Method and system for predicting remaining life of wheel set tread

Abstract

The invention provides a method and system for predicting the remaining life of a wheel set tread, and relates to the technical field of rail transit fault prediction, in particular to the method for predicting the remaining life of the wheel set tread, and the method comprises the steps: obtaining a current health assessment result of a to-be-detected wheel set tread, the health assessment result is at least divided into a first state used for representing normal operation of the tread of the to-be-tested wheel set and a second state used for representing that the tread of the to-be-tested wheel set has a tread out-of-round fault feature; according to the health assessment result, a residual life prediction model of the tread of the wheel set to be tested is determined, and the residual life prediction model is one of a first life prediction model corresponding to the first state and a second life prediction model corresponding to the second state; and utilizing the residual life prediction model to perform residual life prediction on the to-be-tested wheel set tread to obtain a residual life prediction result. According to the invention, the reliability of wheel set tread life prediction can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of rail transit fault prediction, and in particular to a method and system for predicting the remaining life of a wheelset tread. Background Art

[0002] At present, the turning and repair of components in the field of rail transit mostly adopt the planned turning and repair method. Although this traditional model has guaranteed the safe operation of equipment to a certain extent, it also has many disadvantages, such as the frequent occurrence of "over-repair", which leads to excessive equipment maintenance, waste of resources and a sharp increase in maintenance costs. With the continuous development of PHM (Prognostics and Health Management) technology, various industries have an increasingly urgent demand for preventive maintenance. They hope to extend the service life of components and optimize maintenance strategies through precise health management, thereby achieving cost reduction and efficiency improvement. As one of the core technologies of PHM, the remaining life prediction technology based on real-time monitoring data has developed rapidly in recent years, driven by the widespread application of intelligent equipment monitoring systems, providing strong support for the realization of precise preventive maintenance.

[0003] The wheelset tread is one of the key components of the running gear of urban rail vehicles. Its performance and status are directly related to the vehicle's operating safety, stability and overall operational efficiency. However, the current remaining life prediction methods for wheelset treads mostly rely on empirical thresholds to divide the degradation stages of components and select corresponding degradation models accordingly. Since the setting of empirical thresholds is mostly based on subjective judgment and is easily affected by personal experience, it will lead to deviations in the division of degradation stages, which does not conform to the actual situation of the application data, reduces the reliability of life prediction, and makes it difficult to meet the precise maintenance needs of wheelset treads.

[0004] Therefore, how to improve the reliability of wheel set tread life prediction is a technical problem that needs to be solved urgently by technical personnel in this field. Summary of the invention

[0005] To solve the above technical problems, the present application provides a method for predicting the remaining life of a wheelset tread, which can improve the reliability of the prediction of the life of the wheelset tread. The present application also provides a system for predicting the remaining life of a wheelset tread, which has the same technical effect.

[0006] The first objective of the present application is to provide a method for predicting the remaining life of a wheelset tread.

[0007] The above-mentioned application objective 1 of the present application is achieved through the following technical solutions: A method for predicting the remaining life of a wheelset tread, comprising: Obtaining a current health assessment result of the wheelset tread to be tested, wherein the health assessment result is divided into at least a first state for characterizing that the wheelset tread to be tested is operating normally, and a second state for characterizing that the wheelset tread to be tested has a tread out-of-round fault characteristic; Determining a remaining life prediction model of the wheelset tread to be tested according to the health assessment result, wherein the remaining life prediction model is one of a first life prediction model corresponding to the first state and a second life prediction model corresponding to the second state; The remaining life prediction model is used to predict the remaining life of the tread of the wheelset to be tested, and a remaining life prediction result is obtained.

[0008] Preferably, in the above-mentioned remaining life prediction method of the wheelset tread, the health assessment results are divided into at least a normal state, a sub-healthy state and a faulty state, wherein the normal state is taken as the first state, and the sub-healthy state and the faulty state are taken as the second state.

[0009] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, the remaining life prediction model is used to predict the remaining life of the wheelset tread to be tested to obtain a remaining life prediction result, including: Acquire real-time monitoring data and operating mileage of the wheelset tread to be tested, wherein the real-time monitoring data includes measured data of out-of-round runout and measured data of vibration dB value; The remaining life prediction model is used to perform remaining life prediction based on the real-time monitoring data and the operated mileage to obtain a remaining life prediction result.

[0010] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, the remaining life prediction is performed using the remaining life prediction model according to the real-time monitoring data and the operating mileage to obtain the remaining life prediction result, including: The out-of-round runout measured data, the vibration dB value measured data and the operated mileage are respectively preprocessed to obtain out-of-round runout data, vibration dB value data and operated mileage data; The out-of-round runout data, the vibration dB value data and the operated mileage data are respectively subjected to feature extraction to obtain out-of-round runout features, vibration effective value features and operated mileage features; Performing feature fusion on the out-of-round runout feature and the vibration effective value feature to obtain a fused degradation feature; The remaining life prediction model is used to perform remaining life prediction according to the fused degradation characteristics and the operated mileage characteristics to obtain a remaining life prediction result.

[0011] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, the feature extraction of the out-of-round runout data to obtain out-of-round runout features includes: According to the out-of-round runout data, the adaptive first-order density weight-quantile is calculated to obtain the out-of-round runout feature. The specific calculation formula is as follows: ; ; In the formula, Indicates feature extraction interval, , is a positive integer, Indicates The data density within the feature extraction interval is Represents the maximum value of data density in all feature extraction intervals, Indicates The first-order weighted density of feature extraction intervals, , Indicates The measured values ​​of feature extraction intervals, is a positive integer, Express Pick Percentile, Represents the out-of-round runout characteristics.

[0012] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, the feature extraction of the vibration dB value data to obtain the vibration effective value feature includes: According to the vibration dB value data, the root mean square value within the preset period is calculated to obtain the vibration effective value characteristics. The specific calculation formula is as follows: ; In the formula, Indicates the first The vibration dB value, , is the number of vibration dB values ​​corresponding to a preset period, represents the average value of the vibration dB value data, Represents the vibration effective value characteristics.

[0013] Preferably, in the above-mentioned method for predicting the remaining life of the wheelset tread, the feature fusion of the out-of-round runout feature and the vibration effective value feature to obtain the fused degradation feature includes: Normalizing the out-of-round runout amount feature and the vibration effective value feature to obtain a normalized out-of-round runout amount feature and a normalized vibration effective value feature; The fuzzy weighted method is used to fuse the normalized out-of-round runout feature and the normalized vibration effective value feature to obtain the fused degradation feature. The specific calculation formula is as follows: ; In the formula, represents the normalized out-of-round runout characteristic The corresponding fuzzy weight is Represents the normalized vibration effective value characteristic The corresponding fuzzy weight is Represents the fusion degradation feature.

[0014] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, the first life prediction model adopts a double exponential model, and the second life prediction model adopts a single exponential model.

[0015] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, the remaining life prediction is performed using the remaining life prediction model according to the fused degradation feature and the operated mileage feature to obtain the remaining life prediction result, including: Obtaining a predicted failure threshold of the wheel tread for turning and repairing, wherein the predicted failure threshold of the wheel tread for turning and repairing is calculated based on a plurality of historical case data of the wheel tread, each of the historical case data comprising radial runout value data, out-of-round runout value historical data, and vibration dB value historical data; The remaining life prediction model is used to perform remaining life prediction according to the turning and repair prediction failure threshold, the fused degradation characteristics and the operating mileage characteristics to obtain a remaining life prediction result.

[0016] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, the step of obtaining a predicted failure threshold for turning and repairing of the wheelset tread comprises: Acquire a plurality of historical case data of wheelset treads; According to the plurality of historical case data, a fused degradation characteristic value sequence, a radial runout value sequence, and a data sequence number sequence corresponding to the radial runout value sequence and the fused degradation characteristic value sequence are calculated; Sorting the fused degenerate eigenvalue sequence in ascending order to obtain a sorted first sequence, and sorting the data sequence number sequence according to the first sequence to obtain a sorted second sequence; Segmenting the first sequence according to a preset interval length to obtain a plurality of segmented sequences; According to the segmented sequences, respectively counting the data sequence number of the second sequence corresponding to each of the segmented sequences; According to the data sequence number of the second sequence corresponding to each segmented sequence, the number of radial jump values ​​of the corresponding radial jump value sequence is counted, so as to obtain a radial jump value number sequence corresponding to the segmented sequence; Draw a frequency histogram of the radial runout value number sequence according to the radial runout value number sequence, the maximum value and the minimum value of the first sequence, and the number of segmented sequences; According to the frequency histogram, data fitting is performed using statistical distribution to obtain a fitting result; According to the fitting result and the preset confidence interval, the coverage interval of the fused degradation characteristic value is determined, and the mean value of the fused degradation characteristic value within the coverage interval of the fused degradation characteristic value is calculated as the turning and repair prediction failure threshold.

