Vehicle emission abnormality detection and tracing method, device and electronic equipment

Abstract

The present invention relates to the field of transportation technology, and in particular to a method, device, and electronic device for detecting and tracing vehicle emissions anomalies, comprising: inputting the current operating conditions and current operating data into a dynamic threshold model to obtain a first detection result, and inputting them into an emission prediction model to obtain a second detection result; when the first detection result and the second detection result do not meet the preset conditions for determining vehicle emissions anomalies, fusing the first detection result and the second detection result to obtain a final detection result; when the final detection result is greater than or equal to a preset threshold, determining that the vehicle emissions are abnormal; and using a Cosine similarity calculation strategy based on a vehicle emissions anomaly knowledge graph library to obtain the cause and probability of the emissions anomaly. This method solves the problems of traditional vehicle emissions anomaly detection and tracing methods, such as low emission anomaly detection efficiency, high misjudgment rate, and difficulty in tracing, achieves accurate detection of emissions anomalies, completes closed-loop management from anomaly triggering to root cause tracing, and improves intelligent operation and maintenance.

Description

Technical Field

[0001] The present invention relates to the field of transportation technology, and in particular to a method, device and electronic equipment for detecting and tracing vehicle emission anomalies. Background Art

[0002] Traditional methods for detecting abnormal vehicle emissions often rely on single threshold detection (such as fixed nitrogen oxide (NOx) limits) or manual judgment, which presents the following technical challenges:

[0003] (1) Data fragmentation and inefficient analysis: Emission anomalies involve the coupling of multiple components such as the engine and post-treatment system (such as the selective catalytic reduction system SCR and the diesel particulate filter DPF). Traditional methods rely on scattered sensor data or manual experience, and it is difficult to integrate multi-source heterogeneous information such as maintenance cases, real-time data from vehicle terminals, and literature knowledge, resulting in low efficiency in locating the root cause of the fault.

[0004] (2) Poor adaptability of static thresholds: Existing threshold models often use fixed limits, ignoring the dynamic differences in the vehicle's actual operating conditions (such as urban congestion, high-speed cruising, and cold start at low temperatures), which can easily lead to false alarms or missed detections due to fluctuations in operating conditions.

[0005] (3) Insufficient causal reasoning capabilities: The root cause of abnormal emissions and the relationship between the manifestations are complex (e.g., "urea pump blockage → NOx exceeding the standard" requires multi-level transmission). Traditional rule engines or statistical methods are difficult to model deep causal relationships, and the traceability accuracy is low.

[0006] (4) Knowledge update lags: Maintenance experience and new fault cases rely on manual induction, and the knowledge base update cycle is long, which cannot adapt to the rapid iteration of emission control technology. Summary of the Invention

[0007] The present invention provides a method, device and electronic equipment for detecting and tracing vehicle emissions anomalies to solve the problems of traditional vehicle emissions anomaly detection and tracing methods, such as low emission anomaly detection efficiency, high misjudgment rate and difficulty in tracing due to data fragmentation and complex causal relationships.

[0008] The first aspect of the present invention provides a method for detecting and tracing vehicle emission anomalies, comprising the following steps: obtaining the current operating condition and current operating data of a vehicle; inputting the current operating condition and the current operating data into a preset dynamic threshold model to obtain a first detection result, and inputting the current operating condition and the current operating data into a preset emission prediction model to obtain a second detection result, wherein the preset dynamic threshold model is constructed by multiple preset operating conditions and a preset emission abnormality threshold corresponding to each preset operating condition, and the preset emission prediction model is obtained by training a long short-term memory network (LSTM) neural network with historical operating data of the vehicle; when the first detection result and the second detection result do not meet the preset conditions for determining vehicle emission anomalies, the first detection result and the second detection result are fused to obtain a final detection result, and when the final detection result is greater than or equal to the preset threshold, the vehicle emission anomaly is determined, and based on a preset vehicle emission anomaly knowledge graph library, the cause and probability of the vehicle emission anomaly are obtained by using a Cosine similarity calculation strategy.

[0009] Optionally, the first detection result includes the actual emissions of the vehicle. When judging whether the first detection result meets the preset conditions for determining vehicle emissions abnormality, it also includes: judging whether the actual emissions of the vehicle are greater than a preset emissions abnormality threshold, and whether the first duration of the actual emissions of the vehicle being greater than the preset emissions abnormality threshold is greater than a first preset duration, wherein the preset emissions abnormality threshold is obtained by the current operating conditions; if the actual emissions of the vehicle are greater than the preset emissions abnormality threshold and the first duration is greater than the first preset duration, then it is determined that the first detection result meets the preset conditions for determining vehicle emissions abnormality, and the vehicle emissions are determined to be abnormal, and based on the preset vehicle emissions abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality; otherwise, it is determined that the first detection result does not meet the preset conditions for determining vehicle emissions abnormality.

[0010] Optionally, before determining whether the first detection result is greater than the preset emission abnormality threshold, it also includes: obtaining historical operating data of the vehicle; constructing the multiple preset operating conditions based on data features in the historical operating data; establishing a mapping relationship between the historical operating data of each preset operating condition and the emission abnormality threshold, so as to obtain the preset emission abnormality threshold according to the current operating condition based on the mapping relationship.

[0011] Optionally, the second detection result includes the actual emissions of the vehicle. After the current operating conditions and the current operating data are input into the preset emission prediction model to obtain the second detection result, it also includes: judging whether the ratio of the actual emissions of the vehicle to the emission prediction value is greater than the preset ratio, and whether the second duration of the actual emissions of the vehicle is greater than the emission prediction value is greater than the second preset duration; if the ratio of the actual emissions of the vehicle to the emission prediction value is greater than the preset ratio, and the second duration is greater than the second preset duration, then it is determined that the second detection result meets the preset vehicle emission abnormality determination condition, and the vehicle emissions are determined to be abnormal, and based on the preset vehicle emission abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emission abnormality; otherwise, it is determined that the second detection result does not meet the preset vehicle emission abnormality determination condition.

[0012] Optionally, the method of obtaining the cause and probability of vehicle emission abnormality based on a preset vehicle emission abnormality knowledge graph library using a Cosine similarity calculation strategy includes: obtaining historical operation data of the vehicle; based on the historical operation data, using the Cosine similarity calculation strategy to calculate the data features of each relationship chain in the preset vehicle emission abnormality knowledge graph library respectively to obtain multiple Cosine similarity calculation results; sorting the multiple Cosine similarity calculation results based on a preset sorting strategy, and using the Cosine similarity calculation results in the sorting results that meet the preset ranking conditions as the cause and probability of the vehicle emission abnormality.

[0013] Optionally, before obtaining the cause and probability of vehicle emission anomaly based on a preset vehicle emission anomaly knowledge graph library and using a Cosine similarity calculation strategy, the method includes: obtaining vehicle emission anomaly cause data; forming a vehicle emission anomaly cause data relationship chain based on the vehicle emission anomaly cause data; using a preset knowledge graph theory to convert the vehicle emission anomaly cause data relationship chain into entities and relationships in a knowledge graph; and using an entity alignment strategy to perform entity fusion on relatively conflicting entities to obtain the preset vehicle emission anomaly knowledge graph library.

[0014] Optionally, fusing the first detection result and the second detection result to obtain a final detection result includes: using a preset fusion formula to fuse the first detection result and the second detection result to obtain the final detection result, wherein the preset fusion formula is:

[0015] ;

[0016] in, For the final test results,E car is the actual emission of the vehicle, E Th is the preset emission abnormality threshold, E Pv is the emission prediction value, a is the weight of the first detection result, b is the weight of the second detection result.

