Railway vehicle index analysis method and system

Abstract

The invention discloses a railway vehicle index analysis method and system. The method comprises the steps of collecting and preprocessing fault index data of a railway vehicle; carrying out weighting processing on the truck fault vocabularies by adopting a weight calculation method, extracting potential fault features, adjusting weighted part-of-speech and part-of-frequency through a weighting method, and carrying out weighting and sorting on the fault features; based on the extracted fault features, inputting an LDA model for training; and after fault features are extracted from to-be-analyzed railway vehicle indexes, the trained LDA model is input, theme modeling of the fault features is carried out, and automatic classification of railway vehicle fault index data is realized. Compared with a traditional fault classification method, the method has higher accuracy and efficiency, and can provide powerful guarantee for safe operation of the rail wagon.

Description

Technical Field

[0001] The present application relates to the field of fault classification, and particularly to a method and system for analyzing railway vehicle indicators. Background Art

[0002] Currently, railway vehicles, especially railway freight vehicles, as one of the important modes of logistics transportation, play a crucial role in the development of the national economy. With the continuous growth of railway freight volume, the safe operation of railway freight cars has become particularly important. Fault classification, as an important link in freight car maintenance management, is of great significance for ensuring freight efficiency and reducing operating costs.

[0003] In recent years, with the rapid development of big data and artificial intelligence technologies, machine learning algorithms have been widely applied in the field of fault classification. Among them, the (Latent Dirichlet Allocation) LDA model, as an unsupervised learning algorithm, has achieved remarkable results in the field of text classification. Drawing on the application experience of the LDA model in the field of text classification, it is applied to railway freight car fault classification, aiming to improve the accuracy and efficiency of fault classification.

[0004] In the current existing technologies, there are the following problems in the analysis of fault index data of railway vehicles:

[0005] (1) Lack of effective fault data analysis technology: Since freight cars are affected by various factors during long-term use, such as equipment aging and improper human operation, it is common for freight cars to have faults. How to quickly and accurately identify and classify these faults is of great significance for improving maintenance efficiency and reducing operating costs;

[0006] (2) Dependence on manual analysis, low efficiency: Traditional fault classification methods mainly rely on manual experience and expert knowledge, and have problems such as strong subjectivity and low efficiency.

[0007] Based on the above analysis, there is an urgent need to propose a method and system for analyzing railway vehicle indicators, which extracts potential fault features from the fault dataset and combines the LDA model to perform topic modeling on the fault features to achieve automatic classification of freight car faults. Summary of the Invention

[0008] In order to solve the above problems of lack of effective fault data analysis technology and dependence on manual analysis with low efficiency, a method for analyzing railway vehicle indicators is proposed.

[0009] In a first aspect, an embodiment of the present application provides a method for analyzing railway vehicle indicators, the method including:

[0010] Data preprocessing step: Collect the fault index data of railway vehicles, and preprocess the fault index data of railway vehicles by using artificial intelligence-based data conversion methods, data normalization methods, and dataset partitioning methods to generate various types of preprocessed railway vehicle fault index data;

[0011] Fault feature extraction step: Based on the preprocessed railway vehicle fault index data, use a weight calculation method to weight the fault index data, extract potential fault features, and according to the word frequency distribution characteristics of the potential fault features, adjust the part-of-speech word frequency after weighting through the weighting method, and perform weighted sum and ranking of the fault features;

[0012] LDA model training step: Based on the extracted fault features, input them into the LDA model for training, and continuously adjust the parameters of the LDA model through iterative optimization to achieve better fitting parameters;

[0013] Fault index analysis step: After extracting the fault features from the railway vehicle index to be analyzed, input them into the trained LDA model to perform topic modeling of the fault features, and realize the automatic classification of the railway vehicle fault index data.

[0014] In a specific embodiment of the present invention, the above data preprocessing step includes:

[0015] Data conversion step: Use feature scaling methods, feature encoding methods, and feature selection methods to convert the fault index data of railway vehicles into a unified railway data format;

[0016] Data normalization step: Use a normalization method to scale the fault index data of railway vehicles in the converted format to a preset range;

[0017] Dataset partitioning step: Use random abstraction and stratified sampling methods to partition the fault index data of railway vehicles after data normalization into training set, validation set, and test set data.

[0018] In a specific embodiment of the present invention, the above fault feature extraction step includes:

[0019] Fault index data feature extraction step: After weighting the preprocessed fault index data of railway vehicles by using the TF-IDF weight calculation method, extract the fault index data features;

[0020] Weight adjustment step: According to the distribution characteristics of the word frequency of the fault index data, adjust the weights of the words through the SIF method to balance the influence of common words and rare words in expressing the text, and then rank according to the word frequency.

