Multi-frequency weak fault enhancement method based on fractional order resonance response atlas reconstruction

Abstract

The invention discloses a multi-frequency weak fault enhancement method based on fractional order resonance response atlas reconstruction, and the method comprises the steps: carrying out the noise reduction, trend term removal, phase compensation and modal decomposition of a mechanical vibration signal, determining a lag phase through a frequency sweep method for phase compensation, adjusting the phase of a same-frequency signal, reconstructing an effective component through the combination of a modal decomposition borrowing geometric method and a Pearson matrix, and carrying out the phase compensation. Constructing a nonlinear resonance model with historical memory; inputting the parameters into the same-direction coupling neuron model, dynamically updating and simulating transmission intensity change through synaptic weight, and solving response and step length; constructing a zero-energy coupling index based on a signal zero-crossing behavior, energy distribution and time-frequency analysis, wherein the time-frequency analysis comprises short-time Fourier transform and wavelet packet energy spectrum calculation to extract energy features; and iteratively optimizing parameters by using a quantum optimization algorithm, adjusting a rotation angle and preserving population diversity to prevent premature convergence, and outputting optimal parameters by taking a zero-energy coupling index as a target.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of signal detection and processing, in particular to a multi-frequency weak fault enhancement method based on fractional order resonance response atlas reconstruction. BACKGROUND

[0002] In the field of mechanical fault diagnosis, the identification and analysis of multi-frequency weak faults are important links to ensure the stable operation of equipment. Such faults often accompany the operation of various mechanical equipment. The signals are often weak in strength, superimposed with multiple frequency components and disturbed by complex environment, which brings many challenges to accurate identification and feature extraction. The related technology has important application value in the condition monitoring and fault warning of rotating machinery, heavy equipment and the like.

[0003] The prior art often has many limitations in processing multi-frequency weak fault signals: in the signal preprocessing stage, it is difficult to effectively suppress noise, eliminate trend items and compensate for phase deviation, and the filtering effect of the interference signal is not good, resulting in inaccurate extraction of effective signal components; the time sequence characteristics and nonlinear features of the signal are not considered, there is a lack of model construction method that can preserve such features, and the quality of the basic signal is difficult to meet the subsequent analysis requirements; in the process of weak signal transmission and identification, the adaptability is poor, it is difficult to cope with the dynamic changes in signal transmission, and the transmission and identification ability of weak signals is limited; there is a lack of evaluation basis that can comprehensively reflect the multi-dimensional features of the signal, and it is difficult to reliably evaluate the fault enhancement effect; the parameter optimization method is prone to premature convergence problem, it is difficult to obtain optimal parameters, and the stability and effectiveness of the method are affected; it is difficult to construct a multi-dimensional response space covering frequency, energy and time, the atlas is not detailed enough, and it is not conducive to capturing weak fault information; in the separation of multi-frequency signals, repeated analysis is easy to occur, different frequency components cannot be separated step by step, and the final presented features lack structured integration, making it difficult to fully and clearly display the comprehensive features of multi-frequency weak faults, affecting the identification and analysis effect, therefore a multi-frequency weak fault enhancement method based on fractional order resonance response atlas reconstruction is proposed. SUMMARY

[0004] The present application provides the following technical scheme: a multi-frequency weak fault enhancement method based on fractional order resonance response atlas reconstruction, comprising the following steps:

[0005] S1, signal preprocessing and model construction:

[0006] Denoising, detrending, phase compensation and modal decomposition are performed on the mechanical vibration signal to construct a nonlinear resonance model with historical memory; the parameters and variable scale factors are initialized, and the preprocessed signal is input; the phase compensation uses the sweep frequency method to determine the lag phase, and the phase of the same frequency signal is adjusted; the modal decomposition uses the symplectic geometry method, and the effective components are reconstructed by using the Pearson matrix;

[0007] S2, model response solving:

[0008] The parameters are input into the homodromic coupling neuron model, the signal transmission and dynamic threshold adjustment are simulated, and the response and step length are solved; the model simulates the change of transmission strength through dynamic updating of synaptic weight;

