Railway wagon fault prediction method, device and equipment

Abstract

The embodiment of the invention relates to the technical field of fault prediction, and discloses a rail wagon fault prediction method, device and equipment, and the method comprises the steps: obtaining the vehicle body attribute information and loading and unloading station attribute information of a target rail wagon from a database according to the identification of the target rail wagon and a preset period; the vehicle body attribute information and the loading and unloading station attribute information of the target rail wagon are input into a pre-trained fault prediction model to obtain a fault code of the target rail wagon, and the fault prediction model takes the vehicle body attribute information and the loading and unloading station attribute information of the rail wagon as input; a multi-task classification model is obtained through training by taking the fault code of the railway wagon as expected output; and according to the fault code of the target rail wagon, predicting a fault type and a fault component of the target rail wagon to be faulted. The fault component and the fault type of the railway train can be pre-judged, predictive maintenance can be conveniently achieved, and the railway transportation efficiency and safety can be improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of fault prediction, and in particular to a method, device and equipment for predicting faults of railway freight cars. Background Art

[0002] With the rapid growth of my country's high-speed railways and transportation volume, there is an urgent need to improve the transportation efficiency and operational safety of railway freight cars. The rapid diagnosis, location and elimination of railway freight car faults are the decisive factors to ensure the transportation efficiency and operational safety of railway freight cars.

[0003] The operating environment of railway freight cars is relatively complex, which brings great challenges to the fault diagnosis of railway freight cars. At present, the fault diagnosis of railway freight cars often adopts fixed judgment logic and set fixed thresholds. When the fault conditions are met, the fault is confirmed to have occurred. It is impossible to quickly and accurately predict the fault of railway freight cars, which reduces the efficiency, safety and service quality of railway transportation. It can be seen that it is extremely important to achieve fast and accurate prediction of railway freight car faults. Summary of the invention

[0004] The purpose of the present invention is to at least provide a railway freight car fault prediction method, device and equipment to solve the problem that existing methods cannot quickly and accurately predict railway freight car faults, thereby reducing railway transportation efficiency and safety.

[0005] To solve the above technical problems, at least one embodiment of the present application provides a railway freight car fault prediction method, comprising:

[0006] According to the identification of the target railway freight car, the vehicle body attribute information and the loading and unloading station attribute information of the target railway freight car are obtained from the database according to a preset period;

[0007] Inputting the body attribute information and loading and unloading station attribute information of the target railway freight car into a pre-trained fault prediction model to obtain the fault code of the target railway freight car. The fault prediction model is a multi-task classification model trained with the body attribute information and loading and unloading station attribute information of the railway freight car as input and the fault code of the railway freight car as expected output;

[0008] The fault type and faulty components of the target railway freight car that will fail are predicted according to the fault code of the target railway freight car.

[0009] In some embodiments, the vehicle body attribute information includes the vehicle body capacity, vehicle body material, service life, life of each part and mileage; the loading and unloading station attribute information includes the loading and unloading station number, the probability of vehicle body failure at the loading and unloading station, the current distance of the loading and unloading station to the vehicle, whether there is someone to guide the loading and unloading process, the loading and unloading time and loading and unloading cargo information.

[0010] In some embodiments, the fault code is generated by encoding the fault type and the faulty component using one-hot encoding.

[0011] In some embodiments, the method further comprises:

[0012] The probability of a corresponding fault type failure occurring in a fault component of a target railway freight car is determined according to the probability of the fault code output by the fault prediction model.

[0013] In some embodiments, the fault prediction model is built based on a lightweight gradient boosting machine and adopts a leaf-node-based decision tree learning algorithm.

[0014] In some embodiments, in the process of training the fault prediction model, combined with cross-validation, multiple cycles of small-step iterations are used to optimize the maximum number of leaf nodes on the tree in the fault prediction model, the learning rate, the number of weak learners, the proportion of samples used in training each tree, the proportion of features used in training each tree, the first regularization parameter, the second regularization parameter, the minimum number of samples required for each leaf node, the maximum depth of each tree, the task type and loss function, the evaluation index, and the number of rounds for stopping training.

[0015] In some embodiments, the fault prediction model is constructed based on a long short-term memory artificial neural network.

