A method for accelerated fatigue testing of crankshaft
By establishing an empirical model of the change of crankshaft crack depth with the number of cycles and using particle filtering algorithms, the problems of long and high cost of crankshaft acceleration fatigue test time are solved, and accurate prediction of the remaining life of crankshaft and shortening of test time are achieved.
Patent Information
- Application Number
- CN202111410973.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-25
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2041-11-25
AI Technical Summary
At this stage, the crankshaft acceleration fatigue test time is long and the experimental cost is high, making it difficult to accurately predict the remaining life of the crankshaft.
An empirical model is established for the change of crankshaft crack depth with the number of cycles, a particle filtering algorithm is used to predict the remaining life of the crankshaft, and a recursive prediction is performed through the state transfer equation and the observation equation to shorten the test time.
Accurate prediction of crankshaft fatigue life is achieved, shortening test time, reducing experimental costs and improving prediction accuracy.
Smart Images

Figure CN114091203B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of crankshaft testing, in particular to a method for accelerated fatigue testing of a crankshaft. Background Art
[0002] At present, a resonant crankshaft bending fatigue test bench is commonly used to study the accelerated fatigue life of the crankshaft. The crankshaft is loaded with an enhanced load. The fatigue test bench studies the life of the crankshaft based on the crack depth and number of cycles of the crankshaft after loading the bending moment.
[0003] However, at present, the number of cycles when the crankshaft reaches fracture is relatively large, which can reach 10 6 Second, the experimental cost is high and the experimental time is long. Based on this situation, it is hoped that a method can be found to predict the accelerated fatigue life of the crankshaft, and only the partial data obtained by the test bench can be used to accurately predict the remaining life before failure. Summary of the invention
[0004] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.
[0005] In view of the above problems and / or the problems existing in the existing crankshaft bending fatigue test, the present invention is proposed.
[0006] Therefore, the object of the present invention is to provide a method for accelerated crankshaft fatigue testing, which can shorten the crankshaft fatigue test time and achieve an accelerated effect.
[0007] To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions:
[0008] A method for accelerated fatigue testing of a crankshaft, comprising:
[0009] S1. Establish an empirical model of the change of crankshaft crack depth with the number of cycles: y = a exp(bx) + c exp(dx);
[0010] S2. Optimize the established empirical model to obtain the recursive formula of the crankshaft remaining life prediction model:
[0011] y k =y k-1 ·exp(b)+c·exp[d(k-1)][1-exp(bd)]+u k ,u k ~N(0,σ w ),
[0012] Among them, y k is the crack depth of the crankshaft in the kth cycle under a certain bending moment; b, c, d are the parameters of the improved model of the remaining life of the crankshaft; u k is process noise, with a mean of 0 and a variance of σ w The standard normal distribution of
[0013] S3. Establish a state transfer equation and an observation equation for predicting the remaining life of the crankshaft, wherein the state transfer equation is a recursive formula of the crankshaft remaining life prediction model, and the observation equation is:
[0014]
[0015] in: is the crack depth of the crankshaft observed at the kth cycle under a certain bending moment; v k is the observation noise, with a mean of 0 and a variance of σ w The standard normal distribution of
[0016] S4. Based on the established state transfer equation and observation equation for predicting the remaining life of the crankshaft as empirical data, a particle filter algorithm is used to predict the fatigue life of the crankshaft.
[0017] As a preferred embodiment of the method of the present invention, the specific steps of establishing an empirical model of the change of crankshaft crack depth with the number of cycles: y=a exp(bx)+c exp(dx) are as follows:
[0018] S101, obtaining test data showing that the crack depth of a crankshaft increases with the number of cycles;
[0019] S102, performing nonlinear fitting on the obtained test data to obtain a nonlinear function;
[0020] S103. Based on the principle of minimizing the fitting error of the change law of the crack depth of the crankshaft with the increase of the number of cycles, the results obtained by different fitting formulas are compared, and the double exponential empirical model is selected.
