Intelligent data processing method and system of vehicle-mounted monitoring device

Abstract

The application relates to the rail transit technical field and provides an intelligent data processing method and system of a vehicle-mounted monitoring device, the method comprising the following steps: receiving original data frames from a plurality of heterogeneous monitoring devices, identifying the communication protocol types and corresponding device types; loading protocol analysis plug-ins based on the identification results and extracting field name and corresponding value pairs; performing standardization processing on the extracted field name and value pairs, encapsulating standardized fields and meta information into preset structured data packets and transmitting the structured data packets to an intelligent decision module; the intelligent decision module effectively filters out redundant fields through a priority and abnormality joint screening mechanism, screens out high-value abnormal fields for intelligent inference, and finally outputs vehicle state judgment results and control suggestions to a master control unit for subsequent control. The application realizes automatic analysis, semantic unification and intelligent inference judgment of heterogeneous vehicle-mounted device data, and improves the operation intelligence and safety of a rail transit system.

Description

Technical Field

[0001] The present application relates to the field of rail transportation technology, and in particular to an intelligent data processing method and system for vehicle-mounted monitoring equipment. Background Art

[0002] In rail transit systems, vehicles are typically equipped with a large number of heterogeneous onboard monitoring devices to monitor operational status, provide fault warnings, and ensure safety. These devices include speed sensors, vibration sensors, temperature and humidity modules, brake system status sensors, current and voltage sensors, and door status switches. These monitoring devices come from different vendors and utilize different communication protocols (such as MVB, CAN, and Ethernet), with varying data formats and physical meanings. This leads to the following technical challenges:

[0003] 1. Serious heterogeneity in data protocols and structures, and a lack of a unified interface layer: Different devices have different data frame structures, field naming, and unit formats, requiring separate adaptation, resulting in high system integration complexity;

[0004] 2. Inconsistent field semantics hinder intelligent analysis and reasoning: Data fields reported by devices are often low-semantic field names (such as TMP, VAL, and FIELD001). The lack of semantic layer standards hinders unified recognition and learning by AI modules.

[0005] 3. Fixed rule parsing lacks scalability: Traditional methods of parsing fields through hard-coding or static templates are difficult to adapt to the needs of device expansion, upgrades, or vehicle migration.

[0006] 4. The reasoning process lacks a mechanism for selecting key fields, resulting in severe redundancy and interference. Only a subset of the massive number of fields is critical for identifying operational status, making it difficult for existing systems to dynamically filter and adjust field weights.

[0007] Therefore, there is an urgent need for an on-board monitoring data processing solution with protocol adaptability, field semantic unification, key field recognition and intelligent reasoning capabilities to achieve intelligent operation of rail vehicles and compatible deployment of multiple vehicle series. Summary of the Invention

[0008] The purpose of this application is to provide an intelligent data processing method and system for vehicle-mounted monitoring equipment to solve the problems described in the background technology section of this application.

[0009] To achieve the above-mentioned purpose, the first aspect of this application provides the following technical solutions:

[0010] The first aspect of the present application provides an intelligent data processing method for a vehicle-mounted monitoring device, comprising:

[0011] Obtaining a raw data frame transmitted by a heterogeneous monitoring device, identifying a communication protocol of the raw data frame according to a data transmission channel corresponding to the raw data frame, obtaining a communication protocol type and a corresponding device type;

[0012] Loading a corresponding protocol analysis plug-in according to the communication protocol type and the device type, extracting fields of the raw data frame through the protocol analysis plug-in, obtaining a field name and a corresponding value pair of the raw data frame, and the protocol analysis plug-in encapsulates a communication format template, a field extraction rule and a semantic mapping rule matched with the communication protocol type and the device type;

[0013] Standardizing the field name and the corresponding value pair respectively, encapsulating the raw data frame into a data packet in a preset data format according to the standardized field name and the corresponding value pair, and transmitting the data packet to an intelligent decision layer;

[0014] The intelligent decision layer receives a plurality of data packets sent by a plurality of heterogeneous monitoring devices, performs priority filtering and abnormality filtering on the fields of each data packet, inputs a first field set with a priority greater than a preset value and marked as abnormal to an intelligent inference module, and the intelligent inference module outputs a first intelligent inference result to a master control unit, wherein the first intelligent inference result includes a vehicle state determination result and a control suggestion.

[0015] Further, the loading of the corresponding protocol analysis plug-in according to the communication protocol type and the device type comprises:

[0016] Splitting the raw data frame based on the data transmission channel, matching the data structure based on the data structure splitting result and each communication protocol template, and determining the communication protocol type corresponding to the raw data frame according to the matching result;

[0017] Extracting a device address or a device identifier from the raw data frame to identify the device type corresponding to the raw data frame;

[0018] Matching the protocol analysis plug-in based on the combination of the communication protocol type, the device type and the data transmission channel, and loading the protocol analysis plug-in to extract fields of the raw data frame.