[0016] At least one embodiment of the present application further provides a railway freight car fault prediction device, comprising:

[0017] An acquisition module, used to acquire the vehicle body attribute information and loading and unloading station attribute information of the target railway freight car from the database according to the identification of the target railway freight car according to a preset period;

[0018] a processing module, for inputting the vehicle body attribute information and the loading and unloading station attribute information of the target railway freight car into a pre-trained fault prediction model to obtain a fault code of the target railway freight car, wherein the fault prediction model is a multi-task classification model obtained by training with the vehicle body attribute information and the loading and unloading station attribute information of the railway freight car as input and the fault code of the railway freight car as expected output;

[0019] The prediction module is used to predict the fault type and faulty components of the target railway freight car according to the fault code of the target railway freight car.

[0020] At least one embodiment of the present application also provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned railway freight car fault prediction method.

[0021] At least one embodiment of the present application further provides a computer-readable storage medium storing a computer program, wherein the computer program implements the above-mentioned railway freight car fault prediction method when executed by a processor.

[0022] The railway freight car fault prediction method, device and equipment provided by the embodiment of the present application obtains the body attribute information and loading and unloading station attribute information of the target railway freight car from the database according to the preset period according to the identification of the target railway freight car; inputs the body attribute information and loading and unloading station attribute information of the target railway freight car into the pre-trained fault prediction model to obtain the fault code of the target railway freight car, and the fault prediction model is a multi-task classification model trained with the body attribute information and loading and unloading station attribute information of the railway freight car as input and the fault code of the railway freight car as the expected output; predicts the fault type and faulty parts of the target railway freight car according to the fault code of the target railway freight car. It is possible to predict the faulty parts and fault types of railway trains, which is convenient for realizing predictive maintenance, from passive repair to active prevention, and is convenient for preparing materials in advance, mobilizing manpower and material resources to solve problems in advance, improving the ability of pre-perception, and helping to improve the efficiency, safety and service quality of railway transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] One or more embodiments are exemplarily described by the pictures in the corresponding drawings, and these exemplary descriptions do not constitute limitations on the embodiments.

[0024] Figure 1 is a flow chart of a railway freight car fault prediction method provided by an embodiment of the present application;

[0025] Figure 2 is a schematic diagram of a railway freight car fault prediction device provided by an embodiment of the present application;

[0026] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the technical solution of the present disclosure, and to fully understand and implement how the present disclosure applies technical means to solve technical problems and achieve the corresponding technical effects, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only embodiments of a part of the present disclosure, not all of the embodiments. The embodiments of the present disclosure and the various features in the embodiments can be combined with each other without conflict, and the technical solutions formed are all within the scope of protection of the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present disclosure.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0029] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0030] Embodiment 1:

[0031] Figure 1 FIG. 1 is a flow chart of a method for predicting railway freight car faults provided by an embodiment of the present application. The method for predicting railway freight car faults of this embodiment can be applied to electronic devices with communication, computing and data storage capabilities. Figure 1 As shown, the railway freight car fault prediction method provided in this embodiment may include:

[0032] S101. Acquire the vehicle body attribute information and loading and unloading station attribute information of the target railway freight vehicle from a database according to the identification of the target railway freight vehicle at a preset period.

[0033] In order to track the entire life cycle of key parts (such as the lower door of an open car), the important fields of key parts can be combined into a unique identifier. Photographing equipment can be deployed at the loading and unloading station to establish an image fault coding library in order to form a data-driven application-guided maintenance. Specifically, under the state maintenance process system based on the "HCCBM" platform, the body theme data ecological chain can be established first to achieve full life data traceability of body parts under the condition of fixed marshaling operation, including marshaling sequence, marshaling load, typical operation failures, failures and pictures at the loading and unloading points during transportation, including maintenance failures, flaw detection defects, and scrapping reasons during maintenance. Through multi-dimensional data, the failure rules of parts that have been formed in the current state maintenance need to be continuously adjusted according to changes in transportation conditions. Each maintenance unit needs to organize production and material preparation according to the new failure rules of parts. Therefore, each unit and department within the equipment company urgently needs to deeply apply data and drive production organization through data.

[0034] Faced with the complex operating environment of railway freight cars, the HCCBM system can be used to achieve the goals of status monitoring, scientific evaluation, precise repair and continuous improvement of railway freight car maintenance. By building car body photo monitoring equipment at the entry and exit loading and unloading points or the entry and exit maintenance bases along the line, and integrating the car body faults and pictures obtained by these photo equipment, it is used to analyze the impact of each loading and unloading station and loading and unloading method on the failure form and failure probability of car body parts, so as to continuously optimize the failure law of status repair car body faults.

