A remote vehicle fault diagnosis method based on the Internet of Vehicles

Abstract

The present invention relates to the field of fault diagnosis technology, specifically disclosing a method for remote vehicle fault diagnosis based on the Internet of Vehicles. The method comprises the following steps: setting edge nodes based on the condition scores of collection points, marking monitored road sections based on the average condition scores; determining a target vehicle based on the distance between the vehicle and the edge node, and obtaining standard data and pending data for the target vehicle; determining whether the target vehicle has a fault based on the pending data and standard data; and, if a fault exists, the cloud platform provides the corresponding fault cause and treatment strategy based on the pending data. The present invention balances real-time diagnosis with reduced system load.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a vehicle fault remote diagnosis method based on the Internet of Vehicles. Background Art

[0002] The Internet of Vehicles (IoV) refers to an intelligent transportation network system built by interconnecting vehicles, roadside facilities, traffic management centers and other smart devices through advanced technologies such as wireless communications, sensors, cloud computing and big data. Based on the powerful data transmission and processing capabilities of the IoV, remote diagnosis technology of vehicle faults can be realized.

[0003] Current vehicle fault diagnosis methods primarily rely on real-time monitoring and periodic testing. While real-time monitoring can instantly capture changes in a vehicle's operating status, the data transmission and processing process generates a large amount of real-time data, leading to a sharp increase in system load and easily causing response delays and information congestion. Furthermore, frequent data exchange poses a severe challenge to onboard hardware and network resources. In contrast, periodic testing alleviates the system's processing pressure to a certain extent, but due to its long detection cycle, it lacks sufficient real-time performance and cannot promptly reflect sudden abnormal conditions during vehicle operation. Furthermore, periodic testing ignores the impact of environmental factors on the vehicle. For example, in poor road conditions, the vehicle may suffer damage such as chassis scratches, and such external impacts are often difficult to effectively reflect in a timely manner through intermittent data collection. Summary of the Invention

[0004] The purpose of the present invention is to provide a vehicle fault remote diagnosis method based on the Internet of Vehicles to solve the above technical problems.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A vehicle fault remote diagnosis method based on the Internet of Vehicles includes the following steps:

[0007] Setting a collection point at a preset location on a road, controlling a drone to collect images of the road above the collection point, determining a condition score of the collection point based on the image, and setting an edge node on the road based on the condition score;

[0008] Marking a road between two adjacent edge nodes as a sub-segment, and marking a sub-segment whose average condition score is less than a preset value as a monitoring segment;

[0009] The two edge nodes corresponding to the monitored road section are marked as edge node B and edge node B' according to the road traffic direction, with the road traffic direction being from edge node B to edge node B'; the distance between the vehicle and the edge node B is periodically obtained, and the distance is sorted in chronological order. When the sorting is monotonically decreasing, the corresponding vehicle is marked as a target vehicle;

[0010] Collecting the operating data of the target vehicle within a time period of t before entering the monitoring section BDY corresponding to the edge node B, and recording it as standard data, where t is a preset time period; when the target vehicle reaches the edge node B', collecting the operating data of the target vehicle at this time and marking it as pending data; the target vehicle transmits the pending data and the standard data to the edge node B;

[0011] The edge node B determines whether the target vehicle has a fault based on the pending data and the standard data. When a fault exists, the edge node B uploads the pending data to the cloud platform, and the cloud platform gives the corresponding fault cause and processing strategy based on the pending data.

[0012] As a further solution of the present invention, determining the condition score of the acquisition point based on the image includes:

[0013] Obtaining the height information of any point on the road in the image, calculating the height difference between any two points, and determining the maximum height difference Y1 and the average height difference Y2;

[0014] Calculating Condition Score ,η is the preset correction coefficient.

[0015] As a further solution of the present invention, setting edge nodes on the road based on the condition score includes:

[0016] Step 1: Mark the road between two adjacent forks as a route segment, mark the two endpoints of the route segment as endpoint A1 and endpoint A2 respectively, and the road traffic direction of the route segment is from endpoint A1 to endpoint A2. Sort the collection points in ascending order according to the distance between them and endpoint A1 to obtain a first sorting;

[0017] Step 2: Obtain the sorting position D of the collection point corresponding to the maximum condition score in a single route segment in the first sorting, determine the first position D1 and the second position D2 in the first sorting based on preset constraints, and set edge nodes at the first position and the second position;

[0018] Step 3: Remove the collection points between the first position D1 and the second position D2 to obtain a new first sorting, repeat steps 1 and 2, and set new edge nodes.