[0017] Optionally, after determining that the vehicle emissions are abnormal and obtaining the cause and probability of the vehicle emissions abnormality based on a preset vehicle emissions abnormality knowledge graph library using a Cosine similarity calculation strategy, it includes: sending vehicle emissions abnormality warning information to a preset terminal; based on the vehicle emissions abnormality warning information, the cause and probability of the vehicle emissions abnormality, using an AI model to generate vehicle maintenance suggestions.

[0018] The second aspect of the present invention provides a vehicle emission anomaly detection and tracing device, including: an acquisition module for acquiring the current operating condition and current operating data of a vehicle; a detection module for inputting the current operating condition and the current operating data into a preset dynamic threshold model to obtain a first detection result, and inputting the current operating condition and the current operating data into a preset emission prediction model to obtain a second detection result, wherein the preset dynamic threshold model is constructed by multiple preset operating conditions and a preset emission abnormality threshold corresponding to each preset operating condition, and the preset emission prediction model is obtained by training an LSTM neural network with historical operating data of the vehicle; a tracing module for fusing the first detection result and the second detection result to obtain a final detection result when the first detection result and the second detection result do not meet the preset vehicle emission abnormality judgment conditions, and when the final detection result is greater than or equal to the preset threshold, judging the vehicle emission abnormality, and based on a preset vehicle emission abnormality knowledge graph library, using the Cosine similarity calculation strategy, obtaining the cause and probability of the vehicle emission abnormality.

[0019] Optionally, the first detection result includes the actual emissions of the vehicle. When judging whether the first detection result meets the preset conditions for determining vehicle emissions abnormality, the detection module is further used to: judge whether the actual emissions of the vehicle are greater than a preset emissions abnormality threshold, and whether the first duration of the actual emissions of the vehicle being greater than the preset emissions abnormality threshold is greater than a first preset duration, wherein the preset emissions abnormality threshold is obtained by the current operating conditions; if the actual emissions of the vehicle are greater than the preset emissions abnormality threshold and the first duration is greater than the first preset duration, then it is determined that the first detection result meets the preset conditions for determining vehicle emissions abnormality, and the vehicle emissions are determined to be abnormal, and based on the preset vehicle emissions abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality; otherwise, it is determined that the first detection result does not meet the preset conditions for determining vehicle emissions abnormality.

[0020] Optionally, before determining whether the first detection result is greater than the preset emission abnormality threshold, the detection module is further used to: obtain historical operating data of the vehicle; construct the multiple preset operating conditions based on data features in the historical operating data; establish a mapping relationship between the historical operating data of each preset operating condition and the emission abnormality threshold, so as to obtain the preset emission abnormality threshold according to the current operating condition based on the mapping relationship.

[0021] Optionally, the second detection result includes the actual emissions of the vehicle. When judging whether the second detection result meets the preset conditions for determining vehicle emissions abnormality, the detection module is further used to: judge whether the ratio of the actual emissions of the vehicle to the emission prediction value is greater than the preset ratio, and whether the second duration of the actual emissions of the vehicle being greater than the emission prediction value is greater than the second preset duration; if the ratio of the actual emissions of the vehicle to the emission prediction value is greater than the preset ratio, and the second duration is greater than the second preset duration, then it is determined that the second detection result meets the preset conditions for determining vehicle emissions abnormality, and the vehicle emissions are determined to be abnormal, and based on the preset vehicle emissions abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality; otherwise, it is determined that the second detection result does not meet the preset conditions for determining vehicle emissions abnormality.

[0022] Optionally, the traceability module is further used to: obtain historical operating data of the vehicle; based on the historical operating data, use the Cosine similarity calculation strategy to calculate the data features of each relationship chain in the preset vehicle emission abnormality knowledge graph library, and obtain multiple Cosine similarity calculation results; based on a preset sorting strategy, sort the multiple Cosine similarity calculation results, and use the Cosine similarity calculation results in the sorting results that meet the preset ranking conditions as the cause and probability of the vehicle emission abnormality.

[0023] Optionally, before obtaining the cause and probability of vehicle emission anomaly based on a preset vehicle emission anomaly knowledge graph library and using a Cosine similarity calculation strategy, the tracing module is further used to: obtain vehicle emission anomaly cause data; form a vehicle emission anomaly cause data relationship chain based on the vehicle emission anomaly cause data; use a preset knowledge graph theory to convert the vehicle emission anomaly cause data relationship chain into entities and relationships in the knowledge graph; use an entity alignment strategy to perform entity fusion on relatively conflicting entities to obtain the preset vehicle emission anomaly knowledge graph library.

[0024] Optionally, the traceability module is further configured to: use a preset fusion formula to fuse the first detection result and the second detection result to obtain the final detection result, wherein the preset fusion formula is:

[0025] ;

[0026] in, For the final test results, E car is the actual emission of the vehicle, E Th is the preset emission abnormality threshold, E Pv is the emission prediction value, a is the weight of the first detection result, b is the weight of the second detection result.

[0027] Optionally, after determining that the vehicle emissions are abnormal and obtaining the cause and probability of the vehicle emissions abnormality based on a preset vehicle emissions abnormality knowledge graph library using a Cosine similarity calculation strategy, the traceability module is also used to: send vehicle emissions abnormality warning information to a preset terminal; and generate vehicle maintenance suggestions based on the vehicle emissions abnormality warning information, the cause and probability of the vehicle emissions abnormality using an AI model.

[0028] An embodiment of the third aspect of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the vehicle emission anomaly detection and tracing method as described in the above embodiment.

[0029] A fourth aspect of the present invention provides a computer program product having a computer program stored thereon, which is executed by a processor to implement the vehicle emission anomaly detection and tracing method as described in the above embodiments.

[0030] In the above embodiment, the current operating condition and current operating data are input into a preset dynamic threshold model to obtain a first detection result, and the current operating condition and current operating data are input into a preset emission prediction model to obtain a second detection result. The preset dynamic threshold model is constructed from multiple preset operating conditions and preset emission abnormality thresholds corresponding to each preset operating condition, and the preset emission prediction model is obtained by training an LSTM neural network using the vehicle's historical operating data. When the first and second detection results do not meet the preset conditions for determining vehicle emission abnormality, the first and second detection results are fused to obtain a final detection result. When the final detection result is greater than or equal to the preset threshold, the vehicle emissions are determined to be abnormal. Based on the vehicle emission abnormality knowledge graph, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emission abnormality. This solves the problems of traditional vehicle emission abnormality detection and tracing methods, which suffer from low emission abnormality detection efficiency, high false positive rate, and difficulty in tracing due to data fragmentation and complex causal relationships. It achieves accurate detection of emission abnormalities and completes closed-loop management from abnormality triggering to root cause tracing, achieving the goals of reducing vehicle emission pollution, improving the level of intelligent operation and maintenance, and ensuring environmental compliance.

