Intelligent vehicle diagnosis method and system based on cloud edge cooperation

Abstract

The invention discloses an intelligent vehicle diagnosis method and system based on cloud-side cooperation, and relates to the field of unmanned vehicles, and the method comprises a collection module which is used for collecting multi-source heterogeneous data of a vehicle operation state, environment perception and a core part working condition, and uploading the data to an edge node and a cloud after the data is preliminarily preprocessed by an edge side; the calculation module is used for acquiring the preprocessed data, executing vehicle collaborative perception calculation and local dynamic path planning in the area, generating an edge side decision instruction and issuing the edge side decision instruction to the vehicle-mounted end; according to the method, the residual life and the fault risk of the component are accurately pre-judged, early warning information is pushed in a graded mode, fault loss is avoided in advance, cross-regional transport capacity supply and demand matching is optimized, vehicle resources are reasonably scheduled, meanwhile, the optimal operation and maintenance resources are rapidly adapted, operation and maintenance demands are responded according to demand priorities, the response time limit is shortened, continuous and stable operation of the vehicle is guaranteed, and the service life of the vehicle is prolonged. And the operation and maintenance cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of unmanned vehicle technology, specifically to a vehicle intelligent diagnostic method and system based on cloud-edge collaboration. Background Technology

[0002] Unmanned vehicles are autonomous vehicles equipped with multiple sensors such as lidar and cameras, as well as intelligent decision-making and control systems. They can autonomously perceive surrounding road conditions, plan driving routes, and control vehicle start-stop and steering without human operation. They can be adapted to various scenarios such as urban roads, parks, and logistics parks, and are widely used in scenarios such as travel services and material delivery.

[0003] Patent application No. 202011328849.3 discloses a high-end equipment fault intelligent diagnosis system based on edge-cloud collaboration. This application aims to address the problem that "current intelligent fault diagnosis methods are mostly geared towards individual equipment or their components, such as ships, windmills, machine tool spindles, rotating machinery, and bearings, and lack fault diagnosis methods for high-end equipment composed of multiple devices. Current intelligent fault diagnosis methods generally involve transmitting data to a server and using various algorithms to diagnose and analyze the fault. This method suffers from real-time fault diagnosis issues; the models are generally large, and server calculations typically require a long time to return results. Furthermore, servers are generally located far from the equipment, leading to practical application problems in data transmission and real-time fault processing. Moreover, existing methods cannot adapt to new situations promptly, lack evolutionary capabilities, and their accuracy, intelligence, and generalization will decrease over time."

[0004] However, for unmanned logistics fleets, current unmanned logistics fleet management back-ends are mostly monitoring-based, and cannot be combined with the back-end to build an intelligent control and fault prediction operation and maintenance closed loop for unmanned logistics fleets based on a cloud-edge-device collaborative architecture, so as to avoid the problems of information fragmentation at multiple nodes, delayed fault response and insufficient operation and maintenance efficiency.

[0005] To address this, we propose a cloud-edge collaborative intelligent vehicle diagnostic method and system. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a vehicle intelligent diagnostic method and system based on cloud-edge collaboration, which can effectively solve the problems of the existing technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions; This invention discloses a cloud-edge collaborative intelligent vehicle diagnostic system, comprising: The data acquisition module collects multi-source heterogeneous data on vehicle operating status, environmental perception, and core component operating conditions. After preliminary preprocessing at the edge, the data is uploaded to edge nodes and the cloud. The computing module acquires the preprocessed data, performs vehicle collaborative perception computing and local dynamic path planning within the region, generates edge-side decision commands, and sends them to the vehicle. The central module collects global data from each edge node, generates cross-regional capacity scheduling commands, and sends them to the edge nodes. The evaluation module builds digital twin models of components based on multi-source data from the cloud, edge, and terminal, predicts the remaining lifespan and failure risk of components, generates tiered early warning information, and pushes it to the generation module. The generation module receives fault early warning information, analyzes standardized maintenance requirements, matches optimal maintenance resources, generates maintenance work orders, and pushes them synchronously to the corresponding maintenance execution nodes. The simulation module dynamically schedules cloud and edge computing resources according to task priority. The acquisition module is interconnected with the computing module via a wireless network. The computing module is interconnected with the central module via a wireless network. The central module is interconnected with the evaluation module via a wireless network. The evaluation module is interconnected with the generation module and the simulation module via a wireless network.

[0008] Furthermore, the multi-source heterogeneous data of the acquisition module includes vehicle operating status data, environmental perception data, and core component operating condition data; Among them, vehicle operating status data includes at least vehicle speed, acceleration, braking frequency, steering angle, and remaining battery power; environmental perception data includes at least road slope, road surface friction coefficient, real-time traffic flow, ambient temperature, and visibility; and core component operating condition data includes at least engine speed, transmission oil temperature, braking system pressure, battery cycle count, and motor winding temperature. The initial preprocessing at the edge includes data denoising, spatiotemporal alignment, and normalization. Data denoising removes interfering data through sliding window mean filtering. Spatiotemporal alignment synchronizes the data acquisition time and spatial location of each sensor based on timestamps. Normalization maps various types of data to a preset unified dimension range. The pre-processed data is prioritized according to the basic order of core component operating condition data, vehicle operating status data, and environmental perception data. The upload order is dynamically adjusted based on the real-time network bandwidth status of edge nodes and the cloud. When the network bandwidth is lower than a preset threshold, abnormal fluctuation data in the core component operating condition data is uploaded first. Abnormal fluctuation data is identified by comparing the data fluctuation amplitude with a preset threshold.

