A method for predicting the remaining life of heavy-duty truck hook based on assumed distribution
Through a method based on hypothesis distribution, combined with non-destructive testing and statistical analysis, the problem of remaining life prediction of the complex structure of the hook body of the heavy-duty railway truck is solved, and the accurate prediction of the crack parts of the hook body is achieved and the maintenance strategy is formulated, which improves the prediction accuracy and reliability.
Patent Information
- Application Number
- CN202111303450.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-05
- Publication Date
- 2024-12-10
- Estimated Expiration
- 2041-11-05
AI Technical Summary
When predicting the remaining life of a heavy-load railway truck hook body, it is difficult to accurately predict the damage location and crack location of the hook body of a complex structure, especially in the case of small data volumes, and the maintenance strategy for the hook body is very difficult. The formulation lacks precision.
Using a method based on hypothesis distribution, the crack data of the hook body is obtained through non-destructive detection, including service mileage, crack position and size, and combining 3PWD, 2PWD, ND, LND, EMVD1 or EMVD2 as the hypothesis distribution under the life fraction, the cumulative failure probability is constructed. The curve predicts the remaining life of the hook body, and performs abnormal data processing through linear interpolation and Grubbs criterion method.
It improves the accuracy of the remaining life prediction of hook body, provides a more accurate maintenance strategy, reduces resource waste, and ensures the stable operation of the structure at a high reliability level.
Smart Images
Figure CN114021347B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of rail transportation, and in particular to a method for predicting the remaining service life of a heavy-duty truck hook body based on assumed distribution. Background Art
[0002] At present, most of the heavy-duty railways use 16 / 17 type cast hooks, and their structure is as follows: Figure 4-5 As shown in the figure, according to the survey results, the hook body is mostly damaged at the root of the two contact parts of the hook body and the hook tongue, and the hook body is also an important component of the heavy-load hook body. The crack locations of the damaged hook bodies of the 10,000-ton and 20,000-ton mixed-running trains on the Shenhuang Line after a section repair period were counted. The results show that the cracks in the damaged hook bodies mainly appear at the root of the lower traction platform.
[0003] Accurately predicting the remaining life can not only ensure that the structure or system works at a higher reliability level, but also avoid the waste of resources caused by too frequent preventive maintenance. The currently widely used remaining life prediction methods can be divided into the following according to the basic theory adopted: 1. Remaining life prediction methods that build models based on the physical, chemical or empirical system mechanisms of the object being tested; 2. Data-driven methods that build models based entirely on historical feature data, such as statistical models, neural networks, Gaussian regression processes, support vector regression, fuzzy reasoning, etc.; 3. Methods that combine mechanism models with data-driven models; 4. Methods that combine multiple data-driven models.
[0004] The remaining life prediction method based on the mechanism model can have higher accuracy under the premise of clarifying the mechanism, but it is difficult to establish its mechanism model for some complex structures or systems. The remaining life prediction method based on data-driven does not need to clarify the true mechanism expression formula, so it is easier to apply to complex structures or systems, but it usually requires a large amount of data. Although there are a large number of processing methods for small data volumes, there are still certain risks.
[0005] Selectively establishing a hybrid model based on different theories can overcome the shortcomings of each theory and improve the prediction accuracy, which has become a research hotspot in remaining life prediction. Therefore, the present invention provides a method for predicting the remaining life of a heavy-duty railway freight car hook body based on an assumed distribution. Summary of the invention
[0006] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for predicting the remaining life of a heavy-duty truck hook body based on assumed distribution. The hook body crack data is obtained through non-destructive testing. When there is a visible fatigue crack on the hook body, the hook body is in the expansion stage after the crack initiation. The service mileage, crack position, crack size, reliability and other parameters of the hook body are statistically analyzed at this time. The remaining life prediction method based on assumed distribution takes into account the common characteristics and current status of the research object, effectively improves the prediction accuracy of the remaining life of the hook body, and is of great significance to the formulation of the hook body maintenance strategy.
