Motor vehicle servicing integrated management and control method for power set servicing

Abstract

The invention discloses a motor vehicle servicing integrated management and control method for dynamic set servicing, and the method comprises the steps: collecting the state data of a dynamic set vehicle, predicting the future servicing demand through big data analysis, and generating an optimal servicing plan through a machine learning algorithm; a unified data exchange standard and an information platform are established, the execution condition of an operation plan is monitored in real time, a deep learning technology is used for training a fault prediction model, a vehicle-mounted diagnosis system is integrated, the vehicle state is monitored in real time, analysis and comparison with the fault prediction model are carried out, and a collaborative operation mechanism is established. Communication and cooperation among professionals are promoted through the information platform, the system is maintained and upgraded regularly, the stability and safety of the system are ensured, and efficiency, quality and cost of servicing work are evaluated and analyzed regularly. According to the invention, the whole process and all-around management and control of the dynamic set servicing work are realized, and the efficiency, accuracy and collaboration of the servicing work are improved through real-time data acquisition, intelligent analysis and decision, and automatic work execution and monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of EMUs, and in particular to an integrated control method for EMU maintenance. Background Art

[0002] During railway transportation, trains and EMUs used for transportation need to be repaired and maintained according to the maintenance schedule and entered into the section to complete daily maintenance operations. At present, domestic train maintenance mainly includes locomotive maintenance, passenger car maintenance and daily maintenance of CR200J centralized power EMUs. Maintenance operations are an important part of the operation and maintenance of CR200J centralized power EMUs, the link between the operation and maintenance of CR200J centralized power EMUs, and the key link in the transformation of traditional railway operation and maintenance to digitalization, informatization and intelligence.

[0003] At present, the daily inspection and maintenance of CR200J centralized power EMUs are carried out in the passenger maintenance station of the vehicle depot, mainly led by vehicle professionals. During the operation, it involves the mutual cooperation and coordination of multiple professions such as vehicle, locomotive, passenger transportation and electrical services. Especially in the current transitional stage, there are few centralized power EMUs and there are cross-operations with ordinary speed vehicles. Due to the complex operating environment, different operating processes, and numerous operating units, integrated maintenance has become one of the important and imperative tasks in the operation and maintenance management of train vehicles and centralized power EMUs. The current maintenance of centralized power EMUs has the following shortcomings: Problems with work efficiency and accuracy: Manual scheduling and on-site monitoring are easily affected by human factors, resulting in inaccurate work plans, unreasonable scheduling, and low execution efficiency. Manual tracking of plan execution is prone to omissions in plan tracking and failure to achieve expected work goals.

[0004] Information islands and lack of transparency: The information systems of different disciplines are relatively independent, making data sharing difficult, forming information islands. This not only affects the efficiency of work coordination, but also makes it difficult for management to grasp the overall picture of maintenance work in real time and effectively monitor the progress of work and vehicle status.

[0005] Delayed fault response and processing: Due to the lack of intelligent fault warning and tracking systems, when a vehicle fails, it is often impossible to quickly locate and handle it, affecting the normal operation of the centralized power EMU vehicles.

[0006] Poor adaptability and scalability: With the continuous advancement of railway technology and the increase in the number of EMUs, the existing maintenance management model is unable to cope with new models, new technologies and higher standards of operation, and it is difficult to effectively expand and adapt. Summary of the invention

[0007] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide an integrated control method for vehicle preparation of EMU preparation.

[0008] The objective of the present invention is achieved through the following technical solutions: The present invention discloses a vehicle preparation integrated control method for EMU preparation, comprising the following steps: S1. Collect the status data of the EMU vehicle, including mileage, wear and tear, and fault history, clean and format the collected status data to ensure the accuracy and consistency of the status data, and store the processed data in the database to provide a basis for subsequent analysis; S2. Based on the collected vehicle status data and maintenance process requirements, use big data analysis algorithms to predict future maintenance needs, consider job priorities, resource availability, and job time windows, use machine learning algorithms to automatically generate the optimal maintenance plan, and automatically publish the maintenance plan to relevant operators and managers to ensure information synchronization; S3. Integrate various professional information systems, establish unified data exchange standards and information platforms, use RFID and sensors to track the progress of operations in real time, combine GIS to display the location and status of operations, monitor the execution of operation plans in real time, issue early warnings for deviations, and automatically adjust plans to deal with emergencies. Regularly evaluate operation efficiency and quality to provide a basis for continuous improvement; S4. Based on historical fault data, deep learning technology is used to train fault prediction models, integrate on-board diagnostic systems, monitor vehicle status in real time, and analyze and compare with fault prediction models. Once an abnormality is found or a fault is predicted, an early warning is triggered immediately, and relevant personnel are notified through the information platform. Maintenance plans are automatically matched according to the fault type, maintenance resources are dispatched, and on-site rapid processing is guided; S5. Establish a collaborative operation mechanism, clarify the responsibilities and collaboration processes among different disciplines, promote communication and collaboration among different disciplines through information platforms, ensure timely transmission and sharing of operation information, monitor the use of resources, and make reasonable arrangements and optimize positions; S6. Regularly maintain and upgrade the system to ensure system stability and security, expand system functions and adjust system adaptability according to new vehicle models, new technologies and demanding operation requirements, collect user feedback and opinions, and continuously optimize system functions and user experience; S7. Regularly evaluate and analyze the efficiency, quality and cost of maintenance operations. Based on the evaluation results, propose improvement measures and suggestions, continuously optimize the maintenance operation process and management methods, track the implementation effect of improvement measures, form a closed-loop management, and continuously improve the level and efficiency of maintenance operations.

[0009] Furthermore, the use of big data analysis algorithms to predict future maintenance needs in step S2 specifically includes the following steps: S211. Clarify the data parameters for predicting future maintenance big data analysis and determine the data source of the big data analysis algorithm, where the data source includes the train's running routes, mileage, historical running on-board data, ticket failures, maintenance status, maintenance cycle, maintenance history, and train history; The operation routes include the running frequency, running time and running cycle; The running kilometers include the cumulative running kilometers and daily running kilometers of the train; The historical running on-board data includes the running status of the train and historical running fault alarm information; The ticketing failures are operational failures and failures found during maintenance work; The maintenance status includes the health status of the current train and the maintenance status of the planned maintenance process; The maintenance cycle is the planned maintenance schedule of the train; The maintenance history includes the maintenance status and maintenance time of the train; The train history includes the status and replacement records of train components; S212. Collect data through ETL tools according to the type and characteristics of data sources; S213. Conduct preliminary quality checks on the data after collection to ensure the completeness, accuracy and consistency of the data; S214, preprocessing the data, specifically including: Data cleaning: handle missing values, outliers and duplicate values; for missing values, use mean filling, median filling and regression filling according to the specific situation; for outliers, identify and process them by setting thresholds; for duplicate values, delete or merge them directly; Data conversion: convert text data into numerical data, standardize or normalize the data to eliminate the impact of different dimensions and magnitudes on data analysis; Data division: Divide the cleaned and converted data set into training set, validation set and test set; use 70% of the data as the training set, 20% of the data as the validation set, and 10% of the data as the test set; S215, performing feature processing on the data, specifically including: Feature selection: Based on domain knowledge and data correlation analysis, select features related to maintenance demand forecasting; Feature extraction: extract new features from raw data to improve the predictive power of the model; Feature dimensionality reduction: If the feature dimension of the data is high, use principal component analysis, linear discriminant analysis or feature selection to reduce the dimension to reduce the computational complexity and improve the generalization ability of the model; S216, perform model selection and training, specifically including: Model selection: Select a model based on the characteristics of the data and the forecast requirements. For the maintenance demand forecast problem, choose to use a neural network model; Hyperparameter tuning: Use grid search, random search, or Bayesian optimization to adjust the model's hyperparameters, including learning rate, number of iterations, and regularization coefficient, to improve model performance; Model training: Train the model on the training set to find the optimal model parameters; also include preventing overfitting and underfitting by using cross-validation during the training process; S217, model evaluation and tuning, including: Model evaluation: Evaluate the performance of the model on the validation set; evaluation indicators include mean square error, mean absolute error and accuracy; at the same time, draw learning curves and validation curves to assist in evaluating model performance; Model tuning: Tune the model based on the evaluation results. If the model performance is poor, adjust it by replacing the model, adding features, adjusting hyperparameters, or improving feature engineering. Prevent overfitting: During the training process, regularization, early stopping, and dropout are used to reduce the risk of overfitting. S218. Predict and monitor the maintenance status, including: Model deployment: Deploy the trained model to the production environment. The model needs to be converted into a format suitable for the production environment and integrated into the system clock. Real-time prediction: In the production environment, real-time data input is received and deployed models are used to predict maintenance demand. The prediction results are used to guide the formulation and execution of maintenance plans. Monitoring and feedback: Monitor the model prediction results to ensure the stability and accuracy of the model. At the same time, collect user feedback and new data samples to continuously optimize model performance. S219. Continuous improvement and optimization of the model; specifically including: Data update: As new data is continuously generated and collected, the training dataset is updated regularly and the model is retrained to improve prediction accuracy; Model Update: As machine learning techniques continue to evolve and new algorithms emerge, model architecture and algorithms are regularly evaluated and updated to improve predictive performance. Domain knowledge fusion: Combine knowledge and experience in this field to further optimize and improve the model.

