A bolt life prediction method and system for a new energy vehicle

Abstract

The application relates to the technical field of part detection, in particular to a bolt life prediction method and system for a new energy vehicle, which comprises the following steps: acquiring bolt vibration signals, vehicle body vibration signals, acoustic emission signals and pre-tightening forces at different moments of a plurality of bolts, taking each bolt as a sample; calculating event activity, fastening instability and degradation characteristic values of each sample; constructing a feature vector, identifying the posterior probability of each sample belonging to each damage type by using a classification model; calculating degradation gain rates of each damage type, determining damage coefficients of each sample belonging to each damage type, obtaining feature weights of each sample, optimizing an explicit proxy model to serve as a bolt life prediction model, and predicting the bolt life. The application introduces the heterogeneous influence of different damage modes into the model, reduces the prediction deviation problem caused by the heterogeneity of damage modes, and improves the accuracy and reliability of bolt life prediction.

Description

Technical Field

[0001] This application relates to the field of parts inspection technology, specifically to a method and system for predicting the life of bolts in new energy vehicles. Background Technology

[0002] New energy vehicles have become the mainstream direction of the automotive industry. In these vehicles, key components require numerous high-strength bolts for structural connection and fixation. The reliability of these bolt connections directly affects the vehicle's safety, durability, and NVH performance. Under complex and variable driving conditions, bolts are subjected to the coupled effects of multiple loads, including vibration, impact, thermal cycling, and electrochemical environments, making them highly susceptible to fatigue damage, preload decay, and even breakage. This can lead to serious safety incidents such as high-pressure leaks and power interruptions. Therefore, it is necessary to predict the service life of critical connecting bolts to ensure the safe operation of new energy vehicles and reduce maintenance costs.

[0003] However, since bolt failure is a gradual process that goes through different stages from material plastic deformation to macroscopic loosening, and its failure modes are diverse, such as fatigue fracture, stress corrosion cracking, and preload relaxation, the deterioration rate and the degree of harm to the structure vary among different failure modes. When using explicit proxy models to predict bolt life, the prediction bias caused by the heterogeneity of damage modes is seriously ignored, which makes it difficult for the accuracy and reliability of the model to meet the requirements of high-safety engineering. Summary of the Invention

[0004] To address the aforementioned technical problems, a method and system for predicting bolt life in new energy vehicles are provided to solve existing issues.

[0005] The solution to the technical problem addressed in this application is to provide a method and system for predicting the life of bolts in new energy vehicles, comprising the following steps: In a first aspect, embodiments of this application provide a method for predicting the life of bolts in new energy vehicles, the method comprising the following steps: The bolt vibration signal, vehicle body vibration signal, acoustic emission signal, and preload force at different times of multiple bolts are acquired, and the data of each bolt is used as a sample. The activity level and average level of signal intensity abrupt changes within the acoustic emission signal of each sample are analyzed, and the event activity of each sample is calculated. The fastening instability of each sample is calculated by analyzing the steepness and periodicity of the preload distribution and the discreteness of the preload distribution in the frequency domain. Based on the synergistic relationship between the changes in bolt vibration signal and vehicle body vibration signal, combined with the event activity and fastening instability, the degradation characterization value of each sample is obtained. The frequency domain energy distribution characteristics of acoustic emission signals, as well as the statistical characteristics of bolt vibration signals, vehicle body vibration signals, and preload, are evaluated. Combined with the deterioration characterization values, feature vectors are constructed, and a classification model is used to identify the posterior probability of a sample belonging to each damage type. By conducting simulation experiments on bolts with different damage types, the rate of change of bolt deterioration under different damage types is evaluated, the deterioration gain rate of each damage type is calculated, and the damage coefficient of each sample belonging to each damage type is determined by combining the posterior probability and the deterioration characterization value. The proportion of each sample under different damage types is analyzed, the feature weight of each sample is obtained, and the explicit surrogate model is optimized to serve as a bolt life prediction model to predict the bolt life.

[0006] Preferably, the calculation of the event activity of each sample includes: For each sample, a segmentation threshold for the acoustic emission signal is obtained, and the moment when the signal intensity within the acoustic emission signal is greater than the segmentation threshold is marked as an acoustic emission event; Divide the acoustic emission signal into multiple signal segments; count the number of acoustic emission events occurring in each signal segment. Calculate the mean signal strength corresponding to all acoustic emission events within each signal segment, denoted as the average strength; calculate the product of the quantity and the average strength; The event activity is the average of the product values ​​of all signal segments.

