Intelligent operation and maintenance system for platform door
By utilizing the data processing and distributed diagnostic mechanism of the platform screen door intelligent operation and maintenance system, and employing support vector machine algorithms and a distributed intelligent diagnostic system, the system can quickly and accurately locate the fault location and cause, generate handling suggestions, and solve the problems of inaccurate fault prediction and superficial diagnosis in existing systems, thereby improving operation and maintenance efficiency and system stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG HAINING RAIL TRANSIT OPERATION MANAGEMENT CO LTD
- Filing Date
- 2025-05-30
- Publication Date
- 2026-08-04
AI Technical Summary
Existing intelligent operation and maintenance systems for platform screen doors are insufficient in terms of the accuracy of fault prediction and the depth of fault diagnosis. They cannot make full use of the complex features in the equipment operation data, resulting in inaccurate fault prediction and inaccurate diagnosis, which affects the normal operation of the equipment.
It employs a data acquisition and processing module, a fault prediction model construction module, a fault early warning module, an intelligent diagnosis module, and a fault decision-making module. Combined with the support vector machine algorithm and a distributed intelligent diagnosis system, it can quickly and accurately locate the fault location and cause through real-time data processing and comprehensive analysis, generate handling suggestions, and form a closed-loop optimization model.
It improves the accuracy of fault prediction and the precision of diagnosis, reduces false alarm rate, lowers operation and maintenance costs and workload, and ensures the operational safety and efficiency of urban rail transit.
Smart Images
Figure CN120598536B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent operation and maintenance technology for rail transit, and in particular to an intelligent operation and maintenance system for platform screen doors. Background Technology
[0002] With the rapid development of urban rail transit, platform screen door systems, as key equipment for ensuring passenger safety and train operation efficiency, are becoming increasingly important. In recent years, with the continuous development of IoT, big data, and artificial intelligence technologies, platform screen door operation and maintenance has gradually moved towards intelligentization. Some intelligent operation and maintenance systems collect equipment operation data by installing sensors, realizing real-time monitoring and fault alarms for platform screen door systems, aiming to reduce maintenance costs, improve system reliability, and extend equipment lifespan.
[0003] Existing intelligent operation and maintenance systems for platform screen doors have significant shortcomings in the accuracy of fault prediction and the depth of fault diagnosis. First, the fault prediction models in existing systems are mostly based on simple statistical analysis, failing to fully utilize the complex features in equipment operating data, resulting in low accuracy in fault prediction. Second, existing systems lack a distributed intelligent diagnostic mechanism during fault diagnosis, hindering rapid fault location and precise analysis, leading to prolonged fault handling time and impacting normal equipment operation. Summary of the Invention
[0004] In view of the aforementioned existing problems, the present invention is proposed.
[0005] Therefore, the present invention provides an intelligent operation and maintenance system for platform screen doors to solve the problems of inaccurate fault prediction and superficial fault diagnosis.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:
[0007] In a first aspect, the present invention provides an intelligent operation and maintenance system for platform screen doors, comprising: a data acquisition and processing module: acquiring real-time operating data and preprocessing it, extracting operating data features and storing them; a fault prediction model construction module: constructing a fault prediction model, training the fault prediction model using operating data features, and generating a trained fault prediction model; a fault early warning module: inputting real-time operating data into the trained fault prediction model, generating early warning information, and sending the early warning information to a distributed intelligent diagnostic system; an intelligent diagnostic module: the distributed intelligent diagnostic system identifies IoT nodes that need to be diagnosed based on the early warning information, and the IoT nodes perform preliminary diagnosis on the collected data according to preset rules, generating preliminary diagnostic results; a fault decision module: comprehensively analyzing the preliminary diagnostic results, determining the location and cause of the fault based on the comprehensive analysis results, and generating fault handling suggestions; and an optimization closed-loop module: the fault prediction model dynamically adjusts its parameters according to the fault handling suggestions, retrains the fault prediction model using the updated parameters, and continues to use it for fault prediction, forming a complete closed loop.
[0008] As a preferred embodiment of the intelligent operation and maintenance system for platform doors described in this invention, the real-time operation data includes electrical operation data, environmental operation data, mechanical operation data, and control operation data.
