Network fault diagnosis method and system applied to car networking service

Abstract

The embodiment of the invention relates to the technical field of Internet of Vehicles, and particularly provides a network fault diagnosis method and system applied to Internet of Vehicles services, which can capture network state information in various communication modes by acquiring a network state operation log of a target Internet of Vehicles service system from a multi-protocol network communication server, and lay a foundation for accurate diagnosis. By mining the network fault diagnosis embedded representation set, multiple rounds of network fault diagnosis can be presented in a specific representation form, and network fault related characteristics are effectively integrated. Linkage fault diagnosis knowledge characterization is obtained in combination with a diagnosis event distribution quantification label, and the incidence relation among all rounds of diagnosis is fully considered, so that diagnosis information is more comprehensive and accurate. According to the method, the target fault diagnosis analysis algorithm is utilized to obtain the fault judgment confidence coefficient, fault features can be analyzed in a hierarchical and multi-dimensional mode, the accuracy of fault diagnosis is improved, the fault car networking service system can be reliably determined, and the efficiency and precision of network fault diagnosis of the car networking service system are integrally improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of vehicle networking, and more particularly, to a network fault diagnosis method and system applied to vehicle networking services. Background Art

[0002] In the context of the rapid development of vehicle networking, the network environment of vehicle networking service systems has become increasingly complex, involving the interaction of multiple communication protocols. However, existing network fault diagnosis technologies face many problems. Traditional methods often can only diagnose for a single protocol or some network states, lacking the ability to integrate comprehensive network state information under multiple communication methods in vehicle networking, resulting in incomplete diagnostic information and easy omission of potential faults.

[0003] At the same time, when performing multiple rounds of network fault diagnosis, existing technologies fail to effectively utilize the correlation between rounds of diagnosis, making the diagnostic results lack coherence and comprehensiveness. Moreover, the accuracy of fault diagnosis in the fault discrimination link is relatively low, making it difficult to accurately determine the faulty vehicle networking service system, thus affecting the normal operation of vehicle networking services and user experience. Summary of the Invention

[0004] The embodiments of the present application at least provide a network fault diagnosis method and system applied to vehicle networking services.

[0005] The embodiment of the present application provides a network fault diagnosis method applied to an Internet of Vehicles service, which is applied to a network fault diagnosis system. The method includes: obtaining a target network state operation log of a target Internet of Vehicles service system in a preceding remote diagnosis task node from a multi-protocol network communication server; wherein the multi-protocol network communication server is configured to match a CAX communication protocol, a YIX communication protocol, an Ethernet communication protocol, and a wireless communication protocol; based on the target network state operation log, mining a network fault diagnosis embedded representation set of the target Internet of Vehicles service system in a preceding remote diagnosis task node; wherein the network fault diagnosis embedded representation set includes X network fault diagnosis embedded representations, each network fault diagnosis embedded representation corresponds to a round of network fault diagnosis implemented by the target Internet of Vehicles service system, and X is a positive integer not less than 1; based on the X network A fault diagnosis embedding representation and X diagnostic event distribution quantization labels are used to obtain X linkage fault diagnosis knowledge representations; wherein one diagnostic event distribution quantization label among the X diagnostic event distribution quantization labels is used to represent the before and after sequence association vectors of a round of network fault diagnosis of the target Internet of Vehicles service system in the X rounds of network fault diagnosis; a fault discrimination confidence is obtained by inputting the X linkage fault diagnosis knowledge representations into a target fault diagnosis analysis algorithm, and in response to the fault discrimination confidence being not less than a set confidence, the target Internet of Vehicles service system is determined as a faulty Internet of Vehicles service system; wherein the target fault diagnosis analysis algorithm includes a cascaded layer of multivariate decision tree branches, and an input of one multivariate decision tree branch of the layer of multivariate decision tree branches is determined based on the outputs of Y multivariate decision tree branches located upstream of the one multivariate decision tree branch, and Y is a positive integer.

[0006] An embodiment of the present application also provides a network fault diagnosis system, including a processor and a memory; the processor and the memory are communicatively connected, and the processor is used to read a computer program from the memory and execute it to implement the above method.

[0007] The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored, and the computer program implements the above method when running.

[0008] The technical solutions provided by the embodiments of the present application may include the following beneficial effects: By obtaining the network status operation logs of the target vehicle networking service system from a multi-protocol network communication server, the network status information under various communication methods in the vehicle networking can be comprehensively covered, laying a foundation for accurate diagnosis. Mining the embedded representation set of network fault diagnosis can present multiple rounds of network fault diagnosis in a specific representation form, effectively integrating the features related to network faults. Combining the diagnostic event distribution quantization tags to obtain the linked fault diagnosis knowledge representation fully considers the correlation between each round of diagnosis, making the diagnostic information more comprehensive and accurate. Using the target fault diagnosis analysis algorithm to obtain the fault discrimination confidence can analyze the fault features hierarchically and multi-dimensionally, improving the accuracy of fault diagnosis, and thus being able to reliably determine the faulty vehicle networking service system, overall improving the efficiency and accuracy of network fault diagnosis for the vehicle networking service system.

[0009] For the description of the effects of the above network fault diagnosis system and computer-readable storage medium, refer to the description of the above method.

[0010] To make the above objects, features, and advantages of the embodiments of the present application more obvious and understandable, the following specific preferred embodiments are given, and in conjunction with the accompanying drawings, the detailed description is as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] To more clearly illustrate the technical solutions of the embodiments of the present application, the accompanying drawings required for use in the embodiments will be briefly introduced below. The accompanying drawings herein are incorporated into the specification and constitute a part of this specification. These drawings show embodiments that conform to the embodiments of the present application and are used together with the specification to illustrate the technical solutions of the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0012] Figure 1 is a block diagram of a network fault diagnosis system shown in an embodiment of the present application.

[0013] Figure 2 is a schematic flowchart of a network fault diagnosis method applied to a vehicle networking service shown in an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] Here, the exemplary embodiments will be described in detail, and the examples are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the embodiments of the present application. On the contrary, they are only examples of devices and methods that are consistent with some aspects of the embodiments of the present application.

[0015] Figure 1 The schematic diagram of the structure of the network fault diagnosis system 10 provided in the embodiment of the present application includes a processor 102, a memory 104, and a bus 106. Among them, the memory 104 is used to store execution instructions, including a memory and an external memory. The memory can also be understood as an internal memory, which is used to temporarily store the operation data in the processor 102 and the data exchanged with the external memory such as the hard disk. The processor 102 exchanges data with the external memory through the memory. When the network fault diagnosis system 10 is running, the processor 102 communicates with the memory 104 through the bus 106, so that the processor 102 executes the network fault diagnosis method applied to the Internet of Vehicles service in the embodiment of the present application.

[0016] Please combine Figure 2 , Figure 2 It is a flow chart of a network fault diagnosis method for a vehicle networking service provided in an embodiment of the present application, which is applied to a network fault diagnosis system. The method may exemplarily include the following steps 202 to 208.

[0017] Step 202: The network fault diagnosis system obtains the target network status operation log of the target Internet of Vehicles service system in the previous remote diagnosis task node from the multi-protocol network communication server.

[0018] Wherein, the multi-protocol network communication server is configured to match the CAX communication protocol, the YIX communication protocol, the Ethernet communication protocol and the wireless communication protocol.

[0019] Step 204: The network fault diagnosis system mines the network fault diagnosis embedded representation set of the target Internet of Vehicles service system in the previous remote diagnosis task node according to the target network status operation log.

[0020] The network fault diagnosis embedded representation set includes X network fault diagnosis embedded representations, each network fault diagnosis embedded representation corresponds to a round of network fault diagnosis implemented by the target Internet of Vehicles service system, and X is a positive integer not less than 1.

[0021] Step 206: The network fault diagnosis system obtains X linkage fault diagnosis knowledge representations according to the X network fault diagnosis embedding representations and the X diagnosis event distribution quantization labels.

[0022] Among them, one diagnostic event distribution quantization label among the X diagnostic event distribution quantization labels is used to characterize the before and after sequence correlation vector of a round of network fault diagnosis of the target Internet of Vehicles service system in the X rounds of network fault diagnosis.

[0023] Step 208: The network fault diagnosis system inputs the X linked fault diagnosis knowledge representations into a target fault diagnosis analysis algorithm to obtain a fault discrimination confidence level. In response to the fault discrimination confidence level being not less than a set confidence level, the target vehicle networking service system is determined as a faulty vehicle networking service system.

[0024] Among them, the target fault diagnosis analysis algorithm includes a cascaded one-layer multi-decision tree branch. The input of one multi-decision tree branch of the one-layer multi-decision tree branch is determined based on the outputs of Y multi-decision tree branches upstream of the one multi-decision tree branch, and Y is a positive integer.

[0025] It can be understood that the network fault diagnosis system is the execution subject in the embodiments of this application. In step 202, the network fault diagnosis system obtains the target network status operation log of the target vehicle networking service system in the previous remote diagnosis task node from the multi-protocol network communication server. The CAX communication protocol, YIX communication protocol, Ethernet communication protocol, and wireless communication protocol matched by the multi-protocol network communication server provide a solid guarantee for comprehensively collecting the network status information of the vehicle networking service system.

[0026] Taking the modern vehicle networking architecture as an example, there are numerous electronic control units (ECUs) inside the vehicle. The communication between these ECUs depends on specific protocols. For example, between the ECUs related to the vehicle's power system, such as the engine control unit and the transmission control unit, the CAX communication protocol may be used for efficient and real-time data interaction. The data transmission under this protocol has strict format requirements and timing regulations to ensure that the engine operation status information (such as parameters like speed and torque) can be accurately transmitted to the transmission control unit, thereby achieving precise control of operations such as gear shifting.

[0027] The YIX communication protocol may be applied to the communication between some auxiliary function modules inside the vehicle, such as the in-vehicle lighting system and the window control system. Although these modules do not have extremely high requirements for real-time performance and accuracy like the power system, stable communication is also required to ensure the realization of normal functions. For example, when the driver presses the window lift button, the window control unit needs to accurately convey the instruction to the window motor through the YIX communication protocol to realize the window lifting system network.

[0028] The Ethernet communication protocol has a wide range of application scenarios in the vehicle networking, especially when vehicles conduct high-speed data interaction with external devices. For example, when a vehicle drives into a maintenance station for fault detection and repair, the maintenance equipment can connect to the vehicle's internal network through an Ethernet interface to quickly read the detailed information of each ECU in the vehicle, including fault codes, configuration parameters, etc. At the same time, data exchange may also occur between the vehicle and some intelligent transportation facilities (such as roadside units) through the Ethernet protocol, such as obtaining traffic condition information and receiving traffic control instructions.