[0021] As a preferred embodiment of the method of the present invention, in step S103, the results obtained by comparing different fitting formulas according to the principle of minimizing the fitting error of the crack depth of the crankshaft as the number of cycles increases are as follows: 2 The root mean square error (RMSE) was used to judge the quality of each model in fitting the test data. 2 The value range is [0,1], R 2 The closer it is to 1, the better the fitting effect is. 2If the RMSE value is close to 0, the model matching is not ideal. The smaller the RMSE value is, the better the model fitting effect is. 2 The calculation formula of RMSE is as follows:
[0022]
[0023] Among them, y i represents the i-th original data point, is the average value of the original data, is the model calculated value, and n is the number of original data points.
[0024] As a preferred embodiment of the method of the present invention, in step S4, based on the established state transfer equation and observation equation for predicting the remaining life of the crankshaft as empirical data, the specific steps of using the particle filter algorithm to predict the fatigue life of the crankshaft are as follows:
[0025] S401, initial state
[0026] Set the initialization iteration number t=0, the number of particles N and various parameter values involved in the algorithm, use a large number of particles to simulate x(t), the particles are evenly distributed in the space, and the initial weight values of all particles in the particle swarm are set to 1 / N;
[0027] S402: Prediction status
[0028] From the known process noise ω t N noise particles are obtained by sampling, and the predicted value of the particle position at the next iteration is calculated according to the state transfer equation
[0029]
[0030] In the formula, x k Represents the current state value of the system, y k represents the current observation value of the system, f(·) represents the state transfer function, which characterizes the functional relationship between the state at the previous moment and the current moment, h(·) represents the observation function, and u k represents the process noise of the system, v k represents the observation noise of the system;
[0031] S403, Importance Sampling
[0032] Set the number of iterations t = t + 1, and from the importance density function To randomly sample particle clusters in , use express;
[0033] S404, calibration phase
[0034] Calculate the weight of each particle and evaluate the weight of the predicted particles. Calculate the weight of all particles in the particle set
[0035]
[0036] S405, weight normalization
[0037] Normalize the weights,
[0038]
[0039] S406, resampling
[0040] Calculate the effective number of particles N eff , set the threshold value N th , if and only if N eff <N th When, yes Resample and set the weight to 1 / N;
[0041] S407, status output
[0042] Perform status output;
[0043]
[0044] S408: Is the iteration finished?
[0045] Determine whether the process is finished, if so, exit; otherwise, go to step S302.
[0046] As a preferred embodiment of the method of the present invention, after obtaining the test data showing that the crankshaft crack depth increases with the number of cycles, the step S101 also includes processing the obtained test data to discard the data on the crankshaft crack depth and the corresponding number of cycles in which the crankshaft crack initiated in the previous part.
[0047] As a preferred embodiment of the method of the present invention, in step S101, the specific steps for obtaining the test data showing that the crack depth of the crankshaft increases with the number of cycles are as follows: the crankshaft fatigue test bench is loaded with an enhanced load to cycle the crankshaft, and the number of cycles until failure is calculated.
[0048] As a preferred embodiment of the method of the present invention, in step S304, the effective particle number N eff The calculation formula is as follows:
[0049]
[0050] As a preferred embodiment of the method of the present invention, the method further comprises the following steps:
[0051] S5. Logarithmically process the fatigue life and stress obtained by predicting the fatigue life of the crankshaft, and fit the obtained fatigue life and stress by the least square method to obtain the logarithmic linear relationship between the stress and fatigue life of the part;
[0052] S6. Based on the linear relationship obtained by fitting, use the loading moment of one fatigue life as Q CI Point, calculate the fatigue limit load:
[0053]
[0054] Where, FS i is the estimated value of the parent fatigue limit measured by the i-th specimen, N i is the number of tests when the i-th specimen fails, S i is the load of the i-th specimen, S A Q is obtained by fitting using the least squares method. CI Stress value at point, N A The S obtained at the stress A The life value of the crankshaft under the horizontal action of;
[0055] S7. Estimate the fatigue limit of the parent body measured for the i-th specimen and calculate the median rank:
[0056]
[0057] in, i k is the average failure number, n is the sample size, and j represents the total failure number;
[0058] On this basis, the normal distribution model is used for analysis to obtain the fatigue limit load of the crankshaft with a 50% survival probability required for actual engineering.