[0009] The preprocessing includes smoothing and noise reduction.
[0010] The operational data characteristics include statistical characteristics, time-series characteristics, and frequency domain characteristics;
[0011] Store runtime data characteristics in a relational database.
[0012] As a preferred embodiment of the intelligent operation and maintenance system for platform screen doors described in this invention, the specific steps for constructing a fault prediction model and training the fault prediction model using operational data features to generate a trained fault prediction model are as follows.
[0013] The support vector machine algorithm is selected to construct the fault prediction model, and the fault prediction model parameters are initialized.
[0014] The operational data features are divided into training and testing sets;
[0015] The fault prediction model adjusts its internal parameters based on the training set data;
[0016] The trained fault prediction model is validated using a test set, and validation results are generated.
[0017] Based on the validation results, the parameters of the fault prediction model were optimized using the grid search method to generate a trained fault prediction model.
[0018] As a preferred solution of the platform door intelligent operation and maintenance system described in the present invention, wherein: inputting the real-time operation data into the trained fault prediction model to generate early warning information and sending the early warning information to the distributed intelligent diagnosis system, the specific steps are as follows,
[0019] Load the trained fault prediction model, input the real-time operation data features into the fault prediction model, and calculate the probability P of fault occurrence;
[0020] Define the fault probability threshold Y. When P>Y, trigger an early warning and generate early warning information;
[0021] When P<Y, it is considered that the operation state of the current device is stable;
[0022] Encapsulate the early warning information into JSON format and send the early warning information to the distributed diagnosis system through the MQTT protocol.
[0023] As a preferred solution of the platform door intelligent operation and maintenance system described in the present invention, wherein: the distributed intelligent diagnosis system identifies the Internet of Things nodes that need to be diagnosed according to the early warning information, the specific steps are as follows,
[0024] The early warning information includes the fault type, the estimated occurrence time, the affected components and the fault severity;
[0025] The distributed intelligent diagnosis system receives the early warning information through the MQTT protocol and analyzes the early warning information;
[0026] Based on the analyzed early warning information and combined with the pre-defined mapping relationship between components and nodes, the intelligent diagnosis system identifies the Internet of Things nodes related to the early warning information.
[0027] As a preferred solution of the platform door intelligent operation and maintenance system described in the present invention, wherein: the Internet of Things node conducts a preliminary diagnosis on the collected data according to the preset rules to generate a preliminary diagnosis result, the specific steps are as follows,
[0028] The preset rules include threshold rules, combination rules, pattern matching rules, time series rules and anomaly detection rules;
[0029] The Internet of Things node collects the operation data of the platform door and key components in real time and extracts the operation data features;
[0030] Conduct a preliminary diagnosis on the operation data features according to the preset rules to generate a preliminary diagnosis result, and send the preliminary diagnosis result to the central diagnosis system.
[0031] As a preferred solution of the platform door intelligent operation and maintenance system described in the present invention, wherein: comprehensively analyzing the preliminary diagnosis result, the specific steps are as follows,<00,00065>
[0032] The central diagnostic system collects preliminary diagnostic results from each IoT node and analyzes the correlation between these preliminary diagnostic results.
[0033] Identify the fault mode based on the fault manifestations in the preliminary diagnostic results;
[0034] By comparing real-time operating data with historical normal operating data, data points that significantly deviate from the normal range can be identified.
[0035] Through correlation analysis, we can explore whether current anomalies in different types of data are accompanied by temperature increases, and whether changes in vibration frequency are related to specific operating cycles.
[0036] As a preferred embodiment of the intelligent operation and maintenance system for platform screen doors described in this invention, the specific steps for determining the location and cause of the fault based on comprehensive analysis results and generating fault handling suggestions are as follows:
[0037] Based on the comprehensive analysis results, the source of the fault is narrowed down, and the location and cause of the problem are identified.
[0038] Based on the equipment failure mode and preliminary diagnostic results, infer the cause of the failure;
[0039] Based on the severity of the fault, determine its urgency, identify the cause of the fault, develop a repair plan, and generate fault handling suggestions.
