New energy vehicle bearing health state online evaluation system based on digital twinning

Abstract

The application discloses a new energy automobile bearing health state online evaluation system based on digital twinning, relates to the automobile fault prediction technical field, and can solve the problem of how to accurately distinguish bearing self-failure and external road surface interference under complex driving road conditions and realize accurate evaluation of bearing health state, and comprises a feature extraction module, a benchmark determination module, a correlation analysis module and a correction fusion module.The feature extraction module determines running fluctuation characteristic parameters and anti-interference fusion characteristic parameters according to collected bearing vibration data.The benchmark determination module determines a road condition self-adaptive benchmark vector according to current vehicle driving road conditions.The correlation analysis module obtains a road surface interference correlation coefficient between bearing vibration data and a pre-stored road surface interference vibration signal.The correction fusion module determines a bearing health index according to the running fluctuation characteristic parameters, the anti-interference fusion characteristic parameters, the road condition self-adaptive benchmark vector and the road surface interference correlation coefficient.The evaluation feedback module compares the bearing health index with a theoretical health index to judge whether the bearing is abnormal.

Description

Technical Field

[0001] This invention relates to the field of automotive fault prediction technology, specifically to an online assessment system for the health status of bearings in new energy vehicles based on digital twins. Background Technology

[0002] The reliability of core components of the drive system in new energy vehicles, such as bearings, directly affects the safety and lifespan of the entire vehicle. Therefore, there is an urgent need for real-time and accurate assessment of bearing health status. Digital twin technology, by constructing high-fidelity virtual models of physical entities, provides a new technological paradigm for predictive health management of equipment and has been introduced into the field of condition monitoring of industrial equipment.

[0003] However, existing bearing health assessment methods based on digital twins or data-driven approaches are easily affected by complex and variable driving environments, especially road bumps, when applied to new energy vehicles in actual operation. This results in serious challenges to the accuracy, reliability, and efficiency of condition assessment under real and complex working conditions. Summary of the Invention

[0004] To address the current technical challenge of accurately distinguishing between bearing failures and external road disturbances under complex driving conditions, and thus achieving precise assessment of bearing health status, this invention aims to provide an online bearing health status assessment system for new energy vehicles based on digital twins. The specific technical solution adopted is as follows:

[0005] In a first aspect, the present invention provides an online health status assessment system for new energy vehicle bearings based on digital twins, comprising: a feature extraction module, used to determine operational fluctuation feature parameters and anti-interference fusion feature parameters based on collected bearing vibration data; wherein, the operational fluctuation feature parameters are used to characterize the operational amplitude fluctuation state of the bearing, and the anti-interference fusion feature parameters are used to characterize fault features in the bearing vibration data that are distinct from road surface interference; a benchmark determination module, used to determine a road condition adaptive benchmark vector corresponding to the current vehicle driving road conditions; a correlation analysis module, used to obtain the road surface interference correlation coefficient between the bearing vibration data and pre-stored road surface interference vibration signals; a correction fusion module, used to determine a bearing health index based on the operational fluctuation feature parameters, the anti-interference fusion feature parameters, the road condition adaptive benchmark vector, and the road surface interference correlation coefficient; and an assessment feedback module, used to compare the bearing health index with the theoretical health index and determine whether the bearing health status is abnormal based on the comparison results; wherein, the theoretical health index is output based on a digital twin model.

[0006] Secondly, this invention provides an online assessment method for the health status of bearings in new energy vehicles based on digital twins, comprising: determining operational fluctuation characteristic parameters and anti-interference fusion characteristic parameters based on collected bearing vibration data; wherein, the operational fluctuation characteristic parameters are used to characterize the operational amplitude fluctuation state of the bearing, and the anti-interference fusion characteristic parameters are used to characterize the fault characteristics in the bearing vibration data that are different from road surface interference; determining a road condition adaptive reference vector corresponding to the current vehicle driving road conditions based on the current vehicle driving road conditions; obtaining the road surface interference correlation coefficient between the bearing vibration data and pre-stored road surface interference vibration signals; determining a bearing health index based on the operational fluctuation characteristic parameters, anti-interference fusion characteristic parameters, road condition adaptive reference vector, and road surface interference correlation coefficient; comparing the bearing health index with the theoretical health index, and judging whether the bearing health status is abnormal based on the comparison results; wherein, the theoretical health index is output based on the digital twin model.

[0007] Thirdly, the present invention provides an electronic device, comprising: a processor and a memory; wherein the memory is used to store one or more programs, the one or more programs including computer-executable instructions, and when the electronic device is running, the processor executes the computer-executable instructions stored in the memory to cause the electronic device to perform the online assessment method for the health status of new energy vehicle bearings based on digital twins as described in the first aspect and any possible implementation thereof.

[0008] This invention offers the following advantages: By combining multi-domain feature fusion extraction with an adaptive dynamic benchmark for road conditions, and innovatively introducing a road disturbance quantification identification and correction mechanism, the system effectively suppresses broadband random vibration interference caused by road bumps, significantly improving the signal-to-noise ratio. This allows early, subtle bearing fault characteristics to be highlighted and accurately captured, greatly improving the accuracy and timeliness of fault diagnosis. Simultaneously, the assessment process, through closed-loop verification and predictive extrapolation with a digital twin model, not only enhances the credibility of the condition assessment but also enables predictive judgment of fault type, severity, and evolution trend, thereby supporting predictive maintenance decisions. Ultimately, while ensuring high accuracy and reliability in the assessment, the system optimizes the utilization efficiency of storage, transmission, and computing resources in the vehicle system by focusing on high-value data. Attached Figure Description

[0009] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1This is a schematic diagram of the architecture of an online health status assessment system for new energy vehicle bearings based on digital twins, provided in one embodiment of the present invention.

[0011] Figure 2 This is a schematic diagram of the architecture of a feature extraction module provided in one embodiment of the present invention;

[0012] Figure 3 This is a flowchart illustrating an online health status assessment method for bearings in new energy vehicles based on digital twins, as provided in one embodiment of the present invention. Detailed Implementation

[0013] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the specific implementation methods, structures, features, and effects of the present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0014] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0015] The following description, in conjunction with the accompanying drawings, details the specific scheme of an online health status assessment system for new energy vehicle bearings based on digital twins provided by this invention.