[0017] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, when the remaining life prediction model is a first life prediction model corresponding to the first state, the remaining life prediction is performed using the remaining life prediction model according to the turning and repair prediction failure threshold, the fused degradation feature and the operating mileage feature to obtain a remaining life prediction result, including: Acquire multiple historical case fused degradation features of the wheelset tread, wherein each of the historical case fused degradation features is extracted based on a corresponding copy of the historical case data; Calculating the similarity between the fused degradation feature and each of the historical case fused degradation features respectively, and selecting the historical case fused degradation feature with the greatest similarity as the first historical case fused degradation feature; Acquire a pre-built double exponential model corresponding to the first historical case fused with the degradation feature as the first life prediction model, wherein the double exponential model is fitted based on the first historical case fused with the degradation feature; The first life prediction model is used to perform a remaining life prediction based on the operating mileage characteristics and the turning and repair prediction failure threshold to obtain a point estimate and an interval estimate of the remaining service life of the wheelset tread to be tested.

[0018] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, when the remaining life prediction model is a second life prediction model corresponding to the second state, the remaining life prediction is performed using the remaining life prediction model according to the turning and repair prediction failure threshold, the fused degradation feature and the operating mileage feature to obtain a remaining life prediction result, including: Obtaining a pre-built single exponential model, and fitting the single exponential model using the fused degradation feature and the operated mileage feature to obtain the second life prediction model; The second life prediction model is used to perform a remaining life prediction based on the operating mileage characteristics and the turning and repair prediction failure threshold to obtain a point estimate and an interval estimate of the remaining service life of the wheelset tread to be tested.

[0019] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, the use of the fused degradation feature and the operated mileage feature to fit the single exponential model to obtain the second life prediction model includes: The single exponential model is fitted by using the fused degradation feature and the operated mileage feature and adopting the least squares method to obtain the second life prediction model.

[0020] Preferably, in the above-mentioned method for predicting the remaining service life of a wheelset tread, the use of the second life prediction model to predict the remaining service life according to the operating mileage characteristics and the turning and repair prediction failure threshold to obtain a point estimate and an interval estimate of the remaining service life of the wheelset tread to be tested includes: Calculating the predicted operating mileage corresponding to the turning and repair prediction failure threshold using the second life prediction model; According to the predicted operating mileage and the operated mileage characteristics, a point estimate of the remaining service life of the wheelset tread to be measured is calculated, and using the Monte Carlo method, an interval estimate of the remaining service life of the wheelset tread to be measured is calculated.

[0021] The second objective of the present application is to provide a remaining life prediction system for wheelset treads.

[0022] The second application objective of the present application is achieved through the following technical solutions: A remaining life prediction system for a wheelset tread, comprising: an acquisition unit, configured to acquire a current health assessment result of the wheelset tread to be tested, wherein the health assessment result is at least divided into a first state for characterizing that the wheelset tread to be tested is operating normally, and a second state for characterizing that the wheelset tread to be tested has a tread out-of-round fault characteristic; a determination unit, configured to determine a remaining life prediction model of the tread of the wheelset to be tested according to the health assessment result, wherein the remaining life prediction model is one of a first life prediction model corresponding to the first state and a second life prediction model corresponding to the second state; The prediction unit is used to use the remaining life prediction model to predict the remaining life of the tread of the wheel pair to be tested, and obtain a remaining life prediction result.

[0023] The above technical scheme selects the optimal segmentation point of the segmentation model from the perspective of component failure development. Specifically, based on the current health assessment result of the wheelset tread to be tested, the remaining life prediction model of the wheelset tread to be tested is determined. The health assessment result is at least divided into a first state for characterizing the normal operation of the wheelset tread to be tested, and a second state for characterizing that the wheelset tread to be tested has a tread out-of-round fault characteristic. The remaining life prediction model is one of a first life prediction model corresponding to the first state and a second life prediction model corresponding to the second state; wherein the second state of the health assessment result is used as the segmentation point of the model. When the current health assessment result of the wheelset tread to be tested is the second state, the wheelset tread to be tested has a tread out-of-round fault characteristic, and the degradation characteristic shows a stable trend with a large amount of data; compared with the existing scheme that relies on empirical thresholds to divide the degradation stages of components and selects the corresponding degradation model accordingly, the above technical scheme effectively avoids the problem of deviation in the degradation stage division caused by subjective judgment and not conforming to the actual situation of the application data, and can improve the reliability of the wheelset tread life prediction.

[0024] In addition, the above technical scheme starts from the failure mode and the failure mechanism of the component monitoring index data, analyzes the physical meaning of the index and the physical characteristics of the fault, selects vibration dB value and out-of-round runout as monitoring data, and predicts the remaining life. Taking into account the number of feature dimensions and the complexity and efficiency of engineering applications, a fused degradation feature extraction method based on vibration dB value and out-of-round runout is proposed, which can reflect the overall degradation characteristics of the component through multi-dimensional fusion, effectively improving the accuracy and efficiency of wheelset tread life prediction.

[0025] In addition, the above technical scheme establishes a segmented life prediction model based on the data characteristics of different stages of the fault. The first stage adopts the set theoretical double exponential model degradation model, and the second stage adopts the single exponential dynamic prediction model, which effectively solves the problem that the parameter distribution fitted by relying on a small number of case samples in the existing life prediction method deviates greatly from the parameter distribution of actual application data; in addition, based on the accumulated component failure cases, the degradation characteristics of the case throughout the life cycle are calculated, and the turning and repair point / replacement point is taken as the end point of a cycle. The distribution is fitted based on all statistical cases, and the value when the confidence level is the set value is taken as the turning and repair prediction failure threshold. It can integrate the accumulated component failure case data information and formulate the optimal turning and repair point / replacement point, so that the remaining life prediction result is more in line with the actual application situation, and effectively solves the problem that the failure threshold is set by relying on subjective experience in the existing life prediction method, which deviates greatly from the actual situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 A schematic diagram of a flow chart of a method for predicting the remaining life of a wheelset tread in an embodiment of the present application; Figure 2 A frequency histogram provided in an embodiment of the present application; Figure 3 It is a multi-peak distribution fitting distribution diagram provided in the examples of the present application; Figure 4 It is a structural schematic diagram of a remaining life prediction system for a wheelset tread in an embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0029] In the embodiments provided in the present application, it should be understood that the disclosed methods and systems can be implemented in other ways. The system embodiments described below are merely schematic. For example, the division of units and modules is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or modules can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0030] In addition, all functional units in the embodiments of the present application may be integrated into one processor, or each unit may be a separate device, or two or more units may be integrated into one device; each functional unit in the embodiments of the present application may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0031] A person skilled in the art can understand that all or part of the steps of the following method embodiments can be completed by program instructions and related hardware. The aforementioned program instructions can be stored in a computer-readable storage medium. When the program instructions are executed, the steps of the following method embodiments are executed; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a magnetic disk or an optical disk, and other media that can store program codes.

[0032] It should be understood that the use of "system", "device", "unit" and / or "module" in this application is only a method for distinguishing different components, elements, parts, parts or assemblies at different levels. However, if other words can achieve the same purpose, the word can be replaced by other expressions.

[0033] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "multiple" and "several" mean two or more, unless otherwise clearly and specifically defined.

[0034] If a flow chart is used in the present application, the flow chart is used to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, each step can be processed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or a certain step or several steps of operations can be removed from these processes.

[0035] It should also be noted that, in this article, terms such as "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that an article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such articles or devices. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the article or device including the above elements.

[0036] The embodiments of the present application are written in a progressive manner.

[0037] like Figure 1 As shown, an embodiment of the present application provides a method for predicting the remaining life of a wheelset tread, comprising: S101. Obtain the current health assessment result of the wheelset tread to be tested; In S101, specifically, for the wheelset tread to be tested that is in working condition, a health monitoring system is generally used to monitor its health status in real time, and the corresponding health assessment results are output in real time. Therefore, each time the remaining life of the wheelset tread is predicted, the health assessment results of the wheelset tread to be tested at the current moment can be obtained from the health monitoring system of the wheelset tread to be tested, such as the health status at the time of this prediction. Taking the wheelset tread of an urban rail vehicle as an example, the wheelset tread is one of the key components of the running gear of the urban rail vehicle. The vehicle depot is equipped with an intelligent operation and maintenance system for urban rail vehicles to monitor its status and health assessment, realizing dynamic monitoring and health assessment of the wheelset tread. Therefore, each time the remaining life of the wheelset tread is predicted, the current health assessment results of the wheelset tread to be tested can be obtained from the intelligent operation and maintenance system of the urban rail vehicle in real time. The present application is not limited to this.