[0040] As a preferred embodiment of the intelligent operation and maintenance system for platform screen doors described in this invention, the fault prediction model dynamically adjusts its parameters based on fault handling suggestions. The specific steps are as follows:
[0041] Collect fault handling suggestions and actual maintenance results from the operation and maintenance management system, and organize the collected data into a structured form;
[0042] Analyze the deviation between the prediction results of the fault prediction model and the actual faults, and identify the fault prediction model parameters that need to be adjusted based on the results of the deviation analysis.
[0043] As a preferred embodiment of the intelligent operation and maintenance system for platform screen doors described in this invention, the steps of retraining the fault prediction model using updated parameters and continuing to use it for fault prediction to form a complete closed loop are as follows.
[0044] The compiled fault handling suggestions and maintenance results data are merged with historical data to form a new training dataset. The fault prediction model is then retrained using the new training dataset and updated parameters.
[0045] The retrained fault prediction model is updated into the fault prediction system, replacing the old fault prediction model, and new fault handling suggestions and maintenance results are collected regularly.
[0046] The beneficial effects of this invention are as follows: Through a distributed intelligent diagnosis and comprehensive analysis mechanism, this invention can quickly and accurately locate the fault location, analyze the cause of the fault, and generate targeted handling suggestions, thereby improving operation and maintenance efficiency. By dynamically adjusting model parameters based on actual fault handling results, the system performance is continuously optimized, enhancing the system's adaptability and stability, accelerating fault response speed, improving the accuracy of fault prediction, reducing false alarm rate, reducing reliance on manual inspection, reducing operation and maintenance costs and workload, and ensuring the operational safety and efficiency of urban rail transit. Attached Figure Description
[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0048] Figure 1 This is a schematic diagram of an intelligent operation and maintenance system for platform screen doors in Example 1.
[0049] Figure 2 This is a flowchart of data acquisition and fault prediction in Example 1.
[0050] Figure 3 This is a flowchart of the distributed intelligent diagnosis process in Example 1.
[0051] Figure 4 This is the optimized closed-loop flowchart in Example 1. Detailed Implementation
[0052] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0053] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0054] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.
[0055] Example 1, referring to Figures 1-4 This is the first embodiment of the present invention, which provides an intelligent operation and maintenance system for platform screen doors, including the following steps:
[0056] Data acquisition and processing module: Acquires real-time running data and preprocesses it, extracts running data features and stores them.
[0057] Real-time operational data includes electrical operational data, environmental operational data, mechanical operational data, and control operational data;
[0058] It should be noted that electrical operating data refers to the current value, voltage value, and power consumption of the platform screen door motor; environmental operating data refers to the temperature of the platform screen door and its key components, the humidity of the platform screen door's operating environment, and the air pressure of the platform screen door's operating environment; mechanical operating data refers to the vibration frequency and amplitude of the platform screen door, the displacement data of the platform screen door, and the opening and closing speed of the platform screen door; control operating data refers to the input and output signals of the platform screen door control unit, the fault codes generated by the platform screen door control unit, and the opening and closing operation records of the platform screen door.
[0059] Preprocessing includes smoothing and noise reduction;
[0060] The characteristics of operational data include statistical characteristics, time-series characteristics, and frequency domain characteristics;
[0061] It should be noted that statistical characteristics refer to mean, variance, standard deviation, maximum value, and minimum value; time series characteristics refer to autocorrelation coefficient, partial autocorrelation coefficient, moving average, and exponentially weighted moving average; frequency domain characteristics refer to Fourier transform and power spectral density.
[0062] Store runtime data characteristics in a relational database;
[0063] It should be noted that data in relational databases is stored in tabular form, each table has a fixed column structure, and tables can be linked together.
[0064] Fault prediction model building module: Builds a fault prediction model, trains the fault prediction model using runtime data features, and generates a trained fault prediction model.
[0065] The support vector machine algorithm is selected to construct the fault prediction model, and the fault prediction model parameters are initialized.
[0066] It should be noted that the Support Vector Machine (SVM) algorithm is a supervised learning algorithm widely used in classification and regression tasks. In fault prediction, it is particularly suitable for handling high-dimensional data and can effectively handle nonlinear problems. The fault prediction model parameters refer to the kernel function type, regularization parameter C, and kernel function parameter γ.