[0059] Compared with the prior art, the present invention has the following beneficial effects: in order to address the defect of a long bending fatigue test cycle of a traditional resonant crankshaft, the present invention determines the crankshaft fatigue failure crack threshold, and based on a time series analysis method, develops a crankshaft bending residual fatigue life prediction model, and experimentally verifies the prediction results based on multiple groups of test data, and compares them with the fatigue limit load obtained from actual data. The remaining life of the crankshaft can be accurately predicted before it fails, thereby achieving an accelerated effect and shortening the crankshaft fatigue test time. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:
[0061] Figure 1 A flow chart of a crankshaft accelerated fatigue test method according to the present invention;
[0062] Figure 2 A flowchart of the specific steps of a crankshaft accelerated fatigue test method of the present invention;
[0063] Figure 3 This is a prediction result diagram of the first set of data from the initial to 5mm prediction of a crankshaft accelerated fatigue test method of the present invention;
[0064] Figure 4 This is a prediction result diagram of the second group of data from the initial to 5mm prediction of a crankshaft accelerated fatigue test method of the present invention;
[0065] Figure 5 This is a prediction result diagram of the third group of data from the initial to 5mm prediction of a crankshaft accelerated fatigue test method of the present invention;
[0066] Figure 6 This is a prediction result diagram of the first set of data from the initial to 6mm prediction of a crankshaft accelerated fatigue test method of the present invention;
[0067] Figure 7 This is a prediction result diagram of the second group of data of a crankshaft accelerated fatigue test method of the present invention from the initial to 6mm prediction;
[0068] Figure 8 This is a prediction result diagram of the second group of data of a crankshaft accelerated fatigue test method of the present invention from the initial to 6mm prediction;
[0069] Fig. 9 This is a prediction result diagram of the first set of data from the initial to 7mm prediction of a crankshaft accelerated fatigue test method of the present invention;
[0070] Fig.10 This is a prediction result diagram of the second group of data of a crankshaft accelerated fatigue test method of the present invention from the initial to 7mm prediction;
[0071] Fig.11 This is a prediction result diagram of the third group of data from the initial to 7mm prediction of a crankshaft accelerated fatigue test method of the present invention;
[0072] Fig.12This is a prediction result diagram of the first set of data from the initial to 8mm prediction of a crankshaft accelerated fatigue test method of the present invention;
[0073] Fig.13 This is a prediction result diagram of the second group of data of a crankshaft accelerated fatigue test method of the present invention from the initial to 8mm prediction;
[0074] Fig.14 This is a prediction result diagram of the third group of data from the initial to 8mm prediction of a crankshaft accelerated fatigue test method of the present invention;
[0075] Fig.15 A schematic diagram of double logarithmic coordinate fitting of load-fatigue life obtained by fitting crankshaft fatigue test data in a crankshaft accelerated fatigue test method of the present invention;
[0076] Fig.16 A schematic diagram of double logarithmic coordinate fitting of load-fatigue life obtained by fitting the first group of predicted crankshaft fatigue test data in a method for crankshaft accelerated fatigue test of the present invention;
[0077] Fig.17 This is a schematic diagram of double logarithmic coordinate fitting of load-fatigue life obtained by fitting the second group of crankshaft fatigue test data obtained by prediction in a method for accelerated crankshaft fatigue testing of the present invention. DETAILED DESCRIPTION
[0078] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0079] Secondly, the present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention in detail, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.
[0080] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.
[0081] Example 1
[0082] The invention provides a method for crankshaft accelerated fatigue test, which can shorten the crankshaft fatigue test time and achieve the acceleration effect.
