Shock absorber data processing method based on digital twinning

Abstract

This invention discloses a data processing method for shock absorbers based on digital twins, comprising the following steps: generating damping force data and hysteresis loops through a digital twin model of the shock absorber; establishing unified hysteresis phase coordinates to form current hysteresis phase co-position residual samples; performing Wasserstein centroid processing on the healthy residual distribution to form a healthy reference sample for operating conditions; forming a Cucconi joint rank pool, calculating the Cucconi biplane rank response and Cucconi statistic to form a rank migration phase distribution; determining candidate correction amounts for candidate correction objects; updating the shock absorber digital twin model based on the established trial model, recalculating the Cucconi statistic and rank migration phase distribution to form a corrected digital twin model; and generating shock absorber correction results based on the corrected digital twin model. This invention improves the operating condition adaptation processing capability and data correction reliability of shock absorber operating data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a shock absorber data processing method based on digital twins. Background Technology

[0002] With the development of vehicle suspension system and shock absorber testing technology, data on displacement, velocity, damping force, oil temperature, load, and excitation frequency generated during shock absorber operation are increasingly being used for damping performance evaluation and operational status analysis. Current technologies typically establish digital twin models of the shock absorber, predict damping force based on shock absorber operating data and condition data, and use the residual between the measured and predicted damping forces to determine model deviation.

[0003] The damping response of shock absorbers exhibits significant hysteresis, making it difficult to directly compare data points corresponding to different operating cycles, excitation frequencies, and motion directions based on sampling time or displacement position. Existing methods, when employing time alignment or displacement normalization, are prone to miscorresponding to different hysteresis states during compression, recovery, and reversal, resulting in residual samples that cannot accurately reflect the deviation in damping response under the same hysteresis motion position. Furthermore, when oil temperature, load, and excitation frequency change, it is difficult to establish a healthy reference benchmark that matches the current operating conditions using fixed healthy samples or fixed judgment thresholds.

[0004] Meanwhile, existing model calibration methods directly adjust model parameters after detecting an increase in residuals, lacking targeted judgment on candidate correction objects, correction directions, and correction magnitudes. They also lack verification of recalculation before and after correction, which can easily lead to an overall decrease in error but an increase in local hysteresis phase deviation, or even the formation of new distribution migrations outside the original abnormal locations. This affects the working condition adaptability of shock absorber data processing, the accuracy of migration positioning, and the reliability of model calibration.

[0005] Therefore, how to provide a data processing method for shock absorbers based on digital twins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a shock absorber data processing method based on digital twins. This invention reduces the interference of different operating conditions and hysteresis paths on the processing results of shock absorber operating data by unifying hysteresis phase alignment, healthy operating condition reference reconstruction, and Cucconi rank migration discrimination. This enables accurate location of abnormal residuals and closed-loop correction of the digital twin model, thereby improving the operating condition adaptability, processing accuracy, and correction reliability of shock absorber operating data.

[0007] A shock absorber data processing method based on digital twin according to an embodiment of the present invention includes the following steps:

[0008] Acquire shock absorber operating data and working condition characterization data, and generate damping force data and hysteresis loop through the digital twin model of the shock absorber;

[0009] A unified hysteresis phase coordinate is established using the hysteresis loop, and the twin residuals of the shock absorber operation data and damping force data are mapped to the unified hysteresis phase coordinate to form the current hysteresis phase co-position residual sample.

[0010] The weights of healthy operating conditions are determined based on the operating condition characterization data, and the Wasserstein centroid processing is performed on the distribution of healthy residuals to form a reference sample of operating condition health.

[0011] The current hysteresis phase isotope residual sample and the working condition health reference sample are used to form a Cucconi joint rank pool. The Cucconi two-wing rank response and Cucconi statistic are calculated to form the rank migration phase distribution.

[0012] Based on the Cucconi statistic persistence state, Cucconi two-winged rank response, and rank transition phase distribution, the candidate correction amount for the candidate correction object is determined.

[0013] Based on the candidate correction values, construct a trial update digital twin model of the shock absorber, recalculate the Cucconi statistic and rank migration phase distribution, retain the candidate correction values ​​when the Cucconi statistic decreases, the rank migration phase distribution shrinks and no new rank migration phase distribution is formed, otherwise withdraw the candidate correction values ​​and form a corrected digital twin model.

[0014] The shock absorber calibration results are generated based on the calibration digital twin model.

[0015] Optionally, obtaining the damping force data and hysteresis loop specifically includes:

[0016] The shock absorber operation data and working condition characterization data are obtained according to a unified sampling period, and the sampling time is processed together to form time-aligned shock absorber operation data and time-aligned working condition characterization data.

[0017] Input the time-aligned shock absorber operating data and time-aligned working condition characterization data into the shock absorber digital twin model, and recursively deduce the hysteresis state of the shock absorber digital twin model according to the sampling time sequence to form a hysteresis state sequence.

[0018] The damping component, stiffness component, and hysteresis component are calculated based on the time-aligned damper operation data, time-aligned working condition characterization data, and hysteresis state sequence. They are then superimposed to form damping force data. The displacement values ​​in the time-aligned damper operation data and the damping force values ​​at the same sampling time in the damping force data are used to form hysteresis coordinate points. The hysteresis coordinate points are then connected in the order of sampling time to form a hysteresis loop.

[0019] Optionally, obtaining the current hysteresis phase co-position residual sample specifically includes:

[0020] Hysteresis cycle branching and energy-consuming phase modeling are performed on the hysteresis loop, and a unified hysteresis phase coordinate is formed based on the ratio of the cumulative damping work in the compressed phase branch and the restored phase branch to the total damping work of the corresponding phase branch.

[0021] The twin residuals between the measured damping force value in the shock absorber operating data and the corresponding damping force value in the damping force data are calculated. The damping degradation direction is made consistent for the twin residuals, and the processed twin residuals are mapped to a unified hysteresis phase coordinate to form a hysteresis phase oriented residual sequence.

[0022] Based on the unified hysteresis phase coordinates, the intra-branch co-position residual construction and periodic equalization processing are performed on the hysteresis phase orientation residual sequence. Co-position twin residuals are formed in the compressed phase branch and the restored phase branch respectively. The co-position twin residuals are arranged according to the complete motion period and co-position phase position to form the current hysteresis phase co-position residual sample.

[0023] Optionally, obtaining the working condition health reference sample specifically includes:

[0024] Perform health hysteresis phase trajectory construction and health residual distribution aggregation processing on the historical health operation data to form a set of health hysteresis phase residual trajectories and health residual distribution;

[0025] Based on the working condition adaptation relationship between the current working condition and the healthy working condition, as well as the phase integrity of the healthy hysteresis phase residual trajectory, the working condition adaptation weight is performed to form the healthy working condition weight, and the same set of healthy working condition weights are shared to all co-phase points in the unified hysteresis phase coordinate.

[0026] Based on the health condition weights, Wasserstein centroid processing under the same quantile index constraint is performed on the health residual distributions corresponding to all co-phase points to form a co-ordered quantile centroid residual sequence.

