Fault early warning system for locomotive brake device

Abstract

The invention relates to the technical field of locomotive braking, and discloses a locomotive braking device fault early warning system which comprises a multi-source data fusion layer, a digital twinning driving fault analysis layer and an intelligent early warning output layer. The multi-source data fusion layer collects and integrates real-time sensing data, historical operation and maintenance data and digital twinning simulation data of a braking device; the fault analysis layer forms a data source of fault analysis, the fault analysis layer constructs a fault risk assessment model based on a feature extraction, threshold judgment and performance degradation prediction fusion technology, the model can identify a potential fault type and a risk level of the braking device, and the intelligent early warning output layer generates graded early warning information according to a fault assessment result. Real-time sensing data, historical operation and maintenance data and digital twinborn simulation data are integrated through a multi-source data fusion layer, and a multi-dimensional fault risk assessment mechanism is constructed in combination with feature extraction, dynamic threshold judgment and intelligent agent model prediction technologies.

Description

Technical Field

[0001] This invention relates to the field of locomotive braking technology, specifically to a locomotive braking device fault early warning system. Background Technology

[0002] Fault early warning of locomotive braking systems is a crucial link in ensuring railway transportation safety. Traditional fault early warning systems often rely on single sensor data with preset thresholds or the accumulated experience of maintenance personnel. This approach has significant limitations: single-threshold-based warnings are susceptible to fluctuations in operating conditions, resulting in a high false alarm rate; while fault judgment relying on human experience is not only inefficient but also fails to cover various potential fault modes under complex operating conditions. With the development of sensing and data analysis technologies, some early warning systems have begun to adopt multi-sensor data fusion methods. However, existing solutions often focus on simple statistical analysis of the data, lacking deep integration of the physical characteristics of the braking system (such as the mechanical behavior of brake calipers and the thermal conduction laws of brake discs). This makes it difficult for early warning models to accurately identify hidden faults such as brake pad wear, stress concentration, and abnormal temperature. Furthermore, existing early warning systems often lack integration with digital twin models of the equipment, making it impossible to achieve fault tracing and trend prediction based on the comparison of virtual simulation data and real-time operating data. The presentation of early warning information is also relatively simple, making it difficult to intuitively show maintenance personnel the fault location, risk level, and evolution trend, thus failing to meet the requirements of intelligent railways for accurate, timely, and interpretable fault early warnings for braking systems.

[0003] To address this issue, those skilled in the art have proposed a locomotive braking device fault early warning system. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a locomotive braking device fault early warning system, which solves the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a locomotive braking device fault early warning system, comprising a multi-source data fusion layer, a digital twin-driven fault analysis layer, and an intelligent early warning output layer. The multi-source data fusion layer collects and integrates real-time sensor data, historical operation and maintenance data, and digital twin simulation data of the braking device to form a data source for fault analysis. The fault analysis layer constructs a fault risk assessment model based on a fusion technology of feature extraction, threshold judgment, and performance degradation prediction. The model can identify potential fault types and risk levels of the braking device. The intelligent early warning output layer generates graded early warning information based on the fault assessment results and accurately displays the fault location, risk level, and maintenance suggestions through a visual interface.

[0006] Through the above technical solution, a multi-source data fusion layer integrates real-time sensor data, historical operation and maintenance data, and digital twin simulation data to provide a comprehensive data source for fault analysis. The fault analysis layer uses feature extraction, threshold judgment, and performance degradation prediction fusion technology to build a risk assessment model, accurately identifying potential fault types and risk levels. The intelligent early warning output layer generates graded early warning information based on the assessment results and accurately displays the fault location, risk level, and maintenance suggestions through a visual interface. Overall, it realizes accurate identification, risk classification, and intuitive presentation of braking device faults, effectively improving the accuracy of fault early warning and the efficiency of operation and maintenance decision-making.

[0007] Preferably, the multi-source data fusion layer includes a sensing unit, a data preprocessing submodule, and a data integration submodule. The sensing unit is deployed at the brake caliper, brake cylinder, and wheel to collect data on brake lever deflection angle, brake cylinder pressure, wheel speed, and temperature. The data preprocessing submodule filters, reduces noise, and standardizes the format of the raw sensing data. The data integration submodule associates and stores the preprocessed data with historical fault data and digital twin simulation data.