[0010] Furthermore, the maintenance plan in step S2 specifically includes: S221. Collect key data, including running routes, train operation diagrams, mileage, historical fault records and maintenance cycles, clean the collected key data, remove duplicate, erroneous or invalid data to ensure the accuracy and completeness of the data, and integrate the cleaned data to form a unified data set for subsequent analysis and processing; S222, according to the running route and train operation diagram of the EMU vehicle, determine the running law and running cycle of the train, and according to the running mileage and maintenance cycle of the train, determine the balanced maintenance cycle of D2 maintenance, which specifically includes monthly maintenance, quarterly maintenance, half-year maintenance and annual maintenance; S223, subcontract and disassemble the D2 balance repair to obtain monthly repair, quarterly repair and semi-annual repair, and merge them into the daily D1 repair work with and without electricity according to the work content; S224. According to the scope of work of monthly repair, quarterly repair and half-year repair, monthly repair is divided into A package, L package and D package according to profession, quarterly repair is divided into B package and 2C package according to profession, and half-year repair is divided into C package; S225. The subcontracting schemes generated in step S224 are systematically and intelligently integrated to form a J package. The longest period of balanced maintenance of the J package is 24 subcontracts, covering a half-year maintenance period. The subcontracting method of the J package is based on a period of six days. Four periods include monthly maintenance content, and six days are reserved as a mitigation period. If there are weekends and holidays, the period is postponed. The subcontracting period is recorded, the subcontracting is automatically arranged and calculated, and the execution of the subcontracting is monitored. If it is necessary to transfer the weekly plan, the weekly plan is automatically adjusted, and the daily operation plan is automatically generated based on the monthly plan and the weekly plan; S226. The system automatically generates a monthly plan based on the logic algorithm. The monthly plan is published after being confirmed by the maintenance scheduling and notified to each work team. The weekly plan and daily plan are automatically generated based on the monthly plan. The daily plan automatically generates the daily plan for tomorrow before 12:00 every day. If each work team has a temporary plan, it needs to submit the temporary plan in the system before 12:00 every day. The system automatically integrates the temporary plan when generating the plan for tomorrow; S227. The generated daily plan is automatically sent to each work team after being confirmed by the preparation and scheduling. Each work team makes personnel scheduling and work preparation work according to the daily plan; S228. The system automatically monitors and updates the execution of the daily plan, accumulates the monthly execution status, provides data support for train plan formulation, and continuously optimizes the scheduling of automatic plans.

[0011] Preferably, the step S2 of automatically generating the optimal maintenance plan using a machine learning algorithm specifically includes the following steps: S231, data preprocessing, specifically including: Collect maintenance data: collect and predict future maintenance status data, monthly overall plan and plan execution status, work team scheduling, holiday time, balanced maintenance cycle, balanced maintenance execution status, train status information and train maintenance status; Data cleaning: Clean the collected data to remove duplicate, erroneous and invalid data to ensure data accuracy and consistency; Data conversion and standardization: convert unstructured data into structured data and standardize numerical data for subsequent model training; S232. Feature engineering, including: Feature selection: Select key features that have an impact on the generation of maintenance plans from the cleaned data, including train mileage, train status information, train maintenance status, overall plan, and balanced maintenance execution status; Feature extraction: Extract new features through statistical methods or domain knowledge, including the wear degree and failure frequency of vehicle components; Feature dimensionality reduction: If the feature dimension is too high, principal component analysis is used to reduce the dimension to improve model training efficiency; S233, model selection and training, specifically including: Select machine learning algorithms: select decision tree, random forest or support vector machine algorithms according to the complexity of the problem and the characteristics of the data to establish an algorithm model for vehicle component prediction and health status; select logistic regression and gradient boosting tree to establish an algorithm model for optimal maintenance plan generation; Model training: Use preprocessed data and the selected algorithm to train the model. During the training process, adjust the model parameters through cross-validation to improve the generalization ability of the model. Model evaluation: Evaluate the trained model through accuracy, recall and F1 score to ensure the effectiveness of the model; S234, generating an optimal maintenance plan, specifically including: Generate maintenance plan: import the train data to be maintained into the trained model and output the optimal maintenance plan, which includes the maintenance time, the operation time and operation content of each team; Plan optimization: Optimize the generated maintenance plan according to the maintenance time and balanced maintenance plan to ensure the feasibility and economy of the plan; Result feedback and iteration: Feedback the generated maintenance plan to the maintenance scheduling personnel, collect feedback, and iterate and optimize the model based on the feedback; The model generates the optimal maintenance plan through the following formula: Optimal_Maintenance_Plan = f(Train_Data, Features, Constraints,Algorithms); Among them, Optimal_Maintenance_Plan is the optimal maintenance plan, including monthly maintenance plan, weekly maintenance plan and daily production plan; Train_Data is the training data, which is the historical maintenance data used for model training, including train status information, maintenance records, mileage and failure rate; Features are the key factors that are extracted from the training data and have an impact on the generation of the maintenance plan, including train mileage, component wear, failure frequency, and balancing repair cycle; Constraints are constraints that need to be met during the maintenance plan generation process, including maintenance time cycle, team scheduling, holiday time and balanced maintenance plan; Algorithms are algorithms, which are machine learning algorithm models used to generate maintenance plans. The algorithm model generates the optimal maintenance plan based on the input training data, features, and constraints.

[0012] Furthermore, the step S3 specifically includes the following steps: S31, when the EMU train enters the depot, the AEI identification device automatically identifies the train arrival and the train model and number. The system collects the information of the AEI device, matches it with the train daily plan, and automatically generates the operation process; S32. Through the Internet of Things technology, the system controls the electric garage door and the driving signal light, determines whether the surrounding environment of the electronic garage door is safe, automatically opens the garage door, and turns the signal light green. After the train enters the garage and goes to the platform, it automatically determines and closes the garage door, and turns off the signal light at the same time; S33. The corresponding work team uses the in-warehouse workstation terminal for face recognition and applies for power-off operation. The approval of the power-off operation is automatically transferred to the on-site dispatch and integrated duty room for approval and authorization. After approval, the system jointly controls the in-warehouse voice dialogue system and the in-warehouse workstation terminal to remind the start of the operation. S34. The system automatically notifies relevant work teams to carry out joint work according to the preset work process. Each team member receives work tasks and specific guidance information through handheld terminals or workstation terminals. The system uses sensors, video surveillance, image algorithms and AI artificial intelligence to conduct all-round monitoring of the work area. S35. If an abnormal situation occurs, the system will immediately sound an alarm and notify relevant personnel to take emergency measures through the voice dialogue system. After the operation is completed, each team member completes the operation through face recognition on a handheld terminal or workstation terminal. The system automatically reviews the operation status, eliminates the operation authority after confirmation, and broadcasts the completion information of the operation; S36. If there is any equipment failure or abnormality during the operation, the operator can quickly report the fault information through the handheld terminal or the workstation terminal. After receiving the fault information, the system will automatically notify the relevant maintenance personnel to handle it and track the progress of the fault handling in real time; S37. After all operations are completed, the system automatically generates a train certificate and operation report based on the operation report and fault handling report submitted by each team. The certificate and operation report contain train information, operation content, operation time and operating personnel.

[0013] Furthermore, the intelligent fault warning and response in step S4 specifically includes the following steps: S411. Collect key train operation data in real time through sensors and on-board data, and clean the collected data to improve the accuracy and efficiency of subsequent data analysis; S412. Use big data analysis technology to conduct in-depth mining and analysis on the cleaned data, identify potential fault signs and abnormal patterns, and train machine learning models to make intelligent judgments and predictions on the operating status of equipment or systems. When the model identifies a fault or abnormality, it will output a corresponding warning signal. S413. Set preset warning thresholds and preset conditions. When the anomaly identified by the data analysis layer reaches the preset warning threshold, the warning mechanism is immediately triggered. The warning information is promptly conveyed to relevant personnel through voice broadcast and system reminders. The warning information includes the fault type, fault location and severity, so that relevant personnel can quickly take corresponding measures. S414. The system also includes an emergency handling guide library to provide maintenance personnel with references and suggestions for fault handling. For predictable faults or abnormal conditions, the system initiates an automated response process, including automatic shutdown or switching to backup equipment, to reduce the impact of the fault on the operation of the equipment or system; S415. After receiving the warning information, the maintenance personnel shall take response measures according to the emergency handling guide and the actual situation, including repairing the fault or eliminating the abnormality; S416. Based on historical fault data, establish a fault prediction model to predict the future operating status of the equipment or system. According to the output results of the fault prediction model, formulate a preventive maintenance plan in advance and conduct regular inspection and maintenance of key components; S417. The system continuously adapts to changes in the operating environment of equipment and systems through continuous learning algorithms, optimizes prediction models and algorithms, improves the accuracy and timeliness of early warnings, collects user feedback and system operation data, and continuously optimizes and improves the system to improve system stability and reliability.

[0014] Preferably, in step S417, the system continuously optimizes the prediction model and algorithm through continuous learning, including the following specific steps: S4171. New data collection and annotation, including: As the vehicle operates and the environment changes, new data is automatically and continuously collected and annotated. The new data includes the health status information of the vehicle in different operating environments; S4172, Model retraining and updating, including: Integrate new data into the existing training set and retrain the model. During the retraining process, gradually introduce new data to avoid causing too much impact on the model. Based on the performance evaluation results of the retrained model, decide whether to update the model parameters and architecture. S4173, Online learning and adaptation, specifically including: Use online learning algorithms to enable the model to continuously learn and adapt to new data and environmental changes during actual operation; monitor the model's predictive performance and trigger the model update mechanism in a timely manner when performance degradation is found; S4174. User feedback data collection, including: Collect user feedback on the prediction results through user feedback channels, including the website's built-in feedback channel, social media, and customer service hotline; the feedback includes the accuracy and timeliness of the prediction results and the user's experience; S4175, system optimization and iteration, including: Optimize and improve the system based on user feedback and data analysis results; continuously optimize prediction models and algorithms to improve the accuracy and timeliness of early warnings; iteratively update system functions and interface design to enhance user experience.