[0019] Further, the standardization of the field name and the corresponding value pair respectively comprises:

[0020] Based on the scenario information and context information corresponding to the field name, a pre-trained semantic recognition model is called to perform semantic recognition on the field name to obtain a standard semantic field, wherein the scenario information includes the device type, data channel type, and communication protocol type corresponding to the field name, and the context information is other fields in the same device or the same data packet as the field name;

[0021] According to the standard semantic field, the numerical unit is identified and the numerical ratio is scaled for the numerical value pair to obtain a standardized numerical value pair;

[0022] Based on the standard semantic field and the standardized value pair, the original data frame is encapsulated into a data packet in the preset data format.

[0023] Furthermore, the performing priority screening and exception screening on the fields of each of the data packets respectively includes:

[0024] The intelligent decision layer reads the priority tag of each field in the data packet, and filters out the fields in the data packet whose priority is greater than the preset value, where the preset value is determined according to the current resource status of the system, wherein before encapsulating the original data frame into the data packet in the preset data format, the priority tag of each field in the original data frame is added as metadata of each field to the encapsulated data packet;

[0025] Performing sliding window detection on each field in the data packet, identifying abnormal fields based on the detection results and marking them as abnormal, filtering fields whose priorities are greater than the preset value and marked as abnormal as the first field set, performing field fusion on associated fields in the first field set to generate corresponding virtual fields, where the virtual fields are physical quantities formed by fusing two or more fields in the first field set by field multiplication, field combination, or functional relationship;

[0026] The first field set and the virtual field are respectively input into the intelligent reasoning module for intelligent reasoning to obtain the first intelligent reasoning result.

[0027] Furthermore, the performing priority screening and exception screening on the fields of each of the data packets respectively further includes:

[0028] The intelligent decision layer reads the priority tag of each field in the data packet, and filters the fields whose priorities are less than or equal to the preset value and are marked as abnormal as the second field set;

[0029] The second field set, the first field set and the virtual field are input into the intelligent inference module for inference respectively to obtain a second intelligent inference result, and the second intelligent inference result includes a vehicle state determination result and a control suggestion.

[0030] Further, a field with a priority greater than the preset value but not marked as an exception is taken as an auxiliary fusion field, and field fusion is performed on an associated field in the first field set in a field multiplication, field combination or function relationship manner to generate the virtual field.

[0031] Further, the intelligent inference module outputs a first intelligent inference result to a master control unit, and further includes:

[0032] The intelligent inference module extracts an involvement degree index of each input field in the first field set according to the first intelligent inference result to generate a contribution score of each input field in the first field set.

[0033] The priority label of the corresponding field is updated based on the contribution score of each input field in the first field set, and the updated priority label and the corresponding field are stored in a field registration table, and the field registration table is used to encapsulate the original data frame into a data packet in the preset data format.

[0034] Further, the intelligent inference module extracts an involvement degree index of each input field in the first field set according to the first intelligent inference result, and further includes:

[0035] The intelligent inference module extracts the involvement degree index of each input field in the first field set from one or more involvement degree indexes including attention weight, marginal influence, model gradient and field participation path frequency to obtain each involvement degree index and a corresponding numerical value of the input field.

[0036] The contribution score of the input field is obtained by normalizing and weighting the each involvement degree index and the corresponding numerical value of the input field.

[0037] The priority label of the input field is adjusted and updated based on the contribution score.

[0038] The second aspect of the present application provides an intelligent data processing system of a vehicle-mounted monitoring device, and the method of the first aspect of the present application is implemented based on the system of the second aspect of the present application. The system includes:

[0039] The general interface layer is used to obtain an original data frame transmitted by a heterogeneous monitoring device, identify a communication protocol of the original data frame according to a data transmission channel corresponding to the original data frame, obtain a communication protocol type and a corresponding device type, load a corresponding protocol analysis plug-in according to the communication protocol type and the device type, perform field extraction on the original data frame through the protocol analysis plug-in, obtain a field name and a corresponding value pair of the original data frame, the protocol analysis plug-in encapsulates a communication format template, a field extraction rule and a semantic mapping rule matched with the communication protocol type and the device type, and standardize the field name and the corresponding value pair respectively, encapsulate the original data frame into a data packet in a preset data format according to the standardized field name and the corresponding value pair, and transmit the data packet to an intelligent decision layer.

[0040] The intelligent decision layer is used to receive a plurality of data packets transmitted by a plurality of heterogeneous monitoring devices, perform priority filtering and abnormality filtering on fields of each data packet respectively, input a first field set with a priority greater than a preset value and marked as abnormal to an intelligent inference module, and output a first intelligent inference result to a master control unit, wherein the first intelligent inference result includes a vehicle state judgment result and a control suggestion.

[0041] Further, the general interface layer further includes:

[0042] The protocol identification module is used to split a data structure of the original data frame based on the data transmission channel, match the data structure splitting result with each communication protocol template, and determine the communication protocol type corresponding to the original data frame according to the matching result.

[0043] The device identification module is used to extract a device address or a device identifier from the original data frame to identify the device type corresponding to the original data frame.

[0044] The protocol analysis module is used to match the protocol analysis plug-in based on the combination of the communication protocol type, the device type and the data transmission channel, and load the protocol analysis plug-in to perform field extraction on the original data frame.

