Electric vehicle reducer gear fatigue life prediction method, system, device and medium considering load spectrum extrapolation and temperature field

Abstract

The application discloses a kind of electric vehicle reducer gear fatigue life prediction method, system, equipment and medium considering load spectrum extrapolation and temperature field.The method first obtains actual driving data of user, target driving condition is constructed in combination with K-means clustering optimized by particle swarm algorithm and Markov chain Monte Carlo method, the torque load history is converted, by rain flow counting and distribution fitting, Copula function and extreme value theory are introduced, and the load spectrum of extrapolated full life cycle is prepared;Calculate time-varying friction heat flux density and convective heat transfer coefficient, establish finite element model for thermal-mechanical coupling analysis to obtain equivalent stress, and predict gear fatigue life in combination with modified material S-N curve and extrapolated load spectrum.The application overcomes the defects of weak adaptability of standard condition, small sample data cannot cover extreme load and ignore thermal effect, significantly improves the fatigue life prediction accuracy, and constructs a multi-dimensional intellectual property protection network of software and hardware.

Description

Technical Field

[0001] This invention belongs to the field of reliability analysis and life prediction technology of electric vehicle transmission systems. Specifically, it relates to a method, system, computer equipment, and storage medium for predicting the fatigue life of electric vehicle reducer gears based on user driving condition construction, parametric rainflow load spectrum extrapolation, and thermo-mechanical coupling analysis. Background Technology

[0002] With the rapid development of the new energy vehicle industry, electric drive systems are gradually evolving towards higher speeds, greater integration, and higher power density. As a core transmission component of the electric drive system, the gears in the reducer often operate under extreme conditions of high speed, high torque, and frequent start-stop cycles during vehicle operation. The heat generated by tooth surface friction accumulates rapidly, easily leading to galling or fatigue failure, severely impacting the vehicle's power and fuel economy. Currently, the main technical challenges in predicting the fatigue life of reducer gears in electric vehicles are as follows: 1. Insufficient representativeness of operating conditions and weak adaptability of standard operating conditions: Existing fatigue durability test specifications (such as GB / T29307) are mostly based on single or phased steady-state standard driving cycles (such as NEDC, WLTC, etc.), which makes it difficult to take into account the differences in different geographical environments (such as mountains and plains) and different user driving habits (such as rapid acceleration). The operating conditions compiled by enterprises themselves are often set with too high a intensity, resulting in over-design of products, or they cannot truly reflect the load conditions of vehicles in actual operation due to insufficient representativeness.

[0003] 2. Small sample data cannot cover the entire life cycle, making extreme load prediction difficult: Due to limitations in testing time and cost, the measured load data often has a small sample size, covering only a small portion of the vehicle's entire life cycle. Directly using short-term data for life assessment cannot cover the extreme high loads that may occur throughout the entire life cycle. In addition, existing load extrapolation methods often face problems such as insufficient capture of the joint distribution characteristics of mean and amplitude, and low fitting accuracy when dealing with complex multi-peak load characteristics, failing to meet high reliability design requirements.

[0004] 3. Ignoring the thermo-mechanical coupling effect leads to low accuracy in life prediction: Existing gear fatigue life studies mostly focus on mechanical stress analysis, with less consideration for the influence of coupling factors such as temperature field. Under the high-speed operating conditions unique to electric vehicles, the frictional heat generated by gear meshing leads to a significant increase in temperature. Ignoring this thermal effect will overestimate gear fatigue life, resulting in a large error between the predicted results and the actual failure situation. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned defects of the prior art and provide a method, system, device and medium for predicting the fatigue life of gears in electric vehicle reducers that considers load spectrum extrapolation and temperature field, so as to solve the problems of weak adaptability to standard working conditions, inability of small sample data to cover extreme loads throughout the entire lifespan and low accuracy of fatigue life prediction due to neglecting thermal effects.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A method for predicting the fatigue life of gears in electric vehicle reducers, considering load spectrum extrapolation and temperature field, includes the following steps: S1. Obtain and preprocess the user's actual driving data, and construct the target driving conditions by combining the particle swarm optimization algorithm with the Markov chain Monte Carlo method. S2. Based on the vehicle driving equation, the target driving condition is converted into a torque-time load history. The distribution characteristics of the load mean and amplitude are obtained by rainflow counting. Based on the parametric rainflow extrapolation technology, the extrapolated full life cycle load spectrum is compiled. S3. Calculate the frictional heat flux density and convective heat transfer coefficient of the reducer gear system, establish a finite element model and perform thermo-mechanical coupled finite element analysis to obtain the equivalent stress under thermo-mechanical coupling. S4. Combining the material SN curve of the reducer gear, the extrapolated full-life cycle load spectrum, and the equivalent stress, the fatigue life of the reducer gear is predicted using the fatigue cumulative damage theory.