[0015] Furthermore, the use of deep learning technology to train the fault prediction model in step S4 specifically includes the following steps: S421. Data preparation, including: Data collection: Collect large amounts of mission-related raw data from databases, sensors, and networks; Data cleaning: Process outliers, missing values, and duplicate values ​​in the data to ensure data quality; Data annotation: providing labels or target values ​​for supervised learning tasks; Data segmentation: divide the dataset into training set, validation set and test set; Data preprocessing: including data standardization, normalization, feature extraction and feature selection, which is used to reduce the learning burden of the model and improve performance; S422. Establishing a data model, specifically including: Select model architecture: Select appropriate deep learning models based on the task type, which include image recognition, natural language processing, and time series prediction. The deep learning models include CNN, RNN, LSTM, and Transformer. Define the model: Use deep learning frameworks such as TensorFlow and PyTorch to define the structure of the model, including input layer, hidden layer, output layer, activation function, and loss function. S423, model training, specifically including: Data input: input the preprocessed training data into the model; Compile the model: Set the optimizer and learning rate in the deep learning framework and compile the model. The optimizer includes Adam and SGD. Iterative training: Calculate the prediction value and loss through forward propagation, and then update the weight and bias of the model through back propagation; this process will be repeated many times until the preset number of training rounds is reached or the loss converges; S424, Model verification and tuning, including: Performance evaluation: Evaluate the performance of the model on the validation set. The evaluation indicators include accuracy, precision, recall, F1 score and mean square error. Hyperparameter tuning: Adjust the model's hyperparameters based on the performance of the validation set to improve model performance. The hyperparameters include learning rate, batch size, number of layers, and number of nodes. Regularization and Optimization: Regularization techniques are used to prevent overfitting, and optimization algorithms are used to accelerate the training process. The regularization techniques include L1, L2 regularization, and Dropout; the optimization algorithms include momentum and RMSprop; S425, Model evaluation and optimization, including: Performance evaluation: Use evaluation metrics to evaluate the performance of the model on the test set; Model optimization: If the model performance does not meet the requirements, optimize it by adjusting the model structure, hyperparameters, increasing the amount of data, and improving data preprocessing; Model fusion: improve overall performance by combining the prediction results of multiple models through ensemble learning; ensemble learning includes bagging and boosting; S426, Model loading and application, including Model export: export the trained model to a deployable format; the deployable formats include TensorFlow SavedModel and PyTorch; Model deployment: deploy the model to the production environment; Performance monitoring: Monitor the performance of the model in real applications and update and optimize it as needed; Data collection and annotation: Continuously collect new data and annotate it to enrich the training set; Model iteration: Continuously iterate and optimize the model based on new data and feedback to improve its performance and accuracy.

[0016] The beneficial effects of the present invention are: 1) Improving the accuracy of work planning and execution: In view of the problems of inaccurate work planning and low execution efficiency caused by manual scheduling in the prior art, the present invention aims to optimize the formulation and execution of maintenance plans through automated and intelligent means, and formulate a set of automated scheduling methods to ensure the scientificity, rationality and efficiency of work plans.

[0017] 2) Enhanced operation monitoring and equipment joint control: In view of the problems of inadequate operation monitoring and poor information transmission in the prior art, which lead to missed operations and inadequate execution, the present invention aims to build a comprehensive information monitoring and equipment joint control system to achieve real-time tracking and monitoring of the operation process and improve the transparency and traceability of the operation. In view of the problems of low efficiency, poor adaptability and scalability of multi-professional collaboration in the prior art, the present invention aims to promote effective collaboration and information sharing among various professions by building an integrated management and control platform, and at the same time design a flexible and scalable system architecture to adapt to the future development of railway technology and the increase in the number of EMUs.

[0018] 3) Strengthen fault warning and response capabilities: In response to the problems of delayed fault response and untimely processing in the existing technology, the present invention aims to integrate advanced fault warning and diagnosis technologies to achieve real-time monitoring and rapid response to vehicle faults, thereby reducing the impact of faults on train operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A schematic flow chart of the steps of a method for integrated control of vehicle maintenance for EMU maintenance according to an embodiment of the present invention; Figure 2 A schematic flow chart of steps for formulating a maintenance plan according to an embodiment of the present invention; Figure 3 A schematic flow chart of the steps of preparing production operation monitoring and equipment joint control according to an embodiment of the present invention; Figure 4 A schematic diagram of the steps of intelligent fault warning and response according to an embodiment of the present invention; Figure 5 A schematic diagram for generating a job plan according to an embodiment of the present invention; Figure 6 This is a schematic diagram of monitoring and managing a first operation according to an embodiment of the present invention; Figure 7 A schematic diagram of monitoring and managing a second operation according to an embodiment of the present invention; Figure 8 A schematic diagram of the third operation monitoring and management according to an embodiment of the present invention; Fig. 9 This is a schematic diagram of tracking the collection of the first operation task according to an embodiment of the present invention; Fig.10 This is a schematic diagram of tracking the collection of the second operation task according to an embodiment of the present invention; Fig.11 A schematic diagram of fault warning and fault handling according to an embodiment of the present invention; Fig.12 This is a flowchart of traditional operations; Fig.13 It is a schematic diagram of the process flow of the dynamic assembly preparation operation according to an embodiment of the present invention; Fig.14 A schematic diagram of a daily plan according to an embodiment of the present invention; Fig.15 A schematic diagram of preparing an electronic map according to an embodiment of the present invention; Fig.16 A schematic diagram of a system according to an embodiment of the present invention; Fig.17 Schematic diagram of job scheduling according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0021] The present invention discloses a method for integrated control of vehicle preparation for EMU preparation, and a schematic flow chart of the steps is shown in FIG. Figure 1 As shown, the following steps are included: S1. Collect the status data of EMU vehicles, including mileage, wear and tear, and fault history, clean and format the collected status data to ensure the accuracy and consistency of the status data, and store the processed data in the database to provide a basis for subsequent analysis; S2. Based on the collected vehicle status data and maintenance process requirements, use big data analysis algorithms to predict future maintenance needs, consider factors such as job priority, resource availability, and job time windows, use machine learning algorithms to automatically generate the optimal maintenance plan, and automatically publish the maintenance plan to relevant operators and managers to ensure information synchronization; S3. Integrate various professional information systems, establish unified data exchange standards and information platforms, use technologies such as RFID and sensors to track the progress of operations in real time, combine GIS to display the location and status of operations, monitor the execution of operation plans in real time, issue early warnings for deviations, and automatically adjust plans to deal with emergencies. Regularly evaluate operation efficiency and quality to provide a basis for continuous improvement; S4. Based on historical fault data, use deep learning and other technologies to train fault prediction models, integrate on-board diagnostic systems, monitor vehicle status in real time, and analyze and compare with fault prediction models. Once an abnormality is found or a fault is predicted, an early warning is triggered immediately, and relevant personnel are notified through the information platform. The maintenance plan is automatically matched according to the fault type, maintenance resources are dispatched, and quick on-site processing is guided; S5. Establish a collaborative operation mechanism, clarify the responsibilities and collaboration processes among different disciplines, promote communication and collaboration among different disciplines through information platforms, ensure timely transmission and sharing of operation information, monitor the use of resources (such as personnel, tools and venues), and make reasonable arrangements and optimize locations; S6. Regularly maintain and upgrade the system to ensure system stability and security, expand system functions and adjust system adaptability according to new vehicle models, new technologies and higher standards of operation requirements, collect user feedback and opinions, and continuously optimize system functions and user experience; S7. Regularly evaluate and analyze the efficiency, quality and cost of maintenance operations. Based on the evaluation results, propose improvement measures and suggestions, continuously optimize the maintenance operation process and management methods, track the implementation effect of improvement measures, form a closed-loop management, and continuously improve the level and efficiency of maintenance operations.

[0022] Specifically, the use of big data analysis algorithms in step S2 to predict future maintenance needs specifically includes the following steps: S211. Clearly define the data parameters for predicting future maintenance big data analysis, and determine the data source of the big data analysis algorithm. The data source includes the train's operating routes, mileage, historical operating on-board data, ticketing faults, maintenance status, maintenance cycle, maintenance history, train history, etc.; the operating routes record the train's operating schedule, operating time and operating cycle; mileage records the train's cumulative mileage and daily mileage; historical operating on-board data records the train's operating status and historical operating fault alarm information; ticketing faults define operating faults and faults found in maintenance operations; maintenance status records the current health status of the train and the maintenance status of the planned maintenance schedule; maintenance cycle records the train's planned maintenance schedule; maintenance history records the train's maintenance status and maintenance time; train history records the status of train components and replacement records.

[0023] S212. Collect data using ETL tools based on the type and characteristics of the data source.

[0024] S213. After collecting data, conduct a preliminary quality check on the data to ensure its completeness, accuracy and consistency.

[0025] S214, preprocessing the data, specifically including: Data cleaning: Process missing values, outliers, and duplicate values. For missing values, mean filling, median filling, regression filling, etc. can be used according to the specific situation; for outliers, they can be identified and processed by setting thresholds; for duplicate values, they can be directly deleted or merged; Data conversion: convert text data into numerical data (such as using one-hot encoding, label encoding, etc.), and standardize or normalize the data to eliminate the impact of different dimensions and magnitudes on data analysis; Data partitioning: Divide the cleaned and converted dataset into training set, validation set, and test set; usually use 70% of the data as the training set, 20% of the data as the validation set, and 10% of the data as the test set.

[0026] S215, performing feature processing on the data, specifically including: Feature selection: Based on domain knowledge and data correlation analysis, select features related to maintenance demand prediction. For example, the usage time of a component, the number of failures, and the maintenance cost may be important prediction features; Feature extraction: extract new features from the original data to improve the predictive ability of the model; for example, statistical features (such as mean, variance, maximum, minimum, etc.) or frequency domain features (such as spectral density, power spectrum, etc.) can be extracted from time series data; Feature dimensionality reduction: If the feature dimension of the data is high, you can use methods such as principal component analysis (PCA), linear discriminant analysis (LDA), or feature selection to reduce the dimension to reduce the computational complexity and improve the generalization ability of the model.

[0027] S216, perform model selection and training, specifically including: Model selection: Choose a suitable model based on the characteristics of the data and the forecasting requirements. For maintenance demand forecasting problems, choose to use models such as neural networks.

[0028] Hyperparameter tuning: Use grid search, random search, or Bayesian optimization to adjust the model’s hyperparameters, such as learning rate, number of iterations, regularization coefficient, etc., to improve the model’s performance.

[0029] Model training: Train the model on the training set to find the optimal model parameters. During the training process, strategies such as cross-validation can be used to prevent overfitting and underfitting.