[0009] Furthermore, when the computing module performs vehicle collaborative perception calculations within the area, it generates reliable perception results through cross-node verification of vehicle data, and its collaborative perception confidence level is: ; In the local dynamic path planning stage, the optimal path is determined based on the collaborative sensing results and the real-time path cost of the inherent paths within the path planning area. The path with the lowest cost and the collaborative sensing results are selected for generating edge-side decision instructions. ; In the formula: To collaboratively perceive confidence; The number of vehicles participating in collaborative sensing within the region; Let be the perception weight coefficient for the i-th vehicle; Let be the single-source perception confidence level of the i-th vehicle; Let be the distance between the i-th vehicle and the center of the target monitoring area; The average distance from all participating collaborative sensing vehicles within the region to the center of the target monitoring area; This represents the path cost value. , , Optimize the weight coefficients for the path; The estimated energy consumption for the target path; The average estimated energy consumption of all feasible paths with the same start and end point within the region; The estimated travel time for the target route; It represents the average estimated travel time for all feasible paths with the same origin and destination within the region.

[0010] Furthermore, the global data collected by the central module from each edge node includes full order data, full vehicle status data, real-time traffic fusion data, regional transportation capacity characteristic data, and environmental perception summary data. In the cross-regional capacity scheduling instruction generation stage, the global data is first standardized, deduplicated, and filtered for outliers to obtain a fused dataset. Then, the regional capacity supply and demand gap is calculated based on the fused dataset. The optimal cross-regional route is planned and matched with suitable vehicles in combination with real-time road conditions and vehicle status. Scheduling instructions are generated according to order priority.

[0011] Furthermore, the component remaining lifespan and failure risk prediction logic in the evaluation module is as follows: ; In the formula: The remaining lifespan of the component; The rated design life of the component; The number of key operating conditions that affect component lifespan; For the k-th type of operating condition, the life loss weighting coefficient is used. This represents the actual cumulative operating time of the component under the k-th type of operating condition; This represents the rated cumulative operating time of the component under the k-th type of operating condition. This is the environmental stress influence coefficient; The deviation between the real-time operating energy loss and the rated energy loss of the component; This represents the average rated energy loss of the component. This represents the component failure risk value. This is the lifespan offset coefficient; The number of component failure monitoring indicators; Let be the weight of the q-th fault monitoring indicator; This is the real-time monitoring value of the q-th fault monitoring indicator; is the normal operating standard value of the q-th fault monitoring indicator.

[0012] Furthermore, when the generation module parses the standardized operation and maintenance requirements, it constructs a three-dimensional operation and maintenance requirement matrix based on the fault risk value, component importance level, and vehicle task urgency output by the evaluation module. Each dimension of the matrix corresponds to a core requirement indicator. The standardized operation and maintenance requirement vector is obtained through matrix normalization. The component importance level is preset. The matching of operation and maintenance resources is determined by calculating a matching score: ; In the formula: Scoring based on the matching of operation and maintenance resources; , , To match the weight coefficients; The response distance from maintenance resources to the faulty vehicle; The average response distance from all candidate maintenance resources to the faulty vehicle; The service quality level of the operation and maintenance resources; The current load rate of the operation and maintenance resources; The average load rate of all candidate operation and maintenance resources; The generated maintenance work order includes faulty vehicle identification, faulty component information, fault risk level, standardized maintenance requirements, optimal maintenance resource identification, response time limit requirements, and maintenance operation specifications. The work order is synchronously pushed to the corresponding maintenance execution node through encrypted transmission, and is also synchronously backed up to the central module.

[0013] Furthermore, the graded early warning information generated by the assessment module is divided into intervals according to the historical maximum and minimum values ​​of the fault risk value, corresponding to three early warning levels; Level 1 warning corresponds to a high fault risk value, Level 2 warning corresponds to a medium fault risk value, and Level 3 warning corresponds to a low fault risk value. After receiving the tiered early warning information, the generation module adapts the corresponding operation and maintenance response priority according to the early warning level: Level 1 warning triggers an emergency maintenance response, prioritizing the allocation of optimal maintenance resources in idle states and simultaneously pushing real-time alert information to vehicle management personnel; Level 2 warning triggers a routine maintenance response, completing maintenance resource matching and work order push within a preset time; Level 3 warning triggers a preventative maintenance response, entering the maintenance queue after the vehicle's current task is completed. Vehicles in the maintenance queue are sorted according to the trigger time of the Level 3 warning and maintenance operations are performed in sequence.

[0014] Furthermore, when the simulation module dynamically schedules cloud-edge computing resources, it allocates computing power based on task priority, data processing complexity, and the current load status of the nodes: ; In the formula: Allocate computing power to the j-th cloud edge node; The total available computing power of the cloud-edge collaborative system; This represents the priority coefficient of the current task. Let be the computing power requirement of the j-th cloud edge node for the current task; Let be the current load rate of the j-th cloud edge node; This represents the total number of cloud-edge nodes participating in computing power scheduling.