[0016] For example, such as Figure 1 The diagram shown is an architectural schematic of an online health status assessment system for new energy vehicle bearings based on digital twins (hereinafter referred to as online assessment system 10) according to an embodiment of the present invention. The online assessment system 10 includes: a feature extraction module 11, a benchmark determination module 12, a correlation analysis module 13, a correction and fusion module 14, and an assessment feedback module 15. The modules are described below in sequence:

[0017] (1) Feature extraction module 11.

[0018] The feature extraction module 11 is responsible for receiving the raw bearing vibration data from the sensor and extracting two types of key feature parameters that can characterize the bearing's state in different dimensions, laying the foundation for subsequent fusion analysis and health assessment.

[0019] Optionally, the feature extraction module 11 is used to determine operational fluctuation characteristic parameters and anti-interference fusion characteristic parameters based on the collected bearing vibration data. The operational fluctuation characteristic parameters characterize the amplitude fluctuation state of the bearing's operation, while the anti-interference fusion characteristic parameters characterize fault features in the bearing vibration data that differ from road surface interference.

[0020] Specifically, the operating fluctuation characteristic parameters are a set of indicators extracted from the time-domain amplitude changes of vibration signals. They are mainly used to quantify the stability and turbulence of bearing operation. The core parameters include the fluctuation index, which reflects the intensity of local amplitude fluctuations, and the motion degree parameter, which reflects the sudden changes in operating state over a longer period.

[0021] Anti-interference fusion characteristic parameters are a set of comprehensive indicators obtained by fusing multi-dimensional information such as frequency domain and time domain statistics of vibration signals. They aim to suppress broadband random vibration interference caused by road bumps and highlight the fault characteristics of the bearing itself. The core of these parameters usually includes the frequency characteristic index, which reflects the energy concentration of the fault frequency, the crest factor, which reflects the transient impact intensity, and the comprehensive fluctuation index, which integrates the aforementioned indicators.

[0022] For example, such as Figure 2 As shown, the feature extraction module 11 can be divided into a fluctuation feature extraction submodule 111 and an anti-interference fusion feature extraction submodule 112 to handle feature extraction tasks with two different technical objectives, which will be described below:

[0023] Optionally, when determining the running fluctuation feature extraction submodule 111, the following steps are specifically performed:

[0024] S101. Calculate the fluctuation index based on the adjacent changes in bearing vibration amplitude within a preset time window. The fluctuation index is used to characterize the degree of local amplitude fluctuation.

[0025] In this step, the bearing vibration amplitude data used for calculation comes from the bearing vibration signal acquired in real time by the data acquisition module 16. After digital processing, this bearing vibration signal forms a bearing vibration data sequence arranged in chronological order. In this sequence, the value of each sampling point is the bearing vibration amplitude.

[0026] Specifically, the volatility index is a time-domain indicator used to quantify the degree of amplitude change of vibration signals in a short period of time. The larger the value, the more unstable the bearing operation is at the microscopic level.

[0027] In this step, the preset time window refers to a fixed-length time segment used when segmenting continuously acquired bearing vibration data. It is the basic analytical unit for extracting time-domain features (such as calculating fluctuation indices and RMS values). The window length needs to be balanced between temporal resolution and statistical stability: if the window is too short, the extracted features will be greatly affected by random noise, resulting in drastic fluctuations; if the window is too long, it may smooth out early transient fault features, reducing monitoring sensitivity. For vibration analysis of bearings in new energy vehicles, considering the medium-to-high speeds of the motor (e.g., several thousand to tens of thousands of revolutions per minute) and the corresponding fault characteristic frequencies, the length of this preset time window can typically be set in the range of 10 milliseconds to 1000 milliseconds (i.e., 0.01 seconds to 1 second). For example, a 100-millisecond window can cover multiple bearing rotation cycles, thus enabling stable calculation of its vibration statistical characteristics. The specific window length can be determined through prior testing or simulation based on the rated speed of the target bearing, the sampling frequency, and the fault characteristic cycles of interest.

[0028] For example, the volatility feature extraction submodule 111 calculates the volatility index using the following formula:

[0029]

[0030] Where P represents the volatility index; This represents the bearing vibration amplitude value at the i-th sampling point within the preset time window; This represents the bearing vibration amplitude value at the (i+1)th sampling point within the preset time window; This represents the normalization function, such as a maximum-minimum normalization function. The maximum and minimum values ​​are preset empirical extreme values ​​derived from a large amount of historical experimental data. If the calculated result exceeds the interval [0, 1], it is restricted to the range [0, 1] by a truncation function (i.e., if the result is less than 0, it is taken as 0; if it is greater than 1, it is taken as 1) to eliminate the influence of outliers on the evaluation index. It should be noted that if vibration signals appear in the formula... and If both are zero, the ratio term can be set to 0 and then included in the calculation.

[0031] The above formula first calculates the ratio of the difference in amplitude of each pair of adjacent sampling points within the window to their sum. This ratio eliminates the influence of absolute dimensions and reflects the local relative change in vibration. Then, this ratio is averaged over all adjacent points to obtain a raw index reflecting the average fluctuation level within the entire window. Finally, this raw index is mapped to the [0,1] interval using a normalization function to obtain the final fluctuation index P. The larger the value of P, the more unstable the bearing operation is in the current stage (i.e., within the preset time window).

[0032] S102. Calculate the motion degree parameter based on the change in the fluctuation index between adjacent preset time windows. The motion degree parameter is used to characterize the running vibration state of the bearing.