[0008] The above technical solution achieves comprehensive monitoring and data integration of the locomotive braking system's operating status through a multi-source data fusion layer. Sensing units are deployed at key locations to collect real-time data. The data preprocessing submodule filters, reduces noise, and standardizes the format of the collected raw data to improve data quality. The data integration submodule then associates and stores the processed data with historical fault data and digital twin simulation data, forming a comprehensive and high-quality data source. This provides reliable data support for subsequent fault analysis and early warning, thereby improving the accuracy and timeliness of fault warnings.

[0009] Preferably, the fault analysis layer includes a feature extraction submodule, a digital twin comparison submodule, and a performance prediction submodule. The feature extraction submodule extracts fault-sensitive features from the fused data, including brake pad wear, abnormal brake caliper stress, brake disc temperature gradient, and brake response delay. The digital twin comparison submodule identifies abnormal operating conditions by comparing the deviation between real-time data and the simulation data of the digital twin. The performance prediction submodule predicts the performance degradation trend of the braking device based on a trained intelligent agent model.

[0010] The above technical solution enables accurate identification and risk assessment of potential faults in locomotive braking systems through a fault analysis layer. The feature extraction submodule extracts key fault-sensitive features from multi-source fusion data, providing a clear basis for fault diagnosis. The digital twin comparison submodule quickly identifies abnormal operating conditions by comparing the deviation between real-time data and the simulation data of the digital twin. The performance prediction submodule uses an intelligent surrogate model to predict the performance degradation trend of the braking system, providing early warnings of potential faults. This significantly improves the accuracy and foresight of fault warnings, providing strong support for predictive maintenance of the braking system.

[0011] Preferably, the wear amount of the brake pads is calculated by analyzing the change in the deflection angle of the brake caliper. Specifically, based on the reference deflection angle of the brake caliper when there is no brake pad wear, the difference in the deflection angle of the brake levers on both sides during actual operation is compared, and the wear amount of the brake pads on both sides is derived by combining the structural parameters of the braking device.

[0012] The above technical solution accurately calculates brake pad wear by analyzing changes in the brake caliper deflection angle, thus enabling real-time monitoring of brake pad wear. This method utilizes the difference between the reference deflection angle under no-wear conditions and the actual deflection angle during operation, combined with the structural parameters of the braking system, to accurately deduce the wear amount of the brake pads on both sides. This provides a crucial fault-sensitive feature for the fault warning system, enabling timely detection of excessive brake pad wear, effectively preventing braking failures caused by excessive brake pad wear, and improving the safety and reliability of locomotive operation.

[0013] Preferably, the intelligent agent model is a prediction model based on radial basis function neural network. The model takes brake cylinder pressure, locomotive speed and braking duration as inputs and outputs predicted values ​​of stress, temperature and wear within a preset time. The model is trained and optimized using multi-condition finite element simulation data.

[0014] The above technical solution utilizes an intelligent surrogate model based on radial basis function neural networks to predict future performance parameters of locomotive braking systems. This model takes brake cylinder pressure, locomotive speed, and braking duration as inputs, and is trained and optimized using multi-condition finite element simulation data to output predicted values ​​within a preset timeframe. This predictive capability can identify potential fault risks in advance, providing a forward-looking basis for maintenance decisions, thereby improving the reliability and safety of the braking system and reducing downtime and maintenance costs.

[0015] Preferably, the intelligent early warning output layer includes a risk level classification submodule, a fault location submodule, and an information display submodule. The risk level classification submodule classifies fault risks into prompt level, warning level, and emergency level, each corresponding to different processing priorities. The fault location submodule, in conjunction with the braking device structural topology and sensor distribution, locates the specific component where the fault occurs. The information display submodule displays early warning information through text, color coding, 3D model highlighting, and dynamic charts.

[0016] The above technical solution enables the classification, precise location, and intuitive display of locomotive braking device fault risks through an intelligent early warning output layer. The risk level classification submodule categorizes fault risks into alert, warning, and emergency levels, clearly defining the handling priorities for different risks. The fault location submodule, combining the braking device's structural topology and sensor distribution, accurately locates the faulty component. The information display submodule intuitively presents early warning information through text, color coding, 3D model highlighting, and dynamic charts.

[0017] Preferably, the system further includes a self-learning optimization module, which continuously iteratively optimizes the weights of fault-sensitive features, prediction model parameters, and dynamic thresholds based on historical fault handling data, thereby improving the accuracy and generalization ability of fault early warning.