[0015] On the other hand, a cloud-edge collaborative vehicle intelligent diagnostic method includes: Multi-source heterogeneous data on vehicle operating status, environmental perception, and core component operating conditions are collected. After denoising, spatiotemporal alignment, and normalization preprocessing, the data is dynamically uploaded to edge nodes and the cloud according to data priority and network bandwidth status. Based on the uploaded data, vehicle collaborative perception calculations within the region are performed to obtain reliable results. Combined with real-time path costs, the optimal path is determined, and edge-side decision instructions are generated and sent to the vehicle terminal. Global data from each edge node is collected and processed through standardization, deduplication, and outlier filtering. The regional capacity supply and demand gap is calculated, and the optimal path is planned. Cross-regional capacity scheduling instructions are generated according to order priority. Digital twin models of components are constructed based on multi-source data from cloud, edge, and terminal. The remaining lifespan and failure risk of components are calculated simultaneously, and graded early warning information is generated. Standardized operation and maintenance requirements are constructed based on graded early warning information and core indicators. The optimal operation and maintenance resources are selected through matching and scoring. Operation and maintenance work orders containing key information are generated and pushed to execution nodes and the central module simultaneously. Cloud and edge computing resources are dynamically scheduled based on task priority, data processing complexity, and node load status.

[0016] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects: This invention achieves dynamic allocation of computing power through cloud-edge collaboration, efficiently processes multi-source heterogeneous data, improves the reliability of perception results and the rationality of path planning, reduces vehicle energy consumption and travel time, accurately predicts the remaining lifespan and failure risk of components, pushes early warning information in a tiered manner, avoids failure losses in advance, optimizes the supply and demand matching of cross-regional transportation capacity, rationally schedules vehicle resources, and quickly adapts to the best maintenance resources, responds to maintenance needs according to demand priority, shortens response time, ensures continuous and stable operation of vehicles, reduces operating and maintenance costs, improves the overall efficiency of transportation and maintenance services, and provides comprehensive support for vehicle operation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0018] Figure 1 This is a schematic diagram of the structure of a cloud-edge collaborative vehicle intelligent diagnostic system; Figure 2 This is a flowchart illustrating a cloud-edge collaborative intelligent vehicle diagnostic method. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] The present invention will be further described below with reference to embodiments.