[0009] S3, evaluation index construction:

[0010] Based on the signal zero-crossing behavior, energy distribution and time-frequency analysis, a zero-energy coupling index is constructed to measure the enhancement effect; the time-frequency analysis includes short-time Fourier transform and wavelet packet energy spectrum calculation to extract energy features in different frequency bands;

[0011] S4, parameter optimization:

[0012] The quantum optimization algorithm is used to iteratively optimize the parameters, and the quantum evolution is simulated combined with an adaptive mutation strategy, with the zero-energy coupling index as the target, and the optimal parameters are output; the algorithm adjusts the rotation angle and preserves the population diversity to prevent premature convergence;

[0013] S5, response atlas generation:

[0014] Based on the optimal parameters, a three-dimensional response space containing frequency, energy and time is constructed to solve the response and generate a multi-scale resonance response atlas; the three-dimensional space is constructed by correlating model parameters and response characteristics;

[0015] S6, iteration termination and atlas fusion:

[0016] Determine whether the optimal zero-energy coupling index is greater than the dynamic termination threshold; if yes, lock the main frequency, trap wave filter to remove the identified signal, and update the input iteration; otherwise, terminate and integrate the multi-round features into a structured atlas according to the frequency; the dynamic threshold is updated according to the exponential change trend; the main frequency is locked combined with strong energy frequency and phase to determine the main frequency; the structured atlas arranges the features according to the frequency, and superimposes the energy and phase to form a multi-level result.

[0017] The mechanical vibration signal is denoised, detrended, phase compensated and modal decomposed to construct a nonlinear resonance model with historical memory; the parameters and variable scale factors are initialized and input into the preprocessed signal; among them, the sweep frequency method is used to determine the lag phase to tune the phase of the same frequency signal; the modal decomposition uses the symplectic geometry method to reconstruct the effective component with the help of the Pearson matrix;

[0018] The parameters are input into the homodromic coupling neuron model, the signal transmission and dynamic threshold adjustment are simulated, and the response and step length are solved; the model simulates the change of transmission strength through dynamic updating of synaptic weight;

[0019] Based on the zero-crossing behavior, energy distribution and time-frequency analysis results of the signal, a zero-energy coupling index is constructed to measure the fault enhancement effect; time-frequency analysis includes short-time Fourier transform and wavelet packet energy spectrum calculation, which is used to extract energy features in different frequency bands;

[0020] The quantum optimization algorithm is used to iteratively optimize the parameters, and the optimal parameters are output by simulating quantum evolution and combining with an adaptive mutation strategy, with the zero-energy coupling index as the optimization target; the algorithm prevents premature convergence by adjusting the rotation angle and maintaining population diversity;

[0021] Based on the optimal parameters, a three-dimensional response space including frequency, energy and time is constructed, and a multi-scale resonance response map is generated by solving the response; the three-dimensional space is constructed by associating model parameters with response characteristics;

[0022] Determine whether the optimal zero-energy coupling index is greater than the dynamic termination threshold; if yes, lock the main frequency, remove the identified signal using a notch filter, update the input and iterate; if not, terminate the iteration and integrate the multi-round features into a structured map according to the frequency; the dynamic threshold is updated according to the trend of the index; the main frequency is determined by combining the strong energy frequency and the phase; the structured map arranges the features according to the frequency, and superimposes the energy and phase to form a multi-level result.

[0023] Preferably, the noise reduction in S1, signal preprocessing and model construction uses a method combining ensemble empirical mode decomposition and wavelet thresholding, which selects effective components by calculating the correlation coefficients of each intrinsic mode function.

[0024] Noise reduction is performed using a method combining ensemble empirical mode decomposition and wavelet thresholding, which first processes the signal through ensemble empirical mode decomposition, then combines the wavelet thresholding method, and then calculates the correlation coefficients of each intrinsic mode function to select effective components.

[0025] Preferably, the detrending in S1, signal preprocessing and model construction uses an adaptive local iterative filtering method, which realizes effective separation of nonlinear trend items by constructing a time-varying filter.

