Train wheel polygon prediction method, system, device, medium and program product
By constructing a train wheel polygon prediction model based on the random forest algorithm, the problem of unpredictable wheel polygon wear was solved, the prediction accuracy was improved, the wheel service life was extended, and the operating cost was reduced.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHUZHOU CSR TIMES ELECTRIC CO LTD
- Filing Date
- 2024-11-07
- Publication Date
- 2026-05-08
AI Technical Summary
In existing technologies, the problem of polygonal wear on wheels is common, which leads to a decline in vehicle dynamics, an increase in vibration and noise, and makes it difficult for maintenance departments to carry out timely repairs, affecting train operation safety and maintenance costs.
A train wheel polygon prediction model based on the random forest algorithm is constructed. By acquiring historical wheel data, correlation analysis and feature index selection are performed, and the model is trained using the random forest algorithm to predict the state of the wheel polygon.
It improves the accuracy of wheel polygon prediction, provides important maintenance reference, extends wheel life, and reduces rail transit operation and maintenance costs.
Smart Images
Figure CN121997197A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of public transportation safety technology, and in particular to methods, systems, devices, media, and program products for predicting polygonal shapes of train wheels. Background Technology
[0002] As high-speed trains continue to increase in speed, the interaction between wheels and rails becomes increasingly intense, making them more prone to problems such as wheel tread fatigue peeling, abrasion, and polygonal wheel wear. During maintenance, various vehicle operating units have found that polygonal wheel wear is prevalent in multiple models of high-speed locomotives. Furthermore, polygonal wheel wear exacerbates wheel-rail interaction, altering the vehicle's dynamic performance and generating high-frequency vibrations and noise during high-speed operation, thus reducing passenger comfort.
[0003] Currently, the main methods for judging the future working condition of wheels rely heavily on the work experience of maintenance personnel. However, this method is highly subjective and cannot meet the requirements for high-precision data prediction. Furthermore, research on wheel polygonal wear and its formation mechanism is still limited. Various factors can lead to polygonal wear on wheels, and maintenance departments may not be able to perform timely turning operations on the corresponding wheels. This results in a significant problem of condition-based maintenance, which greatly reduces the service life of wheels and increases the operating and maintenance costs of rail transit. At the same time, wheel polygonal wear seriously affects train operation safety. Summary of the Invention
[0004] The purpose of this application is to provide a method, system, device, medium, and program product for predicting the polygonal shape of train wheels. By constructing a polygonal prediction model for train wheels that predicts wear, the accuracy of prediction can be significantly improved.
[0005] To address the aforementioned technical problems, this application provides a method for predicting the polygon shape of train wheels, the specific technical solution of which is as follows:
[0006] Historical wheel data of locomotives and trains are acquired, and data samples corresponding to the historical wheel data are constructed. The historical wheel data includes turning data of wheel turning machine tools, detection data of locomotive wheelset fault dynamic detection system, and daily mileage data of trains.
[0007] The data sample is divided into sets of influencing factors, and correlation analysis is performed on these sets of influencing factors to determine characteristic indicators whose correlation coefficients meet a set threshold.
[0008] The feature indicators are used as training data, and the training data is randomly sampled to generate a training set;
[0009] The training set is trained using the random forest algorithm to obtain a train wheel polygon prediction model;
[0010] Obtain the wheel polygon feature vector of the train to be predicted, input the wheel polygon feature vector into the train wheel polygon prediction model, and output the wheel polygon prediction result of the train to be predicted.
[0011] Optionally, acquiring historical wheel data of locomotives and trains, and constructing data samples corresponding to the historical wheel data includes:
[0012] Acquire historical wheel data of locomotives and trains, and determine the continuous features, nominal features, and target variables in the historical wheel data;
[0013] The continuous features include: wheel diameter, wheel flange thickness machining deviation, predicted maximum value of the wheel pair polygon before the last turning, wheel diameter machining deviation, total vehicle mileage, and cumulative mileage since the last turning repair; the nominal features include carriage type, axle number, and season; the target variable is the presence or absence of wheel polygon faults, and the features are divided into training and testing sets for the random forest model according to a preset ratio.