[0029] Wireless communication protocols (such as 4G / 5G) are the key to establishing communication connections between vehicles and cloud servers or other remote devices. In the context of autonomous driving, vehicles need to upload their sensor data (such as camera images, lidar data, etc.) to the cloud for real-time analysis and processing, and at the same time receive traffic condition information, navigation instructions, etc. sent by the cloud. These large amounts of data transmissions rely on high-speed and stable wireless communication protocols.

[0030] The multi-protocol network communication server can comprehensively monitor and record the network states under these different protocols to form an operation log of the target network state. This log contains rich information. For example, for the data packets under the CAX communication protocol, the log will record information such as the packet identifier, data length, sending time, source address, destination address, etc.; for the connections under the Ethernet communication protocol, it will record the connection establishment time, connection duration, data transmission rate, packet loss situation, etc.; for the communications under the wireless communication protocol, it will record information such as signal strength, signal frequency band, network connection type (such as 4G LTE or 5G NR), data traffic, etc. These detailed network state information provides a solid data foundation for subsequent fault diagnosis.

[0031] Enter step 204, and the network fault diagnosis system mines the network fault diagnosis embedding representation set of the target vehicle networking service system in the previous remote diagnosis task nodes based on the operation log of the target network state. The network fault diagnosis embedding representation set in the embodiments of this application contains X network fault diagnosis embeddings (X is a positive integer not less than 1), and each network fault diagnosis embedding corresponds to a round of network fault diagnosis implemented by the target vehicle networking service system.

[0032] To better understand the network fault diagnosis embedding representation, take the linear feature vector as an example. For example, in vehicle networking, six key network state features, namely network bandwidth, network latency, packet loss rate, signal strength, protocol compliance, and data transmission frequency, are selected to construct the network fault diagnosis embedding representation. For a round of network fault diagnosis, its network fault diagnosis embedding representation can be a six-dimensional linear feature vector.

[0033] Consider a situation where, in a network fault diagnosis, it is detected that the network bandwidth is 8 Mbps, the network latency is 40 ms, the packet loss rate is 1.5%, the signal strength is -65 dBm, and a minor violation is found in terms of protocol compliance (which can be represented by 0 or 1, where 0 means compliant and 1 means violated, and here it is 1), and the data transmission frequency is once every 5 seconds. Then the corresponding network fault diagnosis embedding representation is [8, 40, 1.5, -65, 1, 5]. Each element in this vector precisely reflects a state characteristic of the network. Such a comprehensive network fault diagnosis embedding representation can only be formed through meticulous analysis and extraction of a large amount of data in the operation logs of the target network state.

[0034] During the mining process, the network fault diagnosis system will employ various technical means. The diagnostic technology based on protocol analysis plays an important role in this process. For example, for the CAX communication protocol, the system will carefully check whether the format of each CAX protocol data packet in the log conforms to the standard. If it is found that the identifier of the data packet exceeds the specified range, or the data length does not match the protocol requirements, this indicates a protocol compliance issue. In the network fault diagnosis embedding representation, a mark will be made on the corresponding protocol compliance dimension (such as 1 in the above example).

[0035] Signal feature analysis technology is also indispensable. Taking the signal strength as an example, if the wireless signal strength of a vehicle in a certain area should normally be between -60 dBm and -70 dBm, and the operation log shows that the signal strength continuously drops below -75 dBm, this may imply problems such as signal interference or antenna failure. The system will reflect this abnormal signal strength situation in the signal strength dimension of the network fault diagnosis embedding representation.

[0036] In addition, the system may also adopt model-based fault diagnosis technology. For example, a physical model of the network state is constructed based on historical data, and the relationships between various network state characteristics in the normal operating state are defined in the model. By substituting the data currently obtained from the operation log into the model for calculation, if there is a large deviation between the calculation result and the actual measurement value, then it can be determined that there are problems in one or some network state characteristics, and corresponding manifestations will be made in the network fault diagnosis embedding representation.

[0037] Next, in step 206, the network fault diagnosis system obtains X associated fault diagnosis knowledge representations based on X network fault diagnosis embedding representations and X diagnostic event distribution quantization labels. One of the diagnostic event distribution quantization labels in the diagnostic event distribution quantization labels is used to represent the pre-and post-sequence correlation vector of a round of network fault diagnosis of the target vehicle networking service system in X rounds of network fault diagnosis.

[0038] Continue to illustrate with the example of linear feature vectors. For example, for a vehicle networking service system, there are four rounds of network fault diagnosis (X = 4). For the first round of network fault diagnosis, the quantization label of its diagnosis event distribution may be a four-dimensional vector representing the relationship with the subsequent three rounds. If the result of the first-round fault diagnosis shows that the network bandwidth is low, after analysis, this low bandwidth situation may have a greater impact on the diagnosis result of the second round (for example, because low bandwidth may cause incomplete data transmission during the second-round diagnosis, affecting the diagnosis accuracy), a smaller impact on the diagnosis result of the third round, and almost no impact on the fourth round. Then this vector may be [0.7, 0.2, 0.1, 0]. The values in the embodiments of the present application represent the degree of association between the diagnosis result of the first round and the diagnosis results of subsequent rounds.

[0039] When constructing the knowledge representation of linked fault diagnosis, the information in the network fault diagnosis embedding representation and the quantization label of the diagnosis event distribution will be integrated. For example, for the knowledge representation of the first round of linked fault diagnosis, for example, the network fault diagnosis embedding representation of the first round is [8, 40, 1.5, -65, 1, 5], and its quantization label of the diagnosis event distribution is [0.7, 0.2, 0.1, 0]. The weighted summation method can be used to construct the knowledge representation of linked fault diagnosis. For example, set the weight of network bandwidth to 0.2, the weight of network latency to 0.15, the weight of packet loss rate to 0.15, the weight of signal strength to 0.2, the weight of protocol compliance to 0.2, and the weight of data transmission frequency to 0.1.

[0040] Then the first element of the knowledge representation of the first round of linked fault diagnosis is: 8 * 0.2 * 0.7 (network bandwidth multiplied by its weight and then multiplied by the association degree with the second round) + 8 * 0.2 * 0.2 (network bandwidth multiplied by its weight and then multiplied by the association degree with the third round) + 8 * 0.2 * 0.1 (network bandwidth multiplied by its weight and then multiplied by the association degree with the fourth round) = 1.12 + 0.32 + 0.16 = 1.6. Calculate other elements in the same way, and finally obtain the knowledge representation of the first round of linked fault diagnosis as a new vector, such as [1.6, 6.3, 0.27, -9.1, 0.14, 0.35].

[0041] This process can be realized by means of fault diagnosis technology based on artificial intelligence models. For example, construct a neural network model, and use the network fault diagnosis embedding representation and the quantization label of the diagnosis event distribution as input features. In the training stage, a large number of sample data in known normal and fault states are used for training. The neural network model adjusts the weights of internal neurons continuously and learns how to accurately generate the knowledge representation of linked fault diagnosis according to the input network fault diagnosis embedding representation and the quantization label of the diagnosis event distribution.

[0042] Finally, in step 208, the network fault diagnosis system obtains the fault discrimination confidence by inputting X linked fault diagnosis knowledge representations into the target fault diagnosis analysis algorithm. The target fault diagnosis analysis algorithm includes a cascaded one-layer multi-decision tree branch, and the input of a multi-decision tree branch in the one-layer multi-decision tree branch is determined based on the outputs of Y multi-decision tree branches upstream of the one multi-decision tree branch (Y is a positive integer).

[0043] For example, X = 5, that is, there are 5 linked fault diagnosis knowledge representations. Taking the network fault diagnosis in the Internet of Vehicles as an example, these 5 linked fault diagnosis knowledge representations are input into the target fault diagnosis analysis algorithm.

[0044] The first multi-decision tree branch may first analyze the network bandwidth-related features in the linked fault diagnosis knowledge representation. If the network bandwidth is lower than a certain threshold (such as 6 Mbps), it outputs a specific intermediate result, which can be a numerical value or a logical state. For example, if the numerical value corresponding to the network bandwidth in the first linked fault diagnosis knowledge representation is 5 Mbps, which is lower than 6 Mbps, then the first multi-decision tree branch may output a logical state "low bandwidth" indicating insufficient network bandwidth.

[0045] This intermediate result, together with the outputs of other upstream multi-decision tree branches, serves as the input for the next multi-decision tree branch. For example, the second multi-decision tree branch focuses on packet loss rate and signal strength-related features and needs to make a judgment in combination with the output of the first multi-decision tree branch. If the first multi-decision tree branch outputs "low bandwidth", and the second multi-decision tree branch finds that the packet loss rate is also high (such as higher than 3%) and the signal strength is weak (such as lower than -70 dBm), then it may output a result more inclined to the existence of a network fault, such as a numerical value of 0.8, indicating an 80% probability of a network fault.

[0046] With the cascaded processing of the multi-decision tree branches, subsequent multi-decision tree branches will comprehensively consider the outputs of the previous branches and other features in the linked fault diagnosis knowledge representation for further judgment. For example, the third multi-decision tree branch may consider features such as protocol compliance and data transmission frequency, and conduct a more in-depth analysis in combination with the outputs of the previous branches. If the previous branches' outputs indicate insufficient network bandwidth, high packet loss rate, and weak signal strength, and the third multi-decision tree branch finds that there are also protocol violations and unstable data transmission frequency, then it may output a higher numerical value, such as 0.9, indicating a higher probability of a network fault.

[0047] Through such cascaded processing of multiple decision tree branches, the fault discrimination confidence is finally obtained. For example, if the confidence is set to 0.8 and the finally obtained fault discrimination confidence is 0.85, this indicates that the network fault diagnosis system has a high degree of certainty that the target vehicle networking service system has a fault. At this time, the target vehicle networking service system is determined as a faulty vehicle networking service system.

[0048] During the network fault diagnosis process of the entire vehicle networking service system, the network fault diagnosis system starts from obtaining the comprehensive network status operation logs from the multi-protocol network communication server, carefully mines the network fault diagnosis embedded representation set, skillfully combines the diagnostic event distribution quantization labels to construct the linked fault diagnosis knowledge representation, and finally uses the target fault diagnosis analysis algorithm to obtain the accurate fault discrimination confidence, so as to achieve an accurate judgment on whether the target vehicle networking service system is a faulty system. This fault diagnosis solution based on multi-technology integration and multi-data comprehensive processing can effectively meet the fault diagnosis requirements in the complex and changeable network environment of the vehicle network, providing a reliable guarantee for the stable operation of the vehicle networking service.

[0049] In an optional technical idea, the mining of the network fault diagnosis embedded representation set of the target vehicle networking service system in the previous remote diagnosis task node according to the target network status operation log includes: obtaining X network status operation logs corresponding to X rounds of network fault diagnosis implemented by the target vehicle networking service system in the previous remote diagnosis task node; where each of the X network status operation logs includes sensing status monitoring data and vehicle status monitoring data; through data mining of the sensing status monitoring data and the vehicle status monitoring data, X network fault diagnosis embedded representations corresponding to the X network status operation logs are obtained; where the network fault diagnosis embedded representation set includes the X network fault diagnosis embedded representations.

