Truck fault identification method, device and equipment based on TFDS, and storage medium
By performing fault matching and weight accumulation in the TFDS system, the problems of low efficiency and insufficient accuracy in truck fault identification are solved, enabling accurate identification and efficient repair of accident vehicles.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUITIE TECH CO LTD
- Filing Date
- 2026-06-22
- Publication Date
- 2026-07-21
AI Technical Summary
The existing TFDS system cannot integrate the results of individual component failures when identifying faults in railway freight cars, resulting in low identification efficiency, high false alarm and false negative rates, and difficulty in accurately identifying accident vehicles.
By acquiring the identification results of truck components and storing them in the database and cache, fault matching is performed based on the preset fault list of accident vehicles. The fault weight score of the truck is calculated by accumulating fault weights, and the accident vehicle is marked when the preset threshold is reached, and a repair work order is pushed.
It improved the accuracy of accident vehicle identification, reduced the probability of missed and false judgments, and realized a closed-loop process from intelligent identification to maintenance work order generation, thereby improving the safety of railway freight car operation and maintenance efficiency.
Smart Images
Figure CN122433015A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of railway freight car fault detection technology, and in particular to a freight car fault identification method, device, equipment and storage medium based on TFDS. Background Technology
[0002] In recent years, the Trouble of Moving Freight Car Detection System (TFDS) has been widely used in the field of railway freight safety. TFDS is a core piece of equipment for ensuring the safe operation of railway freight cars. It collects vehicle images through trackside equipment and uses intelligent recognition models to complete component fault detection.
[0003] In existing technologies, the TFDS intelligent identification model can only output discrete and independent single component fault results, and fault prediction relies on simple threshold judgments for single components. However, freight trains consist of a large number of vehicles, and each vehicle has a large number of components to be detected, resulting in a massive volume of fault identification data. Furthermore, in batch processing of multiple vehicles, the identification results of multiple components of the same vehicle arrive asynchronously, which can easily lead to accident vehicles being marked repeatedly.
[0004] In addition, since vehicles in freight trains that pose an accident risk are usually accompanied by multiple components and various types of faults, discrete results usually cannot be fused for judgment, making it difficult to accurately identify accident vehicles and resulting in high false alarm and false negative rates; the identification results of multiple components arrive asynchronously, leading to low identification efficiency and insufficient data consistency and accuracy. Summary of the Invention
[0005] The main purpose of this application is to provide a freight car fault identification method, device, equipment and storage medium based on TFDS, which aims to solve the problem that the faulty parts of a single freight car in a freight train are numerous, single and discrete, and the results cannot be fused and judged, resulting in low identification efficiency and high false alarm rate and false negative rate.
[0006] To achieve the above objectives, in a first aspect, this application provides a truck fault identification method based on TFDS, comprising: Obtain the component identification results of the truck output by the TFDS intelligent recognition system for truck operation faults, and store the component identification results in the database and cache respectively; Based on a preset list of faults in accident vehicles, the identification results of the components are matched to obtain target fault information. The fault weight score of the truck is calculated based on the target fault information using a fault weight accumulation method. When the fault weight score reaches a preset weight threshold and there is no accident vehicle identifier corresponding to the truck in the cache, the truck is marked as an accident vehicle and the corresponding accident vehicle identifier is written into the cache. The identified accident vehicle is pushed to the maintenance platform to generate a maintenance work order to be processed.
[0007] Optionally, the target fault information includes: faulty components and their corresponding weight values; the calculation of the truck's fault weight score based on the target fault information using a fault weight accumulation method includes: Based on the faulty components of the truck and their corresponding weight values, the weight values of all matching faults of the truck are summed using the following weight accumulation formula to obtain the truck's fault weight score: S=ΣW i Where S is the fault weight score, W i Let Σ be the weight value corresponding to the i-th matched fault, and let Σ be the summation symbol; the weight value is set based on the severity of the fault of the truck component.
[0008] Optionally, the method further includes: Collect TFDS fault data of multiple historical accident vehicles, calculate the fault weight score of each accident vehicle, and obtain the fault weight score distribution of the accident vehicles. Collect false alarm fault data from multiple normally operating trucks caused by occasional factors, calculate the fault weight score for each normally operating truck, and form the fault weight score distribution of the normally operating trucks. Based on the fault weight score distribution of normal vehicles and the fault weight score distribution of accident vehicles, the optimal score is selected as the preset weight threshold according to at least one of the ROC curve, precision-recall curve or Youden index maximization.
[0009] Optionally, storing the component identification results in a database and a cache respectively includes: Fault detection information is extracted from the component identification results. The fault detection information includes vehicle passage time, vehicle sequence, fault name, fault code, and component location information. The extracted fault detection information is stored in a database in a structured form; The extracted fault detection information is cached in a high-speed cache.
[0010] Optionally, the fault list of the accident vehicle includes: fault name, fault code, fault type, fault level, fault weight value, fault classification, and fault range; the step of performing fault matching on the component identification results based on the preset fault list of the accident vehicle to obtain target fault information includes: Extract the fault name or fault code from the fault detection information in the component identification results of the truck as the information to be matched; The information to be matched is compared with the pre-stored fault entries in the preset fault list of accident vehicles, and the fault information belonging to the preset fault list of accident vehicles is selected as the target fault information.