[0067] The operational data features are divided into training and testing sets;
[0068] It should be noted that the operational data features are divided into training and test sets in an 8:2 ratio;
[0069] The fault prediction model adjusts its internal parameters based on the training set data;
[0070] It should be noted that optimization algorithms (such as gradient descent and Sequence Minimum Optimization (SMO)) adjust the model's internal parameters based on the training set data to find the optimal parameter combination, and the optimization algorithm updates the model's parameters in each iteration.
[0071] The trained fault prediction model is validated using a test set, and validation results are generated.
[0072] It should be noted that the test set is used to evaluate the performance of the fault prediction model on unseen data, thereby verifying the model's generalization ability. The fault prediction model will output the fault probability or fault category based on the feature data in the test set. The validation results typically include the model's accuracy, recall, and F1 score.
[0073] Based on the verification results, the parameters of the fault prediction model were optimized using the grid search method to generate a trained fault prediction model.
[0074] It should be noted that grid search is a commonly used parameter optimization method. By traversing a predefined parameter grid, the optimal combination of parameters is found to optimize the model performance. The fault prediction model is then retrained using the optimal parameter combination. This training uses the entire training set to ensure that the fault prediction model makes full use of all available data, thus generating the final fault prediction model.
[0075] Fault early warning module: Inputs real-time operating data into the trained fault prediction model, generates early warning information, and sends the early warning information to the distributed intelligent diagnostic system.
[0076] Load the trained fault prediction model, input the real-time running data features into the fault prediction model, and the fault prediction model calculates the probability of fault occurrence P, expressed as:
[0077]
[0078] Among them, P is the failure probability, x represents the sample, f(x) is the decision function, which is used to represent the distance from the sample x to the decision boundary, α is the scaling parameter, used to adjust the sensitivity of the decision function output, usually α = 0.5, and β is the scaling parameter, used to adjust the offset of the probability curve;
[0079] f(x) = w·x + b;
[0080] Among them, w is the normal vector of the hyperplane, used to determine the direction of the hyperplane, and b is the bias term of the hyperplane, used to determine the position of the hyperplane;
[0081] It should be noted that the trained fault prediction model is usually saved as a file for loading and use in the subsequent prediction stage. The purpose of loading the fault prediction model is to restore the trained model parameters and structure to the current running environment so that it can perform fault prediction on new data;
[0082] Define the fault probability threshold Y. When P > Y, a warning is triggered and a warning message is generated;
[0083] When P < Y, it is considered that the current operating state of the device is stable;
[0084] It should be noted that the fault probability threshold is used to determine when to trigger a fault warning. The selection of this threshold needs to comprehensively consider the performance of the model, business requirements, the costs of false alarms and missed alarms, etc. By evaluating the performance of the model on the validation set, a threshold that can balance accuracy and recall is selected. A false alarm means that the model wrongly predicts a fault, and a missed alarm means that the model fails to predict an actually occurring fault. If the cost of false alarms is high, a higher threshold can be selected to reduce false alarms; if the cost of missed alarms is high, a lower threshold can be selected to reduce missed alarms; when P > Y, it is considered that the device has a high fault risk and a warning needs to be triggered. When P < Y, it is considered that the current operating state of the device is stable and there is no obvious fault risk. In this case, the operating state of the device will continue to be monitored and the fault probability will be calculated regularly to ensure that any state change can be captured in a timely manner;
[0085] The warning message is encapsulated in JSON format and sent to the distributed diagnostic system through the MQTT protocol;
[0086] It should be noted that JSON is a lightweight data exchange format that is easy to read and write, and also easy for machines to parse and generate; MQTT is a lightweight message transmission protocol widely used in the Internet of Things (IoT) field. It supports message publishing and subscription between devices and features low bandwidth, high reliability, and ease of implementation. When sending alert information, a connection needs to be established with the MQTT broker server first, and then the alert information in JSON format is published to the specified MQTT topic. The distributed diagnostic system needs to subscribe to the relevant MQTT topic to receive and process these alert information in real time, and finally confirm message delivery.
[0087] Intelligent Diagnosis Module: The distributed intelligent diagnosis system identifies IoT nodes that need diagnosis based on early warning information. The IoT nodes perform preliminary diagnosis on the collected data according to preset rules and generate preliminary diagnosis results.
