A vehicle over-discharging risk assessment, over-discharging identification and traceability tracking system

By constructing a multi-dimensional vehicle emission characteristic benchmark and risk classification mechanism, combined with a semi-supervised isolated forest model and source tracing analysis, the problems of low regulatory efficiency and insufficient judgment accuracy in existing technologies are solved, enabling accurate identification and source tracing of vehicle over-emissions and supporting efficient on-site law enforcement.

CN123116053APending Publication Date: 2026-11-17CATARC AUTOMOTIVE TEST CENT TIANJIN CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611222860.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-12
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies for vehicle emission regulation suffer from low efficiency, insufficient accuracy, and lack of root cause analysis, making it impossible to achieve precise and efficient emission regulation and on-site enforcement support.

Method used

By establishing modules for characterizing normal emissions, assessing vehicle over-emission risks, accurately identifying emissions per vehicle, and tracing the causes of over-emissions, a benchmark for vehicle emissions characteristics under multi-dimensional and segmented scenarios is constructed. A risk grading mechanism for vehicle over-emissions is set up, a semi-supervised isolated forest model is used to determine the emission status, and a tracing analysis of system-level and component-level root causes is performed.

Benefits of technology

It enables accurate identification and tracing of high-risk vehicles, improves regulatory efficiency and judgment accuracy, and supports efficient closed-loop operation of on-site law enforcement.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116053A_ABST
    Figure CN123116053A_ABST
Patent Text Reader

Abstract

This application provides a vehicle emission over-emission risk assessment, over-emission identification, and source tracing system, relating to the field of transportation technology. It includes a normal emission characteristic characterization module that acquires driving data of vehicles with normal emission status and establishes a multi-dimensional, segmented statistical map of vehicle emission characteristics; a vehicle emission over-emission risk assessment module that acquires driving data of all vehicles on the platform and calculates the comprehensive over-emission risk of all vehicles; a vehicle emission over-emission risk classification mechanism based on the comprehensive over-emission risk to screen out high-risk vehicles; a vehicle-by-vehicle emission precise identification module that establishes a dual emission status determination mechanism for high-risk vehicles to determine whether the vehicle's final emission status is over-emission; and an over-emission cause tracing and analysis module that establishes an analysis strategy for vehicles determined to be over-emission to trace the source of over-emissions. This system solves the problems of poor regulatory efficiency, low judgment accuracy, and lack of root cause tracing in existing technologies.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of transportation technology, and in particular to a system for assessing, identifying, and tracing the source of vehicle emissions exceeding limits. Background Technology

[0002] Vehicle pollution control is a core area of ​​air pollution management in my country, with vehicle emission control being a key focus. According to the requirements of the China VI emission standards, in-use vehicles must be equipped with onboard remote emission monitoring devices and connected to a unified remote monitoring platform to achieve real-time monitoring of vehicle emission status, identification and location of vehicles exceeding emission standards, and source analysis of emission anomalies. Against this backdrop, relying on vehicle-to-everything (V2X) data transmission for monitoring vehicles exceeding emission standards has become a core technological approach for environmental regulatory departments.

[0003] However, existing technologies still have three limitations in practical engineering applications, making it difficult to support the needs of accurate, efficient, and closed-loop emission supervision. First, the efficiency of comprehensive supervision is low, lacking a pre-emptive risk screening mechanism. The current regulatory system needs to cover a massive number of vehicles registered on the network, but a scientific vehicle emission risk classification system has not yet been established. Regulatory enforcement resources are evenly distributed, failing to accurately focus on high-risk vehicles, resulting in high enforcement costs and a low hit rate for problematic vehicles. Second, the accuracy of emission exceedance judgment is insufficient, failing to fully consider the impact of load conditions. Existing emission exceedance identification mostly uses fixed emission threshold judgment logic, failing to fully consider the impact of vehicle loading status, and cannot adapt to the differences in emission characteristics under different load conditions. This easily leads to misjudgments of normal emissions under heavy load conditions and missed detections of hidden emission exceedances under light load conditions. The accuracy of single-vehicle identification is insufficient to meet the requirements of law enforcement compliance. Third, the root cause analysis mechanism is lacking, failing to effectively support on-site enforcement. Existing root cause analyses of excessive emissions mostly rely on purely data-driven black box models, without combining the physical mechanisms of the post-treatment system for deduction. This makes it difficult to output verifiable root causes and rectification guidelines on-site. Enforcement personnel still need to conduct a full investigation of the emission system during on-site inspections, resulting in low inspection efficiency and poor implementation. Summary of the Invention

[0004] This application provides a vehicle over-emission risk assessment, over-emission identification, and source tracing system to address the problems of poor regulatory efficiency, low judgment accuracy, and lack of root cause tracing in the prior art.

[0005] The system includes: A normal emission feature characterization module is configured to acquire driving data of vehicles with normal emission status and establish a benchmark statistical map of vehicle emission features under multi-dimensional subdivided scenarios based on the driving data. The vehicle over-emission risk assessment module is configured to acquire driving data of all vehicles in the platform, calculate the comprehensive over-emission risk of all vehicles based on the driving data, set a vehicle over-emission risk classification mechanism based on the comprehensive over-emission risk, and filter out high-risk vehicles from all vehicles in the platform based on the vehicle over-emission risk classification mechanism. The vehicle-by-vehicle emission precision identification module is configured to establish a dual emission status determination mechanism for high-risk vehicles; determine whether the final emission status of the vehicle is excessive based on the dual emission status determination mechanism; the dual emission status determination mechanism includes a baseline threshold initial judgment and anomaly score verification; different types of vehicles correspond to different vehicle excessive emission risk classification mechanisms; The over-emission cause tracing and analysis module is configured to establish an analysis strategy, including system-level positioning and component root cause positioning, for vehicles identified as over-emission vehicles, and to trace the over-emission sources of vehicles based on the analysis strategy, generating a standardized diagnostic report; the standardized diagnostic report is used to guide on-site law enforcement.