[0030] S217, model evaluation and tuning, including: Model evaluation: Evaluate the performance of the model on the validation set. Common evaluation indicators include mean square error (MSE), mean absolute error (MAE), accuracy (for classification problems), etc. At the same time, you can also draw learning curves, validation curves and other charts to assist in evaluating model performance.

[0031] Model tuning: Tune the model based on the evaluation results. If the model performance is poor, you can try changing the model, adding features, adjusting hyperparameters, or improving feature engineering.

[0032] Prevent overfitting: During the training process, be careful to prevent the model from overfitting. You can use strategies such as regularization, early stopping, and dropout (for neural networks) to reduce the risk of overfitting.

[0033] S218. Predict and monitor the maintenance status, including: Model deployment: Deploy the trained model to the production environment. The model needs to be converted into a format suitable for the production environment (such as PMML, ONNX, etc.) and integrated into the system clock.

[0034] Real-time prediction: In the production environment, real-time data input is received and deployed models are used to predict maintenance demand. The prediction results can be used to guide the formulation and execution of maintenance plans.

[0035] Monitoring and feedback: Monitor the model prediction results to ensure the stability and accuracy of the model. At the same time, collect user feedback and new data samples to continuously optimize model performance.

[0036] S219. Continuous improvement and optimization of the model; specifically including: Data update: As new data is continuously generated and collected, the training dataset is updated regularly and the model is retrained to improve prediction accuracy.

[0037] Model updates: As machine learning techniques continue to advance and new algorithms emerge, model architectures and algorithms should be regularly evaluated and updated to improve predictive performance.

[0038] Domain knowledge integration: Combine the knowledge and experience of domain experts to further optimize and improve the model. For example, adjust feature selection or introduce new feature engineering methods based on expert advice.

[0039] Through the above steps, a comprehensive and operational big data analysis algorithm model can be constructed to predict future maintenance needs. In practical applications, it is necessary to adjust and optimize according to specific circumstances to meet actual needs.

[0040] Specifically, this automated and intelligent maintenance plan formulation method comprehensively considers factors such as the train's operating routes, operation diagrams, and mileage to automatically formulate an accurate operation and inspection plan. The core of this method is to carefully subcontract and disassemble the balancing repair of the D2 maintenance of the EMU, and integrate it into the daily D1 maintenance without power and with power operations to reduce the downtime of the EMU train and improve operation efficiency. The D2 maintenance and balancing repair is further subdivided into monthly maintenance, quarterly maintenance, semi-annual maintenance, and annual maintenance. The flowchart of the steps in the maintenance plan formulation is shown in the figure below. Figure 2 As shown, the preparation plan in step S2 specifically includes: S221. Collect key data, including running routes, train operation diagrams, mileage, historical fault records and maintenance cycles, etc. Clean the collected key data, remove duplicate, erroneous or invalid data to ensure the accuracy and completeness of the data, and integrate the cleaned data to form a unified data set for subsequent analysis and processing; S222, according to the running route and train operation diagram of the EMU vehicle, determine the running law and running cycle of the train, and according to the running mileage and maintenance cycle of the train, determine the balanced maintenance cycle of D2 maintenance, which specifically includes monthly maintenance, quarterly maintenance, half-year maintenance and annual maintenance; S223, subcontract and disassemble the D2 balance repair to obtain monthly repair, quarterly repair and half-year repair, etc., and merge them into the daily D1 repair work with and without electricity according to the work content; S224. According to the scope of monthly maintenance, quarterly maintenance and semi-annual maintenance, the maintenance scope of monthly maintenance is divided into Package A (vehicle), Package L (vehicle) and Package D (cleaning) according to the profession, the quarterly maintenance is divided into Package B (vehicle) and Package 2C (machine maintenance) according to the profession, and the semi-annual maintenance is divided into Package C (vehicle); S225. The subcontracting schemes generated in step S224 are systematically and intelligently integrated to form a J package. The longest cycle of the J package balanced maintenance plan is 24 subcontracts, covering a half-year maintenance period. The subcontracting method of the J package is based on a cycle of six days. Four cycles include monthly maintenance content, and six days are reserved as a mitigation period. If there are weekends and holidays, the maintenance period will be postponed. The subcontracting cycle is recorded, the subcontracting is automatically arranged and calculated, and the execution of the subcontracting is monitored. If it is necessary to transfer the weekly plan, the weekly plan is automatically adjusted, and the daily operation plan is automatically generated based on the monthly plan and the weekly plan; S226. The system automatically generates a monthly plan based on the logic algorithm. The monthly plan is published after being confirmed by the maintenance scheduling and notified to each work team. The weekly plan and daily plan are automatically generated based on the monthly plan. The daily plan automatically generates the daily plan for tomorrow before 12:00 every day. If each work team has a temporary plan, it needs to submit the temporary plan in the system before 12:00 every day. The system automatically integrates the temporary plan when generating the plan for tomorrow; S227. The generated daily plan is automatically sent to each work team after being confirmed by the preparation and scheduling. Each work team makes personnel scheduling and work preparation work according to the daily plan; S228. The system automatically monitors and updates the execution of the daily plan, accumulates the monthly execution status, provides data support for train plan formulation, and continuously optimizes the scheduling of automatic plans.

[0041] Exemplarily, the maintenance plan arranges the operation and inspection plan according to the factors such as the operation route, train operation diagram, and mileage. The D2 maintenance and balancing maintenance are subcontracted and disassembled, and merged into the daily D1 maintenance without electricity and electricity operation. The D2 maintenance and balancing maintenance are divided into monthly maintenance, quarterly maintenance, semi-annual maintenance and annual maintenance. Among them, the monthly, quarterly and semi-annual maintenance are split by time. The annual maintenance is based on the time and mileage that come first, combined with the four rounds of centralized maintenance. Monthly maintenance includes special projects such as power-centralized EMU wheel pair identification, bogie depth inspection, electrical integrated control cabinet inspection, brake system drainage and dust removal, special brake test train network inspection, filter cleaning, lower body and bogie wiping. According to the definition of each post operation, it is divided into A, L, and D packages, and divided into 4 operation packages according to the number of vehicles. The maintenance needs to be completed within one month. Quarterly repairs include electric water heaters, electric water heaters, sliding doors, toilets, batteries and boxes, DC600V power supply devices under the vehicle, safety loops, catering kitchen equipment, summit inspection and maintenance. According to the subcontract definition, it is divided into 12 work packages and needs to be renovated within one quarter. Semi-annual repairs include double-layer folding shed windshields, according to the subcontract definition, it is divided into 4 work packages and needs to be renovated within one quarter. Annual repairs include axle box devices, brake devices, wheelset inspection and maintenance, which are carried out in a centralized manner. The specific unpacking method is shown in Table 1: Table 1: Specific unpacking methods

[0042]

[0043]

[0044]

[0045] The J package is formed by integrating all subcontracts in a systematic and intelligent manner. The longest period of the J package's balanced maintenance plan is 24, covering monthly, quarterly, and annual balanced maintenance plans. In addition, the C1-package is the locomotive operation content, which is divided into 6 packages, covering the locomotive quarterly maintenance, which is mainly completed through running kilometers, and the quarterly maintenance is completed from 70,000 to 100,000 kilometers, and the quarterly maintenance subcontract is merged into the daily operation.

[0046] The J package subcontracting method is based on a 6-day cycle, which is postponed when encountering weekends. The average package cycle is recorded, and the monthly plan is automatically generated according to the J package whole package plan. The weekly and daily execution plans are generated based on the monthly plan to form a daily operation plan for tracking and execution. The subcontracting plan is shown in Table 2: Table 2: Subcontracting Scheme

[0047]

[0048]

[0049] By introducing big data analysis and machine learning technologies, we can intelligently predict future maintenance needs based on historical operating data, fault records, maintenance cycles, etc., dynamically adjust maintenance plans, monitor train operation status in real time, respond quickly to sudden faults or abnormal situations, and automatically adjust maintenance plans to ensure efficient use of maintenance resources.

[0050] According to the monthly plan, weekly plan and work team, the daily preparation plan is automatically generated according to the preparation operation cycle and the standard operation time of each operation link. The next day's daily production plan is generated in advance and released in advance to notify each work team to prepare for the operation in advance. The daily plan diagram is as follows Fig.14 shown.

[0051] Specifically, the optimal maintenance plan includes a monthly maintenance plan, a weekly maintenance plan and a daily production plan. Through logical parameters and the established algorithm model, the monthly maintenance plan, the weekly maintenance plan and the daily production plan are automatically generated by inputting data, and the generated plan is automatically sent to the maintenance scheduling personnel for confirmation and then released. The step S2 in which the optimal maintenance plan is automatically generated by the machine learning algorithm specifically includes the following steps: S231, data preprocessing, specifically including: Collect maintenance data: collect data on predicted future maintenance status, monthly overall plan and plan execution, work team scheduling, holiday time, balanced maintenance cycle, balanced maintenance execution, train status information, train maintenance status, etc.

[0052] Data cleaning: Clean the collected data to remove duplicate, erroneous and invalid data to ensure data accuracy and consistency.

[0053] Data conversion and standardization: Convert unstructured data into structured data and standardize numerical data for subsequent model training.

[0054] S232. Feature engineering, including: Feature selection: Select key features that have an impact on the generation of maintenance plans from the cleaned data, such as train mileage, train status information, train maintenance status, overall plan, and balanced maintenance execution status.

[0055] Feature extraction: Extract new features through statistical methods or domain knowledge, such as the degree of wear and failure frequency of vehicle parts.

[0056] Feature dimensionality reduction: If the feature dimension is too high, it can be reduced through methods such as principal component analysis (PCA) to improve model training efficiency.

[0057] S233, model selection and training, specifically including: Select machine learning algorithms: According to the complexity of the problem and the characteristics of the data, select decision trees, random forests, and support vector machines (SVM) to establish algorithm models for vehicle component prediction and health status. For maintenance plan generation, select algorithms such as logistic regression and gradient boosting tree (GBDT) to establish an optimal maintenance plan algorithm model.