[0035] In order to establish a car body fault diagnosis information system, based on data assets, develop car body failure law models, car body inspection and repair methods, and material recommendation models, etc., to form a data-driven car body inspection and repair model, and use the predicted data to guide the inspection and repair, it is necessary to first build a database to form car body data assets. Specifically, by obtaining car body fault pictures before and after loading and unloading, the failure of the car body in the entire life cycle under the transportation environment is studied, especially the impact of the continuous increase in loading and unloading and heavy-load marshaling on the use and inspection of the car body, and form data assets with the car body as the theme.

[0036] The vehicle body data assets included in the database in this embodiment are specifically as follows:

[0037] 1. Formulate the corresponding rules between the unique vehicle ID and the vehicle number: Since some vehicle models of the equipment company have a special management mode that can change the vehicle number, it is necessary to formulate the corresponding rules between the unique vehicle ID and the vehicle number. When the vehicle number is changed, the vehicle body information of the original vehicle number can still be recorded, thereby ensuring the continuous tracking of the vehicle body usage data.

[0038] 2. Vehicle body fault image information: Based on the research on the vehicle's unique ID and the vehicle body image fault coding library, receive and parse the vehicle body photo monitoring equipment information to form vehicle body fault image information.

[0039] 3. Basic vehicle body information: The basic vehicle body information database mainly includes vehicle unique ID, assembly information, size information, loading information, mileage information, status information, loading and unloading point information, loading and unloading method information, marshaling information, and vehicle body fault image information.

[0040] 4. Vehicle maintenance and operation fault information: The vehicle maintenance and operation fault information database refers to the association of vehicle faults and mileage during the vehicle operation and maintenance process. Specifically, it includes the vehicle unique ID, vehicle number, vehicle model, vehicle manufacturing time, vehicle manufacturing unit, maintenance time, maintenance unit, maintenance process, fault description, repair method, number of fault bits, fault severity, size unit, major fault part, minor fault part, specific fault part, fault status, last maintenance process, last maintenance time, last maintenance unit, mileage after maintenance, number of operating months, number of operating days, and the unit and time of the previous maintenance process at all levels.

[0041] 5. Vehicle operation monitoring fault information: The vehicle operation monitoring fault information database is based on the vehicle's unique ID, and associates TF operation process monitoring data, vehicle body photo monitoring fault data with operation mileage, loading and unloading methods, and loading and unloading points. It contains vehicle unique ID, TF operation process monitoring data, vehicle body photo monitoring fault data, operation mileage, loading and unloading methods, loading and unloading points and other information.

[0042] 6. Special position running mileage: The special position running mileage database tracks the running mileage of the vehicle body in the marshaling position and the running mileage of the railway line based on the vehicle's unique ID. It includes unique identification, vehicle number, vehicle sequence, number of marshaling vehicles, number of digits, vehicle type, up and down, position, starting station, ending station, running mileage and other information.

[0043] 7. Car body fault image archive: Based on the unique vehicle ID, the HCCBM system collects car body fault image information, basic car body information, car body maintenance and operation fault information, car body operation monitoring fault information (TF operation process monitoring data, car body photo monitoring fault data), special position operation mileage (marshaling position operation mileage, railway line operation mileage) and other car body full life cycle data to form a car body fault image archive.

[0044] 8. Establish measurement indicators: Establish the body fault image file completion index. The calculation process is as follows: First, count the number of body fault collections and the number of records in the body fault image library respectively; Second, calculate the body fault image file completion rate. The formula is: number of records in the body fault image library / number of body fault collections.

[0045] Among them, data assets can be formed by using Huawei's big data platform Spark Streaming + Hbase to calculate the mileage of vehicles and vehicle bodies. Based on Huawei's big data platform, a data warehouse with vehicle bodies as the theme is built to form the above 8 data assets.

[0046] In this embodiment, the vehicle attribute information includes the vehicle body capacity, vehicle body material, service life, life of each part and mileage; the loading and unloading station attribute information includes the loading and unloading station number, the probability of vehicle body failure at the loading and unloading station, the current distance between the loading and unloading station and the vehicle, whether there is someone guiding the loading and unloading process, loading and unloading time and loading and unloading cargo information. The details are shown in Table 1:

[0047] Table 1

[0048]

[0049] S102. Input the body attribute information and loading and unloading station attribute information of the target railway freight car into a pre-trained fault prediction model to obtain the fault code of the target railway freight car. The fault prediction model is a multi-task classification model that is trained with the body attribute information and loading and unloading station attribute information of the railway freight car as input and the fault code of the railway freight car as the expected output.