[0026] The detrending is removed using an adaptive local iterative filtering method, which constructs a time-varying filter to separate the nonlinear trend items.

[0027] Preferably, the nonlinear resonance model with historical memory in S1, signal preprocessing and model construction realizes the historical memory function through a time series caching mechanism, and sets a loop feedback link inside the model to integrate the response results at previous time into the current model input according to the time weight, and the weight decays over time to highlight the recent memory effect.

[0028] The history memory function is realized through a time series cache mechanism, a loop feedback link is set in the model, the response result at a previous time is integrated into the current model input according to a time weight, and the weight decays with time to highlight the influence of recent memory.

[0029] Preferably, the dynamic threshold adjustment in the model response solving S2 comprises a neuron firing rate feedback mechanism, and the threshold parameter is dynamically adjusted by monitoring the neuron cluster firing rate.

[0030] The neuron firing rate feedback mechanism is included, the neuron cluster firing rate is monitored, and the threshold parameter is dynamically adjusted according to the monitored firing rate.

[0031] Preferably, the construction of the zero energy coupling index in the evaluation index construction S3 comprises three sub-steps of signal zero-crossing rate feature extraction, energy entropy calculation and time-frequency feature fusion.

[0032] The three sub-steps of signal zero-crossing rate feature extraction, energy entropy calculation and time-frequency feature fusion are included, and the three steps are completed in sequence to construct the zero energy coupling index.

[0033] Preferably, the adaptive mutation strategy in the parameter optimization S4 comprises a gradient adjustment mechanism of quantum bit mutation probability, and the mutation strength is dynamically adjusted according to the gradient value of the objective function.

[0034] The gradient adjustment mechanism of quantum bit mutation probability is included, the gradient value of the objective function is calculated, and the mutation strength is dynamically adjusted according to the gradient value of the objective function.

[0035] Preferably, the three-dimensional response space construction in the response atlas generation S5 comprises a parameter sensitivity analysis step, and the key parameter dimension is determined by calculating the influence degree of parameter change on system response.

[0036] The parameter sensitivity analysis step is included, the influence degree of parameter change on system response is calculated, the key parameter dimension is determined according to the influence degree, and then the three-dimensional response space is constructed.

[0037] Preferably, the notch filtering in the iteration termination and atlas fusion S6 adopts an adaptive notch filter, the center frequency of which is dynamically adjusted according to the main frequency feature locking result, and the bandwidth is adaptively determined according to the signal frequency resolution.

[0038] The adaptive notch filter is used for notch filtering, the center frequency of the adaptive notch filter is dynamically adjusted according to the main frequency feature locking result, and the bandwidth is adaptively determined according to the signal frequency resolution

[0039] Preferably, the structured atlas construction in the iteration termination and atlas fusion S6 comprises a feature clustering analysis step, and the multiple iteration results are divided into different fault mode categories by calculating the similarity between features.

[0040] The feature clustering analysis step is included, the similarity between features is calculated, the multiple iteration results are divided into different fault mode categories according to the similarity, and a structured atlas is constructed.

[0041] Compared with the prior art, the application provides a multi-frequency weak fault enhancement method based on fractional order resonance response atlas reconstruction, which has the following beneficial effects:

[0042] The application can effectively reduce the influence of interference signals, accurately extract and retain effective components in the signal, and construct a nonlinear resonance model with historical memory, can fully consider the time sequence characteristics and nonlinear features of the signal, and provide high-quality basic signals for subsequent analysis, by denoising, detrending and phase compensation of mechanical vibration signals, modal decomposition of effective components based on symplectic geometry combined with Pearson matrix reconstruction.

[0043] The signal transmission and dynamic threshold adjustment are simulated by means of the homodirectional coupling neuron model, and the transmission strength change is simulated by dynamic updating of synaptic weights, which can adapt to changes in the signal transmission process, enhance the transmission and recognition ability of weak signals, and improve the accuracy of response solution.