[0011] Optionally, the method further includes: In response to the pending maintenance work order, maintenance operations are performed on the truck corresponding to the accident vehicle identifier, and the maintenance results are fed back to the operation and maintenance platform.
[0012] Secondly, this application provides a truck fault identification device based on TFDS, comprising: The data storage module is used to acquire the component identification results of the truck output by TFDS intelligent recognition, and store the component identification results in the database and the cache respectively; The fault matching module is used to perform fault matching on the component identification results based on a preset fault list of accident vehicles in order to obtain target fault information; The scoring calculation module is used to calculate the fault weight score of the truck based on the target fault information using a fault weight accumulation method. The accident determination module is used to mark the truck as an accident vehicle and write the corresponding accident vehicle identifier into the cache when the fault weight score reaches a preset weight threshold and there is no accident vehicle identifier corresponding to the truck in the cache. The result push module is used to push the accident vehicle identifier to the operation and maintenance platform to generate a maintenance work order to be processed.
[0013] Thirdly, this application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the truck fault identification method based on TFDS provided in the foregoing embodiments.
[0014] Fourthly, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the truck fault identification method based on TFDS provided in the foregoing embodiments.
[0015] The beneficial effects of the embodiments of this application are as follows: Based on the component identification results automatically detected for key parts of freight cars by TFDS intelligent identification output, and combined with a preset accident vehicle fault list matching mechanism and fault weight cumulative score, judgment can be made. This abandons the subjective judgment method of single faults, and calculates the vehicle's fault weight score by quantifying fault levels and weights, thus quantifying the fault risk of freight cars. Accident vehicles can be automatically marked when a preset weight threshold is reached, significantly improving the accuracy of accident vehicle identification and judgment, and effectively reducing the probability of missed or false judgments. Furthermore, by introducing a high-speed cache to store fault identification results and accident vehicle identifiers, the uniqueness, accuracy, and timeliness of accident vehicle identifiers can be ensured. Additionally, by pushing and generating pending maintenance work orders, a closed-loop processing system can be achieved from intelligent identification, quantitative judgment, accurate identification to work order issuance, improving the safety of railway freight car operation and the efficiency of maintenance work. Attached Figure Description
[0016] Figure 1 A flowchart illustrating a truck fault identification method based on TFDS provided in this application embodiment; Figure 2 A flowchart of another truck fault identification method based on TFDS provided for embodiments of this application; Figure 3 A block diagram of a truck fault identification device based on TFDS provided in this application embodiment; Figure 4 This is a block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation
[0017] The technical solutions of 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 a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0018] In related technologies, TFDS is a safety inspection equipment that automatically identifies faults in train components by acquiring images of freight cars in operation via trackside devices. The system uses linear or area array high-speed cameras to automatically capture images of the bottom and sides of trains as they pass, transmitting these images via network to the train inspection center for manual or intelligent judgment models to determine the degree of fault based on the images. The system is suitable for train speeds of 5–120 km / h, typically capturing 53 images per car, covering key components such as couplers, side frames, bogies, and underframes.
[0019] Freight trains consist of locomotives and freight cars coupled together in a fixed sequence to transport goods by rail. Vehicle components include dozens of key parts such as bogies, brake beams, couplers and buffers, side frames, load-bearing saddles, crossbars, braking systems, wheel axles, underframes, intermediate sections, and suspension components, which are the primary targets of TFDS (Traffic Transaction Disposal System) testing.
[0020] Figure 1 A flowchart illustrating a TFDS-based truck fault identification method is provided for embodiments of this application. This method can be applied to a truck accident vehicle identification system. (Refer to...) Figure 1 The method includes: Step S101: Obtain the component identification results of the truck output by TFDS intelligent recognition, and store the component identification results in the database and cache respectively; It's important to note that the cache can be a Redis (Remote Dictionary Server) cache. Redis, as a high-performance in-memory database, is used to temporarily store frequently accessed data in memory, reducing direct access to the backend database. Redis caching can quickly store, retrieve, and update vehicle weight scores and accident vehicle identifiers, supporting high concurrency.
[0021] In this embodiment, the truck accident vehicle identification system receives the truck component identification results output by TFDS intelligent identification in real time. These component identification results are generated by capturing images of the truck from a high-speed trackside camera, which are then detected by the intelligent identification model. The results include vehicle identification information, basic component information, fault identification information, and image information. The vehicle identification information includes station number, passing time, and vehicle sequence. The basic component information includes component number, component type, and component location. The fault identification information includes fault name, fault code, fault area, and fault level. The image information includes a high-resolution image of the component, a fault marker image, and a unique image identifier. The system pre-establishes an interactive connection with a relational database and performs structured parsing of the component identification results according to a preset table structure. The vehicle identification information, basic component information, fault identification information, and image information are persistently stored in the corresponding data tables of the database. The system constructs a unique cache key-value pair using the combination of station number, passing time, and vehicle sequence, and synchronously stores key data from the component identification results in a Redis cache. This key data includes component number, fault name, fault weight, and cumulative vehicle fault score. The database can be a relational database such as MySQL or Oracle; the Redis cache can be a Redis cache cluster.