[0007] Furthermore, in step S1, after acquiring and preprocessing the user's actual driving data, the following steps are performed: S11. Divide the preprocessed user actual driving data into kinematic segments, with the following division rules: acceleration greater than a preset first acceleration threshold indicates an acceleration state; acceleration less than a preset second acceleration threshold indicates a deceleration state; acceleration within a closed interval between the preset second acceleration threshold and the preset first acceleration threshold, and vehicle speed lower than a preset vehicle speed threshold, indicates an idling state; acceleration within a closed interval between the preset second acceleration threshold and the preset first acceleration threshold, and vehicle speed not lower than the preset vehicle speed threshold, indicates a constant speed state. S12. Extract the feature parameters of the kinematic segment, and use principal component analysis to reduce the dimensionality of the feature parameters, and select the principal components whose cumulative contribution rate reaches the preset contribution rate threshold as feature vectors.

[0008] Furthermore, in step S1, the specific steps for constructing the target driving condition using the K-means clustering algorithm optimized by the particle swarm optimization algorithm and the Markov chain Monte Carlo method are as follows: S13. Encode the cluster center coordinates into the particle positions of the particle swarm optimization algorithm, and use the clustering effectiveness index as the fitness function for iterative updates to find the optimal initial cluster centers. S14. Based on the optimal initial cluster centers, the K-means algorithm is used to divide the dimensionality-reduced kinematic segments into multiple categories with different driving characteristics; S15. Calculate the state transition probability matrix between various driving segments, and use the Markov chain Monte Carlo method to randomly extract representative segments of each driving feature category and splice them together according to the time sequence to generate the target driving condition.

[0009] Furthermore, in step S2, the specific steps for compiling the extrapolated full-life-cycle load spectrum based on the parametric rainflow extrapolation technique are as follows: S21. Fit the load amplitude and load mean of the torque-time load history using the Weibull distribution model and the Gaussian mixture distribution model respectively, and use the chi-square test to determine the independence of the load mean and the load amplitude; S22. If the two sides are determined to be independent, a joint distribution model is directly constructed; if they are determined to be non-independent, a Copula function is introduced to construct a joint probability density function of mean and magnitude. S23. Based on the Conover probability criterion, the extreme loads within the entire life cycle are inferred from the measured data, and the load cycle frequency is extrapolated and extended by combining the joint probability density function to generate a two-dimensional extrapolated load spectrum of mean-amplitude-frequency.

[0010] Furthermore, after generating the two-dimensional extrapolated load spectrum of mean-amplitude-frequency, the following steps are also included: S24. Apply Goodman's theory and the variable mean method to transform the load amplitudes of different means in the two-dimensional extrapolated load spectrum into equivalent amplitudes with zero mean. S25. The equivalent amplitude is classified according to the principle of potential damage equivalence to generate a one-dimensional full-life cycle load spectrum.

[0011] Furthermore, in step S3, the specific steps for calculating the frictional heat flux density and convective heat transfer coefficient of the reducer gear system, establishing a finite element model, and performing thermo-mechanical coupled finite element analysis are as follows: S31. Calculate the time-varying frictional heat flux density based on gear meshing parameters, and calculate the convective heat transfer coefficients corresponding to the meshing area, gear end face area and non-meshing area based on the boundary conditions of different regions of the gear teeth. S32. In steady-state thermal analysis, the time-varying frictional heat flux density is used as the heat source, and the convective heat transfer coefficient is applied as the boundary condition to the corresponding region to solve for the temperature field of the gear body. S33. The temperature field of the gear body is imported into the static analysis module as a body load, and a mechanical torque load is applied at the same time for solution, so as to obtain the tooth surface contact stress under thermo-mechanical coupling as the equivalent stress.

[0012] Furthermore, in step S4, the specific steps for predicting the fatigue life of the reducer gear using fatigue cumulative damage theory, based on the material SN curve of the reducer gear, the extrapolated full-life cycle load spectrum, and the equivalent stress, are as follows: S41. Convert the tensor history of the equivalent stress into a scalar and perform statistical counting to obtain the load mean and stress amplitude. S42. Correct the material SN curve of the reducer gear using the average load value; S43. Combining the modified material SN curve of the reducer gear, the stress amplitude, and the number of cycles corresponding to the extrapolated full-life cycle load spectrum, calculate the total fatigue damage accumulation based on Miner's linear fatigue cumulative damage theory. When the total damage accumulation reaches 1, it is determined that the gear has experienced fatigue failure.