[0006] In some embodiments, the normal emission characteristic characterization module is further configured to: The driving data is preprocessed to remove data from periods of malfunction or abnormal operation. The preprocessed data was segmented using a sliding time window method to obtain several driving samples; Establish the multi-dimensional segmented scenarios that consider vehicle cargo status, vehicle operating environment and vehicle driving conditions, and classify all driving samples into the corresponding segmented scenarios in the multi-dimensional segmented scenarios; For each of the aforementioned sub-scenarios, based on the normal distribution theory, the upper limit of the confidence interval for the nitrogen oxide emission values ​​of all driving samples in the corresponding sub-scenarios is calculated, and this is used as the emission benchmark threshold for the corresponding sub-scenarios, thereby forming the vehicle emission characteristic benchmark statistical map.

[0007] In some embodiments, the vehicle over-emission risk assessment module is further configured to: Historical remote transmission data of the target vehicle within a certain period of time is obtained and divided into units of natural months. The historical remote transmission data of each month is processed into driving samples and assigned to the corresponding sub-scenarios. Within each calendar month, the percentage of driving samples whose nitrogen oxide emissions exceed the emission benchmark threshold for each of the aforementioned sub-scenarios is calculated to obtain the monthly single-scenario risk score. The average score of all the subdivided scenarios is calculated to obtain the monthly risk score for each month, and this is used to form the annual monthly risk score sequence for the target vehicle. The monthly risk scores in the annual monthly risk score sequence are assigned decreasing weights according to their time proximity and then summed in a weighted manner to obtain the comprehensive over-emission risk of the target vehicle.

[0008] In some embodiments, the vehicle over-emission risk classification mechanism further includes: Vehicles whose overall emission risk does not reach the high-risk level, but whose overall emission risk has increased continuously over the past three months and whose rate of increase exceeds the preset warning threshold, are included in the high-risk level.

[0009] In some embodiments, the vehicle-by-vehicle emission precision identification module is further configured to: In the initial judgment stage of the benchmark threshold, the driving samples of the vehicle to be judged are divided into the corresponding sub-scenarios, the average risk score of all sub-scenarios is calculated and compared with the first preset threshold; when the average risk score is greater than the first preset threshold, the corresponding vehicle is initially judged as over-emission. In the abnormal score review stage, the abnormal scores of all driving samples of the vehicle to be judged are calculated using a pre-built semi-supervised isolated forest model. When the proportion of driving samples with abnormal scores greater than the second preset threshold exceeds the third preset threshold, the corresponding vehicle is judged as over-evaluated. When the vehicle is judged as over-evaluated in both the baseline threshold initial judgment stage and the abnormal score review stage, the vehicle is judged as over-evaluated.

[0010] In some embodiments, the method for constructing the semi-supervised isolated forest model includes: A basic isolated forest was constructed using the driving samples with normal emission status to characterize the distribution of normal emissions; Semi-supervised threshold calibration is performed on the basic isolated forest using driving samples with abnormal emission status; through an optimization algorithm, the recall rate of abnormal samples is maximized while ensuring that the recognition accuracy of normal samples is not lower than a preset value, thereby determining the second preset threshold. Establish a periodic update mechanism for the semi-supervised isolated forest model, and periodically use newly acquired law enforcement confirmation of over-emission data and re-inspection qualified data to iteratively optimize the second preset threshold or the structure of the semi-supervised isolated forest model.

[0011] In some embodiments, the over-discharge cause tracing and analysis module is further configured to: In system-level positioning, the average value of the nitrogen oxide conversion efficiency of the aftertreatment selective catalytic reduction system of the vehicle with excessive emissions is calculated; when the average value is lower than the theoretical value, the cause of excessive emissions is determined to be an abnormality of the aftertreatment selective catalytic reduction system. In the root cause localization of components, for vehicles that are determined to have abnormal selective catalytic reduction systems in the aftertreatment, the causes of excessive emissions are further subdivided into urea depletion faults, selective catalytic reduction catalyst faults, or poor urea quality faults based on the characteristic distribution of urea level, selective catalytic reduction inlet temperature, and nitrogen oxide conversion efficiency.

[0012] In some embodiments, the subdivision of over-discharge causes into urea depletion failures includes: Calculate the percentage of driving samples where the urea level of the over-emission vehicle is below a preset low level threshold; if this state occurs continuously and the duration or number of consecutive samples exceeds the preset threshold, then the cause of the over-emission of the vehicle is determined to be the urea depletion fault.

[0013] In some embodiments, the causes of over-emission can be further subdivided into selective catalytic reduction catalyst failure, including: For vehicles not identified as having urea depletion issues, if their nitrogen oxide conversion efficiency is lower than the fourth preset threshold, and the inlet temperature of the aftertreatment selective catalytic reduction system is higher than the fifth preset threshold and the duration exceeds the sixth preset threshold, then the cause of the excessive emissions will be identified as a failure of the selective catalytic reduction catalyst.

[0014] In some embodiments, the subdivision of excessive discharge causes into urea quality failures includes: For vehicles that are not identified as having a urea depletion fault or a selective catalytic reduction catalyst fault, if the urea level is not lower than a preset low level threshold, the nitrogen oxide conversion efficiency is lower than a seventh preset threshold, the selective catalytic reduction inlet temperature is within a preset normal operating temperature range, and the vehicle's urea consumption state is highly consistent with that of a normal vehicle, then the vehicle is identified as having a poor urea quality fault.