[0058] Model training: Use the preprocessed data and the selected algorithm to train the model. During the training process, you can adjust the model parameters through methods such as cross-validation to improve the generalization ability of the model.

[0059] Model evaluation: Evaluate the trained model through indicators such as accuracy, recall, and F1 score to ensure the effectiveness of the model.

[0060] S234, generating an optimal maintenance plan, specifically including: Generate maintenance plan: Import the train data to be maintained into the trained model and output the optimal maintenance plan. The maintenance plan should include information such as maintenance time, operation time of each team, and operation content.

[0061] Plan optimization: Optimize the generated maintenance plan according to constraints such as maintenance time and balanced maintenance plan to ensure the feasibility and economy of the plan.

[0062] Result feedback and iteration: The generated maintenance plan will be fed back to the maintenance scheduling personnel, feedback will be collected, and the model will be iteratively optimized based on the feedback to improve the accuracy and practicality of the maintenance plan.

[0063] When choosing a specific machine learning algorithm, factors such as the characteristics of the data, the complexity of the problem, and computing resources need to be considered. Usually, the performance of different algorithms on specific tasks can be evaluated through methods such as cross-validation, so as to select the most appropriate algorithm. In addition, combining professional knowledge in the field of train maintenance, interpreting and optimizing the output of the algorithm is also a key step in generating the optimal maintenance plan.

[0064] In the process of generating the optimal maintenance plan, the optimal maintenance plan can be obtained through the formula: Optimal_Maintenance_Plan = f(Train_Data, Features, Constraints,Algorithms) Among them, Optimal_Maintenance_Plan: optimal maintenance plan, which is our target output, including monthly maintenance plan, weekly maintenance plan and daily production plan.

[0065] Train_Data: training data, i.e. historical maintenance data used for model training, including train status information, maintenance records, mileage, failure rate, etc.

[0066] Features: Features are key factors extracted from the training data that affect the generation of maintenance plans, such as train mileage, component wear, failure frequency, and balancing repair cycle.

[0067] Constraints: Constraints that need to be met during the maintenance plan generation process, such as maintenance time cycle, team scheduling, holiday time, balanced maintenance plan, etc.

[0068] Algorithms: Algorithms are machine learning algorithm models used to generate maintenance plans. The algorithm model will generate the optimal maintenance plan based on the input training data, features, and constraints.

[0069] Specifically, the monitoring and equipment joint control method for the preparation of production operations realizes the refined management of the entire process from job preparation to job completion. This method integrates a variety of technologies and equipment, including AEI recognition equipment, sensors, face recognition equipment, handheld terminals, workstation terminals, etc., and combines the Internet of Things, big data, artificial intelligence and other technologies to collect, monitor, predict and remind the status of production operations in real time. This method effectively improves the accuracy, efficiency and safety of production operations, and provides strong support for subsequent production operations and quality management. The flowchart of the steps for the preparation of production operation monitoring and equipment joint control is shown below. Figure 3 As shown, step S3 specifically includes the following steps: S31, when the EMU train enters the depot, the AEI identification device automatically identifies the train arrival and the train model and number. The system collects the information of the AEI device, matches it with the train daily plan, and automatically generates the operation process; S32. Through the Internet of Things technology, the system controls the electric garage door and the driving signal light, determines whether the surrounding environment of the electronic garage door is safe, automatically opens the garage door, and turns the open signal light green. After the train enters the garage and goes to the platform, it automatically determines and closes the garage door, and turns off the signal light (set to red); S33. The corresponding work team uses the in-warehouse workstation terminal for face recognition and applies for power-off operation. The approval of the power-off operation is automatically transferred to the on-site dispatch and integrated duty room for approval and authorization. After approval, the system jointly controls the in-warehouse voice dialogue system and the in-warehouse workstation terminal to remind the start of the operation. S34. The system automatically notifies relevant work teams to carry out joint work according to the preset work process. Each team member receives work tasks and specific guidance information through handheld terminals or workstation terminals. The system uses sensors, video surveillance, image algorithms and AI artificial intelligence to conduct all-round monitoring of the work area. S35. If an abnormal situation occurs, the system will immediately sound an alarm and notify relevant personnel to take emergency measures through the voice dialogue system. After the operation is completed, each team member completes the operation through face recognition on a handheld terminal or workstation terminal. The system automatically reviews the operation status, eliminates the operation authority after confirmation, and broadcasts the completion information of the operation; S36. If there is any equipment failure or abnormality during the operation, the operator can quickly report the fault information through the handheld terminal or the workstation terminal. After receiving the fault information, the system will automatically notify the relevant maintenance personnel to handle it and track the progress of the fault handling in real time; S37. After all operations are completed, the system automatically generates a train certificate and operation report based on the operation report and fault handling report submitted by each team. The certificate and operation report contain detailed information such as train information, operation content, operation time and operation personnel.

[0070] For example, the flow chart of traditional operation is as follows: Fig.12 As shown; the schematic diagram of the dynamic set preparation operation process of the present invention is as shown Fig.13 As shown, the monitoring and execution tracking of the preparation and production operations integrates AEI recognition equipment, sensors, face recognition equipment, handheld terminals, workstation terminals and other equipment, combines the Internet of Things, big data, artificial intelligence and other technologies, to collect, monitor, predict and remind the production operation status in real time, accurately track the operation status, and effectively improve the accuracy and efficiency of production operations.

[0071] When a train enters the depot, the AEI identification device automatically determines the arrival of the train and identifies the train model and number. The system collects the AEI equipment information and automatically generates the operation process by matching the train daily plan. The operation process is pushed to the scheduling page, and the scheduling page and the large screen automatically display the train entry status.

[0072] Through the Internet of Things technology, the system controls the electric garage door, and uses image algorithms to determine whether the surrounding environment of the electronic garage door is safe, automatically opens the electric garage door, and controls the driving signal light at the same time. The open signal light is green, and the train driver drives the train into the garage according to the signal light and the garage door opening status. When the train enters the garage and goes to the platform, the system tests the arrival of the train through the ultrasonic sensor installed in the garage, and automatically determines whether the train is on the platform. The system controls the garage door to close according to the status of the train on the platform, and turns off the signal light at the same time (sets the signal light to red). The system controls the voice dialogue system in the garage to announce that the train has entered the platform and prepares to generate operations. The system pushes the train status to the dispatching page and displays the train status on the large screen.

[0073] Through voice broadcast, system reminders and daily plan arrangements, the corresponding work team performs facial recognition through the workstation terminal in the warehouse and applies for power-off operations. The approval of power-off operations is automatically transferred to the on-site dispatch and integrated duty room. After approval and authorization, the system jointly controls the voice dialogue system in the warehouse and the workstation terminal in the warehouse to remind them to start the operation. At the same time, the train status is pushed to the dispatching page and the large screen displays the train status. Each position and decision-making management can monitor the production operation status in real time through the system.

[0074] After the power-off operation is approved, the system controls the safety interlocking equipment to power off the contact network in the depot, automatically obtains the power-off status of the contact network, and pushes the contact network status to the scheduling page and large-screen display. The system automatically notifies the relevant work teams to carry out joint operations according to the preset operation process. Each team member receives the operation tasks and specific guidance information, including operation content, operation sequence, safety precautions, etc. through handheld terminals or workstation terminals. The system's image algorithm and video monitoring track the progress of each team's operations in real time to ensure that each operation link is closely connected without omissions. During the operation, the system uses sensors, video monitoring, image algorithms, and AI artificial intelligence to monitor the operation area in all directions to ensure operation safety. In the event of an abnormal situation, such as personnel entering a dangerous area or abnormal operation of equipment, the system immediately issues an alarm and notifies relevant personnel to take emergency measures through the voice dialogue system.

[0075] After the operation is completed, each team member completes the operation through face recognition on a handheld terminal or workstation terminal. The system automatically reviews the operation status, and the system assists in decision-making by providing a suggestion to the dispatcher for decision-making. After confirming that the operation is correct, the operation authority is automatically eliminated to ensure that subsequent operation teams cannot repeat the same operation, avoiding operation conflicts and safety hazards. At the same time, the system controls the voice dialogue system in the warehouse, broadcasts the completion information of the operation, and updates the dispatching page and the train status on the large screen to prepare for the next operation link.

[0076] During the operation, if there is equipment failure or abnormal situation, the operator can quickly report the fault information through the handheld terminal or workstation terminal, including the fault location, fault phenomenon, preliminary judgment, etc. After receiving the fault information, the system automatically notifies the relevant maintenance personnel to handle it, and tracks the progress of the fault handling in real time through the Internet of Things technology. After the maintenance personnel handle the fault, they need to submit a fault handling report, which will be reviewed and confirmed by the system. After the fault handling is completed, the system updates the train status and restores the relevant operation permissions to ensure the smooth progress of production operations.

[0077] After all operations are completed, the system automatically generates a train certificate and operation report based on the operation reports and fault handling reports submitted by each team. The certificate contains detailed information such as train information, operation content, operation time, and operator, and the operation report records in detail the key nodes, operation conditions, fault handling conditions, etc. in the entire operation process. The system sends the certificate and operation report to relevant decision-making managers for review through the Internet of Things technology. After the review is passed, the system automatically stores the certificate and operation report in the database, and can be viewed and printed through the large screen or system interface, providing strong support for subsequent production operations and quality management. Through the above-mentioned monitoring and equipment joint control measures, the preparation production operation realizes the refined management of the entire process from operation preparation to operation completion, effectively improving the accuracy, efficiency and safety of production operations.