[0083] See also Figure 1-Figure 2 The specific steps of the crankshaft accelerated fatigue test method are as follows:
[0084] S1. Establish an empirical model of the change of crankshaft crack depth with the number of cycles: y = a exp(bx) + c exp(dx),
[0085] The specific steps of establishing the empirical model of the change of crankshaft crack depth with the number of cycles: y = a exp(bx) + c exp(dx) are as follows:
[0086] S101, obtaining test data showing that the crack depth of a crankshaft increases with the number of cycles, specifically, the crankshaft is cycled with a strengthening load on a shaft fatigue test bench, and the number of cycles when the crankshaft fails is calculated. Preferably, in this step, the obtained test data is also processed, and the data on the crack depth of the crankshaft and the corresponding number of cycles of the crankshaft crack initiation in the previous part are discarded.
[0087] S102, performing nonlinear fitting on the obtained test data to obtain a nonlinear function;
[0088] S103, according to the principle of minimum fitting error of the law of crack depth of the crankshaft with the increase of cycle number, the results obtained by different fitting formulas are compared, and the double exponential empirical model is selected. Specifically, in step S103, according to the principle of minimum fitting error of the law of crack depth of the crankshaft with the increase of cycle number, the results obtained by different fitting formulas are compared as follows: Introducing the goodness of fit value R 2 The root mean square error (RMSE) was used to judge the quality of each model in fitting the test data. 2 The value range is [0,1], R 2 The closer it is to 1, the better the fitting effect is. 2 If the RMSE value is close to 0, the model matching is not ideal. The smaller the RMSE value is, the better the model fitting effect is. 2 The calculation formula of RMSE is as follows:
[0089]
[0090] Among them, y i represents the i-th original data point, is the average value of the original data, is the model calculated value, and n is the number of original data points.
[0091] S2. Optimize the established empirical model and obtain the recursive formula of the crankshaft remaining life prediction model: k =y k-1 ·exp(b)+c·exp[d(k-1)][1-exp(bd)]+u k ,u k ~N(0,σ w ),
[0092] Among them, y k is the crack depth of the crankshaft in the kth cycle under a certain bending moment; b, c, d are the parameters of the improved model of the remaining life of the crankshaft; u k is process noise, with a mean of 0 and a variance of σ w The standard normal distribution of
[0093] S3. Establish a state transfer equation and an observation equation for predicting the remaining life of the crankshaft, wherein the state transfer equation is a recursive formula of the crankshaft remaining life prediction model, and the observation equation is:
[0094]
[0095] in: is the crack depth of the crankshaft observed at the kth cycle under a certain bending moment; v k is the observation noise, with a mean of 0 and a variance of σ w The standard normal distribution of
[0096] S4. Based on the established state transfer equation and observation equation for predicting the remaining life of the crankshaft as empirical data, a particle filter algorithm is used to predict the fatigue life of the crankshaft; wherein, based on the established state transfer equation and observation equation for predicting the remaining life of the crankshaft as empirical data, the specific steps of using the particle filter algorithm to predict the fatigue life of the crankshaft are as follows:
[0097] S401, initial state
[0098] Set the initialization iteration number t=0, the number of particles N and the various parameter values involved in the algorithm, use a large number of particles to simulate x(t), the particles are evenly distributed in the space, and the initial weight values of all particles in the particle swarm are set to 1 / N.
[0099] S402: Prediction status
[0100] From the known process noise ω t N noise particles are obtained by sampling, and the predicted value of the particle position at the next iteration is calculated according to the state transfer equation
[0101]
[0102] In the formula, x k Represents the current state value of the system, y k represents the current observation value of the system, f(·) represents the state transfer function, which characterizes the functional relationship between the state at the previous moment and the current moment, h(·) represents the observation function, and u k represents the process noise of the system, v k represents the observation noise of the system;
[0103] S403, Importance Sampling
[0104] Set the number of iterations t = t + 1, and from the importance density function To randomly sample particle clusters in , use express;
[0105] S404, calibration phase
[0106] Calculate the weight of each particle and evaluate the weight of the predicted particles. Calculate the weight of all particles in the particle set
[0107]
[0108] S405, weight normalization
[0109] Normalize the weights,
[0110]
[0111] S406, resampling
[0112] Calculate the effective number of particles N eff , set the threshold value N th , if and only if N eff <N th When, yes Resample and set the weight to 1 / N;
[0113] S407, status output
[0114] Perform status output;
[0115]
[0116] S408: Is the iteration finished?