[0015] As described above, this application provides a vehicle emission over-emission risk assessment, over-emission identification, and source tracing system. The system includes a normal emission characteristic characterization module, configured to acquire driving data of vehicles with normal emission status and establish a multi-dimensional, segmented vehicle emission characteristic benchmark statistical map based on the driving data; a vehicle emission over-emission risk assessment module, configured to acquire driving data of all vehicles on the platform, calculate the comprehensive over-emission risk of all vehicles based on the driving data; set a vehicle emission over-emission risk grading mechanism based on the comprehensive over-emission risk; and filter out high-risk vehicles from all vehicles on the platform based on the vehicle emission over-emission risk grading mechanism; and perform vehicle-by-vehicle emission precision... The system includes a quasi-identification module, configured to establish a dual emission status determination mechanism for high-risk vehicles; determine whether the vehicle's final emission status is excessive based on this mechanism; the dual emission status determination mechanism includes initial judgment based on a baseline threshold and verification of abnormal scores; different types of vehicles correspond to different vehicle excessive emission risk classification mechanisms; and an excessive emission cause tracing and analysis module, configured to establish an analysis strategy including system-level positioning and component root cause positioning for vehicles identified as excessively emitting, and perform excessive emission tracing for these vehicles based on the analysis strategy, generating a standardized diagnostic report; this standardized diagnostic report is used to guide on-site enforcement. This application solves the problems of poor regulatory efficiency, low judgment accuracy, and lack of root cause tracing in the prior art through the above system. Attached Figure Description

[0016] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of a vehicle over-emission risk assessment, over-emission identification, and source tracing system according to this application; Figure 2 This is a flowchart illustrating the operational logic of a vehicle over-emission risk assessment, over-emission identification, and source tracing system according to this application. Figure 3 A graph showing the NOx emission baseline thresholds for each specific scenario; Figure 4 A typical vehicle emission overload risk diagram; Figure 5 A typical example chart showing the proportion of vehicles exceeding the abnormal score threshold; Figure 6 The recognition effect of a semi-supervised isolated forest model. Detailed Implementation

[0018] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be noted that the brief descriptions of terms in this application are only for the convenience of understanding the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise stated, these terms should be understood in their ordinary and common meaning.

[0020] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0021] See Figure 1 and Figure 2 As can be seen, this embodiment provides a vehicle over-emission risk assessment, over-emission identification, and source tracing system, the system comprising: The normal emission feature characterization module is configured to acquire driving data of vehicles with normal emission status and establish a benchmark statistical map of vehicle emission features under multi-dimensional subdivided scenarios based on the driving data.

[0022] Specifically, in this embodiment, 2,000 vehicles with normal emission status are selected as benchmark sample vehicles through the normal emission characteristic characterization module, and the driving data of the vehicles throughout 2025 are obtained. The sources of the driving data include, but are not limited to: calibration data of models that have passed emission compliance certification, normal operation data of vehicles that have passed annual inspection, historical remote transmission data of vehicles in operation with no fault records, and compliance emission test data under standard working conditions of chassis dynamometer.

[0023] Driving data parameters include, but are not limited to: time, vehicle speed, acceleration, engine torque, engine speed, intake air volume, inlet and outlet temperatures of the aftertreatment SCR system (selective catalytic reduction system), upstream and downstream NOx (nitrogen oxides) emission rates of the aftertreatment SCR system, NOx conversion efficiency of the aftertreatment SCR system, latitude and longitude, etc.

[0024] Among them, engine torque, engine speed, and intake air volume are obtained by the vehicle engine controller or vehicle CAN bus; SCR system inlet and outlet temperatures, upstream and downstream NOx emission rates, and NOx conversion efficiency are obtained by sensors of the vehicle aftertreatment system; latitude, longitude, and altitude are obtained by the vehicle positioning module; and OBD fault codes and aftertreatment regeneration status indicators are obtained by the vehicle remote emission monitoring terminal.

[0025] The above data is imported into the computer processing unit, and the vehicle's cargo status and road slope parameters are calculated based on this.

[0026] The vehicle's cargo status is calculated and obtained by a pre-built vehicle cargo status identification model based on vehicle dynamics theory, including three states: empty, half-loaded, and fully loaded.

[0027] Among them, the road slope is obtained based on the vehicle's latitude and longitude coordinates to identify uphill, flat, and downhill slopes.

[0028] The normal emission characteristic characterization module also removes data from the OBD fault code period, the after-processing regeneration period, and the sensor abnormal period based on the vehicle's driving data mentioned above, ensuring that all remaining data are in the normal operation phase of the vehicle, thereby obtaining preprocessed vehicle driving data.

[0029] On the computer processing end, the sliding time window method is used to divide all preprocessed vehicle driving data into fixed-duration windows to obtain driving samples of vehicles with normal emission status.

[0030] On the computer processing end, a multi-dimensional segmented scenario is established that considers vehicle cargo status, vehicle operating environment, and vehicle driving conditions, and all vehicle driving samples are divided into segmented scenarios. For each segmented scenario, based on the normal distribution theory, the upper limit of the 95% confidence interval of the NOx emission values ​​of all samples in the scenario is calculated as the emission benchmark threshold. The emission benchmark thresholds of all segmented scenarios are integrated to form a universal vehicle emission characteristic benchmark statistical map.

[0031] In the process of establishing multi-dimensional subdivided scenarios, the vehicle's cargo status is divided into empty, half-loaded, and fully loaded; the vehicle's operating environment is divided into urban roads, suburban roads, highways, and mountain roads; and the vehicle's driving conditions are divided into idling, acceleration, constant speed, and climbing.

[0032] The aforementioned multi-dimensional segmentation method and the resulting universally applicable vehicle emission characteristic benchmark statistical map are stored in the distributed storage device of the vehicle remote monitoring platform.

[0033] like Figure 3As shown in the figure, the upper limit of the 95% confidence interval for NOx emissions from normal vehicle driving samples in each sub-scenario is presented, serving as the benchmark threshold for NOx emissions in the corresponding sub-scenario. Figure 3 It is evident that the NOx emission baseline threshold generally increases with increasing vehicle load, more complex road operating conditions, and increased engine load. Specifically, the baseline threshold for an unloaded urban road idling scenario is approximately 0.28 g / s, while the baseline threshold for a fully loaded mountain road climbing scenario is approximately 1.38 g / s. This demonstrates that the normal emission levels of vehicles vary significantly under different load conditions, road environments, and driving conditions, necessitating the use of emission baseline thresholds corresponding to specific scenarios for over-emission risk assessment and identification.