[0007] Preferably, the calculation of the fastening instability of each sample includes: For each sample, calculate the kurtosis of the preload at all times; perform frequency domain analysis on the preload at all times and obtain the power spectrum, calculate the root mean square bandwidth of the power spectrum and normalize it; calculate the sum of the kurtosis and the normalized root mean square bandwidth; obtain the autocorrelation coefficients of the preload at multiple preset hysteresis orders at all times, and obtain the maximum autocorrelation coefficient. The fastening instability is the ratio of the sum to the maximum autocorrelation coefficient.

[0008] Preferably, obtaining the degradation characterization values ​​for each sample includes: Calculate the correlation between bolt vibration signals and vehicle body vibration signals, and perform a negative mapping on the correlation. The degradation characterization value is the product of the result of the negative mapping, the event activity, and the tightness instability.

[0009] Preferably, the process of constructing the feature vector is as follows: Wavelet packet transform is used to decompose the acoustic emission signal into multiple decomposition layers. The energy of each wavelet packet coefficient vector in the last decomposition layer is calculated. The sub-frequency band corresponding to the wavelet packet coefficient vector with the maximum energy is selected and denoted as the main frequency band. The maximum energy is taken as the energy of the main frequency band. Calculate the bolt vibration signal, vehicle body vibration signal, and preload statistics for each sample at all times. The feature vector is composed of all the statistics, main frequency band, main frequency band energy, and degradation characterization values ​​corresponding to each sample.

[0010] Preferably, the calculation of the degradation gain rate for each damage type includes: selecting multiple bolts to conduct simulation experiments for different damage types, recording the complete life data of the bolts under each simulation experiment for different damage types, and dividing the complete life of the bolts into multiple time periods; calculating the degradation characterization value for each time period based on the bolt vibration signal, vehicle body vibration signal, acoustic emission signal, and preload at different times under each simulation experiment; performing linear fitting on the degradation characterization value for all time periods under each simulation experiment to obtain the slope of the fitted line as the degradation rate for each simulation experiment; and taking the average degradation rate of all simulation experiments under each damage type as the degradation gain rate for each damage type.

[0011] Preferably, determining the damage coefficient of each sample belonging to each damage type includes: multiplying the posterior probability of each sample belonging to each damage type by the degradation gain rate of the corresponding damage type as the damage risk degree of each sample belonging to each damage type; and multiplying the damage risk degree by the degradation characterization value as the damage coefficient of each sample belonging to each damage type.

[0012] Preferably, obtaining the feature weights of each sample includes: evaluating the degree of harm of each damage type to the bolt through experiments to obtain the prior weights of each damage type; calculating the sum of the damage coefficients of all samples belonging to each damage type; recording the ratio of the damage coefficient of each sample belonging to each damage type to the sum as the relative proportion, and using the prior weights as weights, performing a weighted summation of the relative proportions of each sample belonging to all damage types as the feature weights of each sample.

[0013] Preferably, optimizing the explicit proxy model specifically includes: constructing a weighted loss function using the feature weights, and training the explicit proxy model by minimizing the weighted loss function.

[0014] Secondly, embodiments of this application also provide a bolt life prediction system for new energy vehicles, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the bolt life prediction methods for new energy vehicles described above.