[0013] This invention also provides a fatigue life prediction system for electric vehicle reducer gears that considers load spectrum extrapolation and temperature field, comprising: The driving condition construction module is used to acquire and preprocess the user's actual driving data, and combine the particle swarm optimization algorithm with the Markov chain Monte Carlo method to construct the target driving condition. The load spectrum extrapolation module is used to convert the target driving condition into a torque-time load history based on the vehicle driving equation, obtain the distribution characteristics of the load mean and amplitude through rain flow counting, and compile the extrapolated full life cycle load spectrum based on parametric rain flow extrapolation technology. The thermo-mechanical coupling analysis module is used to calculate the frictional heat flux density and convective heat transfer coefficient of the reducer gear system, establish a finite element model and perform thermo-mechanical coupling finite element analysis to obtain the equivalent stress under thermo-mechanical coupling. The life prediction module is used to combine the extrapolated full life cycle load spectrum, the material SN curve of the reducer gear, and the equivalent stress to predict the fatigue life of the reducer gear using the fatigue cumulative damage theory.

[0014] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0016] Compared with the prior art, the present invention has the following significant advantages: 1. More realistic driving scenario construction, significantly enhanced algorithm generalization ability and robustness: This invention innovatively combines a particle swarm optimization-optimized K-means clustering algorithm with Markov chain Monte Carlo method to achieve high-precision reconstruction of real driving scenarios, effectively avoiding deviations caused by distortion of standard driving scenarios. Specifically, in the kinematic segment division logic, this invention abandons the absolute limitation of fixed values ​​and introduces adjustable "acceleration threshold" and "closed interval" judgment rules. This technique creates a rigorous mathematical closed loop in the state judgment logic, effectively improving the algorithm's generalization ability and robustness. This mechanism can flexibly adapt the corresponding driving scenario extraction threshold according to different vehicle parameters and complex actual driving environments, thereby further improving the accuracy and wide applicability of driving scenario construction.

[0017] 2. More accurate extreme load prediction, effectively overcoming the limitations of small sample data: This invention introduces parametric rainflow extrapolation technology, which not only improves fitting accuracy by utilizing models such as Gaussian mixtures, but also flexibly introduces the Copula function to construct a joint probability density function based on independence testing. Combined with Conover's extreme value theory, this invention successfully achieves a reliable extension from small sample measured data to the entire life cycle load history. This method not only solves the problems of long durability testing cycles and high costs, but also effectively predicts extreme loads that may occur but do not appear during the test, preventing prediction biases caused by missing extreme loads that could lead to early fatigue failure of gears.

[0018] 3. Introducing the thermo-mechanical coupling effect significantly improves the accuracy of fatigue life prediction: This invention overcomes the limitation of traditional life prediction that only considers mechanical loads, and establishes a thermo-mechanical coupling fatigue life analysis model that considers temperature effects. By quantitatively calculating the time-varying frictional heat flux density and the convective heat transfer coefficient of different regions of the gear, this invention accurately simulates the superimposed effect of gear temperature rise on stress state under high-speed operation. This technological breakthrough makes the final predicted life result more consistent with the actual service environment of the reducer under actual operating conditions, thereby effectively improving the overall reliability and durability assessment accuracy of the electric drive system.

[0019] 4. Constructing a multi-dimensional execution platform facilitates systematic integration and engineering deployment: This invention not only provides a high-precision lifetime prediction method but also simultaneously constructs a hardware and software execution architecture that includes a virtual system, computer equipment, and readable storage media. This modular system design enables fully automated calculations from data acquisition, operating condition construction, load extrapolation to lifetime prediction, greatly improving the data processing efficiency of complex thermo-mechanical coupling analysis. It allows for flexible deployment on cloud servers or enterprise R&D workstations, accelerating the engineering implementation and verification of the algorithm.

[0020] In summary, this invention improves the accuracy of fatigue life prediction for electric vehicle reducer gears through a systematic approach of "real-world working condition construction—full life cycle load extrapolation—thermomechanical coupling analysis," which has significant engineering application value for optimizing reducer design, reducing development costs, and shortening the R&D cycle. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method for constructing the driving conditions of the gear load spectrum of the reducer based on clustering and Markov chains in this invention.

[0022] Figure 2 This is a flowchart illustrating the extrapolation and compilation process of the full-life-cycle load spectrum of reducer gears based on parameter rainflow extrapolation in this invention.

[0023] Figure 3 This is a flowchart of the temperature field analysis and fatigue life prediction under thermal coupling of the reducer gear in this invention. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to the embodiments.

[0025] This embodiment provides a method for predicting the fatigue life of gears in electric vehicle reducers that considers load spectrum extrapolation and temperature field. The method mainly includes three core steps: constructing driving conditions based on user data, compiling and extrapolating the full life cycle load spectrum, and predicting the fatigue life of gears considering thermo-mechanical coupling.