[0045] The intelligent data processing method of the vehicle-mounted monitoring device provided by the application can at least achieve the following technical effects:

[0046] This application realizes the unified parsing and standardized packaging of heterogeneous on-board monitoring equipment data through communication protocol identification and plug-in parsing mechanism; effectively screens out redundant fields through priority and exception joint screening mechanism, ensuring the accuracy and criticality of intelligent reasoning input; the status and recommended results output by the reasoning module can be directly used for the main control strategy linkage, improving the intelligence level of rail vehicle operation status judgment and safety control. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the specific implementation methods or the description of the prior art. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0048] Figure 1 A schematic flow chart of an intelligent data processing method for a vehicle-mounted monitoring device provided in an embodiment of the present application;

[0049] Figure 2 A schematic diagram of the architecture of an intelligent data processing system for a vehicle-mounted monitoring device provided in an embodiment of the present application;

[0050] Figure 3 A schematic diagram of a computer device provided in an embodiment of the present application.

[0051] Reference numerals: 200 , intelligent data processing system of vehicle-mounted monitoring equipment; 201 , general interface layer; 202 , intelligent decision-making layer; 203 , main control unit; 301 , memory; 302 , processor. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions of this application in conjunction with the embodiments. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0053] The embodiment of the present application provides an intelligent data processing method for on-board monitoring equipment, realizes unified parsing and standardized packaging of heterogeneous on-board monitoring equipment data through communication protocol identification and plug-in parsing mechanism; effectively screens out redundant fields through priority and exception joint screening mechanism, ensuring the accuracy and criticality of intelligent reasoning input; the status and recommended results output by the reasoning module can be directly used for the main control strategy linkage, improving the intelligent level of rail vehicle operation status judgment and safety control. Figure 1The figure shows a flow chart of an intelligent data processing method for vehicle-mounted monitoring equipment according to this embodiment. Figure 2 The intelligent data processing system 200 of the vehicle-mounted monitoring device shown in FIG. 1 includes the following steps:

[0054] Step S100: Acquire original data frames transmitted by heterogeneous monitoring devices, identify the communication protocol of the original data frames according to the data transmission channel corresponding to the original data frames, and obtain the communication protocol type and the corresponding device type;

[0055] Specifically, the method provided in this embodiment is applied to data processing for heterogeneous on-board monitoring devices. The system receives raw data frames uploaded by multiple heterogeneous on-board monitoring devices (such as speed sensors, temperature and humidity modules, and door switches). Each monitoring device is connected to the main control unit 203 via a different communication channel (such as MVB, CAN, RS485, Ethernet, etc.). The main control unit 203 dispatches the received raw data frames and forwards them to the system's universal adaptation layer. The universal adaptation layer first calls the protocol identification module to identify the data channel type of the received raw data frame. Based on the data structure corresponding to the data channel type, the raw data frame is decomposed into its data structure. The decomposed protocol feature fields or data frame header information, frame length, address field, and other structural information are matched against a preset communication protocol template library. If the match is successful, the communication protocol used by the raw data frame (such as CAN_J1939, MVB-FTE, etc.) can be determined. If the match fails, the raw data frame can be marked as an unknown protocol; the operation and maintenance personnel can be requested to manually upload a template; or the AI ​​model can be triggered to perform self-learning and field extraction.

[0056] Secondly, the universal interface layer 201 calls the device identification module to extract the device address or identifier (such as the ID field of CAN, the address segment of MVB) in the original data frame according to the protocol format corresponding to the determined communication protocol type, and matches it with the preset device template library to identify the type of monitoring device corresponding to the original data frame (such as speed sensor, vibration module, temperature module, etc.).

[0057] Step S200: Loading a corresponding protocol parsing plug-in according to the communication protocol type and the device type, performing field extraction on the original data frame by the protocol parsing plug-in to obtain field name and corresponding value pairs of the original data frame, wherein the protocol parsing plug-in encapsulates a communication format template, field extraction rules, and semantic mapping rules that match the communication protocol type and the device type;

[0058] Furthermore, the step S200 further includes:

[0059] Step S210: splitting the original data frame into a data structure based on the data transmission channel, matching the data structure with each preset communication protocol template according to the data structure splitting result, and determining the communication protocol type corresponding to the original data frame according to the matching result;

[0060] Step S220: extracting a device address or a device identifier from the original data frame to identify a device type corresponding to the original data frame;

[0061] Step S230: Match the protocol parsing plug-in based on the combination of the communication protocol type, the device type, and the data transmission channel, and load the protocol parsing plug-in to perform field extraction on the original data frame.