[0015] This application has at least the following beneficial effects: This application analyzes the frequency and intensity of sudden events in acoustic emission signals to calculate the event activity of each sample. Its advantage lies in its ability to keenly capture the energy release activity of microscopic damage within the bolt material. It also calculates the tightening instability of each sample, taking into account the random changes in bolt preload and the discreteness of the spectral distribution. Through comprehensive time and frequency domain analysis, it assesses the instability of the preload state during bolt connections, effectively distinguishing between healthy periodic fluctuations and instability caused by loosening. Furthermore, it obtains the degradation characterization value for each sample, considering the differences between bolt vibration signals and vehicle body vibration signals, and combining event activity and tightening instability to comprehensively assess the overall degradation degree of the bolt from multiple dimensions. Finally, it constructs a feature vector, uses a classification model to identify the damage type of the sample, and obtains the posterior probability of belonging to each damage type. Its advantage lies in considering the changing characteristics of various data to construct a high-dimensional, information-rich feature vector, which, through the classification model, can accurately identify the bolt damage type and obtain the posterior probability. This method reflects the confidence level of samples belonging to different damage types; calculates the degradation gain rate for each damage type, and determines the damage coefficient of each sample belonging to each damage type. Its beneficial effect lies in evaluating the deterioration rate of degradation characteristics under different damage types through simulation experiments, assessing the hazard of different damage types to bolts, and reflecting the high risk of rapid deterioration faced by bolts, as well as the characteristics of bolt working conditions conforming to a certain high-hazard damage mode, by combining posterior probability. The damage coefficient is used to assess the degree of damage risk of bolts under different damage types. The feature weights of each sample are obtained and the explicit surrogate model is optimized to serve as a bolt life prediction model. This method predicts bolt life by quantifying the differences in bolt damage type and degree among different samples into sample feature weights, and using these feature weights to optimize the explicit surrogate model. This introduces the heterogeneous influence of different damage modes into the model, reduces prediction bias caused by damage mode heterogeneity, improves the accuracy and reliability of bolt life prediction, and enhances the protection of bolt safety in new energy vehicles. Attached Figure Description

[0016] The following section provides a more detailed description of a bolt life prediction method for new energy vehicles, in conjunction with the accompanying drawings.

[0017] Figure 1 A flowchart illustrating the steps of a bolt life prediction method for new energy vehicles provided in this application embodiment; Figure 2 A flowchart illustrating the steps of the method for obtaining feature weights provided in this application embodiment. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of a bolt life prediction method and system for new energy vehicles, in conjunction with the accompanying drawings and implementation examples, is provided. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] 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 application pertains.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a bolt life prediction method for new energy vehicles according to an embodiment of this application. The method includes the following steps: Step 1: Obtain bolt vibration signals, vehicle body vibration signals, acoustic emission signals, and preload at different times for multiple bolts.

[0021] The primary function of bolts is to connect, secure, and seal components in new energy vehicles. During use, bolts are subjected to various complex operating conditions, such as vibration, temperature changes, load fluctuations, and yield deformation. These factors can cause bolt preload to decrease, loosen, or even break, leading to serious safety hazards. Therefore, accurately predicting bolt life allows for advance maintenance and replacement, preventing sudden malfunctions and improving vehicle reliability and safety.

[0022] For new energy vehicles, triaxial vibration acceleration sensors are deployed on the surface of the connected parts near the bolt head and on the vehicle body to collect bolt vibration signals and vehicle body vibration signals, respectively. The three axes of the two triaxial vibration acceleration sensors are aligned. Acoustic emission signals from the bolts are collected by deploying acoustic emission sensors on the connected parts near the bolt holes. An ultrasonic micro-sensor is deployed on the bolt head, and a portable ultrasonic axial force meter is used to collect the bolt preload force in real time; In this embodiment, the acquisition frequency of the triaxial vibration acceleration sensor is 50kHz, the acquisition frequency of the acoustic emission sensor and the ultrasonic axial force meter is 5MHz, and the data acquisition duration is 1 hour. As other implementation methods, the implementer can set the frequency according to the actual situation.

[0023] Collect bolt vibration signals, vehicle body vibration signals, acoustic emission signals, and preload at different times for different bolts, and record all data for each bolt as a sample; In this embodiment, data from 150 bolts on new energy vehicles are collected. As another implementation method, the implementer can set it according to the actual situation. Secondly, the data is time-aligned, and the maximum-minimum normalization method is used to normalize the collected data to eliminate dimensions. The time alignment algorithm and the maximum-minimum normalization method are well-known technologies and will not be described in detail here. As another implementation method, the implementer can use other methods of existing technology, such as Z-score normalization, etc. This embodiment does not impose any special restrictions on this.

[0024] Thus, bolt vibration signals, vehicle body vibration signals, acoustic emission signals, and preload at different times were obtained for multiple bolts.

[0025] Step 2: Analyze the activity of sudden changes in signal intensity within the acoustic emission signal of each sample and its average level, and calculate the event activity of each sample; calculate the fastening instability of each sample by analyzing the steepness and periodicity of the preload distribution and the discreteness of the preload distribution in the frequency domain; based on the synergistic changes of bolt vibration signal and vehicle body vibration signal, combined with the event activity and fastening instability, obtain the degradation characterization value of each sample.