[0027] Virtual health residual trajectories are reconstructed along a unified hysteresis phase coordinate system for centroid residual values ​​with the same quantile number among all co-phase points to form virtual health residual trajectories. The virtual health residual trajectories corresponding to different quantile numbers are then arranged to form a working condition health reference sample.

[0028] Optionally, obtaining the rank-transfer phase distribution specifically includes:

[0029] The current hysteresis phase isotope residual sample and the operating condition health reference sample are subjected to source-preserving joint rank pooling to form Cucconi joint rank pool;

[0030] According to the co-phase position, the natural rank and reverse rank of the current hysteresis phase co-phase residual sample in the Cucconi joint rank pool are collected. For each co-phase position, the two-wing square rank response is calculated to form the Cucconi two-wing rank response. The Cucconi two-wing rank responses corresponding to all co-phase positions are accumulated to form the full-phase natural square rank response and the full-phase reverse square rank response.

[0031] The rank correlation correction coefficient is determined based on the number of residual values ​​in the Cucconi joint rank pool. The correlation correction joint calculation is performed on the full-phase natural square rank response and the full-phase inverse square rank response to form the Cucconi statistic.

[0032] The Cucconi double-wing rank response corresponding to each isophase position is decomposed with the full-phase natural square rank response, the full-phase reverse square rank response, and the rank correlation correction coefficient to form the rank migration phase contribution value. The rank migration phase contribution value is then mapped to the unified hysteresis phase coordinate to form the rank migration phase distribution.

[0033] Optionally, obtaining the candidate correction amount of the candidate correction object specifically includes:

[0034] Perform a persistence transition determination on the Cucconi statistic corresponding to the continuous processing window to form the persistence state of the Cucconi statistic.

[0035] When the Cucconi statistic is in a continuous migration state, the Cucconi two-wing rank response corresponding to each isophase position is extracted according to the corresponding rank migration phase ratio to form the current rank migration response feature.

[0036] The current rank migration response characteristics are matched with the corresponding positive unit phase rank response template and negative unit phase rank response template in the digital twin model of the shock absorber. Based on the template matching results, candidate correction objects and correction directions are determined. The candidate correction amount is determined by combining the response scaling degree, template matching confidence and Cucconi statistic persistence state.

[0037] Optionally, obtaining the corrected digital twin model specifically includes:

[0038] Read the candidate correction object, candidate correction amount, and current model parameters of the shock absorber digital twin model. Add the candidate correction amount to the current model parameters corresponding to the candidate correction object, keep the model parameters other than the candidate correction object unchanged, form the trial update model parameters, and write the trial update model parameters into a model copy with the same model structure as the shock absorber digital twin model to construct the trial update shock absorber digital twin model.

[0039] Based on the digital twin model of the test-updated shock absorber, perform the same source residual reconstruction and Cucconi statistical recalculation to form the test-updated Cucconi statistic and the test-updated rank migration phase distribution;

[0040] The Cucconi statistics before and after applying the candidate correction amount are subjected to rank migration shrinkage and newborn exclusion judgment to form the candidate correction amount verification results;

[0041] Based on the verification results of the candidate corrections, a trial update retention process is performed. When the candidate corrections meet the trial update rank migration closure condition, the candidate corrections are retained, and the trial update model parameters corresponding to the candidate correction object are written into the shock absorber digital twin model. The model parameters other than the candidate correction object remain unchanged to form a calibration digital twin model. When the candidate corrections do not meet the trial update rank migration closure condition, the candidate corrections are revoked, the trial update model parameters are deleted, the current model parameters corresponding to the candidate correction object are restored, and a calibration digital twin model is formed based on the restored model parameters.

[0042] Optionally, obtaining the shock absorber correction result specifically includes:

[0043] Based on the corrected digital twin model, the corrected response reconstruction process is performed to generate corrected damping force data and corrected hysteresis loop;

[0044] The corrected damping force data are subjected to the same source residual verification and Cucconi correction effect verification to form the corrected Cucconi statistic, the corrected rank migration phase distribution and the correspondence before and after correction.

[0045] The model state, correction response reconstruction results, and Cucconi correction effect verification results of the correction digital twin model are encapsulated to form the shock absorber correction results.

[0046] The beneficial effects of this invention are:

[0047] Based on the characteristics of shock absorber damping response, such as motion direction correlation, historical path dependence, and operating condition sensitivity, this invention establishes a unified hysteresis phase coordinate by the proportion of cumulative damping work in the hysteresis loop to the total damping work in the phase branch. Corresponding twin residuals are constructed in the compression and restoration phase branches respectively, enabling comparison of residuals in different complete motion cycles under the same motion direction and the same damping energy dissipation process. This avoids hysteresis state mismatch and commutation position overlap that occur when directly aligning using sampling time and displacement values, thereby improving the hysteresis phase alignment accuracy and residual sample comparability of shock absorber operating data.

[0048] This invention addresses the problem of fixed health reference benchmark failure caused by changes in oil temperature, load, and excitation frequency. It employs Wasserstein centroid processing to fuse health residual distributions corresponding to different health conditions, and shares the same set of health condition weights across all co-located phase points, reconstructing virtual health residual trajectories under the same quantile constraint. The Wasserstein centroid processing generates a health reference sample matching the current operating condition while preserving the location and scale characteristics of the health residual distribution. Sharing health condition weights and co-located quantile reconstruction avoids reference trajectory jumps caused by adjacent hysteresis phase points being dominated by different health condition combinations, thereby improving the operating condition adaptability, cross-phase continuity, and statistical comparison stability of the health reference sample.

[0049] This invention uses Cucconi statistics to perform joint rank comparison between the current hysteresis phase co-position residual sample and the operating condition health reference sample, so that the position migration and scale migration of the residual distribution can be identified synchronously within the same statistical framework. The overall distribution migration is mapped to a unified hysteresis phase coordinate by Cucconi biwing rank response and phase contribution decomposition, and the location and concentration of abnormal migration in the compressed phase branch and the restored phase branch are determined.

[0050] This invention matches the current rank migration response characteristics with the positive and negative unit phase rank response templates corresponding to the candidate correction objects to determine the candidate correction objects, correction directions, and candidate correction amounts with physical correspondence. It also performs recalculation by updating the digital twin model of the shock absorber, retaining the candidate correction amount only when the Cucconi statistic decreases, the original rank migration phase distribution shrinks, and no new rank migration phase is formed. This avoids the problem of only reducing the overall error while amplifying the local hysteresis phase deviation, thus improving the accuracy of migration identification of shock absorber operating data, the pertinence of model parameter correction, and the reliability of data correction. Attached Figure Description

[0051] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0052] Figure 1 This is a flowchart of a shock absorber data processing method based on digital twin proposed in this invention;

[0053] Figure 2 This is a schematic diagram of the closed-loop correction of the digital twin model based on Cucconi rank migration discrimination in the shock absorber data processing method based on digital twin proposed in this invention. Detailed Implementation

[0054] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0055] refer to Figures 1-2 A method for processing shock absorber data based on digital twins includes the following steps:

[0056] Acquire shock absorber operating data and working condition characterization data, and generate damping force data and hysteresis loop through the digital twin model of the shock absorber;

[0057] A unified hysteresis phase coordinate is established using the hysteresis loop, and the twin residuals of the shock absorber operation data and damping force data are mapped to the unified hysteresis phase coordinate to form the current hysteresis phase co-position residual sample.