[0019] As a further solution of the present invention: the first position D1 satisfies the following constraints:

[0020] C D -C i ≤ΔC, i∈[D1, D-1], C i represents the condition score corresponding to the i-th collection point in the first ranking, and ΔC represents a preset condition score difference;

[0021] The first position D2 satisfies the following constraints:

[0022] C j -C D ≤ΔC,j∈[D+1,D2].

[0023] As a further solution of the present invention: the cloud platform provides corresponding fault causes and processing strategies based on the pending data, including:

[0024] Obtaining an average value of the standard data of a single type, and determining that the target vehicle has a fault when a difference between the pending data of a certain type and the corresponding average value is greater than a preset difference threshold;

[0025] The cloud platform is configured with a pre-trained fault identification model. The pending data is input into the fault identification model, and the fault identification model outputs the corresponding fault cause and processing strategy.

[0026] As a further solution of the present invention: obtaining a pre-trained fault recognition model includes:

[0027] Establishing a data set, wherein the data set stores data samples, wherein each data sample is the operating data annotated with a fault cause and a processing strategy, and dividing the data set into a training set and a validation set according to a preset ratio;

[0028] A fault recognition model is established based on deep learning, and the fault recognition model is trained and verified based on the training set and the verification set to obtain a pre-trained fault recognition model.

[0029] As a further solution of the present invention: when the confidence level of the fault cause and processing strategy given by the fault identification model is less than a preset confidence threshold, it is sent to a preset management personnel for error reporting.

[0030] As a further solution of the present invention: in step 3, when the distance between the new edge node and any determined edge node B1 is less than a preset distance threshold, the edge node is not set, and the new edge node is replaced by the edge node B1.

[0031] Beneficial effects of the present invention:

[0032] 1) This invention uses road condition scores to filter vehicles, performing relatively intensive monitoring only on monitored sections with poor road conditions. It also uses a target vehicle identification strategy to prioritize vehicles that are truly at risk of failure, significantly reducing unnecessary data collection and transmission. This balances timely detection with efficient use of system resources.

[0033] 2) Within the monitored road section, edge nodes perform preliminary fault assessments on the operational data uploaded by vehicles, and only upload abnormal or suspicious pending data to the cloud, achieving an effective division of labor between the edge and cloud. This layered architecture not only reduces the cloud platform's burden on processing massive amounts of data, but also significantly reduces network bandwidth usage, fundamentally avoiding network congestion and response delays caused by global real-time data synchronization.

[0034] 3) Relying on imagery collected by drones, the system uses road height differences and scores to identify areas with poor road conditions and deploy edge nodes. This allows for precise demarcation of monitored sections. This allows for earlier detection of potential failures caused by road impacts or scratches when a vehicle is on such high-risk road sections. In response to road repairs or changes in traffic conditions, the system can dynamically adjust based on the latest scores, avoiding the waste of resources caused by redundant edge node deployment while maintaining effective monitoring coverage.

[0035] 4) When the edge node detects signs of a vehicle failure, it uploads pending data to the cloud platform, where a pre-trained fault identification model performs further accurate diagnosis. This leverages the cloud's high computing power and big data models to provide a more comprehensive and accurate fault cause analysis and treatment strategy without overloading the cloud platform.

[0036] In summary, the present invention significantly reduces the pressure of data uploading and processing while taking into account the timeliness of detection, and uses the collaboration between edge nodes and cloud platforms to achieve efficient, reliable and scalable remote diagnosis of vehicle faults, providing a more targeted and effective solution for vehicle management and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The present invention will be further described below with reference to the accompanying drawings.