[0033] Specifically, the motion degree parameter is an index used to characterize the stability of bearing operation over a longer time scale. It reflects abrupt changes or continuous turbulence trends in bearing operation by analyzing the changes in the fluctuation index between continuous time windows. Optionally, the operation fluctuation feature extraction submodule 111 calculates the motion degree parameter using the following formula. :

[0034]

[0035] in, Indicates the degree of motion; This represents the volatility index corresponding to the j-th preset time window; This represents the volatility index corresponding to the (j+1)th (i.e., the next adjacent) preset time window; This indicates the minimum value operation, which takes the smaller value of the two parameters inside the parentheses; This represents the softsign activation function, specifically expressed as follows: In this formula, its role is to map the value of the motion degree parameter to the interval (-1, 1). It should be noted that if the fluctuation index of adjacent time windows is zero, i.e. =0, then directly define the motion level parameter. =0.

[0036] In the above formula, the absolute value of the difference between the volatility indices of two adjacent time windows is first calculated. This is used to measure the magnitude of change in volatility. Then, this magnitude is divided by the smaller volatility index between the previous and next windows. This yields the relative rate of change of the volatility index, a ratio that more effectively measures the severity of state changes. Finally, the softsign function is used to process this ratio, smoothly mapping the result to the interval (-1, 1) to obtain the final motion parameter. .

[0037] Optionally, the anti-interference fusion feature extraction submodule 112 performs the following steps when selecting anti-interference fusion feature parameters:

[0038] S103. Calculate the frequency characteristic index based on the spectral characteristics of the bearing vibration data. The frequency characteristic index is used to characterize the energy proportion of the bearing's own fault characteristic frequencies.

[0039] First, the anti-interference fusion feature extraction submodule 112 performs a fast Fourier transform on the bearing vibration data within the current time window to obtain its spectrum.

[0040] In this step, the bearing vibration data is the original time-domain vibration signal sequence acquired by the data acquisition module 16 and transmitted to this module. First, a fast Fourier transform is performed on the vibration data to obtain its spectrum.

[0041] Specifically, the formula for calculating the frequency characteristic index F is as follows: .in, This indicates a specific fault characteristic frequency in the bearing. (The passing frequency of the inner ring, outer ring, or rolling elements is calculated from the bearing geometry and shaft speed) a preset fixed frequency band range [e.g., the passing frequency of the inner ring, outer ring, or rolling elements]. - , + The sum of the squares of the amplitudes of all frequency components within the range. The preset half-width of the frequency band, for example, can be set to 2% or 5%, this range is used to focus the characteristic frequency energy associated with bearing failure; E represents the sum of the squares of the amplitudes of all frequency components in the vibration signal spectrum within the current time window.

[0042] It is understandable that the frequency characteristic index F is calculated by measuring the vibration signal energy in the neighborhood of the fault characteristic frequency. The proportion of the total signal energy E is used to determine the overall vibration. This proportion directly reflects the weight of vibrations with specific frequency components caused by structural damage to the bearing itself (such as spalling or cracks) in the overall vibration.

[0043] S104. Calculate the crest factor based on the peak and effective values ​​of the bearing vibration data. The crest factor is used to characterize the impact state of the bearing.

[0044] Specifically, the crest factor (CF) is a time-domain statistical indicator used to characterize the presence of sharp peaks or transient impact components in the waveform of a vibration signal. This is very effective in identifying whether a bearing has been subjected to a transient load impact (such as passing through pits). A higher CF value generally indicates a more significant impact.

[0045] For example, the anti-interference fusion feature extraction submodule 112 calculates the peak factor. The formula is: .in, A represents the peak value (maximum absolute value) of the bearing vibration data within the current time window; A represents the effective value (root mean square value) of the bearing vibration data within the current time window. This is a very small positive number, such as 10 to the power of negative 5, to avoid the denominator being zero. In this formula, the peak value of the signal is calculated. The ratio of the peak value to the effective value A, which characterizes the average energy level of the signal. When a transient impulse occurs in the signal, the peak value increases significantly, while the effective value changes relatively little, thus leading to... The value suddenly increased.

[0046] S105. Calculate the anti-interference fusion characteristic parameters based on the motion degree parameters, peak factor, and frequency characteristic index.

[0047] It should be noted that the anti-interference fusion characteristic parameter is a multi-domain fusion index that integrates the characteristics of operating bumps, transient impact intensity, and fault frequency. It aims to enhance the fault characteristics of the bearing itself while suppressing the interference of broadband random vibrations of the road surface.

[0048] For example, the anti-interference fusion feature extraction submodule 112 calculates the anti-interference fusion feature parameters using the following formula. :

[0049]

[0050] in, Indicates the anti-interference fusion characteristic parameters; This represents the sigmoid activation function, specifically expressed as follows: Its role in this formula is to map the value of the motion degree parameter to the interval (0, 1); The normalized reference value for the crest factor is usually the average value of the CF obtained from long-term statistical analysis of the bearing under stable and healthy operating conditions.

[0051] In the above formula, firstly, the crest factor CF is normalized to eliminate sensor baseline differences, thus obtaining the relative impact intensity. Then, the relative impact intensity, representing the turbulent operating condition, is... Adding this to the normalized impact intensity yields a comprehensive "state-impact" feature. Next, this comprehensive feature is multiplied by the frequency characteristic exponent F, thus weighting the time-domain feature using the concentration of fault frequencies: if the vibration energy is significant at the fault frequency (high F value), the feature value is amplified; if it is not caused by a fault frequency, the feature value is suppressed. Finally, the result is mapped to the (0,1) interval using the sigmoid function to obtain numerically stable, interference-resistant fusion feature parameters. .

[0052] Based on the above S101-S105, the feature extraction module 11 achieves the purification of high-value information from the bearing vibration signal through a hierarchical, multi-domain fusion feature extraction process. Specifically, firstly, it extracts the fluctuation index, which quantifies the degree of local fluctuation, and the motion degree parameter, which characterizes the macroscopic operational turbulence state, from the time-domain amplitude sequence, constructing basic time-domain features reflecting the bearing's operational stability. Then, it further extracts the frequency feature index, which reflects the concentration of fault characteristic frequency energy, from the frequency domain, and extracts the crest factor, which reflects the transient impact intensity, from time-domain statistics. Finally, by multiplying and coupling the motion degree parameter reflecting the operational turbulence state, the normalized impact intensity, and the fault frequency concentration, and normalizing them using the sigmoid function, it generates anti-interference fusion feature parameters.

