynamic slope factor are mapped onto this two-dimensional framework as intensity indicators, thus obtaining a distribution map of energy transitions on the two-dimensional plane. To make this distribution map more clearly characterize the anomalous trend, a color coding method is further introduced, where the intensity of the color represents the magnitude of the dynamic slope factor, and the color change trend represents the evolution direction of energy amplification. The final two-dimensional matrix is ​​the fingerprint matrix characterizing the frequency amplification trend. This fingerprint matrix not only contains the location and range of the energy concentration interval but also the intensity and direction of the energy transitions, thus fully recording the dynamic amplification characteristics of the frequency superposition trajectory under anomalous conditions.

[0073] The fingerprint matrix is ​​used as a constraint in natural frequency matching analysis to improve the accuracy of high-order resonance identification. In practice, modal analysis is first performed on the natural frequencies of the axial flow structure to obtain its natural frequency distribution range. Then, the abnormal frequency intervals corresponding to the frequency coordinates in the fingerprint matrix are matched one by one with the natural frequency ranges obtained from the modal analysis. To prevent false matches, two conditions must be met simultaneously during matching: first, the overlap between the two frequency ranges must exceed a preset overlap threshold, for example, exceeding 50%; second, the corresponding dynamic slope factor must be at a high intensity level, for example, greater than 0.7. When these conditions are met, the abnormal frequency interval is considered to have a high-risk coupling with the natural frequency. Through this matching method, potential dangerous frequency neighborhoods can be quickly identified under the constraint of the fingerprint matrix, avoiding the ambiguity caused by relying solely on natural frequency calculations. Ultimately, the fingerprint matrix not only records the frequency amplification trend but also directly serves as a constraint for natural frequency matching analysis, thus enabling the entire early warning method to achieve a closed loop from signal extraction to risk assessment.

[0074] This implementation method completes the entire process from frequency superposition trajectory to abnormal trend characterization and then to natural frequency matching by extracting energy concentration intervals, calculating dynamic slope factors, generating fingerprint matrices, and matching them with natural frequencies. It can not only reveal the details of energy accumulation and transition on a unified time baseline, but also intuitively present the frequency amplification trend in the form of fingerprint matrices. Furthermore, it can introduce this as a constraint condition into natural frequency analysis to ensure the accuracy of dangerous frequency interval identification.

[0075] S103. Based on the fingerprint matrix, the fingerprint features are matched one by one with the modal characteristics of the axial flow structure, and a modal response phase field is constructed during the matching process to achieve real-time coupling analysis under unified conditions, thereby locking the dangerous frequency neighborhood that overlaps with the natural frequency.

[0076] The specific steps for constructing the modal response phase field are as follows:

[0077] Based on the obtained fingerprint matrix, fingerprint features are extracted and matched one by one with the modal characteristics of the axial flow structure. In the specific operation, the fingerprint matrix is ​​plotted with time on the x-axis and frequency on the y-axis, with a dynamic slope factor as the intensity identifier; therefore, each continuous energy trajectory can be considered a fingerprint feature. For each fingerprint feature, its center frequency, frequency bandwidth, and corresponding energy intensity distribution are first calculated to form a parameterized description of the fingerprint feature. Simultaneously, modal analysis of the axial flow structure is performed using the finite element method to obtain modal characteristic parameters including natural frequencies, mode shapes, and damping ratios. During the matching process, the center frequency of each fingerprint feature is compared with the natural frequency of the modal characteristic. When the frequency difference is less than a preset frequency difference threshold (e.g., 5% of the natural frequency) and the energy intensity of the fingerprint feature exceeds three times the background energy level, the fingerprint feature is considered to have a high correlation with the corresponding modal characteristic. In this way, fingerprint features coupled with the inherent modes of the axial flow structure can be screened one by one, providing accurate input for the subsequent construction of the modal response phase field.

[0078] After matching fingerprint features with modal characteristics, a modal response phase field is constructed to achieve real-time mapping of coupling relationships under unified conditions. In the specific implementation, time is used as one dimension and spatial modal morphology as another, mapping the dynamic slope factor and energy intensity corresponding to the matched fingerprint features to the spatial distribution of modal morphology, thus obtaining a three-dimensional time-space-energy distribution framework. To further reveal the temporal synchronization or phase differences of different modal responses, phase information is introduced into this three-dimensional distribution framework. Specifically, a Hilbert transform is performed on the energy envelope signal of each fingerprint feature to extract the instantaneous phase, and the instantaneous phase is combined with the spatial modal morphology distribution to form a time-space-phase mapping relationship. This mapping relationship is the modal response phase field, which can display the phase evolution of different modes during energy amplification under a unified time baseline. If a certain mode maintains a locked or continuously approaching phase relationship with the fingerprint feature throughout the entire energy concentration range, it indicates a strong coupling effect between the mode and the fingerprint feature. In this way, the phase relationship between modal response and fingerprint features can be tracked in real time, rather than just static matching at the frequency level.