[0044] The zero-energy coupling index constructed based on signal zero-crossing behavior, energy distribution and time-frequency analysis can comprehensively reflect the multi-dimensional characteristics of the signal, and provide reliable and comprehensive evaluation basis for fault enhancement effect.

[0045] The quantum optimization algorithm is combined with an adaptive mutation strategy, the rotation angle is adjusted and the population diversity is maintained to avoid premature convergence, and the zero-energy coupling index is used as the target iteration optimization parameter, which can effectively find the optimal parameters and ensure that the model runs in the best state, improving the stability and effectiveness of the method.

[0046] Based on the optimal parameters, a three-dimensional response space containing frequency, energy and time is constructed, and a multi-scale resonance response atlas is generated, which can comprehensively present the characteristics of the signal in different frequency, energy and time dimensions, and facilitate more detailed capture of weak fault information.

[0047] The relationship between the optimal zero-energy coupling index and the dynamic termination threshold is judged for iteration, the identified signals are removed by combining the main frequency locking and notch filtering to update the input, which can gradually separate and identify the multi-frequency weak fault signals, avoid repeated analysis, and finally integrate the multiple features into a structured atlas according to the frequency, which can comprehensively and clearly present the comprehensive characteristics of the multi-frequency weak fault, thereby effectively enhancing the identification and analysis effect of the multi-frequency weak fault. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is a schematic diagram of the method of the application. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the protection scope of the present application.

[0050] Please refer to Figure 1 The present application provides a technical solution, a multi-frequency weak fault enhancement method based on fractional order resonance response atlas reconstruction, comprising the following steps:

[0051] S1, signal preprocessing and model construction:

[0052] Denoising, detrending, phase compensation and modal decomposition are performed on the mechanical vibration signal to construct a nonlinear resonance model with historical memory; the initialization parameters and variable scale factor are inputted into the preprocessed signal; the phase compensation is performed by using the sweep frequency method to determine the lagging phase and adjust the phase of the same frequency signal; the modal decomposition is performed by using the symplectic geometry method to reconstruct the effective components by using the Pearson matrix;

[0053] S2, model response solving:

[0054] The parameters are inputted into the homodromous coupling neuron model to simulate signal transmission and dynamic threshold adjustment, and the response and step length are solved; the model simulates the change of transmission strength by dynamic updating of synaptic weight;

[0055] S3, evaluation index construction:

[0056] Based on the signal zero-crossing behavior, energy distribution and time-frequency analysis, the zero-energy coupling index is constructed to measure the enhancement effect; the time-frequency analysis includes short-time Fourier transform and wavelet packet energy spectrum calculation to extract energy characteristics in different frequency bands;

[0057] S4, parameter optimization:

[0058] The quantum optimization algorithm is used to iteratively optimize the parameters, the quantum evolution is simulated combined with an adaptive mutation strategy, the zero-energy coupling index is taken as the target, and the optimal parameters are outputted; the algorithm prevents premature convergence by adjusting the rotation angle and preserving population diversity;

[0059] S5, response atlas generation:

[0060] Based on the optimal parameters, a three-dimensional response space containing frequency, energy and time is constructed to solve the response and generate a multi-scale resonance response atlas; the three-dimensional space is constructed by associating the model parameters and the response characteristics;

[0061] S6, iteration termination and atlas fusion:

[0062] determine whether the optimal zero-energy coupling index is greater than a dynamic termination threshold; if yes, lock the main frequency, remove the identified signal by notch filtering, and update the input iteration; if no, terminate, integrate the multi-round features into a structured atlas according to the frequency; the dynamic threshold is updated according to the exponential change trend; the main frequency locking determines the main frequency in combination with the strong energy frequency and phase; the structured atlas arranges the features according to the frequency, and superimposes the energy and phase to form a multi-level result.