[0021] Example 1: This embodiment presents a cloud-edge collaborative vehicle intelligent diagnostic system, such as... Figure 1 As shown, it includes: The data acquisition module is used to collect multi-source heterogeneous data on vehicle operating status, environmental perception, and core component operating conditions. After preliminary preprocessing at the edge, the data is uploaded to the edge nodes and the cloud. The multi-source heterogeneous data collected by the acquisition module includes vehicle operating status data, environmental perception data, and core component operating condition data. Among them, vehicle operating status data includes at least vehicle speed, acceleration, braking frequency, steering angle, and remaining battery power; environmental perception data includes at least road slope, road surface friction coefficient, real-time traffic flow, ambient temperature, and visibility; and core component operating condition data includes at least engine speed, transmission oil temperature, braking system pressure, battery cycle count, and motor winding temperature. Preliminary preprocessing at the edge includes data denoising, spatiotemporal alignment, and normalization. Data denoising removes interfering data through sliding window mean filtering. Spatiotemporal alignment synchronizes the data acquisition time and spatial location of each sensor based on timestamps. Normalization maps various types of data to a preset unified dimension range. The pre-processed data is prioritized according to the basic order of core component operating condition data, vehicle operating status data, and environmental perception data. The upload order is dynamically adjusted based on the real-time network bandwidth status of edge nodes and the cloud. When the network bandwidth is lower than a preset threshold, abnormal fluctuation data in the core component operating condition data is uploaded first. Abnormal fluctuation data is identified by comparing the data fluctuation amplitude with a preset threshold. The computing module is used to acquire preprocessed data, perform vehicle cooperative perception computing and local dynamic path planning within the area, generate edge-side decision commands and send them to the vehicle terminal; When the computing module performs collaborative perception calculations for vehicles within a region, it generates reliable perception results through cross-node verification of vehicle data. The confidence level of its collaborative perception is as follows: ; The above formula integrates the number of vehicles participating in collaborative sensing within the region, the sensing weight coefficient of each vehicle, the credibility of single-source sensing, and the distance relationship between the vehicle and the target monitoring area. By weighted averaging combined with the distance attenuation factor, it highlights the sensing contribution of vehicles with high sensor accuracy, accurate historical data, and close proximity to the monitoring area, effectively filtering out interference from low-credibility and distant vehicles, making the collaborative sensing results more in line with actual monitoring needs. In the local dynamic path planning stage, the optimal path is determined based on the collaborative sensing results and the real-time path cost of the inherent paths within the path planning area. The path with the lowest cost is selected and used in conjunction with the collaborative sensing results to generate decision instructions on the edge side. ; In the formula: To collaboratively perceive confidence; The number of vehicles participating in collaborative sensing within the region; Let be the perception weight coefficient for the i-th vehicle; Let be the single-source perception confidence level of the i-th vehicle; Let be the distance between the i-th vehicle and the center of the target monitoring area; The average distance from all participating collaborative sensing vehicles within the region to the center of the target monitoring area; The above formula takes into account the three core factors of the target path: estimated energy consumption, travel time and collaborative perception confidence. By setting path optimization weight coefficients, it integrates the ratio of energy consumption, time and regional average level and the inverse indicator of perception confidence. It considers both the economic efficiency of the path and relies on reliable perception data, thereby ensuring that the selected path can provide reliable support for edge-side decision-making. This represents the path cost value. , , Optimize the weight coefficients for the path; The estimated energy consumption for the target path; The average estimated energy consumption of all feasible paths with the same start and end point within the region; The estimated travel time for the target route; The average estimated travel time for all feasible paths with the same origin and destination within the region; in, ∈(0,1], the higher the accuracy level of the vehicle sensor and the higher the accuracy of historical data, the larger the value will be, and vice versa. It is obtained by inversely mapping the ratio of the standard deviation to the mean of its locally preprocessed data; , , All are positive numbers, and their sum is 1; , All are preset values. Based on the vehicle's current real-time power system efficiency and remaining battery / fuel levels, combined with the slope grade, road friction coefficient, altitude change and path length of each segment of the target path, the energy consumption is accumulated through a piecewise integral vehicle energy consumption physical model. Based on the total mileage of the target path, the real-time traffic flow obtained by the collaborative perception of each road segment, combined with the traffic light timing rules at intersections, the safe speed limit standards for road segments, and the estimated frequency of vehicle start and stop, the result is obtained by calculating the travel time of each road segment and the waiting time at intersections in segments and then summing them up. The central module is used to collect global data from each edge node, generate cross-regional capacity scheduling instructions, and send them to the edge nodes. The central module collects global data from various edge nodes, including full order data, vehicle status data, real-time traffic fusion data, regional transportation capacity characteristic data, and environmental perception summary data. In the cross-regional capacity dispatch instruction generation stage, the global data is first standardized, deduplicated, and filtered for outliers to obtain a fused dataset. Then, the regional capacity supply and demand gap is calculated based on the fused dataset. Combined with real-time road conditions and vehicle status, the optimal cross-regional route is planned and suitable vehicles are matched. Dispatch instructions are generated according to order priority. In the outlier filtering stage, outliers are filtered by comparing them with a preset outlier threshold, and the optimal path is based on... The formula is determined so that the regional transportation capacity supply and demand gap is represented by the difference between the total transportation capacity demand of the orders to be allocated and the effective supply capacity of available vehicles. The total transportation capacity demand of the orders to be allocated is calculated by multiplying the weight / volume of the goods of all orders to be allocated in the region by the transportation mileage. The effective supply capacity of available vehicles is the sum of the product of the upper limit of the single vehicle carrying capacity of the available vehicles in the region that are fault-free and unoccupied and the remaining driving mileage. The assessment module is used to build digital twin models of components based on multi-source data from cloud, edge, and device, predict the remaining lifespan and failure risk of components, generate graded early warning information, and push it to the generation module; The logic for predicting the remaining life of components and failure risk in the assessment module is as follows: ; In the formula: The remaining lifespan of the component; The rated design life of the component; The number of key operating conditions that affect component lifespan; For the k-th type of operating condition, the life loss weighting coefficient is used. This represents the actual cumulative operating time of the component under the k-th type of operating condition; This represents the rated cumulative operating time of the component under the k-th type of operating condition. This is the environmental stress influence coefficient; The deviation between the real-time operating energy loss and the rated energy loss of the component; This represents the average rated energy loss of the component. This represents the component failure risk value. This is the lifespan offset coefficient; The number of component failure monitoring indicators; Let be the weight of the q-th fault monitoring indicator; This is the real-time monitoring value of the q-th fault monitoring indicator; Let q be the standard value under normal operating conditions for the qth fault monitoring indicator; The above formula is based on the rated design