[0034] The system also includes: The vehicle over-emission risk assessment module is configured to acquire driving data of all vehicles in the platform, calculate the comprehensive over-emission risk of all vehicles based on the driving data, set a vehicle over-emission risk classification mechanism based on the comprehensive over-emission risk, and select high-risk vehicles from all vehicles in the platform based on the vehicle over-emission risk classification mechanism.

[0035] Specifically, in this embodiment, on the computer processing end, through the vehicle over-emission risk assessment module in the vehicle remote monitoring platform, a two-dimensional over-emission risk assessment framework of "monthly static basic score + annual time-series trend correction" is established for all vehicles connected to the platform to calculate the comprehensive over-emission risk of the vehicles. The calculation method for the comprehensive over-emission risk of any vehicle connected to the platform is as follows: (1) Obtain the remote transmission data of the target vehicle in the past 12 months and divide the vehicle remote transmission data into units of natural months.

[0036] (1.1) Preprocess, window, and subdivide all remote transmission data of the vehicle according to the same method as above, and then obtain the driving sample of the target vehicle and classify it into the corresponding subdivided scenario.

[0037] (1.2) Within each natural month, for all target vehicle driving samples in each sub-scenarios, calculate the proportion of driving samples with NOx emission values ​​higher than the emission benchmark threshold, and divide it by 0.95 to obtain the monthly single-scenario risk score.

[0038] (1.3) Within each natural month, calculate the average monthly risk score of all subdivided scenarios to obtain the monthly risk score.

[0039] (1.4) Integrate the monthly risk scores of all natural months to form the annual monthly risk score sequence of the target vehicle.

[0040] (2) Following the objective law of "nearer is larger and farther is smaller", the monthly risk scores of the 12 months of the year are assigned differentiated weights. Specifically, the closer the month is to the current time, the higher the weight (the recommended value is 1); the farther the month is from the current time, the lower the weight (it is recommended to reduce the weight by 1 / 12 for each month farther away). This method can highlight the dominant role of the vehicle's recent operating status on the current risk.

[0041] Based on the monthly risk score sequence of the target vehicle throughout the year, the comprehensive over-emission risk of the target vehicle can be calculated.

[0042] like Figure 4 As shown in the figure, six representative case vehicles, A1 to A6, are displayed, showing their monthly risk scores for 12 consecutive months and the overall over-emission risk of the vehicles. Figure 4 In (b), the comprehensive emission risk of vehicles A1 to A6 are 0.85, 0.91, 0.95, 0.97, 1.00, and 1.02, respectively. This shows that the comprehensive emission risk not only reflects the overall degree of vehicle emission anomalies but also highlights vehicles with rapidly deteriorating emission performance in the near future through time-series trend correction, providing a basis for subsequent vehicle risk classification and high-risk vehicle screening.

[0043] The vehicle over-emission risk assessment module uses comprehensive over-emission risk data from all vehicles connected to the platform to establish a vehicle over-emission risk classification mechanism, screening out high-risk vehicles, mainly including: Establish a vehicle emission over-emission risk classification mechanism, among which... Vehicles with high risk levels are defined as those ranking in the top 10% of the comprehensive emission risk values. These vehicles are included in the key supervision list and will directly enter the next stage of precise vehicle-by-vehicle identification. Vehicles at medium risk are defined as those ranking in the top 10% to 40% of the comprehensive emission risk values. These vehicles are subject to regular monitoring and their risk status is updated monthly. Vehicles with low risk levels are defined as those ranking in the bottom 60% of the overall emission risk index. These vehicles are subject to low-frequency monitoring, and their risk status is updated quarterly.

[0044] Based on this, a forward-looking supplementary rule is incorporated: for vehicles that are not in the top 10% but whose comprehensive emission risk value has been rising continuously over the past 3 months and whose slope exceeds the warning threshold (1 is recommended), they are additionally included in the "high-risk level" to achieve early interception of vehicles with rapidly deteriorating emission performance.

[0045] Based on the above mechanism, among all vehicles connected to the platform, high-risk vehicles are selected to carry out the next stage of precise vehicle-by-vehicle identification, thereby narrowing the scope of supervision from all vehicles to a small number of high-risk vehicles.

[0046] The system also includes: The vehicle-by-vehicle emission precision identification module is configured to establish a dual emission status determination mechanism for high-risk vehicles; determine whether the final emission status of the vehicle is excessive based on the dual emission status determination mechanism; the dual emission status determination mechanism includes a baseline threshold initial judgment and anomaly score verification; different types of vehicles correspond to different vehicle excessive emission risk classification mechanisms.

[0047] Specifically, in this embodiment, The vehicle-by-vehicle emission precision identification module retrieves remotely transmitted vehicle data from the platform for vehicles selected for high-risk emissions levels and imports it into the computer processing unit. A dual emission status determination mechanism is established, consisting of initial judgment based on a baseline threshold and verification of anomaly scores, to determine the final emission status of the vehicle, including: (1) Vehicles with high hyperbaric emission risk levels are selected as vehicles to be judged. For each vehicle to be judged, the recent remote transmission data of the vehicle to be judged is obtained from the platform and imported into the computer processing terminal. All remote transmission data of the vehicle are preprocessed, windowed and subdivided into scenarios according to the same method as above, so as to obtain the driving samples of the vehicles to be judged.

[0048] (2) Preliminary judgment stage of benchmark threshold: For all driving samples in each sub-scenario, calculate the proportion of driving samples with NOx emission values ​​higher than the emission benchmark threshold, and divide it by 0.95 to obtain the single-scenario risk score of the vehicle to be judged; based on this, calculate the average of the single-scenario risk scores of all sub-scenarios and compare it with 0.95.

[0049] If the average value is greater than 0.95, the vehicle to be judged is considered to be over-emission in the initial judgment stage of the baseline threshold; otherwise, the emission status is normal.

[0050] (3) Abnormal score review stage: Construct a driving sample status judgment model based on semi-supervised isolated forest and obtain the optimal score threshold of the model; Based on this, use the model to calculate the abnormal score of all driving samples of the vehicle to be judged. When the proportion of samples with abnormal scores greater than the optimal score threshold exceeds the abnormal score review threshold (recommended value is 50%), the vehicle to be judged is considered to be over-emission in the abnormal score review stage; otherwise, the emission status is normal.