[0079] Based on the modal response phase field, dangerous frequency neighborhoods overlapping with natural frequencies are further identified and locked. In the specific implementation, a phase-locked region is first searched in the modal response phase field; this is a region where the phase difference remains consistently below a certain phase difference threshold, such as 30 degrees. Then, it is checked whether this region corresponds to the natural frequency matched in the previous step. If the condition is met, the region is determined to be a potential dangerous frequency neighborhood. To further improve the reliability of the identification, the relationship between energy intensity and phase needs to be examined simultaneously: when the energy intensity continuously exceeds three times the background energy level, the background energy level refers to the normal energy reference value exhibited by the valve group opening / closing cycle signal and the axial flow aerodynamic response signal on a unified time baseline before significant nonlinear coupling and abnormal energy amplification occur. This is usually obtained through statistical analysis of the frequency superposition trajectory under stable operating conditions; for example, taking the average or median energy value of the trajectory over a long period and removing short-term spikes to form a stable reference value. The detected energy intensity is then compared with this benchmark value. When the energy intensity significantly exceeds the background energy level, it indicates an abnormal energy accumulation in the current frequency range, potentially forming a dangerous frequency neighborhood. Furthermore, if the phase difference remains continuously within the phase difference threshold for a specified duration (e.g., 10 sampling periods), the frequency range is formally identified as a dangerous frequency neighborhood. Finally, the identified dangerous frequency neighborhoods are marked on a unified time baseline for subsequent energy coupling path calculations and risk warnings. This process not only clearly identifies which frequency ranges overlap with the natural frequencies of the axial flow structure but also combines the dynamic evolution characteristics of energy and phase to achieve high-precision locking of dangerous frequency neighborhoods.

[0080] This implementation method realizes the entire process of identification from the fingerprint matrix to the dangerous frequency neighborhood. First, through parameterized description and frequency matching, the fingerprint features are mapped one-to-one with the modal characteristics of the axial flow structure. Second, by constructing a modal response phase field, energy evolution and phase evolution are fused on a unified time baseline, and the coupling relationship is plotted in real time. Third, by jointly determining the phase-locked region and energy intensity, the dangerous frequency neighborhood is accurately identified and locked. It considers not only frequency and energy but also introduces phase relationships, making the identification of dangerous frequency neighborhoods more comprehensive and reliable.

[0081] S104. Based on the modal response phase field, force distribution weights are introduced into the neighborhood of the dangerous frequency, and the amplification effect of the energy coupling path is dynamically calculated to form a continuously evolving energy amplification curve, which is used to characterize the local response risk of the structure.

[0082] The specific steps for generating an energy amplification curve are as follows:

[0083] Based on the modal response phase field, the identified danger frequency neighborhoods are further processed, and force distribution weights are introduced into these neighborhoods. In practice, firstly, time-series data and spatial modal morphology data corresponding to the danger frequency neighborhoods are extracted from the modal response phase field. The spatial modal morphology data is then divided into multiple discrete points according to the element nodes obtained from finite element analysis. For each discrete point, a corresponding force weight is assigned based on its relative displacement amplitude in the modal shape. The weight is calculated as the ratio of the relative displacement amplitude at that point to the maximum displacement amplitude of the overall mode, ensuring that the greater the displacement contribution in the modal response, the higher the assigned force weight. Subsequently, these force weights are weighted and superimposed on the time series of the danger frequency neighborhoods, so that the modal response phase field not only contains time and phase information but also reflects the contribution of spatial distribution to the energy response. By introducing force distribution weights, the originally uniformly processed modal phase information can be transformed into a weighted distribution reflecting the sensitivity of local structures, providing more refined input conditions for subsequent energy coupling path calculations.