[0026] Due to the influence of the operating conditions and usage behavior of new energy vehicles, the bolt fatigue conditions of different vehicles are not exactly the same, resulting in significant differences in the types of bolt damage, and the degree of harm caused by different types of damage to the bolts is also not exactly the same.

[0027] Secondly, when bolts are used, their deterioration process generally goes through two stages: the plastic deformation stage and the bolt loosening stage. During the plastic deformation of the material, the bolt preload will decrease. However, since the initial plastic deformation of the material is relatively slight and mainly occurs in local locations, the bolt will hardly slip or slip to a small degree at this time. As the bolt preload gradually decreases to the critical value, the bolt will rotate relative to the ground, resulting in poor structural stability. Therefore, during the plastic deformation stage, the working state of the bolt is mainly related to the local plastic deformation of the material. In the initial stage of plastic deformation, the deformation is relatively small, causing slight fluctuations in the preload. At this time, the change in preload exhibits a certain periodicity, and the peak value decreases. Simultaneously, in the frequency domain, the energy distribution of the preload is mainly concentrated near the fundamental frequency, but the high frequencies show a gradually increasing trend. Since the bolt hardly slips relative to the ground during this stage, the bolt vibration signal is basically consistent with the change in the vehicle body vibration signal. Secondly, as the bolt preload decreases, it gradually enters the bolt loosening stage. At this time, due to bolt vibration, the preload exhibits irregular vibration and periodic instability. Simultaneously, the high-frequency energy of the preload in the frequency domain increases significantly, the spectrum widens, and the randomness of the energy distribution increases. Furthermore, due to the bolt… When the bolt loosens, its vibration primarily originates from its own impact and shaking, rather than simply following the vibration of the vehicle body. While the bolt vibration signal is significantly influenced by the vehicle's own vibrations, a certain difference emerges between the two vibration signals. Simultaneously, the high-frequency energy of the preload signal in the frequency domain increases significantly, the spectrum widens, and the randomness of the energy distribution increases. Furthermore, the bolt's acoustic emission signal also fluctuates with plastic deformation. Under normal circumstances, the acoustic emission signal exhibits only a few random low-energy fluctuations. In the early stages of plastic deformation, localized plastic deformation leads to intermittent, high-amplitude, and high-energy acoustic emission signal fluctuations. As plastic deformation intensifies, cracks propagate on the bolt, causing more frequent and continuous signal fluctuations.

[0028] Based on the above analysis, the event activity is calculated by analyzing the fluctuation of the acoustic emission signal of the bolt, specifically as follows: For each sample, a segmentation threshold for the acoustic emission signal is obtained, and the moment when the signal intensity within the acoustic emission signal is greater than the segmentation threshold is marked as an acoustic emission event; In this embodiment, the Otsu threshold segmentation method is used to obtain the segmentation threshold of the signal intensity at all times within the acoustic emission signal. The Otsu threshold segmentation method is a well-known technique and will not be described in detail here.

[0029] It should be noted that the segmentation threshold is used to distinguish between valid acoustic emission events and noise.

[0030] The acoustic emission signal is divided into multiple signal segments; In this embodiment, the duration of the signal segment is 10ms. As for other implementation methods, the implementer can set it according to the actual situation.

[0031] Calculate the mean signal intensity corresponding to all acoustic emission events within each signal segment, and denot it as the average intensity; Count the number of acoustic emission events occurring in each signal segment; calculate the product of the number and the average intensity; The mean of the product values ​​of all signal segments within the acoustic emission signal is used as the event activity of each sample. It should be noted that a larger number indicates a more frequent occurrence of acoustic emission events within the signal segment, reflecting a higher frequency of energy release within the bolt material within that signal segment; a larger average intensity indicates a higher average energy of acoustic emission events within the signal segment, reflecting a more intense energy release within the bolt material, which may be related to more severe defects or crack propagation; a higher event activity indicates very active acoustic emission activity, reflecting intense lattice slip and dislocation movement within the material, as well as macroscopic relative sliding and impact between the bolt and the connected parts, and between threaded pairs, generating a large number of high-intensity friction and impact-type acoustic emission events, indicating intensified internal material damage.

[0032] Secondly, the periodic fluctuations of the preload force under each sample and its energy distribution in the frequency domain are analyzed to calculate the fastening instability, specifically: For each sample, calculate the kurtosis of the preload at all times; It should be noted that kurtosis calculation is a well-known technique and will not be elaborated upon here.