[0051] (4) If a vehicle is judged to be over-emission in both the initial judgment of the baseline threshold and the review of abnormal scores, then the vehicle is considered to have an over-emission problem.

[0052] like Figure 5As shown, this case study focuses on the six typical vehicles A1 to A6 mentioned above. A semi-supervised isolated forest model was used to calculate the anomaly scores for each vehicle's driving samples, and the proportion of driving samples with anomaly scores greater than the optimal score threshold of 0.55 was calculated. The results show that the anomaly sample proportion for vehicle A1 was 0%, for vehicle A2 it was 17%, for vehicle A3 it was 67%, for vehicle A4 it was 58%, and for vehicles A5 and A6 it was 100%. Therefore, this case study ultimately identifies vehicles A3 to A6, which passed both stages of the assessment, as vehicles exceeding emission limits.

[0053] (3.1) The model training dataset includes: a sample set with normal emission status (the main training data, accounting for 95%) and a sample set with abnormal emission status (for calibration, accounting for 5%).

[0054] The sample set with normal emissions mainly consists of driving samples of several vehicles with normal emissions.

[0055] Among them, the sample set of abnormal emission status mainly comes from the remote transmission data of vehicles that exceed emission limits found by law enforcement agencies. The remote transmission data of vehicles that exceed emission limits is preprocessed, windowed, and subdivided into different scenarios in the same way as described above, so as to obtain the driving samples of vehicles that exceed emission limits, which serve as the sample set of abnormal emission status.

[0056] The model input features include, but are not limited to: vehicle speed, acceleration, engine speed, engine torque, intake air volume, vehicle load status, road gradient, SCR system inlet temperature, SCR system outlet temperature, upstream NOx emission rate of the SCR system, downstream NOx emission rate of the SCR system, and SCR system NOx conversion efficiency. These features are combined in a preset order to form a fixed-dimensional driving sample feature vector, which serves as the input to the semi-supervised isolated forest-based driving sample state determination model.

[0057] (3.2) The model validation dataset consists of a small number of sample data with normal emission status and a small number of sample data with abnormal emission status.

[0058] (3.3) The model training process includes two stages: "basic forest construction" and "semi-supervised threshold calibration", which are mainly as follows: In the basic forest construction phase, a large dataset of samples with normal emission status is used for model training. First, arbitrary samples are randomly sampled from the normal emission status dataset. A feature parameter is randomly selected, and a splitting threshold is randomly generated within its value range, dividing all samples into left and right nodes. Second, the newly generated left and right nodes are processed again in this manner, recursively splitting until the final node contains only one sample or the maximum depth of the isolated tree is reached. Finally, the above process is repeated to generate N isolated trees (100-200 trees are recommended), which together form the basic isolated forest, characterizing the distribution of vehicles with normal emission status.

[0059] In this case, a basic isolated forest is constructed using driving samples with normal emission status from the model training dataset. First, 256 driving samples are randomly selected without replacement from the normal emission status sample set as the training sample set for a single isolated tree.

[0060] Next, a feature parameter is randomly selected from the feature vector of the driving samples, and a splitting threshold is randomly generated between the minimum and maximum values ​​of the current range of the feature parameter. All driving samples within the current node are then assigned to left or right nodes. For driving samples whose feature parameter values ​​are less than the splitting threshold, they are assigned to the left node; for driving samples whose feature parameter values ​​are greater than or equal to the splitting threshold, they are assigned to the right node.

[0061] The newly generated left and right nodes are recursively split using the method described above until the number of driving samples in a node is 1, the feature parameter values ​​of all driving samples in a node are the same, or the depth of the isolated tree reaches the preset maximum depth. In this case, the maximum depth of the isolated tree is set to 8.

[0062] Finally, the isolation tree construction process was repeated to generate 150 independent isolation trees, which together formed a basic isolated forest. This basic isolated forest was used to characterize the distribution of driving sample features of vehicles with normal emissions.

[0063] In the semi-supervised threshold calibration stage, a basic isolated forest is trained using a small sample set of samples with abnormal emission states. First, for any sample in the abnormal emission state sample set, its average path length across all isolated trees in the basic isolated forest is calculated and mapped to an anomaly score between 0 and 1 using a normalization formula, as shown in the following equation: ; In the formula, Let x be the average path length of sample x across all isolated trees; It is the normalization factor for the average path length of n samples; the closer the score is to 1, the higher the probability of abnormal emission status of the sample.

[0064] Secondly, the model's score threshold is optimized using samples from the model validation dataset. Under the constraint of ensuring that the accuracy of identifying samples with normal emission status is ≥95%, the recall rate (true rate) of samples with abnormal emission status is maximized to obtain the optimal score threshold. The grid search method is mainly used to traverse the candidate score thresholds within the score threshold range of 0.3 to 0.8, calculate the accuracy of identifying samples with normal emission status and the recall rate of samples with abnormal emission status under the corresponding score threshold, and select the score threshold that meets the constraint and has the highest recall rate as the final optimal score threshold.

[0065] In this case, the anomaly score threshold was calibrated based on the model validation dataset. A grid search method was used to traverse the candidate threshold range of 0.30 to 0.80 with a step size of 0.01. The recognition accuracy of normal samples and the recognition effect of anomaly samples under each candidate threshold were compared. Finally, 0.55 was selected as the optimal anomaly score threshold for the semi-supervised isolated forest model.

[0066] Figure 6 This demonstrates the predictive performance of the semi-supervised isolated forest model when 0.55 is selected as the optimal anomaly score threshold. Figure 6 As shown in (a), when the anomaly score threshold is 0.55, the anomaly scores of most normal samples are below this threshold, while the anomaly scores of most abnormal samples are above this threshold, with only a small number of samples being misjudged. Figure 6 As shown in (b), the model has an accuracy of 96.7% for normal samples, an accuracy of 91.3% for abnormal samples, an abnormal sample recall rate of 93.3%, and an overall accuracy of 95.8%. This indicates that selecting 0.55 as the abnormal score threshold can effectively identify driving samples with abnormal emission status while ensuring the accuracy of normal sample identification.