[0026] 1. Construction of driving conditions based on improved clustering and Markov chains like Figure 1 As shown, the purpose of this step is to address the shortcoming that standard operating conditions cannot reflect the driving characteristics of real users. The specific process is as follows: 1. Data Acquisition and Preprocessing: Collect actual driving time-series data from target electric vehicle users, including information such as time, vehicle speed, and acceleration. Use statistical methods to remove outliers caused by sensor malfunctions, such as abnormal acceleration / deceleration or abnormal long-term parking. Specifically, an acceleration greater than 3.97 m / s² is specified. 2 Or deceleration less than -8m / s2 The data was deemed abnormal, and reasonable data was obtained through linear interpolation. Data exceeding 180 seconds was defined as idle data segments and was therefore discarded. The cleaned continuous driving data was divided into several small kinematic segments according to kinematic definitions. Each segment consisted of a single state or a combination thereof, such as idle, acceleration, constant speed, and deceleration, thus establishing an original driving sample database. Specifically, acceleration was defined as acceleration greater than a preset first acceleration threshold; deceleration was defined as deceleration; idling was defined as acceleration within a closed interval between the preset second acceleration threshold and the preset first acceleration threshold, and vehicle speed lower than a preset vehicle speed threshold; and constant speed was defined as acceleration within a closed interval between the preset second acceleration threshold and the preset first acceleration threshold, and vehicle speed not lower than the preset vehicle speed threshold. It should be noted that 'within the closed interval' means the range including the values ​​at both ends of the interval. In this embodiment, the preset first acceleration threshold is preferably 0.15 m / s². 2 The preset second acceleration threshold is preferably -0.15 m / s². 2 The preset vehicle speed threshold is preferably 0.5 km / h. This closed-interval rule also applies when the acceleration is exactly equal to the threshold endpoint.

[0027] That is, in this embodiment, the acceleration is specified to be greater than 0.15 m / s². 2 In acceleration state; acceleration less than -0.15 m / s² 2 The vehicle is in a deceleration state; the absolute value of the acceleration does not exceed 0.15 m / s². 2 Furthermore, a vehicle speed below 0.5 km / h is considered idling; the absolute value of acceleration does not exceed 0.15 m / s². 2 Furthermore, a vehicle speed of at least 0.5 km / h is considered a constant speed.

[0028] 2. Feature Parameter Extraction and Dimensionality Reduction: For each kinematic segment, its feature parameters are calculated, including average velocity, average acceleration, average deceleration, idle time ratio, acceleration time ratio, velocity standard deviation, and acceleration standard deviation. Given the strong correlation among feature parameters, Principal Component Analysis (PCA) is used to reduce the dimensionality of the feature parameter matrix. Principal components with a cumulative contribution rate of over 85% are selected as new eigenvectors to reduce computational dimensionality. The PCA analysis steps include data standardization, calculation of the correlation coefficient matrix, calculation of eigenvalues ​​and eigenvectors, calculation of the cumulative contribution rate of principal components, determination of the number of principal components, and calculation of principal component scores and loading matrices.

[0029] 3. PSO-based K-means Clustering: The K-means algorithm is used to cluster the dimensionality-reduced kinematic segments. To overcome the shortcomings of the traditional K-means algorithm, which is sensitive to initial cluster centers and prone to getting trapped in local optima, Particle Swarm Optimization (PSO) is introduced to find the optimal initial cluster centers. Specifically, the coordinates of the K cluster centers are encoded as particle positions, and a clustering effectiveness index (such as the CH index, Calinski-Harabasz Index) is used as the fitness function. Through iterative updates of the PSO algorithm, the position that maximizes the CH index is found as the optimal initial cluster center for K-means. Based on the optimal initial centers, K-means clustering is performed to classify all kinematic segments into categories with different driving characteristics (such as congestion, smooth traffic, and high speed).

[0030] 4. Markov Chain-Based Driving Condition Synthesis: Synthetic driving conditions are constructed using the Markov Chain Monte Carlo (MCMC) method. First, the number of transitions between different types of driving segments is counted, and the state transition probability matrix is ​​calculated. Based on this matrix, representative segments from each category are randomly extracted using the Monte Carlo method and spliced ​​together in a time series to obtain driving conditions reflecting specific user driving habits and geographical environment characteristics. The effectiveness of the driving conditions is verified by comparing the relative error and the joint probability distribution of velocity and acceleration between the synthetic driving conditions and the original data.

[0031] 2. Compilation of the full-life-cycle load spectrum based on parametric rainflow extrapolation like Figure 2 As shown, the purpose of this step is to address the problem that measured small sample data cannot cover extreme loads throughout the entire life cycle. The specific process is as follows: 1. Load Spectrum Conversion and Rainflow Counting: Based on the vehicle driving equation (including rolling resistance, air resistance, gradient resistance, and acceleration resistance), the driving conditions (speed-time curve) synthesized in step one are converted into torque-time load histories at the input of the drive motor and reducer. Rainflow counting is used to count the torque histories and extract load cycles, thereby obtaining the rainflow matrix of load mean and amplitude.

[0032] 2. Distribution function fitting and independence test: To adapt to the distribution characteristics of the mean and amplitude, the Weibull distribution model is used to fit the load amplitude, and the Gaussian mixture distribution model is used to fit the load mean, so as to improve the fitting accuracy.