[0053] Therefore, the feature extraction module 11 deeply integrates the vibration signal in the time domain, frequency domain, and time domain statistical features, so that the final feature parameters can effectively suppress the broadband random vibration energy caused by road bumps that is unrelated to bearing fault characteristics, while enhancing and highlighting the periodic impact and modulation features related to bearing damage itself, thus providing high signal-to-noise ratio and high discrimination feature input for subsequent accurate and reliable health status assessment.

[0054] (2) Benchmark determination module 12.

[0055] The benchmark determination module 12 is responsible for establishing a road condition adaptive evaluation benchmark for the system. Specifically, based on the current driving environment of the vehicle, it dynamically selects or calculates a set of normal characteristic value ranges that match the bearing health status, so that the health evaluation standard can be adaptively adjusted with changes in road conditions. This is the key to improving the evaluation accuracy under complex working conditions.

[0056] Optionally, the reference determination module 12 is used to determine the road condition adaptive reference vector corresponding to the current road conditions based on the current road conditions of the vehicle.

[0057] For example, when determining the road condition adaptive reference vector, the reference determination module 12 specifically performs the following steps:

[0058] S201. Obtain a preset baseline mapping relationship of characteristic parameters of bearings under healthy conditions, covering multiple typical road conditions.

[0059] The characteristic parameter benchmark mapping relationship is specifically a data structure or database table stored in the system. It establishes a correspondence between "road condition type" and "normal value range of characteristic parameters corresponding to the bearing's health status under that road condition." Its function is to serve as a dynamic and adaptive reference benchmark for assessing whether the current bearing condition is abnormal. The specific acquisition process may include the following two sub-steps:

[0060] (a) Acquire historical sensor data of the vehicle bearings under healthy conditions and covering various typical road conditions with different combinations of vehicle speed and torque, and statistically analyze the numerical distribution range of the historical sensor data for each road condition type.

[0061] Specifically, the baseline determination module 12 retrieves historical datasets accumulated from the database for the vehicle or similar vehicles under known health conditions (such as during factory testing or long-term fault-free monitoring periods). This dataset must cover various preset typical road conditions, such as flat roads, continuous speed bumps, and gravel roads, and include different combinations of vehicle speed and drive torque for each road condition.

[0062] Subsequently, the baseline determination module 12 analyzes this historical dataset and extracts the time series of the running fluctuation characteristic parameters and anti-interference fusion characteristic parameters calculated by the feature extraction module from all historical data segments of each typical road condition that has been marked or identified.

[0063] Finally, the baseline determination module 12 performs statistical analysis on these characteristic parameter sequences, calculating their numerical distribution range. For example, it can calculate their mean (μ) and standard deviation (σ), and use [μ-2σ, μ+2σ] as the healthy baseline range for this characteristic parameter under the road condition. This range can theoretically cover approximately 95% of healthy data. If the characteristic parameter value falls within this range, its fluctuation is considered to be within the normal range for the road condition.

[0064] (b) Determine the baseline mapping relationship of characteristic parameters based on the numerical distribution range of historical sensor data under each road condition type.

[0065] Furthermore, the baseline determination module 12 stores the results obtained in step (a) in a structured manner. For the kth road condition type... and its corresponding m-th feature parameter (For example To run the fluctuation characteristic parameters, (For anti-interference fusion characteristic parameters), record the lower limit of its health benchmark range. and upper limit Ultimately, the correspondence between all road condition types and the health baseline ranges of all characteristic parameters constitutes the characteristic parameter baseline mapping relationship M, which can be formally expressed as the following formula:

[0066]

[0067] In the above formula, M is the feature parameter baseline mapping relationship, which is a set where each element corresponds to a road condition type. For each road condition, a subset is associated, which contains the various feature parameters. and their corresponding upper and lower limits of health value range [ ].

[0068] S202. Based on the vehicle's current driving status information, determine the current road condition type of the vehicle.

[0069] The purpose of this step is to identify the vehicle's driving environment in real time so that the corresponding health benchmark can be used for subsequent evaluation. The vehicle's current driving status information may include longitudinal acceleration, lateral acceleration, suspension travel, wheel speed signals, etc., from the vehicle's sensors. These signals together reflect the road surface unevenness and the vehicle's motion state.

[0070] Furthermore, the baseline determination module 12 receives relevant signals from the vehicle's CAN bus to construct a current-time state feature vector. This state feature vector is then input into a pre-defined classification model (e.g., a road condition classifier pre-trained based on Support Vector Machine (SVM) or K-means clustering). This classifier maps the current state to a specific class in a pre-defined set of typical road condition types (e.g., "flat road surface," "moderately bumpy road surface," "severely bumpy road surface"), thereby determining the road condition type.

[0071] S203. Determine the adaptive reference vector for road conditions based on the road condition type and the reference mapping relationship of characteristic parameters.

[0072] Specifically, the road condition adaptive baseline vector is a list containing all the baseline values ​​required for the current assessment. Its elements correspond one-to-one with the feature parameters that need to be compared (operational fluctuation feature parameters and anti-interference fusion feature parameters). Each element gives the normal range or typical value that the corresponding feature parameter should have when it is in a healthy state under the current road conditions.

[0073] For example, the benchmark determination module 12 performs the following steps:

[0074] First, obtain the currently determined road condition type from S202. Then, in the feature parameter benchmark mapping relationship M established in S201, find the corresponding... The corresponding record. Then, the various feature parameters are extracted from this record. Health benchmark range [ Finally, these extracted baseline ranges (e.g., in the form of range midpoints or range upper and lower bounds) are arranged in order of feature parameters to form the road condition adaptive baseline vector. :

[0075]

[0076] In another possible implementation, to simplify subsequent calculations, a range (i.e., a health baseline range) can also be used. The median constitutes the baseline vector.