[0088] The warning information includes the type of fault, the expected time of occurrence, the affected components, and the severity of the fault;
[0089] It should be noted that the fault type refers to the specific manifestation of the fault, including abnormal current, abnormal temperature, abnormal vibration, and abnormal displacement; the expected occurrence time refers to the specific time point at which the fault is predicted to occur, including the time format and time range; the fault severity refers to the degree of impact of the fault on the operation and stability of the equipment, which is usually divided into three ranges: "low", "medium", and "high"; the affected equipment components include "motors", "controllers", "sensors", etc.
[0090] The distributed intelligent diagnostic system receives and parses warning information via the MQTT protocol.
[0091] It should be noted that the received warning information is data encapsulated in JSON format. The distributed intelligent diagnostic system needs to use parsing tools to parse this data and extract fields such as fault type, expected occurrence time, and fault severity.
[0092] Based on the parsed early warning information and combined with the predefined mapping relationship between components and nodes, the intelligent diagnostic system identifies IoT nodes related to the early warning information;
[0093] It should be noted that in the distributed intelligent diagnostic system, a mapping relationship is predefined to associate device components and IoT nodes. This mapping relationship is usually a table or database that records the IoT node information corresponding to each component. This mapping relationship can quickly locate the IoT nodes related to the warning information. The distributed intelligent diagnostic system identifies the IoT nodes related to the warning information based on the "affected component" field in the parsed warning information and the predefined mapping relationship.
[0094] The preset rules include threshold rules, combination rules, pattern matching rules, time series rules, and anomaly detection rules;
[0095] It should be noted that threshold rules determine the presence of anomalies based on thresholds for specific features. If a feature value exceeds a set threshold, a fault is considered possible. These thresholds include current thresholds, temperature change rate thresholds, and vibration frequency thresholds. Combination rules combine multiple features for comprehensive judgment. If multiple features simultaneously meet specific conditions, a fault is considered possible. For example, if both current and temperature exceed thresholds, a severe overload is considered possible. Pattern matching rules identify specific data patterns, such as periodic fluctuations or abrupt changes. If a specific pattern appears in the data, a fault is considered possible. This includes periodic fluctuation and abrupt change detection. Time series rules analyze the time series trend of the data, such as continuous rise or fall. If the time series trend of the data is abnormal, a fault is considered possible. Anomaly detection rules identify outliers in the data, such as abrupt changes in current or temperature. If an outlier appears in the data, a fault is considered possible.
[0096] IoT nodes collect real-time operational data of platform doors and key components, and extract operational data features;
[0097] Based on preset rules, the characteristics of the running data are preliminarily diagnosed, a preliminary diagnosis result is generated, and the preliminary diagnosis result is sent to the central diagnostic system.
[0098] It should be noted that the preliminary diagnostic results usually include the diagnostic status, such as "normal", "abnormal current", "abnormal temperature", "vibration problem", etc., and related data, such as the average current, the rate of temperature change, the frequency of vibration, etc. The IoT node encapsulates the preliminary diagnostic results into JSON format and sends them to the central diagnostic system via the MQTT protocol.
[0099] Fault Decision Module: Performs comprehensive analysis on the preliminary diagnostic results, determines the location and cause of the fault based on the comprehensive analysis results, and generates fault handling suggestions.
[0100] The central diagnostic system collects preliminary diagnostic results from each IoT node and analyzes the correlation between these preliminary diagnostic results.
[0101] Identify the fault mode based on the fault manifestations in the preliminary diagnostic results;
[0102] It should be noted that the fault manifestation form refers to the specific description of the potential problem judged by the IoT node according to preset rules, including abnormal current, abnormal temperature, abnormal vibration, and abnormal displacement. The fault manifestation form is extracted from the preliminary diagnosis results and matched with known fault modes. If the preliminary diagnosis results mention "abnormal current", the matched fault mode may be "excessive current" or "current fluctuation". Based on the matching results, a specific fault mode description is generated. If multiple nodes report "abnormal current", the generated fault mode description may be "power supply system failure". If a node reports "abnormal temperature" and "vibration problem", the generated fault mode description may be "mechanical component overload".
[0103] By comparing real-time operating data with historical normal operating data, data points that significantly deviate from the normal range can be identified.