[0062] Specifically, based on the determined combination of the original data frame and its corresponding communication protocol type and device type, the universal interface layer 201 retrieves and loads a protocol parsing plug-in matching this combination from the system's built-in protocol parsing plug-in library. Each protocol parsing plug-in encapsulates the communication format template, field extraction rules, and semantic mapping rules that match the corresponding communication protocol type and device type. The protocol parsing plug-in is hot-swappable and dynamically updateable. After matching and loading the protocol parsing plug-in corresponding to the original data frame, the protocol parsing plug-in performs a structured parsing of the original data frame, extracting valid fields, deciphering data values, and performing necessary encoding conversions. It then outputs the original protocol frame's field name and corresponding value pairs (e.g., "T_val": 325, "unit_code": 1). As described above, this embodiment dynamically loads protocol parsing plug-ins using the triplet of channel identification, communication protocol type, and device type, achieving unified adaptation for heterogeneous devices. The plug-in structure supports runtime loading and replacement, enabling the same system to be deployed and migrated across different vehicle series.

[0063] Step S300: standardize the field names and their corresponding value pairs respectively, encapsulate the original data frame into a data packet in a preset data format according to the standardized field names and their corresponding value pairs, and transmit the data packet to the intelligent decision layer;

[0064] Furthermore, in step S300, the normalization of the field names and their corresponding value pairs further includes:

[0065] Step S310: Based on the scenario information and context information corresponding to the field name, a pre-trained semantic recognition model is called to perform semantic recognition on the field name to obtain a standard semantic field, wherein the scenario information includes the device type, data channel type, and communication protocol type corresponding to the field name, and the context information is other fields in the same device or the same data packet as the field name;

[0066] Step S320: Identify the numerical unit and perform numerical scaling on the numerical value pair according to the standard semantic field to obtain a standardized numerical value pair;

[0067] Step S330: Encapsulate the original data frame into a data packet in the preset data format based on the standard semantic field and the standardized value pair.

[0068] Specifically, in step S300, the system performs standardization processing on the field names and corresponding value pairs of the original protocol frame obtained by the protocol parsing plug-in, including:

[0069] 1. For field names, the pre-trained semantic recognition model and rule mapping table are used to uniformly name the field names based on the field name, field length, data range and unit information, device type, data channel type, and communication protocol type, as well as the other fields in the data packet or device containing the field name. When ambiguity exists, the field association relationships are summarized based on contextual information to assist in identification, and the standard semantic field corresponding to the field name is obtained. For example, the semantic recognition model can identify "T_val" as "Temperature" and "unit_code=1" as "℃". Different from the traditional method of using field names for regular matching, this embodiment uses a semantic recognition model for collaborative recognition, which can handle fuzzy, ambiguous, and non-standard field names.

[0070] 2. For the value pairs corresponding to the field names, perform unit conversion, scaling, and physical value normalization operations (such as 16-bit unsigned / 10, converted to actual temperature) on the value pairs to obtain standard semantic fields and standardized value pairs.

[0071] Furthermore, in this embodiment, the preset format is preferably the IR (IR-Frame) format. The IR-Frame is a structured data frame used to uniformly describe signals collected by different devices and protocols. Based on an IR format generator, this embodiment first assigns a priority tag to each standard semantic field based on a preset rule base or a dynamic evaluation mechanism. The standardized standard semantic fields and standardized value pairs are then encapsulated into a unified intermediate format and corresponding metadata is added. The IR format generator's input includes standard semantic fields, standard standardized value pairs, device type (device ID), communication protocol type, and timestamp. The IR format generator outputs data packets in the IR-Frame format. Finally, the general interface layer 201 sends each generated data packet to the intelligent decision layer 202. By encapsulating standard fields into data packets in a unified structured preset data format (e.g., IR frames) and carrying metadata such as device identification, timestamp, and communication protocol type, this embodiment enables module decoupling in the system, providing a standard interface for invoking multiple models and multiple decision logics, improving project maintainability and the foundation for intelligent scheduling.

[0072] In step S400, the intelligent decision-making layer 202 receives the plurality of data packets sent by the plurality of heterogeneous monitoring devices, performs priority screening and exception screening on the fields of each data packet, and inputs the first set of fields whose priority is greater than a preset value and is marked as abnormal into the intelligent reasoning module, and the intelligent reasoning module outputs the first intelligent reasoning result to the main control unit 203, wherein the first intelligent reasoning result includes a vehicle status determination result and a control suggestion.

[0073] Specifically, in this embodiment, the intelligent decision layer 202 integrates an AI intelligent decision module for performing protocol conversion, signal scheduling, intelligent reasoning, and function mapping on the data encapsulated in the data packet. The AI ​​intelligent decision module receives each data packet sent by the general interface layer 201, reads each field and its priority tag, calls the signal scheduling module to classify and prioritize the fields, and simultaneously calls the exception identification module to identify the fields for exceptions, obtaining priority screening and exception identification results respectively. After the screening is completed, the fields with a priority greater than the preset value and identified as exceptions are added to the first field set and enter the main path of "fusion" + "reasoning"; the fields with a priority less than or equal to the preset value and identified as exceptions are added to the second field set as auxiliary fields for auxiliary reasoning; the fields with a priority greater than the preset value but not identified as exceptions are used for auxiliary fusion with the fields in the first field set, but are not input into the intelligent reasoning module for reasoning. This embodiment introduces a priority scoring and anomaly detection mechanism to support the screening of "high priority and abnormal" key field sets from massive fields, improving reasoning efficiency and accuracy.