[0084] After introducing the force distribution weights, the initial form of the energy coupling path is dynamically calculated based on the modal response phase field. In the specific implementation, firstly, modal phase difference sequences and corresponding energy intensity sequences are extracted for each time segment in the neighborhood of the hazardous frequency. These sequences are then weighted using the force distribution weights to form a four-dimensional dataset of time, space, phase difference, and energy. In this four-dimensional dataset, the continuous evolution trajectory of the phase difference and the changing trend of the energy intensity jointly determine the direction of energy propagation and the accumulation path. Therefore, the energy coupling path needs to be derived using a phase gradient calculation method. Specifically, the partial derivative of the phase difference at adjacent time points is calculated to obtain the direction vector of energy propagation. This direction vector is then multiplied by the force distribution weights to obtain the weighted energy coupling direction. This method allows us to obtain the energy coupling path that evolves over time in the neighborhood of the hazardous frequency, thus visually characterizing the accumulation and diffusion of energy in the structure. This process elevates the originally static energy distribution into a dynamic energy coupling path with directionality and temporality, laying the foundation for subsequent quantitative analysis of the amplification effect.

[0085] After obtaining the energy coupling path, the dynamic amplification effect of the path is quantitatively calculated. Specifically, the instantaneous energy amplification factor is first calculated at each discrete point along the energy coupling path. This instantaneous energy amplification factor is defined as the ratio of the weighted energy intensity at that point to the baseline energy intensity outside the danger frequency neighborhood. If this ratio is greater than a preset threshold, such as greater than 2, it indicates that energy amplification exists at that point. If the transient energy amplification factor at several consecutive points along the entire path exceeds the threshold, it is determined to be an energy amplification interval. Subsequently, the transient energy amplification factor within the energy amplification interval is integrated over time to obtain the cumulative value of the amplification effect. This cumulative value is then combined with the corresponding phase difference evolution trend to form a dynamic amplification effect curve. This method not only reveals the localized locations of energy enhancement in space but also reflects the energy evolution process over time. Especially when the phase difference gradually converges, the dynamic amplification effect curve exhibits a rapid upward trend, thus providing a quantitative basis for identifying local response risks.

[0086] After calculating the energy coupling path amplification effect, a continuously evolving energy amplification curve is generated to characterize the local response risk of the structure. In practice, the dynamic amplification effect curves of all energy amplification intervals are stitched together chronologically to obtain a complete energy amplification curve. This curve, with time on the x-axis and the cumulative amplification effect on the y-axis, continuously displays the growth trend of energy in the neighborhood of the danger frequency. To more intuitively characterize the local response risk of the structure, the energy amplification curve is jointly labeled with the force distribution weights. This involves color-coding or numerically labeling the amplification contribution of different spatial locations on the curve, clearly showing which locations bear the main risk during the energy amplification process. In this way, the energy amplification curve not only reflects the overall risk evolution trend but also highlights the spatial distribution characteristics of local risks. When the energy amplification curve shows a sharp rise in a short period, and the corresponding spatial weights are concentrated in key structural locations, it can be determined that there is a high-risk local response in that location. The final energy amplification curve, as a continuously evolving quantitative indicator, provides a precise basis for subsequent risk warning and proactive control.

[0087] This implementation method achieves dynamic calculation and risk characterization of energy amplification effects in the neighborhood of dangerous frequencies. First, by introducing force distribution weights, the modal response phase field is transformed into a spatially sensitive weighted distribution. Second, the energy coupling path is obtained through phase gradient and weighting processing, revealing the dynamic propagation direction of energy in the structure. Third, by calculating the instantaneous amplification coefficient and cumulative value, a dynamic amplification effect curve is formed, realizing the quantification of the amplification process. Finally, by splicing and spatial labeling, a continuously evolving energy amplification curve is obtained, intuitively characterizing the local response risk of the structure. This not only achieves high-resolution risk identification in time and frequency but also precise risk location in spatial distribution.

[0088] S105. Based on the energy amplification curve, calculate the coupling risk coefficient between the high-order resonant peak and the natural frequency, and generate a dynamic risk curve by combining the frequency evolution trend on a unified time baseline, so as to realize the quantitative identification and real-time early warning of coupling risk.

[0089] The risk dynamic curve is generated by combining the frequency evolution trend on a unified time baseline. The specific steps are as follows:

[0090] After constructing the energy amplification curve, characteristic parameters of the higher-order resonant peaks are extracted, and their coupling degree with the natural frequency is calculated to obtain the coupling risk coefficient. In the specific operation, the peak of the energy amplification curve is first identified, and the local maxima of the energy curve are found using the first derivative zero-intersection method. The frequency corresponding to the local maxima is recorded as the higher-order resonant peak frequency. Then, this higher-order resonant peak frequency is compared with the natural frequency obtained from modal analysis, the frequency difference is calculated, and a normalization method is used to convert the frequency difference into a frequency proximity index. When the frequency proximity index is higher than a set proximity threshold, it indicates a potential coupling risk between the higher-order resonant peak and the natural frequency. Based on this, an energy amplification factor, i.e., the ratio of the peak value of the energy amplification curve to the baseline energy level, is introduced. The frequency proximity and energy amplification factor are weighted and fused to obtain the coupling risk coefficient. The larger the value of this coupling risk coefficient, the stronger the coupling between the higher-order resonant peak and the natural frequency, and the higher the risk of local resonance in the structure. In this way, frequency information and energy information can be combined to quantitatively characterize the strength of the coupling risk, rather than relying solely on frequency proximity.