[0021] In a specific embodiment of the present invention, the above LDA model training step includes:

[0022] Initialize and set the LDA model parameters, where the model parameters include: the number of topics K, the learning rate α, and β;

[0023] Use the extracted fault features as training data and input them into the LDA model for training;

[0024] During the training process, the LDA model automatically discovers the latent topics in the training data and represents the training data as a mixture of latent topics, completing the topic modeling of the fault features;

[0025] Through the iterative optimization algorithm, continuously adjust the LDA model parameters so that the model can better fit the input data;

[0026] When the model converges or reaches the preset number of iterations, stop the training.

[0027] In the specific embodiments of the present invention, the above-mentioned fault index analysis steps include:

[0028] The LDA model describes the topic structure in the document collection through the probability relationships among these three levels of random variables;

[0029] The three levels of random variables include: the document level, the topic level, and the vocabulary level; among them, the random variable at the document level represents each document in the document collection, the random variable at the topic level represents each topic in the document, and the random variable at the vocabulary level represents each vocabulary in the topic.

[0030] In the specific embodiments of the present invention, the above-mentioned fault index analysis steps include:

[0031] Perform preprocessing and feature extraction operations on the fault data to be analyzed, and the feature extraction obtains a feature vector in the same format as the training data; among them, the format of the feature vector is the faults that occur in the components of the vehicle model;

[0032] Input the fault data to be analyzed into the trained LDA model, and calculate the probability distribution of the fault data belonging to each topic;

[0033] According to the probability distribution result, divide the fault data into the topic category with the highest probability to complete the fault classification.

[0034] In the specific embodiments of the present invention, the above-mentioned document level processing includes: selecting a specific topic for a given document, and the topic distribution of each document in the LDA model follows the Dirichlet prior method. Among them, the hyperparameter of the prior method affects the sparsity of the topic distribution. According to the value of the hyperparameter, determine whether the topic distribution tendency of the document covers multiple topics;

[0035] For the probability distribution of the topic, in the associated scenario allocation model LDA of the Dirichlet distribution method and the multinomial distribution, determine the distribution probability of the topic;

[0036] The sum of the topic probability distributions uses the Dirichlet distribution in the Latent Dirichlet Allocation (LDA) topic model to characterize the prior situation of the probability distributions between documents and topics, and between topics and terms.

[0037] Based on different word types (verbs, nouns, idiomatic expressions, etc.), the co-occurrence times, co-occurrence values, frequencies, total frequencies, and number of occurrences of each word are obtained. Using the tf-idf weighting method, a multi-topic fault dataset with information such as the correlation, imbalance, average distance, and minimum distance of each word type is obtained.

[0038] In a second aspect, an embodiment of the present application provides a method for analyzing railway vehicle indicators, which uses the above-mentioned railway vehicle indicator analysis system. The system includes:

[0039] Data preprocessing module: Collect fault indicator data of railway vehicles, and preprocess the fault indicator data of railway vehicles using artificial intelligence-based data conversion methods, data normalization methods, and dataset partitioning methods to generate various types of preprocessed railway vehicle fault indicator data.

[0040] Fault feature extraction module: Based on the preprocessed railway vehicle fault indicator data, use a weight calculation method to weight the freight vehicle fault vocabulary, extract potential fault features, and according to the word frequency distribution characteristics of the potential fault features, adjust the part-of-speech word frequency after weighting through a weighting method, and perform weighted sum and sorting of the fault features.

[0041] LDA model training module: Based on the extracted fault features, input them into the LDA model for training, and continuously adjust the parameters of the LDA model through iterative optimization to achieve better fitting parameters.

[0042] Fault indicator analysis module: After extracting the fault features from the railway vehicle indicators to be analyzed, input them into the trained LDA model for topic modeling of the fault features, and realize automatic classification of the railway vehicle fault indicator data.

[0043] In a third aspect, an embodiment of the present application provides an electronic device, which includes: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the above-mentioned railway vehicle indicator analysis method.

[0044] In a fourth aspect, an embodiment of the present application provides a computer storage medium, which stores computer instructions that, when called, are used to execute the above-mentioned railway vehicle indicator analysis method.