[0050] A model can be established to regress the fault information between the vehicle body attributes and the loading and unloading station information, so the data set is classified and processed with vehicle body fault and loading and unloading transportation conditions as keywords. In order to predict the fault of railway freight cars, in this embodiment, the vehicle body attribute information and loading and unloading station attribute information of railway freight cars are used as input, and the fault code of railway freight cars is used as the expected output to train a multi-task classification model - the fault prediction model.

[0051] In this embodiment, one-hot encoding can be used to encode the prediction results of the fault prediction model, and the fault type and the fault component are uniformly encoded to form a fault code. The vehicle body attributes and loading and unloading station attributes associated with a fault will be used as the training input of the model, and the fault code will be used as their label to participate in the prediction training, and a decision tree will be constructed based on information gain. In other words, the fault code is generated by encoding the fault type and the fault component using one-hot encoding.

[0052] In order to predict the fault of the target railway freight car, in this embodiment, it is only necessary to input the body attribute information and loading and unloading station attribute information of the target railway freight car into the pre-trained fault prediction model, and the fault prediction model will output the fault code of the target railway freight car.

[0053] S103: predicting the type of fault and the faulty component of the target railway freight car that will fail according to the fault code of the target railway freight car.

[0054] Since one-hot encoding is performed for both the fault location and the failure form in this embodiment, after obtaining the fault code of the target railway freight car, the result needs to be transcoded. One-hot encoding can solve the problem of disordered result categories, and map each case of the result category to a fixed position of the one-hot encoding. Therefore, it is only necessary to convert according to the reference table during encoding to obtain the fault type and faulty component that is about to fail. Furthermore, the probability of the corresponding fault type failure of the faulty component of the target railway freight car can also be determined based on the probability of the fault code output by the fault prediction model. The probability of the fault code is the probability of a fault in the body of that part.

[0055] In this embodiment, a fault prediction model is obtained based on data asset training, which is used to predict the faulty parts and fault types of railway freight cars, forming a data-driven car body maintenance model, and the prediction data can be used to guide maintenance. The model predicts the failure mode and failure probability of car body parts, and predicts the failure parts of each vehicle body in advance, so as to achieve predictive maintenance and fault diagnosis, from passive repair to active prevention. At the same time, materials can be prepared in advance, manpower and material resources can be mobilized in advance to solve problems, and the ability of pre-perception can be improved. Truly realize the digitalization and intelligentization of railway information technology to improve the efficiency, safety and service quality of the railway transportation system.

[0056] The railway freight car fault prediction method provided in this embodiment obtains the target railway freight car's body attribute information and loading and unloading station attribute information from the database according to the preset period according to the target railway freight car's identification; inputs the target railway freight car's body attribute information and loading and unloading station attribute information into a pre-trained fault prediction model to obtain the target railway freight car's fault code, and the fault prediction model is a multi-task classification model trained with the railway freight car's body attribute information and loading and unloading station attribute information as input and the railway freight car's fault code as the expected output; predicts the target railway freight car's fault type and faulty component according to the target railway freight car's fault code. It is possible to predict the faulty components and fault types of railway trains, which is convenient for realizing predictive maintenance, from passive repair to active prevention, and is convenient for preparing materials in advance, mobilizing manpower and material resources to solve problems in advance, improving the ability of pre-perception, and helping to improve railway transportation efficiency, safety and service quality.

[0057] Embodiment 2:

[0058] On the basis of the above embodiments, the following will further explain in detail how to train a fault prediction model. In some optional implementations, the fault prediction model can be built based on a lightweight gradient boosting machine (LightGradientBoosting Framework, LightGBM) and adopt a leaf-node-based decision tree learning algorithm. In manufacturing and industrial production, a large amount of data collected by sensors and monitoring systems is often used to predict equipment failures, optimize maintenance plans, and improve production efficiency. LightGBM can process complex time series data and multidimensional features to achieve accurate prediction of the health status of equipment and spare parts; the diversity and applicability of LightGBM in industrial prediction. It can not only provide a high-precision prediction model, but also achieve real-time prediction and decision support in large-scale data scenarios by optimizing training efficiency and memory consumption. Therefore, we select a tree model with strong interpretability as the underlying model LightGBM for classification prediction. LightGBM belongs to a gradient boosting framework. Its main idea is to train a series of weak classifiers (generally decision trees) in series, and each tree tries to correct the prediction error of the previous tree. During the training process, the establishment of each tree depends on the conclusions of all previous trees, and the accuracy of the overall model is improved by minimizing the loss function through gradient descent. LightGBM uses a leaf-based decision tree learning algorithm (Leaf-wise) by default. Compared with the traditional level-wise splitting (Level-wise), it can grow trees more efficiently, reduce the number of splits, and thus reduce training time. The Leaf-wise algorithm selects the leaf with the largest split gain from all current leaves each time to split, making the tree deeper, thereby improving the training speed and efficiency. LightGBM supports two parallel modes: feature parallelism and data parallelism. In feature parallelism, the features of the data set are divided into different machines for parallel processing; in data parallelism, different machines process different subsets of the data set and then summarize the results. LightGBM uses a histogram algorithm to preprocess the data, discretize continuous floating-point eigenvalues ​​into a specified number of integer eigenvalues ​​(i.e., histogram binning), and then construct a histogram to reduce memory usage and computational overhead. LightGBM can handle large-scale sparse data and uses a sparse feature optimization algorithm to reduce storage space and computational overhead.

[0059] LightGBM (hereinafter referred to as LGBM) is a gradient boosting decision tree with data compression function and more optimized decision tree modeling process. The data compression method can greatly compress the actual training data while still maintaining relatively complete information. The optimization of the decision tree modeling process can not only accelerate the decision tree modeling speed, but also process the compressed data very well, thus ultimately greatly improving the training efficiency of each tree. It has better understanding ability and prediction accuracy for the rich and complex input data in this application, and also meets the requirements for detection speed in industrial applications.

[0060] In this embodiment, training the LightGBM fault prediction model may include the following steps:

[0061] 1. Data preparation: prepare training data sets and validation data sets. Make sure that the data set contains the required feature columns and target columns (for example, labels or prediction targets), and perform necessary data cleaning and preprocessing (such as missing value processing, feature encoding, etc.). Abnormal data and data with too many missing values ​​should be deleted; for attributes containing a single feature, they need to be normalized and standardized so that their data distribution is Gaussian; for data with fewer missing values, models such as random forests can be used for prediction filling or statistical methods can be used to fill them directly; for attributes containing multiple features, the correlation between each feature and the target output should be considered, and methods such as the chi-square test can be used to remove features with extremely small correlations and retain the remaining features; the original data is processed to generate derivative variables that are relevant to vehicle body failures.

[0062] 2. Define parameters: Set the parameters of the LightGBM model, including but not limited to the number of trees (num_iterations or num_trees), learning rate (learning_rate), maximum tree depth (max_depth), number of leaves (num_leaves), feature column sampling rate (feature_fraction), row sampling rate (bagging_fraction), etc. The choice of parameters directly affects the performance and training speed of the model.

[0063] 3. Create a Dataset object: Convert the data into a LightGBM-specific data structure, usually a Dataset object. The Dataset object can process data more efficiently, improve training speed and memory utilization. You can use the lgb.Dataset() function to create a Dataset object and pass in training data and corresponding parameters.

[0064] 4. Model training: Use the lgb.train() function to train the model. In the training function, pass in the Dataset object, model parameters, and other optional parameters, such as validation datasets, evaluation indicators, etc. The model will perform multiple rounds of iterations based on the set parameters and datasets, gradually optimizing the tree structure and parameters to minimize the loss function.

[0065] 5. Model tuning: You can adjust model parameters based on the performance of the validation dataset, such as increasing the number of trees, adjusting the learning rate, or increasing regularization parameters to prevent overfitting. LightGBM provides a wealth of tuning parameters and techniques, such as early stopping and cross-validation, to help optimize the model.

[0066] 6. Model saving: After training is completed, the trained model can be saved to a local file for subsequent loading and use. The model can be saved using the model.save_model() method for prediction in a production environment.

[0067] 7. Model evaluation: Use the trained model to predict test data or independent validation data and evaluate the performance of the model. Various indicators such as accuracy, mean square error (MSE), logarithmic loss function (logloss), etc. can be used to evaluate the predictive ability of the model.