[0104] It should be noted that real-time operational data refers to the device operating status data collected from IoT nodes at the current moment, while historical normal operation data refers to historical data accumulated by the device under normal operating conditions. Comparative analysis typically includes: data alignment to ensure that real-time operational data and historical normal operation data are aligned in terms of time, sampling frequency, and data format; difference calculation: for each data point, calculate the difference between real-time operational data and historical normal operation data; threshold setting: set a threshold based on the statistical characteristics of historical data; and outlier data identification: if the difference of a data point exceeds the set threshold, the data point is considered to have significantly deviated from the normal range.
[0105] Through correlation analysis, we can explore whether current anomalies in different types of data are accompanied by temperature increases, and whether changes in vibration frequency are related to specific operating cycles.
[0106] It should be noted that correlation analysis is a data analysis method used to explore the relationships between different variables. In intelligent operation and maintenance systems, correlation analysis can help identify potential connections between different operational data characteristics, thereby providing a more comprehensive understanding of the equipment's operating status and potential faults. During equipment operation, excessive current or current fluctuations may affect the equipment's temperature. By collecting current and temperature data and calculating their correlation, it can be determined whether abnormal current is accompanied by a rise in temperature. The vibration frequency of the equipment may be related to specific operating cycles (such as equipment startup, shutdown, acceleration, deceleration, etc.). By collecting vibration frequency data and operating cycle data and using time series analysis methods, it can be determined whether changes in vibration frequency are related to specific operating cycles.
[0107] Based on the comprehensive analysis results, the source of the fault is narrowed down, and the location and cause of the problem are identified.
[0108] Based on the equipment failure mode and preliminary diagnostic results, infer the cause of the failure;
[0109] It should be noted that if "abnormal current" occurs and the equipment has an overload history, it is because the load is too large, resulting in excessive current. If "abnormal temperature" is accompanied by "vibration problem", it is because the mechanical parts are loose or worn, resulting in poor heat dissipation. If voltage fluctuations occur at multiple nodes at the same time, it is a problem with the power supply.
[0110] Based on the severity of the fault, determine its urgency, and develop a maintenance plan and fault handling recommendations based on the cause of the fault.
[0111] It should be noted that the troubleshooting recommendations include maintenance time, components to be inspected, parts to be replaced, parameter adjustments, and cleaning. If the fault involves multiple components, the repair sequence needs to be coordinated to ensure that the repair process is efficient and orderly.
[0112] Optimize the closed-loop module: The fault prediction model dynamically adjusts the model parameters based on fault handling suggestions, retrains the fault prediction model using the updated parameters, and continues to use it for fault prediction, forming a complete closed loop.
[0113] Collect fault handling suggestions and actual maintenance results from the operation and maintenance management system, and organize the collected data into a structured form;
[0114] It should be noted that the operation and maintenance management system is a system used to record and manage equipment operation and maintenance information. The actual maintenance results include the actual maintenance time, the replaced parts, and the equipment status after maintenance. Structured data is usually stored in tabular form, with each row representing a record and each column representing an attribute, which facilitates subsequent data analysis and model optimization.
[0115] Analyze the deviation between the prediction results of the fault prediction model and the actual faults, and identify the fault prediction model parameters that need to be adjusted based on the results of the deviation analysis.
[0116] It should be noted that the accuracy of the fault prediction model is identified by comparing its prediction results with the actual fault conditions. The parameters are then adjusted using a grid search method, a commonly used parameter optimization method that finds the optimal combination of parameters for the fault prediction model by traversing a predefined parameter grid.
[0117] The compiled fault handling suggestions and maintenance results data are merged with historical data to form a new training dataset. The fault prediction model is then retrained using the new training dataset and updated parameters.
[0118] It should be noted that historical data refers to data that has been previously used to train the fault prediction model. This data includes past fault records, maintenance records, and related operational data features. Merging the organized fault handling suggestions and maintenance results data with historical data can increase the quantity and diversity of training data, improve the generalization ability of the fault prediction model, ensure that the fault prediction model can adapt to new data distributions, and improve prediction performance.
[0119] The retrained fault prediction model is updated into the fault prediction system to replace the old fault prediction model, and new fault handling suggestions and maintenance results are collected regularly.