[0038] Among them, the health assessment result is divided into at least a first state for characterizing the normal operation of the wheelset tread to be tested, and a second state for characterizing that the wheelset tread to be tested has a tread out-of-round fault characteristic. According to the application scenario of the wheelset tread, the characteristics of a tread out-of-round fault cycle data are analyzed. When a wheelset tread just starts to operate normally, the amount of degradation characteristic data monitored is small, the degradation trend is uncertain, and all are normal data, and the model is not easy to fix. Accordingly, the health assessment result includes at least the first state for characterizing the normal operation of the wheelset tread to be tested; when the wheelset tread has early tread out-of-round characteristics, the degradation characteristics have a basic trend and a certain amount of data, and the model can basically be fixed. When the tread out-of-round fault characteristics appear, the degradation characteristics show a stable trend and a large amount of data, and the model can be determined. Accordingly, the health assessment result includes at least the second state for characterizing that the wheelset tread to be tested has a tread out-of-round fault characteristic.

[0039] The health assessment results can also be divided into more levels of status. In some embodiments, the health assessment results are divided into at least normal state, sub-healthy state and fault state, wherein the normal state is taken as the first state, and the sub-healthy state and fault state are taken as the second state; in other embodiments, the health assessment results are divided into at least normal state, sub-healthy state, minor fault state, moderate fault state and severe fault state, wherein the normal state is taken as the first state, and the sub-healthy state, minor fault state, moderate fault state and severe fault state are taken as the second state; the present application is not limited to this.

[0040] Among them, sub-health is defined as abnormal signs appearing, component performance begins to decline, but can perform its designed functions normally. Sub-health status can be expressed as the appearance of tread out-of-round fault characteristics, and degradation characteristics are characterized by a stable trend and a large amount of data. Therefore, the second state in the health assessment results, that is, the sub-health state, can be used as the segmentation point of the model.

[0041] S102. Determine a remaining life prediction model of the wheelset tread to be tested according to the health assessment result, wherein the remaining life prediction model is one of a first life prediction model corresponding to the first state and a second life prediction model corresponding to the second state; In S102, specifically, according to different degradation stages, a corresponding degradation model is selected; when the health assessment result is the first state, that is, the tread of the wheelset to be tested is operating normally, the first life prediction model corresponding to the first state is selected; when the health assessment result is the second state, that is, the tread of the wheelset to be tested has a tread out-of-round fault characteristic, the second life prediction model corresponding to the second state is selected. Among them, the first life prediction model and the second life prediction model can be selected based on actual application requirements, such as the first life prediction model adopts a double exponential model or other types of degradation models, and the second life prediction model adopts a single exponential model or other types of degradation models, but the present application is not limited thereto.

[0042] S103. Using the remaining life prediction model, the remaining life of the tread of the wheel pair to be measured is predicted to obtain a remaining life prediction result.

[0043] In S103, specifically, after determining the remaining life prediction model based on the current health assessment result of the wheelset tread to be tested, the remaining life prediction model can be used to perform remaining life prediction based on the real-time monitoring data of the wheelset tread to be tested and the operating mileage to obtain the remaining life prediction result.

[0044] At present, the remaining life prediction methods for wheelset treads mostly rely on empirical thresholds to divide the degradation stages of components and select the corresponding degradation model accordingly. Since the setting of empirical thresholds is mostly based on subjective judgment, it is easily affected by personal experience and there are differences, which will lead to deviations in the degradation stage division and do not conform to the actual situation of the application data, making the reliability of life prediction reduced and difficult to meet the precise maintenance needs of wheelset treads.

[0045] In the above embodiment, the optimal segmentation point of the segmentation model is selected from the perspective of component failure development. Specifically, based on the current health assessment result of the wheelset tread to be tested, the remaining life prediction model of the wheelset tread to be tested is determined. The health assessment result is at least divided into a first state for characterizing the normal operation of the wheelset tread to be tested, and a second state for characterizing that the wheelset tread to be tested has a tread out-of-round fault characteristic. The remaining life prediction model is one of a first life prediction model corresponding to the first state and a second life prediction model corresponding to the second state. The second state of the health assessment result is used as the segmentation point of the model. When the current health assessment result of the wheelset tread to be tested is the second state, the wheelset tread to be tested has a tread out-of-round fault characteristic, and the degradation characteristic is a stable trend with a large amount of data. Compared with the existing solution that relies on empirical thresholds to divide the degradation stages of components and selects the corresponding degradation model accordingly, the above embodiment effectively avoids the problem of deviation in the degradation stage division caused by subjective judgment and not conforming to the actual situation of the application data, and can improve the reliability of the wheelset tread life prediction.

[0046] The current remaining life prediction methods for wheelset treads often use feature extraction methods such as average values ​​and minimum values ​​for key tread indicators, and only consider features of a single dimension. Data features are not analyzed from the perspective of failure mode mechanisms, and multi-dimensional fusion features are not considered, making it impossible to reflect the overall degradation characteristics of components.

[0047] In order to solve the above technical problems, in other embodiments of the present application, a remaining life prediction model is used to predict the remaining life of the tread of the wheel pair to be measured, and one implementation method of the step of obtaining the remaining life prediction result specifically includes: S201. Obtaining real-time monitoring data and operating mileage of the wheelset to be tested; In S201, specifically, taking the wheelset tread of an urban rail vehicle as an example, each time the remaining life of the wheelset tread is predicted, the real-time monitoring data and the operating mileage of the wheelset tread to be tested can be obtained from the urban rail vehicle intelligent operation and maintenance system. Among them, the monitoring data includes vibration and impact data, specifically radial runout, out-of-round runout, vibration effective value, vibration dB value, impact waveform duty cycle, impact dB value and other trend data. The detailed description of each physical quantity is shown in the following table: Table 1 Description of physical quantities

[0048] The data characteristics and physical meanings of the monitoring indicators are analyzed in combination with the failure mode of the wheelset tread. Taking the wheelset tread out-of-round failure mode as an example, the analysis of the physical quantities of the monitoring indicators is as follows: (1) Radial runout: The physical value measured during wheel turning and repair can directly reflect the tread out-of-roundness of the wheelset under the tread out-of-round fault mode. However, this indicator is an offline measurement value and only has a value when the tread is turned. The number is small and discontinuous, and it is not suitable for life prediction scenarios; (2) Out-of-round runout: This is an online monitoring quantity that reflects the tread out-of-roundness and is strongly correlated with the tread out-of-round failure mode. In addition, there will be monitoring data when the vehicle is in normal operation, which is suitable for life prediction scenarios. (3) Shock waveform duty cycle: It can reflect the out-of-round caused by continuous peeling and is strongly correlated with the tread out-of-round failure mode. However, due to the data storage mechanism, the data record is incomplete and it is impossible to obtain an accurate shock waveform duty cycle. In addition, expert experience shows that if the train runs in a serpentine manner during driving, the shock waveform duty cycle calculation is inaccurate, so it is not suitable for life prediction scenarios. (4) Impact dB value: It reflects the impact information generated after the tread surface peels off. It cannot reflect the tread state under the tread out-of-round failure mode, so it is not suitable for life prediction scenarios. (5) Vibration effective value: reflects the vibration energy of the wheelset tread, which is strongly correlated with the tread out-of-round failure mode. In addition, there will be monitoring data when the vehicle is in normal operation, which is suitable for life prediction scenarios; (6) Vibration dB value: reflects the vibration intensity of the wheelset tread and is strongly correlated with the tread out-of-round failure mode. In addition, there will be monitoring data when the vehicle is operating normally, which is suitable for life prediction scenarios.

[0049] To sum up the above analysis, the indicators that can be used to predict the life of tread out-of-round fault are out-of-round runout, effective vibration value, and vibration dB value. The calculation of the vibration dB value uses the effective vibration value, and the vibration dB value covers the energy of the effective vibration value. Therefore, the indicators used to predict the life of tread out-of-round fault are vibration dB value and out-of-round runout. Accordingly, the real-time monitoring data includes the measured data of out-of-round runout and the measured data of vibration dB value.

[0050] S202. Use the remaining life prediction model to predict the remaining life according to the real-time monitoring data and the mileage in operation to obtain the remaining life prediction result.

[0051] In S202, specifically, after determining the remaining life prediction model based on the current health assessment result of the wheelset tread to be tested, the remaining life prediction model can be used to perform remaining life prediction based on the actual measured data of the out-of-round runout of the wheelset tread to be tested, the actual measured data of the vibration dB value and the operating mileage to obtain the remaining life prediction result.

[0052] In some embodiments, one implementation of this step specifically includes: S2021. Preprocess the measured data of the out-of-round runout, the measured data of the vibration dB value and the operated mileage to obtain the out-of-round runout data, the vibration dB value data and the operated mileage data; In S2021, specifically, outlier processing is performed on the measured data of the out-of-round runout and the measured data of the vibration dB value, respectively, to obtain out-of-round runout data and vibration dB value data. For example, for the out-of-round runout and the vibration dB value exceeding the specified range, the out-of-round runout is greater than or less than , vibration dB value greater than or less than , directly adopt the elimination method to avoid abnormal data affecting the accuracy of subsequent feature calculations.