[0078] Specifically, the intelligent fault warning and response method is a comprehensive solution that integrates advanced sensor technology, the Internet of Things, big data analysis, machine learning, and artificial intelligence. It aims to monitor the equipment or system in real time, warn of faults, and respond quickly to them to improve operational efficiency and safety. Through real-time data collection and transmission, real-time monitoring and fault warning of equipment or systems are achieved. By using big data analysis technology and machine learning algorithms, the collected data is deeply mined and analyzed to improve the accuracy of fault identification. Through automated response processes and remote monitoring functions, the efficiency and timeliness of fault handling are improved. The system has good scalability and can be customized and expanded according to actual needs. The flowchart of the steps of intelligent fault warning and response is shown below. Figure 4 As shown, the intelligent fault warning and response in step S4 specifically includes the following steps: S411. Collect key train operation data in real time through sensors and on-board data, and clean the collected data to improve the accuracy and efficiency of subsequent data analysis; S412. Use big data analysis technology to conduct in-depth mining and analysis on the cleaned data, identify potential fault signs and abnormal patterns, and train machine learning models to make intelligent judgments and predictions on the operating status of equipment or systems. When the model identifies a fault or abnormality, it will output a corresponding warning signal. S413. Set preset warning thresholds and preset conditions. The preset warning thresholds and preset conditions are preset ranges according to the working environment and are not limited to a specific range. When the abnormality identified by the data analysis layer reaches the preset warning threshold, the warning mechanism is immediately triggered, and the warning information is promptly conveyed to relevant personnel through voice broadcast and system reminders. The warning information includes key information such as fault type, fault location and severity, so that relevant personnel can quickly take corresponding measures; S414. The system also includes an emergency handling guide library to provide maintenance personnel with references and suggestions for fault handling. For predictable faults or abnormal conditions, the system initiates an automated response process, including automatic shutdown or switching to backup equipment, to reduce the impact of faults on equipment or system operation; S415. After receiving the warning information, the maintenance personnel shall take response measures according to the emergency handling guide and the actual situation, including repairing the fault or eliminating the abnormality; S416. Based on historical fault data, establish a fault prediction model to predict the future operating status of the equipment or system. According to the output results of the fault prediction model, formulate a preventive maintenance plan in advance and conduct regular inspection and maintenance of key components; S417. The system continuously adapts to changes in the operating environment of equipment and systems through continuous learning algorithms, optimizes prediction models and algorithms, improves the accuracy and timeliness of early warnings, collects user feedback and system operation data, and continuously optimizes and improves the system to improve system stability and reliability.

[0079] For example, the system collects train status in real time through sensors, on-board download data, and personnel operation monitoring, and uses image recognition, voice recognition and other technologies to monitor production operations in real time. The system uses big data analysis technology and machine learning algorithms to conduct in-depth mining of the collected data and identify the vehicle status and production operation status. When a fault or abnormal situation occurs, the system immediately triggers the early warning mechanism and pushes early warning information to ensure that the information is transmitted immediately. The system has built-in emergency handling guidelines and automated response processes to guide maintenance personnel to take countermeasures quickly to reduce the impact of faults on train operations. Based on historical fault data and expert knowledge, the system establishes a prediction model for the health status of the vehicle, predicts the types and time of possible future faults, and reduces the occurrence of sudden faults by formulating preventive maintenance plans in advance, thereby improving the reliability and stability of the vehicle. The system continuously optimizes the prediction model and algorithm through continuous learning and adaptation to changes in the vehicle's operating environment, improves the accuracy and timeliness of early warnings, supports user feedback and the collection and analysis of system operation data, and provides data support for the continuous improvement and optimization of the system.

[0080] The Internet of Things technology is used to realize the interconnection between vehicles and monitoring systems, collect vehicle operation data in real time and transmit it to the monitoring center for analysis and processing. Big data analysis technology is used to deeply mine and analyze the collected data, identify potential fault modes and trends, and use machine learning algorithms to train fault prediction models to improve the accuracy and timeliness of fault warnings. The integration of artificial intelligence technology and expert systems enables the system to simulate the reasoning behavior of experts, make intelligent judgments and decisions on vehicle faults, support natural language processing and human-computer interaction technology, and facilitate communication and exchange between users and the system. Cloud computing technology is used to achieve efficient storage and processing of data, support the fault warning and response needs of large-scale vehicles, and edge computing technology is used to bring data processing closer to the device end, reduce delays and improve the response speed and reliability of the system. Provide a visual monitoring interface and operation interface to facilitate users to understand the vehicle operation status and fault warning information in real time, support remote monitoring and diagnosis functions, improve maintenance efficiency and reduce labor costs. The intelligent fault warning and response system integrates advanced sensors, the Internet of Things, big data, artificial intelligence and other technical means to achieve real-time monitoring, rapid positioning and efficient processing of vehicle faults, effectively improving the operation safety and reliability of trains.

[0081] Specifically, in step S417, the system continuously optimizes the prediction model and algorithm through continuous learning, including the following specific steps: S4171. New data collection and annotation, including: As the vehicle operates and the environment changes, new data is automatically and continuously collected and annotated. The new data includes the health status information of the vehicle in different operating environments; S4172, Model retraining and updating, including: Integrate new data into the existing training set and retrain the model. During the retraining process, gradually introduce new data to avoid causing too much impact on the model. Based on the performance evaluation results of the retrained model, decide whether to update the model parameters and architecture. S4173, Online learning and adaptation, specifically including: Use online learning algorithms to enable the model to continuously learn and adapt to new data and environmental changes during actual operation; monitor the model's predictive performance and trigger the model update mechanism in a timely manner when performance degradation is found; S4174. User feedback data collection, including: Collect user feedback on the prediction results through user feedback channels, including the website's built-in feedback channel, social media, and customer service hotline; the feedback includes the accuracy and timeliness of the prediction results and the user's experience; S4175, system optimization and iteration, including: Optimize and improve the system based on user feedback and data analysis results; continuously optimize prediction models and algorithms to improve the accuracy and timeliness of early warnings; iteratively update system functions and interface design to enhance user experience.

[0082] Through the above steps, the system can continuously learn and adapt to changes in the vehicle's operating environment, continuously optimize prediction models and algorithms, and improve the accuracy and timeliness of railway power centralized vehicle health status prediction. At the same time, by supporting user feedback and the collection and analysis of system operation data, it provides data support for the continuous improvement and optimization of the system.

[0083] Specifically, deep learning technology can automatically extract and analyze features in data, and perform efficient data classification and prediction. By collecting a large amount of train operation data and using deep learning technology to build a train component fault prediction model, the accuracy of fault prediction can be significantly improved. For example, using a convolutional neural network (CNN) to analyze the train operation status image, or using a recurrent neural network (RNN) and its variants (such as LSTM, GRU) to process time series data, potential fault modes can be effectively identified. Deep learning technology has a strong real-time processing capability and can monitor and warn the operating status of train components in real time. By deploying a fault prediction model based on deep learning, a warning signal can be issued in time before a fault occurs, providing maintenance personnel with a valuable time window, thereby effectively avoiding or reducing the impact of the fault on train operation. By predicting train component faults through deep learning technology, maintenance plans can be formulated more accurately. According to the prediction results, preventive maintenance can be performed on key components to avoid unnecessary maintenance and replacement, thereby reducing operating costs. At the same time, maintenance resources can be reasonably allocated according to the type and severity of the fault to improve maintenance efficiency. The use of deep learning technology to train the fault prediction model in step S4 specifically includes the following steps: S421. Data preparation, including: Data collection: Collect large amounts of task-related raw data from relevant sources (such as databases, sensors, networks, etc.).

[0084] Data cleaning: Process outliers, missing values, duplicate values, etc. in the data to ensure data quality.

[0085] Data annotation: providing labels or target values ​​for supervised learning tasks such as classification and regression.

[0086] Data segmentation: Divide the dataset into training set, validation set and test set, usually in a ratio of 70%-80% (training set), 10%-15% (validation set), and 10%-15% (test set).

[0087] Data preprocessing: including data standardization, normalization, feature extraction, feature selection, etc., to reduce the learning burden of the model and improve performance.

[0088] S422. Establishing a data model, specifically including: Select model architecture: Choose a suitable deep learning model (such as CNN, RNN, LSTM, Transformer, etc.) according to the task type (such as image recognition, natural language processing, time series prediction, etc.).

[0089] Define the model: Use a deep learning framework (such as TensorFlow, PyTorch, etc.) to define the structure of the model, including the input layer, hidden layer, output layer, activation function, loss function, etc.

[0090] S423, model training, specifically including: Data input: Input the preprocessed training data into the model.

[0091] Compile the model: Set the optimizer (such as Adam, SGD, etc.) and hyperparameters such as learning rate in the deep learning framework, and compile the model.

[0092] Iterative training: Calculate the prediction value and loss through forward propagation, and then update the weight and bias of the model through back propagation. This process is repeated many times (called epochs) until the predetermined number of training rounds is reached or the loss converges.

[0093] S424, Model verification and tuning, including: Performance evaluation: Evaluate the performance of the model on the validation set. Common evaluation indicators include accuracy, precision, recall, F1 score, mean square error, etc.

[0094] Hyperparameter tuning: Adjust the model’s hyperparameters (such as learning rate, batch size, number of layers, number of nodes, etc.) based on the performance of the validation set to improve model performance.

[0095] Regularization and Optimization: Use regularization techniques (such as L1, L2 regularization, Dropout, etc.) to prevent overfitting, and use optimization algorithms (such as momentum, RMSprop, etc.) to speed up the training process.

[0096] S425, Model evaluation and optimization, including: Performance evaluation: Evaluate the performance of the model on the test set, using evaluation indicators including accuracy, precision, recall, F1 score, etc.

[0097] Model optimization: If the model performance does not meet the requirements, optimize it by adjusting the model structure, hyperparameters, increasing the amount of data, improving data preprocessing, etc.

[0098] Model fusion: Combine the prediction results of multiple models through ensemble learning (such as Bagging and Boosting) to improve the overall performance.

[0099] S426, model loading and application, specifically including: Model export: Export the trained model to a deployable format (such as TensorFlow SavedModel, PyTorch model file, etc.).

[0100] Model deployment: Deploy the model to a production environment.

[0101] Performance monitoring: Monitor the performance of the model in real-world applications and update and optimize it as needed.

[0102] Data collection and annotation: Continuously collect new data and annotate it to enrich the training set.

[0103] Model iteration: Continuously iterate and optimize the model based on new data and feedback to improve its performance and accuracy.

[0104] The entire process of training a deep learning model is an iterative and optimization process that requires continuous adjustment of the model structure, hyperparameters, and data preprocessing strategies to obtain the best performance and accuracy.