[0120] It should be noted that the retrained model should be saved to a file for later loading and use. In order to continuously optimize the fault prediction model, it is necessary to collect new fault handling suggestions and maintenance results regularly. These data will be used for subsequent fault prediction model training and optimization. Define the data collection time interval, such as collecting new fault handling suggestions and maintenance results once a week or once a month.
[0121] In summary, this invention, through a distributed intelligent diagnosis and comprehensive analysis mechanism, quickly and accurately locates faults, analyzes their causes, and generates targeted handling suggestions, thereby improving operation and maintenance efficiency. It also dynamically adjusts model parameters based on actual fault handling results, continuously optimizing system performance, enhancing system adaptability and stability, accelerating fault response speed, improving fault prediction accuracy, reducing false alarm rates, and decreasing reliance on manual inspections. This reduces operation and maintenance costs and workload, ensuring the operational safety and efficiency of urban rail transit.
[0122] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A platform screen door intelligent operation and maintenance system, characterized in that: include, Data acquisition and processing module: Acquires real-time running data and preprocesses it, extracts running data features and stores them; Fault prediction model building module: Builds a fault prediction model, trains the fault prediction model using runtime data features, and generates a trained fault prediction model; Fault early warning module: Inputs real-time operating data into the trained fault prediction model, generates early warning information, and sends the early warning information to the distributed intelligent diagnostic system; Intelligent Diagnosis Module: The distributed intelligent diagnosis system identifies IoT nodes that require diagnosis based on early warning information. These IoT nodes then perform preliminary diagnosis on the collected data according to preset rules, generating preliminary diagnostic results. The specific steps are as follows: The warning information includes the fault type, the expected time of occurrence, the affected components, and the severity of the fault; Fault types include abnormal current, abnormal temperature, abnormal vibration, and abnormal displacement; predicted occurrence time includes the predicted time point and time range of the fault; fault severity is categorized as low, medium, and high; affected equipment components include motors, controllers, and sensors. The distributed intelligent diagnostic system receives and parses warning information via the MQTT protocol. Based on the parsed early warning information and combined with the predefined mapping relationship between components and nodes, the intelligent diagnostic system identifies IoT nodes related to the early warning information; The mapping relationship between components and nodes is used to associate device components and IoT nodes, and to record the IoT node information corresponding to each component; The distributed intelligent diagnostic system identifies IoT nodes related to the warning information based on the affected component fields in the parsed warning information and the mapping relationship between components and nodes. The preset rules include threshold rules, combination rules, pattern matching rules, time series rules, and anomaly detection rules; Threshold rules determine the presence of anomalies based on thresholds for specific features. If a feature value exceeds a set threshold, a fault risk is identified. These thresholds include current thresholds, temperature change rate thresholds, and vibration frequency thresholds. Combination rules combine multiple features for comprehensive judgment. If multiple features simultaneously meet specific conditions, a fault risk is identified. For example, when both current and temperature exceed their corresponding thresholds, a severe overload risk is identified. Pattern matching rules identify specific data patterns, including periodic fluctuations and abrupt changes. Time series rules analyze the time series trend of data, including continuous upward and continuous downward trends. If the time series trend of the data is abnormal, a fault risk is identified. Anomaly detection rules identify outliers in the data, including sudden current and temperature changes. IoT nodes collect real-time operational data of platform doors and key components, and extract operational data features; Based on preset rules, the characteristics of the running data are preliminarily diagnosed, a preliminary diagnosis result is generated, and the preliminary diagnosis result is sent to the central diagnostic system. The preliminary diagnostic results include diagnostic status and related data. The diagnostic status includes normal, abnormal current, abnormal temperature, and vibration problems. The related data includes the mean current, the rate of temperature change, and the frequency of vibration. Fault Decision Module: Performs comprehensive analysis on the preliminary diagnosis results, determines the location and cause of the fault based on the comprehensive analysis results, and generates fault handling suggestions; The fault manifestations are extracted from the preliminary diagnostic results. The fault manifestations refer to the specific descriptions of potential problems judged by the IoT nodes according to preset rules. The fault manifestations include abnormal current, abnormal temperature, abnormal vibration, and abnormal displacement. The extracted fault manifestations are matched with known fault modes, and fault mode descriptions are generated based on the matching results. Specifically, when the fault manifestation extracted from