[0053] Mileage correction is performed on the operated mileage to obtain the operated mileage data. Specifically, the operated mileage data of the vehicles recorded by the urban rail vehicle online monitoring system may have problems such as downward jumps, upward mutations, and zero values ​​due to operations such as software replacement with new machines. Combined with the daily operating mileage experience of the vehicle, the operated mileage is corrected point by point to ensure the monotonically increasing nature of the operated mileage data.

[0054] S2022. Extract features from the out-of-round runout data, the vibration dB value data and the operated mileage data, respectively, to obtain out-of-round runout features, vibration effective value features and operated mileage features; In S2022, specifically, based on the data characteristics, effective features of the out-of-round runout data, vibration dB value data, and operated mileage data are respectively extracted by sliding to obtain out-of-round runout features, vibration effective value features, and operated mileage features.

[0055] Among them, considering the non-uniform density characteristics of the out-of-round runout data, one of the implementation methods of the step of extracting features from the out-of-round runout data to obtain the out-of-round runout features specifically includes: calculating the adaptive first-order density weight-quantile according to the out-of-round runout data to obtain the out-of-round runout features; the adaptive first-order density weight-quantile is insensitive and relatively stable in the fault-free stage, but is more sensitive to the fault development stage, and can more accurately describe the fault evolution process of the tread. The specific calculation formula is as follows: ; ; In the formula, Indicates feature extraction interval, , is a positive integer, Indicates The data density of out-of-round runout in the feature extraction interval is: It represents the maximum value of the out-of-round runout data density within all feature extraction intervals. Indicates The first-order weighted density of feature extraction intervals, , Indicates The out-of-round runout measurement value for the feature extraction interval, is a positive integer, Express Pick Percentile, the operating mileage corresponding to each interval can be , Indicates the out-of-round runout characteristics.

[0056] Among them, considering that the effective value of vibration reflects the vibration intensity of the wheelset tread, the root mean square value within the calculation period reflects the energy of its vibration intensity. Accordingly, one implementation method of the step of extracting features of the vibration dB value data to obtain the effective value of vibration features specifically includes: calculating the root mean square value within a preset period according to the vibration dB value data to obtain the effective value of vibration features. The specific calculation formula is as follows: ; In the formula, Indicates the first The vibration dB value, , is the number of vibration dB values ​​corresponding to a preset period, It represents the average value of the vibration dB value data. The operating mileage corresponding to a preset cycle can be , Indicates the effective value characteristics of vibration.

[0057] Among them, one implementation method of the step of extracting features from the operated mileage data to obtain the operated mileage features specifically includes: slidingly calculating the mean of the operated mileage data to obtain the operated mileage features. Specifically, the sliding interval length can be 500km, and the fixed interval length can be 2000km. The present application is not limited to this.

[0058] S2023. Perform feature fusion on the out-of-round runout feature and the vibration effective value feature to obtain a fused degradation feature; In S2023, specifically, the extracted out-of-round runout features and vibration effective value features are normalized, and at the same time, considering the number of feature dimensions and the complexity and efficiency of engineering applications, combined with manual analysis experience, a fuzzy weighted method is used to extract fusion features to obtain fusion degradation features. One implementation of this step specifically includes: Out-of-round runout characteristics and vibration effective value characteristics , perform normalization processing to obtain the normalized out-of-round runout characteristics And the normalized vibration effective value characteristics ; The fuzzy weighted method is used to normalize the out-of-round runout characteristics. And the normalized vibration effective value characteristics , perform feature fusion and obtain fused degenerate features. The specific calculation formula is as follows: ; In the formula, Represents the normalized out-of-round runout characteristics The corresponding fuzzy weight is Represents the normalized vibration effective value characteristics The corresponding fuzzy weight is Indicates fusion degradation characteristics.

[0059] S2024. Use the remaining life prediction model to perform remaining life prediction based on the fusion degradation characteristics and the operated mileage characteristics to obtain the remaining life prediction result.

[0060] In S2024, specifically, after determining the remaining life prediction model based on the current health assessment result of the wheelset tread to be tested, the remaining life prediction model can be used to perform remaining life prediction based on the fused degradation characteristics and operated mileage characteristics of the wheelset tread to be tested, and obtain a remaining life prediction result.

[0061] The above embodiment starts from the fault mode and the fault mechanism of the component monitoring index data, analyzes the physical meaning of the index and the physical characteristics of the fault, selects vibration dB value and out-of-round runout as monitoring data, and predicts the remaining life. Taking into account the number of feature dimensions and the complexity and efficiency of engineering applications, a fused degradation feature extraction method based on vibration dB value and out-of-round runout is proposed, which can reflect the overall degradation characteristics of the component through multi-dimensional fusion, effectively improving the accuracy and efficiency of wheelset tread life prediction.

[0062] At present, the remaining life prediction method for wheelset treads often uses the Bayesian principle and the Monte Carlo method to estimate the parameters of the tread degradation model and obtain the predicted degradation trajectory. However, components may fail only after running for a certain period of time, and when the components start to run, the amount of monitoring data is small. When the amount of data is small, the Bayesian theory parameter solution method relies on the set initial parameters and is prone to fall into local optimality. Moreover, the initial parameters are the parameter distribution fitted by a small number of case samples, which deviates greatly from the parameter distribution of actual application data.

[0063] In order to solve the above technical problems, in other embodiments of the present application, a corresponding degradation model is selected according to the two degradation stages divided by the first state and the second state. Specifically: When the health assessment result is the first state, that is, the tread of the wheelset to be tested is operating normally, the first life prediction model corresponding to the first state is selected, and the first life prediction model adopts a double exponential model; specifically, when the tread of the wheelset to be tested is operating normally, the degradation trend is characterized by insufficient accumulation of data, and the double exponential model is used to fit the historical tread out-of-round failure case data to obtain an initial prediction model, thereby forming a set double exponential integrated model; wherein, the expression of the double exponential model is as follows: ; In the formula, represents the degenerate characteristic variable, represents the mileage variable, , , , Represents model parameters.

[0064] When the health assessment is combined with the second state, that is, the tread of the wheelset to be tested has the tread out-of-round fault characteristics, the second life prediction model corresponding to the second state is selected, and the second life prediction model adopts a single exponential model. Specifically, when the tread of the wheelset to be tested has the tread out-of-round fault characteristics, the degradation characteristics show a stable trend and the data has a certain amount. A single exponential model is established as the remaining life prediction model. The expression of the model is as follows: ; In the formula, represents the degenerate characteristic variable, represents the mileage variable, and Represents model parameters.

[0065] Based on the above embodiment and the construction of the above segmented life prediction model, a remaining life prediction is performed by using the remaining life prediction model according to the fusion degradation characteristics and the operating mileage characteristics, and one implementation method of the step of obtaining the remaining life prediction result specifically includes: S301. Obtaining a predicted failure threshold for wheel tread turning and repair; In S301, specifically, for the segmented life prediction model, when performing dynamic life prediction, it is necessary to first determine the turning prediction failure threshold corresponding to the optimal turning point. The turning prediction failure threshold can be calculated based on multiple historical case data of the wheelset tread, each of which includes radial runout value data, out-of-round runout historical data, and vibration dB value historical data.

[0066] In some embodiments, one implementation of this step specifically includes: S3011. Obtain multiple historical case data of wheelset tread; In S3011, specifically, multiple historical case data of wheelset treads can be obtained from an existing expert knowledge base or case library, and the historical case data can specifically be historical tread out-of-round fault case data, wherein each historical case data includes at least one radial runout value data, out-of-round runout historical data, and vibration dB value historical data of a turning and repair cycle. The historical case data can also be obtained by other reasonable methods, and the present application is not limited thereto.

[0067] S3012. Based on multiple historical case data, calculate and obtain a fused degradation characteristic value sequence, a radial runout value sequence, and a data sequence number sequence corresponding to the radial runout value sequence and the fused degradation characteristic value sequence; In S3012, specifically, based on the historical data of the out-of-roundness and the historical data of the vibration dB value of a turning and repair cycle in each historical case data, refer to the specific implementation details in steps S2021-S2023 to calculate the fused degradation characteristic value corresponding to each historical case data; for example, the historical data of the out-of-roundness and the historical data of the vibration dB value are preprocessed to obtain the preprocessed out-of-roundness data and the preprocessed vibration dB value data; the preprocessed out-of-roundness data and the preprocessed vibration dB value data are feature extracted to obtain the historical features of the out-of-roundness and the historical features of the effective vibration values; the historical features of the out-of-roundness and the historical features of the effective vibration values ​​are feature fused to obtain the fused degradation characteristic value. Further, based on the fused degradation characteristic value corresponding to each historical case data, a fused degradation characteristic value sequence is constructed, which is recorded as ; According to the diameter runout value data of a turning and repair cycle in each historical case data, a diameter runout value sequence is constructed and recorded as ; According to the radial runout value sequence and fused degenerate eigenvalue sequence , construct the data sequence corresponding to the two sequences, recorded as .