[0014] Optionally, a correlation analysis is performed on the set of influencing factors to determine the characteristic indicators whose correlation coefficients meet a set threshold, including:
[0015] Pearson correlation analysis was used to analyze the correlation between the continuous features of the influencing factor set and the target variable to determine the continuous features whose correlation coefficients meet a set threshold.
[0016] Optionally, the training set is trained using the random forest algorithm to obtain a train wheel polygon prediction model, including:
[0017] Multiple random sample sets are generated by randomly extracting feature information from the training set.
[0018] A random attribute selection method is used to split the nodes of the training subset to construct a single CART regression tree;
[0019] The above steps are repeated to generate several CART regression trees. Each CART regression tree is allowed to grow freely to obtain a train wheel polygon prediction model composed of all the CART regression trees. The average output value of all the CART regression trees is used as the output result of the train wheel polygon prediction model.
[0020] Optionally, randomly extracting feature information from the training set to generate multiple random sample sets includes:
[0021] Determine the number of feature information in the training set;
[0022] A set number of samples are randomly drawn with replacement from the training set to form a bootstrap sample;
[0023] For each bootstrap sample, calculate the statistic of interest and obtain the empirical distribution of the statistic;
[0024] A training subset is generated based on the empirical distribution.
[0025] Optionally, after training the training set based on the random forest algorithm to obtain the train wheel polygon prediction model, the method further includes:
[0026] The feature information of the wheel polygon feature vector is randomly sampled multiple times to obtain a sample set; wherein, after each random sampling, the sampled feature information is put back into the test set for the next random sampling.
[0027] A prediction model is constructed based on the sample set, and the accuracy score of the prediction model is evaluated.
[0028] Add the feature information corresponding to the prediction model with the highest accuracy score to the optimal feature subset;
[0029] The train wheel polygon prediction model is optimized based on the optimal feature subset.
[0030] This application also provides a train wheel polygon prediction system, including:
[0031] The data acquisition module is used to acquire historical wheel data of locomotives and trains and construct data samples corresponding to the historical wheel data; the historical wheel data includes turning data of wheel turning machine tools, detection data of locomotive wheelset fault dynamic detection system and daily mileage data of trains;
[0032] The feature analysis module is used to divide the data sample into sets of influencing factors and perform correlation analysis on the sets of influencing factors to determine feature indicators whose correlation coefficients meet a set threshold.
[0033] The training data sampling module is used to randomly sample the training data, using the feature indicators as training data, to generate a training set.
[0034] The model training module is used to train the training set based on the random forest algorithm to obtain a train wheel polygon prediction model.
[0035] The wheel prediction module is used to obtain the wheel polygon feature vector of the train to be predicted, input the wheel polygon feature vector into the train wheel polygon prediction model, and output the wheel polygon prediction result of the train to be predicted.
[0036] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the train wheel polygon prediction method as described above.
[0037] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor, when calling the computer program in the memory, implements the steps of the train wheel polygon prediction method as described above.
[0038] This application also provides a computer program product, including a computer program that, when executed, implements the steps of the train wheel polygon prediction method as described above.
[0039] This application provides a method for predicting the polygonal shape of train wheels, comprising: acquiring historical wheel data of locomotives and trains, and constructing data samples corresponding to the historical wheel data; the historical wheel data includes turning data of a turning machine tool, detection data of a dynamic detection system for locomotive wheelset faults, and daily mileage data of the train; dividing the data samples into a set of influencing factors, performing correlation analysis on the set of influencing factors to determine feature indicators whose correlation coefficients meet a set threshold; using the feature indicators as training data, randomly sampling the training data to generate a training set; training the training set based on a random forest algorithm to obtain a train wheel polygonal prediction model; acquiring the wheel polygonal feature vector of the train to be predicted, inputting the wheel polygonal feature vector into the train wheel polygonal prediction model, and outputting the wheel polygonal prediction result of the train to be predicted.