[0038] Figure 1 It is a flow chart of a vehicle fault remote diagnosis method based on the Internet of Vehicles of the present invention. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0040] See also Figure 1 As shown, the present invention is a vehicle fault remote diagnosis method based on the Internet of Vehicles, comprising the following steps:

[0041] Setting a collection point at a preset location on a road, controlling a drone to collect images of the road above the collection point, determining a condition score of the collection point based on the image, and setting an edge node on the road based on the condition score;

[0042] In a preferred embodiment of the present invention, determining the condition score of the acquisition point based on the image includes:

[0043] Obtaining the height information of any point on the road in the image, calculating the height difference between any two points, and determining the maximum height difference Y1 and the average height difference Y2;

[0044] Calculating Condition Score , η is the preset correction coefficient;

[0045] In another preferred embodiment of the present invention, setting edge nodes on the road based on the condition score includes:

[0046] Step 1: Mark the road between two adjacent forks as a route segment, mark the two endpoints of the route segment as endpoint A1 and endpoint A2 respectively, and the road traffic direction of the route segment is from endpoint A1 to endpoint A2. Sort the collection points in ascending order according to the distance between them and endpoint A1 to obtain a first sorting;

[0047] Step 2: Obtain the sorting position D of the collection point corresponding to the maximum condition score in a single route segment in the first sorting, determine the first position D1 and the second position D2 in the first sorting based on preset constraints, and set edge nodes at the first position and the second position;

[0048] Step 3: Remove the collection points between the first position D1 and the second position D2 to obtain a new first sorting, repeat steps 1 and 2, and set a new edge node;

[0049] In a preferred embodiment of this invention, the first position D1 satisfies the following constraints:

[0050] C D -C i ≤ΔC, i∈[D1, D-1], C i represents the condition score corresponding to the i-th collection point in the first ranking, and ΔC represents a preset condition score difference;

[0051] The first position D2 satisfies the following constraints:

[0052] C j -C D ≤ΔC, j∈[D+1, D2];

[0053] It is worth noting that a suitable collection point layout can be selected based on the actual characteristics of the road (such as straightness, traffic flow, historical fault data, etc.), and auxiliary signs can be set on the ground or coordinates can be preset in the UAV navigation system to ensure that the UAV can accurately reach the designated location and hover stably for shooting. During the shooting process, the UAV can use multi-view image acquisition or carry a laser ranging device to capture detailed information covering the road surface and a certain range around it; then, the height distribution data of the target area can be generated using image fusion or three-dimensional reconstruction algorithms, and the possible depressions, bulges or large fluctuations on the road can be portrayed by comparing the height difference between any two points, and the maximum height difference Y1 and the average height difference Y2 are extracted as core indicators. The maximum height difference Y1 is usually used to measure whether there are more significant protrusions or potholes in the local area, while the average height difference Y2 can reflect the overall degree of fluctuation in the area. After combining these two indicators with the pre-set correction coefficient η, the maximum height difference Y1 is used to measure whether there are more significant protrusions or potholes in the local area, and the average height difference Y2 can reflect the overall degree of fluctuation in the area. After combining these two indicators with the pre-set correction coefficient η, the maximum height difference Y1 is used to measure whether there are more significant protrusions or potholes in the local area, and the average height difference Y2 can reflect the overall degree of fluctuation in the area. Obtain condition scores for collection points to quantify the impact that roads may have on vehicles. After scoring all collection points, we can identify which road sections are more likely to generate significant load impacts on vehicle chassis, safety systems, or suspension components based on the scoring results. These high-scoring areas can then be designated as key monitoring targets, with edge nodes deployed around them.

[0054] In another preferred embodiment of this embodiment, in step 3, when the distance between the new edge node and any determined edge node B1 is less than a preset distance threshold, the edge node is not set, and the new edge node is replaced by the edge node B1.