[0074] Furthermore, the intelligent reasoning module outputs a first intelligent reasoning result, including vehicle status determination results and control recommendations. Vehicle status, such as stable operation, door abnormality, and brake warning, and control recommendations, such as triggering an alarm, recommending speed reduction, and maintaining monitoring, are examples of control logic. The intelligent decision-making layer 202 encapsulates the first intelligent reasoning result into standard logical fields, forming a unified logical data structure for unified reading and access by the main control unit 203 (main control CPU). For example, the main control unit 203's access operations for the first intelligent reasoning result include:

[0075] 1. If the current vehicle speed is judged to be inconsistent with the stability risk, a "speed limit" warning will be automatically issued;

[0076] 2. If the vibration risk increases, the notification web service will be highlighted on the human-machine interface;

[0077] 3. If abnormalities occur multiple times, the maintenance log will be recorded and uploaded to the ground maintenance system. The Web service port also supports operation and maintenance personnel to remotely view the current equipment operation status, support remote diagnosis, monitoring equipment restart and other operations.

[0078] Furthermore, in step S400, the priority screening and exception screening are performed on the fields of each data packet, including:

[0079] Step S410: The intelligent decision layer 202 reads the priority tag of each field in the data packet, and filters out fields whose priority is greater than a preset value, where the preset value is determined according to the current resource status of the system. Before encapsulating the original data frame into a data packet in the preset data format, the priority tag of each field in the original data frame is added as metadata of each field to the encapsulated data packet;

[0080] Step S420: Perform sliding window detection on each field in the data packet, identify abnormal fields based on the detection results and mark them as abnormal, filter the fields whose priorities are greater than the preset value and are marked as abnormal to form the first field set, perform field fusion on the associated fields in the first field set, and generate corresponding virtual fields, where the virtual fields are physical quantities formed by fusion of two or more fields in the first field set by field multiplication, field combination, or functional relationship;

[0081] Step S430: Input the first field set and the virtual field into the intelligent reasoning module respectively for intelligent reasoning to obtain the first intelligent reasoning result.

[0082] Specifically, after the general interface layer 201 completes field parsing and semantic mapping, it assigns a priority tag to each field based on a preset rule base or dynamic evaluation mechanism, and then encapsulates them into an IR-structured data packet for transmission. The intelligent decision-making layer 202 calls the signal scheduling module to identify the priority tags of the fields in the data packet, performs priority determination, and classifies them based on preset values. Simultaneously, it calls the anomaly identification module to identify anomalies in each field using a sliding window detection method. Sliding window detection is used to analyze the changing trends of fields over time. After the signal scheduling module completes field standardization and screening, the system identifies multiple key fields with high priority and marked as anomalies, forming a first field set. For example, in the monitoring data from a certain carriage, the fields "Vibration_X" and "Speed" are both marked as anomalies, with priority scores of 0.92 and 0.85, respectively.

[0083] Furthermore, the signal scheduling module queries the system's pre-set fusion rule library and performs "physical logical combination" on multiple related fields in the first field set according to pre-defined fusion rules. Fusion rules include field multiplication, field combination (concatenation / weighted sum), functional relationships, logical rules (conditional reasoning), statistical model fusion (covariance / trend), and small neural network models (data-driven modeling). Each fusion rule is a pre-set physical meaning field fusion rule. Each fusion rule includes a description of a field combination pattern, a list of input fields, a calculation method, and the semantics of the output field. The field fusion module matches the fusion rule library and performs field fusion on each related field in the first field set according to the fusion calculation method (function / rule / model) defined in the fusion rule, generating a new virtual field. This virtual field is then added to the current data packet structure as a new field eligible for inference and pushed to the intelligent inference module. It is used as input along with other abnormal fields to trigger a comprehensive judgment of the vehicle's operating status. Virtual fields are high-order indicators based on several basic physical quantities, which are used to provide stronger decision-making reference signals and indicator dimensions for the intelligent reasoning module. For example, the first field set includes two related fields, "vibration" and "speed", and then the virtual field "impact energy" can be generated to amplify the combined effect of multiple weak abnormal signals, so as to jointly perform subsequent reasoning.

[0084] Furthermore, the first field set and the virtual field are each input into the intelligent reasoning module for intelligent reasoning, obtaining the first intelligent reasoning result. The intelligent reasoning module uses deep learning to determine the vehicle status based on each high-priority and abnormal field and generates control recommendations. The intelligent reasoning module includes a classification model and a regression model. The classification model is used to determine vehicle status, such as whether the vehicle is abnormal or at risk of derailment. The regression model is used to generate anomaly severity scores, risk levels, and recommended actions.

[0085] Furthermore, step S430 further includes:

[0086] Step S431: The intelligent decision layer 202 reads the priority tag of each field in the data packet, and filters the abnormal fields whose priorities are less than or equal to the preset value and are marked as abnormal as a second field set;

[0087] Step S432: Input the second field set, the first field set and the virtual field into the intelligent reasoning module for reasoning respectively to obtain a second intelligent reasoning result, which includes a vehicle state determination result and a control suggestion.