[0040] This application constructs data samples using turning machine tool data, locomotive wheelset fault dynamic detection system data, and daily train mileage data. Feature indicators such as wheel radius, polygon amplitude of the wheelset before the last turning, and wheel diameter machining deviation are extracted as feature vectors for the training samples. Next, a training set is generated through random sampling, and the training set is trained using a random forest algorithm to obtain a train wheel polygon prediction model. The random forest consists of multiple independent decision trees. When predicting a new sample set, the classification of the sample set depends on the judgment of each decision tree. The prediction model itself is composed of multiple classifiers, and the final prediction result is obtained through voting and averaging. This gives the train wheel polygon prediction model good generalization performance, effectively predicting wheel polygon states and providing important reference for train maintenance.
[0041] This application also provides a train wheel polygon prediction system, device, medium, and program product, which have the above-mentioned beneficial effects, and will not be repeated here. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0043] Figure 1 A flowchart illustrating a method for predicting polygonal shapes of train wheels provided in an embodiment of this application;
[0044] Figure 2 A schematic diagram illustrating the training process of the random forest prediction model provided in this application embodiment;
[0045] Figure 3 This is a schematic diagram of a train wheel polygon prediction system provided in an embodiment of this application;
[0046] Figure 4 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation
[0047] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0048] The object information involved in this application, including but not limited to object device information, object personal information, and data, including but not limited to data used for analysis, stored data, and displayed data, are all information and data authorized by the object or fully authorized by all parties, and the collection, use and processing of related data shall comply with the laws, regulations and standards of relevant countries and regions.
[0049] See Figure 1 , Figure 1 A flowchart of a train wheel polygon prediction method provided in this application embodiment, the method including:
[0050] S101: Acquire historical wheel data of locomotives and trains, and construct data samples corresponding to the historical wheel data; the historical wheel data includes turning data of wheel turning machine tools, detection data of locomotive wheelset fault dynamic detection system, and daily mileage data of trains;
[0051] S102: Divide the data sample into sets of influencing factors, and perform correlation analysis on the sets of influencing factors to determine characteristic indicators whose correlation coefficients meet a set threshold.
[0052] S103: Using the feature indicators as training data, randomly sample the training data to generate a training set;
[0053] S104: Train the training set based on the random forest algorithm to obtain a train wheel polygon prediction model;
[0054] S105: Obtain the wheel polygon feature vector of the train to be predicted, input the wheel polygon feature vector into the train wheel polygon prediction model, and output the wheel polygon prediction result of the train to be predicted.
[0055] First, historical wheel data of the locomotive and train is obtained. The locomotive and train mentioned in this application refers to railway transportation equipment consisting of a locomotive and carriages. The time span for the historical wheel data is not limited here; for example, data from the locomotive and train within the past two years can be selected. This historical wheel data mainly includes turning data from a wheel turning machine, detection data from a dynamic detection system for locomotive wheelset faults, and data samples constructed from daily train mileage data. It may also include other wheel data, including basic wheel data such as wheel diameter and material.
[0056] Specifically, the selected historical wheel data can be divided into continuous features and nominal features. Continuous features include: wheel diameter, flange thickness machining deviation, predicted maximum polygon value of the wheel pair before the last turning, wheel diameter machining deviation, total vehicle mileage, and cumulative mileage since the last turning. Nominal features include: carriage type, axle number, and season. The target variable for prediction is the presence or absence of wheel polygon faults.
[0057] Subsequently, the historical wheel data was divided into a set of influencing factors. Based on correlation analysis, feature indicators with high correlation coefficients were selected: wheel radius, polygon amplitude of the wheel pair before last turning, wheel flange thickness machining deviation, wheel diameter machining deviation, and mileage as feature vectors for training samples. Training data was obtained by random sampling as the training set.