[0068] The data set used for training includes 509,759 data items, with a total of 256 categories of vehicle body fault images, including but not limited to: handrail cracks, lower door buckle deformation, horizontal strut damage inside the vehicle, side wall cracks, lower door panel deformation, side wall deformation, side wall damage, handrail deformation, lower door damage, horizontal strut damage inside the vehicle, lower door buckle bolt wear, escalator damage, lower door panel damage, lower door panel cracks, side wall corrosion, lower door buckle damage, escalator deformation, lower door buckle seat damage, lower door buckle damage, lower door panel outer account, handrail damage, lower door hinge round pin break, handrail break, pedal deformation, lower door hinge open welding, lower door buckle break, lower door hinge crack, strut seat corrosion, strut corrosion, side wall panel open welding... It can be seen that the faulty parts and fault types are very detailed, and fault repair can be accurately guided according to the predictions of the fault prediction model.

[0069] Optionally, in the process of training the fault prediction model, combined with cross-validation, multiple cycles of small-step iterations are used to optimize the maximum number of leaf nodes (num_leaves) on the tree in the fault prediction model, the learning rate (learning_rate, used to control the update amplitude of the model weights in each iteration), the number of weak learners (n_estimators), the proportion of samples used in training each tree (subsample), the proportion of features used in training each tree (colsample_bytree), the first regularization parameter (L1 regularization parameter, reg_alpha), the second regularization parameter (L2 regularization parameter, reg_lambda), the minimum number of samples required for each leaf node (min_child_samples), the maximum depth of each tree (max_depth), the task type and loss function (objective:), the evaluation indicator (metric), and the number of rounds for stopping training (early_stopping_rounds).

[0070] Finally, one-hot encoding is used to encode the prediction results, and the fault type and faulty parts are uniformly encoded to form a fault code. The car body attributes and loading and unloading station attributes associated with a fault will be used as the training input of the model, and the fault code will participate in the prediction training as their labels, and a decision tree will be built based on information gain. After the optimization is completed, the decision model can be applied to car body fault prediction. With the attributes of the loading and unloading station and the car body attributes as input, the LGBM model will predict the possible fault codes of the train, and the confidence of the result in the classification model will be regarded as the fault probability. Since the system performs one-hot encoding for both the fault location and the failure form, the results need to be transcoded in the result output part. One-hot encoding solves the problem of disordered result categories, and maps each case of the result category to a fixed position of the one-hot encoding. After the LightGBM prediction model predicts the one-hot encoding value and probability, it only needs to be converted into the location of the car body fault according to the comparison table during encoding. The probability of the encoding value is the probability of the car body fault at that location.

[0071] After the fault prediction model is trained, it can be integrated with the HCCBM application system to regularly send the train number Train_code to the Kafka message queue of the big data platform, and then consume the Kafka data through the SpringBoot microservice, query the information of the vehicle body in the data management database, and send it to the LightGBM model loaded by the Flask service through the Post request of the HTTP protocol. After the model calculation, the result is returned to the SpringBoot microservice. The SpringBoot microservice stores the result data in the data management database and sends it to the Kafka message queue at the same time. Finally, the HCCBM application system consumes the result data returned by Kafka, determines the failure mode and failure probability of the vehicle body accessories, guides vehicle maintenance, prepares materials in advance and arranges maintenance plans.

[0072] In other optional implementations, the fault prediction model may also be constructed based on a long short-term memory artificial neural network.

[0073] In summary, through data governance means, we integrate basic car body information, car body fault image information, car body inspection and operation fault information, car body operation monitoring fault information, special location mileage, car body fault image archives, measurement information and establish feature engineering, train the LightGBM prediction model based on existing data, and use the model to predict the failure form and failure probability of car body parts, and continuously optimize the failure law of state repair car body faults. Through the model, we predict the failure form and failure probability of car body parts, predict the failure parts of each vehicle body in advance, realize predictive maintenance and fault diagnosis, from passive repair to active prevention, and prepare materials in advance, mobilize manpower and material resources to solve problems in advance, and improve the ability of pre-perception. We can truly realize the digitalization and intelligent information technology of railways to improve the efficiency, safety and service quality of railway transportation systems.

[0074] Embodiment three:

[0075] Another embodiment of the present application relates to a railway freight car fault prediction device. The implementation details of the railway freight car fault prediction device of this embodiment are specifically described below. The following content is only for the convenience of understanding the implementation details provided, and is not necessary for the implementation of this solution. The schematic diagram of the railway freight car fault prediction device of this embodiment can be as follows Figure 2 As shown, it includes: an acquisition module 201, a processing module 202 and a prediction module 203.