[0033] Frequency domain analysis was performed on the preload at all times, and the power spectrum was obtained. The root mean square bandwidth of the power spectrum was calculated and normalized. In this embodiment, a fast Fourier transform is used for frequency domain analysis. The fast Fourier transform and the calculation of the root mean square bandwidth are well-known techniques and will not be described in detail here. Secondly, the process of normalizing the root mean square bandwidth is as follows: the ratio of the root mean square bandwidth to a preset frequency is used as the result of normalization. The preset frequency is the Nyquist frequency, which is half of the sampling frequency. Since the sampling frequency of the preload is 5MHz, the preset frequency is 2.5MHz.

[0034] Obtain the autocorrelation coefficients of the preload force at multiple preset hysteresis orders at all times, and obtain the maximum autocorrelation coefficient; In this embodiment, the autocorrelation function is used to calculate the autocorrelation coefficient, and the preset lag order is an integer from 1 to 30. The autocorrelation function is a well-known technique and will not be described in detail here.

[0035] Calculate the sum of the kurtosis and the normalized root mean square bandwidth, and use the ratio of the sum to the maximum autocorrelation coefficient as the tightness instability of each sample. It should be noted that a larger kurtosis indicates more and more severe instantaneous spikes or impacts in the preload of the bolted connection represented by the sample, reflecting macroscopic loosening and impact in the bolted connection; a larger root mean square bandwidth indicates that the energy of the preload has spread from a few concentrated frequency points to a wide frequency band, with more high-frequency components, reflecting significant irregular changes in the bolt preload, possibly due to bolt loosening; a smaller maximum autocorrelation coefficient indicates lower similarity of the preload under different time delays, reflecting random changes or irregularities in the preload, indicating that the bolt condition may be relatively unstable; a larger fastening instability indicates a wider irregular change and spectral distribution width of the preload, and lower autocorrelation, which may reflect a higher degree of bolt loosening or deterioration, and an unstable bolt condition.

[0036] Furthermore, by analyzing the differences between bolt vibration signals and vehicle body vibration signals, and combining this with event activity and fastening instability, the degradation characterization value is determined, specifically as follows: Calculate the correlation between bolt vibration signals and vehicle body vibration signals, and perform a negative mapping on the correlation. In this embodiment, the correlation is calculated by the Pearson correlation coefficient between the bolt vibration signal and the vehicle body vibration signal. The Pearson correlation coefficient is a well-known technique and will not be elaborated upon here. As for other implementations, the implementer can use other methods from the prior art, such as the Spearman correlation coefficient, etc. This embodiment does not impose any special restrictions on this. Secondly, the specific process of negative mapping is as follows: negative mapping is performed using an exponential function. Let the correlation be denoted as... ,Will The result is used as the result of the negative mapping, where, It is an exponential function with the natural constant as the base.

[0037] The product of the negative mapping result, event activity, and tightness instability is used as the degradation characterization value for each sample. It should be noted that when the bolt's working condition is better and the degree of degradation is lower, there are fewer irregular fluctuations in the bolt preload, and the periodicity is stronger. Simultaneously, the energy of the preload in the frequency domain is mainly concentrated near the fundamental frequency, with relatively low spectral randomness, meaning lower fastening instability. Secondly, the higher the synchronicity between the bolt vibration signal and the vehicle body vibration signal, the smaller the negative mapping result. Furthermore, fewer acoustic emission events occur, and the corresponding acoustic emission signal intensity is lower, indicating lower event activity, resulting in a smaller degradation characterization value. Conversely, a larger degradation characterization value indicates a worse working condition and more severe degradation of the bolt represented by the sample.

[0038] Thus, the degradation characterization values ​​for each sample are obtained.

[0039] Step 3: Evaluate the frequency domain energy distribution characteristics of the acoustic emission signal and the statistical characteristics of the bolt vibration signal, vehicle body vibration signal, and preload. Combine the degradation characterization values ​​to construct a feature vector, use a classification model to identify the damage type of the sample, and obtain the posterior probability of each damage type. Through simulation experiments on bolts with different damage types, evaluate the rate of change of bolt degradation under different damage types, calculate the degradation gain rate of each damage type, and combine the posterior probability and degradation characterization values ​​to determine the damage coefficient of each sample belonging to each damage type.