[0077] Based on S201-S203 above, the benchmark determination module 12 significantly improves the accuracy and reliability of the bearing health status assessment system in complex driving environments by constructing and applying a road condition adaptive dynamic benchmark mechanism. Specifically, this module establishes a health benchmark mapping relationship between characteristic parameters and road condition types in advance through statistical analysis of a large amount of historical health data (covering various typical road conditions, vehicle speed, and torque combinations), ensuring that the benchmark referenced for the assessment originates from objective and comprehensive health status statistics, rather than fixed or empirical thresholds. Furthermore, it analyzes the vehicle's current driving status information (such as acceleration and suspension data) in real time and uses a classification model to determine the current road condition type, achieving online perception and identification of the vehicle's external driving environment. Finally, based on the real-time determined road condition type, it quickly retrieves and assembles the corresponding "road condition adaptive benchmark vector" from the preset benchmark mapping relationship, thereby providing a dynamic assessment scale that matches the current environment for subsequent health index calculations.

[0078] This allows the system's evaluation benchmark to be automatically and accurately adjusted according to changes in road conditions (such as flat or bumpy). This fundamentally avoids the problem of misjudging normal broadband vibration responses as bearing failures on bumpy roads or missing weak fault characteristics on smooth roads due to the use of a single fixed benchmark. It ensures the robustness of the evaluation results to road disturbances and the sensitivity to the actual bearing condition.

[0079] In practical applications, the benchmark determination module 12 can be integrated into the in-vehicle infotainment system (IVI) or the in-vehicle gateway, utilizing its data access and processing capabilities.

[0080] (3) Correlation analysis module 13.

[0081] The correlation analysis module 13 is responsible for quantifying the similarity between the currently collected bearing vibration data and interference signals caused purely by road surface bumps. Its output correlation coefficient is the core basis for distinguishing between "fault" and "interference" in subsequent correction stages, directly affecting the system's effectiveness in suppressing road surface interference.

[0082] Optionally, the correlation analysis module 13 is used to obtain the road interference correlation coefficient between the bearing vibration data and the pre-stored road interference vibration signal.

[0083] For example, when obtaining the correlation coefficient of road surface disturbance, the correlation analysis module 13 specifically performs the following steps:

[0084] S301. Obtain the spectral characteristics of the bearing vibration data within the current preset time window.

[0085] It should be noted that the current preset time window is the same data segment of the same length used by the feature extraction module 11 when calculating parameters such as the fluctuation index and frequency characteristic index. To ensure the temporal alignment and consistency of the analysis data across the entire system, this window length is consistent with the setting in the feature extraction module, for example, it can be set to 100 milliseconds, 500 milliseconds, or 1 second. This ensures that the spectrum targeted by the subsequent correlation analysis originates from the same set of vibration data as the various characteristic parameters calculated within the same time period, thereby ensuring the temporal synchronization of the evaluation logic.

[0086] In this step, the correlation analysis module 13 performs a frequency domain transformation on the original bearing vibration data sequence within the current preset time window. A standard and efficient implementation method is to use the Fast Fourier Transform (FFT) algorithm. The module first preprocesses the time-domain vibration signal, such as removing the DC component and windowing (e.g., Hanning window) to reduce spectral leakage, and then performs FFT calculations to obtain the complex spectrum of the vibration signal within the time window. Finally, the amplitude spectrum or power spectrum is taken as the "spectral feature" required for subsequent analysis. This spectral feature reflects the distribution of vibration energy at various frequency components within the current time period.

[0087] S302. Perform correlation analysis between the spectral characteristics of the bearing vibration data within the current preset time window and the pre-stored reference spectrum template, and calculate the road surface interference correlation coefficient based on the results of the correlation analysis.

[0088] The purpose of this step is to quantify the similarity between the current vibration signal and a typical vibration mode caused purely by road surface disturbances. By calculating the correlation coefficient of their spectral characteristics, a metric between 0 and 1 can be obtained. The closer the value is to 1, the lower the probability that the current vibration is mainly dominated by road surface disturbances and the abnormal characteristics of the bearing itself; conversely, the lower the value is, the lower the probability that the vibration mode deviates from typical road surface disturbances and may contain bearing fault information.

[0089] Furthermore, the correlation analysis module 13 first reads the pre-stored reference spectrum template from the system memory. This template, obtained through experimental data collection or big data analysis, characterizes the low-frequency stimulated resonance spectrum of a vehicle suspension system under typical bumpy road surface excitation, and it embodies the frequency domain characteristic patterns of road surface interference.

[0090] Next, the module will use the spectral characteristics of the current time window obtained in S301. With reference spectrum template A comparison is performed. A widely used algorithm for measuring the linear correlation between two signals is the Pearson correlation coefficient algorithm. The correlation analysis module 13 calculates using this algorithm. and The Pearson correlation coefficient is used as the correlation coefficient of road surface disturbance. The specific calculation process of the Pearson correlation coefficient algorithm is existing technology and will not be elaborated here.

[0091] Based on S301-S302 above, the correlation analysis module 13 introduces and calculates a quantified road disturbance correlation coefficient, providing the system with crucial road disturbance identification and quantification capabilities. This transforms the complex pattern recognition problem of "whether the current vibration is mainly caused by road bumps" into a precise and calculable scalar indicator. This coefficient directly and quantitatively reflects the similarity between the current vibration signal and the pure road disturbance pattern, thus providing a vital decision-making basis for the subsequent correction and fusion module: when... A high value indicates that the current data is dominated by road surface disturbances. The system can automatically reduce the weight of characteristic deviations caused by broadband vibrations, effectively avoiding misjudging normal road impact responses as bearing failures; conversely, when... When the value is low, it indicates that the vibration may be caused by an abnormality in the bearing itself, and the relevant characteristic deviation will be retained and used for health assessment.

[0092] In practical applications, the correlation analysis module 13 can be implemented by a microprocessor unit (MCU) that supports floating-point operations, and its spectrum calculation and correlation analysis can be completed by software algorithms.

[0093] (4) Modify the fusion module 14.

[0094] The correction fusion module 14 is responsible for receiving the current value of the feature parameters from the feature extraction module 11, the road condition adaptive reference vector from the reference determination module 12, and the road surface interference correlation coefficient from the correlation analysis module 13, performing a series of calculations, and finally outputting a quantitative health index that can accurately reflect the health status of the bearing itself.