[0076] An acquisition module 201 is used to acquire the vehicle body attribute information and loading and unloading station attribute information of the target railway freight car from a database according to the identification of the target railway freight car according to a preset period;

[0077] The processing module 202 is used to input the vehicle body attribute information and the loading and unloading station attribute information of the target railway freight car into a pre-trained fault prediction model to obtain the fault code of the target railway freight car. The fault prediction model is a multi-task classification model that is trained with the vehicle body attribute information and the loading and unloading station attribute information of the railway freight car as input and the fault code of the railway freight car as expected output;

[0078] The prediction module 203 is used to predict the fault type and faulty component of the target railway freight car according to the fault code of the target railway freight car.

[0079] The device of this embodiment can be used to perform Figure 1 The technical solution of the method embodiment shown has similar implementation principles and technical effects, which will not be repeated here.

[0080] In some optional implementations, the vehicle body attribute information includes the vehicle body capacity, vehicle body material, service life, life of each part and mileage; the loading and unloading station attribute information includes the loading and unloading station number, the probability of vehicle body failure at the loading and unloading station, the current distance of the loading and unloading station to the vehicle, whether there is someone to guide the loading and unloading process, the loading and unloading time and loading and unloading cargo information.

[0081] In some optional implementations, the fault code is generated by encoding the fault type and the faulty component using one-hot encoding.

[0082] In some optional implementations, the prediction module 203 is further configured to determine the probability of a corresponding fault type fault occurring in a fault component of the target railway freight car according to the probability of the fault code output by the fault prediction model.

[0083] In some optional implementations, the fault prediction model is built based on a lightweight gradient boosting machine and adopts a leaf-node-based decision tree learning algorithm.

[0084] In some optional implementations, in the process of training the fault prediction model, combined with cross-validation, multiple cycles of small-step iterations are used to optimize the maximum number of leaf nodes on the tree in the fault prediction model, the learning rate, the number of weak learners, the proportion of samples used in training each tree, the proportion of features used in training each tree, the first regularization parameter, the second regularization parameter, the minimum number of samples required for each leaf node, the maximum depth of each tree, the task type and loss function, the evaluation index, and the number of rounds for stopping training.

[0085] In some optional implementations, the fault prediction model is constructed based on a long short-term memory artificial neural network.

[0086] It is worth mentioning that all modules involved in this embodiment are logic modules. In practical applications, a logic unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by this application, but this does not mean that there are no other units in this embodiment.

[0087] Embodiment 4:

[0088] Another embodiment of the present application relates to an electronic device, such as Figure 3 As shown, it includes: at least one processor 301; and a memory 302 that is communicatively connected to the at least one processor 301; wherein the memory 302 stores instructions that can be executed by the at least one processor 301, and the instructions are executed by the at least one processor 301 so that the at least one processor 301 can execute the railway freight car fault prediction method in the above-mentioned embodiments.

[0089] Among them, the memory and the processor are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor is transmitted on a wireless medium via an antenna, and further, the antenna also receives data and transmits the data to the processor.

[0090] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management and other control functions. The memory can be used to store data used by the processor when performing operations. The processor can include, but is not limited to, for example, one or more processors or microprocessors. Each processor can be an application specific integrated circuit (Application Specific Integrated Circuit, referred to as ASIC), a digital signal processor (Digital Signal Processor, referred to as DSP), a digital signal processing device (Digital Signal Processing Device, referred to as DSPD), a programmable logic device (Programmable Logic Device, referred to as PLD), a field programmable gate array (Field Programmable Gate Array, referred to as FPGA), a controller, a microcontroller, a microprocessor or other electronic components to implement the method in the above embodiment.

[0091] Embodiment five:

[0092] Another embodiment of the present application relates to a computer-readable storage medium storing a computer program. The computer program is executed by the processor to implement the above method embodiment. That is, those skilled in the art can understand that all or part of the steps in the above method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as: ROM), random access memory (Random Access Memory, referred to as: RAM), disk or optical disk and other media that can store program code.