[0058] Specifically, for multiple features, Pearson correlation analysis can be used to analyze the correlation between each feature and the magnitude of higher-order polygons. Pearson correlation analysis can quantitatively reflect the linear relationship between two sets of data, and its calculation formula is as follows:
[0059] ;
[0060] If ρ xy If the calculation result is positive, then variables X and Y are positively correlated; if the calculation result is negative, then variables X and Y are negatively correlated.xy The closer the absolute value is to 1, the stronger the correlation between the variables; the closer it is to 0, the weaker the correlation between the variables. When ρ xy When the absolute value ranges from [0.8, 1], variables X and Y are highly correlated; when the range is [0.6, 0.8), they are strongly correlated; when the range is [0.4, 0.6], they are moderately correlated; when the range is [0.2, 0.4), they are weakly correlated; and when the range is [0, 0.2), they are very weakly correlated or uncorrelated. In application, one variable should be the wheelset polygon amplitude, and the other variable should be a single feature from historical wheel data. A single feature refers to a feature in historical wheel data that can independently affect the wheelset polygon amplitude.
[0061] Calculations show that the absolute values of the correlation between the polygon amplitude and wheel radius of the previous wheel alignment and the polygon amplitude of the current alignment are all above 0.4, indicating a moderate degree of correlation. The machining deviations of the wheel flange thickness and wheel diameter show a weak correlation with the polygon amplitudes of the current alignment.
[0062] For example, feature indicators with high correlation coefficients can be selected as training data, such as wheel radius, polygon amplitude of the wheel pair before last turning, wheel flange thickness machining deviation, wheel diameter machining deviation, and mileage. Random sampling can be used to obtain training data as the training set.
[0063] Subsequently, the training set was trained using the random forest algorithm to obtain a train wheel polygon prediction model. Random forest is a machine learning algorithm commonly used in various academic research fields. It combines the weak model CART (Classification and Regression Trees), the bagging algorithm, and random feature subspaces. CART is suitable not only for regression prediction but also for data classification prediction; therefore, random forest can also be used for both types of analysis, and its basic structure shows good results in regression prediction problems. Figure 2 This is a schematic diagram of the training process of the random forest prediction model provided in the embodiments of this application, as shown below. Figure 2 As shown, training the prediction model includes the following steps:
[0064] Step 1: Randomly extract feature information from the training set to generate multiple random sample sets;
[0065] This step can be achieved using other sampling methods, such as the Bootstrap self-sampling method.
[0066] The second step is to generate CART regression trees for multiple random sample sets based on the classification regression algorithm, and to calculate the Gini coefficients for multiple random sample sets.
[0067] The third step is to calculate the Gini coefficient of each feature in the random sample set where the current node of the classification and regression tree is located, if the Gini coefficient of the random sample set is less than the Gini coefficient threshold, the classification and regression algorithm returns to the sub-decision tree and stops recursion; or, if the Gini coefficient of the random sample set is not less than the Gini coefficient threshold, calculate the Gini coefficient of each feature in the random sample set where the current node is located.
[0068] Fourth step: Take the feature information corresponding to the smallest Gini coefficient among the Gini coefficients of each feature information as the optimal feature information, divide the corresponding random sample set into the first feature set and the second feature set based on the optimal feature information, and take the first feature set and the second feature set as the left child node and the right child node of the current node, respectively.
[0069] Step 5: Take the average of the output values of the T CART regression trees as the output result H(x) of the random forest.
[0070] The method of random sampling is not limited here; one feasible implementation is the Bootstrap sampling method. Bootstrap sampling is a statistical method used to estimate the distribution of a statistic or to perform hypothesis testing. It generates multiple new datasets by repeatedly sampling (with replacement) from the original dataset, then calculates the statistic for each new dataset, thereby obtaining an estimate of the distribution of the original statistic. When applying Bootstrap sampling, the number of Bootstrap iterations to be performed needs to be determined (typically between several hundred and several thousand). Samples are randomly drawn from the original dataset (with replacement), with the same number of samples drawn each time as in the original dataset. This process is repeated multiple times according to the set number of iterations. For each resampled dataset, the statistic of interest (such as mean, median, variance, etc.) is calculated.