life of the component, combined with the life loss weight of each key operating condition and the actual cumulative operating time, and also incorporates the accelerating effect of environmental stress on aging. Through the product and exponential adjustment of multi-dimensional factors, the life loss is accurately quantified, and the remaining life is estimated. On this basis, the deviation of multiple fault monitoring indicators from standard values ​​is integrated, and a fault risk assessment method is constructed in combination with the remaining life to comprehensively and accurately reflect the potential probability of component failure. in, ∈(0,1), and the sum of all working condition dimension coefficients is 1. The working condition that contributes more to the component's life loss takes a larger value, and the working condition that has a smaller impact on life loss takes a smaller value. ∈(0,2], the larger the temperature fluctuation, the higher the humidity, the higher the concentration of corrosive medium in the environment, the stronger the aging acceleration effect on the component, and the smaller the value when the environmental conditions are closer to the rated working environment of the component and the weaker the aging effect. ∈(0.1,1], the larger the error in the remaining life prediction of the component and the more drastic the fluctuation of the operating conditions, the larger the value; the more accurate the remaining life prediction and the more stable the operating conditions, the smaller the value. ∈(0,1), and the sum of the weights of all monitoring indicators is 1. The stronger the correlation with the core failure mode of the component and the higher the probability of failure after the indicator is abnormal, the larger the value is. The weaker the correlation with the failure and the smaller the value is when it only reflects the fluctuation of secondary operating conditions. The generation module is used to receive fault warning information, parse standardized operation and maintenance requirements, match the optimal operation and maintenance resources, generate operation and maintenance work orders, and push them synchronously to the corresponding operation and maintenance execution nodes. When the generation module parses standardized operation and maintenance requirements, it constructs a three-dimensional operation and maintenance requirement matrix based on the fault risk value, component importance level, and vehicle task urgency output by the evaluation module. Each dimension of the matrix corresponds to a core requirement indicator. The standardized operation and maintenance requirement vector is obtained through matrix normalization. The component importance level is preset. Operation and maintenance resource matching is determined by calculating a matching score: ; In the formula: Scoring based on the matching of operation and maintenance resources; , , To match the weight coefficients; The response distance from maintenance resources to the faulty vehicle; The average response distance from all candidate maintenance resources to the faulty vehicle; The service quality level of the operation and maintenance resources; The current load rate of the operation and maintenance resources; The average load rate of all candidate operation and maintenance resources; The above formula combines the response distance, service quality level, and current load status of joint operation and maintenance resources. Based on the response time requirements of the faulty vehicle, the complexity of the fault type, and the overall load level of the candidate operation and maintenance resources, the weight of each evaluation dimension is dynamically adjusted. This not only meets the differentiated needs under different fault scenarios, but also achieves reasonable matching and efficient utilization of operation and maintenance resources, ensuring the timeliness of operation and maintenance services. The generated maintenance work order includes faulty vehicle identification, faulty component information, fault risk level, standardized maintenance requirements, optimal maintenance resource identification, response time limit requirements, and maintenance operation specifications. The work order is synchronously pushed to the corresponding maintenance execution node through encrypted transmission, and is also synchronously backed up to the central module. in, , , The values ​​of are all within the range of (0,1), and the higher the requirement for the timeliness of maintenance response for faulty vehicles, the better. The larger the value, the lower the requirement for response timeliness. The smaller the value, the more complex the fault type and the higher the requirement for professional operation and maintenance. The larger the value, the simpler the fault type and the lower the requirement for professional operation and maintenance. The smaller the value, the higher the overall load level of the candidate operation and maintenance resources, and the more important it is to avoid resource overload. The larger the value, the more sufficient the overall load of the candidate operation and maintenance resources are, and there is no need to focus on load balancing. The smaller the value; The calculation is based on the weighted summation and normalization of the historical operation and maintenance success rate of the operation and maintenance resources and the quantitative coefficients corresponding to the qualification certification level within the near preset period. The weight of the historical operation and maintenance success rate is larger than that of the quantitative coefficient of the qualification certification level. It is represented by the ratio of the number of uncompleted maintenance work orders currently accepted by the maintenance resource to the maximum number of work orders that the resource can accept. The tiered early warning information generated by the assessment module is divided into intervals based on the historical maximum and minimum values ​​of the fault risk value, corresponding to three early warning levels; Level 1 warning corresponds to a high fault risk value, Level 2 warning corresponds to a medium fault risk value, and Level 3 warning corresponds to a low fault risk value. After receiving the tiered early warning information, the generation module adapts the corresponding operation and maintenance response priority according to the early warning level: Level 1 warning triggers emergency maintenance response, prioritizing the allocation of optimal maintenance resources in idle state and simultaneously pushing real-time alert information to vehicle management personnel; Level 2 warning triggers routine maintenance response, completing maintenance resource matching and work order push within a preset time; Level 3 warning triggers preventive maintenance response, after the vehicle's current task is completed, it enters the maintenance queue, and vehicles in the maintenance queue are sorted according to the trigger time of Level 3 warning, and maintenance operations are performed in sequence. The simulation module is used to dynamically schedule cloud-edge computing resources based on task priority; When the simulation module dynamically schedules cloud-edge computing resources, it allocates computing power based on task priority, data processing complexity, and the current load status of the nodes. ; In the formula: Allocate computing power to the j-th cloud edge node; The total available computing power of the cloud-edge collaborative system; This represents the priority coefficient of the current task. Let be the computing power requirement of the j-th cloud edge node for the current task; Let be the current load rate of the j-th cloud edge node; The total number of cloud-edge nodes participating in computing power scheduling; The above formula is based on the total available computing power of the cloud-edge collaborative system. It combines the urgency of the current task, the computing power demand of each cloud-edge node, and the current load status. By weighting the allocation ratio of each node, computing power resources are preferentially tilted to nodes with high task priority, large computing power demand, and low current load. Ultimately, it realizes the dynamic optimization and scheduling of cloud-edge computing power resources, ensuring the efficiency and reliability of system data processing and decision execution. in, ∈(0,1], the higher the urgency of the corresponding deduction task, the larger the value; the lower the urgency of the corresponding deduction task, the smaller the value. ∈(0, The larger the amount of task data that the corresponding node needs to process or the higher the complexity of the scenario, the larger the value; the smaller the amount of inference task data that the corresponding node needs to process or the lower the complexity of the inference scenario, the smaller the value. The acquisition module is interconnected with the computing module via a wireless network. The computing module is interconnected with the central module via a wireless network. The central module is interconnected with the evaluation module via a wireless network. The evaluation module is interconnected with the generation module and the simulation module via a wireless network.