[0033] It should be noted that, in other alternative embodiments, when fitting the distribution characteristics of the mean and amplitude of the load in the rainflow matrix, a normal distribution, a single Weibull distribution, or a mixed distribution model composed of them (such as a mixed Weibull distribution) can be flexibly selected for equivalent replacement based on the distribution characteristics of the actual load data.

[0034] Specifically, the maximum likelihood (MLE) method is used to estimate the parameters of the Weibull distribution, and the probability density function of the amplitude is: In the formula, y is the load amplitude, α is the shape parameter, β is the scale parameter, and ε is the position parameter, which is usually taken as 0.

[0035] A mixture Gaussian distribution is composed of multiple Gaussian distributions with different characteristic parameters, and its probability density function is: In the formula, This represents the mean of a Gaussian distribution. This represents the variance of the Gaussian distribution. These are the weighting coefficients. ≥0 and .

[0036] The Akaike Information Criterion (AIC criterion) is used to evaluate the model's fit and complexity, thereby determining the optimal number of basis functions, expressed as: In the formula, The number of basis functions in the mixed distribution model. Let be the maximum log-likelihood function value in the mixed distribution model. The value of n corresponding to the minimum AIC value in the solution results is selected as the optimal number of basis functions.

[0037] The EM algorithm is used to solve for the parameters in the mixture distribution. The E-step refers to solving for the conditional expectation of the log-likelihood function, and the M-step refers to maximizing the conditional expectation obtained in the E-step. The calculation steps are as follows: Calculate the likelihood function and take its logarithm: Solving for the conditional expectation of the log-likelihood function, the above equation is transformed into: Maximize the conditional expectation obtained from the solution: Repeat the above process iteratively until convergence.

[0038] The chi-square test is used to determine the independence of the mean and amplitude. If the mean and amplitude are independent, a joint distribution model is directly established; if they are not independent, a Copula function is introduced to construct a joint probability density function of the mean and amplitude.

[0039] According to Fisher's theorem, if the mean and amplitude are independent, then they follow a sequence of degrees of freedom (r-1)(s-1). distributed: In the formula, This represents the total number of cyclic loading cycles in the stress spectrum. , These are the average and amplitude levels, respectively, classified by the rainflow counting method; The mean is at the th Frequency of level; For the amplitude in the th Frequency of level; For the first Level mean and the first The frequency of cyclic loads corresponding to the amplitude level.

[0040] The upper a quantile of the distribution is: m=(r-1)(s-1) In the formula, m represents the degrees of freedom of the chi-square distribution. This can be obtained from the standard normal distribution. If the solution yields... Less than Then it can be determined that the mean and amplitude are independent of each other at the (1-a) confidence level.

[0041] 3. Extreme Value and Frequency Extrapolation: Based on the Conover probability criterion, extreme loads that may occur during the entire life cycle are inferred using measured data; combined with the joint probability density function, the frequency of load cycles is extrapolated and extended, and finally a two-dimensional extrapolated load spectrum containing mean, amplitude and frequency is generated.

[0042] The extreme values ​​of the mean and amplitude can be obtained from the extreme value probability expression, expressed as: In the formula, The extreme load probability is 10. -6 ; It is the maximum value of the mean; It is the minimum value of the mean; This is the maximum amplitude value; It is the mean probability density function. This represents the amplitude probability density.

[0043] The mean was divided into 8 equal levels, and the amplitude was divided at unequal intervals according to the Conover scaling factor. The number of load spectrum cycles was extrapolated to 10. 6 Assuming the mean and magnitude are independent, the joint probability density function, when double integraled, is expressed as: In the formula, This represents the number of load cycles. and These are the upper and lower limits of the mean integral, respectively. and These are the upper and lower limits of the amplitude integral, respectively.

[0044] 4. One-dimensional load spectrum transformation: Using the variable mean method and Goodman's theory, the amplitudes of different means in the two-dimensional load spectrum are transformed into equivalent amplitudes with zero mean, thus obtaining the one-dimensional load spectrum. The Goodman formula is: In the formula, For the two-dimensional load spectrum Amplitude; For the two-dimensional load spectrum Grade mean; For the equivalent number of Amplitude; This refers to the torsional limit strength of the material.

[0045] Based on the principle of potential damage equivalence, the equivalent load amplitudes are divided into 8 levels with unequal intervals, thus obtaining a one-dimensional full-life-cycle load spectrum suitable for SN curve analysis, expressed as: In the formula, The torque amplitudes at each stage of the one-dimensional load spectrum; This is the Conover scaling factor; This represents the highest amplitude of the equivalent torque.

[0046] Since loads below 60% of the material's fatigue limit cause relatively little damage, these loads are removed to improve fatigue test efficiency. Considering the influence of load interaction, a low-high-low loading sequence is used to sort the one-dimensional load spectrum.