[0091] After obtaining the coupling risk coefficient, the dynamic changes of risk are analyzed by combining the frequency evolution trend on a unified time baseline to generate a risk dynamic curve. In the specific implementation, the frequency evolution trajectory under the unified time baseline is first synchronized with the energy amplification curve, ensuring a one-to-one correspondence between frequency information and energy amplification information at each time point. Subsequently, the coupling risk coefficient is recalculated at each time point, forming a continuous curve over time. To ensure the smoothness of the curve, a weighted moving average method is introduced during the calculation to eliminate the influence of short-term fluctuations and retain the overall trend of risk changes. Furthermore, the slope information of the frequency evolution trend is superimposed on the risk dynamic curve. If the frequency evolution trend rapidly approaches the inherent frequency, additional weights are added to the coupling risk coefficient, enabling the risk dynamic curve to more sensitively reflect the proximity of the risk. Through this method, the risk dynamic curve not only shows the intensity of the risk but also reflects the speed and direction of its evolution, thus providing a more comprehensive risk assessment than static indicators.

[0092] After the risk dynamic curve is generated, quantitative identification and real-time early warning of coupling risks are achieved based on the real-time evolution of the curve. In practice, multi-level thresholds are first set for the risk dynamic curve, such as normal range, warning range, and danger range, each corresponding to different risk levels. When the value of the risk dynamic curve is within the normal range, it indicates a low coupling risk; when the value enters the warning range, it indicates a gradually increasing risk level requiring continuous monitoring; when the value exceeds the danger range, it indicates that the higher-order resonant peak has significantly coupled with the natural frequency, and an early warning should be triggered immediately. When an early warning is triggered, not only the risk level is output, but the corresponding time and frequency ranges are also marked, indicating the specific frequency neighborhood that may lead to local response risks. To improve real-time performance, the update frequency of the risk dynamic curve should be consistent with the signal sampling frequency; for example, when the signal sampling frequency is 20kHz, the update interval of the risk curve should not exceed 0.05 seconds. In this way, quantitative identification and real-time early warning of coupling risks can be achieved, providing reliable data support for proactively adjusting valve group opening and closing strategies.

[0093] This step transforms the energy amplification curve into a risk dynamic curve. First, by comparing the high-order resonant peak with the natural frequency and combining it with the energy amplification factor, the coupling risk coefficient is calculated, thus quantitatively characterizing the risk intensity. Second, by combining it with a unified time baseline, the coupling risk coefficient evolves over time into a risk dynamic curve, reflecting the trend and speed of risk change. Third, through real-time monitoring of the risk dynamic curve and the determination of graded thresholds, quantitative identification and real-time early warning of risks are achieved. This method integrates the three dimensions of frequency, energy, and time, overcoming the limitations of static analysis and single-frequency determination in existing technologies, and enabling earlier and more accurate capture of potential risk signals.

[0094] S106. Based on the risk dynamic curve, when the coupling risk between the high-order resonance peak and the natural frequency is identified, the spectral characteristics of the valve group opening and closing cycle signal are dynamically monitored, and the frequency domain window of the valve group opening and closing cycle signal is shifted under the condition of continuous enhancement of the resonance factor, so that the overall operating frequency of the valve group is pushed away from the natural frequency neighborhood.

[0095] Based on the risk dynamic curve, when a coupling risk is identified between the high-order resonant peak and the natural frequency, the operating frequency of the valve group is pushed away from the natural frequency neighborhood as a whole. The specific steps are as follows:

[0096] Based on real-time calculations of the risk dynamic curve, the coupling risk between higher-order resonant peaks and natural frequencies is identified, and the dangerous frequency neighborhood requiring monitoring is pinpointed accordingly. In practice, when the risk dynamic curve value reaches a preset warning threshold, the corresponding frequency range is immediately extracted; this range represents the potentially dangerous neighborhood. At this point, the valve group opening / closing cycle signal corresponding to this neighborhood is acquired and subjected to a Fast Fourier Transform (FFT) to obtain the spectral characteristics of the valve group opening / closing cycle signal. These spectral characteristics not only record the dominant frequency of the current valve group opening / closing cycle signal but also include harmonic components and energy distribution. By comparing the spectral characteristics with the dangerous frequency neighborhood, it can be determined whether the valve group opening / closing action has approached or entered the natural frequency range. If the dominant frequency or higher-order harmonic components of the valve group opening / closing cycle signal overlap with the dangerous frequency neighborhood, a substantial coupling risk is identified between the valve group opening / closing characteristics and the structural modes. This process combines the risk dynamic curve with spectral monitoring, achieving a seamless transition from macroscopic risk assessment to detailed signal analysis.