[0007] The objective of the present invention is achieved through the following technical solutions:
[0008] A method for predicting the remaining life of a heavy-duty truck hook body based on an assumed distribution is used to predict the remaining life of a hook body with a visible crack at the root of a traction flange under the hook body. The method specifically comprises: obtaining crack data of the hook body through nondestructive testing, wherein the crack data includes the full life of the hook body, and the corresponding service mileage, crack size and reliability when there is a visible crack at the root of the traction flange under the hook body; then calculating the surface length of the crack at the root of the traction flange under the hook body at different life fractions, wherein the life fraction refers to the ratio of the current service life to the full life, using 3PWD, 2PWD, ND, LND, EMVD1 or EMVD2 as the assumed distribution of the crack length at a certain life fraction, performing a goodness of fit test on the crack data, determining a good assumed distribution type, and obtaining a cumulative failure probability curve of the crack surface length at different life fractions; based on the cumulative failure probability curve, predicting the remaining life of the target hook body with a visible crack.
[0009] Furthermore, based on the cumulative failure probability curve, when predicting the remaining life of a target hook body with visible cracks, a certain reliability is selected according to the service mileage of the target hook body and the surface length of the crack at the root of the traction flange under the hook body, and the evaluation point of the target hook body in the cumulative failure probability curve can be obtained. According to the life score corresponding to the evaluation point, the remaining life of the target hook body under the reliability is calculated.
[0010] Furthermore, the surface length of the crack at the root of the traction flange under the hook body at different life fractions was calculated by linear interpolation.
[0011] Furthermore, after calculating the surface length of the crack at the root of the traction flange under the hook body at different life fractions by linear interpolation, the Grubbs criterion method was used to judge and eliminate abnormal data.
[0012] Furthermore, EMVD1 is used as the assumed distribution of crack length under a certain life fraction, and the reliability curves of crack surface length under different life fractions are obtained.
[0013] Furthermore, the cumulative failure probability=(1-reliability)×100%.
[0014] Furthermore, the reliability life index is calculated through the data obtained from the crack propagation at the root of the traction flange under the hook body, and the statistical distribution of its life is obtained; if the probability density function of the structural service life t is f(t), the reliability R(t) is defined as the probability that the service life is greater than a certain specified value t, that is,
[0015]
[0016] The cumulative failure probability F(t), also known as unreliability, can be calculated using the following formula:
[0017] F(t)=1-R(t)
[0018] The key to calculating structural reliability is to obtain the structural failure probability density function f(t). For large sample problems (capacity exceeds 50 or 70), the probability density function can be directly determined based on the data. For small sample problems, the overall fitting effect, fatigue failure mechanism and algorithm security can be comprehensively considered, and a good hypothesis distribution can be selected from the commonly used statistical distribution types, namely three-parameter Weibull distribution (3PWD), two-parameter Weibull distribution (2PWD), normal distribution (ND), lognormal distribution (LND), minimum value distribution (EMVD1), maximum value distribution (EMVD2) and exponential distribution (ED). The basic process is as follows:
[0019] a). Sort the data from small to large and calculate the empirical failure probability P based on the median rank em (x i );
[0020]
[0021] b). Based on linear regression, the above common statistical distribution types are used to fit (x i , P em (x i )) data, and obtain the statistical distribution parameters and fitting correlation coefficient R XY ;
[0022] c) For high reliability problems, by calculating d F1 and d F2 To judge the safety of tail prediction.
[0023] For the left tail problem, we have
[0024] d F1 =P em (x1)-P th (x1)
[0025] d F2 =P em (x2)-Pth (x2)
[0026] For the right tail problem, we have
[0027] d F1 =P em (x n )-P th (x n )
[0028] d F2 =P em (x n-1 )-P th (x n-1 )
[0029] The fitting correlation coefficient, fatigue failure mechanism and tail prediction safety are comprehensively considered to determine the good statistical distribution type.
[0030] The beneficial effects of the present invention are: a method for predicting the remaining life of a heavy-duty truck hook body based on assumed distribution is provided in the present invention. The hook body crack data is obtained through non-destructive testing. When there is a visible fatigue crack on the hook body, the hook body is in the expansion stage after the crack initiation. The service mileage, crack position, crack size, reliability and other parameters of the hook body are statistically analyzed at this time to construct a remaining life prediction model. A remaining life prediction method based on assumed distribution is proposed. At the same time, the common characteristics and current status of the research object are taken into account, which effectively improves the prediction accuracy of the remaining life of the hook body, which is of great significance to the formulation of the hook body maintenance strategy. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 The graph showing the mean change of crack length at the root of the traction flange under the hook body at different life fractions;
[0032] Figure 2 It is the coefficient of variation diagram under different lifespan scores;
[0033] Figure 3 It is the EMVD1 cumulative failure probability curve of crack size at different life fractions;
[0034] Figure 4 It is a schematic diagram of the structure of the 16 and 17 type casting hook bodies;
[0035] Figure 5 It is a schematic diagram of the structure of the 17-type casting hook body; DETAILED DESCRIPTION
[0036] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited to the following.