[0022] Therefore, the component identification results can be permanently stored in a database for subsequent historical data queries, fault tracing, and model training. Furthermore, Redis's high-speed read / write capabilities can be leveraged to quickly retrieve component identification results and fault score data, reducing the pressure on the backend database and improving the processing efficiency of the accident vehicle assessment process.
[0023] Step S102: Based on the preset fault list of the accident vehicle, perform fault matching on the component identification results to obtain the target fault information; In this embodiment, based on a pre-built and stored list of accident vehicle faults, the fault matching processing of the freight car component identification results output by TFDS intelligent identification can be performed to obtain the corresponding target fault information. The list of accident vehicle faults can be pre-built based on TFDS fault samples of historical railway accident vehicles and combined with an expert decision-making model. The list defines the fault name, fault code, fault type, fault level, fault weight value, fault classification and fault range corresponding to each fault. The key fields such as fault name, fault type and component location parsed from the component identification results are compared and matched one by one with the items in the pre-built list of accident vehicle faults.
[0024] By traversing each fault record in the component identification results, valid fault entries that match the fault names in the fault list of the accident vehicle can be filtered out. Corresponding parameters such as fault name, fault level, fault weight value, and fault classification corresponding to the valid fault entries can be extracted to generate structured target fault information. New unknown faults in the component identification results that do not match the corresponding entries in the fault list of the accident vehicle can be identified separately and reported to the operation and maintenance platform.
[0025] Therefore, it can be seen that the standardized pre-set fault list of accident vehicles is used to standardize the fault matching of the component identification results output by TFDS intelligent identification, unify the fault judgment criteria and evaluation standards, and avoid the bias problem caused by subjectivity. By comparing each field one by one, valid faults are quickly screened and normal interference information is filtered out. Key parameters such as fault level and fault weight value are accurately extracted, providing reliable data support for subsequent weight accumulation scoring and pre-set weight threshold judgment. At the same time, unknown faults outside the list are automatically reported to manual verification and the fault list is iteratively updated to realize the continuous optimization of fault matching rules. While ensuring the accuracy and consistency of fault matching, the fully automatic processing capability of the truck accident vehicle identification system is improved.
[0026] Step S103: Calculate the fault weight score of the truck based on the target fault information using the fault weight accumulation method; Step S104: When the fault weight score reaches the preset weight threshold and there is no accident vehicle identifier corresponding to the truck in the cache, mark the truck as an accident vehicle and add the accident vehicle identifier corresponding to the accident vehicle to the cache. In a high-speed cache such as Redis, three types of key-value data can be maintained: the component failure weight key: weight:{fault code}:{vehicle order}, which stores the current component failure weight value of a single freight car; the failure weight score key: score:{passage time}:{vehicle order}, which stores the failure weight score of a single freight car after accumulating the component failure weight values; and the accident vehicle identifier key: accident_flag:{passage time}:{vehicle order}, which is used to store a unique identifier for whether the freight car is marked as an accident vehicle.
[0027] The key-value structure parameters for the accident vehicle identifier include a key format, a value, and an expiration time (TTL, Time to Live). The key format is, for example, `accident_flag:{passage time}:{vehicle sequence}`, which uniquely identifies a vehicle's passage record using the passage time and vehicle sequence. A value of 1 or `true` indicates that the truck has been marked as an accident vehicle, while 0 or `false` represents a non-accident vehicle. The TTL can be dynamically adjusted based on the marshalling cycle to automatically clear expired identifiers and avoid data redundancy; for example, it can be configured to 86400 seconds.
[0028] For steps S103 and S104, after the truck accident vehicle identification system obtains the matched target fault information, it extracts the fault weight score corresponding to each target fault information. Taking a single truck as a unit, a fault weight accumulation calculation model is established, and the weight scores of all target fault information corresponding to that truck are accumulated one by one to calculate the fault weight score of that truck. If the fault weight score does not reach the preset weight threshold, the accident vehicle marking process is not triggered, and the fault weight score calculated this time is only updated to the Redis cache for subsequent fault accumulation calculation. If the fault weight score reaches the preset weight threshold, the system further queries the Redis cache to verify whether the accident vehicle identifier corresponding to that truck already exists.
[0029] Specifically, the truck accident identification system first reads the truck's currently accumulated fault weight score from the Redis cache (if the truck is being assessed for fault for the first time and has no historical accumulated fault weight score, it is initialized to 0). Then, it iterates through each target fault information of the truck, sequentially adding the weight score corresponding to each fault to the current accumulated score, and updating the truck's fault weight score in real time. During the accumulation process, the system synchronously records the weight score, accumulation time, and corresponding fault name of each fault, forming a complete weight accumulation record for subsequent fault tracing and verification, ensuring the traceability and accuracy of fault weight score calculation, and avoiding the occurrence of missed or incorrect weight additions.
[0030] If the truck's accident vehicle identifier already exists in the Redis cache, it means the truck has been marked as an accident vehicle. To avoid duplicate marking and data redundancy, the system directly terminates the current marking process. If the truck's accident vehicle identifier does not exist in the Redis cache, the system immediately marks the truck as an accident vehicle, synchronously generates a unique accident vehicle identifier corresponding to the accident truck (the identifier includes key information such as basic vehicle information, fault weight score, and marking time), and writes the accident vehicle identifier into the Redis cache according to a preset key-value format. At the same time, a reasonable cache expiration time is configured to enable fast querying and automatic cleanup of accident vehicle identifiers. This ensures the uniqueness and accuracy of accident vehicle marking, avoids Redis cache data redundancy, and provides a reliable identification basis for the generation of subsequent maintenance work orders.