[0018] Through the above technical solution, by continuously learning from historical fault handling data, the system dynamically adjusts the weights of fault-sensitive features, optimizes prediction model parameters, and updates dynamic thresholds. This process continuously improves the system's accuracy in identifying different fault types and enhances its generalization ability under complex operating conditions, thereby ensuring that the fault early warning system maintains a high-efficiency and accurate operating state in the long term, and better adapts to the actual operating needs of locomotive braking devices.

[0019] Preferably, the fault analysis layer driven by the digital twin can call the three-dimensional model of the digital twin, highlight the fault location through the model when an early warning is issued, and overlay and display the real-time performance parameters of the location and the historical data comparison curve.

[0020] The above technical solution uses a digital twin's 3D model to visually display fault information. When an alert is issued, the system highlights the faulty area and overlays a curve comparing the real-time performance parameters of that area with historical data. This visualization method allows maintenance personnel to quickly and accurately locate the fault and intuitively understand the performance change trend of the faulty area, thereby improving fault diagnosis efficiency.

[0021] Preferably, the intelligent early warning output layer further includes a maintenance suggestion generation submodule. The submodule automatically generates targeted maintenance strategy suggestions based on fault type, risk level and historical maintenance plans, including maintenance timing, inspection parts and operation guidelines.

[0022] The maintenance suggestion generation submodule, based on the aforementioned technical solution, automatically generates targeted maintenance strategy recommendations according to the fault type, risk level, and historical maintenance plans. These recommendations cover maintenance timing, inspection locations, and operational guidelines. This provides maintenance personnel with clear and specific maintenance references, facilitating quick and accurate fault handling, improving maintenance efficiency, and reducing downtime.

[0023] This invention provides a locomotive braking device fault early warning system. It has the following beneficial effects: 1. This invention integrates real-time sensor data, historical operation and maintenance data, and digital twin simulation data through a multi-source data fusion layer. It combines feature extraction, dynamic threshold judgment, and intelligent agent model prediction technologies to construct a multi-dimensional fault risk assessment mechanism. By comparing the deviations between real-time and simulation data using digital twins, it accurately identifies potential faults such as brake pad wear and abnormal stress. Furthermore, it dynamically adjusts the judgment threshold based on braking conditions, significantly improving the accuracy and relevance of fault warnings.

[0024] 2. This invention achieves graded display, precise location, and maintenance suggestion generation of fault risks through an intelligent early warning output layer. Combined with 3D model highlighting and dynamic chart visualization technology, it intuitively presents the fault location, risk level, and performance degradation trend. Simultaneously, through a self-learning optimization module, it iteratively optimizes model parameters and judgment rules based on historical data, continuously improving the early warning generalization capability. This provides a scientific basis for predictive maintenance of braking devices, helping to reduce downtime and safety hazards. Attached Figure Description

[0025] Figure 1 This is a flowchart of the present invention; Figure 2 This is a schematic diagram of the pose changes of the brake caliper in multiple states according to the present invention. Detailed Implementation

[0026] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Please see the appendix Figure 1 - Appendix Figure 2This invention provides a locomotive braking device fault early warning system, including a multi-source data fusion layer, a digital twin-driven fault analysis layer, and an intelligent early warning output layer. The multi-source data fusion layer collects and integrates real-time sensor data, historical operation and maintenance data, and digital twin simulation data of the braking device to form a data source for fault analysis. The fault analysis layer constructs a fault risk assessment model based on feature extraction, threshold judgment, and performance degradation prediction fusion technology. The model can identify the potential fault types and risk levels of the braking device. The intelligent early warning output layer generates graded early warning information based on the fault assessment results and accurately displays the fault location, risk level, and maintenance suggestions through a visual interface.

[0028] The multi-source data fusion layer includes a sensing unit, a data preprocessing submodule, and a data integration submodule. The sensing unit is deployed in the brake caliper, brake cylinder, and wheel to collect data on brake lever deflection angle, brake cylinder pressure, wheel speed, and temperature. The data preprocessing submodule filters, reduces noise, and standardizes the format of the raw sensing data. The data integration submodule associates and stores the preprocessed data with historical fault data and digital twin simulation data.