[0088] Specifically, fields whose priorities are less than or equal to a preset value and are marked as abnormal are filtered into a second field set, and input into the intelligent reasoning module together with the first field set and the virtual field for further reasoning.

[0089] Furthermore, step S420 further includes:

[0090] Step S421: Use fields whose priorities are greater than the preset value but are not marked as abnormal as auxiliary fusion fields, and assist the associated fields in the first field set in field fusion by field multiplication, field combination or functional relationship to generate the virtual field.

[0091] Specifically, high-priority fields that are not marked as abnormal are used to merge with the fields in the first field set to assist in generating virtual fields and participate in trend judgment. However, these high-priority fields that are not marked as abnormal are not used for input into the intelligent reasoning module.

[0092] Furthermore, in step S400, the intelligent reasoning module outputs the first intelligent reasoning result to the main control unit 203, further comprising:

[0093] The intelligent reasoning module extracts the participation index of each input field in the first field set based on the first intelligent reasoning result, and generates a contribution score for each input field in the first field set; updates the priority label of the corresponding field based on the contribution score corresponding to each input field in the first field set, and stores the updated priority label and its corresponding field in the field registration table, which is used to encapsulate the original data frame into a data packet in the preset data format.

[0094] Specifically, after completing state recognition or control suggestion output, the intelligent reasoning module records the input field set involved in this reasoning, such as F = {f1, f2, ..., f n}. Based on one or more participation indicators including attention weight (Attn), marginal influence (SHAP), model gradient (applicable to neural networks) and field participation path frequency (applicable to tree models), the participation index of each input field in the first field set is extracted to obtain multiple participation indicators and corresponding values ​​corresponding to each input field. Afterwards, the multiple participation indicators and corresponding values ​​are normalized and weighted fused to generate an input field contribution score (contribution_score∈[0,1]) for priority feedback. Finally, based on the contribution score, the priority label of the corresponding input field in the field registry is adjusted and updated, using a sliding average or replacement update method. The updated priority participates in the field scheduling of the next round of signal scheduling module.

[0095] This embodiment uses an intelligent reasoning module to extract field participation indicators (such as attention weight and contribution score), feeds them back to the field priority management module, dynamically updates the field weight, and realizes closed-loop feedback control of data scheduling and AI model collaborative optimization.

[0096] This embodiment also provides an intelligent data processing system for vehicle-mounted monitoring equipment. Since the principle of solving problems by the intelligent data processing system for vehicle-mounted monitoring equipment is similar to that of the intelligent data processing method for vehicle-mounted monitoring equipment, the implementation of the intelligent data processing system for vehicle-mounted monitoring equipment can refer to the implementation of the intelligent data processing system for vehicle-mounted monitoring equipment, and the repeated parts will not be repeated. Figure 2 As shown, an intelligent data processing system 200 for vehicle-mounted monitoring equipment specifically includes:

[0097] The general interface layer 201 is used to obtain raw data frames transmitted by heterogeneous monitoring devices, identify the communication protocol of the raw data frames based on the data transmission channel corresponding to the raw data frames, obtain the communication protocol type and the corresponding device type, load the corresponding protocol parsing plug-in based on the communication protocol type and the device type, perform field extraction on the raw data frames through the protocol parsing plug-in, and obtain the field name and corresponding value pairs of the raw data frames. The protocol parsing plug-in encapsulates the communication format template, field extraction rules, and semantic mapping rules that match the communication protocol type and the device type; standardize the field names and their corresponding value pairs respectively, and encapsulate the raw data frames into data packets in a preset data format based on the standardized field names and corresponding value pairs, and transmit them to the intelligent decision layer 202;

[0098] The intelligent decision-making layer 202 is used to receive multiple data packets sent by multiple heterogeneous monitoring devices, perform priority screening and exception screening on the fields of each data packet, and input the first set of fields whose priority is greater than a preset value and is marked as abnormal into the intelligent reasoning module. The intelligent reasoning module outputs the first intelligent reasoning result to the main control unit 203, wherein the first intelligent reasoning result includes the vehicle status judgment result and control suggestions.

[0099] The main control unit 203 is used to receive the intelligent reasoning result and execute the corresponding control strategy or alarm processing.

[0100] Furthermore, the general interface layer 201 also includes:

[0101] A channel identification module is used to receive raw data frames sent by multiple heterogeneous monitoring devices, and identify their communication protocol type and corresponding device type based on the transmission channel of the raw data frame; a device identification module is used to extract the device address or device identifier from the raw data frame to identify the device type corresponding to the raw data frame; a protocol identification module is used to split the data structure of the raw data frame based on the data transmission channel, match the data structure with the preset communication protocol templates according to the data structure splitting result, and determine the communication protocol type corresponding to the raw data frame according to the matching result; a protocol parsing module is used to match the protocol parsing plug-in based on the combination of the communication protocol type, the device type and the data transmission channel, and load the protocol parsing plug-in to extract fields from the raw data frame; a data encapsulation module is used to encapsulate the standardized fields and metadata into a data packet in a preset format, and send it to the intelligent decision-making module.