[0063] The mechanical vibration signal is denoised, detrended, phase compensated and modal decomposed to construct a nonlinear resonance model with historical memory; the parameters and variable scale factors are initialized and input into the preprocessed signal; the phase compensation adopts the sweep frequency method to determine the lag phase to tune the phase of the same frequency signal; the modal decomposition adopts the symplectic geometry method to reconstruct the effective component by means of the Pearson matrix;

[0064] The parameters are input into the homodirectional coupling neuron model to simulate the signal transmission process and dynamic threshold adjustment process, and to solve the response and step length; the model simulates the change of transmission strength through dynamic update of synaptic weight;

[0065] Based on the zero-crossing behavior, energy distribution and time-frequency analysis result of the signal, a zero-energy coupling index is constructed to measure the fault enhancement effect; the time-frequency analysis includes short-time Fourier transform and wavelet packet energy spectrum calculation, which are used to extract energy features in different frequency bands;

[0066] A quantum optimization algorithm is used to iteratively optimize the parameters, and the optimal parameters are output by simulating quantum evolution and combining an adaptive mutation strategy, with the zero-energy coupling index as the optimization objective; the algorithm prevents premature convergence by adjusting the rotation angle and maintaining population diversity;

[0067] Based on the optimal parameters, a three-dimensional response space including frequency, energy and time is constructed, and a multi-scale resonance response atlas is generated by solving the response; the three-dimensional space is constructed by associating the model parameters and response characteristics;

[0068] determine whether the optimal zero-energy coupling index is greater than a dynamic termination threshold; if yes, lock the main frequency, remove the identified signal by notch filtering, and update the input iteration; if no, terminate, integrate the multi-round features into a structured atlas according to the frequency; the dynamic threshold is updated according to the exponential change trend; the main frequency locking determines the main frequency in combination with the strong energy frequency and phase; the structured atlas arranges the features according to the frequency, and superimposes the energy and phase to form a multi-level result;

[0069] The quality of mechanical vibration signals can be effectively improved by combining various processing methods such as noise reduction, detrending, phase compensation and modal decomposition, and interference factors can be removed to provide a reliable signal basis for subsequent model construction and analysis. The construction of a nonlinear resonance model with historical memory helps to better capture the dynamic characteristics and historical information of the signal, and improves the sensitivity to weak fault signals.

[0070] The homodirectional coupling neuron model simulates signal transmission and dynamic threshold adjustment, which can better simulate the actual signal transmission process. The dynamically adjusted threshold can adapt to signals of different intensities and characteristics, enhancing the response capability to weak fault signals. The dynamic updating of synaptic weights simulates the change of transmission intensity, further improving the simulation accuracy of the model to the signal transmission process.

[0071] The zero-energy coupling index comprehensively considers the zero-crossing behavior, energy distribution and time-frequency analysis results of the signal, which can accurately measure the fault enhancement effect and provide reliable evaluation basis for subsequent parameter optimization and fault identification. The energy features of different frequency bands are extracted in time-frequency analysis, which helps to better understand the energy distribution of the signal and provides more information for fault feature identification.

[0072] The quantum optimization algorithm combined with adaptive mutation strategy can efficiently optimize parameters. Simulating the process of quantum evolution helps to find the optimal parameters in a larger range. By adjusting the rotation angle and maintaining population diversity to prevent premature convergence, the accuracy and stability of parameter optimization are improved, ensuring that the optimal parameters obtained can better meet the needs of fault enhancement.

[0073] The three-dimensional response space contains frequency, energy and time information, which can comprehensively display the characteristics of the signal. The generated multi-scale resonance response atlas can intuitively present the response under different frequencies, energies and times, which helps to more clearly identify weak fault features.

[0074] The dynamic termination threshold is updated according to the exponential trend, which can more reasonably determine whether the iteration is terminated, avoiding excessive iteration or premature termination. The main frequency locking combines strong energy frequency and phase to determine the main frequency, which improves the accuracy of main frequency identification. The notch filter removes the identified signal and updates the input iteration, which helps to gradually extract more fault features. The multi-round features are integrated into a structured atlas by frequency, and the energy and phase are superimposed to form multi-level results, making the fault features more systematic and comprehensive, and improving the accuracy and integrity of fault identification.