[0031] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0033] Figure 1 A flowchart of a method for detecting and tracing vehicle emission anomalies according to an embodiment of the present invention;

[0034] Figure 2 (a) is a flow chart of a detection mechanism of a dynamic threshold model and an LSTM machine learning model according to one embodiment of the present invention. Figure 2(b) is a flowchart of establishing a collaborative detection mechanism integrating a dynamic threshold model and an LSTM machine learning model according to one embodiment of the present invention;

[0035] Figure 3 A flowchart of a method for detecting and tracing vehicle emissions anomalies according to another embodiment of the present invention;

[0036] Figure 4 Schematic diagram of a knowledge graph of causes of abnormal vehicle emissions according to one embodiment of the present invention;

[0037] Figure 5 This is an example diagram of a vehicle emission anomaly detection and source tracing device according to an embodiment of the present invention;

[0038] Figure 6 FIG. 1 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0040] The following describes a method, device, and electronic device for detecting and tracing vehicle emissions anomalies according to an embodiment of the present invention with reference to the accompanying drawings. In response to the problems of the conventional vehicle emissions anomaly detection and tracing method mentioned in the background art, which suffer from low efficiency, high misjudgment rate, and difficulty in tracing emissions anomaly due to data fragmentation and complex causal relationships, the present invention provides a method for detecting and tracing vehicle emissions anomalies. In this method, the current operating condition and current operating data are input into a preset dynamic threshold model to obtain a first detection result, and the current operating condition and current operating data are input into a preset emission prediction model to obtain a second detection result. The preset dynamic threshold model is constructed by a plurality of preset operating conditions and a preset emission anomaly threshold corresponding to each preset operating condition, and the preset emission prediction model is obtained by training an LSTM neural network with historical vehicle operating data. When the first detection result and the second detection result do not meet the preset conditions for determining vehicle emissions anomaly, the first detection result and the second detection result are fused to obtain a final detection result. When the final detection result is greater than or equal to the preset threshold, the vehicle emissions are determined to be abnormal. Based on a vehicle emissions anomaly knowledge graph, the cause and probability of the vehicle emissions anomaly are determined using a Cosine similarity calculation strategy. This solves the problems of traditional vehicle emission anomaly detection and tracing methods, such as low efficiency, high misjudgment rate and difficult tracing due to data fragmentation and complex causal relationships. It achieves accurate detection of emission anomalies and completes closed-loop management from anomaly triggering to root cause tracing, achieving the goals of reducing vehicle emission pollution, improving the level of intelligent operation and maintenance, and ensuring environmental compliance.

[0041] Specifically, Figure 1 A flow chart of a method for detecting and tracing vehicle emission anomalies provided by an embodiment of the present invention.

[0042] like Figure 1 As shown, the vehicle emission abnormality detection and tracing method includes the following steps:

[0043] In step S101 , the current operating condition and current operating data of the vehicle are acquired.

[0044] Among them, the current operating data includes vehicle speed, acceleration, engine net output torque, engine speed, engine fuel flow, engine coolant temperature, DPF pressure difference, DPF regeneration frequency, ambient temperature, intake pressure, exhaust temperature, exhaust back pressure, air-fuel ratio, exhaust gas recirculation EGR valve opening, urea injection amount, particulate matter concentration and SCR downstream NOx transmitter output value data. The current operating condition is one of urban and suburban conditions, high-speed cruising conditions, low-temperature cold start conditions, and climbing conditions.

[0045] In step S102, the current operating condition and the current operating data are input into a preset dynamic threshold model to obtain a first detection result, and the current operating condition and the current operating data are input into a preset emission prediction model to obtain a second detection result, wherein the preset dynamic threshold model is constructed by multiple preset operating conditions and a preset emission abnormality threshold corresponding to each preset operating condition, and the preset emission prediction model is obtained by training an LSTM neural network with the vehicle's historical operating data.

[0046] Among them, the preset emission prediction model is also an LSTM model.

[0047] Optionally, in some embodiments, before determining whether the first detection result is greater than a preset emission abnormality threshold, it also includes: obtaining historical operating data of the vehicle; constructing multiple preset operating conditions based on data features in the historical operating data; establishing a mapping relationship between the historical operating data of each preset operating condition and the emission abnormality threshold, so as to obtain the preset emission abnormality threshold according to the current operating condition based on the mapping relationship.

[0048] Specifically, the steps of constructing a preset dynamic threshold model are:

[0049] The first step is to establish four types of preset operating conditions: urban and suburban conditions (vehicle speed less than 50 km / h, idle ratio greater than 18%), high-speed cruising conditions (vehicle speed greater than 70 km / h, accelerator pedal opening fluctuation rate less than 10%), climbing conditions (engine torque greater than 70%, slope greater than 6%), and low-temperature cold start conditions (coolant temperature less than 40°C, ambient temperature less than 10°C).

[0050] In the second step, the historical operating data was processed using the 30-second moving window averaging method. The data was divided into urban and suburban operating conditions, high-speed cruising conditions, climbing conditions, and low-temperature cold start conditions according to data characteristics. The 95th percentile of NOx emissions under each operating condition was calculated as the preset emission anomaly threshold (for example, the NOx emission anomaly threshold for urban and suburban conditions is 500 ppm). This established a mapping relationship between the historical operating data for each preset operating condition and the emission anomaly threshold.

[0051] Optionally, in some embodiments, the first detection result includes the actual emissions of the vehicle. When judging whether the first detection result meets the preset conditions for determining vehicle emissions abnormality, it also includes: judging whether the actual emissions of the vehicle are greater than a preset emissions abnormality threshold, and whether the first duration of the actual emissions of the vehicle being greater than the preset emissions abnormality threshold is greater than a first preset duration, wherein the preset emissions abnormality threshold is obtained by the current operating conditions; if the actual emissions of the vehicle are greater than the preset emissions abnormality threshold and the first duration is greater than the first preset duration, then it is determined that the first detection result meets the preset conditions for determining vehicle emissions abnormality, and the vehicle emissions are determined to be abnormal, and based on the preset vehicle emissions abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality; otherwise, it is determined that the first detection result does not meet the preset conditions for determining vehicle emissions abnormality.

[0052] In the embodiment of the present invention, the first preset duration is set to 2 hours.

[0053] It should be understood that the current operating condition of the vehicle is determined based on the current operating data, and the preset emission abnormality threshold corresponding to the current operating condition is determined based on the mapping relationship. If the actual emissions of the vehicle are greater than the preset emission abnormality threshold corresponding to the current operating condition, and the first duration is greater than 2 hours, then the first detection result is determined to meet the preset conditions for determining vehicle emission abnormality, and the vehicle emission abnormality is directly determined. Further, based on the preset vehicle emission abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emission abnormality. Otherwise, it is determined that the first detection result does not meet the preset conditions for determining vehicle emission abnormality, that is, the first detection result cannot directly determine the vehicle emission abnormality. Among them, based on the preset vehicle emission abnormality knowledge graph library, the steps of obtaining the cause and probability of vehicle emission abnormality using the Cosine similarity calculation strategy are specifically explained below.

[0054] For example, if the actual emissions of a vehicle exceed the preset emission abnormality threshold corresponding to the current operating conditions for two consecutive hours during operation, there is no need to integrate the second detection result for judgment, and the vehicle emissions can be directly judged as abnormal.

[0055] Optionally, in some embodiments, the second detection result includes the actual emissions of the vehicle, and determining whether the second detection result meets the preset conditions for determining vehicle emissions abnormality includes: determining whether the ratio of the vehicle's actual emissions to the emission prediction value is greater than the preset ratio, and whether the second duration of the vehicle's actual emissions being greater than the emission prediction value is greater than the second preset duration; if the ratio of the vehicle's actual emissions to the emission prediction value is greater than the preset ratio, and the second duration is greater than the second preset duration, then determining that the second detection result meets the preset conditions for determining vehicle emissions abnormality, and determining that the vehicle emissions are abnormal, and based on the preset vehicle emissions abnormality knowledge graph library, using the Cosine similarity calculation strategy, obtaining the cause and probability of the vehicle emissions abnormality, otherwise, determining that the second detection result does not meet the preset conditions for determining vehicle emissions abnormality.

[0056] In the embodiment of the present invention, the second preset duration is set to 2 hours, and the preset ratio is set to 1.5.

[0057] Specifically, the steps of building a preset emission prediction model and detecting whether vehicle emissions are abnormal are as follows:

[0058] The first step is to use the historical operating data obtained as training data, with the vehicle speed, acceleration, torque, speed, and fuel consumption rate in the historical operating data as input, and the NOx emission rate as output. Utilizing the LSTM neural network, we train and build an LSTM machine learning model to predict vehicle emissions.