[0055] It is understandable that if the newly planned edge node is too close to the existing node, it will be replaced by the existing node as appropriate, so as to balance resource utilization efficiency and monitoring coverage;

[0056] Marking a road between two adjacent edge nodes as a sub-segment, and marking a sub-segment whose average condition score is less than a preset value as a monitoring segment;

[0057] The two edge nodes corresponding to the monitored road section are marked as edge node B and edge node B' according to the road traffic direction, with the road traffic direction being from edge node B to edge node B'; the distance between the vehicle and the edge node B is periodically obtained, and the distance is sorted in chronological order. When the sorting is monotonically decreasing, the corresponding vehicle is marked as a target vehicle;

[0058] Collecting the operating data of the target vehicle within a time period of t before entering the monitoring section BDY corresponding to the edge node B, and recording it as standard data, where t is a preset time period; when the target vehicle reaches the edge node B', collecting the operating data of the target vehicle at this time and marking it as pending data; the target vehicle transmits the pending data and the standard data to the edge node B;

[0059] To promptly detect vehicles that may be experiencing anomalies within the monitored section, the distance traveled between the vehicle and edge node B is collected at regular intervals. This distance data is arranged in chronological order and continuously updated. If the distance from the vehicle to node B shows a monotonically decreasing trend over multiple consecutive collections—that is, each measured distance is smaller than the previous one—the vehicle is determined to be rapidly approaching the endpoint of the monitored section and is marked as a target vehicle. Next, to more accurately analyze the vehicle's operating status before and after it passes through the monitored section, sensor data such as engine speed, speed, acceleration, temperature, and pressure (i.e., operating data) is continuously collected for a preset time period t before the vehicle enters the monitored section. The data aggregated during this period is defined as standard data. When the target vehicle is about to reach node B', the vehicle's operating status at that time is obtained and generated as pending data. The vehicle's onboard terminal then transmits this pending data along with the previously generated standard data to edge node B. Through this periodic distance monitoring and data extraction at key moments, the edge node can compare the vehicle's status during normal driving with that when it approaches the fault-risk section to determine whether a fault has occurred within the monitored section.

[0060] The edge node B determines whether the target vehicle has a fault based on the pending data and the standard data. When a fault occurs, the edge node B uploads the pending data to the cloud platform, and the cloud platform provides a corresponding fault cause and processing strategy based on the pending data;

[0061] In a preferred embodiment of the present invention, the cloud platform provides corresponding fault causes and processing strategies based on the pending data, including:

[0062] Obtaining an average value of the standard data of a single type, and determining that the target vehicle has a fault when a difference between the pending data of a certain type and the corresponding average value is greater than a preset difference threshold;

[0063] The cloud platform is configured with a pre-trained fault identification model, the pending data is input into the fault identification model, and the fault identification model outputs the corresponding fault cause and processing strategy;

[0064] In a preferred embodiment of this embodiment, obtaining a pre-trained fault recognition model includes:

[0065] Establishing a data set, wherein the data set stores data samples, wherein each data sample is the operating data annotated with a fault cause and a processing strategy, and dividing the data set into a training set and a validation set according to a preset ratio;

[0066] Establishing a fault recognition model based on deep learning, and training and verifying the fault recognition model based on the training set and the verification set to obtain a pre-trained fault recognition model;

[0067] In another preferred embodiment of the present invention, when the confidence level of the fault cause and processing strategy given by the fault identification model is less than a preset confidence threshold, an error message is sent to a preset management personnel for prompting;

[0068] Notably, the cloud platform first compares the pending data with its corresponding standard data. If the deviation of a particular category of data from the standard data average exceeds a preset threshold, the vehicle is preliminarily identified as having an anomaly. Furthermore, to more accurately locate the cause of the fault and generate a corresponding strategy, the pending data is fed into a pre-deployed fault identification model on the cloud platform. This model, constructed using deep learning methods and trained and validated on a sufficient number of labeled samples (including fault causes and solutions), is capable of identifying both common and specific fault patterns. If the model's output is highly reliable, it directly identifies the specific fault type and recommends a repair or resolution plan. Conversely, if the model's confidence in the pending data falls below a preset threshold, the system uploads the relevant information to management for manual review or secondary confirmation, thus avoiding excessive false positives or missed detections. This layered diagnostic strategy, combining edge nodes with cloud-based models, reduces the cloud's burden on indiscriminately processing massive amounts of real-time data while leveraging the cloud's big data and high-performance computing capabilities when necessary, providing rapid and accurate analysis and response plans for vehicle fault diagnosis.