[0058] The weights of healthy operating conditions are determined based on the operating condition characterization data, and the Wasserstein centroid processing is performed on the distribution of healthy residuals to form a reference sample of operating condition health.

[0059] The current hysteresis phase isotope residual sample and the working condition health reference sample are used to form a Cucconi joint rank pool. The Cucconi two-wing rank response and Cucconi statistic are calculated to form the rank migration phase distribution.

[0060] Based on the Cucconi statistic persistence state, Cucconi two-winged rank response, and rank transition phase distribution, the candidate correction amount for the candidate correction object is determined.

[0061] Based on the candidate correction values, construct a trial update digital twin model of the shock absorber, recalculate the Cucconi statistic and rank migration phase distribution, retain the candidate correction values ​​when the Cucconi statistic decreases, the rank migration phase distribution shrinks and no new rank migration phase distribution is formed, otherwise withdraw the candidate correction values ​​and form a corrected digital twin model.

[0062] The shock absorber calibration results are generated based on the calibration digital twin model.

[0063] In this embodiment, obtaining the damping force data and hysteresis loop specifically includes:

[0064] The shock absorber operating data and working condition characterization data are acquired according to a unified sampling period and then processed together to form time-aligned shock absorber operating data and time-aligned working condition characterization data. The shock absorber operating data is a set of data used to characterize the motion state and damping response state of the shock absorber, including displacement data, velocity data, and measured damping force data. The working condition characterization data is a set of data used to characterize the current operating environment and external excitation conditions of the shock absorber, including oil temperature data, load data, and excitation frequency data. The sampling time correspondence processing involves dividing the shock absorber operating data and working condition characterization data to the same sampling time according to the data timestamp, and establishing a correspondence between the data records belonging to the same sampling time.

[0065] The time-aligned shock absorber operating data and time-aligned working condition characterization data are input into the shock absorber digital twin model. The hysteresis state of the shock absorber digital twin model is recursively deduced according to the sampling time sequence to form a hysteresis state sequence. The hysteresis state is an internal state that characterizes the degree of dependence of the shock absorber's damping response on the historical motion path.

[0066] The digital twin model of the shock absorber includes a motion state input layer, a working condition parameter mapping layer, a hysteresis state recursion layer, a damping force synthesis layer, and a hysteresis loop output layer. The motion state input layer receives a fixed-length motion state matrix corresponding to the time-aligned shock absorber operating data. The number of rows in the fixed-length motion state matrix corresponds to the number of consecutive sampling moments, and the number of columns corresponds to the motion state dimension composed of displacement and velocity values. The working condition parameter mapping layer receives a fixed-length working condition matrix corresponding to the time-aligned working condition characterization data and outputs damping parameters, stiffness parameters, hysteresis contribution parameters, and hysteresis evolution parameters. The hysteresis state recursion layer is connected to both the motion state input layer and the working condition parameter mapping layer, and is used to recursively deduce the hysteresis state at the current sampling moment based on the velocity value, hysteresis evolution parameters, and the hysteresis state at the previous sampling moment. The damping force synthesis layer is connected to the motion state input layer, the working condition parameter mapping layer, and the hysteresis state recursion layer, and is used to superimpose damping components, stiffness components, and hysteresis. The components are combined to form damping force data. The hysteresis loop output layer is connected to the damping force synthesis layer to form hysteresis coordinate points and hysteresis loops by combining displacement values ​​with damping force values ​​at the same sampling time. The training data comes from shock absorber bench test data and actual operating history data of shock absorbers. The training samples include shock absorber operating data, working condition characterization data and corresponding measured damping force labels. The measured damping force labels are determined by the force sensor acquisition results. The training loss includes damping force prediction loss, hysteresis loop structure loss, hysteresis state period consistency loss and model parameter smoothing loss. The trainable parameters of the working condition parameter mapping layer and the damping force synthesis layer are updated according to the training loss. The training parameters include learning rate, batch sample size, training window length, loss weight and maximum number of training rounds. Model training stops when the change in validation loss corresponding to consecutive training rounds is not greater than the training convergence threshold and the damping force prediction error is not greater than the preset prediction error threshold.

[0067] Damping, stiffness, and hysteresis components are calculated based on time-aligned damper operating data, time-aligned working condition characterization data, and hysteresis state sequences. These components are then superimposed to form damping force data. Hysteresis coordinate points are formed by combining displacement values ​​from the time-aligned damper operating data with damping force values ​​at the same sampling time from the damping force data. These hysteresis coordinate points are then connected in the order of sampling times to form a hysteresis loop. The damping component is the product of the damping parameter corresponding to the current working condition and the velocity value at the current sampling time. The stiffness component is the product of the stiffness parameter corresponding to the current working condition and the displacement value at the current sampling time. The hysteresis component is the product of the hysteresis contribution parameter corresponding to the current working condition and the hysteresis state. The damping parameter is used to characterize the response relationship of the damper's damping force as the motion speed changes. The stiffness parameter is used to characterize the response relationship of the damper's damping force as the displacement changes. The hysteresis contribution parameter is used to characterize the degree of contribution of the hysteresis state to the damping force. The damping force data is all damping force values ​​arranged in the order of sampling time. The hysteresis loop is a closed or quasi-closed response trajectory formed by connecting the displacement value as the horizontal coordinate and the damping force value as the vertical coordinate according to the movement process of the damper.

[0068] The unified sampling period is set to 1 ms, the duration of a single processing window is set to 10 s, each processing window contains 10,000 sampling moments, the size of the fixed-length motion state matrix is ​​set to 10,000×2, the size of the fixed-length working condition matrix is ​​set to 10,000×3, the acquisition range of displacement data is set to -50 mm to 50 mm, the acquisition range of velocity data is set to -1.2 m / s to 1.2 m / s, and the acquisition range of measured damping force data is set to -6000 N to 6000 N.

[0069] In this embodiment, obtaining the current hysteresis phase co-position residual sample specifically includes:

[0070] Hysteresis loops are subjected to hysteresis cycle branching and energy dissipation phase modeling. Based on the proportion of cumulative damping work in the compression phase branch and the recovery phase branch to the total damping work of the corresponding phase branch, a unified hysteresis phase coordinate is formed. The hysteresis cycle branching is performed by determining the complete motion cycle according to the velocity direction and displacement extreme value corresponding to the hysteresis loop, and dividing the complete motion cycle into compression phase branch and recovery phase branch. The energy dissipation phase modeling is performed by calculating the damping work increment between adjacent hysteresis coordinate points, and determining the hysteresis phase value according to the proportion of cumulative damping work. The compression phase branch corresponds to the hysteresis phase range of zero to half, and the recovery phase branch corresponds to the hysteresis phase range of half to one. The damping work is the energy consumption formed by the damping force acting on the displacement change of the shock absorber in the hysteresis loop. The compression phase branch is the hysteresis loop branch corresponding to the movement of the shock absorber from the extension position to the compression position, and the recovery phase branch is the hysteresis loop branch corresponding to the movement of the shock absorber from the compression position to the extension position.