[0045] Compared with the related prior art, it has the following outstanding beneficial effects:

[0046] 1) The method of the present invention extracts potential fault features from the fault dataset and performs topic modeling on these features in combination with the LDA model, realizing the automatic classification of freight car faults. Compared with traditional fault classification methods, this method has higher accuracy and efficiency, and can provide strong guarantee for the safe operation of railway freight cars;

[0047] 2) The experimental results show that the freight car fault classification method of the railway vehicle index analysis method and its system of the present invention has a high classification accuracy rate. On the test set, the accuracy rate of the model reaches more than 90%;

[0048] 3) The method of the present invention has strong generalization ability and versatility, and can be applied to the fault classification tasks of different railway vehicle models and components. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation of the present application. In the drawings:

[0050] Figure 1 It is a schematic diagram of the railway vehicle index analysis method of the present invention;

[0051] Figure 2 It is a schematic diagram of the railway vehicle index analysis method in the specific embodiment of the present invention;

[0052] Figure 3a It is a schematic diagram of the data with adjectives as the theme in the specific embodiment of the present invention;

[0053] Figure 3b It is a schematic diagram of the data with railway freight car models as the theme in the specific embodiment of the present invention;

[0054] Figure 4 It is a schematic diagram of the TF-IDF equal weight calculation formula of the railway freight car fault data in the specific embodiment of the present invention;

[0055] Figure 5 It is a schematic diagram of the noun / word frequency ranking in the specific embodiment of the present invention;

[0056] Figure 6 It is a schematic diagram of the LDA model calculation method for the railway freight car component fault index data in the specific embodiment of the present invention;

[0057] Figure 7 It is a schematic diagram of the fault classification probability distribution result graph in the specific embodiment of the present invention;

[0058] Figure 8 It is a schematic diagram of the multi-topic fault dataset in the specific embodiment of the present invention;

[0059] Figure 9Schematic diagram of the distribution ratio of crossbeam words of specific embodiments of the present invention in each theme;

[0060] Figure 10 Schematic diagram of the theme presented by the word adjacent spacing of specific embodiments of the present invention;

[0061] Figure 11 Schematic diagram of the associated diagram of the crossbeam and faulty parts of specific embodiments of the present invention;

[0062] Figure 12 Schematic diagram of the railway vehicle index analysis system according to the embodiments of the present invention;

[0063] Figure 13 Schematic diagram of the computer hardware of the present invention. Specific implementation manners

[0064] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0065] The terms "first", "second", etc. in the specification, claims and above-mentioned drawings of the present invention are used to distinguish different objects, rather than to describe a specific order. In addition, the term " / and" in this article is only a description of the associated relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. These three situations, where A and B can be singular or plural. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after, but it may also represent an " / and" relationship, which can be specifically understood with reference to the context.

[0066] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one (item)" or its similar expression means any combination of these items, including any combination of single (item) or plural items (items). For example, at least one (item) of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, c can be single or multiple.

[0067] It should also be understood that in various embodiments of the present invention, the order numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0068] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0069] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0070] In addition, in each embodiment of the present invention, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0071] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or this part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, and other media that can store program codes.

[0072] To make the above features and effects of the present invention more clearly understandable, the following specific examples are given and detailed descriptions are made in conjunction with the accompanying drawings of the specification as follows. This specification discloses one or more embodiments incorporating the features of the present invention. The disclosed embodiments are only for illustrative purposes. The protection scope of the present invention is not limited to the disclosed embodiments, and the present invention is defined by the appended claims.

[0073] The following is a system embodiment corresponding to the above method embodiment, and this embodiment can be implemented in cooperation with the above embodiment. The relevant technical details mentioned in the above embodiment are still valid in this embodiment. To avoid repetition, they will not be elaborated here. Correspondingly, the relevant technical details mentioned in this embodiment can also be applied to the above embodiment.

[0074] The present invention aims to provide a method and system for analyzing railway vehicle indicators. By extracting potential fault features from a fault dataset and combining with the LDA model for topic modeling of fault features, automatic classification of freight car faults is achieved.

[0075] Experimental results show that the freight car fault classification method of the method and system for analyzing railway vehicle indicators proposed by the present invention has a high classification accuracy. On the test set, the accuracy of the model reaches more than 90%. At the same time, this method has strong generalization ability and can be applied to fault classification tasks of different vehicle types and components.

[0076] Embodiment 1

[0077] Please refer to Figure 1 , Figure 1 which is a method for analyzing railway vehicle indicators of the present invention. It collects fault data of the maintenance workshop of the freight car depot, extracts potential fault features, combines with the LDA model for topic modeling of fault features, and realizes automatic classification of freight car faults. The specific process steps include:

[0078] Data preprocessing step 101: Collect fault index data of railway vehicles, and preprocess the fault index data of railway vehicles by using an artificial intelligence-based data conversion method, a data normalization method, and a dataset partitioning method to generate various types of preprocessed railway vehicle fault index data;

[0079] Fault feature extraction step 102: Based on the preprocessed railway vehicle fault index data, use a weight calculation method to weight the fault index data, extract potential fault features, and according to the word frequency distribution characteristics of the potential fault features, adjust the part-of-speech word frequency after weighting by the weighting method to perform weighted sum and sorting of fault features;

[0080] LDA model training step 103: Based on the extracted fault features, input them into the LDA model for training, and continuously adjust the parameters of the LDA model through iterative optimization to achieve better fitting parameters;

[0081] Fault index analysis step 104: After extracting fault features from the railway vehicle indicators to be analyzed, input them into the trained LDA model for topic modeling of fault features to achieve automatic classification of railway vehicle fault index data.