[0071] When constructing a single CART regression tree, the optimal splitting attribute can be selected at each node by calculating different splitting criteria (such as mean squared error, Gini impurity, etc.). Based on the selected attribute and threshold, several training subsets are generated from the feature information. This process is repeated for each training subset until a stopping condition is met (such as reaching the maximum depth, minimum number of samples, or the mean squared error of the node falling below a certain threshold). When further splitting is impossible, the current node is designated as a leaf node, resulting in a single CART regression tree. At this point, its predicted values (such as mean, median, etc.) can be calculated.
[0072] Since the Random Forest algorithm is an ensemble learning artificial intelligence algorithm, it consists of many classifiers. These classifiers vote or average the final output, thus improving the accuracy and generalization performance of the Random Forest algorithm. Therefore, multiple classifiers are used to learn and classify multiple decision trees to obtain the train wheel polygon prediction model. The output of the Random Forest regression is the average value H(x) of the output values of T CART regression trees.
[0073] ;
[0074] After training the training set using the random forest algorithm to obtain the train wheel polygon prediction model, the train wheel polygon prediction model can be optimized. The specific process is as follows:
[0075] The first step is to randomly sample the feature information of the wheel polygon feature vector test set multiple times to obtain a sample set; wherein, after each random sampling, the sampled feature information is put back into the test set for the next random sampling.
[0076] The second step is to generate multiple weight values for the sampled sample set based on multiple decision trees;
[0077] The third step is to calculate the sum of multiple weight values, divide the sum of multiple weight values by the number of random samplings, and obtain the prediction result of the wheel polygon. The feature information of the test set is randomly sampled multiple times to obtain the sampled sample set.
[0078] After each random sampling, the sampled feature information is put back into the test set for the next random sampling. Multiple weight values are generated based on multiple decision trees for the sampled sample set. The sum of the multiple weight values is then calculated. The sum of the multiple weight values is divided by the number of random samplings to obtain the prediction result of the wheel polygon.
[0079] See Figure 3 , Figure 3 This application provides a schematic diagram of a train wheel polygon prediction system according to an embodiment of the present application. The present application also provides a train wheel polygon prediction system, comprising:
[0080] The data acquisition module is used to acquire historical wheel data of locomotives and trains and construct data samples corresponding to the historical wheel data; the historical wheel data includes turning data of wheel turning machine tools, detection data of locomotive wheelset fault dynamic detection system and daily mileage data of trains;
[0081] The feature analysis module is used to divide the data sample into sets of influencing factors and perform correlation analysis on the sets of influencing factors to determine feature indicators whose correlation coefficients meet a set threshold.
[0082] The training data sampling module is used to use the feature indicators as training data and randomly sample the training data to generate a training sample set.
[0083] The model training module is used to train the training set based on the random forest algorithm to obtain a train wheel polygon prediction model.
[0084] The wheel prediction module is used to obtain the wheel polygon feature vector of the train to be predicted, input the wheel polygon feature vector into the train wheel polygon prediction model, and output the wheel polygon prediction result of the train to be predicted.
[0085] Based on the above embodiments, as a preferred embodiment, the data acquisition module is used to acquire historical wheel data of locomotives and trains, and determine the continuous features, nominal features, and target variables in the historical wheel data; wherein, the continuous features include: wheel diameter, flange thickness machining deviation, predicted maximum value of the wheelset polygon before the last turning, wheel diameter machining deviation, total vehicle mileage, and cumulative mileage since the last turning repair; the nominal features include car type, axle number, and season; the target variable is the presence or absence of wheel polygon faults;
[0086] Based on the above embodiments, as a preferred embodiment, the feature analysis module is a module used to perform correlation analysis on the continuous features in the set of influencing factors and the target variable using the Pearson correlation analysis method, so as to determine the continuous features whose correlation coefficients meet a set threshold.
[0087] Based on the above embodiments, as a preferred embodiment, the model training module includes:
[0088] A feature extraction unit is used to randomly extract feature information from the training set to generate a training subset.
[0089] The regression tree generation unit is used to split the training subset into nodes by adopting a random attribute selection method to construct a single CART regression tree.