[0097] After identifying the overlap between the spectral characteristics of the valve group's opening and closing cycle signal and the dangerous frequency neighborhood, the spectral characteristics are dynamically monitored to track the changing trend of the resonance factor in real time. In the specific implementation, the spectrum of the valve group's opening and closing cycle signal is continuously sampled and updated, and the spectral result of each sample is compared with the previous one to obtain the evolution trajectory of the resonance factor. The resonance factor is calculated by ratioing the energy value within the dangerous frequency neighborhood to the reference energy value outside the dangerous frequency neighborhood. If the ratio continues to rise, it indicates that the resonance effect is increasing. To avoid misjudgments caused by occasional disturbances, a sliding window averaging method is used to smooth the resonance factor sequence, ensuring that the monitoring results reflect the true trend. In this way, after the risk dynamic curve indicates coupled risk, further dynamic spectral monitoring can be used to confirm whether the resonance factor continues to increase, thereby determining whether the risk has developed from a potential state to an actual state. This step ensures that risk identification is not only based on static frequency overlap but also considers the dynamic energy evolution process, improving the reliability of the early warning.

[0098] When dynamic monitoring results show a continuous increase in the resonance factor, the frequency domain window of the valve group's opening and closing cycle signal is shifted to actively push the valve group's operating frequency away from its natural frequency neighborhood. In specific operations, the dominant frequency position of the valve group's opening and closing cycle signal is first determined and compared with the center frequency of the danger frequency neighborhood, calculating the offset between the two. If the offset is less than a preset safety distance, such as less than 2% of the natural frequency, the current operating frequency is determined to be in a danger zone, requiring shifting. The shift is achieved by adjusting the valve group's opening and closing cycle parameters, such as extending or shortening the valve's opening and closing duration, or changing the phase difference distribution of the valves in the cycle, thereby shifting the overall opening and closing frequency of the valve group away from its natural frequency. To ensure a smooth shift process, a gradual adjustment is adopted, with each adjustment not exceeding 0.5% of the dominant frequency. During the adjustment process, the risk dynamic curve and spectral characteristics are monitored in real time to ensure the shift direction is correct and no new danger frequency components are introduced. This active shift method effectively avoids direct overlap between the valve group's opening and closing cycle signal and its natural frequency, fundamentally reducing coupling risks.

[0099] It should be noted that the operating frequency refers to the set frequency of the valve group opening and closing cycle signal, which is usually adjusted through the valve group opening and closing cycle parameters; the dominant frequency refers to the main frequency component in the system response obtained through spectrum analysis, representing the most influential frequency in the system.

[0100] After completing the frequency domain window migration, the migration effect is verified, and a closed-loop risk warning and avoidance mechanism is formed. In the specific implementation process, firstly, the spectral characteristics of the valve group opening and closing cycle signal are recalculated in the time period after the migration is completed, and compared with the neighborhood of the dangerous frequency to confirm that the dominant frequency has moved far away from the inherent frequency. Subsequently, combined with the real-time update results of the risk dynamic curve, it is checked whether the coupling risk coefficient has decreased significantly. If the coupling risk coefficient falls back to the normal range, it indicates that the migration measure is effective. During the verification process, the stability of the valve group opening and closing actions is evaluated to ensure that adjusting the cycle parameters will not adversely affect the flow field regulation performance. If the risk dynamic curve still shows a high risk after migration, the frequency domain migration amplitude needs to be further increased until the risk is significantly reduced. Finally, the migration and verification form a closed-loop process, enabling risk warning to not only remain at the identification level but also achieve proactive intervention and real-time avoidance, thereby ensuring the stability and safety of the axial flow structure under complex operating conditions.

[0101] This step enables spectral monitoring and proactive avoidance based on a risk dynamic curve. First, the risk dynamic curve identifies the neighborhood of dangerous frequencies and extracts spectral features. Second, the evolution trend of the resonance factor is dynamically monitored to confirm the persistence of the risk. Next, the frequency domain window of the valve group's opening and closing cycle signal is shifted to push the operating frequency away from the inherent frequency. Finally, verification after the shift forms a closed loop, ensuring that the risk is substantially eliminated. This method not only provides early warning before the risk occurs but also proactively adjusts the valve group's operating characteristics as the risk continues to increase, thus transforming passive monitoring into active control.