[0082] In a specific embodiment of the present invention, such asFigure 2 As shown in Figure 2 , the above data preprocessing step 101 includes:

[0083] Data conversion step: Using feature scaling method, feature encoding method and feature selection method, convert the fault index data of railway vehicles into a unified railway data format;

[0084] Data normalization step: Using a normalization method, scale the fault index data of railway vehicles in the converted format to a preset range;

[0085] Dataset division step: Using random abstraction and stratified sampling methods, divide the fault index data of railway vehicles after data normalization into training set, validation set and test set data.

[0086] Specifically, the preliminary processing of the fault index data of railway freight cars in the specific embodiment of the present invention includes:

[0087] 1) Before using the LDA model for railway freight car fault classification, it is necessary to perform preliminary processing on the original fault data. The preliminary processing steps mainly include data cleaning, word segmentation, stop word removal, etc. Data cleaning is to check the original fault data and delete duplicate, incorrect or irrelevant data records. Word segmentation is to split each fault description text into several lexical units for subsequent feature extraction and model training. Stop word removal is to remove common words, function words and other meaningless words in the text to reduce the model complexity.

[0088] 2) Preprocess the fault data of railway freight car components by AI artificial intelligence method

[0089] The preprocessing of the fault data of railway freight car components by AI artificial intelligence consists of several parts such as data conversion, data normalization, and dataset division.

[0090] The data conversion of the fault data of railway freight car components refers to converting the original data into a form more suitable for machine learning algorithms, which can be carried out in the following ways: feature scaling, feature encoding and feature selection

[0091] The normalization of the fault data of railway freight car components is to scale the data to a specific range so that they can be processed by machine learning algorithms, and it is carried out in the following ways: min-max normalization and Z-score normalization;

[0092] Min-max normalization: Min-max normalization is to scale the data to a range between 0 and 1. This can maintain the relative size relationship of the data.

[0093] Z-score normalization: Z-score normalization is to scale the data to a range with a mean of 0 and a standard deviation of 1. This can make the data distribution more normal so that they can be processed by machine learning algorithms.

[0094] The division of the fault data dataset of railway freight car components is the process of dividing the original dataset into a training set, a validation set, and a test set, which is to evaluate the performance and accuracy of machine learning models and can be carried out in the following ways:

[0095] Random sampling: Random sampling is to randomly select a part of the data from the original dataset as the training set, validation set, and test set.

[0096] Stratified sampling: Stratified sampling is to select a certain proportion of data in the original dataset and stratify it according to its characteristics to ensure that the data in the training set, validation set, and test set have a similar characteristic distribution.

[0097] Preprocessing the fault data of railway freight car components through AI artificial intelligence can improve data quality, reduce model complexity, and improve classification effects. For example Figure 3a The classification effects of adjectives such as corrosion perforation and wear Figure 3b represent classification by 4 types of railway freight car models such as C70 and C64K.

[0098] In the specific embodiments of the present invention, the above-mentioned fault feature extraction step 102 includes:

[0099] Fault index data feature extraction step: After weighting the fault index data of the preprocessed railway vehicle by using the TF-IDF weight calculation method, extract the fault index data features;

[0100] Weight adjustment step: According to the distribution characteristics of the word frequencies of the fault index data, adjust the weights of the words through the SIF method to balance the influence of common words and rare words in expressing the text, and then rank according to the word frequencies.

[0101] Specifically, in the specific embodiments of the present invention, the extraction of freight car fault index data features by TF-IDF includes:

[0102] 1) After the preprocessing is completed, it is necessary to extract effective features from the fault description text. Considering that the LDA model is a vocabulary-based model, the vocabulary in the railway freight car fault text can be directly used as features. To improve the effectiveness of the features, the TF-IDF (Term Frequency-Inverse Document Frequency, TF-IDF) weight calculation method can be used to weight the freight car fault vocabulary. The weight calculation formula such as TF-IDF is as Figure 4 shown; the word represents the component type, and the weight, correlation, imbalance, average distance, and minimum distance are obtained by weighting according to the common times, co-occurrence values, times, total times, and number of items of the component type by using the TF-IDF weight calculation method.

[0103] 2) Weight the characteristics of the truck fault index data using the SIF weighting method

[0104] The SIF (Smooth Inverse Frequency) method takes into account the distribution characteristics of word frequencies. By adjusting the weights of words, it balances the influence of common words and rare words in expressing text.