[0031] For example, the preset weight threshold can be 2.5. When the fault weight score of the truck obtained through historical accumulation is 1, the accident vehicle marking operation is not triggered because 1 < 2.5. If the axle end bolt is currently identified as broken, loose, or missing, the level is high and the weight is 2. Then the fault weight score obtained by accumulating the current score and the historical score is 3. Since 3 > 2.5, the accident vehicle marking operation is triggered. The system starts to query whether there is an accident vehicle identifier for the truck in the Redis cache. If there is no accident vehicle identifier for the truck, the truck can be marked as an accident vehicle and the accident vehicle identifier can be added to the Redis cache.
[0032] Step S105: Push the accident vehicle identification of the truck to the operation and maintenance platform to generate a maintenance work order to be processed.
[0033] In this embodiment, after the freight car accident identification system marks the accident car and writes the accident car identifier to the Redis cache, it immediately parses and extracts the corresponding key information of the freight car based on the accident car identifier, including the station number, passing time, vehicle sequence, component location, fault name, cumulative fault weight score, and judgment time. The system encapsulates the above information according to the message format specified by the operation platform and actively pushes it to the railway operation and maintenance platform through a preset communication interface. After receiving the accident car push information, the operation and maintenance platform automatically parses the key data and generates a pending maintenance work order containing vehicle information, fault location, fault level, and handling priority according to the maintenance business process. Simultaneously, it completes the work order number allocation, responsibility team assignment, and processing time limit setting, and updates the work order status to pending acceptance. This is used to remind maintenance personnel to conduct on-site verification, fault confirmation, and targeted maintenance work on the freight car in a timely manner, realizing closed-loop management of the entire process from intelligent identification and accident judgment to maintenance handling.
[0034] Optionally, step S101 further includes: Sub-step S1011: Extract fault detection information from the component identification results. The fault detection information includes vehicle passage time, vehicle sequence, fault name, fault code, and component location information.
[0035] Sub-step S1012: Store the extracted fault detection information in a structured form in the database.
[0036] Sub-step S1013: Cache the extracted fault detection information into the cache.
[0037] For sub-steps S1011 to S1013, the component identification results output by TFDS intelligent identification are first analyzed to extract key fault detection information. This information uniquely identifies each fault detection record, providing basic data support for subsequent fault matching, weight calculation, and accident vehicle identification. The truck accident vehicle identification system persistently stores the extracted fault detection information in a structured format in a relational database according to a preset data table structure, achieving permanent storage, unified management, and historical traceability of fault data, facilitating subsequent query statistics, fault analysis, and model optimization. Simultaneously, unique key-value pairs can be constructed based on vehicle passage time and vehicle sequence, synchronously caching identical fault detection information in Redis. Utilizing Redis's high-speed read / write capabilities enables rapid access and retrieval of fault data, reducing frequent access to the backend database, improving overall system processing efficiency, and providing efficient data support for subsequent fault matching, weight accumulation, and accident vehicle identification verification.
[0038] Optionally, the fault list of the accident vehicle may include: fault name, fault code, fault type, fault level, fault weight value, fault classification, and fault range.
[0039] Step S102 further includes: Sub-step S1021: Extract the fault name or fault code from the fault detection information in the truck component identification results as the information to be matched; Sub-step S1022: Compare the information to be matched with the pre-stored fault entries in the preset accident vehicle fault list, and filter out the fault information that belongs to the accident vehicle fault list as the target fault information.
[0040] For sub-steps S1021 and S1022, the identification results of freight car components output from TFDS intelligent identification can be parsed and the fault name or fault code in the fault detection information can be extracted and used as the matching information to be matched for fault matching. Among them, the fault name is a professional name such as slight deformation of side frame, loose fastener, excessive gap between coupler and buffer, etc., and the fault code is a unique fault identification code uniformly compiled by the railway industry. Either the fault name or the fault code can be used as the matching basis.
[0041] The extracted information to be matched is searched and compared with each standard fault item in the pre-stored fault list of accident vehicles. Based on the component range, vehicle type applicability conditions (such as the 367 type truck and other regular truck models) and fault type rules defined in the fault list, valid fault records that are fully included in the fault list specifications are selected and identified as target fault information. Abnormal identification information, regular minor wear information and non-applicable vehicle model fault information that are not included in the fault list are filtered out and not included in the subsequent weight accumulation and accident vehicle judgment process, so as to ensure that the matching results are compliant and accurate and reduce interference from invalid data.
[0042] Optionally, the target fault information includes: the faulty component and the corresponding weight value of the faulty component; step S103 further includes: Sub-step S1031: Based on the faulty components of the truck and their corresponding weight values, sum the weight values of all matching faults of the truck using the following weight accumulation formula to obtain the truck's fault weight score: S=ΣW i Where S is the fault weight score; W i Let Σ be the weight value corresponding to the i-th matched fault, and let Σ be the summation symbol. The weight value is set based on the severity of the fault of the truck component.