[0059] In the second step, during the operation of the vehicle, the preset emission prediction model continuously predicts the emission prediction value. If the ratio of the vehicle's actual emissions to the emission prediction value is greater than 1.5 for two consecutive hours, the second detection result is determined to meet the preset vehicle emission abnormality determination conditions. There is no need to integrate the second detection result with the first detection result. The heavy-duty vehicle emission abnormality is directly determined. Based on the preset vehicle emission abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emission abnormality. Otherwise, it is determined that the second detection result does not meet the preset vehicle emission abnormality determination conditions. In other words, the second detection result cannot directly determine the vehicle emission abnormality.

[0060] In summary, the preset emission prediction model and the preset dynamic threshold model can independently determine whether vehicle emissions are abnormal.

[0061] In step S103, when the first detection result and the second detection result do not meet the preset conditions for determining vehicle emissions abnormality, the first detection result and the second detection result are fused to obtain a final detection result. When the final detection result is greater than or equal to a preset threshold, the vehicle emissions are determined to be abnormal. Based on the preset vehicle emissions abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality.

[0062] It should be noted that the fact that the first test result and the second test result do not meet the preset conditions for determining vehicle emissions abnormality can be understood as the first test result cannot directly determine vehicle emissions abnormality and the second test result cannot directly determine vehicle emissions abnormality. At this time, the first test result and the second test result need to be integrated.

[0063] Optionally, in some embodiments, fusing the first detection result and the second detection result to obtain the final detection result includes: using a preset fusion formula to fuse the first detection result and the second detection result to obtain the final detection result, wherein the preset fusion formula is:

[0064] ; (1)

[0065] in, For the final test results, E car is the actual emission of the vehicle, E Th is the preset emission abnormality threshold, E Pv is the emission prediction value, a is the weight of the first detection result, b is the weight of the second detection result.

[0066] In formula (1), parameter a defaults to 0.6, and parameter b defaults to 0.4. If the final detection result is greater than or equal to the preset threshold c, the vehicle is considered to have an abnormal emission condition, triggering an abnormal emission source analysis. The preset threshold c needs to balance false positives and false negatives. Therefore, the preset threshold c is set to 1.2 by default to cover slight fluctuations while avoiding oversensitivity, thereby effectively preventing false positives and false negatives.

[0067] Specifically, the embodiment of the present invention establishes a collaborative detection mechanism of a preset dynamic threshold model and a preset emission prediction model to accurately trigger the source analysis of emission anomalies. The specific steps are as follows:

[0068] In the first step, the preset dynamic threshold model and the preset emission prediction model are used to detect whether there are abnormal vehicle emissions. If the vehicle emissions are determined to be abnormal, the emission abnormality tracing analysis is directly triggered.

[0069] When the first detection result and the second detection result do not meet the preset conditions for determining vehicle emissions are abnormal, it is necessary to fuse the first detection result output by the preset dynamic threshold model and the second detection result output by the preset emission prediction model.

[0070] In the second step, a preset fusion formula is used to fuse the first detection result output by the preset dynamic threshold model and the second detection result output by the preset emission prediction model to obtain a final detection result.

[0071] For example, Figure 2 As shown in (a), if the actual emission of the vehicle E car >Preset emission abnormality threshold E Th If the first duration of the vehicle emission is greater than the first preset duration, the vehicle emission is directly determined to be abnormal, and the abnormal emission source analysis is directly triggered. Otherwise, the vehicle emission is not directly determined to be abnormal. If the actual emission of the vehicle E car >Emission forecast E Pv If the second duration of the vehicle emission is greater than the second preset duration, the vehicle emission is directly determined to be abnormal, and the abnormal emission source analysis is directly triggered; otherwise, the vehicle emission is not directly determined to be abnormal;

[0072] If the actual emission amount is not one of the above two situations, that is, the first test result and the second test result do not meet the preset emission abnormality determination conditions, that is, the vehicle emission abnormality is not determined by the first test result and the second test result, then it is necessary to merge the first test result and the second test result to obtain the final test result, such as Figure 2 As shown in (b), if the final test result If the value is greater than or equal to the preset threshold value c, the vehicle’s emissions are judged to be abnormal, directly triggering the emission abnormality tracing analysis. If the final test result is less than the preset threshold value c, the vehicle’s emissions are judged to be normal.

[0073] It can be seen from this that the preset dynamic threshold model and the preset emission prediction model can independently detect whether there is an abnormal emission situation in the vehicle, and can also integrate the results of the separate detections to determine whether there is an abnormal emission situation. A collaborative detection mechanism of the preset dynamic threshold model and the preset emission prediction model is established, which effectively and reliably supports the detection of abnormal emissions of heavy-duty vehicles and avoids false alarms and missed reports of a single model.

[0074] Optionally, in some embodiments, before obtaining the cause and probability of vehicle emission anomaly based on a preset vehicle emission anomaly knowledge graph library using a Cosine similarity calculation strategy, the method includes: obtaining vehicle emission anomaly cause data; forming a vehicle emission anomaly cause data relationship chain based on the vehicle emission anomaly cause data; using a preset knowledge graph theory to convert the vehicle emission anomaly cause data relationship chain into entities and relationships in the knowledge graph; using an entity alignment strategy to perform entity fusion on relatively conflicting entities to obtain a preset vehicle emission anomaly knowledge graph library.

[0075] The historical operation data of the vehicle is obtained through remote monitoring, and cluster analysis is performed using the K-Means clustering method to obtain multi-dimensional cluster centers. The cluster centers are used to represent each relationship chain in the preset vehicle emission anomaly knowledge graph library.

[0076] First, historical operating data is obtained, which includes vehicle speed, acceleration, engine net output torque, engine speed, engine fuel flow, engine coolant temperature, DPF pressure difference, DPF regeneration frequency, ambient temperature, intake pressure, exhaust temperature, exhaust back pressure, air-fuel ratio, EGR valve opening, urea injection amount, particulate matter concentration and SCR downstream NOx transmitter output value parameters.

[0077] Then, using historical operation data as input, the K-Means clustering method is used to perform cluster analysis, obtain multi-dimensional cluster centers, and extract the multi-dimensional features of each cluster center.

[0078] Finally, based on the multi-dimensional characteristics of each cluster center, it is connected with each relationship chain in the preset vehicle emission anomaly knowledge graph library to support the tracing of the cause of heavy-duty vehicle emission anomalies.

[0079] Among them, the steps of using knowledge graph theory to construct a preset vehicle emission anomaly knowledge graph library are as follows:

[0080] First, data on the causes of abnormal vehicle emissions are obtained through multiple channels and dimensions. The data sources include: automobile maintenance manuals, heavy-duty vehicle maintenance case data, operation and emission data collected by heavy-duty vehicle onboard terminals, and existing technical literature.

[0081] Then, from the data on the causes of abnormal vehicle emissions, the frequent correlation patterns between the root causes of the abnormal emissions and the manifestations of the abnormal emissions are extracted to form a relationship chain of the causes of abnormal vehicle emissions (for example, this relationship chain can be expressed as: "urea pump blockage" → "reduced urea injection volume" → "reduced SCR catalytic efficiency" → "NOx exceeds the standard").

[0082] Subsequently, knowledge graph theory was used to transform the cause chain of vehicle emissions anomalies into entities and relationships within the knowledge graph. Based on expert knowledge, entity alignment technology was used to resolve the issue of identical entities with different names (for example, "SCR catalytic efficiency anomaly" and "SCR system catalytic rate reduction" both refer to the entity "SCR catalytic efficiency reduction"). Entity alignment technology was used to fuse conflicting entities, achieving a unified knowledge representation within the knowledge graph.