[0037] Example
[0038] A method for predicting the remaining life of a heavy-duty truck hook body based on an assumed distribution is used to predict the remaining life of a hook body with a visible crack at the root of a traction flange under the hook body. Specifically, based on a fatigue bench test, crack data of the hook body is obtained through nondestructive testing, wherein the crack data includes the full life of the hook body, and the corresponding service mileage, crack size and reliability when there is a visible crack at the root of the traction flange under the hook body; then, the surface length of the crack at the root of the traction flange under the hook body at different life fractions is calculated, wherein the life fraction refers to the ratio of the current service life to the full life, and 3PWD, 2PWD, ND, LND, EMVD1 or EMVD2 is used as the assumed distribution of the crack length at a certain life fraction, and a goodness of fit test is performed on the crack data to determine a good assumed distribution type, and a cumulative failure probability curve of the crack surface length at different life fractions is obtained; based on the cumulative failure probability curve, the remaining life of the target hook body with a visible crack is predicted.
[0039] Specifically, based on the cumulative failure probability curve, when predicting the remaining life of a target hook body with visible cracks, a certain reliability is selected according to the service mileage of the target hook body and the surface length of the crack at the root of the traction flange under the hook body, and the evaluation point of the target hook body in the cumulative failure probability curve can be obtained. According to the life score corresponding to the evaluation point, the remaining life of the target hook body under the reliability is calculated.
[0040] Specifically, the crack surface length at the root of the traction flange under the hook body at different life fractions was calculated by linear interpolation.
[0041] Specifically, after calculating the surface length of the crack at the root of the traction flange under the hook body at different life fractions by linear interpolation, the Grubbs criterion method was used to judge and eliminate abnormal data.
[0042] Specifically, EMVD1 is used as the assumed distribution of crack length under a certain life fraction, and the reliability curves of crack surface length under different life fractions are obtained.
[0043] Specifically, the cumulative failure probability=(1-reliability)×100%.
[0044] Specifically, for reliability, the reliability life index is calculated through the data obtained from the crack extension of the hook tongue, and the statistical distribution of its life is obtained. If the probability density function of the service life t of the structure is f(t), the reliability R(t) is defined as the probability that the service life is greater than a certain specified value t, that is,
[0045]
[0046] The cumulative failure probability F(t), also known as unreliability, can be calculated using the following formula:
[0047] F(t)=1-R(t)
[0048] The key to calculating structural reliability is to obtain the structural failure probability density function f(t). For large sample problems (capacity exceeds 50 or 70), the probability density function is generally determined directly based on the data. For small sample problems, the overall fitting effect, fatigue failure mechanism and algorithm security can be comprehensively considered, and a good hypothesis distribution can be selected from the commonly used statistical distribution types, namely three-parameter Weibull distribution (3PWD), two-parameter Weibull distribution (2PWD), normal distribution (ND), lognormal distribution (LND), minimum value distribution (EMVD1), maximum value distribution (EMVD2) and exponential distribution (ED). The basic process is as follows:
[0049] a). Sort the data from small to large and calculate the empirical failure probability P based on the median rank em (x i );
[0050]
[0051] b). Based on linear regression, the above common statistical distribution types are used to fit (x i , P em (x i )) data, and obtain the statistical distribution parameters and fitting correlation coefficient R XY ;
[0052] c) For high reliability problems, by calculating d F1 and d F2 To judge the safety of tail prediction.
[0053] For the left tail problem, we have
[0054] d F1 =P em (x1)-P th (x1)
[0055] d F2 =P em (x2)-P th (x2)
[0056] For the right tail problem, we have
[0057] d F1 =P em (x n )-P th (x n )
[0058] d F2 =P em (x n-1 )-P th (x n-1 )
[0059] The fitting correlation coefficient, fatigue failure mechanism and tail prediction safety are comprehensively considered to determine the good statistical distribution type.