[0117] Determine whether the process is finished, if so, exit; otherwise, go to step S402.
[0118] The above crankshaft accelerated fatigue test method is scientifically verified in the following scientific demonstration.
[0119] Firstly, a type of crankshaft is selected and cycled under strengthening load on a crankshaft fatigue test bench. The number of cycles until failure is calculated. Then three sets of test data are obtained, showing that the crack depth of the crankshaft increases with the increase of the number of cycles. The data are shown in Table 1-1.
[0120] Table 1-1 Test data
[0121]
[0122]
[0123]
[0124]
[0125] The goodness of fit value R 2 The root mean square error (RMSE) is used to judge the quality of each model in fitting the test data. calculate
[0126] The calculation results of the two data obtained to evaluate the fitting effect and the fitting result evaluation are shown in Table 1-2:
[0127] Table 1-2 Fitting results
[0128]
[0129] By comparing the results obtained from different fitting formulas, the empirical model selected is the double exponential model, that is, y=aexp(bx)+c exp(dx).
[0130] Then the parameters b, c, and d of the state transfer equation and observation equation for predicting the remaining life of the crankshaft are fitted using the selected empirical model. When the crack depth from the beginning of the crankshaft cycle to the crack depth of the crankshaft growing to 5 mm is selected, the values of the parameters b, c, and d are shown in Table 1-3 as follows:
[0131] Table 1-3 State equation parameters
[0132]
[0133] When the crack depth from the beginning of the crankshaft cycle to the crankshaft crack depth growing to 6mm, the values of parameters b, c, and d are shown in Table 1-4:
[0134] Table 1-4 State equation parameters
[0135]
[0136] When the crack depth from the beginning of the crankshaft cycle to the crankshaft crack depth growing to 7mm, the values of parameters b, c, and d are shown in Table 2-5:
[0137] Table 1-5 State equation parameters
[0138]
[0139] When the crack depth from the beginning of the crankshaft cycle to the crankshaft crack depth growing to 8mm, the values of parameters b, c, and d are shown in Table 1-6:
[0140] Table 1-6 State equation parameters
[0141]
[0142] The results obtained by using the particle filter algorithm based on these parameters are as follows:
[0143] like Figure 3-Figure 5 As shown, the prediction results when the data when the crankshaft crack depth grows from the initial to 5mm are selected for the first group of data, the second group of data and the third group of data are predicted through particle filtering.
[0144] like Figure 6-8 As shown, the prediction results when the first group of data, the second group of data and the third group of data are selected for the data when the crankshaft crack depth grows from the initial to 6mm are predicted through particle filtering.
[0145] like Figure 9-11 As shown, the prediction results when the data when the crankshaft crack depth grows from the initial to 7 mm are selected for the first group of data, the second group of data and the third group of data are predicted through particle filtering.
[0146] like Figure 12-14 As shown, the prediction results when the data when the crankshaft crack depth grows from the initial to 7 mm are selected for the first group of data, the second group of data and the third group of data are predicted through particle filtering.
[0147] From the above prediction results, it can be seen that the prediction results of these three sets of data are in line with the rules. The data selected at the beginning of these three sets are from the initial value to the data when the crack reaches 5mm. Due to the small amount of data, the prediction results when the crankshaft fails have a large error compared with the actual number of cycles when the failure occurs. When the data from the initial value to the crack reaches 8mm is selected for prediction, the prediction results have a smaller error than the above three predictions. This shows that the prediction results are related to the amount of input data. Only the third set of data has a large error.