[0050] When the network fault diagnosis system mines the network fault diagnosis embedding representation set of the target vehicle networking service system in the previous remote diagnosis task node according to the target network status operation log, there is a specific operation process. First, the network fault diagnosis system obtains X network status operation logs corresponding to X rounds of network fault diagnosis implemented by the target vehicle networking service system in the previous remote diagnosis task node. Each network status operation log in the embodiments of the present application includes sensing status monitoring data and vehicle status monitoring data. For example, the sensing status monitoring data may include the monitoring values of various vehicle sensors (such as temperature sensors, pressure sensors, etc.), such as the temperature value monitored by the engine temperature sensor is 90 degrees Celsius, and the pressure value monitored by the tire pressure sensor is 2.5 standard atmospheric pressures, etc.; the vehicle status monitoring data can cover information such as the vehicle speed, gear position, steering angle, etc., such as the vehicle speed is 60 kilometers per hour, the gear position is in D gear, and the steering angle is 15 degrees, etc.

[0051] Then, by performing data mining on these sensing status monitoring data and vehicle status monitoring data, X network fault diagnosis embeddings corresponding to the X network status operation logs are obtained. In this process, taking the construction of a network fault diagnosis embedding representation in the form of a linear feature vector as an example, suppose five features, namely vehicle speed, engine temperature, tire pressure, network bandwidth, and packet loss rate, are selected to construct the network fault diagnosis embedding representation. For the network status operation log corresponding to a certain round of network fault diagnosis, where the vehicle speed is 60 kilometers per hour, the engine temperature is 90 degrees Celsius, the tire pressure is 2.5 standard atmospheric pressures, the network bandwidth is 10 Mbps, and the packet loss rate is 1%, then the network fault diagnosis embedding representation corresponding to this round may be [60, 90, 2.5, 10, 1]. Each value in the embodiments of the present application accurately reflects the relevant features mined from the sensing status monitoring data and vehicle status monitoring data in the network status operation log. For different rounds of network fault diagnosis, since the sensing status monitoring data and vehicle status monitoring data will change, the corresponding network fault diagnosis embeddings will also be different. For example, in another round, the vehicle speed becomes 80 kilometers per hour, the engine temperature rises to 95 degrees Celsius, the tire pressure remains 2.5 standard atmospheric pressures, the network bandwidth drops to 8 Mbps, and the packet loss rate rises to 2%, then the network fault diagnosis embedding representation for this round may be [80, 95, 2.5, 8, 2]. These network fault diagnosis embeddings for different rounds together constitute the network fault diagnosis embedding representation set.

[0052] During the data mining process, various technical means are used to accurately extract features for constructing network fault diagnosis embedded representations from sensor status monitoring data and vehicle status monitoring data. For example, for the sensor values in the sensor status monitoring data, they are analyzed based on the normal operating range and accuracy requirements of the sensors. If the normal operating range of a certain temperature sensor is 80 - 100 degrees Celsius, when the mined temperature value is close to the boundary of this range (such as 98 degrees Celsius), when constructing the network fault diagnosis embedded representation, the potential impact of this value on network fault diagnosis will be considered, and it may be reflected in the network fault diagnosis embedded representation in a special coding or weighting manner. For continuously changing data such as vehicle speed in the vehicle status monitoring data, its change trend and relationship with other relevant data are analyzed. For instance, a sudden change in vehicle speed may be associated with the network status. If the network bandwidth drops from 10 Mbps to 8 Mbps while the vehicle speed suddenly increases from 60 km / h to 80 km / h, then this association will be taken into account during the construction of the network fault diagnosis embedded representation.

[0053] Meanwhile, for data related to network status (such as network bandwidth and packet loss rate), mining is carried out in combination with the network architecture and communication protocol of the vehicle network. In the vehicle network, different network devices and communication links have different bandwidth requirements and packet loss tolerances. For example, for the communication link between the vehicle and the cloud server, a higher network bandwidth is required to ensure the transmission of a large amount of data (such as real-time upload of sensor data). If the actually mined network bandwidth is lower than the expected value, this will be reflected in the network fault diagnosis embedded representation, and its weight or coding method in the vector will be determined according to the actual network architecture. The same is true for the packet loss rate. If the packet loss rate is too high in some key communication links (such as the communication between the vehicle control system and safety-related devices), this will be an important consideration factor when constructing the network fault diagnosis embedded representation.

[0054] In this way, the network fault diagnosis system accurately mines a set of network fault diagnosis embedded representations from the network status operation logs containing rich information, which can reflect the network fault diagnosis situation of the target vehicle network service system, providing an important data basis for subsequent further fault diagnosis analysis.

[0055] It can be seen that a set of network fault diagnosis embedded representations is constructed through a specific data mining method. On the one hand, it comprehensively considers the sensing state monitoring data and the vehicle state monitoring data, and this comprehensiveness can more accurately reflect the actual operating state of the vehicle networking system. For example, when considering network faults, various factors such as vehicle speed and engine temperature are combined, avoiding the limitations of single-factor analysis. On the other hand, constructing network fault diagnosis embedded representations with specific numerical values makes the fault diagnosis process more quantitative and operable. Different numerical examples can clearly reflect the characteristic differences in different states, helping to perform more accurate fault diagnosis analysis based on these representations subsequently, thereby improving the accuracy and reliability of network fault diagnosis in the entire vehicle networking service system.

[0056] In a preferred embodiment, obtaining the X linkage fault diagnosis knowledge representations based on the X network fault diagnosis embedded representations and the X diagnosis event distribution quantization labels includes: obtaining the X diagnosis event distribution quantization labels according to the event feature relationships of each of the X network fault diagnosis embedded representations in the network fault diagnosis embedded representation set; performing feature integration on the X network fault diagnosis embedded representations and the X diagnosis event distribution quantization labels to obtain the X linkage fault diagnosis knowledge representations.

[0057] Further, obtaining the X diagnosis event distribution quantization labels according to the event feature relationships of each of the X network fault diagnosis embedded representations in the network fault diagnosis embedded representation set includes: sequentially obtaining each network fault diagnosis embedded representation from the X network fault diagnosis embedded representations as the current network fault diagnosis embedded representation; performing quantization mapping on the current event feature relationship of the current network fault diagnosis embedded representation in the network fault diagnosis embedded representation set to obtain the current diagnosis event distribution quantization label.

[0058] Further, quantifying and mapping the current event feature relationship of the current network fault diagnosis embedded representation in the set of network fault diagnosis embedded representations to obtain a current diagnosis event distribution quantization label includes: in response to the feature size of the current network fault diagnosis embedded representation being Z×1, using a first normalization algorithm to perform a first quantization mapping on the current event feature relationship to obtain a first set of quantization mapping variables; wherein, the first set of quantization mapping variables are the quantization mapping variables in the label region where the distribution feature value in the current diagnosis event distribution quantization label is a first preset value, and Z is a positive integer not less than 2; using a second normalization algorithm to perform a second quantization mapping on the current event feature relationship to obtain a second set of quantization mapping variables; wherein, the second set of quantization mapping variables are the quantization mapping variables in the label region where the distribution feature value in the current diagnosis event distribution quantization label is a second preset value; determining the current diagnosis event distribution quantization label based on the first set of quantization mapping variables and the second set of quantization mapping variables; wherein, the feature size of the current diagnosis event distribution quantization label is Z×1.

[0059] In the process that the network fault diagnosis system obtains X linkage fault diagnosis knowledge representations based on X network fault diagnosis embedded representations and X diagnosis event distribution quantization labels, first, let's look at the link of obtaining X diagnosis event distribution quantization labels according to the event feature relationship of each network fault diagnosis embedded representation in the set of X network fault diagnosis embedded representations.

[0060] Sequentially obtain each network fault diagnosis embedded representation from the X network fault diagnosis embedded representations as the current network fault diagnosis embedded representation. This process is like analyzing a series of complex data one by one. For example, when X = 5, there are five network fault diagnosis embedded representations, and start analyzing from the first one. For example, the network fault diagnosis embedded representation is a vector used to describe the multi-faceted network state information of the vehicle networking system, and each vector contains multiple feature elements.

[0061] Perform a quantization mapping on the current event feature relationship of this current network fault diagnosis embedded representation in the set of network fault diagnosis embedded representations to obtain a current diagnosis event distribution quantization label. This includes multiple detailed operation steps.

[0062] For example, the feature size of the current network fault diagnosis embedded representation is Z×1 (Z is a positive integer not less than 2). Taking a specific numerical example, for instance, Z = 8. The current network fault diagnosis embedded representation may be [12, 18, 25, 30, 35, 40, 45, 50]. Each numerical value in the embodiments of the present application may represent different characteristic values related to the network state. For example, the first numerical value 12 may represent a network bandwidth of 12 Mbps, 18 may represent a network latency of 18 ms, 25 may represent a packet loss rate of 25%, 30 may represent a signal strength of -30 dBm, 35 may represent a specific parameter value under a certain protocol, 40 may represent a data transmission frequency of 40 times per second, 45 may represent a network connection stability index, and 50 may represent another network-related energy consumption index, etc.

[0063] Use the first normalization algorithm to perform the first quantization mapping on the current event feature relationship to obtain the first set of quantization mapping variables. The first normalization algorithm operates based on some internal laws and predefined rules in the network fault diagnosis embedded representation set. For example, for the feature of network bandwidth, the value range in the entire network fault diagnosis embedded representation set may be 10 - 20 Mbps. Then, for the current network bandwidth value of 12 Mbps, through the first normalization algorithm, it may be mapped to a quantization mapping variable related to its relative position in this value range. For example, this variable is calculated according to a specific function, such as (12 - 10) / (20 - 10) = 0.2. Similar operations are also performed on other features such as network latency and packet loss rate according to their respective value ranges in the entire set and the corresponding calculation rules. These quantization mapping variables obtained for different features constitute the first set of quantization mapping variables, which are the quantization mapping variables in the label area where the distribution characteristic value in the current diagnosis event distribution quantization label is the first preset value.

[0064] Next, the second normalization algorithm is used to perform a second quantization mapping on the current event feature relationship, obtaining a second set of quantization mapping variables. The second normalization algorithm is different from the first normalization algorithm. It may consider more the relationship between the current network fault diagnosis embedding representation and other specific network fault diagnosis embedding representations. For example, for the feature of packet loss rate, after considering the comparison relationship with other embedding representations, such as the packet loss rates in the other four network fault diagnosis embedding representations are 20%, 30%, 22%, 28% respectively, and the current packet loss rate is 25%. Through a calculation method based on relative position and difference (which is a possible form of the second normalization algorithm), a specific quantization mapping variable is calculated. For example, calculate the proportion of the average difference from other values in a certain reference range to obtain a quantization mapping variable, such as this variable being 0.3. For each feature, in this way, according to the relationship with other network fault diagnosis embedding representations, quantization mapping is performed, and all these obtained quantization mapping variables constitute the second set of quantization mapping variables, which are the quantization mapping variables in the label area where the distribution feature value in the current diagnostic event distribution quantization label is the second preset value.