[0083] Finally, the above knowledge graph is stored in a graph database (such as Neo4j) to build a preset vehicle emission anomaly knowledge graph library to support efficient path query and reasoning.

[0084] In addition, the preset vehicle emission anomaly knowledge graph library is dynamic and will be continuously optimized with the accumulation of new cases and maintenance experience. The knowledge base can be continuously improved based on maintenance records, emission test results, etc. to enhance the accuracy of the knowledge base.

[0085] Optionally, in some embodiments, based on a preset vehicle emission anomaly knowledge graph library, a Cosine similarity calculation strategy is used to obtain the cause and probability of vehicle emission anomaly, including: obtaining historical operation data of the vehicle; based on the historical operation data, using the Cosine similarity calculation strategy to calculate the data features of each relationship chain in the preset vehicle emission anomaly knowledge graph library respectively to obtain multiple Cosine similarity calculation results; based on a preset sorting strategy, sorting the multiple Cosine similarity calculation results, and using the Cosine similarity calculation results that meet the preset ranking conditions in the sorting results as the cause and probability of vehicle emission anomaly.

[0086] Specifically, based on a preset vehicle emission anomaly knowledge graph library, the present invention uses a Cosine similarity calculation strategy to calculate the potential causes of vehicle emission anomalies and their probability of occurrence, thereby realizing a multi-dimensional tracing and source analysis of vehicle emission anomalies.

[0087] First, remotely extract historical operating data where emission anomalies occurred;

[0088] Then, the Cosine similarity calculation strategy is used to calculate the data features of each relationship chain in the historical operation data of emission anomalies and the preset vehicle emission anomaly knowledge graph library respectively to obtain multiple Cosine similarity calculation results.

[0089] Based on a preset sorting strategy, multiple Cosine similarity calculation results are sorted, and the Cosine similarity calculation results that meet the preset ranking conditions in the sorting results are used as the causes and probabilities of vehicle emission abnormalities. Specifically, in an embodiment of the present invention, each high emission instance and each cause in a preset vehicle emission abnormality knowledge graph library are represented as vectors, and the Cosine similarity between them is calculated to measure the similarity between the instance and each cause.

[0090] A similarity threshold of 0.8 is set, and causes with a similarity greater than 0.8 to the instance are considered likely to contribute to high emissions. For each identified cause, a weighted average method is then used to calculate the probability of each cause contributing to high emissions, based on their similarity. Higher similarity values ​​are associated with greater weights, indicating a greater contribution to high emissions.

[0091] Therefore, the relationship chains from the multiple Cosine similarity calculation results are sorted by their similarity, and the top three relationship chains are selected as potential causes of vehicle emissions abnormality, and the occurrence probability of each potential cause is calculated. Here, the top three can also be understood as referring to the three most similar relationship chains from the multiple Cosine similarity calculation results.

[0092] Among them, the Cosine similarity calculation strategy specifically evaluates the similarity between two vectors by calculating the cosine value of the angle between them. Given two non-zero vectors a and b, the Cosine similarity is defined as: .

[0093] in, is a vector a and b The dot product of are vectors a and b The mold length.

[0094] Optionally, in some embodiments, after determining that vehicle emissions are abnormal and obtaining the cause and probability of the vehicle emissions abnormality based on a preset vehicle emissions abnormality knowledge graph library using a Cosine similarity calculation strategy, it includes: sending vehicle emissions abnormality warning information to a preset terminal; and generating vehicle maintenance suggestions using an AI model based on the vehicle emissions abnormality warning information, the cause and probability of the vehicle emissions abnormality.

[0095] Specifically, it sends vehicle emission abnormality warning information to the preset terminal to remind users or vehicle monitoring personnel of vehicle emission abnormalities, visualizes the potential causes of vehicle emission abnormalities and their probability of occurrence, and combines with large AI models such as Deepseek or ChatGPT to simultaneously generate vehicle maintenance suggestions, which supports viewing on PC and mobile terminals.

[0096] In summary, multi-dimensional tracing and source analysis of abnormal heavy-duty vehicle emissions can be achieved.

[0097] In order to enable those skilled in the art to further understand the vehicle emission abnormality detection and tracing method according to the embodiment of the present invention, the following is a detailed description with reference to specific embodiments. Figure 3 shown.

[0098] 1. Collect multi-source data and extract association patterns to form relationship chains.

[0099] First, data on the causes of abnormal vehicle emissions are obtained through multiple channels and dimensions. The data sources include: automobile maintenance manuals, heavy-duty vehicle maintenance case data, operation and emission data collected by heavy-duty vehicle onboard terminals, and existing technical literature.

[0100] For vehicle maintenance manuals, the team obtained all information on the causes of abnormal vehicle emissions from the Heavy Truck Maintenance Technical Manual. For heavy-duty vehicle maintenance case data, the team researched heavy-duty vehicle maintenance case data from a specific region over the past three years. For operational and emission data collected by onboard terminals, the team selected 500 heavy-duty vehicles with a cumulative mileage exceeding 50,000 kilometers and downloaded all OBD fault information generated by these vehicles since leaving the factory. For existing technical literature, the team searched and downloaded several academic papers that studied abnormal heavy-duty vehicle emissions.

[0101] From the data on causes of abnormal vehicle emissions, the frequent correlation patterns between the root causes of abnormal emissions and the manifestations of abnormal emissions are extracted to form a relationship chain of causes of abnormal vehicle emissions (for example, this relationship chain can be expressed as: "urea pump blockage" → "reduced urea injection volume" → "reduced SCR catalytic efficiency" → "NOx exceeds the standard").

[0102] 2. Build and optimize the knowledge graph and store it in the graph database to form a dynamic knowledge graph library.

[0103] Using knowledge graph theory, we transform the cause chain of vehicle emissions anomalies into entities and relationships within the knowledge graph. Based on expert knowledge, we use entity alignment to resolve the issue of identical entities with different names (for example, "abnormal SCR catalytic efficiency" and "decreased SCR system catalytic rate" both refer to the entity "decreased SCR catalytic efficiency"). We also use entity alignment to fuse conflicting entities, achieving a unified knowledge representation within the knowledge graph.

[0104] Finally, the above knowledge graph is stored in a graph database (such as Neo4j) to build a preset vehicle emission anomaly knowledge graph library, which supports efficient path query and reasoning. The completed vehicle emission anomaly cause knowledge graph is constructed. The vehicle emission anomaly knowledge graph library is as follows: Figure 4 shown.

[0105] 3. Obtain remote monitoring data for cluster analysis and establish a connection between cluster center features and knowledge graph relationships.

[0106] For example, the historical operating data of 500 heavy-duty vehicles with abnormal emissions for one month were remotely obtained. The cumulative mileage of all vehicles exceeded 50,000 kilometers. The K-Means clustering method was used for cluster analysis to obtain multi-dimensional cluster centers. The cluster centers were used to represent each relationship chain in the preset vehicle emission abnormality knowledge graph library.

[0107] First, remote monitoring data of heavy-duty vehicles with a sampling frequency of 1 Hz is obtained, including vehicle speed, acceleration, engine net output torque, engine speed, engine fuel flow, engine coolant temperature, DPF pressure difference, DPF regeneration frequency, ambient temperature, intake pressure, exhaust temperature, exhaust back pressure, air-fuel ratio, EGR valve opening, urea injection amount, particulate matter concentration, and SCR downstream NOx transmitter output value parameters.