[0071] The twin residuals between the measured damping force value in the shock absorber operating data and the corresponding damping force value in the damping force data are calculated. The damping degradation direction is consistentized by performing damping degradation direction unification processing on the twin residuals, and the processed twin residuals are mapped to a unified hysteresis phase coordinate to form a hysteresis phase oriented residual sequence. The twin residuals are the differences between the measured damping force value in the shock absorber operating data and the damping force value generated by the digital twin model of the shock absorber at the same sampling time. They are used to characterize the degree of deviation between the actual damping response and the digital twin damping response of the shock absorber. The damping degradation direction unification processing is to adjust the positive and negative directions of the twin residuals according to the direction of the corresponding damping force value, so that the twin residuals representing the reduction of damping capacity in the compression phase branch and the recovery phase branch change in the same direction. The hysteresis phase oriented residual sequence is a residual sequence in which the twin residuals processed in the same direction correspond to the complete motion cycle, phase branch and hysteresis phase position.

[0072] Based on the unified hysteresis phase coordinates, the hysteresis phase orientation residual sequence is subjected to intra-branch isotope residual construction and period equalization processing. Inotope twin residuals are formed in the compressed phase branch and the restored phase branch respectively. The isotope twin residuals are arranged according to the complete motion cycle and the isotope phase position to form the current hysteresis phase isotope residual sample. The intra-branch isotope residual construction is as follows: isotope phase points are set in the compressed phase branch and the restored phase branch respectively. The isotope twin residuals are determined according to the twin residuals corresponding to the adjacent hysteresis phase positions on both sides of the isotope phase point and the energy dissipation phase distance ratio. It is prohibited to construct isotope twin residuals across the reversal position between the compressed phase branch and the restored phase branch. The period equalization processing is as follows: a continuous complete motion cycle in which all isotope phase points have effective isotope twin residuals is selected, so that each isotope phase point retains the same number of isotope twin residuals from the same complete motion cycle. The isotope twin residuals are twin residuals in the same phase branch and the same cumulative damping energy dissipation ratio in different complete motion cycles.

[0073] The compressed phase branch has 50 co-phase points, the restored phase branch has 50 co-phase points, and the unified hysteresis phase coordinates contain a total of 100 co-phase points. The hysteresis phase interval between adjacent co-phase points in the compressed phase branch is set to 0.01, and the hysteresis phase interval between adjacent co-phase points in the restored phase branch is set to 0.01.

[0074] This invention establishes a unified hysteresis phase coordinate by accumulating the damping energy dissipation ratio, and performs degradation direction unification, intra-branch isotope construction, and period equalization on the twin residuals in the compression phase branch and the restoration phase branch. This avoids residual mismatch caused by different motion cycles, motion directions, and reversal positions, and enables residual samples to be compared under the same hysteresis motion position and energy dissipation process, thereby improving the phase alignment accuracy, residual comparability, and anomaly identification reliability of the shock absorber operating data.

[0075] In this embodiment, obtaining the working condition health reference sample specifically includes:

[0076] The health operation history data is processed by constructing health hysteresis phase trajectories and aggregating health residual distributions to form a set of health hysteresis phase residual trajectories and a health residual distribution. The health hysteresis phase trajectory is constructed by establishing a unified hysteresis phase coordinate based on the complete motion cycle of the hysteresis loop, the compressed phase branch, the restored phase branch, and the cumulative damping energy consumption ratio. The twin residuals in the health operation history data are processed by unifying the damping degradation direction and constructing intra-branch co-position residuals. The health co-position twin residuals corresponding to all co-position phase points within the same complete health motion cycle are arranged along the unified hysteresis phase coordinates to form a health hysteresis phase residual trajectory. The health residual distribution aggregating process is to aggregate health co-position twin residuals according to the health condition and co-position phase points to form a health residual distribution corresponding to each co-position phase point for each health condition.

[0077] Based on the operating condition adaptation relationship between the current operating condition and the healthy operating condition, and the phase integrity of the healthy hysteresis phase residual trajectory, operating condition adaptation weighting is performed to form healthy operating condition weights. The same set of healthy operating condition weights is shared to all co-phase points in the unified hysteresis phase coordinate system. The operating condition adaptation relationship is determined based on the differences in oil temperature, load, and excitation frequency between the current operating condition characterization data and the corresponding operating condition characterization data of the healthy operation history data. The phase integrity is determined based on the proportion of co-phase points with effective healthy co-twin residuals in the healthy hysteresis phase residual trajectory to the total number of co-phase points. The operating condition adaptation weighting makes the healthy operating condition weight decrease as the operating condition difference increases and increase as the phase integrity increases. The sharing of healthy operating condition weights means that all co-phase points use the same healthy operating condition weight combination.

[0078] Based on health condition weights, Wasserstein centroid processing under the same quantile number constraint is performed on the health residual distributions corresponding to all co-phase points to form a co-order quantile centroid residual sequence. The same quantile number constraint is to set the quantile position sequence according to the number of complete motion cycles contained in the current hysteresis phase co-phase residual sample, and to make all co-phase points adopt the same quantile position sequence. The Wasserstein centroid processing is to fuse the health co-phase twin residuals corresponding to each health residual distribution according to the health condition weights for the same co-phase point and the same quantile number to form the corresponding centroid residual value. The co-order quantile centroid residual sequence is a sequence of centroid residual values ​​arranged according to the quantile number within the same co-phase point.

[0079] Virtual health trajectory reconstruction is performed on the centroid residual values ​​with the same quantile number among all co-phase points along a unified hysteresis phase coordinate to form a virtual health residual trajectory. The virtual health residual trajectories corresponding to different quantile numbers are arranged to form a working condition health reference sample. The virtual health trajectory reconstruction is as follows: under the same health working condition weight combination and the same quantile number constraint, the corresponding centroid residual values ​​in the compressed phase branch and the restored phase branch are connected in the order of co-phase points. Each row of the working condition health reference sample corresponds to a virtual health residual trajectory, and each column corresponds to a co-phase point in the unified hysteresis phase coordinate. The working condition health reference sample and the current hysteresis phase co-phase residual sample have the same number of rows, columns, and hysteresis phase position correspondence.

[0080] The historical data of healthy operation is divided into 18 healthy operating conditions according to oil temperature, load, and excitation frequency. The oil temperature is set to 30 ℃, 60 ℃, and 90 ℃, the load is set to 1 kN, 3 kN, and 5 kN, and the excitation frequency is set to 2 Hz and 5 Hz. Each healthy operating condition retains 30 complete healthy motion cycles. The normalization benchmark for oil temperature difference is set to 10 ℃, the normalization benchmark for load difference is set to 0.5 kN, and the normalization benchmark for excitation frequency difference is set to 1 Hz. From the 18 healthy operating conditions, the 6 healthy operating conditions with the smallest operating condition difference and a phase integrity of not less than 0.95 are selected to determine the weight of the healthy operating conditions. The sum of the normalized weights of the 6 healthy operating conditions is set to 1, and the weights of the same group of healthy operating conditions are shared among 100 co-phase points.