[0047] 3. Gear fatigue life prediction considering temperature field like Figure 3 As shown, the purpose of this step is to address the low prediction accuracy caused by neglecting the effects of thermal coupling and load spectrum under high-speed operating conditions. The specific process is as follows: 1. Tooth profile discretization and establishment of dimensionless coordinates: The parameters on the gear tooth profile change with the contact position. To facilitate subsequent calculations, the tooth profile is discretized. In addition, dimensionless coordinates are established to represent the position of any meshing point on the gear, thereby calculating variables such as radius of curvature, relative sliding speed, contact stress, frictional heat flow, and convective heat transfer coefficient.

[0048] The dimensionless coordinates of any point Y on the meshing line are: In the formula, The pressure angle of the drive wheel at any point; It is the gear meshing angle.

[0049] The combined radius of curvature R of the gear is: In the formula, The transmission ratio; Let be the radius of the base circle of the drive wheel.

[0050] Treating the contact between gear teeth as the contact between two cylinders, the maximum tooth surface contact stress is calculated using Hertz's formula, expressed as: In the formula, The tangential force on the base circle; This refers to the total bearing width of the contact wire; , These are the Poisson's ratios of the driving and driven wheels, respectively. , These are the elastic moduli of the driving wheel and the driven wheel, respectively.

[0051] The average contact stress of the gear is expressed as: The relative sliding velocity V of the gears is expressed as: In the formula, The rotational speed (rpm) of the drive wheel; The pressure angle of the gear end face.

[0052] Based on the above calculations, the frictional heat generated at the gear meshing point can be obtained, expressed as: In the formula, The coefficient of friction for gears; It is the energy conversion coefficient from frictional energy to thermal energy.

[0053] Frictional heat flux refers to the heat transferred to the gear per unit time and area. In finite element analysis, its average frictional heat flux density needs to be calculated, expressed as: In the formula, This is the heat distribution coefficient; The rotational speed (rpm) of the driven wheel; The contact half-width is represented as: , Let be the absolute velocities in the tangential direction at the meshing point of the driving and driven gears, respectively, and express them as: In the formula, , These are the radii of curvature of the driving and driven gears at any meshing point, respectively.

[0054] From this, the average frictional heat flux density of the driving and driven wheels can be calculated. , The value of .

[0055] To calculate the convective heat transfer coefficient in different regions of a gear tooth, boundary conditions must first be established. The computational domain of a single gear tooth is divided into the meshing region, the gear end face region, the tooth tip, tooth root, and non-meshing region of the tooth surface, and the gear cross-section region. The boundary conditions for each region are as follows: For the meshing zone of the tooth surface, due to the relative sliding between the teeth in this region, there is both the generation and exchange of frictional heat, which belongs to the second and third types of boundary conditions, expressed as: For the gear end face region, this region only has convective heat transfer boundaries, belonging to the third type of boundary conditions, expressed as: For the non-meshing regions of the tooth tip, tooth root, and tooth surface, these regions only have convective heat transfer boundaries, belonging to the third type of boundary condition, expressed as: For the gear cross-section region, where heat transfer is relatively low, the boundary conditions are expressed as follows: In the formula, Thermal conductivity; The temperature gradient of each surface; , , These are the convective heat transfer coefficients for each region; , , These represent the temperatures of the convective heat transfer medium in each region; denoted as heat flux density at the meshing surface.

[0056] Based on the above boundary conditions, the convective heat transfer coefficient can be calculated as follows: The convective heat transfer coefficient of the tooth meshing zone is expressed as: In the formula, The Reynolds number of the convective heat transfer medium; The Prandt number is the value of the convective heat transfer medium. The pitch circle diameter of the gear; The density of the convective heat transfer medium; The kinematic viscosity of the mixed medium; Specific heat capacity of the convective heat transfer medium; The gear angular velocity; Let be the pitch circle radius of the gear.

[0057] The heat transfer coefficient between the gear end faces is expressed as: In the formula, It is an exponential constant. =2.

[0058] The convective heat transfer coefficients of the non-meshing areas of the tooth tip, tooth root, and tooth surface are expressed as: For the gear cross-section region, its convective heat transfer coefficient can be considered to be zero.

[0059] This completes the calculation of the input parameters for the gear thermal coupling analysis. By combining these parameters, the finite element analysis of the gear temperature field can be performed.

[0060] 2. Gear Temperature Field and Static Analysis: A three-dimensional finite element model of the reducer gear system is established. Based on the frictional heat flux density and convective heat transfer coefficient calculated by heat transfer and tribology theories, steady-state or transient thermal analysis is performed in finite element software. The meshing tooth surface is uniformly divided into multiple curved surfaces, and the heat flux density of each curved surface is determined based on its distance from the gear center. Combined with boundary conditions, the convective heat transfer coefficient is applied to calculate the temperature field of the gear body under high-speed conditions.