[0022] In this embodiment, the acquisition module collects multi-source heterogeneous data on vehicle operating status, environmental perception, and core component operating conditions. After preliminary preprocessing at the edge, the data is uploaded to edge nodes and the cloud. The computing module simultaneously acquires the preprocessed data, performs vehicle collaborative perception computing and local dynamic path planning within the region, generates edge-side decision instructions, and sends them to the vehicle terminal. The central module runs in the background, collects global data from each edge node, generates cross-regional capacity scheduling instructions, and sends them to the edge nodes. The evaluation module further constructs a component digital twin model based on multi-source data from the cloud, edge, and terminal, predicts the component's remaining lifespan and failure risk, generates graded early warning information, and pushes it to the generation module. The generation module then receives the failure early warning information, analyzes standardized operation and maintenance requirements, matches the optimal operation and maintenance resources, generates operation and maintenance work orders, and pushes them synchronously to the corresponding operation and maintenance execution nodes. Finally, the simulation module dynamically schedules cloud and edge computing resources according to task priority.

[0023] In the above embodiments, the system can accurately capture key information related to vehicle operation, environment and components, optimize route planning to reduce energy consumption and travel time, intelligently schedule cross-regional transportation resources, predict component failure risks in advance and quickly match and adapt maintenance forces, and reasonably allocate computing power to support efficient operation, effectively improving vehicle operation efficiency, driving safety and maintenance response timeliness, and reducing failure losses and overall operating costs.

[0024] Application example: A logistics company in XX city has adopted this system for the operation and management of 20 new energy logistics vehicles.

[0025] During vehicle operation, the data acquisition module collects multi-source data in real time: operating status data (vehicle speed 60km / h, acceleration 0.3m / s², braking frequency 2 times / km, remaining battery power 75%), environmental perception data (road gradient 3°, road surface friction coefficient 0.7, real-time traffic flow 20 vehicles / km, ambient temperature 25℃, visibility 10km), and core component operating condition data (engine speed 2500rpm, transmission oil temperature 85℃, braking system pressure 0.6MPa, battery cycle count 320 times, motor winding temperature 62℃). After edge-side noise reduction, spatiotemporal alignment, and normalization preprocessing, and because the network bandwidth meets the preset threshold, the data is uploaded to the edge node and the cloud in the order of core component operating condition data, vehicle operating status data, and environmental perception data.

[0026] After acquiring the preprocessed data, the calculation module selects three nearby logistics vehicles within the area to participate in collaborative perception. Cross-node verification yields a collaborative perception confidence level of 0.92. Combining this confidence level with the real-time cost calculation of each feasible path, a certain path is determined to have an estimated energy consumption 10% lower than the average for vehicles with the same origin and destination, an estimated travel time 8% shorter, and a final path cost of 0.85. This path is then identified as the optimal path, and a decision command is generated and sent to the vehicle's onboard unit to guide the vehicle's movement.

[0027] The central module collects global data from each edge node, including three pending orders in region A (total transport capacity requirement of 120 tons). (km), 5 available vehicles (effective supply capacity 150 tons) Information such as distance (km) and real-time road conditions was collected. After data processing, a capacity shortage of 30 tons was calculated. Based on the kilometers and road conditions, two suitable vehicles are matched, and cross-regional dispatch instructions are generated and sent to edge nodes according to order priority.

[0028] The evaluation module constructs a digital twin model of the engine. Based on its rated design life of 1 million kilometers and data from three key operating conditions, it calculates the remaining life of 650,000 kilometers. Combining data from 12 fault monitoring indicators, it calculates a fault risk value of 0.12, determines it as a level three warning, and pushes it to the generation module.

[0029] The generation module constructs a three-dimensional maintenance requirement matrix based on early warning information, engine importance level, and vehicle task urgency, and processes it in a standardized manner. The optimal maintenance point is selected through matching scores (response distance 20% closer to average, service quality level 0.95, load rate 15% lower than average, matching score 0.91). A work order is generated containing the vehicle number (WL-017), faulty component (engine), risk level (Level 3), maintenance requirements, maintenance point number (YW-08), response time limit (within 24 hours after the current task ends), and operation instructions. This work order is encrypted and pushed to the maintenance node and synchronized with the central module backup.

[0030] The simulation module, for the current engine diagnostic task (priority coefficient 0.8), combined with the computing power requirement of a certain edge node (0.3), load rate (0.2), and the total available computing power of the system (100 units), allocates 35 units of computing power to the node to ensure efficient task processing.

[0031] Example 2: At the implementation level, based on Example 1, this example refers to... Figure 2 A further detailed description of the cloud-edge collaborative vehicle intelligent diagnostic system in Example 1 is provided below: A cloud-edge collaborative intelligent vehicle diagnostic method includes: Multi-source heterogeneous data on vehicle operating status, environmental perception, and core component operating conditions are collected, and after preprocessing such as noise reduction, spatiotemporal alignment, and normalization, the data is dynamically uploaded to edge nodes and the cloud according to data priority and network bandwidth status. Based on the uploaded data, reliable results are obtained by performing vehicle collaborative perception calculations within the region. The optimal path is determined by combining real-time path cost, and edge-side decision commands are generated and sent to the vehicle terminal. Collect global data from each edge node and perform standardization, deduplication, and outlier filtering; calculate the regional capacity supply and demand gap and plan the optimal route; generate cross-regional capacity scheduling instructions according to order priority. Digital twin models of components are built based on multi-source data from cloud, edge, and device, and the remaining lifespan and failure risk of components are calculated synchronously to generate graded early warning information. Based on the hierarchical early warning information and core indicators, standardized operation and maintenance requirements are constructed. The optimal operation and maintenance resources are selected by matching and scoring, and operation and maintenance work orders containing key information are generated and pushed to the execution nodes and central modules simultaneously. Cloud-edge computing resources are dynamically scheduled based on task priority, data processing complexity, and node load status.