[0065] Based on the first set of quantization mapping variables and the second set of quantization mapping variables, the current diagnostic event distribution quantization label is determined. For example, each variable in the first set of quantization mapping variables is combined with the corresponding variable in the second set of quantization mapping variables according to a certain logic. A possible combination method is weighted summation. For example, the weight corresponding to the first set of quantization mapping variables is 0.4, and the weight corresponding to the second set of quantization mapping variables is 0.6. Taking the feature of network bandwidth as an example, the value in the first set of quantization mapping variables is 0.2, and the value in the second set of quantization mapping variables is, for example, 0.3. Then the combined result is 0.2 * 0.4 + 0.3 * 0.6 = 0.26. Calculate in this way for each feature, and finally obtain a current diagnostic event distribution quantization label with a feature size of Z×1 (here Z = 8), such as [0.26, 0.32, 0.28, 0.30, 0.25, 0.33, 0.29, 0.31]. Each numerical value in this label represents the quantization result of the event feature relationship of the current network fault diagnosis embedding representation in the entire set of network fault diagnosis embedding representations. In this way, each network fault diagnosis embedding representation is operated on, thus obtaining X diagnostic event distribution quantization labels.

[0066] Next, feature integration is performed on the X network fault diagnosis embedding representations and the X diagnostic event distribution quantization labels to obtain X coupled fault diagnosis knowledge representations.

[0067] Continuing with the previous numerical example, for instance, five network fault diagnosis embedded representations have been obtained, which are [12, 18, 25, 30, 35, 40, 45, 50], [15, 20, 28, 32, 38, 42, 48, 52], [10, 16, 22, 28, 30, 36, 40, 44], [18, 22, 26, 34, 36, 44, 46, 50], [14, 19, 24, 31, 33, 38, 43, 47] respectively, and the corresponding five diagnostic event distribution quantization labels are [0.26, 0.32, 0.28, 0.30, 0.25, 0.33, 0.29, 0.31], [0.28, 0.30, 0.26, 0.32, 0.27, 0.31, 0.28, 0.30], [0.24, 0.28, 0.22, 0.26, 0.23, 0.29, 0.25, 0.27], [0.30, 0.32, 0.28, 0.34, 0.29, 0.33, 0.30, 0.32], [0.25, 0.29, 0.24, 0.31, 0.26, 0.30, 0.27, 0.29].

[0068] The process of feature integration is a process of complex combination of each element in the network fault diagnosis embedded representation with the corresponding element in the diagnostic event distribution quantization label. A possible way is to first multiply each element in the network fault diagnosis embedded representation by the corresponding element in the diagnostic event distribution quantization label to obtain a new intermediate vector. Taking the first network fault diagnosis embedded representation and its corresponding diagnostic event distribution quantization label as an example, the intermediate vector obtained after multiplication is [12 * 0.26, 18 * 0.32, 25 * 0.28, 30 * 0.30, 35 * 0.25, 40 * 0.33, 45 * 0.29, 50 * 0.31] = [3.12, 5.76, 7.0, 9.0, 8.75, 13.2, 13.05, 15.5].

[0069] Then, further processing is performed on this new vector. For example, the weighted sum of all elements in this vector can be calculated. For example, the weight of each element is preset according to its importance in network fault diagnosis. For example, if the weight vector is [0.1, 0.12, 0.15, 0.13, 0.1, 0.15, 0.12, 0.13], then the weighted sum is as follows: 3.12 * 0.1 + 5.76 * 0.12 + 7.0 * 0.15 + 9.0 * 0.13 + 8.75 * 0.1 + 13.2 * 0.15 + 13.05 * 0.12 + 15.5 * 0.13 = 9.526. This result is a possible form of the first co - failure diagnosis knowledge representation. Operate on each network fault diagnosis embedded representation and its corresponding diagnostic event distribution quantization label in this way to obtain X co - failure diagnosis knowledge representations.

[0070] In this way, co - failure diagnosis knowledge representations are obtained through an accurate quantization mapping and feature integration process. When obtaining the diagnostic event distribution quantization label, deeply explore the internal connections between network fault diagnosis embedded representations to make the diagnostic event distribution quantization label more accurate. Consider factors from multiple perspectives through different normalization algorithms to avoid the limitations of a single algorithm. The feature integration process synthesizes the information of network fault diagnosis embedded representations and diagnostic event distribution quantization labels, making the co - failure diagnosis knowledge representation contain more comprehensive network state information, providing a more accurate data basis for subsequent fault diagnosis, and improving the accuracy and reliability of network fault diagnosis.

[0071] In an alternative embodiment, the feature integration of the X network fault diagnosis embedded representations and the X diagnostic event distribution quantization labels to obtain the X co - failure diagnosis knowledge representations includes: sequentially obtaining each network fault diagnosis embedded representation from the X network fault diagnosis embedded representations as the current network fault diagnosis embedded representation; sequentially obtaining each diagnostic event distribution quantization label from the X diagnostic event distribution quantization labels as the current diagnostic event distribution quantization label; performing knowledge feature cross - ing on the current network fault diagnosis embedded representation and the current diagnostic event distribution quantization label to obtain the current co - failure diagnosis knowledge representation.

[0072] In the process of the network fault diagnosis system performing feature integration on X network fault diagnosis embedded representations and X diagnostic event distribution quantization labels to obtain X co - failure diagnosis knowledge representations, a rigorous and detailed operation process is followed.

[0073] First, sequentially obtain each network fault diagnosis embedded representation from the X network fault diagnosis embedded representations as the current network fault diagnosis embedded representation. The X network fault diagnosis embedded representations in the embodiments of this application are obtained through a series of previous complex operations and data mining, and they carry multi - aspect information of the network state of the vehicle - to - everything service system. For example, X is a relatively large value, such as X = 10, which means there are 10 network fault diagnosis embedded representations to be processed.

[0074] Consider one of the network fault diagnosis embedding representations, which is a multi-dimensional feature vector. For example, this vector is [5, 12, 18, 22, 25, 30, 32, 35, 40, 45]. Each numerical value in the embodiments of the present application corresponds to a specific network state feature. Among them, 5 can represent that the network bandwidth is 5 Mbps, which is a relatively low bandwidth value and may imply limited network transmission capacity; 12 may represent that the network latency is 12 ms, and this latency value may be a value that needs attention in some vehicle networking application scenarios with high real-time requirements; 18 represents that the packet loss rate is 18%, and a high packet loss rate may affect the integrity and accuracy of data; 22 can represent that the signal strength is -22 dBm, and the magnitude of the signal strength is directly related to the stability of wireless communication; 25 can represent the status value of a specific parameter under a certain protocol; 30 represents that the data transmission frequency is 30 times per second, and this frequency reflects the activity of data transmission; numerical values such as 32, 35, 40, 45 may respectively correspond to other different network state features, such as the degree of network congestion, the number of device connections, etc.

[0075] Meanwhile, each diagnostic event distribution quantization label is sequentially obtained from X diagnostic event distribution quantization labels as the current diagnostic event distribution quantization label. These diagnostic event distribution quantization labels are also carefully constructed to represent the quantization result of the event feature relationship of the network fault diagnosis embedding representation in the entire set. Taking one of the diagnostic event distribution quantization labels as an example, for example, it is [0.05, 0.1, 0.15, 0.2, 0.2, 0.15, 0.1, 0.03, 0.01, 0.01]. Each numerical value in the embodiments of the present application has a corresponding relationship with the features in the network fault diagnosis embedding representation. For example, 0.05 is related to the event feature relationship of the network bandwidth in the entire set. It can represent a quantization relationship of the current bandwidth value relative to other bandwidth values in all network fault diagnosis embedding representations, and may imply a relatively low level or specific distribution characteristics of this bandwidth value in the overall.

[0076] Next, perform knowledge feature cross on the current network fault diagnosis embedding representation and the current diagnostic event distribution quantization label to obtain the current associated fault diagnosis knowledge representation. This process is not a simple numerical operation, but a deep fusion based on the internal logic of vehicle networking network fault diagnosis.

[0077] Taking the previously mentioned network fault diagnosis embedding representation [5, 12, 18, 22, 25, 30, 32, 35, 40, 45] and the diagnostic event distribution quantization label [0.05, 0.1, 0.15, 0.2, 0.2, 0.15, 0.1, 0.03, 0.01, 0.01] as an example. One possible way of knowledge feature cross is to first perform a specific operation combination on each corresponding element.

[0078] For the value corresponding to the network bandwidth, 5 × 0.05 gives 0.25. This result is not just a simple multiplication result. It actually fuses the actual value of the network bandwidth with its event feature relationship in the entire network fault diagnosis embedding representation set. From the perspective of network fault diagnosis, this value of 0.25 synthesizes information such as the magnitude of the network bandwidth and its relative position or importance in the whole.

[0079] For the value corresponding to the network latency, 12 × 0.1 gives 1.2. This means combining the actual value of the network latency with its quantization result in the entire event feature relationship, thereby reflecting the comprehensive impact of network latency on network fault diagnosis in a new dimension.

[0080] In this way, the values corresponding to other features such as packet loss rate and signal strength are calculated in turn. 18 × 0.15 gives 2.7, 22 × 0.2 gives 4.4, 25 × 0.2 gives 5, 30 × 0.15 gives 4.5, 32 × 0.1 gives 3.2, 35 × 0.03 gives 1.05, 40 × 0.01 gives 0.4, 45 × 0.01 gives 0.45.

[0081] Then, these calculated values are further integrated. One way is to sum them up, 0.25 + 1.2 + 2.7 + 4.4 + 5 + 4.5 + 3.2 + 1.05 + 0.4 + 0.45 = 23.15. This 23.15 is a numerical result of the current associated fault diagnosis knowledge representation. In actual operation, these values may also be weighted and summed or other more complex operations may be performed according to different network fault diagnosis requirements. For example, according to the different importance of network state features for network fault diagnosis, different weights are set for the calculation results corresponding to each feature. If the weight of the network bandwidth is 0.1, the weight of the network latency is 0.12, the weight of the packet loss rate is 0.15, etc., and then summed according to the weights, the result obtained in this way will more accurately reflect the comprehensive impact of each feature in network fault diagnosis.

[0082] According to such an operation process, knowledge feature cross-operation is performed on each group of current network fault diagnosis embedding representations and current diagnosis event distribution quantization labels. For example, for the next group of network fault diagnosis embedding representations [6, 13, 16, 20, 24, 28, 30, 33, 38, 42] and the corresponding diagnosis event distribution quantization labels [0.06, 0.12, 0.14, 0.18, 0.2, 0.16, 0.08, 0.04, 0.01, 0.01].