[0075] S1, the noise reduction in signal preprocessing and model construction adopts the method of combining ensemble empirical mode decomposition and wavelet threshold, and the effective components are selected by calculating the correlation coefficient of each intrinsic mode function.

[0076] The method combining ensemble empirical mode decomposition and wavelet threshold is adopted to carry out noise reduction, the signal is processed through ensemble empirical mode decomposition, then the wavelet threshold method is combined, and then the correlation coefficient of each intrinsic mode function is calculated, and the effective component is selected according to the correlation coefficient;

[0077] The method combining ensemble empirical mode decomposition and wavelet threshold can give full play to the advantages of the two methods in noise reduction, and the effective information in the signal can be accurately retained by calculating the correlation coefficient to select the effective component, so that the noise reduction effect is improved.

[0078] S1, the adaptive local iterative filtering method is adopted for the detrending item in signal preprocessing and model construction, and the time-varying filter is constructed to realize effective separation of the nonlinear trend item.

[0079] The adaptive local iterative filtering method is adopted to remove the trend item, the time-varying filter is constructed, and the time-varying filter is used to realize separation of the nonlinear trend item;

[0080] The adaptive local iterative filtering method has the adaptive characteristic, and the time-varying filter constructed can well adapt to the characteristics of the nonlinear trend item, so that the nonlinear trend item is effectively separated, and the accuracy of signal preprocessing is improved.

[0081] S1, the nonlinear resonance model with historical memory in signal preprocessing and model construction realizes the historical memory function through the time series cache mechanism, the loop feedback link is set in the model, the response result at the previous time is integrated into the current model input according to the time weight, and the weight decays with time to highlight the influence of recent memory.

[0082] The time series cache mechanism realizes the historical memory function, the loop feedback link is set in the model, the response result at the previous time is integrated into the current model input according to the time weight, and the weight decays with time to highlight the influence of recent memory.

[0083] The setting of the time series cache mechanism and the loop feedback link enables the model to retain historical information, the response result at the previous time is integrated according to the time weight, and the weight decays with time to highlight the influence of recent memory, thereby enhancing the processing ability of the model to the continuity and correlation of signal changes.

[0084] S2, the dynamic threshold adjustment in model response solving includes a neuron firing rate feedback mechanism, which dynamically adjusts the threshold parameter by monitoring the neuron cluster firing rate.

[0085] The neuron firing rate feedback mechanism is included, the neuron cluster firing rate is monitored, and the threshold parameter is dynamically adjusted according to the monitored firing rate;

[0086] The neuron firing rate feedback mechanism can monitor the activity state of the neuron cluster in real time, and based on this, dynamically adjust the threshold parameter, so that the threshold setting is more in line with the actual situation, and the accuracy of the model response solution is improved.

[0087] The construction of the evaluation index includes three sub-steps of signal zero-crossing rate feature extraction, energy entropy calculation and time-frequency feature fusion.

[0088] The construction of the evaluation index includes three sub-steps of signal zero-crossing rate feature extraction, energy entropy calculation and time-frequency feature fusion.

[0089] The construction of the evaluation index includes three sub-steps of signal zero-crossing rate feature extraction, energy entropy calculation and time-frequency feature fusion.

[0090] The adaptive mutation strategy in parameter optimization includes a gradient adjustment mechanism for quantum bit mutation probability, which dynamically adjusts the mutation strength according to the gradient value of the objective function.

[0091] The adaptive mutation strategy in parameter optimization includes a gradient adjustment mechanism for quantum bit mutation probability, which dynamically adjusts the mutation strength according to the gradient value of the objective function.

[0092] The adaptive mutation strategy in parameter optimization includes a gradient adjustment mechanism for quantum bit mutation probability, which dynamically adjusts the mutation strength according to the gradient value of the objective function.

[0093] The three-dimensional response space construction in response map generation includes a parameter sensitivity analysis step, which determines the key parameter dimension by calculating the influence of parameter changes on system response.

[0094] The three-dimensional response space construction in response map generation includes a parameter sensitivity analysis step, which determines the key parameter dimension by calculating the influence of parameter changes on system response.