the preliminary diagnosis results is abnormal current, the fault manifestation is matched with the high current fault mode and the current fluctuation fault mode. When multiple nodes report abnormal current, a fault mode description of the power supply system fault is generated. When the same node reports abnormal temperature and vibration problems, a fault mode description of mechanical component overload is generated. If the preliminary diagnosis shows abnormal current and the device has a history of overload, it is determined that the load is too large and the current is too high; if the preliminary diagnosis shows abnormal temperature accompanied by vibration problems, it is determined that the mechanical parts are loose or worn and the heat dissipation is poor; if the data collected by multiple IoT nodes shows voltage fluctuations, it is determined that there is a problem with the power supply. Optimize the closed-loop module: The fault prediction model dynamically adjusts its parameters based on fault handling suggestions. The updated parameters are then used to retrain the fault prediction model, which continues to be used for fault prediction, forming a complete closed loop. The specific steps are as follows. Collect fault handling suggestions and actual maintenance results from the operation and maintenance management system, and organize the collected data into a structured form; Analyze the deviation between the prediction results of the fault prediction model and the actual faults, and identify the fault prediction model parameters that need to be adjusted based on the results of the deviation analysis. The compiled troubleshooting suggestions and maintenance results data are merged with historical data to form a new training program. The dataset was used to retrain the fault prediction model using the new training dataset and updated parameters. The retrained fault prediction model is updated into the fault prediction system, replacing the old fault prediction model, and new fault handling suggestions and maintenance results are collected regularly.
2. The intelligent operation and maintenance system for platform screen doors as described in claim 1, characterized in that: The real-time operating data includes electrical operating data, environmental operating data, mechanical operating data, and control operating data; The preprocessing includes smoothing and noise reduction. The operational data characteristics include statistical characteristics, time-series characteristics, and frequency domain characteristics; Store runtime data characteristics in a relational database.
3. The intelligent operation and maintenance system for platform screen doors as described in claim 2, characterized in that: The process of constructing a fault prediction model involves training the model using operational data features to generate a trained fault prediction model. The specific steps are as follows: The support vector machine algorithm is selected to construct the fault prediction model, and the fault prediction model parameters are initialized. The operational data features are divided into training and testing sets; The fault prediction model adjusts its internal parameters based on the training set data; The trained fault prediction model is validated using a test set, and validation results are generated. Based on the validation results, the parameters of the fault prediction model were optimized using the grid search method to generate a trained fault prediction model.
4. The intelligent operation and maintenance system for platform screen doors as described in claim 3, characterized in that: The process of inputting real-time operational data into a trained fault prediction model to generate early warning information, and then sending the early warning information to a distributed intelligent diagnostic system, involves the following specific steps: Load the trained fault prediction model, input the real-time running data features into the fault prediction model, and calculate the fault occurrence probability P; Define a fault probability threshold Y. When P > Y, an early warning is triggered and an early warning message is generated. If P < Y, the current operating state of the device is considered stable. The warning information is encapsulated in JSON format and sent to the distributed diagnostic system via the MQTT protocol.
5. The intelligent operation and maintenance system for platform screen doors as described in claim 4, characterized in that: The specific steps for comprehensively analyzing the preliminary diagnostic results are as follows. The central diagnostic system collects preliminary diagnostic results from each IoT node and analyzes the correlation between these preliminary diagnostic results. Identify the fault mode based on the fault manifestations in the preliminary diagnostic results; By comparing real-time operating data with historical normal operating data, data points that significantly deviate from the normal range can be identified. Through correlation analysis, we can explore whether current anomalies in different types of data are accompanied by temperature increases, and whether changes in vibration frequency are related to specific operating cycles.
6. The intelligent operation and maintenance system for platform screen doors as described in claim 5, characterized in that: The steps for determining the location and cause of the fault based on the comprehensive analysis results and generating fault handling suggestions are as follows. Based on the comprehensive analysis results, the source of the fault is narrowed down, and the location and cause of the problem are identified. Based on the equipment failure mode and preliminary diagnostic results, infer the cause of the failure; Based on the severity of the fault, determine its urgency, identify the cause of the fault, develop a repair plan, and generate fault handling suggestions.