[0105] In this method, the weight calculation formula for a word is a / (a + p(w)), where a is a constant, usually ranging from 1e-3 to 1e-4, and p(w) is the word frequency of the word in the corpus. The SIF method proposes to perform weighted averaging on each word vector and then subtract their first principal component (an application of PCA) from this average value. This usually can remove some common components in the corpus, such as stop words like "is" and "of", and retain the semantic information of the document.

[0106] The part-of-speech and word-frequency analysis after adjusting the weights by SIF is as Figure 5 shown. According to the SIF weight calculation method, the occurrence frequencies (i.e., word frequencies) of railway nouns such as crossbeam, crack, bolt, and perforation are obtained; according to the SIF weight calculation method, the word frequencies of verbs such as wear, corrosion, traction, braking, stress, and overlimit are obtained.

[0107] Among them, Figure 5 the format of the feature vector in the left part of the figure is that the vertical coordinate represents the part type, and the horizontal coordinate represents the occurrence frequency of the vocabulary of this part type; Figure 5 the format of the feature vector in the right part of the figure is that the vertical coordinate represents the fault type, and the horizontal coordinate represents the occurrence frequency of the vocabulary of this fault type.

[0108] In the specific embodiment of the present invention, the above-mentioned LDA model training step 103 includes:

[0109] Initialize and set the LDA model parameters, and the model parameters include: the number of topics K, the learning rate α, and β;

[0110] Take the extracted fault features as training data and input them into the LDA model for training;

[0111] During the training process, the LDA model automatically discovers the latent topics in the training data and represents the training data as a mixture of latent topics, completing the topic modeling of the fault features;

[0112] Through the iterative optimization algorithm, continuously adjust the LDA model parameters so that the model can better fit the input data;

[0113] When the model converges or reaches the preset number of iterations, stop the training.

[0114] Specifically, in the specific embodiments of the present invention, the LDA model training for the fault index data of railway freight car components includes:

[0115] After the extraction and weighting of the text features of freight car faults are completed, the LDA model can be used to train the freight car fault data. The specific steps are as follows:

[0116]

[0117] The formula explanation is as follows: θ d ~Dir(α) indicates that θ d is the topic distribution of document d and follows the Dirichlet distribution with parameter α.

[0118] β k ~Dir(β) indicates that β k is the word distribution of topic k and follows the Dirichlet distribution with parameter β.

[0119] w~θ d (z)×β z (w) indicates that w is the generated word and follows the product distribution of the topic distribution θ d and the word distribution β z . α and β are the concentration parameters of the Dirichlet distribution.

[0120] count(topick,documentd) is the number of times topic k appears in document d. K is the number of topics. This formula describes the probability of generating a word given a document and a topic, and is smoothed by adding a pseudo-count α to avoid the zero-probability problem.

[0121] Set the parameters of the LDA model, including the number of topics K, the learning rates α and β, etc. The selection of these parameters will affect the training effect and classification performance of the model.

[0122] Use the preprocessed fault data as the input data and input it into the LDA model for training. During the training process, the LDA model will automatically discover the latent topics in the document and represent the document as a mixture of these topics.

[0123] Through the iterative optimization algorithm, continuously adjust the parameters of the LDA model so that the model can better fit the input data. When the model converges or reaches the preset number of iterations, stop the training.

[0124] The potential "Latent Dirichlet Allocation (LDA)" adopted by the present invention is a currently the most superior and mature topic model algorithm, which is a data-driven unsupervised learning method. It not only does not rely on the precise mathematical model of the industrial system, but also, compared with other data-driven fault diagnosis methods such as neural networks and support vector machines, it does not require prior knowledge of the data. All data distribution patterns and association rules come from the information implicit in the data itself, rather than being determined by specific meanings (such as fault labels) specified by experience.

[0125] The LDA model is a probabilistic graphical model based on Bayesian theory and is used for topic modeling of a document collection. This model assumes that each document is a mixture of multiple topics, and each topic is generated by a set of words according to a certain probability distribution. By modeling the document collection, the LDA model can automatically discover the latent topics in the documents and represent the documents as a mixture of these topics.