[0090] The model generation module is used to repeatedly execute the above steps to generate several CART regression trees, and each CART regression tree is allowed to grow freely to obtain a train wheel polygon prediction model composed of all the CART regression trees; wherein, the average output value of all the CART regression trees is used as the output result of the train wheel polygon prediction model.
[0091] Based on the above embodiments, as a preferred embodiment, the feature extraction unit is a unit for performing the following steps:
[0092] Determine the number of feature information in the training set;
[0093] A set number of samples are randomly drawn with replacement from the training set to form a bootstrap sample;
[0094] For each bootstrap sample, calculate the statistic of interest and obtain the empirical distribution of the statistic;
[0095] A training subset is generated based on the empirical distribution.
[0096] Based on the above embodiments, as a preferred embodiment, it further includes:
[0097] The model optimization module is used to perform the following steps:
[0098] The feature information of the wheel polygon feature vector is randomly sampled multiple times to obtain a sample set; wherein, after each random sampling, the sampled feature information is put back into the test set for the next random sampling.
[0099] A prediction model is constructed based on the sample set, and the accuracy score of the prediction model is evaluated.
[0100] Add the feature information corresponding to the prediction model with the highest accuracy score to the optimal feature subset;
[0101] The train wheel polygon prediction model is optimized based on the optimal feature subset.
[0102] This application also provides a computer program product, including a computer program that, when executed, implements the steps of the train wheel polygon prediction method as described in the above embodiments.
[0103] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed, can perform the steps provided in the above embodiments. The storage medium may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0104] This application also provides an electronic device, see [link to document]. Figure 4 The present application provides a structural diagram of an electronic device, such as... Figure 4 As shown, it may include a processor 1410 and a memory 1420.
[0105] The processor 1410 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 1410 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1410 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 1410 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 1410 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.
[0106] The memory 1420 may include one or more computer-readable storage media, which may be non-transitory. The memory 1420 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 1420 is used to store at least the following computer program 1421, which, after being loaded and executed by the processor 1410, is capable of implementing the relevant steps in the methods executed by the electronic device side as disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 1420 may also include an operating system 1422 and data 1423, etc., and the storage method may be temporary storage or permanent storage. The operating system 1422 may include Windows, Linux, Android, etc.
[0107] In some embodiments, the electronic device may further include a display screen 1430, an input / output interface 1440, a communication interface 1450, a sensor 1460, a power supply 1470, and a communication bus 1480.
[0108] certainly, Figure 4 The structure of the electronic device shown does not constitute a limitation on the electronic device in the embodiments of this application. In practical applications, the electronic device may include more than [other components]. Figure 4 More or fewer components as shown, or combinations of certain components.
[0109] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. As the system provided in the embodiments corresponds to the method provided in the embodiments, the description is relatively simple; relevant parts can be found in the method section.
[0110] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.
[0111] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
Claims
1. A method for predicting polygonal shapes of train wheels, characterized in that, include: Historical wheel data of locomotives and trains are acquired, and data samples corresponding to the historical wheel data are constructed. The historical wheel data includes turning data of wheel turning machine tools, detection data of locomotive wheelset fault dynamic detection system, and daily mileage data of trains. The data sample is divided into sets of influencing factors, and correlation analysis is performed on these sets of influencing factors to determine characteristic indicators whose correlation coefficients meet a set threshold. The feature indicators are used as training data, and the training data is randomly sampled to generate a training sample set; The training set is trained using the random forest algorithm to obtain a train wheel polygon prediction model; Obtain the wheel polygon feature vector of the train to be predicted, input the wheel polygon feature vector into the train wheel polygon prediction model, and output the wheel polygon prediction result of the train to be predicted.
2. The train wheel polygon prediction method according to claim 1, characterized in that, Acquire historical wheel data of locomotives and trains, and construct data samples corresponding to the historical wheel data, including: Acquire historical wheel data of locomotives and trains, and determine the continuous features, nominal features, and target variables in the historical wheel data; The continuous features include: wheel diameter, wheel flange thickness machining deviation, predicted maximum value of wheel pair polygon before last turning, wheel diameter machining deviation, total vehicle mileage, and cumulative mileage since last turning repair; the nominal features include carriage type, axle number, and season; and the target variable is the presence or absence of wheel polygon faults.