[0068] In a specific embodiment, taking 26 historical case data collected based on the expert knowledge base as an example, the fusion degradation feature value of a turning and repair cycle of each historical case data is extracted, and the fusion degradation feature value sequence of the turning and repair is counted. =[0.59, 0.82, 0.83, 0.67, 0.76, 0.91, 0.58, 0.89, 0.86, 0.67, 0.74, 1.17, 0.92, 0.75, 0.76, 0.53, 0.38, 0.42, 0.26, 0.41, 0.34, 0.55, 0.35, 0.44, 0.35, 0.85], and the series of diameter runout values ​​during turning and repair is counted. =[0.53, 1.05, 1.11, 1.07, 1.11, 1.28, 1.04, 0.8, 1.17, 1.26, 0.86, 0.96, 1.20, 1.40, 1.10, 0.52, 0.53, 0.86, 0.68, 0.76, 0.62, 0.82, 0.91, 0.66, 0.71, 0.93], and with the two sequences and Corresponding data sequence number =[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26].

[0069] S3013. Sort the fused degenerate eigenvalue sequence in ascending order to obtain a sorted first sequence, and sort the data sequence number sequence based on the first sequence to obtain a sorted second sequence; In S3013, specifically, the fused degenerate feature value Sort from small to large to get the first sequence after sorting, recorded as ; and according to the first sequence , for the data sequence Sort and get the second sorted sequence, recorded as .

[0070] Taking the above specific embodiment as an example, the fused degradation feature value Sort from small to large to get the first sequence =[0.26, 0.34, 0.35, 0.35, 0.38, 0.41, 0.42, 0.44, 0.53, 0.55, 0.58, 0.59, 0.67, 0.67, 0.74, 0.75, 0.76, 0.76, 0.82, 0.83, 0.85, 0.86, 0.89, 0.91, 0.92, 1.17], according to the first sequence , for the data sequence Sort and get the sorted second sequence =[19, 21, 23, 25, 17, 20, 18, 24, 16, 22, 7, 1, 4, 10, 11, 14, 5, 15, 2, 3, 26, 9, 8, 6, 13, 12]. It should be noted that for the first sequence The same values ​​appearing in the same sequence, such as two 0.35, two 0.67 and two 0.76, are , for the data sequence When sorting, you can sort the data sequence numbers from small to large according to the data sequence numbers corresponding to these values. Sort.

[0071] S3014. Segment the first sequence according to the preset interval length to obtain multiple segmented sequences; In S3014, specifically, the preset interval length can be set based on actual needs. Preferably, the preset interval length can be set to 0.1. Segmentation is performed to obtain multiple segment sequences after segmentation.

[0072] Taking the above specific embodiment as an example, the first sequence The maximum and minimum values ​​are 1.17 and 0.26 respectively. After segmentation, the following 10 segment sequences can be obtained: the segment sequence of the interval [0.2, 0.3) = [0.26], the segment sequence of the interval [0.3, 0.4) = [0.34, 0.35, 0.35, 0.38], the segment sequence of the interval [0.4, 0.5) = [0.41, 0.42, 0.44], the segment sequence of the interval [0.5, 0.6) = [0.53, 0.55, 0.58, 0.59], the segment sequence of the interval [0.6, 0.7) = [0. sequence = [0.67, 0.67], the segmented sequence of the interval [0.7, 0.8) = [0.74, 0.75, 0.76, 0.76], the segmented sequence of the interval [0.8, 0.9) = [0.82, 0.83, 0.85, 0.86, 0.89], the segmented sequence of the interval [0.9, 1.0) = [0.91, 0.92], the segmented sequence of the interval [1.0, 1.1) is empty, the segmented sequence of the interval [1.1, 1.2) = [1.17].

[0073] S3015. According to the segmented sequence, respectively count the data sequence number of the second sequence corresponding to each segmented sequence; In S3015, specifically, according to the segmented sequence, the second sequence within the range of each segmented sequence is counted respectively. The data sequence number.

[0074] Taking the above specific embodiment as an example, the second sequence in each segment of the above 10 segment sequences is segmented and counted. The data sequence number is shown in the following table: Table 2 Statistics of segmented sequence and second sequence

[0075] S3016. According to the data sequence number of the second sequence corresponding to each segmented sequence, count the number of radial jump values ​​of the corresponding radial jump value sequence, so as to obtain a radial jump value number sequence corresponding to the segmented sequence; In S3016, specifically, according to the second sequence within the range of each segment sequence The data sequence number is used to count the radial runout value sequence within the corresponding range. The number of data, that is, the number of radial runout values, finally obtains the segmented radial runout value sequence The sequence of radial runout values ​​is denoted as .

[0076] Taking the above specific embodiment as an example, according to the second sequence in each segment range of the above 10 segment sequences The data sequence number is used to count the radial runout value sequence within the corresponding range. The number of data is finally obtained, and the segmented radial jump value sequence is obtained. The sequence of the number of radial runout values , as shown in the following table: Table 3 Statistics of segmented sequence and radial runout value sequence

[0077] From the above table, we can see that the sequence of radial runout values ​​is =[1, 4, 3, 4, 2, 4, 5, 2, 0, 1].

[0078] S3017. Draw a frequency histogram of the diameter runout value number sequence according to the diameter runout value number sequence, the maximum value and the minimum value of the first sequence, and the number of segmented sequences; In S3017, specifically, the first sequence is screened out The maximum and minimum values ​​of the radial runout value sequence , first sequence The maximum and minimum values, as well as the number of segmented sequences, plot the number of radial runout values Taking the above specific embodiment as an example, according to the above jump value number sequence , first sequence The maximum value is 1.17 and the minimum value is 0.26. The number of segmented sequences is 10. Draw the sequence of the number of radial runout values. The frequency histogram of (i.e., the point diagram of fused degradation eigenvalues ​​during maintenance) is as follows: Figure 2 shown.

[0079] S3018. According to the frequency histogram, data fitting is performed using statistical distribution to obtain a fitting result; In S3018, specifically, according to the frequency histogram, data fitting is performed using statistical distribution to obtain a fitting result, wherein the statistical distribution method can be selected based on actual needs, preferably, multi-modal distribution is adopted. Taking the above specific embodiment as an example, a multi-modal distribution fitting distribution diagram based on multi-modal distribution fitting is as follows: Figure 3 shown.

[0080] S3019. According to the fitting results and the preset confidence interval, determine the coverage interval of the fused degradation characteristic value, and calculate the mean value of the fused degradation characteristic value within the coverage interval of the fused degradation characteristic value as the failure threshold for turning and repair prediction.

[0081] In S3019, specifically, the preset confidence interval can be set based on actual needs. Preferably, the 95% confidence interval of the multi-peak distribution can be taken as the preset confidence interval; according to the multi-peak distribution fitting result and the preset confidence interval, the fused degradation characteristic value coverage interval corresponding to the preset confidence interval is determined, and the mean of the fused degradation characteristic value within the fused degradation characteristic value coverage interval is calculated as the turning and repair prediction failure threshold.

[0082] Taking the above specific embodiment as an example, taking the 95% confidence interval of the multi-peak distribution, it is obtained that the coverage interval of the fused degradation characteristic value during maintenance is [0.4, 1]. The fused degradation characteristic values ​​within the coverage interval [0.4, 1] of the fused degradation characteristic value can be determined, specifically [0.41, 0.42, 0.44, 0.53, 0.55, 0.58, 0.59, 0.67, 0.67, 0.74, 0.75, 0.76, 0.76, 0.82, 0.83, 0.85, 0.86, 0.89, 0.91, 0.92]. The average value of all values ​​in the sequence is 0.6975, which is used as the optimal turning and repair prediction failure threshold.

[0083] In other embodiments, according to application experience, when the radial runout value is greater than 0.5, the tread needs to be repaired. According to the radial runout value sequence corresponding to the interval [0.4, 1] of the fusion degradation characteristic value obtained during the repair, The values ​​greater than 0.5 are deleted, and the radial runout values ​​less than or equal to 0.5 are eliminated. The radial runout value sequence after deletion is [0.76, 0.86, 0.66, 0.52, 0.82, 1.04, 0.53, 1.07, 1.26, 0.86, 1.40, 1.11, 1.10, 1.05, 1.11, 0.93, 1.17, 0.80, 1.28, 1.20] and its corresponding data sequence number sequence [20, 18, 24, 1 6, 22, 7, 1, 4, 10, 11, 14, 5, 15, 2, 3, 26, 9, 8, 6, 13], and the corresponding fused degradation eigenvalue sequence for calculating the optimal turning and repair prediction failure threshold [0.41, 0.42, 0.44, 0.53, 0.55, 0.58, 0.59, 0.67, 0.67, 0.74, 0.75, 0.76, 0.76, 0.82, 0.83, 0.85, 0.86, 0.89, 0.91, 0.92].