[0095] Optionally, the correction fusion module 14 is used to determine the bearing health index based on the operating fluctuation characteristic parameters, anti-interference fusion characteristic parameters, road condition adaptive reference vector, and road surface interference correlation coefficient.

[0096] For example, when the correction fusion module 14 determines the bearing health index, it specifically performs the following steps:

[0097] S401. Determine the current values ​​of the operation fluctuation characteristic parameters and the anti-interference fusion characteristic parameters, respectively, and the characteristic deviation between them and the corresponding reference values ​​in the road condition adaptive reference vector.

[0098] The purpose of this step is to quantify the deviation of the current bearing condition from the health baseline under the corresponding road conditions. The larger the characteristic deviation, the further the current characteristic parameter value deviates from the normal range, and the higher the probability of an anomaly.

[0099] Specifically, the correction fusion module 14 receives the current values ​​(denoted as ) of each feature parameter from the feature extraction module 11. Where m represents the feature type, such as To run the fluctuation characteristic parameters, These are the anti-interference fusion feature parameters, and the road condition adaptive reference vector from the reference determination module 12. This reference vector contains the health reference values ​​(denoted as ) corresponding to each feature parameter under the current road condition. ).

[0100] Next, the correction fusion module 14 calculates the feature deviation of each feature parameter, and the calculation formula is as follows: , This represents the feature deviation of the m-th feature parameter (e.g., operational fluctuation feature parameter or anti-interference fusion feature parameter). It is a very small positive number, such as 10 to the power of negative 5, to avoid the denominator being zero.

[0101] In this formula, the absolute difference between the current value of each feature parameter and its health baseline value is first calculated. | This value reflects the absolute magnitude of the deviation. Then, this absolute difference is divided by its corresponding health baseline value. The absolute value of the deviation is used to obtain the relative deviation. The reason for using relative deviation is that it eliminates the influence of differences in the dimensions and orders of magnitude of different characteristic parameters, allowing the degree of deviation of different characteristics (such as volatility index and composite volatility index) to be compared and integrated on a unified, unitless scale. This characteristic deviation quantity... The larger the value, the further the current characteristic deviates from the health benchmark.

[0102] S402. For each feature deviation, the feature deviation weight is corrected according to the road surface interference correlation coefficient.

[0103] This step is crucial for suppressing road surface interference and improving assessment accuracy. Its purpose is to utilize the road surface interference correlation coefficient calculated in S302. The characteristic deviations obtained in S401 are dynamically weighted. When the bearing vibration spectrum characteristics used to calculate the correlation coefficient in the current time window are highly similar to typical road surface disturbances (…),… When the value is high, the weight of the characteristic deviation should be reduced accordingly to avoid misjudging normal fluctuations caused by road impacts as faults; otherwise, the value should be retained or given a higher weight.

[0104] Furthermore, the correction fusion module 14 performs a correction operation on each feature deviation. Let the feature deviation of the m-th feature parameter be denoted as... The correlation coefficient of road surface interference is The corrected eigenbia weights Calculated using the following formula:

[0105]

[0106] in, This represents the feature deviation weight after correction of the m-th feature parameter; This represents the feature deviation calculated in S401 for the m-th feature parameter; This represents the pavement disturbance correlation coefficient calculated in S302. The formula includes the characteristic deviation of each feature. Multiply by a correction factor ( ).because The range of is [0,1], therefore the range of the correction coefficient is also [0,1]. When When the value is close to 1, it indicates that the current vibration is mainly caused by road surface disturbance. At this time, the correction coefficient is close to 0, making the corrected weight... It is significantly suppressed (approaching 0). When When the value is close to 0, it indicates a large difference between the current vibration and road surface disturbance modes, and the correction coefficient is close to 1. At this point, the characteristic deviation is... Almost entirely preserved as weights This achieves the effect of adaptively attenuating the original feature deviation based on the possibility of road surface disturbance.

[0107] S403. Determine the bearing health index based on all corrected characteristic deviation weights.

[0108] Finally, the correction and fusion module 14 merges all the corrected feature deviation weights obtained in S402 to generate a single quantitative index that comprehensively represents the overall health status of the bearing, namely the bearing health index H.

[0109] Alternatively, a direct and effective fusion method is to perform a weighted summation of the modified weights of each feature. The bearing health index H is calculated using the following formula:

[0110]

[0111] in, This represents the final determined bearing health index; n represents the total number of characteristic parameters involved in the fusion (for example, if the operating fluctuation characteristic parameter and the anti-interference fusion characteristic parameter are combined, then n=2). This represents the feature deviation weight after the i-th feature parameter is corrected (calculated by S402); This represents the weight coefficient of the i-th feature parameter.

[0112] It should be noted that the above formula applies a corrected bias weight to each feature parameter. Multiply by its corresponding weighting coefficient Then, the weighted results of all features are summed to obtain the total, which is the bearing health index H. Weighting coefficients. This reflects the relative importance of different characteristic parameters in the comprehensive evaluation. Their values ​​can be set through historical data or historical cases, or automatically learned through machine learning methods (such as logistic regression, support vector machine, and other models trained based on historical failure data). The H value is a non-negative scalar; the larger the value, the worse the health condition of the bearing is after comprehensive evaluation of various characteristics, and the higher the risk of failure.

[0113] Furthermore, by setting a health index threshold (for example, a threshold value of 0.7), automatic early warning of bearing anomalies can be achieved. The rule for determining this health index threshold is: it is determined by analyzing a large amount of historical data on bearings in a healthy state. Typically, a high percentile (such as the 95th percentile) of the historical health index distribution is selected as the threshold. The core principle is to achieve an engineering balance between controlling the false alarm rate (avoiding normal fluctuations being judged as faults) and reducing the false alarm rate (avoiding missing real faults).

[0114] Based on the above S401-S402, the correction and fusion module 14 intelligently corrects and weights the feature deviation by introducing the road surface interference correlation coefficient, and finally outputs a comprehensive health index that can both keenly capture the intrinsic characteristics of early bearing failure and strongly suppress the external interference of broadband random vibration of the road surface, thereby ensuring the accuracy and reliability of the evaluation results in complex real driving environments.