[0043] As an example, if the system identifies a minor side frame deformation fault in the truck's bogie, the corresponding weight score is 2; a loose brake beam component, the corresponding weight score is 1; and an excessive coupler-buffer clearance fault, the corresponding weight score is 0.5. Therefore, W1=2, W2=1, and W3=0.5. Based on the above weighted summation formula, the truck's fault weight score S=ΣW is calculated. i =2+1+0.5=3.5.
[0044] Optionally, the above method further includes: Step S106: Collect TFDS fault data of multiple historical accident vehicles, calculate the fault weight score of each accident vehicle, and obtain the fault weight score distribution of the accident vehicles.
[0045] Step S107: Collect false alarm fault data of multiple normally operating trucks caused by occasional factors, and calculate the fault weight score of each normally operating truck to form the fault weight score distribution of the normal trucks.
[0046] Step S108: Based on the fault weight score distribution of normal vehicles and the fault weight score distribution of accident vehicles, select the optimal score as the preset weight threshold according to at least one of the ROC curve, precision-recall curve or Youden index maximization.
[0047] It should be noted that the ROC (Receiver Operating Characteristic) curve is an evaluation curve plotted based on different weight thresholds, with the proportion of normal vehicles misclassified as accident vehicles on the horizontal axis and the proportion of real accident vehicles correctly identified on the vertical axis. The closer the curve is to the upper left corner, the better the threshold's discrimination effect, and it can be used to select the optimal fault weight cutoff value. Precision represents the proportion of vehicles judged as accident vehicles that are actually accident vehicles, while recall represents the proportion of all real accident vehicles that are effectively detected. The precision-recall curve reflects the relationship between the two under different weight thresholds. The score corresponding to the higher overall value of the curve is selected as the judgment threshold, balancing recognition accuracy and full coverage. The Youden index is the difference between the true positive rate and the false positive rate, used to comprehensively measure the discrimination effectiveness of the fault weight threshold in distinguishing accident vehicles from normal trucks. By traversing all candidate weight scores, the score corresponding to the maximum value of the Youden index is selected as the preset weight threshold, achieving the optimal combination of false negative and false positive rates.
[0048] In the embodiments of this application, actual TFDS fault data of multiple historically identified accident vehicles can be collected. For each truck, fault weight scores are calculated for each accident vehicle according to the faults of key components such as bogie, brake beam, and coupler buffer device. The fault weight scores of multiple accident vehicle samples are summarized to form the fault weight score distribution of accident vehicles. At the same time, TFDS false alarm fault data of multiple normally operating trucks caused by occasional factors such as image interference, lighting deviation, and normal wear of components are collected. Similarly, the cumulative value of fault weight is calculated for each vehicle, and the fault weight score distribution of normal vehicles is obtained.
[0049] For example, statistics on 500 historical accident vehicles show that their fault weight scores are mostly concentrated in the range of 2.5 to 5.0; statistics on 3,000 false alarm samples of normally operating trucks show that their fault weight scores are mostly concentrated in the range of 0 to 2.4.
[0050] Based on two sets of score distributions, different weighted scores were used as candidate weight thresholds to plot ROC curves and precision-recall curves. The Youden's index was maximized for a comprehensive screening process. By comparing the positive and negative detection rates under different thresholds, the optimal cutoff point was selected, balancing the highest accident vehicle recognition rate, the lowest false alarm rate for normal vehicles, and the peak Youden's index. After verification using multi-dimensional indicators, this application ultimately selected 2.5 as the system's preset weight threshold. At this point, the Youden's index reaches its peak, achieving the best balance between false alarm and negative detection rates, and enabling accurate quantitative differentiation between accident vehicles and normal vehicles.
[0051] This application achieves precise and quantitative determination of preset weight thresholds through the scientific screening of historical data statistical distribution and ROC / Yorden index, effectively improving the overall accuracy of accident vehicle identification while reducing the false alarm rate, and providing a reliable benchmark for subsequent fault quantification.
[0052] Optionally, the method further includes: Step S109: In response to the pending maintenance work order, perform maintenance operations on the truck corresponding to the accident vehicle identifier, and feed back the maintenance results to the operation and maintenance platform.
[0053] In the embodiments of this application, in response to the maintenance platform issuing a pending maintenance work order for the truck corresponding to the accident vehicle identification, on-site maintenance personnel can carry out targeted maintenance operations on the truck corresponding to the accident vehicle identification based on the vehicle passage time, vehicle sequence, location of faulty parts, fault type and risk level information carried in the work order; focusing on on-site verification, fault confirmation, component repair, fastening and replacement and performance verification of core faulty components such as bogies, brake beams and coupler buffer devices, to identify safety hazards and complete fault rectification.
[0054] After the maintenance work is completed, the actual maintenance results on site are collected, including whether the fault is real, the fault handling method, the component replacement status, the vehicle recovery status and re-inspection qualification information. The data is then structured and packaged according to a preset data format. The complete maintenance results, handling records and vehicle verification status are simultaneously fed back and uploaded to the operation and maintenance platform in real time. After receiving the data, the operation and maintenance platform automatically updates the work order status, retains the maintenance records, and forms a fault and closed-loop ledger. This realizes a closed-loop business process from TFDS intelligent identification, accident vehicle judgment, work order issuance, on-site maintenance to result feedback and archiving, which facilitates subsequent fault tracing, operation and maintenance statistics and rule iteration optimization.