[0084] Through steps S3011-S3019, the formulation of the optimal predicted failure point of the component is realized. The accumulated component failure case data information can be integrated, and the optimal turning point / replacement point is taken as the end point of a cycle to formulate the optimal turning point / replacement point, so that the remaining life prediction result is more in line with the actual application situation.

[0085] S302. Utilize the remaining life prediction model to perform remaining life prediction according to the failure threshold predicted by turning and repair, fusion degradation characteristics and operating mileage characteristics, and obtain the remaining life prediction result.

[0086] In S302, specifically, after determining the remaining life prediction model based on the current health assessment result of the wheelset tread to be tested, the remaining life prediction model can be used to perform current remaining life prediction point estimation and interval estimation value calculation based on the fused degradation characteristics, operating mileage characteristics and turning and repair prediction failure threshold of the wheelset tread to be tested as the remaining life prediction result.

[0087] In some embodiments, when the health assessment result is the first state, that is, the tread of the wheelset to be tested is operating normally, a first life prediction model corresponding to the first state is selected. Accordingly, one implementation of this step specifically includes: S3021. Obtain multiple historical case fusion degradation features of wheelset treads; In S3021, specifically, each historical case fusion degradation feature is extracted based on a corresponding set of historical case data of the wheelset tread, and each set of historical case data includes historical data of out-of-round runout and historical data of vibration dB value.

[0088] In some embodiments, one implementation of this step specifically includes: obtaining multiple historical case data of wheelset treads; and calculating corresponding historical case fusion degradation features based on each historical case data.

[0089] Among them, multiple historical case data of wheelset treads can be obtained from the existing expert knowledge base or case library, and the historical case data can specifically be historical tread out-of-round fault case data, wherein each historical case data includes out-of-round runout historical data and vibration dB value historical data. The historical case data can also be obtained by other reasonable means, and the present application is not limited thereto. According to the out-of-round runout historical data and vibration dB value historical data in each historical case data, refer to the specific implementation details in steps S2021-S2023, and calculate the historical case fusion degradation features corresponding to each historical case data; for example, the out-of-round runout historical data and the vibration dB value historical data are respectively preprocessed to obtain preprocessed out-of-round runout data and preprocessed vibration dB value data; the preprocessed out-of-round runout data and the preprocessed vibration dB value data are respectively feature extracted to obtain the out-of-round runout historical features and the vibration effective value historical features; the out-of-round runout historical features and the vibration effective value historical features are feature fused to obtain the historical case fusion degradation features.

[0090] S3022. Calculate the similarity between the fused degradation feature and each historical case fused degradation feature respectively, and select the historical case fused degradation feature with the greatest similarity as the first historical case fused degradation feature; In S3022, specifically, the existing similarity calculation method can be used to calculate the similarity between the fused degradation feature and each historical case fused degradation feature, and the historical case fused degradation feature corresponding to the maximum similarity value is selected as the first historical case fused degradation feature.

[0091] S3023. Obtaining a pre-built dual exponential model corresponding to the first historical case fusion degradation feature as a first life prediction model; In S3023, specifically, a pre-constructed double exponential model corresponding to the first historical case fused degradation feature is obtained based on the fitting of the first historical case fused degradation feature; in some embodiments, the double exponential model can be used to fit each historical case fused degradation feature separately to obtain a double exponential integrated model, wherein the double exponential integrated model includes a double exponential model corresponding to each historical case fused degradation feature; after determining the first historical case fused degradation feature, the pre-constructed double exponential model corresponding to the first historical case fused degradation feature can be obtained from the double exponential integrated model as the first life prediction model.

[0092] It should be noted that in the process of pre-constructing the double exponential model, a parameter solution method based on Bayesian theory or a same-round alternating iterative parameter solution method can be used; but in the segmented model, the first stage is to directly use the historical case data extracted from the expert knowledge base or case library to fuse the degradation features of the historical cases and perform double exponential function fitting to obtain the theoretical initial prediction model, which does not require a parameter updating method.

[0093] S3024. Using the first life prediction model, according to the operating mileage characteristics and the turning and repair prediction failure threshold, the remaining life prediction is performed to obtain the point estimate and interval estimate of the remaining service life of the wheelset tread to be tested.

[0094] In S3024, specifically, the first life prediction model is used to calculate the remaining service life point estimate and interval estimate of the tread of the wheelset to be tested at the current moment based on the operating mileage characteristics and the turning and repair prediction failure threshold as the remaining service life prediction result.

[0095] In other embodiments, when the health assessment is combined with the second state, that is, the tread of the wheelset to be tested has a tread out-of-round fault characteristic, a second life prediction model corresponding to the second state is selected. Accordingly, one implementation of this step specifically includes: S3025. Obtain a pre-built single index model, and fit the single index model by integrating degradation characteristics and operating mileage characteristics to obtain a second life prediction model; In S3025, specifically, the parameters of the single exponential model at the current moment are estimated by utilizing the fused degradation characteristics and the operated mileage characteristics to obtain a second life prediction model; preferably, the single exponential model can be fitted by utilizing the fused degradation characteristics and the operated mileage characteristics using the least squares method to obtain a second life prediction model, wherein the least squares method can accurately solve parameters when the amount of data is large, and is often used for models with a small number of parameters and simple maximum likelihood.

[0096] S3026. Using the second life prediction model, according to the operating mileage characteristics and the turning and repair prediction failure threshold, the remaining life prediction is carried out to obtain the point estimate and interval estimate of the remaining service life of the wheelset tread to be tested.

[0097] In S3026, specifically, the predicted operating mileage corresponding to the turning and repair prediction failure threshold is calculated using the second life prediction model; the point estimate of the remaining service life of the wheelset tread to be tested is calculated based on the predicted operating mileage and the operating mileage characteristics, and the interval estimate of the remaining service life of the wheelset tread to be tested is calculated using the Monte Carlo method. Among them, the predicted operating mileage can be subtracted from the operating mileage characteristics to obtain the point estimate of the remaining service life of the wheelset tread to be tested at the current moment; the Monte Carlo method is a numerical calculation method based on random sampling. Through the Monte Carlo method, a large number of possible degradation trajectories can be generated to estimate the future state of the components and obtain the interval estimate of the remaining service life of the wheelset tread to be tested at the current moment.

[0098] In the above embodiment, a segmented life prediction model is established based on the data characteristics of different fault stages. The first stage adopts a set theoretical double exponential model degradation model, and the second stage adopts a single exponential dynamic prediction model. This effectively solves the problem that the existing life prediction method relies on a small number of case samples to fit the parameter distribution, which deviates greatly from the parameter distribution of actual application data.

[0099] In addition, the above embodiment also calculates the degradation characteristics of the case throughout its life cycle based on the accumulated component failure cases, takes a turning / replacement point as the end point of a cycle, performs distribution fitting based on all statistical cases, and takes the value when the confidence level is the set value as the turning prediction failure threshold. It can integrate the accumulated component failure case data information and formulate the optimal turning / replacement point, so that the remaining life prediction results are more in line with the actual application situation, and effectively solves the problem that the existing life prediction method relies on subjective experience to set the failure threshold, which deviates greatly from the actual situation.

[0100] like Figure 4 As shown, in another embodiment of the present application, a system for predicting the remaining life of a wheelset tread is also provided, comprising: An acquisition unit 10 is used to acquire a current health assessment result of the wheelset tread to be tested, wherein the health assessment result is divided into at least a first state for characterizing that the wheelset tread to be tested is operating normally, and a second state for characterizing that the wheelset tread to be tested has a tread out-of-round fault characteristic; A determination unit 11 is used to determine a remaining life prediction model of the tread of the wheelset to be tested according to the health assessment result, wherein the remaining life prediction model is one of a first life prediction model corresponding to the first state and a second life prediction model corresponding to the second state; The prediction unit 12 is used to use the remaining life prediction model to predict the remaining life of the tread of the wheel pair to be measured, and obtain a remaining life prediction result.

[0101] In other embodiments of the present application, in the remaining life prediction system of the above-mentioned wheelset tread, the health assessment results are divided into at least normal state, sub-health state and fault state, wherein the normal state is taken as the first state, and the sub-health state and fault state are taken as the second state.

[0102] Preferably, in the above-mentioned method for predicting the remaining life of a wheelset tread, the prediction unit 12 is specifically used to: Obtain the real-time monitoring data and operating mileage of the wheelset to be tested, where the real-time monitoring data includes the measured data of the out-of-round runout and the measured data of the vibration dB value; The remaining life prediction model is used to predict the remaining life according to the real-time monitoring data and the mileage in operation to obtain the remaining life prediction result.