[0060] Test example
[0061] The crack data of the hook body is obtained by nondestructive testing. The crack data includes the full life of the hook body, and the corresponding service mileage, crack size and reliability when there is a visible crack at the root of the traction flange under the hook body; then the surface length of the crack at the root of the traction flange under the hook body is calculated under different life scores. The life score refers to the ratio of the current service life to the full life, and the Grubbs criterion method is used to judge and eliminate abnormal data. The results are shown in Table 1.
[0062] Table 1 Evolution data of crack length of traction flange under hook body with life fraction
[0063]
[0064] The variation law of the mean value and coefficient of variation of crack length is as follows: Figure 1 and Figure 2 As shown in the figure, it can be seen that the mean value of the crack length of the traction flange under the hook body increases with the increase of the life fraction, while the coefficient of variation gradually decreases and tends to be stable; when the life fraction is greater than 0.7, the dispersion is small. The reason is that in the early stage of crack propagation, multiple small cracks are generated at the stress concentration part of the root of the traction flange, and the randomness of the position and size of the small cracks causes a large dispersion; under the action of cyclic load, after the small cracks gradually merge into a fatigue main crack, the crack position is relatively fixed, and the data dispersion is small at this time.
[0065] 3PWD, 2PWD, ND, LND, EMVD1, and EMVD2 are used as the hypothetical distribution of crack length under a certain life fraction, and the linear correlation coefficient R under each distribution function is compared. XY , so as to judge whether the assumed distribution is reasonable; the fitting correlation coefficients under each assumed distribution are shown in Figure 2. Since the crack length data under life fractions of 0.1, 0.2, and 0.3 are all 0, Table 2 only gives the fitting correlation coefficients of the crack length data under other life fractions.
[0066] Table 2 Fitting correlation coefficients of crack length data under various assumed distributions
[0067]
[0068] The marked data in Table 2 are the absolute values of the fitting correlation coefficients at each lifespan fraction | R XY |Maximum value, from |R XY| In terms of the degree of closeness to 1, for the 4th group of data, EMVD2 has the best fitting effect, for the 5th and 9th groups of data, ND is the best, for the 6th and 8th groups of data, EMVD1 has the highest fitting degree, and for the 7th and 10th groups of data, 3PWD has the highest fitting degree. From the perspective of the safety of tail prediction, Table 3 shows the difference between the empirical value and the predicted value of the right tail failure probability of the above six hypothetical distributions (i.e., d F1 and d F2 ).
[0069] Table 3 Right tail fitting errors of six hypothesized distributions for crack length
[0070]
[0071] d F1 <d F2 This indicates that as reliability increases, there is a tendency to be conservative. F1 <0, then it is certain that when x>x n According to Table 3, when d F1 <d F2 For the data, 3PWD has 2 groups, 2PWD has 3 groups, ND has 3 groups, LND has 1 group, EMVD1 has 6 groups, and EMVD2 has 2 groups. F1 <0, 3PWD has 1 group, 2PWD has 3 groups, ND has 3 groups, LND has 1 group, EMVD1 has 6 groups, and EMVD2 has 2 groups. Considering the degree of fit and the safety of tail prediction, EMVD1, which takes into account both the degree of fit and the safety of tail prediction, can be selected as a good hypothesis distribution of the crack length data at the root of the traction flange under the hook body; Table 4 gives the EMVD1 cumulative failure probability function parameters of the surface crack length data under different life fractions, and the cumulative failure probability curves of the crack surface length under different life fractions are obtained, such as Figure 3 shown.
[0072] Table 4 Parameters of EMVD1 cumulative failure probability function for crack length data
[0073]
[0074] based on Figure 3, given the surface length and reliability (or cumulative failure probability) of the crack at the root of the lower traction flange of the hook body, the remaining life of the hook body can be predicted; the specific method is as follows: assuming that a certain hook body has been in service for k kilometers, and the flaw detection shows that the crack at the root of the lower traction flange is 30mm long (corresponding to the horizontal axis in the figure), when the cumulative failure probability is 0.05 (corresponding to the vertical axis in the figure), an evaluation point can be made in the figure, and this point falls in the middle of the curve with a life fraction of 0.6 to 0.7. Through linear interpolation, it can be obtained that the service life is about 67% of the full life, so the full life should be 1.49k, that is, the remaining mileage of this hook body is 0.49k kilometers, and this conclusion has a reliability of 95%.