[0083] Similarly, following the above operation steps of knowledge feature intersection, calculate each corresponding element first. 6×0.06 gives 0.36, 13×0.12 gives 1.56, 16×0.14 gives 2.24, 20×0.18 gives 3.6, 24×0.2 gives 4.8, 28×0.16 gives 4.48, 30×0.08 gives 2.4, 33×0.04 gives 1.32, 38×0.01 gives 0.38, and 42×0.01 gives 0.42.

[0084] Then perform integration. If in the way of summation: 0.36 + 1.56 + 2.24 + 3.6 + 4.8 + 4.48 + 2.4 + 1.32 + 0.38 + 0.42 = 21.56, which is also the result of the next associated fault diagnosis knowledge representation.

[0085] Continue this process, process the X sets of network fault diagnosis embedded representations and diagnosis event distribution quantization labels one by one, and finally obtain X associated fault diagnosis knowledge representations. These associated fault diagnosis knowledge representations will serve as an important basis for subsequent fault diagnosis analysis. They integrate the key information in the network fault diagnosis embedded representation and diagnosis event distribution quantization label, and can more comprehensively and accurately reflect the network fault status of the vehicle networking service system.

[0086] In this way, the associated fault diagnosis knowledge representation is obtained through the knowledge feature intersection operation. This method effectively deeply integrates the network fault diagnosis embedded representation and diagnosis event distribution quantization label. From the specific numerical example, it can be seen that by performing specific operations on the values corresponding to each feature, the actual value of the network state feature and its event feature relationship in the whole are comprehensively considered, avoiding the one-sidedness of only considering information in one aspect. This enables the associated fault diagnosis knowledge representation to more comprehensively and accurately reflect the information related to network faults, provides a more comprehensive and accurate data basis for subsequent fault diagnosis based on these representations, thereby improving the accuracy and effectiveness of the entire network fault diagnosis, and helping to more accurately judge the network fault situation of the vehicle networking service system.

[0087] In an exemplary embodiment, before inputting the X associated fault diagnosis knowledge representations into the target fault diagnosis analysis algorithm to obtain the fault discrimination confidence, the method further includes: obtaining an algorithm debugging data set; wherein each algorithm debugging data in the algorithm debugging data set includes an associated fault diagnosis knowledge representation set determined according to the network state operation log of the vehicle networking service system sample in the multi-protocol network communication server within the previous remote diagnosis task node; using the algorithm debugging data set to debug the initial fault diagnosis analysis algorithm to obtain the debugged target fault diagnosis analysis algorithm.

[0088] In the next step, the initial fault diagnosis analysis algorithm is debugged by using the algorithm debugging data set, including: performing the u-th debugging on the initial fault diagnosis analysis algorithm through the following steps; where u is a positive integer not less than 1, and when u is 1, the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging is the initial fault diagnosis analysis algorithm that has not passed the debugging: selecting the u-th set of associated fault diagnosis knowledge representation example sets corresponding to the u-th set of vehicle networking service system examples in the algorithm debugging data set; inputting the u-th set of associated fault diagnosis knowledge representation example sets into the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging to obtain the u-th set of fault discrimination confidence levels determined by the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging based on the u-th set of associated fault diagnosis knowledge representation sets; determining the value of the training error of the initial fault diagnosis analysis algorithm in the u-th debugging based on the u-th set of fault discrimination confidence levels and the prior discrimination annotations of the u-th set of vehicle networking service system examples; in the case where the value of the training error in the u-th debugging does not meet the set training requirements, updating the neural network weights in a layer of multi - decision tree branches in the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging to obtain the fault diagnosis analysis algorithm obtained from the u-th debugging.

[0089] In a detailed technical solution, the step of inputting the u-th set of associated fault diagnosis knowledge representation example sets into the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging to obtain the u-th set of fault discrimination confidence levels determined by the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging based on the u-th set of associated fault diagnosis knowledge representation sets includes: inputting the u-th set of associated fault diagnosis knowledge representation example sets into a layer of multi - decision tree branches in the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging to obtain the processed associated fault diagnosis knowledge representation generated by the last multi - decision tree branch in the layer of multi - decision tree branches; inputting the processed associated fault diagnosis knowledge representation into the discrimination activation module in the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging to obtain the u-th set of fault discrimination confidence levels; where the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging includes the layer of multi - decision tree branches and the discrimination activation module.

[0090] Before the network fault diagnosis system obtains the fault discrimination confidence level by inputting X associated fault diagnosis knowledge representations into the target fault diagnosis analysis algorithm, a series of complex and critical algorithm debugging operations are performed, and this process is of crucial significance to the accuracy and effectiveness of the entire network fault diagnosis.

[0091] First, obtain the algorithm debugging data set. Each algorithm debugging data in this set is determined based on the linkage fault diagnosis knowledge representation set obtained from the operation logs of the network states in the previous remote diagnosis task nodes of the vehicle networking service system sample in the multi-protocol network communication server. The vehicle networking service system sample includes numerous different network state situations, which are recorded through the network state operation logs. For example, for a certain vehicle networking service system sample, its network state operation logs may record the fluctuations in network bandwidth. For instance, the network bandwidth may be stable at 10 Mbps during some periods, while it may drop to 5 Mbps during other periods; the network latency will also be recorded accordingly. For example, the average latency is 20 ms, but it may increase to 50 ms when the network is busy; the packet loss rate will also be detailedly recorded. It may be 1% under normal circumstances, but it will increase to 5% when the network becomes unstable, etc. Based on this rich network state operation log information, through a series of complex operations such as mining, quantization, and feature integration, the linkage fault diagnosis knowledge representation set is finally determined. The elements in this set are vectors with specific numerical structures, which comprehensively reflect the network state characteristics of the vehicle networking service system sample. For example, one of the linkage fault diagnosis knowledge representation vectors is [8, 18, 2, 22, 30]. The numerical values in the embodiments of this application may respectively correspond to the results after the previous processing of different network state characteristics. For example, 8 may be related to the adjusted network bandwidth numerical value, 18 may be the comprehensive numerical value related to network latency, 2 may be the specific numerical value related to the packet loss rate, 22 can represent the quantized numerical value of signal strength, and 30 may be related to other network state characteristics such as data transmission frequency.

[0092] Next, use this algorithm debugging data set to debug the initial fault diagnosis analysis algorithm, so as to obtain the debugged target fault diagnosis analysis algorithm. This debugging process is a step-by-step iterative and delicate process. Specifically, the following steps are used to perform the u-th debugging on the initial fault diagnosis analysis algorithm (u is a positive integer not less than 1. When u is 1, the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging is the initial fault diagnosis analysis algorithm that has not passed the debugging).

[0093] When performing the u-th debugging, first, select the u-th set of linked fault diagnosis knowledge representation example sets corresponding to the u-th set of Internet of Vehicles service system examples from the algorithm debugging data set. For example, the algorithm debugging data set contains many such data sets, such as 20 sets. When u = 5, the linked fault diagnosis knowledge representation example set corresponding to the 5th set of Internet of Vehicles service system examples will be selected. The elements in this set are also vectors with specific numerical values. For example, one of the vectors is [10, 22, 3, 25, 35]. The 10 in the embodiments of the present application may represent a network bandwidth-related numerical value after specific processing, which reflects the network bandwidth status of the Internet of Vehicles service system example under specific circumstances; 22 may be a numerical value related to network latency, and this value may be the result of synthesizing the influences of different factors; 3 may be a numerical value related to the packet loss rate, which reflects the packet loss situation of this example; 25 may represent a numerical value related to signal strength, and the magnitude of this value is related to the signal strength status in the Internet of Vehicles service system example; 35 may be related to the data transmission frequency or other network status characteristics.

[0094] Then, input the u-th set of linked fault diagnosis knowledge representation example sets into the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging to obtain the u-th set of fault discrimination confidence levels determined by the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging based on the u-th set of linked fault diagnosis knowledge representation sets. This process involves multiple detailed operation steps.

[0095] First, input the u-th set of linked fault diagnosis knowledge representation example sets into a multi-way decision tree branch in the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging. A multi-way decision tree branch contains multiple multi-way decision tree branches, and these multi-way decision tree branches process the input data in a certain order. For example, the first multi-way decision tree branch may analyze the network bandwidth-related values (such as 10 in the above vector) in the linked fault diagnosis knowledge representation example set. According to pre-set rules or based on previously learned patterns, this multi-way decision tree branch determines whether there is a fault tendency in the network state characteristics represented by this value. If this value is lower than a certain threshold (for example, based on historical data or experience, it is set that when the network bandwidth is lower than 8 Mbps, there may be a fault tendency), then it may pass a preliminary judgment result indicating a possible fault to the next multi-way decision tree branch. The next multi-way decision tree branch will combine this result with other elements in the linked fault diagnosis knowledge representation example set (such as 22, 3, 25, 35, etc.) for further analysis. For example, it may consider whether the network latency (22) is within the normal range, and the relationships between factors such as packet loss rate (3), signal strength (25), and data transmission frequency (35) and the preliminary judgment result of low network bandwidth. If the network latency is high, the packet loss rate is low but the signal strength is weak, and at the same time the data transmission frequency is unstable, then this multi-way decision tree branch may further strengthen the judgment result of a fault and pass this more in-depth judgment result to the next multi-way decision tree branch. In this way, after passing through each multi-way decision tree branch in the multi-way decision tree branch layer in turn, finally, the processed linked fault diagnosis knowledge representation generated by the last multi-way decision tree branch in the multi-way decision tree branch layer is obtained.

[0096] For example, this processed linked fault diagnosis knowledge representation may be a vector recombined or adjusted after passing through multiple decision tree branches, such as [12, 20, 4, 20, 30]. The values in the embodiments of the present application have changed compared with the values in the original input linked fault diagnosis knowledge representation example set, and this change reflects the analysis and adjustment results of the multi-way decision tree branch layer on these data. For example, the value 12 may be an adjustment of the network bandwidth-related value after comprehensively considering multiple factors such as network bandwidth, network latency, and packet loss rate; 20 may be a re-evaluation of the network latency-related value; 4 may be an adjustment of the packet loss rate-related value, which may be because during the comprehensive analysis, it is found that although the initial packet loss rate is low, the risk assessment of the packet loss rate has increased after considering other factors; 20 may be an adjustment of the signal strength-related value, reflecting the mutual influence between signal strength and other factors during the entire analysis process; 30 may be an adjustment of the data transmission frequency-related value.