[0102] Furthermore, the intelligent decision-making layer 202 also includes:

[0103] a signal scheduling module and an anomaly identification module, configured to respectively perform priority screening and anomaly screening on fields contained in a plurality of received data packets, and construct a first field set, the first field set including fields having a priority higher than a preset value and marked as anomalies;

[0104] The intelligent reasoning module is used to perform intelligent state reasoning based on the first field set, output intelligent reasoning results including vehicle state judgment and control suggestions, and transmit them to the main control unit 203.

[0105] In this embodiment, a computer device is also provided, such as Figure 3 As shown, it includes a memory 301, a processor 302 and a computer program stored in the memory 301 and capable of running on the processor 302. When the processor 302 executes the computer program, any one of the above-mentioned intelligent data processing methods for vehicle-mounted monitoring equipment is implemented.

[0106] Specifically, the computer device may be a computer terminal, a server or a similar computing device.

[0107] In this embodiment, a computer-readable storage medium is further provided, wherein the computer-readable storage medium stores a computer program for executing any one of the above-mentioned intelligent data processing methods for vehicle-mounted monitoring equipment.

[0108] Specifically, computer-readable storage media include permanent and non-permanent, removable and non-removable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer-readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable storage media does not include transitory media such as modulated data signals and carrier waves.

[0109] The embodiments of the present invention achieve the following technical effects:

[0110] 1. This application dynamically loads protocol parsing plug-ins through the channel identification + communication protocol type + device type triple, achieving unified adaptation for heterogeneous devices. The plug-in structure supports runtime loading and replacement, allowing the same AI host to be deployed and migrated between different car series.

[0111] 2. This application uses a semantic recognition module and a field unit normalization mechanism to standardize the original field name (such as T_val) and the original unit code, and construct an intermediate semantic field (IR field) that can be used for subsequent intelligent decision analysis;

[0112] 3. This application introduces a priority scoring and anomaly detection mechanism to support the screening of "high priority and abnormal" key field sets from massive fields, improving reasoning efficiency and accuracy;

[0113] 4. This application fuses abnormal fields with high-value fields based on physical relationships to generate virtual fields with clear physical meanings, enhancing the system's ability to perceive risk situations;

[0114] 5. The intelligent reasoning module of this application extracts field participation indicators (such as attention weight and contribution score), feeds them back to the field priority management module, dynamically updates the field weight, and realizes closed-loop feedback control for coordinated optimization of data scheduling and AI models;

[0115] 6. This application significantly improves the intelligent processing capability, system versatility and data value extraction efficiency of on-board monitoring data of rail transit vehicles, and has high technological advancement and engineering practical value.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent data processing method for vehicle-mounted monitoring equipment, characterized in that: include: Obtaining original data frames transmitted by heterogeneous monitoring devices, identifying the communication protocol of the original data frames according to the data transmission channel corresponding to the original data frames, and obtaining the communication protocol type and the corresponding device type; Loading a corresponding protocol parsing plug-in according to the communication protocol type and the device type, performing field extraction on the original data frame through the protocol parsing plug-in to obtain field name and corresponding value pairs of the original data frame, wherein the protocol parsing plug-in encapsulates a communication format template, field extraction rules, and semantic mapping rules that match the communication protocol type and the device type; The field names and their corresponding value pairs are respectively standardized, and the original data frame is encapsulated into a data packet in a preset data format according to the standardized field names and the corresponding value pairs, and the data packet is transmitted to the intelligent decision layer, wherein the priority tag corresponding to each field name is encapsulated into the data packet as metadata; The intelligent decision layer receives the plurality of data packets sent by the plurality of heterogeneous monitoring devices, performs priority screening and exception screening on the fields of each of the data packets, and screens the fields whose priority is greater than a preset value and is marked as abnormal into a first field set, wherein the preset value is determined according to the current resource status of the system, and uses the fields whose priority is greater than the preset value but is not marked as abnormal as auxiliary fusion fields to assist the associated fields in the first field set in field fusion, and generates corresponding virtual fields, wherein the virtual fields are physical quantities, and the virtual fields are formed by fusion of two or more fields in the first field set by field multiplication, field combination or functional relationship, and the auxiliary fusion fields assist the associated fields in field fusion by field multiplication, field combination or functional relationship; The first field set and the virtual field are respectively input into an intelligent reasoning module, and the intelligent reasoning module outputs a first intelligent reasoning result to a main control unit, wherein the first intelligent reasoning result includes a vehicle state determination result and a control suggestion.

2. The intelligent data processing method for vehicle-mounted monitoring equipment according to claim 1, characterized in that: The loading of the corresponding protocol parsing plug-in according to the communication protocol type and the device type includes: Performing data structure splitting on the original data frame based on the data transmission channel, performing data structure matching with preset communication protocol templates according to the data structure splitting results, and determining the communication protocol type corresponding to the original data frame according to the matching results; extracting a device address or a device identifier from the original data frame to identify a device type corresponding to the original data frame; The protocol parsing plug-in is matched based on a combination of the communication protocol type, the device type and the data transmission channel, and the protocol parsing plug-in is loaded to perform field extraction on the original data frame.