[0102] This invention performs multi-scale phase difference mapping on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal on a unified time baseline, and constructs a fingerprint matrix by combining the frequency superposition trajectory. Then, the fingerprint matrix is ​​matched with the modal characteristics one by one. This can accurately identify high-order resonance peaks and lock the neighborhood of dangerous frequencies, thereby clarifying the coupling relationship between aerodynamic disturbances and structural modes. This allows potential abnormal risks to be detected in the early stages and ensures the ability to provide early warning of aerodynamic instability trends under complex operating conditions.

[0103] This invention introduces force distribution weights into the modal response phase field, dynamically calculates the energy coupling path and generates an energy amplification curve, and further realizes the active migration of the valve group's opening and closing frequency domain window under the drive of the risk dynamic curve, so that the overall operating frequency of the valve group is far away from the natural frequency. This enables rapid avoidance measures to be taken after risk identification, effectively reducing the risk of local structural response and significantly improving the operational stability and safety of the device.

[0104] This invention provides, for example Figure 4 The anti-surge and pressure relief valve group collaborative anomaly early warning system shown in the axial flow structure includes a time baseline construction module, a fingerprint matrix generation module, a modal phase field construction module, an energy amplification curve formation module, a risk dynamic curve generation module, and a frequency domain window migration module.

[0105] The time baseline construction module constructs a unified time baseline based on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal, and performs multi-scale phase difference mapping on the unified time baseline. It extracts the frequency superposition trajectory under nonlinear conditions from the multi-scale phase difference mapping and identifies the high-order resonance peak.

[0106] The fingerprint matrix generation module extracts the energy concentration interval through frequency superposition trajectory and calculates the dynamic slope factor to characterize the energy transition intensity, thereby generating a fingerprint matrix that characterizes the frequency amplification trend.

[0107] The modal phase field construction module, based on the fingerprint matrix, matches the fingerprint features in it with the modal characteristics of the axial flow structure one by one, and constructs the modal response phase field during the matching process to achieve real-time coupling analysis under unified conditions and lock the dangerous frequency neighborhood that overlaps with the natural frequency.

[0108] The energy amplification curve forming module, based on the modal response phase field, introduces force distribution weights into the neighborhood of the danger frequency and dynamically calculates the amplification effect of the energy coupling path to form a continuously evolving energy amplification curve;

[0109] The risk dynamic curve generation module calculates the coupling risk coefficient between the higher-order resonant peak and the natural frequency through the energy amplification curve, and generates a risk dynamic curve by combining the frequency evolution trend on a unified time baseline.

[0110] The frequency domain window migration module, based on the risk dynamic curve, dynamically monitors the spectral characteristics of the valve group opening and closing cycle signal when it identifies a coupling risk between the high-order resonance peak and the natural frequency. Under the condition that the resonance factor continues to increase, it migrates the frequency domain window of the valve group opening and closing cycle signal, so that the overall operating frequency of the valve group is pushed away from the natural frequency neighborhood.

[0111] The anti-surge and pressure relief valve group collaborative anomaly early warning method for axial flow structures provided in this embodiment of the invention is implemented through the aforementioned anti-surge and pressure relief valve group collaborative anomaly early warning system for axial flow structures. For details of the specific methods and processes of the anti-surge and pressure relief valve group collaborative anomaly early warning system for axial flow structures, please refer to the embodiments of the aforementioned anti-surge and pressure relief valve group collaborative anomaly early warning method for axial flow structures, which will not be repeated here.

[0112] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for early warning of abnormality in a group of anti-surge and pressure relief valves with an axial flow structure, characterized in that, Includes the following steps: A unified time baseline is constructed based on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal. Multi-scale phase difference mapping is performed on the unified time baseline. Frequency superposition trajectory under nonlinear conditions is extracted from the multi-scale phase difference mapping to identify high-order resonance peaks. The energy concentration interval is extracted by frequency superposition trajectory, and the dynamic slope factor is calculated to characterize the energy transition intensity, thereby generating a fingerprint matrix that characterizes the frequency amplification trend. Based on the fingerprint matrix, the fingerprint features are matched one by one with the modal characteristics of the axial flow structure, and the modal response phase field is constructed during the matching process. Through real-time coupling analysis, the dangerous frequency neighborhood that overlaps with the natural frequency is locked. Based on the modal response phase field, a force distribution weight is introduced into the neighborhood of the danger frequency, and the amplification effect of the energy coupling path is dynamically calculated to form a continuously evolving energy amplification curve. The coupling risk coefficient between the higher-order resonant peak and the natural frequency is calculated by using the energy amplification curve, and a dynamic risk curve is generated by combining the frequency evolution trend on a unified time baseline. Based on the risk dynamic curve, when the risk of coupling between the high-order resonance peak and the natural frequency is identified, the spectral characteristics of the valve group opening and closing cycle signal are dynamically monitored. Under the condition that the resonance factor continues to increase, the frequency domain window of the valve group opening and closing cycle signal is shifted, and the operating frequency of the valve group is adjusted to move away from the natural frequency neighborhood.