[0055] Therefore, it can be seen that the identification results of key parts of freight cars automatically detected by TFDS intelligent recognition can be combined with the fault list matching of accident vehicles and the cumulative score of fault weights to make judgments. This abandons the single fault identification method, and calculates the vehicle's fault weight score by quantifying the fault level and weight, thus quantifying the fault risk of freight cars and significantly improving the accuracy of accident vehicle identification and judgment, effectively reducing the probability of missed and false judgments. Furthermore, by introducing a high-speed cache to store fault identification results and accident vehicle identifiers, the uniqueness, accuracy, and timeliness of accident vehicle identifiers can be ensured. When a preset weight threshold is reached, accident vehicles can be automatically marked and a pending maintenance work order can be generated, realizing a closed-loop process from intelligent identification, quantitative judgment, accurate identification to work order issuance, ensuring the safe operation of railway freight cars and the efficient implementation of maintenance work.
[0056] Figure 2A flowchart illustrating another TFDS-based truck fault identification method provided in this application embodiment. This method can be executed by a truck accident identification system, such as... Figure 2 As shown, the process is as follows: Step S201: Obtain the component identification results of the truck.
[0057] For example, a freight train consisting of 50 freight cars can have its passing images and component identification completed by TFDS.
[0058] In the embodiments of this application, component images of the 12th freight car of the freight train can be acquired via TFDS. After automatic identification by the intelligent recognition model, three component identification results are output: First identification result: crossbar deformation (fault); Second identification result: locking plate scraping (fault); Third identification result: axle end bolt broken (fault). All identification results are synchronously written to the MySQL database and temporarily cached in Redis for fast retrieval.
[0059] Step S202: Determine whether there are any faults in the preset fault list of accident vehicles.
[0060] It should be noted that the preset fault list for accident vehicles is a TFDS intelligent identification fault list, which is generated by analyzing historical TFDS images of accident vehicles and incorporating opinions from various experts. The preset fault list for accident vehicles may look like the following: Table 1: List of Faults in Accident Vehicles
[0061] The fault list of the accident vehicles can be stored as structured fields in a basic static standard database. Simultaneously, commonly used fault entries, standard weight values, and matching rules can be pre-loaded into a Redis cache for fast matching.
[0062] In this embodiment, fault matching can be performed on the faults in the component identification results to obtain target fault information related to the corresponding fault names in the preset accident vehicle fault list. The target fault information may include fault name, level, and weight. The three component identification results can be matched with the preset accident vehicle fault list. If it is determined that crossbar deformation, locking plate scratches, and shaft end bolt breakage all belong to fault types within the preset accident vehicle fault list, then these fault information are identified as target fault information. If the fault identification result is a new type of fault outside the preset accident vehicle fault list (e.g., atypical component cracks), it is directly reported to the maintenance platform for manual verification, and the fault level and weight are supplemented before being included in the judgment process. Therefore, for new types of faults not included in the preset accident vehicle fault list, the fault judgment model can be automatically reported to the maintenance platform for manual verification and supplementation of fault level and weight, continuously iterating and optimizing the fault judgment model.
[0063] Step S203: Retrieve the weight values of the component failures of the current truck from the Redis cache.
[0064] It's important to note that key-value pairs are data structures consisting of a unique key and a corresponding value, typically in the format {key:value}. For example, a key-value pair like weight:{fault code}:{vehicle sequence} can be used to store the weight values of truck component faults. Commonly used fault entries, their corresponding weight values, and matching rules can be extracted and pre-loaded into the Redis cache.
[0065] In this embodiment, the weight value of a component failure can be obtained from the corresponding Redis cache by reading the weight key corresponding to the weight value of the current truck's component failure. The weight value of the component failure corresponds to the weight value of the failure entry in the preset accident vehicle failure list. As an example, the weight of the failure code corresponding to the crossbar deformation for this truck model can be 0.5; the weight of the failure code corresponding to the locking plate scratch can be 0.5; and the weight of the failure code corresponding to the axle end bolt breakage / loosening / missing can be 2.
[0066] Step S204: Accumulate the weight values of the component failures of the current truck to obtain the truck's failure weight score, and save it to the Redis cache.
[0067] It should be noted that the key can be defined as score: {passage time}:{vehicle sequence}, and the key value corresponding to the score key is used to store the fault weight score of the vehicle sequence corresponding to the truck.
[0068] In this embodiment, the weight values of the component faults represented by the three component identification results can be accumulated to obtain the fault weight score of the current truck. The weight values of 0.5 corresponding to the fault code of crossbar deformation, 0.5 corresponding to the fault code of locking plate scratch, and 2 corresponding to the fault code of axle end bolt breakage / loosening / missing can be accumulated to obtain the fault weight score of 3 for the 12th truck. In other words, after obtaining the first component fault identification result, the value of score:20240429103000:12 is updated to 0.5 and written to the Redis cache; after obtaining the second component fault identification result, the fault weight score corresponding to the score key is read as 0.5, and the weight value of the fault code corresponding to the locking plate scratch is added as 0.5 to obtain a new fault weight score of 1, and the value of score:20240429103000:12 is updated to 1; after obtaining the third component fault identification result, the fault weight score corresponding to the score key is read as 1, and the weight value of the fault code corresponding to the shaft end bolt breakage / loosening / missing is added as 2 to obtain a new fault weight score of 3, and the value of the score key in the Redis cache is updated to 3.