[0105] The present invention also builds an integrated information platform, combining automated and intelligent technical means to achieve full-process and all-round control of the dynamic assembly preparation operation. Through real-time data collection, intelligent analysis and decision-making, automated operation execution and monitoring, the efficiency, accuracy and coordination of the preparation operation are improved. The schematic diagram of the operation plan generation is shown in the figure. Figure 5 As shown; the first operation monitoring and management diagram is as follows Figure 6 As shown; the second operation monitoring and management diagram is as follows Figure 7 As shown; the third operation monitoring and management diagram is as follows Figure 8 As shown; the first task collection tracking diagram is as follows Fig. 9 As shown; the second task collection tracking diagram is as follows Fig.10 As shown; the fault warning and fault handling diagram is as shown Fig.11 As shown; the schematic diagram of the preparation electronic map is as follows Fig.15 As shown; the system diagram is as Fig.16 As shown; the job scheduling diagram is as follows Fig.17 shown.

[0106] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A method for integrated control of vehicle preparation for EMU preparation, characterized in that: The following steps are involved: S1. Collect the status data of the EMU vehicle, including mileage, wear and tear, and fault history, clean and format the collected status data to ensure the accuracy and consistency of the status data, and store the processed data in the database to provide a basis for subsequent analysis; S2. Based on the collected vehicle status data and maintenance process requirements, use big data analysis algorithms to predict future maintenance needs, consider job priorities, resource availability, and job time windows, use machine learning algorithms to automatically generate the optimal maintenance plan, and automatically publish the maintenance plan to relevant operators and managers to ensure information synchronization; S3. Integrate various professional information systems, establish unified data exchange standards and information platforms, use RFID and sensors to track the progress of operations in real time, combine GIS to display the location and status of operations, monitor the execution of operation plans in real time, issue early warnings for deviations, and automatically adjust plans to deal with emergencies. Regularly evaluate operation efficiency and quality to provide a basis for continuous improvement; S4. Based on historical fault data, deep learning technology is used to train fault prediction models, integrate on-board diagnostic systems, monitor vehicle status in real time, and analyze and compare with fault prediction models. Once an abnormality is found or a fault is predicted, an early warning is triggered immediately, and relevant personnel are notified through the information platform. Maintenance plans are automatically matched according to the fault type, maintenance resources are dispatched, and on-site rapid processing is guided; S5. Establish a collaborative operation mechanism, clarify the responsibilities and collaboration processes among different disciplines, promote communication and collaboration among different disciplines through information platforms, ensure timely transmission and sharing of operation information, monitor the use of resources, and make reasonable arrangements and optimize positions; S6. Regularly maintain and upgrade the system to ensure system stability and security, expand system functions and adjust system adaptability according to new vehicle models, new technologies and demanding operation requirements, collect user feedback and opinions, and continuously optimize system functions and user experience; S7. Regularly evaluate and analyze the efficiency, quality and cost of maintenance operations. Based on the evaluation results, propose improvement measures and suggestions, continuously optimize the maintenance operation process and management methods, track the implementation effect of improvement measures, form a closed-loop management, and continuously improve the level and efficiency of maintenance operations.

2. The integrated vehicle preparation control method for EMU preparation according to claim 1 is characterized in that: The use of big data analysis algorithms in step S2 to predict future maintenance needs specifically includes the following steps: S211. Clarify the data parameters for predicting the future maintenance big data analysis and determine the data source of the big data analysis algorithm, wherein the data source includes the train's running routes, mileage, historical running on-board data, ticket failures, maintenance status, maintenance cycle, maintenance history and train history; The operation routes include the running frequency, running time and running cycle; The running kilometers include the cumulative running kilometers and daily running kilometers of the train; The historical running on-board data includes the running status of the train and historical running fault alarm information; The ticketing failures are operational failures and failures found during maintenance work; The maintenance status includes the health status of the current train and the maintenance status of the planned maintenance process; The maintenance cycle is the planned maintenance schedule of the train; The maintenance history includes the maintenance status and maintenance time of the train; The train history includes the status and replacement records of train components; S212. Collect data through ETL tools according to the type and characteristics of data sources; S213. Conduct preliminary quality checks on the data after collection to ensure the completeness, accuracy and consistency of the data; S214, preprocessing the data, specifically including: Data cleaning: handle missing values, outliers and duplicate values; for missing values, use mean filling, median filling and regression filling according to the specific situation; for outliers, identify and process them by setting thresholds; for duplicate values, delete or merge them directly; Data conversion: convert text data into numerical data, standardize or normalize the data to eliminate the impact of different dimensions and magnitudes on data analysis; Data division: Divide the cleaned and converted data set into training set, validation set and test set; use 70% of the data as the training set, 20% of the data as the validation set, and 10% of the data as the test set; S215, performing feature processing on the data, specifically including: Feature selection: Based on domain knowledge and data correlation analysis, select features related to maintenance demand forecasting; Feature extraction: extract new features from raw data to improve the predictive power of the model; Feature dimensionality reduction: If the feature dimension of the data is high, use principal component analysis, linear discriminant analysis or feature selection to reduce the dimension to reduce the computational complexity and improve the generalization ability of the model; S216, perform model selection and training, specifically including: Model selection: Select a model based on the characteristics of the data and the forecast requirements. For the maintenance demand forecast problem, choose to use a neural network model; Hyperparameter tuning: Use grid search, random search, or Bayesian optimization to adjust the model's hyperparameters, including learning rate, number of iterations, and regularization coefficient, to improve model performance; Model training: Train the model on the training set to find the optimal model parameters; also include preventing overfitting and underfitting by using cross-validation during the training process; S217, model evaluation and tuning, including: Model evaluation: Evaluate the performance of the model on the validation set; evaluation indicators include mean square error, mean absolute error and accuracy; at the same time, draw learning curves and validation curves to assist in evaluating model performance; Model tuning: Tune the model based on the evaluation results. If the model performance is poor, adjust it by replacing the model, adding features, adjusting hyperparameters, or improving feature engineering. Prevent overfitting: During the training process, regularization, early stopping, and dropout are used to reduce the risk of overfitting. S218. Predict and monitor the maintenance status, including: Model deployment: Deploy the trained model to the production environment. The model needs to be converted into a format suitable for the production environment and integrated into the system clock. Real-time prediction: In the production environment, real-time data input is received and deployed models are used to predict maintenance demand. The prediction results are used to guide the formulation and execution of maintenance plans. Monitoring and feedback: Monitor the model prediction results to ensure the stability and accuracy of the model. At the same time, collect user feedback and new data samples to continuously optimize model performance. S219. Continuous improvement and optimization of the model; specifically including: Data update: As new data is continuously generated and collected, the training dataset is updated regularly and the model is retrained to improve prediction accuracy; Model Update: As machine learning techniques continue to evolve and new algorithms emerge, model architecture and algorithms are regularly evaluated and updated to improve predictive performance. Domain knowledge fusion: Combine knowledge and experience in this field to further optimize and improve the model.

3. The integrated vehicle preparation control method for EMU preparation according to claim 1 is characterized in that: The maintenance plan in step S2 specifically includes: S221. Collect key data, including running routes, train operation diagrams, mileage, historical fault records and maintenance cycles, clean the collected key data, remove duplicate, erroneous or invalid data to ensure the accuracy and completeness of the data, and integrate the cleaned data to form a unified data set for subsequent analysis and processing; S222, according to the running route and train operation diagram of the EMU vehicle, determine the running law and running cycle of the train, and according to the running mileage and maintenance cycle of the train, determine the balanced maintenance cycle of D2 maintenance, which specifically includes monthly maintenance, quarterly maintenance, half-year maintenance and annual maintenance; S223, subcontract and disassemble the D2 balance repair to obtain monthly repair, quarterly repair and semi-annual repair, and merge them into the daily D1 repair work with and without electricity according to the work content; S224. According to the scope of work of monthly repair, quarterly repair and half-year repair, monthly repair is divided into A package, L package and D package according to profession, quarterly repair is divided into B package and 2C package according to profession, and half-year repair is divided into C package; S225. The subcontracting schemes generated in step S224 are systematically and intelligently integrated to form a J package. The longest period of balanced maintenance of the J package is 24 subcontracts, covering a half-year maintenance period. The subcontracting method of the J package is based on a period of six days. Four periods include monthly maintenance content, and six days are reserved as a mitigation period. If there are weekends and holidays, the period is postponed. The subcontracting period is recorded, the subcontracting is automatically arranged and calculated, and the execution of the subcontracting is monitored. If it is necessary to transfer the weekly plan, the weekly plan is automatically adjusted, and the daily operation plan is automatically generated based on the monthly plan and the weekly plan; S226. The system automatically generates a monthly plan based on the logic algorithm. The monthly plan is published after being confirmed by the maintenance scheduling and notified to each work team. The weekly plan and daily plan are automatically generated based on the monthly plan. The daily plan automatically generates the daily plan for tomorrow before 12:00 every day. If each work team has a temporary plan, it needs to submit the temporary plan in the system before 12:00 every day. The system automatically integrates the temporary plan when generating the plan for tomorrow; S227. The generated daily plan is automatically sent to each work team after being confirmed by the preparation and scheduling. Each work team makes personnel scheduling and work preparation work according to the daily plan; S228. The system automatically monitors and updates the execution of the daily plan, accumulates the monthly execution status, provides data support for train plan formulation, and continuously optimizes the scheduling of automatic plans.