[0126] In a specific embodiment of the present invention, the above-mentioned fault index analysis step 104 includes:

[0127] Specifically, as Figure 6 shown, the LDA model describes the topic structure in the document collection through the probability relationships between these three levels of random variables; in the LDA model, documents, topics, and words respectively correspond to three levels of random variables. Among them, the random variable at the document level represents each document in the document collection, the random variable at the topic level represents each topic in the document, and the random variable at the word level represents each word in the topic. The LDA model describes the topic structure in the document collection through the probability relationships between these three levels of random variables;

[0128] In a specific embodiment of the present invention, the above-mentioned document-level processing includes: selecting a specific topic for a given document. In the LDA model, the topic distribution of each document follows the Dirichlet prior method, where the hyperparameter of the prior method affects the sparsity of the topic distribution. According to the value of the hyperparameter, it is determined whether the topic distribution tendency of the document covers multiple topics;

[0129] The probability distribution of the topic. In the association scenario allocation model LDA of the Dirichlet distribution method and the multinomial distribution, the distribution probability of the topic is determined;

[0130] The total of the topic probability distributions. In the topic model LDA, the Dirichlet distribution is used to characterize the prior situation of the probability distributions between documents and topics, and between topics and terms.

[0131] 1) Selecting a specific topic from given documents: In the LDA model, the topic distribution p(z∣d) of each document follows a Dirichlet prior Dirichlet(α). The hyperparameter α affects the sparsity of the topic distribution: if α takes a small value, the topic distribution of the document tends to concentrate on a few topics; if α takes a large value, the topic distribution will cover multiple topics more evenly. By introducing this prior, LDA can learn the topic structure of the corpus and achieve unsupervised topic modeling. p(z∣d) represents "the distribution probability of topic z given document d", that is, the proportion of each topic in a certain document d.

[0132] In the embodiments of the present invention, when the given documents are "the assembly bolts of the drive shaft hanger of KM70 are not tightened", "the bolster spring at C701 is broken", "the wear plate of the coupler support seat at C64 two positions is welded open", "the side beam at C70E one position is corroded", "the roof plate of C70EH is cracked", the hyperparameter α takes values on three topics such as "vehicle type", "location", and "fault content", and the topic distribution of the document also tends to concentrate on these three topics.

[0133] 2) Probability distribution of topics: In the associated scenario of the Dirichlet distribution and the multinomial distribution (such as the Latent Dirichlet Allocation model LDA), p(w∣z) represents "the probability of generating word w given topic z", and its essence is the parameter of the multinomial distribution. The distribution probability details of the topics "vehicle type", "location", and "fault content" in this article are shown as cylinders in the figure.

[0134] 3) Total of topic probability distributions: "Generating the prior distribution of p = {z∣d}" means generating a prior probability distribution for the variable z that satisfies the condition d. "Dirichlet(α)" refers to the Dirichlet distribution, which is a commonly used multivariate probability distribution in probability theory and statistics and is often used as the prior distribution of the parameters of the multinomial distribution in Bayesian inference. α is the parameter vector of the Dirichlet distribution, and different values of α will determine the specific form of the distribution. For each topic z, p(w∣z) is a probability vector: p(w∣z) = (p(w1|z), p(w2|z), …, p(wV|z)), where V is the size of the vocabulary, and ∑wp(w|z) = 1. For example, in a topic model (such as LDA), the Dirichlet distribution is often used to describe the prior situation of probability distributions such as document-topic and topic-term. The summary graph of the distribution probabilities of the topics "vehicle type", "location", and "fault content" in this article is as Figure 6 shown by the columnar bodies.

[0135] After the document level, topic level, and vocabulary level go through preprocessing and feature extraction operations, the common frequency, co-occurrence value, frequency, total frequency, and number of items of a certain type of vocabulary classified according to vocabulary are obtained. By adding the weighting coefficient tf-idf and the above parameters, the correlation, imbalance ratio, average distance, and minimum value information of a certain type of vocabulary are obtained.

[0136] In a specific embodiment of the present invention, the above-mentioned fault index analysis step 104 further includes:

[0137] As Figure 7 shown, preprocessing and feature extraction operations are performed on the new fault data to be analyzed, and feature vectors in the same format as the training data are obtained through feature extraction; among them, the format of the feature vector is the probability value of a certain vocabulary appearing in every 100 lines (increasing by 300 lines from zero for each node).

[0138] The new fault data to be analyzed is input into the trained LDA model, and the probability distribution of the fault data belonging to each topic is calculated;

[0139] According to the probability distribution result, the fault data is classified into the topic category with the highest probability to complete fault classification.

[0140] In order to verify the effectiveness of the railway vehicle index analysis method and its system, the present invention conducts experimental verification. The experimental data is sourced from the typical fault record data of freight cars in Shijiazhuang Vehicle Depot. After preprocessing and feature extraction operations, a fault data set containing multiple topics is obtained as Figure 8 shown.

[0141] The above figure can illustrate the proportion of the crossbeam as the theme in the word frequencies of other themes. From this, we can obtain the distribution proportion of the crossbeam word in each theme as Figure 9 shown.