[0108] Then, the heavy-duty vehicle remote monitoring data was used as input, and the K-Means clustering method was used to perform cluster analysis to obtain multi-dimensional cluster centers and extract the multi-dimensional features of each cluster center (such as "vehicle speed v < 25 km / h; acceleration a ≥ 0.32 m / s 2 ” → “Low-speed acceleration condition” → “NOx exceeds the standard”; “Vehicle speed v≥50 km / h; Acceleration a≤-0.68 m / s 2 ” → “High-speed deceleration condition” → “NOx exceeds the standard”).

[0109] Finally, based on the multi-dimensional characteristics of each cluster center, it is connected to each relationship chain in the preset vehicle emission anomaly knowledge graph library to support the tracing of the cause of heavy-duty vehicle emission anomalies.

[0110] 4. Establish a dynamic threshold model and LSTM model, and integrate the detection results to determine whether the emission is abnormal.

[0111] First, a preset dynamic threshold model is built to detect whether heavy-duty vehicle emissions are abnormal. Specifically, it includes:

[0112] The first step is to establish four types of heavy-duty vehicle operating conditions: urban and suburban conditions (vehicle speed less than 50 km / h, idle ratio greater than 18%), high-speed cruising conditions (vehicle speed greater than 70 km / h, accelerator pedal opening fluctuation rate less than 10%), climbing conditions (engine torque greater than 70%, slope greater than 6%), and low-temperature cold start conditions (coolant temperature less than 40°C, ambient temperature less than 10°C).

[0113] In the second step, the remote monitoring data of heavy-duty vehicles was processed using the 30-second moving window averaging method. The data was divided into urban and suburban conditions, high-speed cruising conditions, climbing conditions, and low-temperature cold start conditions according to the data characteristics. The 95th percentile of NOx emissions under each condition was calculated as the emission anomaly threshold, and a mapping relationship between the preset operating conditions and the preset emission anomaly threshold was established.

[0114] The third step is to determine the corresponding preset emission abnormality threshold based on the current operating conditions. If the emissions of the heavy-duty vehicle exceed the preset emission abnormality threshold corresponding to the current operating conditions for two consecutive hours during operation, the first detection result meets the preset vehicle emission abnormality judgment conditions, and the heavy-duty vehicle is directly judged to have abnormal emissions, and the emission abnormality tracing analysis is directly triggered.

[0115] Then, we build a preset emission prediction model (i.e., LSTM model) to detect whether heavy-duty vehicle emissions are abnormal. The steps are as follows:

[0116] The first step is to use the historical data of heavy-duty vehicles obtained remotely as training data, with the vehicle speed, acceleration, torque, speed, and fuel consumption rate in the data as input, and the NOx emission rate as output. Using the LSTM neural network, we train and build an LSTM machine learning model to predict heavy-duty vehicle emissions.

[0117] In the second step, if the emissions of a heavy-duty vehicle exceed 1.5 times the predicted value output by the preset emission prediction model for two consecutive hours during operation, the second test result meets the preset conditions for determining vehicle emission abnormalities, and the heavy-duty vehicle is directly judged to have abnormal emissions, and the emission abnormality tracing analysis is directly triggered.

[0118] Finally, a collaborative detection mechanism is established between the preset dynamic threshold model and the preset emission prediction model to accurately trigger the source analysis of emission anomalies. Specifically, it includes:

[0119] In the first step, a first detection result is obtained by detecting using a preset dynamic threshold model, and a second detection result is obtained by detecting using a preset emission prediction model.

[0120] In the second step, if the first detection result and the second detection result do not meet the preset conditions for determining vehicle emissions are abnormal, the first detection result output by the preset dynamic threshold model is combined with the second detection result output by the preset emission prediction model to obtain the final detection result.

[0121] 5. Calculate and visualize potential causes and probabilities of abnormal data, and generate maintenance recommendations with the help of AI models.

[0122] Based on the preset vehicle emission anomaly knowledge graph library and the Cosine similarity calculation strategy, the potential causes and occurrence probabilities of heavy-duty vehicle emission anomalies are calculated, realizing multi-dimensional tracing and source analysis of heavy-duty vehicle emission anomalies.

[0123] Taking a 25-ton heavy-duty vehicle of a certain model known to have abnormal emissions as an example, the present invention was used to conduct a multi-dimensional root cause tracing analysis. First, the current operating data of the remotely monitored target heavy-duty vehicle was extracted. The data sampling frequency was 1 Hz, and the data content included the vehicle VIN, acquisition time, vehicle speed, atmospheric pressure, engine net output torque, engine speed, engine fuel flow, etc. The data was then pre-processed by removing null values ​​and abnormal values ​​whose characteristic values ​​did not fit the value range.

[0124] Then, using the Cosine similarity calculation method, the data features of each relationship chain in the remote monitoring data of heavy-duty vehicles with abnormal emissions are calculated against the preset vehicle emission anomaly knowledge graph, obtaining the Cosine similarity results. The top three relationship chains in the results are selected as potential causes of abnormal heavy-duty vehicle emissions, and the occurrence probabilities of these potential causes are calculated. Here, the top three can also be understood as the three most similar relationship chains in the results.

[0125] Finally, the potential causes of abnormal emissions from heavy-duty vehicles and their probability of occurrence are visualized, and vehicle maintenance recommendations are generated synchronously in combination with large AI models such as Deepseek or ChatGPT. This method supports viewing on both PC and mobile devices, and the output visualization results are shown in Table 1.

[0126] Table 1

[0127]

[0128] In summary, the beneficial effects of the embodiments of the present invention are as follows:

[0129] (1) Establish a collaborative detection mechanism that integrates dynamic thresholds and LSTM time series prediction models. Through weight collaboration, the false alarm rate and missed detection rate of heavy-duty vehicle emission anomaly detection are significantly reduced, ensuring the accuracy and stability of anomaly identification.

[0130] (2) Relationship chain modeling in the emission anomaly knowledge graph library built based on the Neo4j graph database is combined with Cosine similarity to quickly match potential fault paths, shorten tracing time, improve accuracy, and support a dynamic update mechanism to continuously absorb maintenance cases and real-time data to enhance the practicality of the knowledge base.

[0131] (3) Remotely monitor multiple parameters of the vehicle. The system can trigger abnormal warnings, trace the source analysis, and automatically generate maintenance suggestions through the AI ​​large model in a short period of time, and push them in real time through the visual interface.

[0132] According to the vehicle emissions anomaly detection and tracing method proposed in an embodiment of the present invention, the current operating condition and current operating data are input into a preset dynamic threshold model to obtain a first detection result, and the current operating condition and current operating data are input into a preset emission prediction model to obtain a second detection result. The preset dynamic threshold model is constructed by multiple preset operating conditions and preset emission anomaly thresholds corresponding to each preset operating condition, and the preset emission prediction model is obtained by training an LSTM neural network using the vehicle's historical operating data. When the first and second detection results do not meet the preset conditions for determining vehicle emissions anomaly, the first and second detection results are fused to obtain a final detection result. When the final detection result is greater than or equal to the preset threshold, the vehicle emissions are determined to be abnormal. Based on the vehicle emissions anomaly knowledge graph, the cause and probability of the vehicle emissions anomaly are determined using a Cosine similarity calculation strategy. This solves the problems of traditional vehicle emissions anomaly detection and tracing methods, which suffer from low emission anomaly detection efficiency, high error rate, and difficulty in tracing due to data fragmentation and complex causal relationships, and achieves accurate detection of emission anomalies.

[0133] Next, the vehicle emission abnormality detection and tracing device proposed in an embodiment of the present invention will be described with reference to the accompanying drawings.

[0134] Figure 5 It is a block diagram of a vehicle emission anomaly detection and tracing device according to an embodiment of the present invention.