[0095] The three-dimensional response space construction in response map generation includes a parameter sensitivity analysis step, which determines the key parameter dimension by calculating the influence of parameter changes on system response.

[0096] The notch filter in the iteration termination and map fusion adopts an adaptive notch filter, the center frequency of which is dynamically adjusted according to the main frequency feature locking result, and the bandwidth is adaptively determined according to the signal frequency resolution.

[0097] The notch filter in the iteration termination and map fusion adopts an adaptive notch filter, the center frequency of which is dynamically adjusted according to the main frequency feature locking result, and the bandwidth is adaptively determined according to the signal frequency resolution.

[0098] The center frequency and bandwidth of the adaptive trap filter can be dynamically adjusted according to the main frequency feature locking result and the signal frequency resolution respectively, thereby enhancing the adaptability of the trap filter to different signals and improving the filtering effect.

[0099] S6, the structured graph construction in the iteration termination and graph fusion includes a feature clustering analysis step, and the multiple iteration results are divided into different fault mode categories by calculating the similarity between features.

[0100] The feature clustering analysis step is included, the similarity between features is calculated, and the multiple iteration results are divided into different fault mode categories according to the similarity, so as to construct a structured graph;

[0101] The multiple iteration results are divided into different fault mode categories through feature clustering analysis, so that the structured graph can clearly present different fault modes, facilitating identification and analysis of the fault modes.

[0102] The scheme: the mechanical vibration signal is denoised, detrended, phase compensated and modal decomposed, a nonlinear resonance model with historical memory is constructed; the parameters and variable scale factors are initialized and the preprocessed signal is input; the phase compensation adopts sweep frequency method to determine the lag phase to tune the phase of the same frequency signal; the modal decomposition adopts the symplectic geometry method, and the effective components are reconstructed by means of the Pearson matrix;

[0103] The parameters are input into the homodirectional coupling neuron model to simulate the signal transmission process and the dynamic threshold adjustment process, and the response and step length are solved; the model simulates the change of transmission strength through dynamic update of synaptic weight;

[0104] Based on the zero-crossing behavior, energy distribution and time-frequency analysis result of the signal, a zero-energy coupling index is constructed to measure the fault enhancement effect; the time-frequency analysis includes short-time Fourier transform and wavelet packet energy spectrum calculation, which is used to extract energy features in different frequency bands;

[0105] The quantum optimization algorithm is used to iteratively optimize the parameters, and the optimal parameters are output by simulating quantum evolution and combining an adaptive mutation strategy with the zero-energy coupling index as the optimization objective; the algorithm prevents premature convergence by adjusting the rotation angle and maintaining population diversity;

[0106] Based on the optimal parameters, a three-dimensional response space including frequency, energy and time is constructed, and a multi-scale resonance response graph is generated by solving the response; the three-dimensional space is constructed by associating the model parameters and the response characteristics;

[0107] determining whether the optimal zero-energy coupling index is greater than a dynamic termination threshold; if yes, locking the dominant frequency, removing the identified signal by notch filtering, updating the input and iterating; if no, terminating the iteration, and integrating the multi-round features into a structured atlas according to frequency; wherein the dynamic threshold is updated according to the index variation trend; the dominant frequency locking determines the dominant frequency in combination with strong energy frequency and phase; the structured atlas arranges features according to frequency, and superimposes energy and phase to form multi-level results.

[0108] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any actual relationship or order between or among such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0109] While the embodiments of the application have been illustrated and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and alterations can be made therein without departing from the spirit and scope of the application, which is defined by the appended claims and their equivalents.