[0142] The conclusion is obtained that the crossbeam has a relatively high proportion in the perforation theme, followed by the web, connection with the crossbeam, wing plate, and the situation of foreign objects carried on the crossbeam, etc.

[0143] Based on the average distance and minimum distance, adjacent word data is further generated. The results show that the difference between the topic word frequency proportion and the word adjacent distance is that the word adjacent distance more intuitively shows strongly related topics. As Figure 10 shown.

[0144] The data set is divided into two parts: a training set and a test set. The training set is used to train the LDA model, and the test set is used to verify the classification performance of the model.

[0145] The experimental results show that the fault classification method of a railway vehicle index analysis method and its system has a high classification accuracy. On the test set, the accuracy of the model reached over 90%. At the same time, this method has strong generalization ability and can be applied to the fault classification tasks of different vehicle types and components. For example, for crossbeams, it can accurately associate with faulty parts without any irrelevant data association. Specifically, as Figure 11 shown.

[0146] Example 2

[0147] As Figure 12 shown, an embodiment of the present application provides a railway vehicle index analysis method, which adopts the above-mentioned railway vehicle index analysis system. The system includes:

[0148] Data preprocessing module 201: Collect the fault index data of railway vehicles, and preprocess the fault index data of railway vehicles by using artificial intelligence data conversion methods, data normalization methods and dataset partitioning methods to generate various types of preprocessed railway vehicle fault index data;

[0149] Fault feature extraction module 202: Based on the preprocessed railway vehicle fault index data, use the weight calculation method to weight the freight car fault vocabulary, extract potential fault features, and adjust the weighted part-of-speech frequency according to the word frequency distribution characteristics of the potential fault features, and perform weighted sum and sorting of the fault features;

[0150] LDA model training module 203: Based on the extracted fault features, input them into the LDA model for training, and continuously adjust the parameters of the LDA model through iterative optimization to achieve better fitting parameters;

[0151] Fault index analysis module 204: After extracting the fault features from the railway vehicle index to be analyzed, input them into the trained LDA model to perform topic modeling of the fault features, and realize the automatic classification of the railway vehicle fault index data.

[0152] Example 3

[0153] An embodiment of the present application provides an electronic device, as Figure 13 shown. The device includes: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the railway vehicle index analysis method as described above.

[0154] Example 4

[0155] An embodiment of the present application provides a computer storage medium. The computer storage medium stores computer instructions, which are used to execute the railway vehicle index analysis method as described above when called.

[0156] In addition, the railway vehicle index analysis method according to the embodiments of the present application described in combination with Figure 1 can be implemented by a computer device. Figure 13 FIG. is a schematic hardware structure diagram of a computer device according to an embodiment of the present application.

[0157] In some embodiments, the computer device may further include a communication interface 83 and a bus 80. Among them, as Figure 13 shown, the processor 81, the memory 82, and the communication interface 83 are connected through the bus 80 and complete communication with each other.

[0158] Specifically, the above-mentioned processor 81 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0159] The memory 82 can be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 81.

[0160] The processor 81 reads and executes the computer program instructions stored in the memory 82 to implement any one of the railway vehicle index analysis methods in the above embodiments.

[0161] In summary, the positive technical effects of the present invention compared with the prior art are:

[0162] The present invention is directed to a railway vehicle index analysis method and system thereof, and the effectiveness of the method is verified by experiments. The method realizes automatic classification of truck faults by extracting potential fault features from the fault dataset and performing topic modeling on these features in combination with the LDA model. Compared with traditional fault classification methods, the method has higher accuracy and efficiency, and can provide a strong guarantee for the safe operation of railway trucks.

[0163] In short, a railway vehicle index analysis method and system thereof is an effective index analysis method. Through continuous research and improvement, the method is expected to provide a more intelligent and efficient solution for the safe operation and maintenance management of the railway freight industry.

[0164] The technical features of the above-described embodiments can be combined arbitrarily, and for the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.

[0165] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method for analyzing railway vehicle indicators, characterized in that, The method includes: Data preprocessing step: Collect the fault index data of railway vehicles, and preprocess the fault index data of the railway vehicles by using an artificial intelligence-based data conversion method, a data normalization method, and a data set partitioning method to generate various types of preprocessed railway vehicle fault index data; Fault feature extraction step: Based on the preprocessed fault index data of the railway vehicles, use a weight calculation method to weight the fault index data, extract potential fault features, and adjust the weighted part-of-speech frequency according to the word frequency distribution characteristics of the potential fault features, and perform weighted sum and ranking of the fault features; LDA model training step: Based on the extracted fault features, input them into the LDA model for training, and continuously adjust the parameters of the LDA model through iterative optimization to achieve better fitting parameters; Fault index analysis step: After extracting the fault features from the railway vehicle index to be analyzed, input them into the trained LDA model to perform topic modeling of the fault features, and realize automatic classification of the railway vehicle fault index data.