[0081] This invention reconstructs the residual distribution under different health conditions into a virtual health residual trajectory that is continuous across hysteresis phases by sharing health condition weights and the same quantile number constraint through Wasserstein centroid processing. This avoids reference benchmark jumps caused by switching of health condition weights between adjacent phase points, and ensures that the health condition reference sample and the current hysteresis phase co-position residual sample maintain structural consistency, thereby improving the health benchmark of the shock absorber in terms of health condition adaptability, phase continuity, and residual comparison reliability.

[0082] In this embodiment, obtaining the rank migration phase distribution specifically includes:

[0083] The current hysteresis phase co-position residual sample and the working condition health reference sample are subjected to source-preserving joint rank pooling construction processing to form Cucconi joint rank pool. The source-preserving joint rank pooling construction processing is as follows: the residual values ​​in the two samples are merged, the sample source and co-position position corresponding to the residual value are preserved, the natural rank and reverse rank are determined according to the size of the residual value, and the average of the ranks occupied by the residual values ​​with the same value are used. The natural rank is determined in ascending order of residual values, and the reverse rank is determined in the reverse order. The Cucconi joint rank pool preserves the sample source and co-position position corresponding to each residual value and is used to determine the natural rank and reverse rank of the residual value.

[0084] The natural rank and reverse rank of the current hysteresis phase in-phase residual samples in the Cucconi joint rank pool are collected according to their in-phase positions. For each in-phase position, a two-winged square rank response is calculated to form a Cucconi two-winged rank response. The Cucconi two-winged rank responses corresponding to all in-phase positions are accumulated to form a full-phase natural square rank response and a full-phase reverse square rank response. The two-winged square rank response is calculated by performing theoretical center correction and uniform scale standardization on the natural rank sum of squares and the reverse rank sum of squares, respectively. The natural square rank response is used to characterize the degree of migration of the current hysteresis phase in-phase residual samples to the high end of the Cucconi joint rank pool, and the reverse square rank response is used to characterize the degree of migration of the current hysteresis phase in-phase residual samples to the low end of the Cucconi joint rank pool.

[0085] The theoretical center correction is achieved by subtracting the theoretical expected value corresponding to the absence of distribution migration between the current hysteresis phase co-position residual sample and the operating condition health reference sample from the natural rank sum of squares and the reverse rank sum of squares, respectively, to eliminate the basic offset caused by the sample size and rank range. The unified scale standardization is achieved by scaling the square rank deviation after theoretical center correction based on the number of residual values ​​of the current hysteresis phase co-position residual sample, the number of reference residual values ​​of the operating condition health reference sample, and the total number of residual values ​​of the Cucconi joint rank pool, so that the natural square rank response and the reverse square rank response corresponding to different co-position phase positions are on a comparable numerical scale.

[0086] The rank correlation correction coefficient is determined based on the number of residual values ​​in the Cucconi joint rank pool. Correlation correction joint calculation is performed on the full-phase natural square rank response and the full-phase backward square rank response to form the Cucconi statistic. The rank correlation correction coefficient is used to eliminate the inherent correlation between the natural square rank response and the backward square rank response. The correlation correction joint calculation is to synthesize the squared deviations of the full-phase natural square rank response and the full-phase backward square rank response and deduct the crossover part determined by the rank correlation correction coefficient. The Cucconi statistic is used to characterize the degree of joint residual position migration and residual scale migration of the current hysteresis phase isotopic residual sample relative to the working condition health reference sample.

[0087] The Cucconi double-wing rank response corresponding to each co-phase position is decomposed into a rank migration phase contribution value by performing phase contribution decomposition with the full-phase natural square rank response, the full-phase reverse square rank response, and the rank correlation correction coefficient. The rank migration phase contribution value is then mapped to a unified hysteresis phase coordinate to form a rank migration phase distribution. The phase contribution decomposition involves distributing the Cucconi statistic to the corresponding co-phase position according to the natural square rank response, the reverse square rank response, and their correlation with the full-phase double-wing rank response, so that the sum of the rank migration phase contribution values ​​corresponding to all co-phase positions is consistent with the Cucconi statistic. The rank migration phase distribution consists of the rank migration phase contribution values ​​and their corresponding proportions arranged according to the unified hysteresis phase coordinate, which are used to determine the occurrence location and concentration of residual position migration and residual scale migration in the compressed phase branch and the restored phase branch.

[0088] This invention uses Cucconi combined with rank pooling to synchronously characterize the positional and scale migration of the current hysteresis phase co-position residual sample and the working condition health reference sample. It uses the biwing rank response to distinguish the deviation between the high-value end and the low-value end, and decomposes the overall statistical deviation into a unified hysteresis phase coordinate. This accurately locates the location and concentration of anomalies in the compression phase branch and the restoration phase branch, improving the comprehensiveness of shock absorber residual distribution identification, phase positioning accuracy and data processing reliability.

[0089] In this embodiment, obtaining the candidate correction amount of the candidate correction object specifically includes:

[0090] A continuous migration determination is performed on the Cucconi statistic corresponding to the continuous processing window to form a continuous state of Cucconi statistic. The continuous migration determination is to calculate the proportion of windows in the continuous processing window where the Cucconi statistic reaches a preset statistical migration threshold, and to calculate the average degree of excess of the Cucconi statistic relative to the preset statistical migration threshold. When the window proportion reaches the preset continuous proportion threshold, it is determined to be a continuous migration state; when it does not reach the preset continuous proportion threshold, it is determined to be a non-continuous migration state.

[0091] When the Cucconi statistic is in a continuous migration state, the Cucconi two-wing rank response corresponding to each isophase position is extracted according to the corresponding rank migration phase ratio to form the current rank migration response feature. The two-wing phase response extraction is to perform joint amplitude normalization on the natural square rank response and the inverse square rank response, and then multiply the normalized natural square rank response and the inverse square rank response by the rank migration phase ratio to form the high-value wing phase response and the low-value wing phase response. The current rank migration response feature is the high-value wing phase response and the low-value wing phase response arranged according to the unified hysteresis phase coordinate.

[0092] The current rank migration response feature is matched with the corresponding positive and negative unit phase rank response templates in the digital twin model of the shock absorber using response scaling. Candidate correction objects and directions are determined based on the template matching results. The candidate correction amount is then determined by combining the response scaling degree, template matching confidence, and Cucconi statistic persistence state. The response scaling matching involves calculating the response scaling coefficient and template matching error between the current rank migration response feature and each unit phase rank response template. The candidate correction object and direction corresponding to the unit phase rank response template with the smallest template matching error are taken as the matching result. The candidate correction objects include damping parameters, stiffness parameters, hysteresis contribution parameters, and hysteresis evolution parameters. The template matching confidence is determined based on the proportion of the template matching error to the total response energy of the current rank migration response feature. The candidate correction amount is obtained by multiplying the unit correction amount corresponding to the candidate correction object by the response scaling coefficient, template matching confidence, and persistence migration severity, and then limiting the correction amount according to the maximum allowable correction amount. When the Cucconi statistic persistence state is a non-persistent migration state, the candidate correction amount is set to zero.