4. The integrated vehicle preparation control method for EMU preparation according to claim 3 is characterized in that: The step S2 in which the optimal maintenance plan is automatically generated using a machine learning algorithm specifically includes the following steps S231, data preprocessing, specifically including: Collect maintenance data: collect and predict future maintenance status data, monthly overall plan and plan execution status, work team scheduling, holiday time, balanced maintenance cycle, balanced maintenance execution status, train status information and train maintenance status; Data cleaning: Clean the collected data to remove duplicate, erroneous and invalid data to ensure data accuracy and consistency; Data conversion and standardization: convert unstructured data into structured data and standardize numerical data for subsequent model training; S232. Feature engineering, including: Feature selection: Select key features that have an impact on the generation of maintenance plans from the cleaned data, including train mileage, train status information, train maintenance status, overall plan, and balanced maintenance execution status; Feature extraction: Extract new features through statistical methods or domain knowledge, including the wear degree and failure frequency of vehicle components; Feature dimensionality reduction: If the feature dimension is too high, principal component analysis is used to reduce the dimension to improve model training efficiency; S233, model selection and training, specifically including: Select machine learning algorithms: select decision tree, random forest or support vector machine algorithms to establish vehicle component prediction and health status algorithm models according to the complexity of the problem and the characteristics of the data; select logistic regression and gradient boosting tree to establish the optimal maintenance plan algorithm model for maintenance plan generation; Model training: Use preprocessed data and the selected algorithm to train the model. During the training process, adjust the model parameters through cross-validation to improve the generalization ability of the model. Model evaluation: Evaluate the trained model through accuracy, recall and F1 score to ensure the effectiveness of the model; S234, generating an optimal maintenance plan, specifically including: Generate maintenance plan: import the train data to be maintained into the trained model and output the optimal maintenance plan, which includes the maintenance time, the operation time and operation content of each team; Plan optimization: optimize the generated maintenance plan according to the maintenance time and balanced maintenance plan to ensure the feasibility and economy of the plan; Result feedback and iteration: Feedback the generated maintenance plan to the maintenance scheduling personnel, collect feedback, and iterate and optimize the model based on the feedback; The model generates the optimal maintenance plan through the following formula: Optimal_Maintenance_Plan = f(Train_Data, Features, Constraints,Algorithms); Among them, Optimal_Maintenance_Plan is the optimal maintenance plan, including monthly maintenance plan, weekly maintenance plan and daily production plan; Train_Data is the training data, which is the historical maintenance data used for model training, including train status information, maintenance records, mileage and failure rate; Features are the key factors that are extracted from the training data and have an impact on the generation of the maintenance plan, including train mileage, component wear, failure frequency, and balancing repair cycle; Constraints are constraints that need to be met during the maintenance plan generation process, including maintenance time cycle, team scheduling, holiday time and balanced maintenance plan; Algorithms are algorithms, which are machine learning algorithm models used to generate maintenance plans. The algorithm model generates the optimal maintenance plan based on the input training data, features, and constraints.

5. The integrated vehicle preparation control method for EMU preparation according to claim 1 is characterized in that: The step S3 specifically comprises the following steps: S31, when the EMU train enters the depot, the AEI identification device automatically identifies the train arrival and the train model and number. The system collects the information of the AEI device, matches it with the train daily plan, and automatically generates the operation process; S32. Through the Internet of Things technology, the system controls the electric garage door and the driving signal light, determines whether the surrounding environment of the electronic garage door is safe, automatically opens the garage door, and turns the signal light green. After the train enters the garage and goes to the platform, it automatically determines and closes the garage door, and turns off the signal light at the same time; S33. The corresponding work team uses the in-warehouse workstation terminal for face recognition and applies for power-off operation. The approval of the power-off operation is automatically transferred to the on-site dispatch and integrated duty room for approval and authorization. After approval, the system jointly controls the in-warehouse voice dialogue system and the in-warehouse workstation terminal to remind the start of the operation. S34. The system automatically notifies relevant work teams to carry out joint work according to the preset work process. Each team member receives work tasks and specific guidance information through handheld terminals or workstation terminals. The system uses sensors, video surveillance, image algorithms and AI artificial intelligence to conduct all-round monitoring of the work area. S35. If an abnormal situation occurs, the system will immediately sound an alarm and notify relevant personnel to take emergency measures through the voice dialogue system. After the operation is completed, each team member completes the operation through face recognition on a handheld terminal or workstation terminal. The system automatically reviews the operation status, eliminates the operation authority after confirmation, and broadcasts the completion information of the operation; S36. If there is any equipment failure or abnormality during the operation, the operator can quickly report the fault information through the handheld terminal or the workstation terminal. After receiving the fault information, the system will automatically notify the relevant maintenance personnel to handle it and track the progress of the fault handling in real time; S37. After all operations are completed, the system automatically generates a train certificate and operation report based on the operation report and fault handling report submitted by each team. The certificate and operation report contain train information, operation content, operation time and operating personnel.

6. The integrated vehicle preparation control method for EMU preparation according to claim 1 is characterized in that: The intelligent fault warning and response in step S4 specifically includes the following steps: S411. Collect key train operation data in real time through sensors and on-board data, and clean the collected data to improve the accuracy and efficiency of subsequent data analysis; S412. Use big data analysis technology to conduct in-depth mining and analysis on the cleaned data, identify potential fault signs and abnormal patterns, and train machine learning models to make intelligent judgments and predictions on the operating status of equipment or systems. When the model identifies a fault or abnormality, it will output a corresponding warning signal. S413. Set preset warning thresholds and preset conditions. When the anomaly identified by the data analysis layer reaches the preset warning threshold, the warning mechanism is immediately triggered. The warning information is promptly conveyed to relevant personnel through voice broadcast and system reminders. The warning information includes the fault type, fault location and severity, so that relevant personnel can quickly take corresponding measures. S414. The system also includes an emergency handling guide library to provide maintenance personnel with references and suggestions for fault handling. For predictable faults or abnormal conditions, the system initiates an automated response process, including automatic shutdown or switching to backup equipment, to reduce the impact of the fault on the operation of the equipment or system; S415. After receiving the warning information, the maintenance personnel shall take response measures according to the emergency handling guide and the actual situation, including repairing the fault or eliminating the abnormality; S416. Based on historical fault data, establish a fault prediction model to predict the future operating status of the equipment or system. According to the output results of the fault prediction model, formulate a preventive maintenance plan in advance and conduct regular inspection and maintenance of key components; S417. The system continuously adapts to changes in the operating environment of equipment and systems through continuous learning algorithms, optimizes prediction models and algorithms, improves the accuracy and timeliness of early warnings, collects user feedback and system operation data, and continuously optimizes and improves the system to improve system stability and reliability.

7. The integrated vehicle preparation control method for EMU preparation according to claim 6 is characterized in that: In step S417, the system continuously optimizes the prediction model and algorithm through continuous learning, including the following specific steps: S4171. New data collection and annotation, including: As the vehicle operates and the environment changes, new data is automatically and continuously collected and annotated. The new data includes the health status information of the vehicle in different operating environments; S4172, Model retraining and updating, including: Integrate new data into the existing training set and retrain the model. During the retraining process, gradually introduce new data to avoid causing too much impact on the model. Based on the performance evaluation results of the retrained model, decide whether to update the model parameters and architecture. S4173, Online learning and adaptation, specifically including: Use online learning algorithms to enable the model to continuously learn and adapt to new data and environmental changes during actual operation; monitor the model's predictive performance and trigger the model update mechanism in a timely manner when performance degradation is found; S4174. User feedback data collection, including: Collecting user feedback on the prediction results through user feedback channels, including the website's built-in feedback channel, social media, and customer service hotline; the feedback includes the accuracy and timeliness of the prediction results and the user's experience; S4175, system optimization and iteration, including: Optimize and improve the system based on user feedback and data analysis results; continuously optimize prediction models and algorithms to improve the accuracy and timeliness of early warnings; iteratively update system functions and interface design to enhance user experience.

8. The integrated vehicle preparation control method for EMU preparation according to claim 1 is characterized in that: The use of deep learning technology to train the fault prediction model in step S4 specifically includes the following steps: S421. Data preparation, including: Data collection: Collect large amounts of mission-related raw data from databases, sensors, and networks; Data cleaning: Process outliers, missing values, and duplicate values ​​in the data to ensure data quality; Data annotation: providing labels or target values ​​for supervised learning tasks; Data segmentation: divide the dataset into training set, validation set and test set; Data preprocessing: including data standardization, normalization, feature extraction and feature selection, which is used to reduce the learning burden of the model and improve performance; S422. Establishing a data model, specifically including: Select model architecture: Select appropriate deep learning models based on the task type, which include image recognition, natural language processing, and time series prediction. The deep learning models include CNN, RNN, LSTM, and Transformer. Define the model: Use deep learning frameworks such as TensorFlow and PyTorch to define the structure of the model, including input layer, hidden layer, output layer, activation function, and loss function. S423, model training, specifically including: Data input: input the preprocessed training data into the model; Compile the model: Set the optimizer and learning rate in the deep learning framework and compile the model. The optimizer includes Adam and SGD. Iterative training: Calculate the prediction value and loss through forward propagation, and then update the weight and bias of the model through back propagation; this process will be repeated many times until the preset number of training rounds is reached or the loss converges; S424, Model verification and tuning, including: Performance evaluation: Evaluate the performance of the model on the validation set. The evaluation indicators include accuracy, precision, recall, F1 score and mean square error. Hyperparameter tuning: Adjust the model's hyperparameters based on the performance of the validation set to improve model performance. The hyperparameters include learning rate, batch size, number of layers, and number of nodes. Regularization and Optimization: Regularization techniques are used to prevent overfitting, and optimization algorithms are used to accelerate the training process. The regularization techniques include L1, L2 regularization, and Dropout; the optimization algorithms include momentum and RMSprop; S425, Model evaluation and optimization, including: Performance evaluation: Use evaluation metrics to evaluate the performance of the model on the test set; Model optimization: If the model performance does not meet the requirements, optimize it by adjusting the model structure, hyperparameters, increasing the amount of data, and improving data preprocessing; Model fusion: improve overall performance by combining the prediction results of multiple models through ensemble learning; ensemble learning includes bagging and boosting; S426, Model loading and application, including Model export: export the trained model to a deployable format; the deployable formats include TensorFlowSavedModel and PyTorch; Model deployment: deploy the model to the production environment; Performance monitoring: Monitor the performance of the model in real applications and update and optimize it as needed; Data collection and annotation: Continuously collect new data and annotate it to enrich the training set; Model iteration: Continuously iterate and optimize the model based on new data and feedback to improve its performance and accuracy.

Images