2. The method for early warning of abnormality of anti-surge and pressure relief valve group in axial flow structure according to claim 1, characterized in that, Extracting the frequency superposition trajectory under nonlinear conditions includes the following steps: The valve group opening and closing cycle signal and axial flow aerodynamic response signal are acquired and synchronized under a unified reference clock through time normalization and cubic spline interpolation. Random noise is eliminated by wavelet threshold denoising algorithm to form a unified time baseline. On a unified time baseline, continuous wavelet transforms were performed on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal to obtain the instantaneous phase at multiple scales, and the phase difference was calculated to construct a multi-scale phase difference mapping. Multi-scale phase difference mapping is transformed into a three-dimensional surface model, and contour lines are drawn on the surface to extract the frequency superposition trajectory under nonlinear conditions. A nonlinear energy density calculation method is introduced on the frequency superposition trajectory, and the high-order resonance peak is identified by combining the phase difference convergence phenomenon.

3. The method for early warning of abnormality of anti-surge and pressure relief valve group in axial flow structure according to claim 2, characterized in that, Generating a fingerprint matrix characterizing the frequency amplification trend includes the following steps: After obtaining the frequency superposition trajectory, instantaneous energy is calculated on the frequency superposition trajectory, and the energy distribution curve is obtained by smoothing through sliding window integration, and the energy concentration interval is extracted. The first derivative of the energy distribution curve is calculated within the energy concentration interval, and the instantaneous growth rate is calculated using the central difference method. The dynamic slope factor is then standardized to characterize the energy transition intensity. Using the time series of the energy concentration interval as the horizontal axis and the frequency series of the frequency superposition trajectory as the vertical axis, and mapping the dynamic slope factor value as the intensity identifier, a two-dimensional matrix is ​​established to generate a fingerprint matrix.

4. The method for early warning of abnormality of anti-surge and pressure relief valve group in axial flow structure according to claim 1, characterized in that, Constructing the modal response phase field includes the following steps: Fingerprint features are extracted based on the obtained fingerprint matrix, and the center frequency, frequency bandwidth and energy intensity of the fingerprint features are matched with the natural frequency, mode shape and damping ratio of the axial flow structure. When the frequency difference is less than the preset frequency difference threshold and the energy intensity exceeds the background energy level, the corresponding fingerprint features are determined to be related to the modal characteristics. After matching is completed, the dynamic slope factor and energy intensity are mapped to a three-dimensional distribution using time and space modal morphology as coordinates, and the modal response phase field is constructed. The instantaneous phase is extracted by performing Hilbert transform on the energy envelope signal to form a mapping relationship between time and space phases. In the modal response phase field, phase-locked regions are identified, and when the energy intensity continuously exceeds three times the background energy level and the phase difference remains within the phase difference threshold range, the continuous frequency range corresponding to the phase-locked region is determined as the dangerous frequency neighborhood.

5. The method for early warning of abnormality of anti-surge and pressure relief valve group in axial flow structure according to claim 4, characterized in that, The background energy level is the average energy intensity obtained from the valve group opening and closing cycle signal and the axial flow aerodynamic response signal on a unified time baseline under stable operating conditions of the axial flow structure.

6. The method for early warning of abnormality of anti-surge and pressure relief valve group in axial flow structure according to claim 4, characterized in that, Based on the modal response phase field, a force distribution weight is introduced into the neighborhood of the danger frequency, and the amplification effect of the energy coupling path is dynamically calculated to form a continuously evolving energy amplification curve, including the following steps: Based on the phase field of the modal response, a force distribution weight is introduced into the neighborhood of the dangerous frequency, and the relative displacement amplitude of each discrete point in the mode shape is converted into a weight value and weighted and superimposed on the time series. After introducing the force distribution weights, the phase difference sequence and energy intensity sequence in the neighborhood of the danger frequency are weighted, and the propagation direction of the energy coupling path is obtained by calculating the phase gradient. The ratio of weighted energy intensity to reference energy intensity is calculated at discrete points along the energy coupling path to obtain the instantaneous energy amplification coefficient and identify the energy amplification interval. The energy amplification interval is then integrated and combined with the phase difference evolution to form a dynamic amplification effect curve. The dynamic amplification effect curves are spliced ​​together in chronological order to form an energy amplification curve, and then labeled with intensity distribution weights.