[0029] Specifically, the efficient integration of multi-dimensional data is achieved through the collaborative work of the sensing unit, the data preprocessing submodule, and the data integration submodule. The sensing unit is specifically deployed in key parts such as brake calipers, brake cylinders, and wheels to accurately collect real-time data reflecting the operating status of the braking device, such as brake lever deflection angle, brake cylinder pressure, wheel speed, and temperature. The data preprocessing submodule filters the collected raw sensor data to remove interference signals, reduces noise to lower data noise, and standardizes the format to unify the data format, ensuring that the data quality meets the needs of subsequent analysis. The data integration submodule associates and stores the preprocessed real-time data with historical fault data (including past fault cases and corresponding characteristics) and digital twin simulation data (various working condition performance data simulated by virtual models), constructing a comprehensive fault analysis dataset to provide multi-source, high-quality data support for subsequent fault feature extraction, risk assessment, and early warning decision-making.

[0030] The fault analysis layer comprises a feature extraction submodule, a digital twin comparison submodule, and a performance prediction submodule. The feature extraction submodule extracts fault-sensitive features from the fused data, including brake pad wear, abnormal brake caliper stress values, brake disc temperature gradient, and brake response delay. The digital twin comparison submodule identifies abnormal operating conditions by comparing the deviation between real-time data and the simulation data of the digital twin. The performance prediction submodule predicts the performance degradation trend of the braking system based on a trained intelligent agent model. Brake pad wear is calculated by analyzing the change in brake caliper deflection angle. Specifically, based on the reference deflection angle of the brake caliper when there is no brake pad wear, the difference in deflection angles of the brake levers on both sides during actual operation is compared, and the wear amount of the brake pads on both sides is derived by combining the structural parameters of the braking system.

[0031] Specifically, the fault analysis layer, as the core analysis component, achieves accurate identification of fault risks through the collaborative operation of the feature extraction submodule, the digital twin comparison submodule, and the performance prediction submodule. The feature extraction submodule extracts fault-sensitive features such as brake pad wear, abnormal values ​​of brake caliper stress, brake disc temperature gradient, and brake response delay from multi-source fusion data. The calculation of brake pad wear is based on the reference deflection angle when there is no brake pad wear, which is derived by comparing the difference in deflection angles of the brake levers on both sides during actual operation and combining the structural parameters of the braking device. The digital twin comparison submodule accurately identifies abnormal states that deviate from normal operating conditions by comparing the deviation between the real-time collected operating data and the simulation data of the digital twin. The performance prediction submodule relies on a trained intelligent agent model to predict the performance degradation trend of the braking device. Together, these three modules provide comprehensive feature basis, anomaly judgment, and trend prediction for fault risk assessment, supporting the system to achieve accurate fault early warning.

[0032] The intelligent agent model is a prediction model based on radial basis function neural networks. The model takes brake cylinder pressure, locomotive speed, and braking duration as inputs and outputs predicted values ​​of stress, temperature, and wear within a preset time. The model is trained and optimized using multi-condition finite element simulation data. The intelligent early warning output layer includes a risk level classification submodule, a fault location submodule, and an information display submodule. The risk level classification submodule divides fault risks into alert, warning, and emergency levels, each corresponding to different processing priorities. The fault location submodule, combining the braking device structural topology and sensor distribution, locates the specific component where the fault occurred. The information display submodule displays early warning information through text, color coding, 3D model highlighting, and dynamic charts.

[0033] Specifically, the intelligent agent model adopts a prediction model based on radial basis function neural networks, using brake cylinder pressure, locomotive speed, and braking duration as input parameters. After training and optimization using multi-condition finite element simulation data, it can output predicted values ​​of stress, temperature, and wear within a preset time, providing forward-looking data support for fault risk assessment. The intelligent early warning output layer achieves accurate output and presentation of early warning information through risk level classification submodule, fault location submodule, and information display submodule: the risk level classification submodule divides fault risks into alert, warning, and emergency levels, corresponding to different processing priorities to clarify response strategies; the fault location submodule combines the structural topology of the braking device and the distribution of sensors to accurately locate the specific component where the fault occurred; the information display submodule intuitively displays early warning information through diverse methods such as text descriptions, color coding, 3D model highlighting, and dynamic charts, ensuring that maintenance personnel can quickly understand the fault status and take targeted measures.

[0034] The system also includes a self-learning optimization module, which continuously iteratively optimizes the weights of fault-sensitive features, prediction model parameters, and dynamic thresholds based on historical fault handling data, thereby improving the accuracy and generalization ability of fault warnings. The digital twin-driven fault analysis layer can call the 3D model of the digital twin, highlighting the fault location on the model during warnings and overlaying a comparison curve of the real-time performance parameters of that location with historical data.