[0135] like Figure 5 As shown, the vehicle emission abnormality detection and tracing device 10 includes: an acquisition module 100, a detection module 200 and a tracing module 300.

[0136] Among them, the acquisition module 100 is used to obtain the current operating condition and current operating data of the vehicle; the detection module 200 is used to input the current operating condition and current operating data into a preset dynamic threshold model to obtain a first detection result, and input the current operating condition and current operating data into a preset emission prediction model to obtain a second detection result, wherein the preset dynamic threshold model is constructed by multiple preset operating conditions and a preset emission abnormality threshold corresponding to each preset operating condition, and the preset emission prediction model is obtained by training the LSTM neural network with the vehicle's historical operating data; the tracing module 300 is used to fuse the first detection result and the second detection result to obtain the final detection result when the first detection result and the second detection result do not meet the preset conditions for determining vehicle emission abnormality, and when the final detection result is greater than or equal to the preset threshold, determine that the vehicle emission is abnormal, and based on the preset vehicle emission abnormality knowledge graph library, use the Cosine similarity calculation strategy to obtain the cause and probability of the vehicle emission abnormality.

[0137] Optionally, in some embodiments, the first detection result includes the actual emissions of the vehicle. When judging whether the first detection result meets the preset conditions for determining vehicle emissions abnormality, the detection module 200 is further used to: judge whether the actual emissions of the vehicle are greater than a preset emissions abnormality threshold, and whether the first duration of the actual emissions of the vehicle being greater than the preset emissions abnormality threshold is greater than a first preset duration, wherein the preset emissions abnormality threshold is obtained by the current operating conditions; if the actual emissions of the vehicle are greater than the preset emissions abnormality threshold and the first duration is greater than the first preset duration, then it is determined that the first detection result meets the preset conditions for determining vehicle emissions abnormality, and the vehicle emissions are determined to be abnormal, and based on the preset vehicle emissions abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality; otherwise, it is determined that the first detection result does not meet the preset conditions for determining vehicle emissions abnormality.

[0138] Optionally, in some embodiments, before determining whether the first detection result is greater than a preset emission abnormality threshold, the detection module 200 is further used to: obtain historical operating data of the vehicle; construct multiple preset operating conditions based on data features in the historical operating data; establish a mapping relationship between the historical operating data of each preset operating condition and the emission abnormality threshold, so as to obtain the preset emission abnormality threshold according to the current operating condition based on the mapping relationship.

[0139] Optionally, in some embodiments, the second detection result includes the actual emissions of the vehicle. When judging whether the second detection result meets the preset conditions for determining vehicle emissions abnormality, the detection module 200 is further used to: judge whether the ratio of the vehicle's actual emissions to the emission prediction value is greater than the preset ratio, and whether the second duration of the vehicle's actual emissions being greater than the emission prediction value is greater than the second preset duration; if the ratio of the vehicle's actual emissions to the emission prediction value is greater than the preset ratio, and the second duration is greater than the second preset duration, then it is determined that the second detection result meets the preset conditions for determining vehicle emissions abnormality, and the vehicle emissions are determined to be abnormal, and based on the preset vehicle emissions abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality; otherwise, it is determined that the second detection result does not meet the preset conditions for determining vehicle emissions abnormality.

[0140] Optionally, in some embodiments, the traceability module 300 is further used to: obtain historical operating data of the vehicle; based on the historical operating data, use the Cosine similarity calculation strategy to calculate the data features of each relationship chain in the preset vehicle emission anomaly knowledge graph library respectively to obtain multiple Cosine similarity calculation results; based on a preset sorting strategy, sort the multiple Cosine similarity calculation results, and use the Cosine similarity calculation results that meet the preset ranking conditions in the sorting results as the cause and probability of the vehicle emission anomaly.

[0141] Optionally, in some embodiments, before obtaining the cause and probability of vehicle emission anomaly based on a preset vehicle emission anomaly knowledge graph library using a Cosine similarity calculation strategy, the tracing module 300 is further used to: obtain vehicle emission anomaly cause data; form a vehicle emission anomaly cause data relationship chain based on the vehicle emission anomaly cause data; use a preset knowledge graph theory to convert the vehicle emission anomaly cause data relationship chain into entities and relationships in the knowledge graph; use an entity alignment strategy to fuse relatively conflicting entities to obtain a preset vehicle emission anomaly knowledge graph library.

[0142] Optionally, in some embodiments, the traceability module 300 is further configured to: use a preset fusion formula to fuse the first detection result and the second detection result to obtain a final detection result, wherein the preset fusion formula is:

[0143] ;

[0144] in, For the final test results, E car is the actual emission of the vehicle, E Th is the preset emission abnormality threshold, E Pv is the emission prediction value, a is the weight of the first detection result, b is the weight of the second detection result.

[0145] Optionally, in some embodiments, after determining that vehicle emissions are abnormal and obtaining the cause and probability of the vehicle emissions abnormality based on a preset vehicle emissions abnormality knowledge graph library using a Cosine similarity calculation strategy, the traceability module 300 is also used to: send vehicle emissions abnormality warning information to a preset terminal; and generate vehicle maintenance recommendations using an AI model based on the vehicle emissions abnormality warning information, the cause and probability of the vehicle emissions abnormality.

[0146] It should be noted that the above explanations of the embodiment of the vehicle emission abnormality detection and tracing method are also applicable to the vehicle emission abnormality detection and tracing device of this embodiment, and will not be repeated here.

[0147] According to the vehicle emission anomaly detection and tracing device proposed in an embodiment of the present invention, the current operating condition and the current operating data are input into a preset dynamic threshold model to obtain a first detection result, and the current operating condition and the current operating data are input into a preset emission prediction model to obtain a second detection result, wherein the preset dynamic threshold model is constructed by multiple preset operating conditions and a preset emission abnormality threshold corresponding to each preset operating condition, and the preset emission prediction model is obtained by training an LSTM neural network with historical operating data of the vehicle; when the first detection result and the second detection result do not meet the preset conditions for determining vehicle emission abnormality, the first detection result and the second detection result are fused to obtain a final detection result, and when the final detection result is greater than or equal to the preset threshold, the vehicle emission abnormality is determined, and based on the vehicle emission abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emission abnormality. This solves the problems of traditional vehicle emission anomaly detection and tracing methods, such as low efficiency, high misjudgment rate and difficult tracing due to data fragmentation and complex causal relationships. It achieves accurate detection of emission anomalies and completes closed-loop management from anomaly triggering to root cause tracing, achieving the goals of reducing vehicle emission pollution, improving the level of intelligent operation and maintenance, and ensuring environmental compliance.

[0148] Figure 6 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device may include:

[0149] A memory 601 , a processor 602 , and a computer program stored in the memory 601 and executable on the processor 602 .

[0150] When the processor 602 executes the program, the vehicle emission anomaly detection and tracing method provided in the above embodiment is implemented.

[0151] Furthermore, the electronic device further includes:

[0152] The communication interface 603 is used for communication between the memory 601 and the processor 602 .

[0153] The memory 601 is used to store computer programs that can be run on the processor 602 .

[0154] The memory 601 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0155] If the memory 601, processor 602, and communication interface 603 are implemented independently, the communication interface 603, memory 601, and processor 602 can be interconnected via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 6 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0156] Optionally, in a specific implementation, if the memory 601, the processor 602 and the communication interface 603 are integrated on a chip, the memory 601, the processor 602 and the communication interface 603 can communicate with each other through an internal interface.

[0157] The processor 602 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention.

[0158] An embodiment of the present invention further provides a computer program product having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned vehicle emission anomaly detection and tracing method.

[0159] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0160] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "N" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0161] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing a custom logical function or step of a process, and the scope of the preferred embodiments of the invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of the invention pertain.