[0103] In other embodiments of the present application, in the above-mentioned remaining life prediction system for wheelset tread, the prediction unit 12 performs the remaining life prediction based on the real-time monitoring data and the operating mileage using the remaining life prediction model to obtain the remaining life prediction result, specifically for: The measured data of out-of-round runout, the measured data of vibration dB value and the operated mileage are preprocessed to obtain the out-of-round runout data, the vibration dB value data and the operated mileage data; The out-of-round runout data, vibration dB value data and operating mileage data are respectively subjected to feature extraction to obtain out-of-round runout features, vibration effective value features and operating mileage features; The out-of-round runout feature and the vibration effective value feature are fused to obtain the fused degradation feature; The remaining life prediction model is used to predict the remaining life based on the fusion degradation characteristics and the operating mileage characteristics to obtain the remaining life prediction result.

[0104] In other embodiments of the present application, in the above-mentioned remaining life prediction system of the wheelset tread, when the prediction unit 12 performs feature extraction on the out-of-round runout data to obtain the out-of-round runout feature, it is specifically used to: According to the out-of-round runout data, the adaptive first-order density weight-quantile is calculated to obtain the out-of-round runout characteristics. The specific calculation formula is as follows: ; ; In the formula, Indicates feature extraction interval, , is a positive integer, Indicates The data density within the feature extraction interval is Represents the maximum value of data density in all feature extraction intervals, Indicates The first-order weighted density of feature extraction intervals, , Indicates The measured values ​​of feature extraction intervals, is a positive integer, Express Pick Percentile, Indicates the out-of-round runout characteristics.

[0105] In other embodiments of the present application, in the above-mentioned remaining life prediction system of the wheelset tread, when the prediction unit 12 performs feature extraction on the vibration dB value data to obtain the vibration effective value feature, it is specifically used to: According to the vibration dB value data, the root mean square value within the preset period is calculated to obtain the vibration effective value characteristics. The specific calculation formula is as follows: ; In the formula, Indicates the first The vibration dB value, , is the number of vibration dB values ​​corresponding to a preset period, Indicates the average value of the vibration dB value data, Indicates the effective value characteristics of vibration.

[0106] In other embodiments of the present application, in the above-mentioned remaining life prediction system of the wheelset tread, when the prediction unit 12 performs feature fusion on the out-of-round runout feature and the vibration effective value feature to obtain the fused degradation feature, it is specifically used to: Normalizing the out-of-roundness runout quantity characteristics and the vibration effective value characteristics to obtain normalized out-of-roundness runout quantity characteristics and normalized vibration effective value characteristics; The fuzzy weighted method is used to fuse the normalized out-of-round runout feature and the normalized vibration effective value feature to obtain the fused degradation feature. The specific calculation formula is as follows: ; In the formula, Represents the normalized out-of-round runout characteristics The corresponding fuzzy weight is Represents the normalized vibration effective value characteristics The corresponding fuzzy weight is Indicates fusion degradation characteristics.

[0107] In other embodiments of the present application, in the above-mentioned remaining life prediction system for the wheelset tread, the first life prediction model adopts a double exponential model, and the second life prediction model adopts a single exponential model.

[0108] In other embodiments of the present application, in the above-mentioned remaining life prediction system for wheelset tread, the prediction unit 12 performs the remaining life prediction by using the remaining life prediction model, based on the fusion degradation characteristics and the operating mileage characteristics, to obtain the remaining life prediction result, specifically for: Obtaining a predicted failure threshold of the wheelset tread for turning and repair, wherein the predicted failure threshold of the wheelset tread for turning and repair is calculated based on multiple historical case data of the wheelset tread, each historical case data including radial runout value data, out-of-round runout value historical data, and vibration dB value historical data; The remaining life prediction model is used to predict the failure threshold of the turning and repair prediction, integrate the degradation characteristics and the operating mileage characteristics, and obtain the remaining life prediction result.

[0109] In other embodiments of the present application, in the above-mentioned remaining life prediction system of the wheelset tread, when the prediction unit 12 executes the acquisition of the prediction failure threshold of the wheelset tread, it is specifically used to: Obtain multiple historical case data of wheelset tread; According to multiple historical case data, a fused degradation characteristic value sequence, a radial runout value sequence, and a data sequence number sequence corresponding to the radial runout value sequence and the fused degradation characteristic value sequence are calculated; The fused degenerate eigenvalue sequence is sorted in ascending order to obtain a sorted first sequence, and the data sequence number sequence is sorted based on the first sequence to obtain a sorted second sequence; Segmenting the first sequence according to a preset interval length to obtain a plurality of segmented sequences; According to the segmented sequences, respectively counting the data sequence number of the second sequence corresponding to each segmented sequence; According to the data sequence number of the second sequence corresponding to each segmented sequence, the number of radial jump values ​​of the corresponding radial jump value sequence is counted, so as to obtain a radial jump value number sequence corresponding to the segmented sequence; According to the radial runout value number sequence, the maximum value and the minimum value of the first sequence, and the number of segmented sequences, a frequency histogram of the radial runout value number sequence is drawn; According to the frequency histogram, the data is fitted using statistical distribution to obtain the fitting result; According to the fitting results and the preset confidence interval, the coverage interval of the fused degradation characteristic value is determined, and the mean value of the fused degradation characteristic value within the coverage interval is calculated as the failure threshold for turning and repair prediction.

[0110] In other embodiments of the present application, in the above-mentioned remaining life prediction system for the wheelset tread, when the remaining life prediction model is the first life prediction model corresponding to the first state, the prediction unit 12 performs the remaining life prediction using the remaining life prediction model according to the turning and repair prediction failure threshold, the fusion degradation characteristics and the operating mileage characteristics, and obtains the remaining life prediction result, specifically for: Acquire multiple historical case fusion degradation features of the wheelset tread, wherein each historical case fusion degradation feature is extracted based on a corresponding historical case data; Calculate the similarity between the fused degraded feature and the fused degraded feature of each historical case respectively, and select the fused degraded feature of the historical case with the largest similarity as the first fused degraded feature of the historical case; Obtaining a pre-built double exponential model corresponding to the first historical case fused with the degradation feature as a first life prediction model, wherein the double exponential model is fitted based on the first historical case fused with the degradation feature; The first life prediction model is used to predict the remaining life according to the operating mileage characteristics and the failure threshold predicted by turning and repair, and the point estimate and interval estimate of the remaining service life of the wheelset tread to be tested are obtained.

[0111] In other embodiments of the present application, in the above-mentioned remaining life prediction system for the wheelset tread, when the remaining life prediction model is the second life prediction model corresponding to the second state, the prediction unit 12 performs the remaining life prediction using the remaining life prediction model according to the turning and repair prediction failure threshold, the fusion degradation characteristics and the operating mileage characteristics, and obtains the remaining life prediction result, specifically for: Obtain a pre-built single index model, and fit the single index model by integrating degradation characteristics and operating mileage characteristics to obtain a second life prediction model; The second life prediction model is used to predict the remaining life according to the operating mileage characteristics and the failure threshold predicted by turning and repair, and the point estimate and interval estimate of the remaining service life of the wheelset tread to be tested are obtained.

[0112] In other embodiments of the present application, in the above-mentioned remaining life prediction system of the wheelset tread, when the prediction unit 12 performs fitting of the single exponential model by integrating the degradation feature and the operating mileage feature to obtain the second life prediction model, it is specifically used to: By integrating the degradation characteristics and the operating mileage characteristics and adopting the least square method, the single exponential model is fitted to obtain the second life prediction model.

[0113] In other embodiments of the present application, in the above-mentioned remaining life prediction system for wheelset treads, the prediction unit 12 performs the remaining life prediction using the second life prediction model according to the operating mileage characteristics and the turning and repair prediction failure threshold to obtain the point estimate and interval estimate of the remaining service life of the wheelset tread to be tested, specifically for: The second life prediction model is used to calculate the predicted operating mileage corresponding to the predicted failure threshold of turning and repair; According to the predicted operating mileage and the operating mileage characteristics, the point estimate of the remaining service life of the wheelset tread to be tested is calculated, and the interval estimate of the remaining service life of the wheelset tread to be tested is calculated using the Monte Carlo method.

[0114] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the remaining life of a wheelset tread, characterized in that: include: Obtaining a current health assessment result of the wheelset tread to be tested, wherein the health assessment result is divided into at least a first state for characterizing that the wheelset tread to be tested is operating normally, and a second state for characterizing that the wheelset tread to be tested has a tread out-of-round fault characteristic; Determining a remaining life prediction model of the wheelset tread to be tested according to the health assessment result, wherein the remaining life prediction model is one of a first life prediction model corresponding to the first state and a second life prediction model corresponding to the second state; The remaining life prediction model is used to predict the remaining life of the tread of the wheelset to be tested, and a remaining life prediction result is obtained.