3. The train wheel polygon prediction method according to claim 2, characterized in that, Correlation analysis is performed on the set of influencing factors to determine the characteristic indicators whose correlation coefficients meet a set threshold, including: Pearson correlation analysis was used to analyze the correlation between the continuous features of the influencing factor set and the target variable to determine the continuous features whose correlation coefficients meet a set threshold.
4. The train wheel polygon prediction method according to claim 1, characterized in that, The training set is trained using the random forest algorithm to obtain a train wheel polygon prediction model, including: Multiple random sample sets are generated by randomly extracting feature information from the training set. CART regression trees for multiple random sample sets are generated based on the classification regression algorithm, and the Gini coefficients for multiple random sample sets are calculated. For multiple random sample sets where the current node of the classification and regression tree is located, if the Gini coefficient of the multiple random sample sets is less than the Gini coefficient threshold, the classification and regression algorithm returns to the sub-decision tree and stops recursion; or, if the Gini coefficient of the multiple random sample sets is not less than the Gini coefficient threshold, the Gini coefficient of each feature information in the random sample set where the current node is located is calculated. The feature information corresponding to the smallest Gini coefficient among the Gini coefficients of each feature information is taken as the optimal feature information. The corresponding random sample set is divided into a first feature set and a second feature set based on the optimal feature information, and the first feature set and the second feature set are respectively taken as the left child node and the right child node of the current node. Calculate the Gini coefficients of the first feature set and the second feature set. When the Gini coefficients of the first feature set and the second feature set are less than the Gini coefficient threshold, the classification and regression algorithm returns to the sub-decision tree and stops recursion, generating T CART regression trees and outputting the learning results. The average of the output values of the T CART regression trees is used as the output result of the random forest.
5. The train wheel polygon prediction method according to claim 4, characterized in that, Randomly extracting feature information from the training set to generate multiple random sample sets includes: Determine the number of feature information in the training set; A set number of samples are randomly drawn with replacement from the training set to form a bootstrap sample; For each bootstrap sample, calculate the statistic of interest and obtain the empirical distribution of the statistic; A training sample set is generated based on the empirical distribution.
6. The train wheel polygon prediction method according to claim 1, characterized in that, After training the training set using the random forest algorithm to obtain the train wheel polygon prediction model, the method further includes: The feature information of the wheel polygon feature vector is randomly sampled multiple times to obtain a sample set; wherein, after each random sampling, the sampled feature information is put back into the test set for the next random sampling. A prediction model is constructed based on the sample set, and the accuracy score of the prediction model is evaluated. Add the feature information corresponding to the prediction model with the highest accuracy score to the optimal feature subset; The train wheel polygon prediction model is optimized based on the optimal feature subset.
7. A train wheel polygon prediction system, characterized in that, include: The data acquisition module is used to acquire historical wheel data of locomotives and trains and construct data samples corresponding to the historical wheel data; the historical wheel data includes turning data of wheel turning machine tools, detection data of locomotive wheelset fault dynamic detection system and daily mileage data of trains; The feature analysis module is used to divide the data sample into sets of influencing factors and perform correlation analysis on the sets of influencing factors to determine feature indicators whose correlation coefficients meet a set threshold. The training data sampling module is used to use the feature indicators as training data and randomly sample the training data to generate a training sample set. The model training module is used to train the training set based on the random forest algorithm to obtain a train wheel polygon prediction model. The wheel prediction module is used to obtain the wheel polygon feature vector of the train to be predicted, input the wheel polygon feature vector into the train wheel polygon prediction model, and output the wheel polygon prediction result of the train to be predicted.
8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the method as claimed in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the steps of the method as described in any one of claims 1 to 6.
10. A computer program product, characterized in that, Includes a computer program, which, when executed, implements the steps of the method as described in any one of claims 1 to 6.