[0069] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A vehicle fault remote diagnosis method based on the Internet of Vehicles, characterized in that: The following steps are involved: Setting a collection point at a preset location on a road, controlling a drone to collect images of the road above the collection point, determining a condition score of the collection point based on the image, and setting an edge node on the road based on the condition score; Marking a road between two adjacent edge nodes as a sub-segment, and marking a sub-segment whose average condition score is less than a preset value as a monitoring segment; The two edge nodes corresponding to the monitored road section are marked as edge node B and edge node B' according to the road traffic direction, with the road traffic direction being from edge node B to edge node B'; the distance between the vehicle and the edge node B is periodically obtained, and the distance is sorted in chronological order. When the sorting is monotonically decreasing, the corresponding vehicle is marked as a target vehicle; Collect the running data of the target vehicle within t time before entering the monitoring section BDY corresponding to the edge node B, and record it as standard data, where t is the preset time; When the target vehicle arrives at the edge node B', the operating data of the target vehicle at this time is collected and marked as pending data; the target vehicle transmits the pending data and the standard data to the edge node B; The edge node B determines whether the target vehicle has a fault based on the pending data and the standard data. When a fault occurs, the edge node B uploads the pending data to the cloud platform, and the cloud platform provides a corresponding fault cause and processing strategy based on the pending data; Setting edge nodes on the road based on the condition score includes: Step 1: Mark the road between two adjacent forks as a route segment, mark the two endpoints of the route segment as endpoint A1 and endpoint A2 respectively, and the road traffic direction of the route segment is from endpoint A1 to endpoint A2. Sort the collection points in ascending order according to the distance between them and endpoint A1 to obtain a first sorting; Step 2: Obtain the sorting position D of the collection point corresponding to the maximum condition score in a single route segment in the first sorting, determine the first position D1 and the second position D2 in the first sorting based on preset constraints, and set edge nodes at the first position and the second position; Step 3: Remove the collection points between the first position D1 and the second position D2 to obtain a new first sorting, repeat steps 1 and 2, and set a new edge node; The first position D1 satisfies the following constraints: C D -C i ≤ΔC, i∈[D1, D-1], C i represents the condition score corresponding to the i-th collection point in the first ranking, and ΔC represents a preset condition score difference; The first position D2 satisfies the following constraints: C j -C D ≤ΔC,j∈[D+1,D2]。 2. A vehicle fault remote diagnosis method based on the Internet of Vehicles according to claim 1, characterized in that: Determining the condition score of the collection point based on the image includes: Obtaining the height information of any point on the road in the image, calculating the height difference between any two points, and determining the maximum height difference Y1 and the average height difference Y2; Calculating Condition Score , η is the preset correction coefficient.

3. The vehicle fault remote diagnosis method based on the Internet of Vehicles according to claim 1, characterized in that: The cloud platform provides corresponding fault causes and processing strategies based on the pending data, including: Obtaining an average value of the standard data of a single type, and determining that the target vehicle has a fault when a difference between the pending data of a certain type and the corresponding average value is greater than a preset difference threshold; The cloud platform is configured with a pre-trained fault identification model. The pending data is input into the fault identification model, and the fault identification model outputs the corresponding fault cause and processing strategy.

4. The vehicle fault remote diagnosis method based on the Internet of Vehicles according to claim 3 is characterized in that: Obtaining pre-trained fault recognition models includes: Establishing a data set, wherein the data set stores data samples, wherein each data sample is the operating data annotated with a fault cause and a processing strategy, and dividing the data set into a training set and a validation set according to a preset ratio; A fault recognition model is established based on deep learning, and the fault recognition model is trained and verified based on the training set and the verification set to obtain a pre-trained fault recognition model.

5. The vehicle fault remote diagnosis method based on the Internet of Vehicles according to claim 3 is characterized in that: When the confidence level of the fault cause and processing strategy given by the fault identification model is less than a preset confidence threshold, it is sent to a preset management personnel for error reporting.

6. The vehicle fault remote diagnosis method based on the Internet of Vehicles according to claim 1, characterized in that: In step 3, when the distance between the new edge node and any determined edge node B1 is less than a preset distance threshold, the edge node is not set, and the new edge node is replaced by the edge node B1.

Images