[0035] Specifically, the self-learning optimization module continuously learns from historical fault handling data, iteratively adjusting the weight allocation of fault-sensitive features, optimizing the parameter configuration of the prediction model, and setting dynamic thresholds. This enhances the system's accuracy in identifying different fault types and its generalization ability under complex operating conditions. Meanwhile, the digital twin-driven fault analysis layer can call the 3D model of the digital twin. When an alert is triggered, the fault location is accurately marked by highlighting the model, and a comparison curve of real-time performance parameters and historical data is overlaid on that location. This allows maintenance personnel to intuitively grasp the performance change trend of the fault location, providing a more concrete reference for fault diagnosis and handling.

[0036] The intelligent early warning output layer also includes a maintenance suggestion generation submodule. Based on the fault type, risk level, and historical maintenance plans, the submodule automatically generates targeted maintenance strategy suggestions, including maintenance timing, inspection points, and operation guidelines.

[0037] Specifically, the maintenance suggestion generation submodule serves as a key component in decision support. Based on the fault types identified by the system, the assessed risk levels, and the accumulated historical maintenance plans, it automatically generates targeted maintenance strategy suggestions. These suggestions cover appropriate maintenance timings to suit the urgency of the fault, key inspection areas, and standardized operating guidelines. This provides maintenance personnel with clear and feasible maintenance references, improving the efficiency and accuracy of fault handling.

[0038] This system analyzes and calculates the degree of brake pad wear based on the Euler angle data output from sensors in three states: brake pad wear-free, brake pad wear-free, and brake pad wear-affected. It then uses preset safety thresholds from the maintenance manual to determine if any abnormalities exist. Furthermore, relevant safety thresholds are set for other monitoring data during braking. When the system detects brake pad wear exceeding safe limits or detects an abnormal braking state, it displays real-time warning information on the UI panel. These warnings are categorized into "abnormal brake pad condition" and "abnormal braking state," aiming to provide a basis for equipment maintenance and fault diagnosis.

[0039] Changes in the position of the brake caliper directly reflect the contact condition of the brake pads. For example... Figure 2 As shown, when the brake caliper has no brake pad wear and is in the released state, the brake pad clearance on both sides should be S, and the rotation angle should be α. When there is no brake pad wear and the brake is in the braking state, the clearance should be 0, and the rotation angle should be... .

[0040] When the brake pads begin to wear, it can be known that: Left brake pad wear and wear of the right-side brake pads It can be represented as: In Unity, multiple empty Panels are added under the Canvas to construct dedicated information panels and alert panels for the warning function. The information panel displays a normal list of monitoring data; its size and position are adjusted, placed appropriately on the interface, and kept enabled. The alert panel displays alert information or highlights when an alert occurs; prominent colors and icons are added to the alert panel to indicate the alert status. By default, the alert panel is disabled and only displays when an alert is triggered. Multiple TMP_Text text components are added within the information panel to display the monitoring data. These TMP_Text objects can be dragged and dropped into an array or list reference in the script, allowing their text content to be dynamically updated at runtime.

[0041] The fault warning script, serving as the primary control script for the monitoring system's warning function, is responsible for updating data and responding to interactions. When writing the script, FixedUpdate is used to update the monitoring data at fixed time intervals every frame and dynamically instantiate information items. A Prefab for each information item is pre-created, and new information items are instantiated as needed within FixedUpdate, becoming sub-objects of the infoPanel. When a data point reaches the warning condition, the JingBao method is called to hide and show the current infoPanel and the jingbaoPanel, switching the display between the information panel and the alarm panel. Simultaneously, HighlightingSystem can be used to highlight the corresponding object. The target object triggering the alarm is retrieved, and its Highlighter component's ConstantOn(Color.red) method is called to make the object flash red to alert the user.