[0069] Step S205: Determine whether the accumulated fault weight score of the current truck has reached the preset weight threshold.
[0070] Step S206: When the fault weight score of the current truck reaches the preset weight threshold, determine whether there is an accident vehicle identifier for the truck in the Redis cache.
[0071] In this embodiment, the weight threshold for determining an accident vehicle is set to 2.5. Through analysis of three component fault identification results, the cumulative fault weight score of the truck is 3, exceeding the preset weight threshold of 2.5. Therefore, the accident vehicle identification verification process begins. First, the Redis cache needs to be queried using the Key accident_flag:20240429103000:12. If the query result is empty, it indicates that the truck section has not yet been marked as an accident vehicle, and the process proceeds to the accident vehicle marking steps S207 and S209. If the query result is true, it indicates that the truck section has already been marked as an accident vehicle, and the process jumps directly to step S208.
[0072] Step S207: Add a suspected accident vehicle identifier to the component identification results of the truck.
[0073] Step S209: Add an accident vehicle identifier to the Redis cache to mark the truck as an accident vehicle.
[0074] For steps S207 and S209, if the accident vehicle identifier for the truck does not exist in the Redis cache, the accident vehicle marking process can be executed. First, in the component identification results, a "suspected accident vehicle fault" flag can be added to the 12th section of the truck, and the three component fault identification results are associated with the corresponding fault weight score of 3. This can be accomplished by writing the key `accident_flag:20240429103000:12`, with a value of `true` and a TTL of 86400 seconds to the Redis cache.
[0075] Step S208: Skip the process of repeatedly adding the truck as an accident vehicle marker.
[0076] For steps S208 and S210, when the weight value of the same truck component failure is repeatedly identified and reaches the preset weight threshold, the value of the accident_flag key can be obtained by querying the Redis cache. If the value of accident_flag:20240429103000:12 is 1 or true, it is determined that the truck has been marked as an accident vehicle. Then, the step of repeatedly adding the truck as an accident vehicle is skipped to avoid duplicate marking.
[0077] Step S210: Send the fault identification result-related data to the operation and maintenance platform.
[0078] In this embodiment, fault identification results-related data, including fault identification records, fault weight scores, and accident vehicle identification information, can be sent to the operation and maintenance platform. This data can be pushed to the operation and maintenance platform via the Hypertext Transfer Protocol (HTTP) interface to generate a maintenance work order. The work order is then distributed to the maintenance work area to notify maintenance personnel to conduct focused inspection and maintenance on vehicle section 12. The completion status of the maintenance work order is then fed back to the operation and maintenance platform.
[0079] In the embodiments of this application, a preset fault list matching method for accident vehicles is adopted, combined with a quantitative judgment mechanism that accumulates fault weights. By extracting fault names, fault types, fault codes, etc. from the component identification results, and traversing the entries in the preset fault list for accident vehicles, the corresponding fault information can be quickly matched. This unifies the fault judgment criteria and evaluation standards, overcomes the subjective bias of single fault judgment, effectively avoids misjudgment and omission of accident vehicles, and significantly improves the accuracy of accident vehicle identification. At the same time, high-speed read and write of the fault weight score of trucks and uniqueness verification of accident vehicle identifiers can be achieved through Redis caching. On the one hand, this greatly reduces the access pressure on the backend database and improves the overall processing efficiency of the system; on the other hand, by controlling the uniqueness of accident vehicle identifiers, the problem of duplicate marking is avoided, ensuring the accuracy and reliability of accident vehicle marking. Ultimately, a closed-loop detection and handling process for trucks is achieved, from fault identification and risk assessment to work order push.
[0080] Figure 3 This is a block diagram of a truck fault identification device based on TFDS, provided as an embodiment of this application.
[0081] like Figure 3 As shown, the truck fault identification device 300 based on TFDS includes: a data storage module 301, a fault matching module 302, a score calculation module 303, an accident determination module 304, and a result push module 305. These modules work together to achieve intelligent identification, quantitative determination, and closed-loop handling of truck faults. The specific implementation method is as follows: The data storage module 301 is used to acquire the component identification results of the truck output by TFDS intelligent recognition, and store the component identification results in the database and the cache respectively; The fault matching module 302 is used to perform fault matching on the component identification results based on a preset fault list of accident vehicles in order to obtain target fault information; The scoring calculation module 303 is used to calculate the fault weight score of the truck based on the target fault information using a fault weight accumulation method. The accident determination module 304 is used to mark the truck as an accident vehicle and write the corresponding accident vehicle identifier into the cache when the fault weight score reaches a preset weight threshold and there is no accident vehicle identifier corresponding to the truck in the cache. The result push module 305 is used to push the accident vehicle identifier to the operation and maintenance platform to generate a maintenance work order to be processed.
[0082] In one possible implementation, such as Figure 3As shown, the device may also include a fault handling module 306, which is used to respond to the pending maintenance work order issued by the operation and maintenance platform, perform maintenance operations on the truck corresponding to the accident vehicle identification, and feed back the maintenance results to the operation and maintenance platform.
[0083] Other relevant information can be found in the above method embodiments, which will not be repeated here.