[0093] In this embodiment, obtaining the corrected digital twin model specifically includes:

[0094] Read the candidate correction object, candidate correction amount, and the current model parameters of the shock absorber digital twin model. Superimpose the candidate correction amount onto the current model parameters corresponding to the candidate correction object, keep the model parameters other than the candidate correction object unchanged, form the trial update model parameters, and write the trial update model parameters into a model copy with the same model structure as the shock absorber digital twin model to construct the trial update shock absorber digital twin model. The trial update model parameters are the sum of the current model parameters corresponding to the candidate correction object and the candidate correction amount. The trial update shock absorber digital twin model only changes the model parameters corresponding to the candidate correction object and continues to use the shock absorber operating data, working condition characterization data, and model input structure used when forming the candidate correction amount, so that the calculation results before and after the trial update only reflect the model response changes caused by the candidate correction amount.

[0095] Based on the digital twin model of the test-updated damper, the same source residual reconstruction and Cucconi statistical recalculation are performed to form the test-updated Cucconi statistic and the test-updated rank migration phase distribution. The same source residual reconstruction is as follows: the damper operating data and working condition characterization data used when forming candidate corrections are input into the digital twin model of the test-updated damper to generate test-updated damping force data and test-updated hysteresis loop. The test-updated twin residual is formed according to the measured damping force value and the test-updated damping force data, and the unified hysteresis phase coordinates and damping degradation direction consistent with the current hysteresis phase co-position residual samples are adopted. The trial update hysteresis phase isotope residual sample is formed by the uniformization method, the intra-branch isotope residual construction method and the periodic balancing method. The Cucconi statistical recalculation process is as follows: the trial update hysteresis phase isotope residual sample and the working condition health reference sample are combined to form the trial update Cucconi joint rank pool. According to the same natural rank and reverse rank determination rules, theoretical center correction rules, unified scale standardization rules, rank correlation correction rules and phase contribution decomposition rules, the trial update Cucconi two-wing rank response, the trial update Cucconi statistic and the trial update rank migration phase distribution are formed in sequence.

[0096] The Cucconi statistics before and after the application of the candidate correction are subjected to rank migration shrinkage and newborn exclusion judgment to form the candidate correction verification result. The rank migration shrinkage and newborn exclusion judgment is as follows: compare the Cucconi statistics before the trial update with the Cucconi statistics before the application of the candidate correction. When the Cucconi statistics before the trial update are small, it is determined that the Cucconi statistics have decreased. According to the same preset phase migration threshold, the original rank migration phase set and the trial update rank migration phase set are extracted from the rank migration phase distribution and the trial update rank migration phase distribution, respectively. The effective rank migration phase contribution values ​​before and after the application of the candidate correction in the original rank migration phase set are accumulated to form the original migration phase set. The total migration contribution and the retained amount of the trial update migration contribution are considered. When the retained amount of the trial update migration contribution is less than the total original migration contribution and the number of co-positional phase positions contained in the trial update rank migration phase set is not greater than the number of co-positional phase positions contained in the original rank migration phase set, the rank migration phase distribution is determined to shrink. The positional inclusion relationship between the trial update rank migration phase set and the original rank migration phase set is compared. When there are no co-positional phase positions outside the original rank migration phase set in the trial update rank migration phase set, it is determined that no new rank migration phase distribution has been formed. The verification results of the candidate correction amount include the Cucconi statistic decrease result, the rank migration phase distribution shrinkage result, and the new rank migration phase distribution formation result.

[0097] Based on the verification results of the candidate correction amount, a trial update retention process is performed. When the candidate correction amount meets the trial update rank migration closure condition, the candidate correction amount is retained, and the trial update model parameters corresponding to the candidate correction object are written into the shock absorber digital twin model. The model parameters other than the candidate correction object remain unchanged to form a calibration digital twin model. When the candidate correction amount does not meet the trial update rank migration closure condition, the candidate correction amount is revoked, the trial update model parameters are deleted, the current model parameters corresponding to the candidate correction object are restored, and a calibration digital twin model is formed based on the restored model parameters. The trial update rank migration closure condition is that the trial update Cucconi statistic is less than the Cucconi statistic before the candidate correction amount is applied, the trial update migration contribution retention amount is less than the original migration contribution total amount, the number of co-phase positions contained in the trial update rank migration phase set is not greater than the number of co-phase positions contained in the original rank migration phase set, and the trial update rank migration phase set does not contain any co-phase positions located outside the original rank migration phase set.

[0098] This invention constructs a trial update model attributable to a single correction factor by applying only the candidate correction amount to the candidate correction object while keeping the other model parameters and input conditions unchanged. Combined with the closed-loop verification of Cucconi statistic decrease, original rank migration phase distribution shrinkage and new rank migration exclusion, it avoids the overall error reduction but the expansion of local hysteresis deviation or abnormal migration, thereby improving the pertinence, stability and reliability of the parameter correction of the shock absorber digital twin model.

[0099] In this embodiment, obtaining the shock absorber correction result specifically includes:

[0100] The correction response reconstruction process is performed based on the correction digital twin model to form correction damping force data and correction hysteresis loop. The correction response reconstruction process involves inputting the shock absorber operating data and working condition characterization data into the correction digital twin model to generate correction damping force data, and forming a correction hysteresis loop based on the displacement value and the corresponding correction damping force value.

[0101] The corrected damping force data is subjected to co-source residual verification and Cucconi correction effect verification to form the corrected Cucconi statistic, the corrected rank migration phase distribution, and the correspondence before and after correction. The co-source residual verification is to form the corrected hysteresis phase isotope residual sample based on the measured damping force value and the corrected damping force data. The Cucconi correction effect verification is to form the corrected Cucconi joint rank pool with the corrected hysteresis phase isotope residual sample and the working condition health reference sample to generate the corrected Cucconi statistic and the corrected rank migration phase distribution, and establish a correspondence with the Cucconi statistic and rank migration phase distribution before the candidate correction amount is applied.

[0102] The model state, correction response reconstruction results, and Cucconi correction effect verification results of the correction digital twin model are encapsulated to form the damper correction results. The damper correction results are used to record candidate correction objects, retained candidate correction amounts, and correction model parameters, characterize the damping force response and hysteresis loop changes generated by the correction digital twin model, and reflect the degree of improvement of residual distribution migration through the correction Cucconi statistic, correction rank migration phase distribution, and correspondence before and after correction, providing data basis for damper damping performance evaluation, operation status analysis, and continuous calibration of the digital twin model.

[0103] Example 1: To verify the feasibility of this invention in practice, it was applied to a scenario involving the processing of operational data and the calibration of a digital twin model for a twin-cylinder hydraulic shock absorber test bench. The rated damping force of the test object was 4500 N, and the maximum working stroke was 100 mm. The damping capacity decay of the shock absorber after long-term operation was simulated by adjusting the internal throttling state. During the test, the displacement amplitude was set to 35 mm, the load to 3.2 kN, the excitation frequency to 4.5 Hz, and the oil temperature was stabilized at around 72.8℃. Displacement data, velocity data, measured damping force data, oil temperature data, load data, and excitation frequency data were collected with a uniform sampling period of 1 ms. Each processing window lasted for 10 s and contained 10,000 sampling moments.