7. The method for early warning of abnormality of anti-surge and pressure relief valve group in axial flow structure according to claim 6, characterized in that, Generate a dynamic risk curve by combining the frequency evolution trend on a unified time baseline, including the following steps: After constructing the energy amplification curve, the characteristic parameters of the high-order resonance peak are extracted and compared with the natural frequency. The coupling risk coefficient is obtained by weighted fusion of frequency proximity and energy amplification factor. After obtaining the coupling risk coefficient, the frequency evolution trajectory under the unified time baseline is synchronized with the energy amplification curve, and the coupling risk coefficient is recalculated at each time point so that the coupling risk coefficient continuously evolves into a risk dynamic curve over time. The risk dynamic curve is then weighted and adjusted in combination with the slope of the frequency evolution trend. After the risk dynamic curve is generated, the risk level is determined based on the preset multi-level thresholds. When the risk dynamic curve exceeds the danger threshold, an early warning is triggered, and the corresponding time interval and frequency interval are output.

8. The method for early warning of abnormality of anti-surge pressure relief valve group in axial flow structure according to claim 7, characterized in that, Adjusting the operating frequency of the valve group to move it away from its natural frequency range includes the following steps: Based on the risk dynamic curve, dangerous frequency neighborhoods are identified and fast Fourier transform is performed on the valve group opening and closing cycle signal to extract spectral features and compare them with dangerous frequency neighborhoods to determine whether there is overlap. After identifying the overlap between the spectral features and the neighborhood of the dangerous frequency, the spectral features are dynamically monitored and the resonance factor is calculated. When the resonance factor shows a continuous increasing trend after smoothing, the risk is determined to change from a potential state to an actual state. After the risk is determined to be in an actual state, the frequency domain window of the valve group opening and closing cycle signal is shifted, the opening and closing cycle parameters or phase difference distribution are adjusted, and the risk dynamic curve and spectrum characteristics are monitored in real time during the shift process. After the migration is completed, the spectral characteristics are recalculated and combined with the risk dynamic curve to confirm whether the dominant frequency has moved away from the inherent frequency and the coupling risk coefficient has decreased, thus forming a risk avoidance mechanism.

9. An early warning system for coordinated anomalies of anti-surge and pressure relief valve groups in an axial flow structure, used to implement the early warning method for coordinated anomalies of anti-surge and pressure relief valve groups in an axial flow structure as described in any one of claims 1-8, characterized in that, It includes a time baseline construction module, a fingerprint matrix generation module, a modal phase field construction module, an energy amplification curve formation module, a risk dynamic curve generation module, and a frequency domain window shifting module; The time baseline construction module constructs a unified time baseline based on the valve group opening and closing cycle signal and the axial flow aerodynamic response signal, and performs multi-scale phase difference mapping on the unified time baseline. It extracts the frequency superposition trajectory under nonlinear conditions from the multi-scale phase difference mapping and identifies the high-order resonance peak. The fingerprint matrix generation module extracts the energy concentration interval through frequency superposition trajectory and calculates the dynamic slope factor to characterize the energy transition intensity, thereby generating a fingerprint matrix that characterizes the frequency amplification trend. The modal phase field construction module, based on the fingerprint matrix, matches the fingerprint features in it with the modal characteristics of the axial flow structure one by one, and constructs the modal response phase field during the matching process. Through real-time coupling analysis, it locks the dangerous frequency neighborhood that overlaps with the natural frequency. The energy amplification curve forming module, based on the modal response phase field, introduces force distribution weights into the neighborhood of the danger frequency and dynamically calculates the amplification effect of the energy coupling path to form a continuously evolving energy amplification curve; The risk dynamic curve generation module calculates the coupling risk coefficient between the higher-order resonant peak and the natural frequency through the energy amplification curve, and generates a risk dynamic curve by combining the frequency evolution trend on a unified time baseline. The frequency domain window migration module, based on the risk dynamic curve, dynamically monitors the spectral characteristics of the valve group opening and closing cycle signal when it identifies a coupling risk between the high-order resonance peak and the natural frequency. Under the condition that the resonance factor continues to increase, it migrates the frequency domain window of the valve group opening and closing cycle signal and adjusts the operating frequency of the valve group to move away from the natural frequency neighborhood.

Images