[0093] The computer-readable storage medium may also store at least one computer executable program / instruction, which may be, for example, a computer-readable instruction. The computer-readable storage medium includes, but is not limited to, for example, a volatile memory and / or a non-volatile memory. The volatile memory may include, for example, a random access memory (RAM) and / or a cache memory (cache), etc. The computer-readable storage medium may include, for example, a read-only memory (ROM), a hard disk, a flash memory, etc. For example, a non-transitory computer-readable storage medium may be connected to a computing device such as a computer, and then, when the computing device runs the computer-readable instructions stored on the computer-readable storage medium, the various methods described above may be performed.

[0094] In addition, the computer device may also include (but not limited to) a data bus, an input / output (I / O) bus, a display, and input / output devices (eg, keyboard, mouse, speaker, etc.), etc.

[0095] The processor may communicate with external devices via an I / O bus via a wired or wireless network.

[0096] In one embodiment, the at least one computer executable instruction may also be compiled into or constitute a software product / computer program product, wherein one or more computer executable instructions are executed by a processor to perform the various functions and / or method steps in the embodiments described in the present technology.

[0097] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

[0098] In the embodiments provided in the present disclosure, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0099] It should be noted that in the present disclosure, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element limited by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0100] Although the embodiments disclosed in the present disclosure are as above, the above contents are only embodiments adopted for facilitating the understanding of the present disclosure and are not intended to limit the present disclosure. Any technician in the technical field to which the present disclosure belongs can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present disclosure, but the scope of patent protection of the present disclosure shall still be subject to the scope defined in the attached claims.

Claims

1. A method for predicting railway freight car failures, characterized in that: include: According to the identification of the target railway freight car, the vehicle body attribute information and the loading and unloading station attribute information of the target railway freight car are obtained from the database according to a preset period; Inputting the body attribute information and the loading and unloading station attribute information of the target railway freight car into a pre-trained fault prediction model to obtain a fault code of the target railway freight car, wherein the fault prediction model is a multi-task classification model trained with the body attribute information and the loading and unloading station attribute information of the railway freight car as input and the fault code of the railway freight car as expected output; The fault type and faulty component of the target railway freight car that will fail are predicted according to the fault code of the target railway freight car.

2. The method according to claim 1, characterized in that The vehicle body attribute information includes the vehicle body capacity, vehicle body material, service life, life of each part and mileage; the loading and unloading station attribute information includes the loading and unloading station number, the probability of vehicle body failure at the loading and unloading station, the current distance of the loading and unloading station to the vehicle, whether there is someone to guide the loading and unloading process, the loading and unloading time and loading and unloading cargo information.

3. The method according to claim 1, characterized in that The fault code is generated by encoding the fault type and the fault component using one-hot encoding.

4. The method according to claim 1, characterized in that: The method further comprises: The probability of a corresponding fault type fault occurring in a fault component of the target railway freight car is determined according to the probability of the fault code output by the fault prediction model.

5. The method according to any one of claims 1 to 4, characterized in that The fault prediction model is built based on a lightweight gradient boosting machine and adopts a leaf-node-based decision tree learning algorithm.

6. The method according to claim 5, characterized in that In the process of training the fault prediction model, combined with cross-validation, multiple cycles of small-step iterations are used to optimize the maximum number of leaf nodes on the tree in the fault prediction model, the learning rate, the number of weak learners, the proportion of samples used in training each tree, the proportion of features used in training each tree, the first regularization parameter, the second regularization parameter, the minimum number of samples required for each leaf node, the maximum depth of each tree, the task type and loss function, the evaluation index and the number of rounds for stopping training.

7. The method according to any one of claims 1 to 4, characterized in that The fault prediction model is constructed based on a long short-term memory artificial neural network.

8. A railway freight car fault prediction device, characterized in that: include: An acquisition module, used for acquiring the vehicle body attribute information and loading and unloading station attribute information of the target railway freight car from a database according to the identification of the target railway freight car according to a preset period; a processing module, for inputting the vehicle body attribute information and the loading and unloading station attribute information of the target railway freight car into a pre-trained fault prediction model to obtain a fault code of the target railway freight car, wherein the fault prediction model is a multi-task classification model obtained by training with the vehicle body attribute information and the loading and unloading station attribute information of the railway freight car as input and the fault code of the railway freight car as expected output; The prediction module is used to predict the fault type and faulty component of the target railway freight car according to the fault code of the target railway freight car.

9. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the railway freight car fault prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the railway freight car fault prediction method according to any one of claims 1 to 7 is implemented.

Images