[0067] (3.4) Establish a periodic model update mechanism to iteratively optimize all constructed driving sample state determination models based on semi-supervised isolated forests, mainly including: First, the remote transmission data of vehicles that exceeded emission standards and those that passed the re-inspection and had normal emission status are aggregated in the current month. The data is then substituted into the model validation dataset to re-optimize the optimal score threshold without modifying the forest structure. Secondly, if the updated optimal score threshold differs too much from the original optimal score threshold, the remote transmission data of vehicles with excessive emissions confirmed by law enforcement in the current month and the remote transmission data of vehicles with normal emission status that have passed re-inspection will be respectively brought into the model training dataset to reconstruct the driving sample status judgment model based on semi-supervised isolated forest.

[0068] The system also includes: The over-emission cause tracing and analysis module is configured to establish an analysis strategy, including system-level positioning and component root cause positioning, for vehicles identified as over-emission vehicles, and to trace the over-emission sources of vehicles based on the analysis strategy, generating a standardized diagnostic report; the standardized diagnostic report is used to guide on-site law enforcement.

[0069] Specifically, in this embodiment, the over-emission cause tracing and analysis module establishes an analysis strategy of "system-level positioning + component root cause positioning" for vehicles that exceed emission limits. On the computer processing end, physical analysis is performed on confirmed over-emission vehicles, generating a standardized diagnostic report, which is then pushed to the enforcement terminal. During on-site inspections, enforcement personnel can directly pinpoint the over-emission problem based on the analysis conclusions in the diagnostic report and issue rectification measures, including: (1) System-level positioning: Based on all driving samples of the vehicles with excessive emissions, calculate the NOx conversion efficiency of the SCR system of the driving samples. After removing outliers according to the box plot principle, calculate the average value of the NOx conversion efficiency. If the average value of the NOx conversion efficiency is significantly lower than the theoretical value, it is determined to be an abnormality of the post-processing SCR system; otherwise, it is determined to be due to other reasons.

[0070] (2) Root cause localization of components: For the abnormality of the aftertreatment SCR system, the causes of excessive emissions are analyzed based on the characteristic distribution of parameters such as urea level, SCR inlet temperature, NOx conversion efficiency, urea-fuel ratio, and urea consumption per 100 kilometers. The causes mainly include three categories: poor urea quality, urea depletion, and SCR catalyst failure.

[0071] First, obtain the urea level, SCR inlet temperature, NOx conversion efficiency, urea-fuel ratio, and urea consumption per 100 kilometers for vehicles with excessive emissions during the diagnostic cycle.

[0072] Secondly, obtain reference values ​​for the urea-fuel ratio and urea consumption per 100 kilometers for vehicles with normal emissions under the same load conditions, road operating environment, vehicle driving conditions, and engine load, and calculate the urea consumption consistency coefficient K for the vehicle to be diagnosed using the following formula: ; In the formula, The coefficient for consistency of urea consumption; The urine-fuel ratio of the vehicle to be diagnosed during the diagnostic period; This is a reference value for the fuel-to-air ratio of a vehicle with normal emissions under the same operating conditions. Urea consumption per 100 kilometers for the vehicle to be diagnosed; This is a reference value for urea consumption per 100 kilometers for vehicles with normal emissions under the same operating conditions. The weight corresponding to the relative deviation of the urine fuel ratio; The weight corresponding to the relative deviation of urea consumption per 100 kilometers. It is recommended that both be set to 0.5.

[0073] The value range is 0 to 1. The closer the value is to 1, the more consistent the urea consumption status of the vehicle under diagnosis is with that of a vehicle with normal emission status.

[0074] (2.1) Urea depletion fault determination: Calculate the percentage of driving samples where the urea level of the over-emission vehicle is below 0.5%. If the state of urea level below 0.5% occurs continuously and the number of driving samples that occur continuously exceeds a preset threshold, or the duration of this state exceeds a preset threshold, then it is determined that the SCR system cannot obtain the normal urea supply required for the NOx reduction reaction, and the cause of the vehicle's over-emission is determined to be urea depletion fault.

[0075] In this case, the threshold for the number of consecutive driving samples is set at 3, and the threshold for the duration is set at 30 seconds. If the vehicle's urea level is below 0.5% for 3 consecutive driving samples, or if the urea level is below 0.5% for more than 30 seconds, the cause of the vehicle's over-discharge will be determined as a urea depletion fault.

[0076] (2.2) SCR catalyst failure determination: For vehicles not identified as having urea depletion failure, the vehicle's SCR inlet temperature data is obtained. If the vehicle's NOx conversion efficiency is less than 50%, and there are instances where the SCR inlet temperature is higher than 600℃, and the continuous or cumulative time for the SCR inlet temperature to be higher than 600℃ exceeds the preset high-temperature duration threshold, then it is considered that the SCR catalyst has been in an excessively high-temperature environment for a long time, causing the SCR catalyst to undergo high-temperature sintering, and thus the cause of the vehicle's excessive emissions is determined to be an SCR catalyst failure.

[0077] In this case, the preset high-temperature duration threshold is 300 seconds. If the vehicle's NOx conversion efficiency is less than 50%, and the continuous or cumulative time for the SCR inlet temperature to be above 600°C exceeds 300 seconds, the cause of the vehicle's excessive emissions will be determined to be an SCR catalyst malfunction.

[0078] (2.3) Determination of Poor Urea Quality Fault: For vehicles not identified as having urea depletion or SCR catalyst faults, the vehicle's inlet temperature is compared with the preset normal operating temperature range of the SCR system. If the vehicle's urea level is not lower than 0.5%, the NOx conversion efficiency is lower than 70%, the SCR inlet temperature is within the preset normal operating temperature range, and the urea consumption consistency coefficient K is greater than 0.8, it indicates that the vehicle's urea consumption is basically consistent with the normal emission conditions, but the actual NOx conversion effect has significantly decreased. Therefore, the cause of the vehicle's excessive emissions is determined to be a poor urea quality fault.