[0115] (5) Evaluation and feedback module 15.

[0116] The evaluation feedback module 15 is responsible for the final interpretation and utilization of the bearing health index output by the correction fusion module 14, completing the transformation from "data" to "decision" and "knowledge", and realizing interaction with the digital twin model to form an evaluation closed loop.

[0117] Optionally, the evaluation feedback module 15 is used to compare the bearing health index with the theoretical health index and determine whether the bearing's health status is abnormal based on the comparison results. The theoretical health index is output based on a digital twin model.

[0118] Specifically, the evaluation feedback module 15 first obtains from the cloud or local digital twin model service the theoretical health index, which is simulated and calculated by the digital twin model under the same vehicle driving conditions (such as the same vehicle speed, torque, and road condition type), representing the fault-free state of the bearing.

[0119] Optionally, the evaluation feedback module 15 then calculates the absolute value of the difference between the actual bearing health index and the theoretical health index, and performs processing such as negative exponential mapping on this absolute value to obtain a state similarity index. Finally, the evaluation feedback module 15 determines whether the bearing's health state is abnormal by comparing this state similarity index with a preset similarity threshold. If it is below the threshold, an abnormality alarm is triggered.

[0120] For example, the aforementioned similarity threshold needs to strike a balance between false alarm rate and false negative rate, and is usually determined based on a large amount of historical health data. For instance, the 95th percentile of the distribution of the state similarity index calculated for a bearing in a healthy state can be selected as the threshold. This means that the probability of a healthy bearing having a similarity index below this threshold is extremely small (e.g., 5%). An exemplary value is 0.85; when the state similarity index is below 0.85, an anomaly alarm is triggered. This value can be learned from historical data of specific vehicle models and bearing types, and is adjustable in application.

[0121] Optionally, the evaluation feedback module 15 is also used to drive the digital twin model to simulate the evolution of one or more preset fault modes (such as inner ring spalling, outer ring cracking, etc.). Then, the evaluation feedback module 15 matches the actual bearing health index change trajectory over a period of time with the various fault evolution trajectories simulated by the digital twin model, thereby further predicting the possible fault type and severity based on the anomaly determination. The evaluation feedback module 15 is typically deployed on a vehicle-to-everything (V2X) cloud platform or edge computing node with stronger computing power to invoke complex digital twin models and perform large-scale trajectory matching calculations.

[0122] (6) Data acquisition module 16.

[0123] Optionally, the online evaluation system 10 also includes a data acquisition module 16.

[0124] The data acquisition module 16 is responsible for collecting the raw vibration data of the bearings of new energy vehicles in real time and accurately during operation, providing a reliable data source for all upstream analysis modules.

[0125] Specifically, the data acquisition module 16 is used to acquire bearing vibration data. Its physical components mainly include a vibration acceleration sensor non-invasively mounted on the bearing housing or a nearby critical structure, as well as a signal conditioning circuit and analog-to-digital converter (ADC) responsible for anti-aliasing filtering, amplification, and conversion of the sensor's analog signals into digital signals. This hardware transmits the synchronously acquired digital vibration data stream to the feature extraction module 11 in real time via an in-vehicle network (such as CAN FD or Ethernet).

[0126] The above describes the online health status assessment system 10 for new energy vehicle bearings based on digital twins and its included modules. Based on the aforementioned technical solution, this system combines multi-domain feature fusion extraction with road condition adaptive dynamic benchmarks, and innovatively introduces a road disturbance quantification identification and correction mechanism. The system effectively suppresses broadband random vibration interference caused by road bumps, significantly improving the signal-to-noise ratio. This allows early, subtle bearing fault characteristics to be highlighted and accurately captured, greatly improving the accuracy and timeliness of fault diagnosis. Simultaneously, the assessment process undergoes closed-loop verification and predictive simulation with the digital twin model, enhancing the credibility of the status assessment and enabling predictive judgment of fault type, severity, and evolution trend, thereby supporting predictive maintenance decisions. Ultimately, while ensuring high accuracy and reliability in the assessment, the system optimizes the utilization efficiency of onboard system storage, transmission, and computing resources by focusing on high-value data.

[0127] For example, such as Figure 3 The diagram shown is a flowchart illustrating an online health status assessment method for bearings in new energy vehicles based on digital twins, according to an embodiment of the present invention. The method includes the following steps:

[0128] S1. Based on the collected bearing vibration data, determine the operational fluctuation characteristic parameters and the anti-interference fusion characteristic parameters. The operational fluctuation characteristic parameters characterize the amplitude fluctuation state of the bearing's operation, while the anti-interference fusion characteristic parameters characterize the fault characteristics in the bearing vibration data that distinguish it from road surface interference.

[0129] Optionally, this step is performed by the feature extraction module 11 described above, and the specific implementation process can be found above.

[0130] S2. Based on the current road conditions, determine the adaptive reference vector for the road conditions.

[0131] Optionally, this step is performed by the benchmark determination module 12 described above, and the specific implementation process can be found above.

[0132] S3. Obtain the road interference correlation coefficient between the bearing vibration data and the pre-stored road interference vibration signal.

[0133] Optionally, this step is performed by the correlation analysis module 13 described above, and the specific implementation process can be found above.

[0134] S4. Determine the bearing health index based on the operational fluctuation characteristic parameters, anti-interference fusion characteristic parameters, road condition adaptive reference vector, and road surface interference correlation coefficient.

[0135] Optionally, this step is performed by the correction fusion module 14 described above, and the specific implementation process can be found above.

[0136] S5. Compare the bearing health index with the theoretical health index, and determine whether the bearing's health status is abnormal based on the comparison results. The theoretical health index is output based on the digital twin model.

[0137] Optionally, this step is performed by the evaluation feedback module 15 described above, and the specific implementation process can be found above.

[0138] The technical effects of the above-mentioned online assessment method for the health status of bearings in new energy vehicles based on digital twins can be found in the previous description of the technical effects of the online assessment system 10, and will not be repeated here.