[0097] Next, by inputting the processed linked fault diagnosis knowledge representation into the discrimination activation module in the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging, the u-th group of fault discrimination confidence levels is obtained. The discrimination activation module, based on the processed linked fault diagnosis knowledge representation, outputs a numerical value representing the fault discrimination confidence level according to its internal calculation logic and parameter settings. For example, this numerical value may be 0.65, indicating a 65% probability of a network fault. The calculation logic of the discrimination activation module may be based on some complex functional relationships, and it comprehensively considers each numerical value in the processed linked fault diagnosis knowledge representation and some pre-set weights and other factors. For example, for a numerical value related to network bandwidth (12), if a relatively high weight is set in the discrimination activation module, then the influence of this numerical value on the final fault discrimination confidence level will be relatively large. If a low network bandwidth is considered an important indicator of a network fault, then when this numerical value is low, even if other numerical values are within the normal range, the fault discrimination confidence level may be relatively high.

[0098] After obtaining the u-th group of fault discrimination confidence levels, based on the u-th group of fault discrimination confidence levels and the prior discrimination annotations of the u-th group of vehicle Internet of Things service system samples, the value of the training error of the initial fault diagnosis analysis algorithm in the u-th debugging is determined. For example, the prior discrimination annotations of the u-th group of vehicle Internet of Things service system samples indicate that this sample actually has a fault (1 can be used to represent the existence of a fault, and 0 represents no fault), and the obtained u-th group of fault discrimination confidence level is 0.65. If the mean square error is used to calculate the training error, then the training error may be a numerical value obtained from (0.65 - 1)² = 0.1225. This numerical value of the training error reflects the degree of difference between the current fault discrimination confidence level and the actual situation (prior discrimination annotations).

[0099] In the case where the value of the training error in the u-th debugging does not meet the set training requirements, the neural network weights in a layer of the multi-way decision tree branch in the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging are updated to obtain the fault diagnosis analysis algorithm obtained from the u-th debugging. For example, if the set training requirement is that the training error should be less than 0.1, and the currently calculated training error is 0.1225, then the neural network weights in a layer of the multi-way decision tree branch need to be updated. This update process may adjust the weights in the neural network according to the magnitude and direction of the training error, following a certain algorithm (such as the principle of backpropagation algorithm, etc.).

[0100] For example, in a certain multi - decision tree branch within a layer of multi - decision tree branches, the weight associated with the network bandwidth feature is 0.3, the weight associated with the network latency feature is 0.2, the weight associated with the packet loss rate feature is 0.1, the weight associated with the signal strength feature is 0.2, and the weight associated with the data transmission frequency feature is 0.2. Since the calculated training error is relatively large and it is found that the fault discrimination confidence is lower than the value expected by the prior discrimination annotation (i.e., there is actually a fault but the discrimination confidence is low), the weight associated with the network bandwidth feature may be increased, for example, increased to 0.4, while appropriately adjusting other weights. For instance, the weight associated with the network latency feature is adjusted to 0.15, the weight associated with the packet loss rate feature is adjusted to 0.12, the weight associated with the signal strength feature is adjusted to 0.18, and the weight associated with the data transmission frequency feature is adjusted to 0.15. Such an adjustment is based on the result of re - evaluating the importance of each network state feature in fault judgment. Through this weight adjustment, in subsequent analyses, more attention can be paid to the impact of the network bandwidth feature on fault judgment, while also taking into account the impacts of other features.

[0101] By continuously selecting different sets of linked fault diagnosis knowledge representation sample sets corresponding to the vehicle - to - everything (V2X) service system samples for debugging and continuously updating the neural network weights in the fault diagnosis analysis algorithm until the set training requirements are met. For example, when after multiple debuggings (such as u gradually increasing from 1 to 10), the training error is finally less than the set 0.1, the resulting fault diagnosis analysis algorithm is the target fault diagnosis analysis algorithm after debugging. When this target fault diagnosis analysis algorithm after debugging inputs X linked fault diagnosis knowledge representations in the subsequent process, it can obtain the fault discrimination confidence more accurately, thereby more effectively diagnosing the network faults of the V2X service system. Because the algorithm after debugging has a more reasonable assessment of the importance of different network state features in fault judgment and can accurately judge the possibility of faults based on the input linked fault diagnosis knowledge representation.

[0102] Thus, it can be seen that debugging the initial fault diagnosis analysis algorithm by using the algorithm debugging data set has beneficial effects in many aspects. From the specific numerical examples, it can be seen that calculating the training error based on the fault discrimination confidence and the prior discrimination annotation during the debugging process can accurately evaluate the performance of the algorithm. Adjusting the neural network weights according to the training error can optimize the algorithm's judgment ability for different network state features. This debugging method helps to improve the accuracy of the fault discrimination confidence, enabling the algorithm to more accurately judge the possibility of network faults. Furthermore, it improves the reliability and effectiveness of the entire network fault diagnosis of the V2X service system, ensuring that network faults can be discovered and processed in a timely and accurate manner in practical applications and guaranteeing the normal operation of the V2X service.

[0103] Based on steps 202 - 208, the method further includes: when the fault discrimination confidence level is greater than the set confidence level, reporting the fault discrimination confidence level to the vehicle networking maintenance platform server, and giving a fault warning to the target vehicle networking service system.

[0104] Further, reporting the fault discrimination confidence level to the vehicle networking maintenance platform server and giving a fault warning to the target vehicle networking service system includes: determining the target vehicle networking service system as a faulty vehicle networking service system through the fault calibration thread of the vehicle networking maintenance platform server; sending a warning prompt or starting a maintenance task to the target vehicle networking service system through the fault warning thread of the vehicle networking maintenance platform server; wherein the maintenance task is used to adjust the execution strategy of the network fault diagnosis of the target vehicle networking service system.

[0105] After the network fault diagnosis system executes steps 202 - 208, there are a series of subsequent operations, which are started when the fault discrimination confidence level is greater than the set confidence level, and are of great significance for the fault management and maintenance of the entire vehicle networking service system.

[0106] When the fault discrimination confidence level is greater than the set confidence level, for example, the set confidence level is 0.8, if the fault discrimination confidence level is 0.85, the network fault diagnosis system will report the fault discrimination confidence level to the vehicle networking maintenance platform server and give a fault warning to the target vehicle networking service system. This process is not a simple information transmission and prompt, but involves complex operations in multiple links.

[0107] First, report the fault discrimination confidence level to the vehicle networking maintenance platform server. In this process, the value of the fault discrimination confidence level contains a lot of information about the possibility of network faults in the target vehicle networking service system. The vehicle networking maintenance platform server is the core part responsible for maintenance and management in the entire vehicle networking system. It receives the status information from each vehicle networking service system and centrally processes this information. The fault discrimination confidence level reported by the network fault diagnosis system is like a detailed fault report, providing an important basis for the vehicle networking maintenance platform server to judge whether the target vehicle networking service system is faulty.

[0108] After receiving the fault discrimination confidence level, the vehicle networking maintenance platform server determines the target vehicle networking service system as a faulty vehicle networking service system through its fault calibration thread. This determination process is based on precise logic and rules. Taking the previously mentioned fault discrimination confidence level of 0.85 as an example, the fault calibration thread of the vehicle networking maintenance platform server will follow a pre-set determination criterion, which may be determined based on a large amount of experimental data, historical fault statistics, and the performance requirements of the vehicle networking service system. When the fault discrimination confidence level exceeds the set confidence level, it indicates that there is a relatively high probability of a fault in the target vehicle networking service system, sufficient to be marked as a faulty vehicle networking service system.

[0109] In the database of the vehicle networking maintenance platform server, the relevant information of the target vehicle networking service system will be updated. For example, the database of the vehicle networking maintenance platform server stores a status information table for each vehicle networking service system, which includes fields such as system identification, network status, and fault history. When the target vehicle networking service system is determined as a faulty vehicle networking service system, its status field will be updated from "normal" to "faulty". At the same time, information such as the current fault discrimination confidence level and the approximate time of the fault occurrence (determined based on the reported time) will be recorded in other relevant fields. These records not only help to track and analyze the current fault but also provide a data basis for subsequent fault statistics and prevention.

[0110] Next, the vehicle networking maintenance platform server will send a warning message or initiate a maintenance task to the target vehicle networking service system through its fault warning thread. Among them, the maintenance task is used to adjust the execution strategy of the network fault diagnosis of the target vehicle networking service system.

[0111] Regarding the sending of warning messages, the purpose of this operation is to promptly notify relevant personnel or devices of the target vehicle networking service system about the possible network fault situation. The forms and contents of the warning messages are diverse. For example, if the target vehicle networking service system is a vehicle's vehicle networking system, there is a dedicated display terminal inside the vehicle for receiving information from the vehicle networking maintenance platform server. The fault warning thread of the vehicle networking maintenance platform server will send a signal containing a specific code to the target vehicle networking service system. After being received at the vehicle end, decoded, and processed, detailed warning information will be displayed on the display terminal.

[0112] For example, the warning information of the display terminal is presented in text form, which may show "Network Failure Warning: According to the latest diagnosis, there is a risk of failure in your vehicle networking service system, and the confidence level of failure discrimination is 0.85. Please check it in time." The value of the failure discrimination confidence level in this information clearly informs the user of the degree of possibility of the failure. At the same time, the warning information may also contain some auxiliary information, such as "The current network bandwidth may be lower than the normal level, which may affect some of your vehicle networking functions, such as obtaining real-time traffic information, remote vehicle control, etc." Such detailed warning prompts can enable users to have a preliminary understanding of the failure situation so as to take corresponding measures.

[0113] In addition to displaying the warning information on the vehicle terminal, the warning prompt may also be conveyed in other ways. For example, if the vehicle is equipped with a voice interaction system, the warning information can be converted into a voice prompt, such as "There may be a failure in your vehicle networking service system, and the possibility of failure is relatively high. Please check it as soon as possible." Such a voice prompt can remind the user more timely during driving, preventing the user from missing the text prompt on the display terminal due to focusing on driving.

[0114] In terms of starting the maintenance task, this operation is to deal with the network failure of the target vehicle networking service system more deeply. The maintenance task is used to adjust the execution strategy of the network failure diagnosis of the target vehicle networking service system, which is a targeted and dynamic process.

[0115] For example, the original network failure diagnosis execution strategy of the target vehicle networking service system is based on certain default settings. For example, in terms of network parameter inspection, it may check the three main network parameters of network bandwidth, packet loss rate, and signal strength every 1 hour, and when checking the network bandwidth, only check the bandwidth situation of a specific frequency band (such as the frequency band for the vehicle to communicate with nearby base stations); in terms of the depth of failure diagnosis, it may only perform a preliminary protocol compliance check, for example, only check whether the basic formats of the most commonly used several communication protocols (such as CAN protocol, Ethernet protocol) are correct.

[0116] After the maintenance task is started, these execution strategies will change. In terms of the time interval of network parameter inspection, it may be shortened from every 1 hour to every 15 minutes. This is because the confidence level of failure discrimination is relatively high, indicating a greater possibility of failure, and it is necessary to check the network status more frequently to timely detect changes in the failure. At the same time, in terms of the types of network parameters to be inspected, in addition to the original network bandwidth, packet loss rate, and signal strength, it may also increase the inspection of other network parameters, such as the detailed distribution of network latency (not only the average latency, but also the latency changes at different times and under different operations), the degree of network congestion (determined by analyzing the distribution of data traffic and queuing situation in the network), etc.