[0061] The nodal temperatures obtained from thermal analysis are imported as volume loads into the static analysis module for solution. This includes material settings, contact settings, mesh settings, and the application of constraints and loads. The stress state of the gear under the combined action of thermal and mechanical stresses is simulated, and the tooth surface contact stress results are obtained, completing the gear static analysis based on the temperature field. Those skilled in the art will understand that steady-state thermal analysis can be used to evaluate the ultimate temperature limit of the gear under continuous high-speed operation, while transient thermal analysis can more accurately simulate the impact of dynamic temperature fluctuations on the stress state under variable operating conditions such as frequent start-stop. In practical applications, the choice can be flexibly made according to engineering calculation requirements.

[0062] 3. Fatigue Life Prediction: During the operation of an electric vehicle, the reducer gears experience a high number and frequency of load cycles. Therefore, the nominal stress method is used for fatigue life analysis. This method is based on the material's SN curve and the linear fatigue cumulative damage theory to predict fatigue life. Miner's linear fatigue cumulative damage theory is the most commonly used fatigue life prediction method, used to assess the cumulative damage of materials under cyclic loading. When the total damage reaches 1, fatigue failure is considered to have occurred, expressed as: In the formula, This represents the total accumulated damage. The number of load blocks experienced before fatigue failure occurs; Let be the fatigue life under the i-th stress level; denoted as the number of cycles at stress level i.

[0063] When performing fatigue damage assessment, the system / computation module calculates the stress tensor history according to the following formula, expressed as: In the formula, Input load spectrum; is the scaling factor for the load spectrum; This is residual stress; For stress tensor; is the scaling factor for the stress tensor.

[0064] After obtaining the stress tensor, it is converted into a scalar and statistically counted. The load mean is used to correct the SN curve of the material, while the stress amplitude is calculated using the Miner criterion based on the corrected SN curve and the corresponding number of cycles.

[0065] In summary, based on the finite element analysis results of temperature field and statics, combined with the SN curve of gear material and the extrapolated full-life cycle load spectrum, the fatigue life of gear under thermal coupling can be calculated.

[0066] Through the above steps, this embodiment can calculate the gear fatigue life without considering the temperature field (mechanical load only) and with considering the thermo-mechanical coupling effect, as well as the difference in life under the load spectrum input before and after extrapolation, thereby achieving a more accurate life prediction that is more in line with the actual service environment.

[0067] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the fatigue life of gears in electric vehicle reducers, considering load spectrum extrapolation and temperature field, characterized in that, Includes the following steps: S1. Obtain and preprocess the user's actual driving data, and construct the target driving conditions by combining the particle swarm optimization algorithm with the Markov chain Monte Carlo method. S2. Based on the vehicle driving equation, the target driving condition is converted into a torque-time load history. The distribution characteristics of the load mean and amplitude are obtained by rainflow counting, and the extrapolated full life cycle load spectrum is compiled based on the parametric rainflow extrapolation technology. S3. Calculate the frictional heat flux density and convective heat transfer coefficient of the reducer gear system, establish a finite element model and perform thermo-mechanical coupled finite element analysis to obtain the equivalent stress under thermo-mechanical coupling. S4. Combining the material SN curve of the reducer gear, the extrapolated full-life cycle load spectrum, and the equivalent stress, the fatigue life of the reducer gear is predicted using the fatigue cumulative damage theory.

2. The method for predicting the fatigue life of electric vehicle reducer gears considering load spectrum extrapolation and temperature field as described in claim 1, characterized in that, In step S1, after obtaining and preprocessing the user's actual driving data, the following steps are performed: S11. Divide the preprocessed user actual driving data into kinematic segments, with the following division rules: acceleration greater than a preset first acceleration threshold indicates an acceleration state; acceleration less than a preset second acceleration threshold indicates a deceleration state; acceleration within a closed interval between the preset second acceleration threshold and the preset first acceleration threshold, and vehicle speed lower than a preset vehicle speed threshold, indicates an idling state; acceleration within a closed interval between the preset second acceleration threshold and the preset first acceleration threshold, and vehicle speed not lower than the preset vehicle speed threshold, indicates a constant speed state. S12. Extract the feature parameters of the kinematic segment, and use principal component analysis to reduce the dimensionality of the feature parameters, and select the principal components whose cumulative contribution rate reaches the preset contribution rate threshold as feature vectors.