[0162] The logic and / or steps represented in a flowchart or otherwise described herein, for example, can be considered a sequenced listing of executable instructions for implementing logical functions, and can be embodied in any computer program product for use with, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer program product" can be any device that can contain, store, communicate, propagate, or transmit a program for use with, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (not exhaustive) of computer program products include: an electrical connection having one or more wires (electronic device), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). Furthermore, the computer program product may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary and then storing it in a computer memory.

[0163] It should be understood that various components of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiment, the N steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one or combination of the following technologies known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0164] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer program product, which, when executed, includes one or a combination of the steps of the method embodiment.

[0165] Furthermore, the functional units in the various embodiments of the present invention may be integrated into a single processing module, each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer program product.

[0166] The computer program product mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and are not to be construed as limiting the present invention. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for detecting and tracing vehicle emission anomalies, characterized in that: The following steps are involved: Obtain the vehicle's current operating conditions and current operating data; Inputting the current operating condition and the current operating data into a preset dynamic threshold model to obtain a first detection result, and inputting the current operating condition and the current operating data into a preset emission prediction model to obtain a second detection result, wherein the preset dynamic threshold model is constructed by a plurality of preset operating conditions and a preset emission abnormality threshold corresponding to each preset operating condition, and the preset emission prediction model is obtained by training an LSTM neural network with historical operating data of the vehicle; When the first detection result and the second detection result do not meet the preset conditions for determining vehicle emissions abnormality, the first detection result and the second detection result are integrated to obtain a final detection result. When the final detection result is greater than or equal to a preset threshold, the vehicle emissions are determined to be abnormal. Based on a preset vehicle emissions abnormality knowledge graph library, a Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality. The fusing the first detection result and the second detection result to obtain a final detection result includes: The first detection result and the second detection result are fused using a preset fusion formula to obtain the final detection result, wherein the preset fusion formula is: ； in, For the final test results, E car is the actual emission of the vehicle, E Th is the preset emission abnormality threshold, E Pv is the emission prediction value, a is the weight of the first detection result, b is the weight of the second detection result.

2. The vehicle emission anomaly detection and tracing method according to claim 1, characterized in that: The first detection result includes the actual emission amount of the vehicle. When determining whether the first detection result meets the preset vehicle emission abnormality determination condition, the following steps are included: determining whether the actual emissions of the vehicle are greater than a preset emissions abnormality threshold, and whether a first duration during which the actual emissions of the vehicle are greater than the preset emissions abnormality threshold is greater than a first preset duration, wherein the preset emissions abnormality threshold is obtained from the current operating condition; If the actual emissions of the vehicle are greater than the preset emissions abnormality threshold and the first duration is greater than the first preset duration, it is determined that the first detection result meets the preset vehicle emissions abnormality determination condition, and the vehicle emissions are determined to be abnormal. Based on the preset vehicle emissions abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality. Otherwise, it is determined that the first detection result does not meet the preset vehicle emissions abnormality determination condition.

3. The vehicle emission anomaly detection and tracing method according to claim 2, characterized in that: Before determining whether the first detection result is greater than the preset emission abnormality threshold, the method further includes: Obtaining historical operating data of the vehicle; constructing the plurality of preset operating conditions according to data features in the historical operating data; A mapping relationship between the historical operating data of each preset operating condition and the emission abnormality threshold is established, so as to obtain the preset emission abnormality threshold according to the current operating condition based on the mapping relationship.

4. The vehicle emission anomaly detection and tracing method according to claim 1, characterized in that: The second detection result includes the actual emission amount of the vehicle. When determining whether the second detection result meets the preset vehicle emission abnormality determination condition, the following is also included: Determining whether a ratio of the actual emissions of the vehicle to the predicted emissions is greater than a preset ratio, and whether a second duration during which the actual emissions of the vehicle are greater than the predicted emissions is greater than a second preset duration; If the ratio of the actual emissions of the vehicle to the predicted emissions is greater than the preset ratio, and the second duration is greater than the second preset duration, it is determined that the second detection result meets the preset conditions for determining vehicle emissions abnormality, and the vehicle emissions are determined to be abnormal. Based on the preset vehicle emissions abnormality knowledge graph library, the Cosine similarity calculation strategy is used to obtain the cause and probability of the vehicle emissions abnormality. Otherwise, it is determined that the second detection result does not meet the preset conditions for determining vehicle emissions abnormality.

5. The vehicle emission anomaly detection and tracing method according to claim 1, characterized in that: The method uses a Cosine similarity calculation strategy based on a preset vehicle emission anomaly knowledge graph library to obtain the cause and probability of vehicle emission anomaly, including: Obtaining historical operating data of the vehicle; Based on the historical operating data, the Cosine similarity calculation strategy is used to calculate the data features of each relationship chain in the preset vehicle emission anomaly knowledge graph library to obtain multiple Cosine similarity calculation results; Based on a preset sorting strategy, the multiple Cosine similarity calculation results are sorted, and the Cosine similarity calculation results that meet the preset ranking conditions in the sorting results are used as the cause and probability of the vehicle's emission abnormality.

6. The vehicle emission anomaly detection and tracing method according to claim 5, characterized in that: Before obtaining the cause and probability of vehicle emission anomaly based on the preset vehicle emission anomaly knowledge graph and using the Cosine similarity calculation strategy, the following steps are included: Obtain data on causes of abnormal vehicle emissions; forming a vehicle emission abnormality cause data relationship chain based on the vehicle emission abnormality cause data; Using the preset knowledge graph theory, the data relationship chain of the vehicle emission abnormality cause is converted into entities and relationships in the knowledge graph; The entity alignment strategy is used to fuse the relatively conflicting entities to obtain the preset vehicle emission anomaly knowledge graph library.

7. The vehicle emission anomaly detection and tracing method according to claim 1, characterized in that: After determining that the vehicle emission is abnormal, and obtaining the cause and probability of the vehicle emission abnormality by using the Cosine similarity calculation strategy based on the preset vehicle emission abnormality knowledge graph library, the following steps are included: Send vehicle emission abnormality warning information to the preset terminal; Based on the vehicle emission abnormality warning information, the cause and probability of the vehicle emission abnormality, an AI model is used to generate vehicle maintenance suggestions.

8. A vehicle emission abnormality detection and tracing device, characterized in that: include: An acquisition module is used to obtain the current operating conditions and current operating data of the vehicle; a detection module, configured to input the current operating condition and the current operating data into a preset dynamic threshold model to obtain a first detection result, and input the current operating condition and the current operating data into a preset emission prediction model to obtain a second detection result, wherein the preset dynamic threshold model is constructed by a plurality of preset operating conditions and a preset emission abnormality threshold corresponding to each preset operating condition, and the preset emission prediction model is obtained by training an LSTM neural network with historical operating data of the vehicle; a tracing module, configured to, when the first and second test results do not satisfy a preset condition for determining vehicle emissions abnormality, fuse the first and second test results to obtain a final test result, determine that the vehicle emissions are abnormal when the final test result is greater than or equal to a preset threshold, and, based on a preset vehicle emissions abnormality knowledge graph and using a Cosine similarity calculation strategy, determine the cause and probability of the vehicle emissions abnormality; The traceability module is further configured to: use a preset fusion formula to fuse the first detection result and the second detection result to obtain the final detection result, wherein the preset fusion formula is: ； in, For the final test results, E car is the actual emission of the vehicle, E Th is the preset emission abnormality threshold, E Pv is the emission prediction value, a is the weight of the first detection result, b is the weight of the second detection result.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the vehicle emission anomaly detection and tracing method as described in any one of claims 1 to 7.

Images