[0040] Furthermore, in actual use, different types of bolt damage result in varying rates of degradation and degrees of harm. Therefore, when predicting bolt life, it is necessary to identify the different damage types represented by the bolts in each sample and assess the corresponding damage levels. Secondly, since different types of bolt damage all induce corresponding acoustic emission events, a feature vector is constructed based on the frequency range and frequency energy of energy concentration in the acoustic emission signal, combined with degradation characterization values, and statistics on bolt vibration signals, vehicle body vibration signals, and preload. Specifically: Wavelet packet transform is used to decompose the acoustic emission signal into multiple decomposition layers. The energy of each wavelet packet coefficient vector in the last decomposition layer is calculated. The sub-frequency band corresponding to the wavelet packet coefficient vector with the maximum energy is selected and denoted as the main frequency band. The maximum energy is taken as the energy of the main frequency band. It should be noted that the decomposition level of the wavelet packet transform is set to 4, and the wavelet basis is the db4 wavelet. The wavelet packet transform is a well-known technique and will not be described in detail here.

[0041] Calculate the bolt vibration signal, vehicle body vibration signal, and preload statistics for each sample at all times. In this embodiment, the bolt vibration signal, vehicle body vibration signal, and preload statistics of each sample are measured by calculating kurtosis, skewness, root mean square, and energy entropy. The calculation of kurtosis, skewness, root mean square, and energy entropy is a well-known technique and will not be described in detail here.

[0042] The feature vector is composed of all the statistics, main frequency band, main frequency band energy and degradation characterization values ​​corresponding to each sample. Furthermore, based on the feature vector, the damage type of the sample is identified, specifically as follows: The feature vectors of each sample are used as input to the classification model to obtain the posterior probability of each sample belonging to each damage type. In this embodiment, the classification model adopts a Gaussian mixture model, which is a well-known technique and will not be described in detail here. The training process of the Gaussian mixture model is as follows: By selecting multiple bolts and conducting simulation experiments for different damage types, the complete life data of the bolts under each simulation experiment for different damage types was recorded, and the complete life of the bolts was divided into multiple time periods. Based on the bolt vibration signal, vehicle body vibration signal, acoustic emission signal, and preload at different times under each simulation experiment, the degradation characterization value of each time period was calculated, and the feature vector of each time period was obtained. Based on the feature vectors of all time periods under all simulation experiments for different damage types, the Gaussian mixture model was trained. The maximum number of iterations in the Gaussian mixture model was set to 200, the initial parameter estimation method was K-means clustering, and the number of classifications was determined by the elbow rule. Thus, the trained Gaussian mixture model was obtained.

[0043] It should be noted that each time period lasts for 1 hour. Secondly, the types of damage include fatigue fracture, stress corrosion, and macroscopic loosening.

[0044] Furthermore, the deterioration rate of the bolts under different damage types is evaluated, and the damage coefficient is calculated by combining the posterior probability and the degradation characterization value, specifically: Linear fitting is performed on the degradation characterization values ​​for all time periods under each simulation experiment, and the slope of the fitted line is obtained as the degradation rate for each simulation experiment. In this embodiment, the least squares method is used for linear fitting. The least squares method is a well-known technique and will not be described in detail here.

[0045] The average degradation rate of all simulation experiments under each damage type is used as the degradation gain rate for each damage type. The product of the posterior probability of each sample belonging to each damage type and the degradation gain rate of the corresponding damage type is used as the damage risk degree of each sample belonging to each damage type. The product of the damage risk level and the degradation characterization value is used as the damage coefficient for each sample belonging to each damage type. It should be noted that the higher the posterior probability, the higher the likelihood of the bolt experiencing this type of damage; the higher the degradation gain rate, the faster the overall degradation rate of the bolt under this damage type, reflecting the stronger destructiveness of this damage type to the bolt; the higher the damage risk, the higher the risk of the bolt facing rapid deterioration, reflecting that the working state of the bolt is very consistent with the characteristics of a certain high-hazard damage mode; the higher the damage coefficient, the more severe the damage to the bolt under this damage type, reflecting that the working connection state of the bolt is more likely to be in a highly critical state, and the higher the risk of early failure.

[0046] Thus, the damage coefficients for each sample belonging to each damage type are obtained.

[0047] Step 4: Analyze the proportion of damage coefficient under different damage types, obtain the feature weights of each sample, optimize the explicit surrogate model, and use it as a bolt life prediction model to predict the bolt life.