3. The intelligent data processing method for vehicle-mounted monitoring equipment according to claim 1, characterized in that: The standardization of the field names and their corresponding value pairs includes: Based on the scenario information and context information corresponding to the field name, a pre-trained semantic recognition model is called to perform semantic recognition on the field name to obtain a standard semantic field, wherein the scenario information includes the device type, data channel type, and communication protocol type corresponding to the field name, and the context information is other fields in the same device or the same data packet as the field name; According to the standard semantic field, the numerical unit of the numerical pair is identified and the numerical ratio is scaled to obtain a standardized numerical pair; based on the standard semantic field and the standardized numerical pair, the original data frame is encapsulated into a data packet in the preset data format.

4. The intelligent data processing method for vehicle-mounted monitoring equipment according to claim 1, characterized in that: The performing priority screening and exception screening on the fields of each of the data packets respectively includes: The intelligent decision layer reads the priority tag of each field in the data packet, and filters out the fields in the data packet whose priority is greater than the preset value; Sliding window detection is performed on the fields in the data packet respectively, abnormal fields are identified and marked as abnormal based on the detection results, and fields with a priority greater than the preset value and marked as abnormal are filtered into the first field set.

5. The intelligent data processing method for vehicle-mounted monitoring equipment according to claim 4, characterized in that: The step of performing priority screening and exception screening on the fields of each of the data packets further includes: The intelligent decision layer reads the priority tag of each field in the data packet, and filters the fields whose priorities are less than or equal to the preset value and are marked as abnormal as the second field set; The second field set, the first field set and the virtual field are respectively input into the intelligent reasoning module for reasoning to obtain a second intelligent reasoning result, which includes a vehicle state determination result and a control suggestion.

6. The intelligent data processing method for vehicle-mounted monitoring equipment according to claim 4, characterized in that: The intelligent reasoning module outputs the first intelligent reasoning result to the main control unit, and further includes: The intelligent reasoning module extracts a participation index of each input field in the first field set according to the first intelligent reasoning result, and generates a contribution score for each input field in the first field set; Based on the contribution score corresponding to each input field in the first field set, the priority label of the corresponding field is updated, and the updated priority label and its corresponding field are stored in a field registry, and the field registry is used to encapsulate the original data frame into a data packet in the preset data format.

7. The intelligent data processing method for vehicle-mounted monitoring equipment according to claim 6, characterized in that: The intelligent reasoning module extracts a participation index of each input field in the first field set according to the first intelligent reasoning result, and further includes: The intelligent reasoning module extracts participation indexes of each input field in the first field set from one or more participation indexes of attention weight, marginal influence, model gradient, and field participation path frequency, and obtains each participation index and corresponding value corresponding to the input field; Normalizing and weighted fusion of the participation level indicators and corresponding values ​​corresponding to the input fields to obtain a contribution score of the input fields; The priority label of the input field is adjusted and updated based on the contribution score.

8. An intelligent data processing system for vehicle-mounted monitoring equipment, characterized in that: The method according to any one of claims 1 to 7 is implemented based on the system, which comprises: A universal interface layer is used to obtain raw data frames transmitted by heterogeneous monitoring devices, identify the communication protocol of the raw data frames according to the data transmission channel corresponding to the raw data frames, obtain the communication protocol type and the corresponding device type, load the corresponding protocol parsing plug-in according to the communication protocol type and the device type, perform field extraction on the raw data frames through the protocol parsing plug-in, and obtain the field name and corresponding value pair of the raw data frames. The protocol parsing plug-in encapsulates the communication format template, field extraction rules and semantic mapping rules that match the communication protocol type and the device type; standardize the field names and their corresponding value pairs respectively, and encapsulate the raw data frames into data packets in a preset data format according to the standardized field names and corresponding value pairs and transmit them to the intelligent decision-making layer; An intelligent decision-making layer is used to receive multiple data packets sent by multiple heterogeneous monitoring devices, perform priority screening and exception screening on the fields of each data packet, and input a first set of fields whose priority is greater than a preset value and is marked as abnormal into an intelligent reasoning module. The intelligent reasoning module outputs a first intelligent reasoning result to a main control unit, wherein the first intelligent reasoning result includes a vehicle status determination result and a control suggestion.

9. The intelligent data processing system for vehicle-mounted monitoring equipment according to claim 8, characterized in that: The general interface layer also includes: a protocol identification module, configured to perform data structure splitting on the original data frame based on the data transmission channel, perform data structure matching on the data structure splitting results with preset communication protocol templates, and determine the communication protocol type corresponding to the original data frame based on the matching results; a device identification module, configured to extract a device address or a device identifier from the original data frame to identify a device type corresponding to the original data frame; The protocol parsing module is used to match the protocol parsing plug-in based on the combination of the communication protocol type, the device type and the data transmission channel, and load the protocol parsing plug-in to perform field extraction on the original data frame.

Images