2. The method according to claim 1, characterized in that The health assessment result is divided into at least a normal state, a sub-healthy state and a faulty state, wherein the normal state is taken as the first state, and the sub-healthy state and the faulty state are taken as the second state.

3. The method according to claim 1, characterized in that The method of using the remaining life prediction model to predict the remaining life of the tread of the wheel pair to be tested to obtain a remaining life prediction result includes: Acquire real-time monitoring data and operating mileage of the wheelset tread to be tested, wherein the real-time monitoring data includes measured data of out-of-round runout and measured data of vibration dB value; The remaining life prediction model is used to perform remaining life prediction based on the real-time monitoring data and the operated mileage to obtain a remaining life prediction result.

4. The method as claimed in claim 3, characterized in that The remaining life prediction model is used to perform remaining life prediction according to the real-time monitoring data and the operated mileage to obtain a remaining life prediction result, including: The out-of-round runout measured data, the vibration dB value measured data and the operated mileage are respectively preprocessed to obtain out-of-round runout data, vibration dB value data and operated mileage data; The out-of-round runout data, the vibration dB value data and the operated mileage data are respectively subjected to feature extraction to obtain out-of-round runout features, vibration effective value features and operated mileage features; Performing feature fusion on the out-of-round runout feature and the vibration effective value feature to obtain a fused degradation feature; The remaining life prediction model is used to perform remaining life prediction according to the fused degradation characteristics and the operated mileage characteristics to obtain a remaining life prediction result.

5. The method as claimed in claim 4, characterized in that The step of extracting features from the out-of-round runout data to obtain out-of-round runout features includes: According to the out-of-round runout data, the adaptive first-order density weight-quantile is calculated to obtain the out-of-round runout feature. The specific calculation formula is as follows: ； ； In the formula, Indicates feature extraction interval, , is a positive integer, Indicates The data density within the feature extraction interval is Represents the maximum value of data density in all feature extraction intervals, Indicates The first-order weighted density of feature extraction intervals, , Indicates The measured values ​​of feature extraction intervals, is a positive integer, Express Pick Percentile, Represents the out-of-round runout characteristics.

6. The method as claimed in claim 5, characterized in that The step of extracting features from the vibration dB value data to obtain vibration effective value features includes: According to the vibration dB value data, the root mean square value within the preset period is calculated to obtain the vibration effective value characteristics. The specific calculation formula is as follows: ； In the formula, Indicates the first The vibration dB value, , is the number of vibration dB values ​​corresponding to a preset period, represents the average value of the vibration dB value data, Represents the vibration effective value characteristics.

7. The method as claimed in claim 6, characterized in that The step of fusing the out-of-round runout feature and the vibration effective value feature to obtain a fused degradation feature includes: Normalizing the out-of-round runout amount feature and the vibration effective value feature to obtain a normalized out-of-round runout amount feature and a normalized vibration effective value feature; The fuzzy weighted method is used to fuse the normalized out-of-round runout feature and the normalized vibration effective value feature to obtain the fused degradation feature. The specific calculation formula is as follows: ； In the formula, represents the normalized out-of-round runout characteristic The corresponding fuzzy weight, Represents the normalized vibration effective value characteristic The corresponding fuzzy weight is Represents the fusion degradation feature.

8. The method as claimed in claim 4, characterized in that The first life prediction model adopts a double exponential model, and the second life prediction model adopts a single exponential model.

9. The method as claimed in claim 8, characterized in that The using the remaining life prediction model to perform remaining life prediction according to the fused degradation characteristics and the operated mileage characteristics to obtain a remaining life prediction result includes: Obtaining a predicted failure threshold of the wheel tread for turning and repairing, wherein the predicted failure threshold of the wheel tread for turning and repairing is calculated based on a plurality of historical case data of the wheel tread, each of the historical case data comprising radial runout value data, out-of-round runout value historical data, and vibration dB value historical data; The remaining life prediction model is used to perform remaining life prediction according to the turning and repair prediction failure threshold, the fused degradation characteristics and the operating mileage characteristics to obtain a remaining life prediction result.

10. The method as claimed in claim 9, characterized in that The step of obtaining the predicted failure threshold of the wheelset tread turning and repairing includes: Acquire a plurality of historical case data of wheelset treads; According to the plurality of historical case data, a fused degradation characteristic value sequence, a radial runout value sequence, and a data sequence number sequence corresponding to the radial runout value sequence and the fused degradation characteristic value sequence are calculated; Sorting the fused degenerate eigenvalue sequence in ascending order to obtain a sorted first sequence, and sorting the data sequence number sequence according to the first sequence to obtain a sorted second sequence; Segmenting the first sequence according to a preset interval length to obtain a plurality of segmented sequences; According to the segmented sequences, respectively counting the data sequence number of the second sequence corresponding to each of the segmented sequences; According to the data sequence number of the second sequence corresponding to each segmented sequence, the number of radial jump values ​​of the corresponding radial jump value sequence is counted, so as to obtain a radial jump value number sequence corresponding to the segmented sequence; Draw a frequency histogram of the radial runout value number sequence according to the radial runout value number sequence, the maximum value and the minimum value of the first sequence, and the number of segmented sequences; According to the frequency histogram, data fitting is performed using statistical distribution to obtain a fitting result; According to the fitting result and the preset confidence interval, the coverage interval of the fused degradation characteristic value is determined, and the mean value of the fused degradation characteristic value within the coverage interval of the fused degradation characteristic value is calculated as the turning and repair prediction failure threshold.

11. The method according to claim 9, characterized in that When the remaining life prediction model is the first life prediction model corresponding to the first state, the remaining life prediction is performed using the remaining life prediction model according to the turning and repair prediction failure threshold, the fused degradation feature and the operating mileage feature to obtain a remaining life prediction result, including: Acquire multiple historical case fused degradation features of the wheelset tread, wherein each of the historical case fused degradation features is extracted based on a corresponding copy of the historical case data; Calculating the similarity between the fused degradation feature and each of the historical case fused degradation features respectively, and selecting the historical case fused degradation feature with the greatest similarity as the first historical case fused degradation feature; Acquire a pre-built double exponential model corresponding to the first historical case fused with the degradation feature as the first life prediction model, wherein the double exponential model is fitted based on the first historical case fused with the degradation feature; The first life prediction model is used to perform a remaining life prediction based on the operating mileage characteristics and the turning and repair prediction failure threshold to obtain a point estimate and an interval estimate of the remaining service life of the wheelset tread to be tested.

12. The method according to claim 9, characterized in that When the remaining life prediction model is the second life prediction model corresponding to the second state, the remaining life prediction is performed using the remaining life prediction model according to the turning and repair prediction failure threshold, the fused degradation feature and the operating mileage feature to obtain a remaining life prediction result, including: Obtaining a pre-built single exponential model, and fitting the single exponential model using the fused degradation feature and the operated mileage feature to obtain the second life prediction model; The second life prediction model is used to perform a remaining life prediction based on the operating mileage characteristics and the turning and repair prediction failure threshold to obtain a point estimate and an interval estimate of the remaining service life of the wheelset tread to be tested.

13. The method as claimed in claim 12, characterized in that The step of fitting the single exponential model by using the fused degradation feature and the operated mileage feature to obtain the second life prediction model includes: The single exponential model is fitted by using the fused degradation feature and the operated mileage feature and adopting the least squares method to obtain the second life prediction model.

14. The method of claim 12, wherein: The method of using the second life prediction model to predict the remaining life according to the operating mileage characteristics and the turning and repair prediction failure threshold to obtain a point estimate and an interval estimate of the remaining service life of the wheelset tread to be tested includes: Calculate the predicted operating mileage corresponding to the turning and repair prediction failure threshold by using the second life prediction model; According to the predicted operating mileage and the operated mileage characteristics, a point estimate of the remaining service life of the wheelset tread to be measured is calculated, and using the Monte Carlo method, an interval estimate of the remaining service life of the wheelset tread to be measured is calculated.

15. A system for predicting the remaining life of a wheelset tread, characterized in that: include: an acquisition unit, configured to acquire a current health assessment result of the wheelset tread to be tested, wherein the health assessment result is at least divided into a first state for characterizing that the wheelset tread to be tested is operating normally, and a second state for characterizing that the wheelset tread to be tested has a tread out-of-round fault characteristic; a determination unit, configured to determine a remaining life prediction model of the tread of the wheelset to be tested according to the health assessment result, wherein the remaining life prediction model is one of a first life prediction model corresponding to the first state and a second life prediction model corresponding to the second state; The prediction unit is used to use the remaining life prediction model to predict the remaining life of the tread of the wheel pair to be tested, and obtain a remaining life prediction result.

Images