[0048] Furthermore, since different damage types pose varying degrees of harm to bolts, bolts with different damage types should have different influence weights when constructing subsequent life prediction models. Therefore, based on the damage coefficient, the feature weights are calculated as follows: The impact of each damage type on the bolt was evaluated through experiments, and the prior weights of each damage type were obtained. In this embodiment, the prior weight represents the degree of harm of each damage type to the bolt. The degree of harm of each damage type to the bolt is evaluated by experiments, and the prior weight is assigned a value by evaluating the average maintenance cost caused by each damage type. In this embodiment, the prior weight of fatigue fracture damage type is 0.65, the prior weight of macro loosening damage type is 0.25, and the prior weight of stress corrosion damage type is 0.1.

[0049] Calculate the sum of the damage coefficients for all samples belonging to each damage type; The ratio of the damage coefficient of each sample belonging to each damage type to the sum is denoted as the relative proportion; Using prior weights as weights, the relative proportions of each sample belonging to all damage types are weighted and summed to serve as the feature weights of each sample. It should be noted that a larger feature weight indicates a more severe degree of bolt deterioration in the sample. The flowchart of the feature weight acquisition method provided in this embodiment is as follows: Figure 2 As shown.

[0050] Obtain the inherent parameters and actual response values ​​of all samples. The inherent parameters include working load, connected component parameters, preload, and fastener parameters. The actual response value is the bolt life. It should be noted that the working load refers to the external force that the bolt bears during operation, the parameters of the connected parts refer to the material stiffness, thickness, contact area and friction coefficient of the parts connected by the bolt, and the fastener parameters refer to the property parameters of the bolt itself, including specifications, material properties, fastener friction coefficient, etc.

[0051] Based on the inherent parameters and actual response values ​​of all samples, the polynomial response surface methodology is selected to construct an explicit surrogate model. Feature weights are introduced into the loss function, and the parameters of the explicit surrogate model are optimized by minimizing the loss function. This model is then used as a bolt life prediction model to predict the bolt life. In this embodiment, the loss function is selected as maximum absolute error and mean squared error to evaluate the accuracy of the explicit proxy model. The maximum absolute error and mean squared error are well-known techniques and will not be described in detail here.

[0052] It should be noted that the explicit proxy model is a well-known technique and will not be elaborated upon here.

[0053] The process of introducing feature weights into the loss function is as follows: in, Indicates the maximum absolute error. Indicates mean square error. Indicates the first The actual response value of each sample Indicates the first Approximate model predictions for each sample. Indicates the first Feature weights of each sample This represents the total number of samples. This represents the function that takes the maximum value.

[0054] Based on the same inventive concept as the above method, this application embodiment also provides a bolt life prediction system for new energy vehicles, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described bolt life prediction methods for new energy vehicles.

[0055] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0056] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0057] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application, without departing from the content of the technical solution of this application, shall fall within the protection scope of the technical solution of this application.

Claims

1. A method for predicting the life of bolts in new energy vehicles, characterized in that, The method includes the following steps: The bolt vibration signal, vehicle body vibration signal, acoustic emission signal, and preload force at different times of multiple bolts are acquired, and the data of each bolt is used as a sample. The activity level and average level of signal intensity abrupt changes within the acoustic emission signal of each sample are analyzed, and the event activity of each sample is calculated. The fastening instability of each sample is calculated by analyzing the steepness and periodicity of the preload distribution and the discreteness of the preload distribution in the frequency domain. Based on the synergistic relationship between the changes in bolt vibration signal and vehicle body vibration signal, combined with the event activity and fastening instability, the degradation characterization value of each sample is obtained. The frequency domain energy distribution characteristics of acoustic emission signals, as well as the statistical characteristics of bolt vibration signals, vehicle body vibration signals, and preload, are evaluated. Combined with the deterioration characterization values, feature vectors are constructed, and a classification model is used to identify the posterior probability of a sample belonging to each damage type. By conducting simulation experiments on bolts with different damage types, the rate of change of bolt deterioration under different damage types is evaluated, the deterioration gain rate of each damage type is calculated, and the damage coefficient of each sample belonging to each damage type is determined by combining the posterior probability and the deterioration characterization value. The proportion of each sample under different damage types is analyzed, the feature weight of each sample is obtained, and the explicit surrogate model is optimized to serve as a bolt life prediction model to predict the bolt life.