[0084] In this embodiment, the truck fault identification device based on TFDS can separate the memory caching and persistent storage of high-frequency data through the data storage module, thereby improving data read and write efficiency; it can achieve standardized fault matching through the fault matching module, avoiding the subjectivity of manual judgment; through the weight accumulation and Redis identifier verification mechanism of the accident judgment module, it can ensure the accuracy of accident vehicle judgment and avoid duplicate marking; and through the result push module and fault processing module, it can achieve closed-loop fault handling, ultimately improving the efficiency and reliability of truck fault detection.
[0085] Figure 4 This is a block diagram of an electronic device provided in an embodiment of this application.
[0086] Reference Figure 4 The electronic device 400 includes a processor 401, a memory 402, and a computer program 4021 stored in the memory and executable on the processor. When executed, the processor implements the TFDS-based truck fault identification method of the aforementioned embodiments.
[0087] This application also provides a computer-readable storage medium that, when the instructions in the computer-readable storage medium are executed by the processor of an electronic device, enables the electronic device to perform the TFDS-based truck fault identification method of the foregoing embodiments.
[0088] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.
[0089] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0090] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, devices, and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0091] These computer program instructions may also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing terminal device to operate in a predictive manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0092] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0093] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.
[0094] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.
[0095] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.
Claims
1. A truck fault identification method based on TFDS, characterized in that, include: Obtain the truck component identification results output by the TFDS intelligent recognition system for truck operation faults, and store the component identification results in the database and cache respectively; Based on a preset list of faults in accident vehicles, the component identification results are matched to obtain target fault information; The fault weight score of the truck is calculated based on the target fault information using a fault weight accumulation method. When the fault weight score reaches a preset weight threshold and there is no accident vehicle identifier corresponding to the truck in the cache, the truck is marked as an accident vehicle and the corresponding accident vehicle identifier is written into the cache. The identified accident vehicle is pushed to the maintenance platform to generate a maintenance work order to be processed.
2. The method according to claim 1, characterized in that, The target fault information includes: the faulty component and its corresponding weight value; the calculation of the truck's fault weight score based on the target fault information using a fault weight accumulation method includes: Based on the faulty components of the truck and their corresponding weight values, the weight values of all matching faults of the truck are summed using the following weight accumulation formula to obtain the truck's fault weight score: S=ΣW i Where S is the fault weight score, W i Let Σ be the weight value corresponding to the i-th matched fault, and let Σ be the summation symbol; the weight value is set based on the severity of the fault of the truck component.
3. The method according to claim 1, characterized in that, The method further includes: Collect TFDS fault data of multiple historical accident vehicles, calculate the fault weight score of each accident vehicle, and obtain the fault weight score distribution of the accident vehicles. Collect false alarm fault data from multiple normally operating trucks caused by occasional factors, calculate the fault weight score for each normally operating truck, and form the fault weight score distribution of the normally operating trucks. Based on the fault weight score distribution of normal vehicles and the fault weight score distribution of accident vehicles, the optimal score is selected as the preset weight threshold according to at least one of receiver operating characteristic ROC curve, precision-recall curve or Youden index maximization.
4. The method according to claim 1, characterized in that, The step of storing the component identification results into a database and a cache respectively includes: Fault detection information is extracted from the component identification results. The fault detection information includes vehicle passage time, vehicle sequence, fault name, fault code, and component location information. The extracted fault detection information is stored in a database in a structured form; The extracted fault detection information is cached in a high-speed cache.
5. The method according to claim 4, characterized in that, The fault list of the accident vehicle includes: fault name, fault code, fault type, fault level, fault weight value, fault classification, and fault range; the fault matching based on the component identification results according to the preset fault list of the accident vehicle to obtain target fault information includes: Extract the fault name or fault code from the fault detection information in the component identification results of the truck as the information to be matched; The information to be matched is compared with the pre-stored fault entries in the preset fault list of accident vehicles, and the fault information belonging to the preset fault list of accident vehicles is selected as the target fault information.
6. The method according to claim 1, characterized in that, The method further includes: In response to the pending maintenance work order, maintenance operations are performed on the truck corresponding to the accident vehicle identifier, and the maintenance results are fed back to the operation and maintenance platform.
7. A truck fault identification device based on TFDS, characterized in that, include: The data storage module is used to acquire the component identification results of the truck output by TFDS intelligent recognition, and store the component identification results in the database and the cache respectively; The fault matching module is used to perform fault matching on the component identification results based on a preset fault list of accident vehicles in order to obtain target fault information; The scoring calculation module is used to calculate the fault weight score of the truck based on the target fault information using a fault weight accumulation method. The accident determination module is used to mark the truck as an accident vehicle and write the corresponding accident vehicle identifier into the cache when the fault weight score reaches a preset weight threshold and there is no accident vehicle identifier corresponding to the truck in the cache. The result push module is used to push the accident vehicle identifier to the operation and maintenance platform to generate a maintenance work order to be processed.
8. The apparatus according to claim 7, characterized in that, Also includes: Fault handling module The fault handling module is used to respond to the pending maintenance work order, perform maintenance operations on the truck corresponding to the accident vehicle identifier, and feed back the maintenance results to the operation and maintenance platform.
9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the truck fault identification method based on TFDS as described in any one of claims 1 to 6.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the truck fault identification method based on TFDS as described in any one of claims 1 to 6.