[0032] In summary, the systems and methods described above achieve dynamic allocation of computing power through cloud-edge collaboration, efficiently process multi-source heterogeneous data, improve the reliability of perception results and the rationality of path planning, reduce vehicle energy consumption and travel time, accurately predict the remaining lifespan and failure risk of components, push early warning information in a tiered manner, avoid failure losses in advance, optimize the supply and demand matching of cross-regional transportation capacity, rationally schedule vehicle resources, and quickly adapt to the optimal operation and maintenance resources. They also respond to operation and maintenance needs according to demand priority, shorten response time, ensure the continuous and stable operation of vehicles, reduce operation and maintenance costs, improve the overall efficiency of transportation and operation and maintenance services, and provide comprehensive support for vehicle operation.

[0033] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A vehicle intelligent diagnostic system based on cloud-edge collaboration, characterized in that, include: The data acquisition module is used to collect multi-source heterogeneous data on vehicle operating status, environmental perception, and core component operating conditions. After preliminary preprocessing at the edge, the data is uploaded to the edge nodes and the cloud. The computing module is used to acquire preprocessed data, perform vehicle cooperative perception computing and local dynamic path planning within the area, generate edge-side decision commands and send them to the vehicle terminal; The central module is used to collect global data from each edge node, generate cross-regional capacity scheduling instructions, and send them to the edge nodes. The assessment module is used to build digital twin models of components based on multi-source data from cloud, edge, and device, predict the remaining lifespan and failure risk of components, generate graded early warning information, and push it to the generation module; The generation module is used to receive fault warning information, parse standardized operation and maintenance requirements, match the optimal operation and maintenance resources, generate operation and maintenance work orders, and push them synchronously to the corresponding operation and maintenance execution nodes. The simulation module is used to dynamically schedule cloud-edge computing resources based on task priority.

2. The vehicle intelligent diagnostic system based on cloud-edge collaboration according to claim 1, characterized in that, The multi-source heterogeneous data collected by the acquisition module includes vehicle operating status data, environmental perception data, and core component operating condition data. Among them, vehicle operating status data includes at least vehicle speed, acceleration, braking frequency, steering angle, and remaining battery power; environmental perception data includes at least road slope, road surface friction coefficient, real-time traffic flow, ambient temperature, and visibility; and core component operating condition data includes at least engine speed, transmission oil temperature, braking system pressure, battery cycle count, and motor winding temperature. The initial preprocessing at the edge includes data denoising, spatiotemporal alignment, and normalization. Data denoising removes interfering data through sliding window mean filtering. Spatiotemporal alignment synchronizes the data acquisition time and spatial location of each sensor based on timestamps. Normalization maps various types of data to a preset unified dimension range. The pre-processed data is prioritized according to the basic order of core component operating condition data, vehicle operating status data, and environmental perception data. The upload order is dynamically adjusted based on the real-time network bandwidth status of edge nodes and the cloud. When the network bandwidth is lower than a preset threshold, abnormal fluctuation data in the core component operating condition data is uploaded first. Abnormal fluctuation data is identified by comparing the data fluctuation amplitude with a preset threshold.

3. The vehicle intelligent diagnostic system based on cloud-edge collaboration according to claim 1, characterized in that, When the computing module performs vehicle collaborative perception calculations within the area, it generates reliable perception results through cross-node verification of vehicle data, and its collaborative perception confidence level is: ； In the local dynamic path planning stage, the optimal path is determined based on the collaborative sensing results and the real-time path cost of the inherent paths within the path planning area. The path with the lowest cost and the collaborative sensing results are selected for generating edge-side decision instructions. ； In the formula: To collaboratively perceive confidence; The number of vehicles participating in collaborative sensing within the region; Let be the perception weight coefficient for the i-th vehicle; Let be the single-source perception confidence level of the i-th vehicle; Let be the distance between the i-th vehicle and the center of the target monitoring area; The average distance from all participating collaborative sensing vehicles within the region to the center of the target monitoring area; This represents the path cost value. , , Optimize the weight coefficients for the path; The estimated energy consumption for the target path; The average estimated energy consumption of all feasible paths with the same start and end point within the region; The estimated travel time for the target route; It represents the average estimated travel time for all feasible paths with the same origin and destination within the region.

4. The vehicle intelligent diagnostic system based on cloud-edge collaboration according to claim 1, characterized in that, The global data collected by the central module from each edge node includes full order data, full vehicle status data, real-time traffic fusion data, regional transportation capacity characteristic data, and environmental perception summary data. In the cross-regional capacity scheduling instruction generation stage, the global data is first standardized, deduplicated, and filtered for outliers to obtain a fused dataset. Then, the regional capacity supply and demand gap is calculated based on the fused dataset. The optimal cross-regional route is planned and matched with suitable vehicles in combination with real-time road conditions and vehicle status. Scheduling instructions are generated according to order priority.