[0148] The error is due to the random resampling process of particles in particle filtering. Different random sampling results will lead to different prediction results. At the same time, as the number of particles increases, the accuracy will increase and the probability density distribution range will decrease. It is also partly because the crankshaft crack growth rate curve is divided into the near-threshold area of crack initiation and initial development, the intermediate area of steady-state development and the critical area of late rapid development. When fitting the data, it is not completely consistent with the actual double exponential curve model.
[0149] Experiments show that when less data is used, there is an error between the predicted number of failure cycles and the actual number of cycles. If the error is required to be smaller, more data must be used. At this time, the data used is not much different from the data of crankshaft failure, and the experimental cost and time saved are relatively small. The expected use of less raw data can accurately predict the data at the time of failure. Therefore, in this method, in order to further reduce the prediction error, the acquired test data can be processed and the data of the crankshaft crack depth and the corresponding number of cycles of the crankshaft crack initiation in the previous part can be discarded.
[0150] Example 2
[0151] When the crankshaft is subjected to bending fatigue test, the crankshaft fatigue load, fatigue life and other data will generally have greater randomness and dispersion during the experiment. When analyzing the experimental results, it is found that under the same loading conditions, the fatigue life obtained from multiple fatigue tests is not the same. Or even if the fatigue life of the crankshaft is the same, the bending load applied to the crankshaft during the test is different. Therefore, if you want to analyze the life of the crankshaft, you need to perform statistical analysis on the data obtained from the test.
[0152] Therefore, unlike the above-mentioned embodiment, the method for accelerated crankshaft fatigue testing further includes the following steps:
[0153] S5. Logarithmically process the fatigue life and stress obtained by predicting the fatigue life of the crankshaft, and fit the obtained fatigue life and stress by the least square method to obtain the logarithmic linear relationship between the stress and fatigue life of the part;
[0154] S6. Based on the linear relationship obtained by fitting, use the loading moment of one fatigue life as Q CI Point, calculate the fatigue limit load:
[0155]
[0156] Where, FS i is the estimated value of the parent fatigue limit measured by the i-th specimen, N i is the number of tests when the i-th specimen fails, S i is the load of the i-th specimen, S A Q is obtained by fitting using the least squares method. CI Stress value at point, N A The S obtained at the stress A The life value of the crankshaft under the horizontal action of;
[0157] S7. Estimate the fatigue limit of the parent body measured for the i-th specimen and calculate the median rank:
[0158]
[0159] in, i k is the average failure number, n is the sample size, and j represents the total failure number;
[0160] On this basis, the normal distribution model is used for analysis to obtain the fatigue limit load of the crankshaft with a 50% survival probability required for actual engineering.
[0161] First, different crankshafts were subjected to bending fatigue tests using the resonant crankshaft fatigue test system. The corresponding test results are shown in Table 4-1:
[0162] Table 4-1 Crankshaft bending fatigue test results
[0163]
[0164] The above method is used to fit the data, and the fitting results are as follows: Fig.15 shown.
[0165] Then the median rank estimation is performed and the results are shown in Table 4-2:
[0166] Table 4-2 Calculation of fatigue limit median rank
[0167]
[0168]
[0169] From the above results, we can see that the correlation coefficient of the fitting is 0.8716. Under this sample size, the fitting accuracy meets the linear correlation requirement at a confidence level of 99%. A =1000, the moment load value is 5046N·m.
[0170] When the improved particle filter is used to predict the accelerated fatigue life of the above crankshaft, since the remaining failure cycles are small, there is no need to predict the service life. 6 The above data is used for prediction, so predictions are made for No. 1, 2, 8, and 9. Since there is a certain error in life prediction, it can be seen from Example 1 that when the data from 2mm to 5mm is selected for prediction, the error range is within ±20%. Therefore, two groups of comparative tests are conducted. The first group is that all the predicted data are 20% larger than the actual life data, and the other values remain unchanged; the second group is that all the predicted data are 20% smaller than the actual life data, and the other values remain unchanged. The first group of data is shown in Table 4-3:
[0171] Table 4-3 The first set of prediction results
[0172]
[0173]
[0174] The above method is used to fit the data, and the fitting results are as follows: Fig.16 shown.