Claims

1. A multi-frequency weak fault enhancement method based on fractional order resonance response map reconstruction, characterized in that, It comprises the following steps: S1, signal preprocessing and model construction: Denoising, detrending, phase compensation and modal decomposition of mechanical vibration signal, construction of nonlinear resonance model with historical memory; initialization of parameters and variable scale factor, input of preprocessed signal; Phase compensation uses sweep frequency method to determine lagging phase, adjusts the phase of the same frequency signal; modal decomposition uses symplectic geometry method, and effective components are reconstructed by Pearson matrix; S2, model response solving: Input parameters into the same direction coupling neuron model, simulate signal transmission and dynamic threshold adjustment, solve response and step length; the model simulates the change of transmission strength through dynamic updating of synaptic weight; S3, evaluation index construction: Based on signal zero crossing behavior, energy distribution and time-frequency analysis, construct zero energy coupling index to measure the enhancement effect; time-frequency analysis includes short-time Fourier transform and wavelet packet energy spectrum calculation, and energy characteristics of different frequency bands are extracted; S4, parameter optimization: Iterative optimization of parameters by quantum optimization algorithm, simulation of quantum evolution combined with adaptive mutation strategy, taking zero energy coupling index as the target, outputting the optimal parameters; the algorithm adjusts the rotation angle and preserves the diversity of population to prevent premature convergence; S5, response atlas generation: Based on the optimal parameters, construct a three-dimensional response space containing frequency, energy and time, solve the response and generate a multi-scale resonance response atlas; the three-dimensional space is constructed by correlating model parameters and response characteristics; S6, iteration termination and atlas fusion: Determine whether the optimal zero energy coupling index is greater than the dynamic termination threshold; if yes, lock the main frequency, trap wave filter to remove the identified signal, and update the input iteration; Otherwise, terminate, integrate the multi-round characteristics into a structured atlas according to the frequency; update the dynamic threshold according to the exponential change trend; determine the main frequency by locking the main frequency, combining the strong energy frequency and phase; arrange the characteristics according to the frequency, superimpose the energy and phase to form a multi-level result.

2. The method of claim 1, wherein the method is characterized by: In the S1, signal preprocessing and model construction, the denoising method combines ensemble empirical mode decomposition and wavelet thresholding, and the effective components are selected by calculating the correlation coefficient of each intrinsic mode function.

3. The method of claim 1, wherein the method is characterized by: In the S1, signal preprocessing and model construction, the detrending method uses adaptive local iterative filtering method to realize effective separation of nonlinear trend item by constructing time-varying filter.

4. The method of claim 1, wherein the method is characterized by: In the S1, signal preprocessing and model construction, the nonlinear resonance model with historical memory realizes the function of historical memory through time series buffer mechanism, and the response result of previous time is integrated into the current model input according to time weight, and the weight decays with time to highlight the influence of recent memory.

5. The method of claim 1, wherein the method is characterized by: The dynamic threshold adjustment in S2, model response solving, includes neuron firing rate feedback mechanism, which dynamically adjusts the threshold parameter by monitoring the neuron cluster firing rate.

6. The method of claim 1, wherein the method is characterized by: The construction of zero energy coupling index in S3, evaluation index construction, includes three sub-steps of signal zero crossing rate feature extraction, energy entropy calculation and time-frequency feature fusion.

7. The method of claim 1, wherein the method is characterized by: The adaptive mutation strategy in S4, parameter optimization, includes gradient adjustment mechanism of quantum bit mutation probability, which dynamically adjusts the mutation strength according to the gradient value of objective function.

8. The method of claim 1, wherein the method is characterized by: The S5, the three-dimensional response space construction in the response atlas generation includes a parameter sensitivity analysis step, and the key parameter dimension is determined by calculating the influence degree of parameter change on system response.

9. The method of claim 1, wherein the method is characterized by: The S6, the trap wave filtering in the iteration termination and atlas fusion adopts an adaptive trap wave filter, the center frequency of which is dynamically adjusted according to the main frequency feature locking result, and the bandwidth is adaptively determined according to the signal frequency resolution.

10. The method of claim 1, wherein the method is based on a fractional order resonance response map reconstruction of a multi-frequency weak fault enhancement. The S6, the structured atlas construction in the iteration termination and atlas fusion includes a feature clustering analysis step, and the multiple iteration results are divided into different fault mode categories by calculating the similarity between features.

Images