[0117] In terms of the depth of fault diagnosis, instead of merely conducting preliminary protocol compliance checks, more in-depth protocol analysis will be carried out. For example, for the CAN protocol, in addition to checking whether the basic format is correct, the accuracy of the identifiers in each data packet, the consistency between the data in the data field and the actual device status, etc. will also be checked; for the Ethernet protocol, more complex network configuration parameters will be checked, such as whether the settings of VLAN (Virtual Local Area Network) are correct, whether the IP address allocation is reasonable, etc. This in-depth fault diagnosis can more comprehensively identify the causes of network faults and improve the accuracy of fault diagnosis.

[0118] In addition, maintenance tasks may also involve adjusting network fault diagnosis tools. For example, the originally used network fault diagnosis tool may only have basic data packet capture and simple analysis functions. After starting the maintenance task, more advanced network fault diagnosis tools may be enabled. These advanced tools can perform real-time network traffic analysis, and by deeply analyzing the data flow direction, traffic volume, detailed content of data packets, etc. in the network traffic, potential fault hazards hidden under the appearance of normal traffic can be discovered.

[0119] In the entire vehicle networking service system, this operation process of reporting the fault discrimination confidence level to the vehicle networking maintenance platform server when the fault discrimination confidence level is greater than the set confidence level and then performing subsequent fault calibration and fault warning is a complete and orderly fault management system. It is not only for the fault handling of a single target vehicle networking service system, but also of great significance for the stable operation and maintenance of the entire vehicle networking system.

[0120] From the perspective of the overall stability of the vehicle networking system, timely and accurately identifying a faulty vehicle networking service system and initiating warning and maintenance tasks can prevent the spread of faults. For example, if there is a network fault in a vehicle's vehicle networking service system that is not dealt with in a timely manner, it may affect the normal communication of other vehicles or infrastructure (such as traffic lights, roadside units, etc.) that communicate with it. Through this fault management system, measures can be taken at the initial stage of the fault to prevent the fault from having a greater impact on the entire vehicle networking system.

[0121] From the perspective of user experience, users can receive early warning prompts about faults in the vehicle networking service system in a timely manner, so that they can take corresponding measures in advance. For example, if users know that there may be a network fault in the vehicle networking service system that affects the remote vehicle control function, they can avoid performing remote operations during the period when the fault has not been repaired, preventing possible operation failures or safety risks. At the same time, by starting maintenance tasks to adjust the network fault diagnosis execution strategy, faults can be repaired more quickly, reducing the inconvenience suffered by users due to network faults and improving users' satisfaction with vehicle networking services.

[0122] In terms of the accuracy and efficiency of fault diagnosis, the adjusted network fault diagnosis execution strategy can check the network status more comprehensively and deeply, helping to more accurately locate the cause of the fault. Frequent network parameter checks and more in-depth protocol analysis can detect fault points in a shorter time, thus improving the efficiency of fault diagnosis. This is crucial for ensuring the normal operation of the vehicle networking service system. In the vehicle networking environment, quick and accurate fault diagnosis can reduce the service interruption time caused by network faults and improve the reliability of the entire vehicle networking service.

[0123] The above technical solution constructs a comprehensive fault management system for the vehicle networking service system. By reporting the fault discrimination confidence level and performing operations such as fault calibration, early warning prompts, and maintenance task initiation, the accuracy and timeliness of fault handling are improved. This is specifically reflected in determining the faulty system based on the clear fault discrimination confidence level, informing users of the fault situation through detailed early warning prompts, and adjusting the diagnostic strategy to deeply investigate the fault. This helps to ensure the stability of the vehicle networking system, enhance the user experience, optimize the fault diagnosis management and repair efficiency, and reduce the impact of faults on the entire vehicle networking.

[0124] In an independent embodiment, the steps of adjusting the execution strategy of the network fault diagnosis of the target vehicle networking service system include: obtaining the initial fault diagnosis execution strategy of the target vehicle networking service system; respectively extracting communication system network diagnosis elements and power system network diagnosis elements from multiple strategy task instruction texts in the initial fault diagnosis execution strategy to obtain a communication system network diagnosis element extraction result set and a power system network diagnosis element extraction result set; performing a first diagnosis element optimization on the communication system network diagnosis element extraction result set through a first set of diagnosis element optimization rules to obtain a first diagnosis element optimization feature set for the communication system network; performing a second diagnosis element optimization on the power system network diagnosis element extraction result set through a second set of diagnosis element optimization rules to obtain a second diagnosis element optimization feature set for the power system network; integrating based on the first diagnosis element optimization feature set and the second diagnosis element optimization feature set to obtain a fault diagnosis strategy optimization feature set that matches the target system network in the initial fault diagnosis execution strategy; the target system network includes at least one of a communication system network system network and a power system network system network; adjusting the initial fault diagnosis execution strategy based on the fault diagnosis strategy optimization feature set.

[0125] In the above embodiments, first, obtain the initial fault diagnosis execution policy of the target vehicle networking service system. This initial policy is a collection of a series of instructions and rules, existing in text form, covering various aspects of network fault diagnosis for the target vehicle networking service system. For example, some contents in the initial fault diagnosis execution policy may be as follows: "Within the first 30 minutes after the vehicle starts, check the network connection status every 10 minutes, including checking whether the network interface is normal and whether the network protocol starts normally; for data transmission in network communication, count the total data transmission volume, average transmission rate, and packet loss situation of data transmission every hour; for the power system network, check the communication link status between the power system control unit and other related units (such as the sensor unit, actuator unit) every day, and check parameters such as the signal strength and transmission delay of the link; when a suspected network fault occurs, perform a simple syntax check on the network configuration file to see if there are configuration errors." Next, respectively perform communication system network diagnosis element extraction and power system network diagnosis element extraction on multiple policy task indication texts in the initial fault diagnosis execution policy. Taking the partial content in the initial fault diagnosis execution policy mentioned above as an example, for the policy task indication text "Within the first 30 minutes after the vehicle starts, check the network connection status every 10 minutes, including checking whether the network interface is normal and whether the network protocol starts normally", when performing communication system network diagnosis element extraction, a communication system network diagnosis element extraction result set containing elements such as network connection status, network interface, and network protocol startup situation will be obtained. For example, it can be represented in the form of a set as {network connection status, network interface, network protocol startup}. For the policy task indication text "For the power system network, check the communication link status between the power system control unit and other related units (such as the sensor unit, actuator unit) every day, and check parameters such as the signal strength and transmission delay of the link", after performing power system network diagnosis element extraction, the obtained power system network diagnosis element extraction result set can be represented as {communication link status between the power system control unit and other units, signal strength, transmission delay}.

[0126] Then, through the first set of diagnostic element optimization rules, the first diagnostic element optimization is performed on the communication system network diagnostic element extraction result set to obtain the first diagnostic element optimization feature set for the communication system network. The first set of diagnostic element optimization rules is formulated based on various characteristics and requirements of the communication system. For example, the optimization rules are determined according to factors such as the influence degree of different elements in the communication system on the overall network function, the probability of failure occurrence, and the importance to the vehicle networking service. In the communication system, whether the network interface is normal is directly related to whether the network can establish a connection, and its importance is very high. If, according to historical data statistics, the probability of the network interface failing is relatively high, and once a failure occurs, it will have a serious impact on multiple services of the vehicle networking (such as information interaction between the vehicle and external servers, communication between various systems inside the vehicle, etc.). Then, in the optimization process, this element of the network interface will be treated with emphasis. For example, the first diagnostic element optimization feature set after optimization is {network interface (high importance), network connection status, network protocol startup}, and the "(high importance)" in the embodiment of the present application indicates the special important position of the network interface in this feature set.

[0127] Similarly, through the second set of diagnostic element optimization rules, the second diagnostic element optimization is performed on the power system network diagnostic element extraction result set to obtain the second diagnostic element optimization feature set for the power system network. The second set of diagnostic element optimization rules is also formulated based on the characteristics of the power system network itself. For example, the communication link status between the power system control unit and other units in the power system network is crucial for the overall operation coordination of the power system. If there is a problem with the link status, it may lead to serious consequences such as unstable power output and abnormal vehicle driving. Moreover, according to past failure records, although the possibility of the communication link status failing is not very high, once a failure occurs, the affected range is wide and the repair difficulty is relatively large. Therefore, in the optimization process, this element will be highlighted. For example, the second diagnostic element optimization feature set after optimization is {communication link status between the power system control unit and other units (critical), signal strength, transmission delay}, and the "(critical)" reflects the key position of this element.

[0128] Integrate the optimized feature set based on the first diagnostic element and the optimized feature set based on the second diagnostic element to obtain an optimized feature set of the fault diagnosis strategy that matches the target system network in the initial fault diagnosis execution strategy. The target system network includes at least one of the communication system network and the power system network. For example, the integrated optimized feature set of the fault diagnosis strategy may be {network interface (high importance), network connection status, network protocol startup, communication link status between the power system control unit and other units (critical), signal strength, transmission delay}. This feature set combines the optimized diagnostic elements in the communication system network and the power system network, forming a more comprehensive set that can better reflect the fault diagnosis requirements of the target vehicle networking service system network.

[0129] Finally, adjust the initial fault diagnosis execution strategy based on the optimized feature set of the fault diagnosis strategy. For example, in the initial fault diagnosis execution strategy, the inspection of the network interface was originally just a simple check of whether it was connected. In the adjusted strategy, since the network interface is marked as highly important in the optimized feature set of the fault diagnosis strategy, the inspection method will become more comprehensive and in-depth. It may increase the detailed detection of the hardware status of the network interface, including whether the pin connections of the interface are loose and whether the operating temperature of the interface chip is normal; conduct a more detailed inspection of the software configuration of the network interface to check for compatibility issues with other devices and unauthorized access settings. For the inspection of the network connection status, which was originally just a simple judgment of whether the connection was successful, the adjusted strategy may increase the continuous monitoring of the connection stability. For example, conduct a sampling detection of the connection status every 5 minutes and record the number and duration of connection interruptions. For the inspection of the network protocol startup, in addition to checking whether it starts normally, it will also check whether each step in the startup process follows the standard process and whether there are abnormal waiting times or error messages.

[0130] Regarding the power system network, since the communication link status between the power system control unit and other units is marked as a key element, in the adjusted strategy, the inspection frequency will be greatly increased. Originally, it was checked once a day, and now it may be changed to once an hour. The inspection content will also be more in-depth. It will not only check the signal strength and transmission delay of the link, but also check the data accuracy in the communication link. For example, check whether the control instruction data sent from the power system control unit to the actuator unit is complete and whether it has been tampered with. For the inspection of the signal strength and transmission delay, it will also be more accurate. For example, use more precise instruments or algorithms to measure the signal strength and record the changes in the transmission delay under different working conditions (such as vehicle acceleration, deceleration, idling, etc.).