[0175] Then the median rank estimation is performed and the results are shown in Table 4-4:
[0176] Table 4-4 Calculation of fatigue limit median rank
[0177]
[0178] From the above results, we can see that the correlation coefficient of the fitting is 0.8741. Under this sample size, the fitting accuracy meets the linear correlation requirement at a confidence level of 99%. A = 1000, the moment load value is 5125N·m. The moment load obtained at this time is obtained using the predicted life, and the error with the actual load is 1.5%.
[0179] The second group is that the predicted data are all 20% smaller than the actual life data, and the other values remain unchanged. The data are shown in Table 4-5:
[0180] Table 4-5 The second set of prediction results
[0181]
[0182]
[0183] The above method is used to fit the data, and the fitting results are as follows: Fig.17 shown.
[0184] Then the median rank estimation is performed and the results are shown in Table 4-6:
[0185] Table 4-6 Calculation of fatigue limit median rank
[0186]
[0187] From the above results, we can see that the correlation coefficient of the fitting is 0.8601. Under this sample size, the fitting accuracy meets the linear correlation requirement at a confidence level of 99%. A= 1000, the moment load value is 4933N·m. The moment load obtained at this time is obtained using the predicted life, and the error with the actual load is 2.2%.
[0188] The above test shows that the remaining life data predicted by the above method using the data from 2mm to 5mm is very close to the data analysis results using the actual remaining life data, with an error within 5%, and the results are very accurate. Therefore, when using particle filtering to predict the remaining life of the crankshaft under accelerated fatigue, the predicted results are close to the actual results, with a small error. The predicted results can be used instead of the actual data for calculation and analysis. The use of particle filtering greatly saves test costs and time, and the result error is also relatively small.
[0189] Although the present invention has been described above with reference to the embodiments, various modifications may be made thereto and parts thereof may be replaced by equivalents without departing from the scope of the present invention. In particular, as long as there is no structural conflict, the various features in the embodiments disclosed in the present invention may be used in combination with each other in any manner, and the fact that these combinations are not exhaustively described in this specification is only for the sake of omitting space and saving resources. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.
Claims
1. A method for accelerated fatigue testing of a crankshaft, characterized in that: include: S1. Establish an empirical model of the change of crankshaft crack depth with the number of cycles: y = a exp(bx) + c exp(dx); S2. Optimize the established empirical model and obtain the recursive formula of the crankshaft remaining life prediction model: k =y k-1 ·exp(b)+c·exp[d(k-1)][1-exp(bd)]+u k ,u k ~N(0,σ w ), Among them, y k is the crack depth of the crankshaft in the kth cycle under a certain bending moment; b, c, d are the parameters of the improved model of the remaining life of the crankshaft; u k is process noise, with a mean of 0 and a variance of σ w The standard normal distribution of S3. Establish a state transfer equation and an observation equation for predicting the remaining life of the crankshaft, wherein the state transfer equation is a recursive formula of the crankshaft remaining life prediction model, and the observation equation is: in: is the crack depth of the crankshaft observed at the kth cycle under a certain bending moment; v k is the observation noise, with a mean of 0 and a variance of σ w The standard normal distribution of S4. Based on the established state transfer equation and observation equation for predicting the remaining life of the crankshaft as empirical data, a particle filter algorithm is used to predict the fatigue life of the crankshaft.
2. The method for crankshaft accelerated fatigue testing according to claim 1, characterized in that: The specific steps to establish the empirical model of crankshaft crack depth changing with cycle number: y = aexp(bx) + cexp(dx) are as follows: S101, obtaining test data showing that the crack depth of a crankshaft increases with the number of cycles; S102, performing nonlinear fitting on the obtained test data to obtain a nonlinear function; S103. Based on the principle of minimizing the fitting error of the change law of the crack depth of the crankshaft with the increase of the number of cycles, the results obtained by different fitting formulas are compared, and the double exponential empirical model is selected.