[0042] The final warning information UI is divided into a title bar, a status display area, and a warning information area. The title bar displays the title "Fault Warning," the status display area shows real-time key data using dynamic charts or numerical labels, and the warning information area is located in the lower right corner of the interface. The judgment formula is written into a script for system judgment. The judgment result calls a UI update function. If the calculation result is not within the safe range, the system determines that the brake pad is abnormal and triggers a "Brake Pad Status Abnormal" warning. Simultaneously, other monitored parameters X are compared. If the difference from the normal value is large, it is determined that "X Status Abnormal." If normal, the text is white; if abnormal, the text is highlighted in red and a warning is displayed. To ensure dynamic interface responsiveness, this paper uses the DOTween script to implement a fade-in or flashing effect for the warning text, making the abnormal status more conspicuous. At the same time, other data display controls are refreshed synchronously, forming a closed-loop feedback between data and warning information.

[0043] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A locomotive braking device fault early warning system, characterized in that, The system comprises a multi-source data fusion layer, a digital twin-driven fault analysis layer, and an intelligent early warning output layer. The multi-source data fusion layer collects and integrates real-time sensor data, historical operation and maintenance data, and digital twin simulation data of the braking device to form a data source for fault analysis. The fault analysis layer constructs a fault risk assessment model based on a fusion technology of feature extraction, threshold judgment, and performance degradation prediction. The model can identify the potential fault types and risk levels of the braking device. The intelligent early warning output layer generates graded early warning information based on the fault assessment results and accurately displays the fault location, risk level, and maintenance recommendations through a visual interface.

2. The locomotive braking device fault early warning system according to claim 1, characterized in that, The multi-source data fusion layer includes a sensing unit, a data preprocessing submodule, and a data integration submodule. The sensing unit is deployed at the brake caliper, brake cylinder, and wheel to collect data on brake lever deflection angle, brake cylinder pressure, wheel speed, and temperature. The data preprocessing submodule filters, reduces noise, and standardizes the format of the raw sensing data. The data integration submodule associates and stores the preprocessed data with historical fault data and digital twin simulation data.

3. The locomotive braking device fault early warning system according to claim 1, characterized in that, The fault analysis layer includes a feature extraction submodule, a digital twin comparison submodule, and a performance prediction submodule. The feature extraction submodule extracts fault-sensitive features from the fused data, including brake pad wear, abnormal brake caliper stress, brake disc temperature gradient, and brake response delay. The digital twin comparison submodule identifies abnormal operating conditions by comparing the deviation between real-time data and the simulation data of the digital twin. The performance prediction submodule predicts the performance degradation trend of the braking device based on a trained intelligent agent model.

4. A locomotive braking device fault early warning system according to claim 3, characterized in that, The wear amount of the brake pads is calculated by analyzing the change in the deflection angle of the brake caliper. Specifically, based on the reference deflection angle of the brake caliper when there is no brake pad wear, the difference in the deflection angle of the brake levers on both sides during actual operation is compared, and the wear amount of the brake pads on both sides is derived by combining the structural parameters of the braking device.

5. A locomotive braking device fault early warning system according to claim 3, characterized in that, The intelligent agent model is a prediction model based on radial basis function neural network. The model takes brake cylinder pressure, locomotive speed and braking duration as inputs and outputs predicted values ​​of stress, temperature and wear within a preset time. The model is trained and optimized using multi-condition finite element simulation data.

6. A locomotive braking device fault early warning system according to claim 1, characterized in that, The intelligent early warning output layer includes a risk level classification submodule, a fault location submodule, and an information display submodule. The risk level classification submodule classifies fault risks into alert level, warning level, and emergency level, each corresponding to different processing priorities. The fault location submodule combines the structural topology of the braking device and the distribution of sensors to locate the specific component where the fault occurred. The information display submodule displays early warning information through text, color coding, 3D model highlighting, and dynamic charts.

7. A locomotive braking device fault early warning system according to claim 1, characterized in that, The system also includes a self-learning optimization module, which continuously iteratively optimizes the weights of fault-sensitive features, prediction model parameters, and dynamic thresholds based on historical fault handling data, thereby improving the accuracy and generalization ability of fault warnings.

8. A locomotive braking device fault early warning system according to claim 1, characterized in that, The fault analysis layer driven by the digital twin can call the three-dimensional model of the digital twin. When an early warning is issued, the fault location is highlighted in the model, and the real-time performance parameters of that location are overlaid and compared with historical data curves.

9. A locomotive braking device fault early warning system according to claim 1, characterized in that, The intelligent early warning output layer also includes a maintenance suggestion generation submodule. Based on the fault type, risk level, and historical maintenance plans, the submodule automatically generates targeted maintenance strategy suggestions, including maintenance timing, inspection locations, and operation guidelines.

Images