[0104] The digital twin model of the shock absorber generates damping force data and hysteresis loops based on displacement, velocity, and operating condition characterization data that progressively hysteresis. Before correction, the root mean square error between the measured damping force and the model-generated damping force was 286 N, and the maximum absolute error was 684 N, with significant deviations in the middle section of the compression phase branch and the front section of the recovery phase branch. When directly comparing sampling times, the reversing positions of different complete motion cycles could not be accurately matched, and some residuals from different hysteresis states were incorrectly assigned to the same data position.

[0105] This invention establishes a unified hysteresis phase coordinate system based on the ratio of the cumulative damping work in the compressed phase branch to the total damping work in the corresponding phase branch. Fifty co-position phase points are set in each of the two phase branches, and ten complete motion cycles are selected from the processing window to form a 10-row, 100-column sample of the current hysteresis phase co-position residuals. After the twin residuals are aligned in the damping degradation direction and constructed as co-positions within the branches, the average phase correspondence error of the continuous cycle is reduced from 0.047 using the sampling time alignment method to 0.012, avoiding residual mismatches caused by different motion directions and reversal positions.

[0106] The historical health operation data includes 18 health conditions, each containing 30 complete motion cycles. Based on the current oil temperature, load, and excitation frequency, six health conditions with a phase integrity of no less than 0.95 are selected, and their normalized health condition weights are 0.28, 0.22, 0.18, 0.14, 0.10, and 0.08, respectively. The weights of the same set of health conditions are shared across all corresponding phase points. Wasserstein centroid processing is performed according to 10 identical quantile numbers, forming 10 virtual health residual trajectories and a 10x100 column health reference sample. When the health condition weights are determined independently for each corresponding phase point, the maximum reference residual jump between adjacent phases is 10⁴ N; this is reduced to 39 N after applying this invention.

[0107] The 1000 co-twin residuals and 1000 reference residuals were combined to form the Cucconi joint rank pool. Before correction, the Cucconi statistic was 12.84, exceeding the preset statistical migration threshold of 5.99. 21 co-phase points reached the phase migration threshold, mainly concentrated in the compressed phase range of 0.18 to 0.31 and the restored phase range of 0.67 to 0.76. Four out of five consecutive processing windows exceeded the statistical migration threshold, resulting in a statistical migration persistence rate of 0.80.

[0108] The current rank-transfer response characteristics are matched with the positive and negative unit-phase rank response templates corresponding to the damping parameter, stiffness parameter, hysteresis contribution parameter, and hysteresis evolution parameter. The negative unit-phase rank response template for the damping parameter has the smallest matching error, with a template matching confidence level of 0.89. The candidate correction is -4.6% of the current value of the damping parameter. Only the candidate correction is applied to the damping parameter while keeping the remaining model parameters unchanged. The Cucconi statistic is updated to 4.76, the effective contribution within the original rank-transfer phase set decreases from 12.84 to 4.21, the number of co-phase points reaching the phase-transfer threshold decreases from 21 to 6, and the number of newly added co-phase points outside the original rank-transfer phase set is 0. Therefore, the candidate correction is retained, and a corrected digital twin model is formed.

[0109] To demonstrate the beneficial effects of this invention, it is compared with conventional and control methods. The conventional method establishes residual correspondence by aligning sampling times, uses fixed healthy reference samples, and directly adjusts the parameters of the digital twin model of the shock absorber based on the mean square error of the damping force. The control method aligns residuals of different motion cycles using displacement normalization, independently determines the weights of the healthy operating conditions at each co-phase point, performs Wasserstein centroid processing, and adjusts model parameters based on global statistical results. However, it does not employ shared healthy operating condition weights, same quantile index constraints, Cucconi double-wing rank response phase contribution decomposition, or trial update rank migration closure verification. The specific comparison results are shown in Table 1.

[0110] Table 1 Comparison of Hysteresis Phase Alignment and Model Correction Performance of Different Methods

[0111] Continuous period average phase correspondence error 0.047 0.031 0.012 Health reference adjacent phase maximum residual jump variable / N 168 104 39 Corrected root mean square error of damping force / N 196 164 118 Corrected Cucconi statistic 8.92 6.41 4.76 Number of co-phase points that reach the migration threshold after correction / 16 11 6 Number of new corresponding phase points added outside the original rank shift phase set / 5 2 0 Average processing time per processing window / s 0.41 0.63 0.86

[0112] As shown in Table 1, the average phase correspondence error of the continuous cycle in this invention is 0.012, which is 74.47% and 61.29% lower than that of the conventional method (0.047) and the control method (0.031), respectively. This indicates that the unified hysteresis phase coordinates established based on the cumulative damping energy consumption ratio can improve the phase correspondence accuracy between different motion cycles. The maximum residual jump variable of adjacent phases in the healthy reference of this invention is 39 N, which is 76.79% and 62.50% lower than that of the conventional method (168 N) and the control method (104 N), respectively. This indicates that the shared healthy condition weights and the same-order quantile Wasserstein centroid processing can maintain the phase continuity of the virtual healthy residual trajectory. After model correction, the root mean square error of damping force in this invention is 118 N, which is lower than that of the conventional method (196 N) and the control method (164 N). The Cucconi statistic after correction is reduced from 12.84 to 4.76, which is lower than the preset statistical migration threshold of 5.99. The number of co-phase points reaching the migration threshold was reduced to 6, and the number of newly added co-phase points outside the original rank migration phase set was 0. This indicates that the present invention can not only reduce the overall damping force prediction error, but also shrink the original rank migration phase distribution and prevent abnormal migration from shifting to other hysteresis phase positions. Although the processing window time of the present invention is 0.86 s, which is higher than the conventional method's 0.41 s and the control method's 0.63 s, it is still significantly lower than the processing window duration of 10 s, and can meet the timeliness requirements of shock absorber operation data processing and continuous correction of digital twin models.

[0113] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A shock absorber data processing method based on digital twins, characterized in that, Includes the following steps: Acquire shock absorber operating data and working condition characterization data, and generate damping force data and hysteresis loop through the digital twin model of the shock absorber; A unified hysteresis phase coordinate is established using the hysteresis loop, and the twin residuals of the shock absorber operation data and damping force data are mapped to the unified hysteresis phase coordinate to form the current hysteresis phase co-position residual sample. The weights of healthy operating conditions are determined based on the operating condition characterization data, and the Wasserstein centroid processing is performed on the distribution of healthy residuals to form a reference sample of operating condition health. The current hysteresis phase isotope residual sample and the working condition health reference sample are used to form a Cucconi joint rank pool. The Cucconi two-wing rank response and Cucconi statistic are calculated to form the rank migration phase distribution. Based on the Cucconi statistic persistence state, Cucconi two-winged rank response, and rank migration phase distribution, the candidate correction amount for the candidate correction object is determined. Based on the candidate correction values, construct a trial update digital twin model of the shock absorber, recalculate the Cucconi statistic and rank migration phase distribution, retain the candidate correction values ​​when the Cucconi statistic decreases, the rank migration phase distribution shrinks and no new rank migration phase distribution is formed, otherwise withdraw the candidate correction values ​​and form a corrected digital twin model. The shock absorber calibration results are generated based on the calibration digital twin model.