[0079] In this case, the default normal operating temperature range of the SCR system is 200–500℃. If the vehicle's urea level is not lower than 0.5%, the NOx conversion efficiency is less than 70%, the SCR inlet temperature is between 200 and 500℃, and the urea consumption consistency coefficient K is greater than 0.8, then the cause of the vehicle's excessive emissions is determined to be a urea quality problem.

[0080] As shown in Table 1, this case study focuses on root cause diagnosis of vehicles A3 to A6, which were confirmed to have excessive emissions after initial assessment based on baseline thresholds and review of abnormal scores. The table lists the dual assessment results, average NOx conversion efficiency of the SCR, SCR inlet temperature, urea level, consistency coefficient K, key diagnostic evidence, and final root cause for each vehicle.

[0081] Table 1. Root cause diagnosis of typical vehicles with excessive emissions

[0082] (3) For each cause of excessive emissions, a standardized diagnostic report is generated and simultaneously pushed to the regulatory computer and the mobile terminal of law enforcement personnel. The report includes: basic vehicle information, facts of excessive emissions, root causes of excessive emissions, key parameter evidence, key points of on-site verification, and rectification suggestions.

[0083] (4) When law enforcement officers conduct on-site inspections, they can directly check the corresponding parts of the vehicle based on the root cause of the over-emission analysis from the computer processing terminal, and quickly verify the over-emission problem without disassembling and investigating the entire system. At the same time, they can directly issue clear rectification requirements to the vehicle owner based on the rectification suggestions in the diagnostic report, thus realizing a complete closed loop of "cloud analysis - on-site verification - on-site rectification".

[0084] In this case, when the vehicle's exhaust system malfunctioned due to poor urea quality, a sample of the urea solution in the vehicle's urea tank was taken to test the urea concentration and impurity content. Based on the test results, the vehicle owner was required to replace the urea solution with one that met the requirements.

[0085] When the vehicle's urea tank fails due to urea depletion, the remaining urea level in the tank is checked on-site, the urea level sensor status is verified, and the owner is asked to replenish the urea solution to meet the requirements.

[0086] When the problem is caused by a faulty SCR catalyst, an on-site inspection will be conducted to check whether the SCR catalyst is sintered, damaged, or has reduced conversion performance. Based on the inspection results, the vehicle owner will be required to repair or replace the SCR catalyst.

[0087] This embodiment has the following advantages: Significantly improve the effectiveness of emission supervision across the entire region. The two-dimensional over-emission risk assessment framework and vehicle over-emission risk classification mechanism established in this embodiment can precisely narrow the scope of regulatory attention from all vehicles registered in the network to the top 10% of high-risk vehicles, promote the targeted allocation of regulatory enforcement resources, effectively reduce enforcement costs, and significantly improve the detection rate of problematic vehicles.

[0088] This embodiment effectively improves the accuracy of identifying vehicles exceeding emission standards. It establishes a two-dimensional risk assessment framework for exceeding emission standards and a vehicle emission risk classification mechanism, which fully ensures the accuracy of identifying vehicles exceeding emission standards, significantly improves the problems of misjudging normal emissions under heavy-load conditions and missing hidden emission exceeding standards under light-load conditions, and the judgment results can meet the requirements of law enforcement compliance.

[0089] Strengthen on-site enforcement support capabilities for tracing the source of excessive emissions. The analytical strategy of "system-level positioning + component root cause positioning" established in this embodiment can output verifiable fault conclusions and rectification directions on-site, providing law enforcement personnel with targeted verification basis and constructing a full-process regulatory closed loop of "cloud analysis - on-site verification - on-site rectification".

[0090] It has excellent engineering adaptability. This solution is fully compatible with existing vehicle-mounted remote emission terminals and cloud platform computing and storage hardware systems, eliminating the need for large-scale installation of vehicle-mounted equipment, resulting in low deployment costs and easy large-scale application in the field of mobile source emission regulation.

[0091] For ease of explanation, the above description has been provided in conjunction with specific embodiments. However, the discussion in some embodiments is not intended to be exhaustive or to limit the embodiments to the specific forms disclosed above. Various modifications and variations can be obtained based on the above teachings. The selection and description of the above embodiments are for the purpose of better explaining the contents of this disclosure, thereby enabling those skilled in the art to better utilize the embodiments.

Claims

1. A system for assessing, identifying, and tracing the source of vehicle emissions exceeding limits, characterized in that: The system includes: A normal emission feature characterization module is configured to acquire driving data of vehicles with normal emission status and establish a benchmark statistical map of vehicle emission features under multi-dimensional subdivided scenarios based on the driving data. The vehicle over-emission risk assessment module is configured to acquire driving data of all vehicles in the platform, calculate the comprehensive over-emission risk of all vehicles based on the driving data, set a vehicle over-emission risk classification mechanism based on the comprehensive over-emission risk, and filter out high-risk vehicles from all vehicles in the platform based on the vehicle over-emission risk classification mechanism. The vehicle-by-vehicle emission precision identification module is configured to establish a dual emission status determination mechanism for high-risk vehicles; determine whether the final emission status of the vehicle is excessive based on the dual emission status determination mechanism; the dual emission status determination mechanism includes a baseline threshold initial judgment and anomaly score verification; different types of vehicles correspond to different vehicle excessive emission risk classification mechanisms; The over-emission cause tracing and analysis module is configured to establish an analysis strategy, including system-level positioning and component root cause positioning, for vehicles identified as over-emission vehicles, and to trace the over-emission sources of vehicles based on the analysis strategy, generating a standardized diagnostic report; the standardized diagnostic report is used to guide on-site law enforcement.