3. The method for crankshaft accelerated fatigue testing according to claim 2, characterized in that: In step S103, the results obtained by comparing different fitting formulas according to the principle of minimum fitting error according to the change law of the crack depth of the crankshaft with the increase of the number of cycles are as follows: 2 The root mean square error (RMSE) was used to judge the quality of each model in fitting the test data. 2 The value range is [0,1], R 2 The closer it is to 1, the better the fitting effect is. 2 If the RMSE value is close to 0, the model matching is not ideal. The smaller the RMSE value is, the better the model fitting effect is. 2 The calculation formula of RMSE is as follows: Among them, y i represents the i-th original data point, is the average value of the original data, is the model calculated value, and n is the number of original data points.
4. The method for crankshaft accelerated fatigue testing according to claim 1, characterized in that: In step S4, based on the established state transfer equation and observation equation for predicting the remaining life of the crankshaft as empirical data, the specific steps of using the particle filter algorithm to predict the fatigue life of the crankshaft are as follows: S401, initial state Set the initialization iteration number t=0, the number of particles N and various parameter values involved in the algorithm, use a large number of particles to simulate x(t), the particles are evenly distributed in the space, and the initial weight values of all particles in the particle swarm are set to 1 / N; S402, prediction status From the known process noise ω t N noise particles are obtained by sampling, and the predicted value of the particle position at the next iteration is calculated according to the state transfer equation In the formula, x k Represents the current state value of the system, y k represents the current observation value of the system, f(·) represents the state transfer function, which characterizes the functional relationship between the state at the previous moment and the current moment, h(·) represents the observation function, and u k represents the process noise of the system, v k represents the observation noise of the system; S403, Importance Sampling Set the number of iterations t = t + 1, and from the importance density function To randomly sample particle clusters in , use express; S404, calibration phase Calculate the weight value of each particle, evaluate the weight of the predicted particles, and calculate the weight values of all particles in the particle set S405, weight normalization Normalize the weights, S406, resampling Calculate the effective number of particles N eff , set the threshold value N th , if and only if N eff <N th When, yes Resample and set the weight to 1 / N; S407, status output Perform status output; S408: Is the iteration finished? Determine whether the process is finished, if so, exit; otherwise, go to step S302.
5. The method for crankshaft accelerated fatigue testing according to claim 3, characterized in that: After obtaining the test data that the crankshaft crack depth increases with the increase of the number of cycles, the step S101 also includes processing the obtained test data, and discarding the data of the crankshaft crack depth and the corresponding number of cycles in which the crankshaft crack initiated in the previous part.
6. The method for crankshaft accelerated fatigue testing according to claim 3, characterized in that: In step S101, the specific steps of obtaining the test data that the crack depth of the crankshaft increases with the increase of the number of cycles are as follows: the crankshaft fatigue test bench is loaded with a strengthening load to cycle the crankshaft, and the number of cycles when it fails is calculated.
7. The method for crankshaft accelerated fatigue testing according to claim 4, characterized in that: In step S304, the effective particle number N eff The calculation formula is as follows:
8. The method for crankshaft accelerated fatigue testing according to claim 1, characterized in that: The following steps are also included: S5. Logarithmically process the fatigue life and stress obtained by predicting the fatigue life of the crankshaft, and fit the obtained fatigue life and stress by the least square method to obtain the logarithmic linear relationship between the stress and fatigue life of the part; S6. Based on the linear relationship obtained by fitting, use one of the fatigue life loading moments as Q CI Point, calculate the fatigue limit load: Where, FS i is the estimated value of the parent fatigue limit measured by the i-th specimen, N i is the number of tests when the i-th specimen fails, S i is the load of the i-th specimen, S A Q is obtained by fitting using the least squares method. CI Stress value at point, N A is the S obtained at the stress A The life value of the crankshaft under the horizontal action of; S7. Sort the estimated values of the parent fatigue limit measured for the i-th specimen and calculate the median rank: in, i k is the average failure number, n is the sample size, and j represents the total failure number; On this basis, the normal distribution model is used for analysis to obtain the fatigue limit load of the crankshaft with a 50% survival probability required for actual engineering.