[0131] In this way, through a comprehensive adjustment of the initial fault diagnosis execution strategy, the network fault diagnosis of the target vehicle networking service system can be carried out more effectively. This adjusted strategy fully considers the importance and particularity of each element in the communication system network and the power system network, improves the accuracy and efficiency of fault diagnosis, helps to timely discover and solve the network fault problems in the vehicle networking service system, and ensures the normal operation of the vehicle networking service.

[0132] In this way, the fault diagnosis execution strategy is adjusted by extracting, optimizing, and integrating detailed diagnosis elements. From the specific diagnosis element processing process, the importance of each element is determined according to its characteristics in the communication and power systems. For example, elements such as the network interface and the communication link status of the power system control unit are key optimized. This adjusted strategy is comprehensive and targeted, can diagnose faults more accurately, improves the effectiveness and timeliness of fault diagnosis, and is of great significance for ensuring the stable operation of the vehicle networking service system.

[0133] In summary, in the embodiment of the present application, by obtaining the network status operation log of the target vehicle networking service system from a multi-protocol network communication server that matches multiple protocols (CAX, YIX, Ethernet, wireless communication protocol), the network status information under various communication methods in the vehicle networking can be comprehensively covered, laying a foundation for accurate diagnosis. Mining the network fault diagnosis embedded representation set can present multiple rounds of network fault diagnosis in a specific representation form, effectively integrating network fault-related features. Combining the diagnostic event distribution quantization label to obtain the associated fault diagnosis knowledge representation fully considers the correlation relationship between each round of diagnosis, making the diagnostic information more comprehensive and accurate. Using the target fault diagnosis analysis algorithm including the cascade multi-decision tree branches to obtain the fault discrimination confidence, the algorithm with this structure can analyze fault features hierarchically and multi-dimensionally, improves the accuracy of fault diagnosis, and thus can reliably determine the faulty vehicle networking service system, overall improving the efficiency and accuracy of the network fault diagnosis of the vehicle networking service system.

[0134] Furthermore, a readable storage medium is also provided, on which a program is stored, and when the program is executed by a processor, the above method is implemented.

[0135] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems and devices described above by way of example can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein. In the several embodiments provided in the embodiments of the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings, direct couplings, or communication connections shown or discussed with each other can be through some communication interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

Claims

1. A network fault diagnosis method applied to vehicle networking services, characterized in that, The method is performed by a network fault diagnosis system, and the method comprises: Obtaining a target network status operation log of a target Internet of Vehicles service system in a preceding remote diagnosis task node from a multi-protocol network communication server; wherein the multi-protocol network communication server is configured to match a CAX communication protocol, a YIX communication protocol, an Ethernet communication protocol, and a wireless communication protocol; According to the target network state operation log, a network fault diagnosis embedded representation set of the target Internet of Vehicles service system in the preceding remote diagnosis task node is mined; wherein the network fault diagnosis embedded representation set includes X network fault diagnosis embedded representations, each network fault diagnosis embedded representation corresponds to a round of network fault diagnosis implemented by the target Internet of Vehicles service system, and X is a positive integer not less than 1; According to the X network fault diagnosis embedding representations and the X diagnostic event distribution quantization labels, X linkage fault diagnosis knowledge representations are obtained; wherein one diagnostic event distribution quantization label among the X diagnostic event distribution quantization labels is used to represent the before and after sequence association vectors of a round of network fault diagnosis of the target Internet of Vehicles service system in the X rounds of network fault diagnosis; The fault diagnosis and analysis algorithm is inputted into the X linkage fault diagnosis knowledge representations to obtain a fault discrimination confidence, and in response to the fault discrimination confidence being not less than a set confidence, the target Internet of Vehicles service system is determined as a faulty Internet of Vehicles service system; wherein the target fault diagnosis and analysis algorithm includes a cascaded layer of multivariate decision tree branches, and an input of a multivariate decision tree branch of the layer of multivariate decision tree branches is determined based on the outputs of Y multivariate decision tree branches located upstream of the one multivariate decision tree branch, and Y is a positive integer.

2. The method according to claim 1, characterized in that, The mining of the network fault diagnosis embedded representation set of the target Internet of Vehicles service system in the preceding remote diagnosis task node according to the target network status operation log includes: Obtaining X network status operation logs corresponding to X rounds of network fault diagnosis implemented by the target Internet of Vehicles service system in the preceding remote diagnosis task node; wherein each of the X network status operation logs includes sensor status monitoring data and vehicle status monitoring data; By performing data mining on the sensor state monitoring data and the vehicle state monitoring data, X network fault diagnosis embedded representations corresponding to the X network state operation logs are obtained; wherein the network fault diagnosis embedded representation set includes the X network fault diagnosis embedded representations.

3. The method according to claim 1, wherein The method of obtaining X linkage fault diagnosis knowledge representations based on the X network fault diagnosis embedding representations and the X diagnostic event distribution quantization labels includes: According to the event feature relationship of each network fault diagnosis embedding representation in the network fault diagnosis embedding representation set, obtaining the X diagnosis event distribution quantization labels; Feature integration is performed on the X network fault diagnosis embedding representations and the X diagnosis event distribution quantization labels to obtain the X linkage fault diagnosis knowledge representations.

4. The method according to claim 3, wherein Obtaining the X diagnostic event distribution quantization labels according to the event feature relationships of each network fault diagnosis embedding representation in the set of network fault diagnosis embedding representations includes: Sequentially obtaining each network fault diagnosis embedding representation from the X network fault diagnosis embedding representations as the current network fault diagnosis embedding representation; Performing quantization mapping on the current event feature relationship of the current network fault diagnosis embedding representation in the set of network fault diagnosis embedding representations to obtain the current diagnostic event distribution quantization label.

5. The method according to claim 4, characterized in that, The performing quantization mapping on the current event feature relationship of the current network fault diagnosis embedding representation in the set of network fault diagnosis embedding representations to obtain the current diagnostic event distribution quantization label includes: In response to the feature size of the current network fault diagnosis embedding representation being Z×1, performing first quantization mapping on the current event feature relationship by using a first normalization algorithm to obtain a first set of quantization mapping variables; wherein, the first set of quantization mapping variables are quantization mapping variables in a label region where the distribution feature value in the current diagnostic event distribution quantization label is a first preset value, and Z is a positive integer not less than 2; Performing second quantization mapping on the current event feature relationship by using a second normalization algorithm to obtain a second set of quantization mapping variables; wherein, the second set of quantization mapping variables are quantization mapping variables in a label region where the distribution feature value in the current diagnostic event distribution quantization label is a second preset value; Determining the current diagnostic event distribution quantization label according to the first set of quantization mapping variables and the second set of quantization mapping variables; wherein, the feature size of the current diagnostic event distribution quantization label is Z×1.

6. The method according to claim 3, wherein The performing feature integration on the X network fault diagnosis embedding representations and the X diagnostic event distribution quantization labels to obtain the X linkage fault diagnosis knowledge representations includes: Sequentially obtaining each network fault diagnosis embedding representation from the X network fault diagnosis embedding representations as the current network fault diagnosis embedding representation; Sequentially obtaining each diagnostic event distribution quantization label from the X diagnostic event distribution quantization labels as the current diagnostic event distribution quantization label; Performing knowledge feature crossing on the current network fault diagnosis embedding representation and the current diagnostic event distribution quantization label to obtain the current linkage fault diagnosis knowledge representation.

7. The method according to claim 1, wherein Before obtaining the fault discrimination confidence by inputting the X linkage fault diagnosis knowledge representations into the target fault diagnosis analysis algorithm, the method further includes: Obtaining an algorithm debugging data set; wherein each algorithm debugging data in the algorithm debugging data set includes a set of linkage fault diagnosis knowledge representations determined according to the network state operation logs of the vehicle networking service system sample in the multi-protocol network communication server in the previous remote diagnosis task node; Debugging the initial fault diagnosis analysis algorithm by using the algorithm debugging data set to obtain the debugged target fault diagnosis analysis algorithm; The debugging the initial fault diagnosis analysis algorithm by using the algorithm debugging data set includes: The u-th debugging of the initial fault diagnosis analysis algorithm is performed through the following steps; where u is a positive integer not less than 1, and when u is 1, the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging is the initial fault diagnosis analysis algorithm that has not passed the debugging: Select the u-th set of linked fault diagnosis knowledge representation example sets corresponding to the u-th set of vehicle networking service system examples in the algorithm debugging data set; Input the u-th set of linked fault diagnosis knowledge representation example sets into the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging, and obtain the u-th set of fault discrimination confidence levels determined by the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging based on the u-th set of linked fault diagnosis knowledge representation sets; Based on the u-th set of fault discrimination confidence levels and the prior discrimination annotations of the u-th set of vehicle networking service system examples, determine the value of the training error of the initial fault diagnosis analysis algorithm in the u-th debugging; In the case where the value of the training error in the u-th debugging does not meet the set training requirements, update the neural network weights in a layer of multi-way decision tree branches in the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging to obtain the fault diagnosis analysis algorithm obtained from the u-th debugging; The step of inputting the u-th set of linked fault diagnosis knowledge representation example sets into the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging and obtaining the u-th set of fault discrimination confidence levels determined by the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging based on the u-th set of linked fault diagnosis knowledge representation sets includes: Input the u-th set of linked fault diagnosis knowledge representation example sets into a layer of multi-way decision tree branches in the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging to obtain the processed linked fault diagnosis knowledge representation generated by the last multi-way decision tree branch in the layer of multi-way decision tree branches; Input the processed linked fault diagnosis knowledge representation into the discrimination activation module in the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging to obtain the u-th set of fault discrimination confidence levels; where the fault diagnosis analysis algorithm obtained from the (u - 1)-th debugging includes the layer of multi-way decision tree branches and the discrimination activation module.

8. The method according to any one of claims 1 to 7, characterized in that, The method further includes: in the case where the fault discrimination confidence level is greater than the set confidence level, reporting the fault discrimination confidence level to the vehicle networking maintenance platform server and giving a fault warning to the target vehicle networking service system; The step of reporting the fault discrimination confidence level to the vehicle networking maintenance platform server and giving a fault warning to the target vehicle networking service system includes: Determine the target vehicle networking service system as a faulty vehicle networking service system through the fault calibration thread of the vehicle networking maintenance platform server; Send a warning prompt or start a maintenance task to the target vehicle networking service system through the fault warning thread of the vehicle networking maintenance platform server; where the maintenance task is used to adjust the execution strategy of network fault diagnosis of the target vehicle networking service system.

9. A network fault diagnosis system, characterized in that, It includes a processor and a memory; the processor and the memory are communicatively connected, and the processor is configured to read and execute a computer program from the memory to implement the method according to any one of claims 1-8 above.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and the computer program implements the method according to any one of claims 1-8 above when running.

Images