[0139] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0140] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A digital twin-based online health status assessment system for bearings in new energy vehicles, characterized in that, The system includes: The feature extraction module is used to determine the operating fluctuation feature parameters and the anti-interference fusion feature parameters based on the collected bearing vibration data. The operating fluctuation feature parameters are used to characterize the operating amplitude fluctuation state of the bearing, and the anti-interference fusion feature parameters are used to characterize the fault features in the bearing vibration data that are different from road surface interference. The benchmark determination module is used to determine the road condition adaptive benchmark vector corresponding to the current road conditions of the vehicle. The correlation analysis module is used to obtain the road interference correlation coefficient between the bearing vibration data and the pre-stored road interference vibration signal; The correction fusion module is used to determine the bearing health index based on the operating fluctuation characteristic parameters, the anti-interference fusion characteristic parameters, the road condition adaptive reference vector, and the road surface interference correlation coefficient. The evaluation feedback module is used to compare the bearing health index with the theoretical health index and determine whether the bearing's health status is abnormal based on the comparison results; wherein, the theoretical health index is output based on a digital twin model.

2. The online health status assessment system for new energy vehicle bearings based on digital twins according to claim 1, characterized in that, The operational fluctuation characteristic parameters include at least a fluctuation index and a motion degree parameter; the feature extraction module, when determining the operational fluctuation characteristic parameters, is specifically used for: The fluctuation index is calculated based on the adjacent changes in bearing vibration amplitude within a preset time window; wherein, the fluctuation index is used to characterize the degree of local amplitude fluctuation. The motion degree parameter is calculated based on the change in the fluctuation index between adjacent preset time windows; wherein, the motion degree parameter is used to characterize the running turbulence state of the bearing.

3. The online health status assessment system for new energy vehicle bearings based on digital twins according to claim 2, characterized in that, When determining the anti-interference fusion feature parameters, the feature extraction module is specifically used for: Based on the spectral characteristics of the bearing vibration data, a frequency characteristic index is calculated; wherein, the frequency characteristic index is used to characterize the energy proportion of the bearing's own fault characteristic frequencies; The crest factor is calculated based on the peak and effective values ​​of the bearing vibration data; wherein the crest factor is used to characterize the impact state of the bearing. The anti-interference fusion characteristic parameters are calculated based on the motion degree parameter, the crest factor, and the frequency characteristic index.

4. The online health status assessment system for new energy vehicle bearings based on digital twins according to claim 1, characterized in that, When determining the adaptive road condition reference vector corresponding to the current vehicle driving conditions based on the current road conditions, the reference determination module is specifically used for: Obtain a preset baseline mapping relationship of characteristic parameters of bearings under healthy conditions, covering various typical road conditions; Based on the vehicle's current driving status information, determine the type of road condition the vehicle is currently in; The adaptive reference vector for road conditions is determined based on the road condition type and the reference mapping relationship of the feature parameters.

5. The online health status assessment system for new energy vehicle bearings based on digital twins according to claim 1, characterized in that, When the correlation analysis module obtains the road interference correlation coefficient between the bearing vibration data and the pre-stored road interference vibration signal, it is specifically used for: Obtain the spectral characteristics of the bearing vibration data within the current preset time window; The spectral characteristics of the bearing vibration data within the current preset time window are compared with a pre-stored reference spectral template for correlation analysis, and the road surface interference correlation coefficient is calculated based on the results of the correlation analysis.

6. The online health status assessment system for new energy vehicle bearings based on digital twins according to claim 1, characterized in that, When determining the bearing health index based on the operational fluctuation characteristic parameters, the anti-interference fusion characteristic parameters, the road condition adaptive reference vector, and the road surface interference correlation coefficient, the correction fusion module is specifically used for: The current values ​​of the operational fluctuation characteristic parameters and the anti-interference fusion characteristic parameters are determined respectively, and the characteristic deviations between them and the corresponding reference values ​​in the road condition adaptive reference vector are calculated. For each of the aforementioned feature deviations, the feature deviation weights are corrected based on the road surface interference correlation coefficient. The bearing health index is determined based on the weights of all corrected eigenvalue deviations.

7. The online health status assessment system for new energy vehicle bearings based on digital twins according to claim 4, characterized in that, The benchmark determination module is also used for: Acquire historical sensor data of the vehicle bearings under various typical road conditions with different combinations of vehicle speed and torque, while the bearings are in good condition, and statistically analyze the numerical distribution range of the historical sensor data for each road condition type. The baseline mapping relationship of the feature parameters is determined based on the numerical distribution range of the historical sensor data under each road condition type.

8. The online health status assessment system for new energy vehicle bearings based on digital twins according to claim 1, characterized in that, The evaluation feedback module is also used for: The digital twin model is driven to simulate the evolution of one or more preset fault modes; The actual change trajectory of the bearing health index is matched with the simulated fault evolution trajectory to predict the fault type and severity.

9. The online health status assessment system for new energy vehicle bearings based on digital twins according to any one of claims 1-8, characterized in that, The system also includes: The data acquisition module is used to collect the bearing vibration data.

10. A method for online assessment of the health status of bearings in new energy vehicles based on digital twins, characterized in that, The method includes: Based on the collected bearing vibration data, operational fluctuation characteristic parameters and anti-interference fusion characteristic parameters are determined; wherein, the operational fluctuation characteristic parameters are used to characterize the operating amplitude fluctuation state of the bearing, and the anti-interference fusion characteristic parameters are used to characterize the fault characteristics in the bearing vibration data that are different from road surface interference; Based on the current road conditions, determine the road condition adaptive reference vector corresponding to the current road conditions; Obtain the road interference correlation coefficient between the bearing vibration data and the pre-stored road interference vibration signal; The bearing health index is determined based on the operational fluctuation characteristic parameters, the anti-interference fusion characteristic parameters, the road condition adaptive reference vector, and the road surface interference correlation coefficient. The bearing health index is compared with the theoretical health index, and the bearing health status is judged to be abnormal based on the comparison results; wherein, the theoretical health index is output based on the digital twin model.

Images