[0075] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.
Claims
1. A method for predicting the remaining life of a heavy-duty truck hook body based on an assumed distribution, characterized in that: The remaining life of the hook body with visible cracks at the root of the traction flange under the hook body is predicted, specifically including: obtaining the crack data of the hook body through non-destructive testing, the crack data including the full life of the hook body, and the corresponding service mileage, crack size and reliability when there is a visible crack at the root of the traction flange under the hook body; then calculating the surface length of the crack at the root of the traction flange under the hook body at different life fractions, the life fraction refers to the ratio of the current service life to the full life, using 3PWD, 2PWD, ND, LND, EMVD1 or EMVD2 as the assumed distribution of the crack length at a certain life fraction, and performing a goodness of fit test on the crack data to determine a good assumed distribution type, and obtaining a cumulative failure probability curve of the crack surface length at different life fractions; Based on the cumulative failure probability curve, the remaining life of the target hook with visible cracks is predicted; Based on the cumulative failure probability curve, when predicting the remaining life of a target hook body with visible cracks, a certain reliability is selected according to the service mileage of the target hook body and the surface length of the crack at the root of the traction flange under the hook body, and the evaluation point of the target hook body in the cumulative failure probability curve can be obtained. According to the life score corresponding to the evaluation point, the remaining life of the target hook body under the reliability is calculated; The surface length of the crack at the root of the traction flange under the hook body at different life fractions is calculated by linear interpolation; Among them, for reliability, the reliability life index is calculated through the data obtained from the crack extension of the hook tongue, and its statistical distribution is obtained. If the probability density function of the service life of the structure is , reliability It is defined as the probability that the service life is greater than a specified value t, that is: The cumulative failure probability F(t), also known as unreliability, can be calculated using the following formula: The key to calculating structural reliability is to obtain the structural failure probability density function For capacity exceeding 50 or 70, the probability density function is determined according to the data. Otherwise, the overall fitting effect, fatigue failure mechanism and algorithm security are comprehensively considered, and a good hypothesis distribution is selected from the commonly used statistical distribution types, namely, three-parameter Weibull distribution (3PWD), two-parameter Weibull distribution (2PWD), normal distribution (ND), lognormal distribution (LND), minimum value distribution (EMVD1), maximum value distribution (EMVD2) and exponential distribution (ED). The basic process is as follows: a). Sort the data from small to large and calculate the empirical failure probability based on the median rank ; b). Based on linear regression, fit the data using the above common statistical distribution types. , get the statistical distribution parameters and fitting correlation coefficients ; c) For high reliability problems, by calculating and To judge the safety of tail prediction; For the left tail problem, we have For the right tail problem, we have Comprehensively consider the fitting correlation coefficient, fatigue failure mechanism and tail prediction safety to determine the good statistical distribution type; Among them, given the surface length and reliability of the crack at the root of the lower traction flange of the hook body, the remaining life of the hook body can be predicted as follows: if a hook body has been in service for k kilometers, and the flaw detection shows that the crack at the root of the lower traction flange is 30 mm long, when the cumulative failure probability is 0.05, an evaluation point can be made in the figure. This point falls in the middle of the curve with a life fraction of 0.6~0.
7. Through linear interpolation, it can be obtained that the service life is about 67% of the full life, so the full life should be 1.49k, that is, the remaining mileage of this hook body is 0.49k kilometers.
2. The method for predicting the remaining life of a heavy-duty truck hook body based on an assumed distribution according to claim 1, characterized in that: After calculating the surface length of the crack at the root of the traction flange under the hook body at different life fractions by linear interpolation, the Grubbs criterion method was used to judge and eliminate abnormal data.
3. The method for predicting the remaining service life of a heavy-duty truck hook body based on an assumed distribution according to claim 1, characterized in that: Based on the degree of fitting and the safety of tail prediction, EMVD1 was selected as a good assumed distribution of the crack length data at the root of the traction flange under the hook body, and the reliability curves of the crack surface length under different life fractions were obtained.
4. The method for predicting the remaining service life of a heavy-duty truck hook body based on an assumed distribution according to claim 3 is characterized in that: The cumulative failure probability = (1-reliability) × 100%.