2. The vehicle over-emission risk assessment, over-emission identification, and source tracing system according to claim 1, characterized in that, The normal emission characteristic characterization module is also configured to: The driving data is preprocessed to remove data from periods of malfunction or abnormal operation. The preprocessed data was segmented using a sliding time window method to obtain several driving samples; Establish the multi-dimensional segmented scenarios that consider vehicle cargo status, vehicle operating environment and vehicle driving conditions, and classify all driving samples into the corresponding segmented scenarios in the multi-dimensional segmented scenarios; For each of the aforementioned sub-scenarios, based on the normal distribution theory, the upper limit of the confidence interval for the nitrogen oxide emission values ​​of all driving samples in the corresponding sub-scenarios is calculated, and this is used as the emission benchmark threshold for the corresponding sub-scenarios, thereby forming the vehicle emission characteristic benchmark statistical map.

3. The vehicle over-emission risk assessment, over-emission identification, and source tracing system according to claim 2, characterized in that, The vehicle over-emission risk assessment module is also configured to: Historical remote transmission data of the target vehicle within a certain period of time is obtained and divided into units of natural months. The historical remote transmission data of each month is processed into driving samples and assigned to the corresponding sub-scenarios. Within each calendar month, the percentage of driving samples whose nitrogen oxide emissions exceed the emission benchmark threshold for each of the aforementioned sub-scenarios is calculated to obtain the monthly single-scenario risk score. The average score of all the subdivided scenarios is calculated to obtain the monthly risk score for each month, and this is used to form the annual monthly risk score sequence for the target vehicle. The monthly risk scores in the annual monthly risk score sequence are assigned decreasing weights according to their time proximity and then summed in a weighted manner to obtain the comprehensive over-emission risk of the target vehicle.

4. The vehicle over-emission risk assessment, over-emission identification, and source tracing system according to claim 1, characterized in that, The mechanism for classifying vehicle over-emission risks also includes: Vehicles whose overall emission risk does not reach the high-risk level, but whose overall emission risk has increased continuously over the past three months and whose rate of increase exceeds the preset warning threshold, are included in the high-risk level.

5. The vehicle over-emission risk assessment, over-emission identification, and source tracing system according to claim 2, characterized in that, The vehicle-by-vehicle emission precision identification module is also configured to: In the initial judgment stage of the benchmark threshold, the driving samples of the vehicle to be judged are divided into the corresponding sub-scenarios, the average risk score of all sub-scenarios is calculated and compared with the first preset threshold; when the average risk score is greater than the first preset threshold, the corresponding vehicle is initially judged as over-emission. In the anomaly score review stage, the anomaly scores of all driving samples of the vehicle to be judged are calculated using a pre-built semi-supervised isolated forest model. When the proportion of driving samples with abnormal scores greater than the second preset threshold exceeds the third preset threshold, the corresponding vehicle will be re-evaluated as exceeding the emission limit; when a vehicle is judged as exceeding the emission limit in both the initial judgment stage of the baseline threshold and the abnormal score review stage, the vehicle will be judged as exceeding the emission limit.

6. The vehicle over-emission risk assessment, over-emission identification, and source tracing system according to claim 5, characterized in that, The method for constructing the semi-supervised isolated forest model includes: A basic isolated forest was constructed using the driving samples with normal emission status to characterize the distribution of normal emissions; Semi-supervised threshold calibration is performed on the basic isolated forest using driving samples with abnormal emission status; through an optimization algorithm, the recall rate of abnormal samples is maximized while ensuring that the recognition accuracy of normal samples is not lower than a preset value, thereby determining the second preset threshold. Establish a periodic update mechanism for the semi-supervised isolated forest model, and periodically use newly acquired law enforcement confirmation of over-emission data and re-inspection qualified data to iteratively optimize the second preset threshold or the structure of the semi-supervised isolated forest model.

7. The vehicle over-emission risk assessment, over-emission identification, and source tracing system according to claim 1, characterized in that, The over-discharge cause tracing and analysis module is also configured as follows: In system-level positioning, the average value of the nitrogen oxide conversion efficiency of the aftertreatment selective catalytic reduction system of the vehicle with excessive emissions is calculated; when the average value is lower than the theoretical value, the cause of excessive emissions is determined to be an abnormality of the aftertreatment selective catalytic reduction system. In the root cause localization of components, for vehicles that are determined to have abnormal selective catalytic reduction systems in the aftertreatment, the causes of excessive emissions are further subdivided into urea depletion faults, selective catalytic reduction catalyst faults, or poor urea quality faults based on the characteristic distribution of urea level, selective catalytic reduction inlet temperature, and nitrogen oxide conversion efficiency.

8. The vehicle over-emission risk assessment, over-emission identification, and source tracing system according to claim 7, characterized in that, The causes of excessive emissions can be further categorized into urea depletion failures, including: Calculate the percentage of driving samples where the urea level of the over-emission vehicle is below a preset low level threshold; if this state occurs continuously and the duration or number of consecutive samples exceeds the preset threshold, then the cause of the over-emission of the vehicle is determined to be the urea depletion fault.

9. A vehicle over-emission risk assessment, over-emission identification, and source tracing system according to claim 7, characterized in that, The causes of over-emission can be further categorized into selective catalytic reduction catalyst failure, including: For vehicles not identified as having urea depletion issues, if their nitrogen oxide conversion efficiency is lower than the fourth preset threshold, and the inlet temperature of the aftertreatment selective catalytic reduction system is higher than the fifth preset threshold and the duration exceeds the sixth preset threshold, then the cause of the excessive emissions will be identified as a failure of the selective catalytic reduction catalyst.

10. A vehicle over-emission risk assessment, over-emission identification, and source tracing system according to claim 7, characterized in that, The subdivision of excessive discharge causes into urea quality problems includes: For vehicles that are not identified as having a urea depletion fault or a selective catalytic reduction catalyst fault, if the urea level is not lower than a preset low level threshold, the nitrogen oxide conversion efficiency is lower than a seventh preset threshold, the selective catalytic reduction inlet temperature is within a preset normal operating temperature range, and the vehicle's urea consumption state is highly consistent with that of a normal vehicle, then the vehicle is identified as having a poor urea quality fault.