2. The railway vehicle index analysis method according to claim 1, characterized in that The data preprocessing step includes: Data conversion step: Use a feature scaling method, a feature encoding method, and a feature selection method to convert the fault index data of the railway vehicles into a unified railway data format; Data normalization step: Use a normalization method to scale the fault index data of the railway vehicles in the converted format to a preset range; Data set partitioning step: Use a random abstraction and stratified sampling method to partition the fault index data of the railway vehicles after data normalization into training set, validation set, and test set data.

3. The railway vehicle index analysis method according to claim 1, wherein The fault feature extraction step includes: Fault index data feature extraction step: After weighting the preprocessed fault index data of the railway vehicles by using the TF-IDF weight calculation method, extract the fault index data features; Weight adjustment step: According to the distribution characteristics of the word frequency of the fault index data, adjust the weights of the words by the SIF method to balance the influence of common words and rare words in expressing the text, and then rank according to the word frequency.

4. The railway vehicle index analysis method according to claim 1, wherein, The LDA model training step includes: Initialize and set the LDA model parameters, and the model parameters include: the number of topics K, the learning rate α, and β; Use the extracted fault features as training data and input them into the LDA model for training; During the training process, the LDA model automatically discovers the potential topics in the training data and represents the training data as a mixture of potential topics to complete the topic modeling of the fault features; Through the iterative optimization algorithm, continuously adjust the parameters of the LDA model so that the model better fits the input data; When the model converges or reaches the preset number of iterations, stop the training.

5. The railway vehicle index analysis method according to claim 1, wherein The fault index analysis step includes: The LDA model describes the topic structure in the document set through the probability relationship between these three levels of random variables; The three levels of random variables include: the document level, the topic level, and the vocabulary level; among them, the random variable at the document level represents each document in the document set, the random variable at the topic level represents each topic in the document, and the random variable at the vocabulary level represents each vocabulary in the topic.

6. The railway vehicle index analysis method according to claim 5, characterized in that, The steps for analyzing the failure indicators include: Preprocessing and feature extraction operations are performed on the failure data to be analyzed, and a feature vector in the same format as the training data is obtained through feature extraction; among them, the feature vector includes: the type of railway vehicle component, the frequency of occurrence of component type vocabulary, the type of failure, and the frequency of occurrence of failure type vocabulary. The failure data to be analyzed is input into the trained LDA model, and the probability distribution of the failure data belonging to each topic is calculated. According to the probability distribution result, the failure data is classified into the topic category with the highest probability to complete the failure classification.

7. The railway vehicle index analysis method according to claim 5, wherein The processing at the document level includes: Selecting a specific topic for a given document. The topic distribution of each document in the LDA model follows the Dirichlet prior method. Among them, the hyperparameter of the prior method affects the sparsity of the topic distribution. According to the value of the hyperparameter, it is determined whether the topic distribution tendency of the document covers multiple topics. The probability distribution of the topic. In the association scenario allocation model LDA of the Dirichlet distribution method and the multinomial distribution, the distribution probability of the topic is determined. The total of the topic probability distributions. In the topic model LDA, the Dirichlet distribution is used to describe the prior situation of the probability distributions between the document and the topic, and between the topic and the term.

8. A method for analyzing railway vehicle indicators, which uses the railway vehicle indicator analysis system described in any one of claims 1-7, and is characterized in that, The system includes: Data preprocessing module: Collecting the failure indicator data of railway vehicles, and preprocessing the failure indicator data of railway vehicles by using an artificial intelligence-based data conversion method, a data normalization method, and a data set partitioning method to generate various types of preprocessed railway vehicle failure indicator data. Failure feature extraction module: Based on the preprocessed failure indicator data of railway vehicles, using a weight calculation method to weight the freight car failure vocabulary, extracting potential failure features, and adjusting the weighted part-of-speech word frequency according to the word frequency distribution characteristics of the potential failure features, and performing weighted sum and sorting of the failure features. LDA model training module: Based on the extracted failure features, inputting them into the LDA model for training, and continuously adjusting the parameters of the LDA model through iterative optimization to achieve better fitting parameters. Failure indicator analysis module: After extracting the failure features from the railway vehicle indicators to be analyzed, inputting them into the trained LDA model to perform topic modeling of the failure features, and realizing automatic classification of the railway vehicle failure indicator data.

9. An electronic device, characterized in that, The device includes: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory and executes the railway vehicle indicator analysis method according to any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which are used to execute the railway vehicle indicator analysis method according to any one of claims 1-7 when the computer instructions are called.

Images