3. The method for predicting the fatigue life of electric vehicle reducer gears considering load spectrum extrapolation and temperature field as described in claim 2, characterized in that, In step S1, the specific steps for constructing the target driving condition by combining the K-means clustering algorithm optimized by the particle swarm optimization algorithm with the Markov chain Monte Carlo method are as follows: S13. Encode the cluster center coordinates into the particle positions of the particle swarm optimization algorithm, and use the clustering effectiveness index as the fitness function for iterative updates to find the optimal initial cluster centers. S14. Based on the optimal initial cluster centers, the K-means algorithm is used to divide the dimensionality-reduced kinematic segments into multiple categories with different driving characteristics; S15. Calculate the state transition probability matrix between various driving segments, and use the Markov chain Monte Carlo method to randomly extract representative segments of each driving feature category and splice them together according to the time sequence to generate the target driving condition.

4. The method for predicting the fatigue life of electric vehicle reducer gears considering load spectrum extrapolation and temperature field as described in claim 1, characterized in that, In step S2, the specific steps for compiling the extrapolated full-life-cycle load spectrum based on the parametric rainflow extrapolation technique are as follows: S21. The load amplitude and load mean of the torque-time load history are fitted using the Weibull distribution model and the Gaussian mixture distribution model, respectively, and the independence of the load mean and the load amplitude is determined by the chi-square test. S22. If the two sides are determined to be independent, a joint distribution model is directly constructed; if they are determined to be non-independent, a Copula function is introduced to construct a joint probability density function of mean and magnitude. S23. Based on the Conover probability criterion, the extreme loads within the entire life cycle are inferred from the measured data, and the load cycle frequency is extrapolated and extended by combining the joint probability density function to generate a two-dimensional extrapolated load spectrum of mean-amplitude-frequency.

5. The method for predicting the fatigue life of electric vehicle reducer gears considering load spectrum extrapolation and temperature field as described in claim 4, characterized in that, After generating the two-dimensional extrapolated load spectrum of mean-amplitude-frequency, the following steps are also included: S24. Apply Goodman's theory and the variable mean method to transform the load amplitudes of different means in the two-dimensional extrapolated load spectrum into equivalent amplitudes with zero mean. S25. The equivalent amplitude is classified according to the principle of potential damage equivalence to generate a one-dimensional full-life cycle load spectrum.

6. The method for predicting the fatigue life of electric vehicle reducer gears considering load spectrum extrapolation and temperature field as described in claim 1, characterized in that, In step S3, the specific steps for calculating the frictional heat flux density and convective heat transfer coefficient of the reducer gear system, establishing a finite element model, and performing thermo-mechanical coupled finite element analysis are as follows: S31. Calculate the time-varying frictional heat flux density based on gear meshing parameters, and calculate the convective heat transfer coefficients corresponding to the meshing area, gear end face area and non-meshing area based on the boundary conditions of different regions of the gear teeth. S32. In steady-state thermal analysis, the time-varying frictional heat flux density is used as the heat source, and the convective heat transfer coefficient is applied as the boundary condition to the corresponding region to solve for the temperature field of the gear body. S33. The temperature field of the gear body is imported into the static analysis module as a body load, and a mechanical torque load is applied at the same time for solution, so as to obtain the tooth surface contact stress under thermo-mechanical coupling as the equivalent stress.

7. The method for predicting the fatigue life of electric vehicle reducer gears considering load spectrum extrapolation and temperature field as described in claim 1, characterized in that, In step S4, the specific steps for predicting the fatigue life of the reducer gear using fatigue cumulative damage theory, based on the material SN curve of the reducer gear, the extrapolated full-life cycle load spectrum, and the equivalent stress, are as follows: S41. Convert the tensor history of the equivalent stress into a scalar and perform statistical counting to obtain the load mean and stress amplitude. S42. Correct the material SN curve of the reducer gear using the average load value; S43. Combining the modified material SN curve of the reducer gear, the stress amplitude, and the number of cycles corresponding to the extrapolated full-life cycle load spectrum, calculate the total fatigue damage accumulation based on Miner's linear fatigue cumulative damage theory. When the total damage accumulation reaches 1, it is determined that the gear has experienced fatigue failure.

8. A fatigue life prediction system for electric vehicle reducer gears considering load spectrum extrapolation and temperature field, characterized in that, include: The driving condition construction module is used to acquire and preprocess the user's actual driving data, and combine the particle swarm optimization algorithm with the Markov chain Monte Carlo method to construct the target driving condition. The load spectrum extrapolation module is used to convert the target driving condition into a torque-time load history based on the vehicle driving equation, obtain the distribution characteristics of the load mean and amplitude through rain flow counting, and compile the extrapolated full life cycle load spectrum based on parametric rain flow extrapolation technology. The thermo-mechanical coupling analysis module is used to calculate the frictional heat flux density and convective heat transfer coefficient of the reducer gear system, establish a finite element model and perform thermo-mechanical coupling finite element analysis to obtain the equivalent stress under thermo-mechanical coupling. The life prediction module is used to combine the extrapolated full life cycle load spectrum, the material SN curve of the reducer gear, and the equivalent stress to predict the fatigue life of the reducer gear using the fatigue cumulative damage theory.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images