2. The shock absorber data processing method based on digital twin according to claim 1, characterized in that, The specific methods for obtaining the damping force data and hysteresis loop include: The shock absorber operation data and working condition characterization data are obtained according to a unified sampling period, and the sampling time is processed together to form time-aligned shock absorber operation data and time-aligned working condition characterization data. Input the time-aligned shock absorber operating data and time-aligned working condition characterization data into the shock absorber digital twin model, and recursively deduce the hysteresis state of the shock absorber digital twin model according to the sampling time sequence to form a hysteresis state sequence. The damping component, stiffness component, and hysteresis component are calculated based on the time-aligned damper operation data, time-aligned working condition characterization data, and hysteresis state sequence. They are then superimposed to form damping force data. The displacement values ​​in the time-aligned damper operation data and the damping force values ​​at the same sampling time in the damping force data are used to form hysteresis coordinate points. The hysteresis coordinate points are then connected in the order of sampling time to form a hysteresis loop.

3. The shock absorber data processing method based on digital twin according to claim 1, characterized in that, The specific steps for obtaining the current hysteresis phase co-position residual sample include: Hysteresis cycle branching and energy-consuming phase modeling are performed on the hysteresis loop, and a unified hysteresis phase coordinate is formed based on the ratio of the cumulative damping work in the compressed phase branch and the restored phase branch to the total damping work of the corresponding phase branch. The twin residuals between the measured damping force value in the shock absorber operating data and the corresponding damping force value in the damping force data are calculated. The damping degradation direction is made consistent for the twin residuals, and the processed twin residuals are mapped to a unified hysteresis phase coordinate to form a hysteresis phase oriented residual sequence. Based on the unified hysteresis phase coordinates, the intra-branch co-position residual construction and periodic equalization processing are performed on the hysteresis phase orientation residual sequence. Co-position twin residuals are formed in the compressed phase branch and the restored phase branch respectively. The co-position twin residuals are arranged according to the complete motion period and co-position phase position to form the current hysteresis phase co-position residual sample.

4. The shock absorber data processing method based on digital twin according to claim 1, characterized in that, The acquisition of the working condition health reference sample specifically includes: Perform health hysteresis phase trajectory construction and health residual distribution aggregation processing on the historical health operation data to form a set of health hysteresis phase residual trajectories and health residual distribution; Based on the working condition adaptation relationship between the current working condition and the healthy working condition, as well as the phase integrity of the healthy hysteresis phase residual trajectory, the working condition adaptation weight is performed to form the healthy working condition weight, and the same set of healthy working condition weights are shared to all co-phase points in the unified hysteresis phase coordinate. Based on the health condition weights, Wasserstein centroid processing under the same quantile index constraint is performed on the health residual distributions corresponding to all co-phase points to form a co-ordered quantile centroid residual sequence. Virtual health residual trajectories are reconstructed along a unified hysteresis phase coordinate system for centroid residual values ​​with the same quantile number among all co-phase points to form virtual health residual trajectories. The virtual health residual trajectories corresponding to different quantile numbers are then arranged to form a working condition health reference sample.

5. The shock absorber data processing method based on digital twin according to claim 1, characterized in that, The rank-transfer phase distribution is obtained specifically by: The current hysteresis phase isotope residual sample and the operating condition health reference sample are subjected to source-preserving joint rank pooling to form Cucconi joint rank pool; According to the co-phase position, the natural rank and reverse rank of the current hysteresis phase co-phase residual sample in the Cucconi joint rank pool are collected. For each co-phase position, the two-wing square rank response is calculated to form the Cucconi two-wing rank response. The Cucconi two-wing rank responses corresponding to all co-phase positions are accumulated to form the full-phase natural square rank response and the full-phase reverse square rank response. The rank correlation correction coefficient is determined based on the number of residual values ​​in the Cucconi joint rank pool. The correlation correction joint calculation is performed on the full-phase natural square rank response and the full-phase inverse square rank response to form the Cucconi statistic. The Cucconi double-wing rank response corresponding to each isophase position is decomposed with the full-phase natural square rank response, the full-phase reverse square rank response, and the rank correlation correction coefficient to form the rank migration phase contribution value. The rank migration phase contribution value is then mapped to the unified hysteresis phase coordinate to form the rank migration phase distribution.

6. The shock absorber data processing method based on digital twin according to claim 1, characterized in that, The specific steps for obtaining the candidate correction amount of the candidate correction object include: Perform a persistence transition determination on the Cucconi statistic corresponding to the continuous processing window to form the persistence state of the Cucconi statistic. When the Cucconi statistic is in a continuous migration state, the Cucconi two-wing rank response corresponding to each isophase position is extracted according to the corresponding rank migration phase ratio to form the current rank migration response feature. The current rank migration response characteristics are matched with the corresponding positive unit phase rank response template and negative unit phase rank response template in the digital twin model of the shock absorber. Based on the template matching results, candidate correction objects and correction directions are determined. The candidate correction amount is determined by combining the response scaling degree, template matching confidence and Cucconi statistic persistence state.

7. The shock absorber data processing method based on digital twin according to claim 1, characterized in that, The specific steps involved in obtaining the corrected digital twin model are as follows: Read the candidate correction object, candidate correction amount, and current model parameters of the shock absorber digital twin model. Add the candidate correction amount to the current model parameters corresponding to the candidate correction object, keep the model parameters other than the candidate correction object unchanged, form the trial update model parameters, and write the trial update model parameters into a model copy with the same model structure as the shock absorber digital twin model to construct the trial update shock absorber digital twin model. Based on the digital twin model of the test-updated shock absorber, perform the same source residual reconstruction and Cucconi statistical recalculation to form the test-updated Cucconi statistic and the test-updated rank migration phase distribution; The Cucconi statistics before and after applying the candidate correction amount are subjected to rank migration shrinkage and newborn exclusion judgment to form the candidate correction amount verification results; Based on the verification results of the candidate corrections, a trial update retention process is performed. When the candidate corrections meet the trial update rank migration closure condition, the candidate corrections are retained, and the trial update model parameters corresponding to the candidate correction object are written into the shock absorber digital twin model. The model parameters other than the candidate correction object remain unchanged to form a calibration digital twin model. When the candidate corrections do not meet the trial update rank migration closure condition, the candidate corrections are revoked, the trial update model parameters are deleted, the current model parameters corresponding to the candidate correction object are restored, and a calibration digital twin model is formed based on the restored model parameters.

8. The shock absorber data processing method based on digital twin according to claim 1, characterized in that, The specific steps to obtain the shock absorber correction results include: Based on the corrected digital twin model, the corrected response reconstruction process is performed to generate corrected damping force data and corrected hysteresis loop; The corrected damping force data are subjected to the same source residual verification and Cucconi correction effect verification to form the corrected Cucconi statistic, the corrected rank migration phase distribution and the correspondence before and after correction. The model state, correction response reconstruction results, and Cucconi correction effect verification results of the correction digital twin model are encapsulated to form the shock absorber correction results.

Images