5. The vehicle intelligent diagnostic system based on cloud-edge collaboration according to claim 1, characterized in that, The component remaining life and failure risk prediction logic in the assessment module is as follows: ； In the formula: The remaining lifespan of the component; The rated design life of the component; The number of key operating conditions that affect component lifespan; For the k-th type of operating condition, the life loss weighting coefficient is used. This represents the actual cumulative operating time of the component under the k-th type of operating condition; This represents the rated cumulative operating time of the component under the k-th type of operating condition. This is the environmental stress influence coefficient; The deviation between the real-time operating energy loss and the rated energy loss of the component; This represents the average rated energy loss of the component. This represents the component failure risk value. This is the lifespan offset coefficient; The number of component failure monitoring indicators; Let be the weight of the q-th fault monitoring indicator; This is the real-time monitoring value of the q-th fault monitoring indicator; is the normal operating standard value of the q-th fault monitoring indicator.

6. The vehicle intelligent diagnostic system based on cloud-edge collaboration according to claim 1, characterized in that, When the generation module parses the standardized operation and maintenance requirements, it constructs a three-dimensional operation and maintenance requirement matrix based on the fault risk value, component importance level and vehicle task urgency output by the evaluation module. Each dimension of the matrix corresponds to a core requirement indicator. The standardized operation and maintenance requirement vector is obtained through matrix normalization. The component importance level is preset. The matching of operation and maintenance resources is determined by calculating a matching score: ； In the formula: Scoring based on the matching of operation and maintenance resources; , , To match the weight coefficients; The response distance from maintenance resources to the faulty vehicle; The average response distance from all candidate maintenance resources to the faulty vehicle; The service quality level of the operation and maintenance resources; The current load rate of the operation and maintenance resources; The average load rate of all candidate operation and maintenance resources; The generated maintenance work order includes faulty vehicle identification, faulty component information, fault risk level, standardized maintenance requirements, optimal maintenance resource identification, response time limit requirements, and maintenance operation specifications. The work order is synchronously pushed to the corresponding maintenance execution node through encrypted transmission, and is also synchronously backed up to the central module.

7. A vehicle intelligent diagnostic system based on cloud-edge collaboration according to claim 5, characterized in that, The tiered early warning information generated by the assessment module is divided into intervals based on the historical maximum and minimum values ​​of the fault risk value, corresponding to three early warning levels; Level 1 warning corresponds to a high fault risk value, Level 2 warning corresponds to a medium fault risk value, and Level 3 warning corresponds to a low fault risk value. After receiving the tiered early warning information, the generation module adapts the corresponding operation and maintenance response priority according to the early warning level: The Level 1 warning triggers an emergency operation and maintenance response, prioritizing the allocation of the best available operation and maintenance resources and simultaneously pushing real-time warning information to vehicle management personnel; Level 2 warnings trigger routine maintenance responses, completing maintenance resource matching and work order push within a preset time. Level 3 warnings trigger preventative maintenance responses, entering the maintenance queue after the vehicle's current task is completed. Vehicles in the maintenance queue are sorted according to the trigger time of the Level 3 warnings and perform maintenance operations in sequence.

8. The vehicle intelligent diagnostic system based on cloud-edge collaboration according to claim 1, characterized in that, When the simulation module dynamically schedules cloud-edge computing resources, it allocates computing power based on task priority, data processing complexity, and the current load status of the nodes. ； In the formula: Allocate computing power to the j-th cloud edge node; The total available computing power of the cloud-edge collaborative system; This represents the priority coefficient of the current task. Let be the computing power requirement of the j-th cloud edge node for the current task; Let be the current load rate of the j-th cloud edge node; This represents the total number of cloud-edge nodes participating in computing power scheduling.

9. A vehicle intelligent diagnostic system based on cloud-edge collaboration according to claim 1, characterized in that, The acquisition module is interconnected with the computing module via a wireless network. The computing module is interconnected with the central module via a wireless network. The central module is interconnected with the evaluation module via a wireless network. The evaluation module is interconnected with the generation module and the simulation module via a wireless network.

10. A vehicle intelligent diagnostic method based on cloud-edge collaboration, wherein the method is an implementation method of the vehicle intelligent diagnostic system based on cloud-edge collaboration as described in any one of claims 1-9, characterized in that, include: Multi-source heterogeneous data on vehicle operating status, environmental perception, and core component operating conditions are collected, and after preprocessing such as noise reduction, spatiotemporal alignment, and normalization, the data is dynamically uploaded to edge nodes and the cloud according to data priority and network bandwidth status. Based on the uploaded data, reliable results are obtained by performing vehicle collaborative perception calculations within the region. The optimal path is determined by combining real-time path cost, and edge-side decision commands are generated and sent to the vehicle terminal. Collect global data from each edge node and perform standardization, deduplication, and outlier filtering; calculate the regional capacity supply and demand gap and plan the optimal route; generate cross-regional capacity scheduling instructions according to order priority. Digital twin models of components are built based on multi-source data from cloud, edge, and device, and the remaining lifespan and failure risk of components are calculated synchronously to generate graded early warning information. Based on the hierarchical early warning information and core indicators, standardized operation and maintenance requirements are constructed. The optimal operation and maintenance resources are selected by matching and scoring, and operation and maintenance work orders containing key information are generated and pushed to the execution nodes and central modules simultaneously. Cloud-edge computing resources are dynamically scheduled based on task priority, data processing complexity, and node load status.

Images