2. The bolt life prediction method for new energy vehicles as described in claim 1, characterized in that, The calculation of event activity for each sample includes: For each sample, a segmentation threshold for the acoustic emission signal is obtained, and the moment when the signal intensity within the acoustic emission signal is greater than the segmentation threshold is marked as an acoustic emission event; Divide the acoustic emission signal into multiple signal segments; count the number of acoustic emission events occurring in each signal segment. Calculate the mean signal strength corresponding to all acoustic emission events within each signal segment, denoted as the average strength; calculate the product of the quantity and the average strength; The event activity is the average of the product values ​​of all signal segments.

3. The bolt life prediction method for new energy vehicles as described in claim 1, characterized in that, The calculation of the fastening instability of each sample includes: For each sample, calculate the kurtosis of the preload at all times; perform frequency domain analysis on the preload at all times and obtain the power spectrum, calculate the root mean square bandwidth of the power spectrum and normalize it; calculate the sum of the kurtosis and the normalized root mean square bandwidth; obtain the autocorrelation coefficients of the preload at multiple preset hysteresis orders at all times, and obtain the maximum autocorrelation coefficient. The fastening instability is the ratio of the sum to the maximum autocorrelation coefficient.

4. The bolt life prediction method for new energy vehicles as described in claim 1, characterized in that, The degradation characterization values ​​obtained for each sample include: Calculate the correlation between bolt vibration signals and vehicle body vibration signals, and perform a negative mapping on the correlation. The degradation characterization value is the product of the result of the negative mapping, the event activity, and the tightness instability.

5. The bolt life prediction method for new energy vehicles as described in claim 1, characterized in that, The process of constructing the feature vector is as follows: Wavelet packet transform is used to decompose the acoustic emission signal into multiple decomposition layers. The energy of each wavelet packet coefficient vector in the last decomposition layer is calculated. The sub-frequency band corresponding to the wavelet packet coefficient vector with the maximum energy is selected and denoted as the main frequency band. The maximum energy is taken as the energy of the main frequency band. Calculate the bolt vibration signal, vehicle body vibration signal, and preload statistics for each sample at all times. The feature vector is composed of all the statistics, main frequency band, main frequency band energy, and degradation characterization values ​​corresponding to each sample.

6. The bolt life prediction method for new energy vehicles as described in claim 1, characterized in that, The calculation of the degradation gain rate for each damage type includes: selecting multiple bolts to conduct simulation experiments for different damage types, recording the complete life data of the bolts under each simulation experiment for different damage types, and dividing the complete life of the bolts into multiple time periods; calculating the degradation characterization value for each time period based on the bolt vibration signal, vehicle body vibration signal, acoustic emission signal, and preload at different times under each simulation experiment; performing linear fitting on the degradation characterization value for all time periods under each simulation experiment, obtaining the slope of the fitted line as the degradation rate for each simulation experiment; and taking the average degradation rate of all simulation experiments under each damage type as the degradation gain rate for each damage type.

7. The bolt life prediction method for new energy vehicles as described in claim 1, characterized in that, The determination of the damage coefficient of each sample belonging to each damage type includes: multiplying the posterior probability of each sample belonging to each damage type with the degradation gain rate of the corresponding damage type as the damage risk degree of each sample belonging to each damage type; and multiplying the damage risk degree with the degradation characterization value as the damage coefficient of each sample belonging to each damage type.

8. The bolt life prediction method for new energy vehicles as described in claim 1, characterized in that, The process of obtaining the feature weights of each sample includes: evaluating the degree of harm of each damage type to the bolt through experiments to obtain the prior weights of each damage type; calculating the sum of the damage coefficients of all samples belonging to each damage type; recording the ratio of the damage coefficient of each sample belonging to each damage type to the sum as the relative proportion; using the prior weights as weights, weighting and summing the relative proportions of each sample belonging to all damage types as the feature weights of each sample.

9. The bolt life prediction method for new energy vehicles as described in claim 1, characterized in that, The optimization of the explicit proxy model specifically includes: constructing a weighted loss function using the feature weights, and training the explicit proxy model by minimizing the weighted loss function.

10. A bolt life prediction system for new energy vehicles, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the bolt life prediction method for new energy vehicles as described in any one of claims 1-9.

Images