A method and system for identifying multi-level coupling failure of SCR based on remote OBD

Abstract

The application provides an SCR multi-level coupling failure recognition method and system based on remote OBD, and belongs to the technical field of automobile aftertreatment system fault diagnosis. The method comprises the following steps: collecting vehicle remote OBD mandatory data through a data acquisition and working condition screening module, performing effective working condition screening, and eliminating invalid data; based on the effective working condition data, calculating a carrier failure index CFI and performing carrier deterioration trend prediction, calculating a liquid level failure index LFI, and calculating an OBD failure index OFI; inputting the CFI, LFI and OFI into a multi-factor fusion decision module for fusion decision, outputting a comprehensive failure index SFI and determining a failure level and a root cause; the modules are communicatively connected to form a closed-loop feedback mechanism, and the output results of any module are fed back to other related modules to adjust calculation parameters, fault recognition and failure chain tracing. The scheme improves the accuracy, real-time performance, forward-looking performance and robustness of SCR failure recognition.

Description

Technical Field

[0001] This application relates to the field of fault diagnosis technology for automotive aftertreatment systems, and in particular to a method and system for identifying multi-level coupled failures of SCR based on remote OBD. Background Technology

[0002] With the full implementation of the China VI emission standards, the SCR (Selective Catalytic Reduction) system, as a core after-treatment device for reducing NOx emissions in heavy-duty vehicles, directly affects whether vehicle emissions meet the standards. Currently, SCR system failure detection mainly relies on fault code feedback from the on-board OBD (On-Board Diagnostics) system and offline testing, but this approach has the following technical limitations, failing to meet the needs of remote online monitoring and precise maintenance:

[0003] (1) Existing failure identification methods rely heavily on detailed parameters such as SCR inlet and outlet temperatures, urea injection volume, and urea tank temperature. However, these parameters are not essential data that must be transmitted by remote OBD for China VI vehicles, and most remote monitoring platforms cannot obtain them, resulting in poor universality of the identification methods and difficulty in implementation. (2) It can only identify hard faults of a single component of the SCR system (such as sensor short circuit, urea depletion), but cannot identify hidden faults such as sensor soft failures (accuracy drift, intermittent abnormalities) and OBD diagnostic system failures (missed reports, monitoring not ready, diagnostic logic stuck). (3) Lack of ability to identify multi-component coupled failures, unable to distinguish the root causes of different failure types such as "carrier degradation", "urea failure" and "OBD misjudgment", resulting in low operation and maintenance efficiency; (4) Existing algorithms mostly use static threshold judgment, without combining data time sequence characteristics, which are easily affected by vehicle operating condition fluctuations, resulting in low recognition accuracy and high false alarm rate.

[0004] Therefore, there is an urgent need for an online identification model algorithm that can accurately identify multiple types of failures in the SCR system, locate the root cause of failure, and be directly deployed, using only the data that must be transmitted remotely by the OBD system of China VI vehicles, to fill the gap in existing technology. Summary of the Invention

[0005] In view of this, the embodiments of this application provide a method and system for identifying multi-level coupled failures of SCR based on remote OBD, which solves the technical problems of existing technologies that rely on difficult-to-obtain detailed parameters, cannot identify latent failures and coupled failures, and have low identification accuracy. It realizes remote online accurate diagnosis and root cause location of SCR system failures without the need to install additional hardware, reduces operation and maintenance costs, and supports remote monitoring of heavy-duty vehicle emissions.

[0006] In a first aspect, embodiments of this application provide a method for identifying multi-level coupling failures in SCRs based on remote OBD, including: Step 1: Collect the mandatory data to be transmitted remotely by the vehicle via the data acquisition and operating condition screening module. Perform effective operating condition screening on the collected mandatory data and remove invalid data. The mandatory data includes exhaust temperature, upstream NOx concentration, downstream NOx concentration, urea level, engine speed, engine load, vehicle speed, SCR related fault codes, OBD monitoring ready status, fault confirmation counter, mileage and cumulative running time. Step 2: Based on the effective operating data obtained after screening, calculate the carrier failure index (CFI) through the NOx conversion characteristic analysis module, and predict the carrier degradation trend. Step 3: Calculate the liquid level failure index (LFI) using the urea liquid level anomaly analysis module; Step 4: Calculate the OBD Failure Index (OFI) using the OBD diagnostic system's failure analysis module; Step 5: Input CFI, LFI and OFI into the multi-factor fusion decision module for fusion decision, output the comprehensive failure index SFI and determine the failure level and root cause; Step 6: Connect the modules to form a closed-loop feedback mechanism. The output of any module is fed back to other relevant modules to adjust calculation parameters, identify faults, and trace the failure chain.

[0007] According to a specific implementation of an embodiment of this application, the method further includes: The real-time effective operating condition data is matched with typical operating conditions in the preset typical effective operating condition library based on the K-means clustering method. If the match is successful, the calculation parameters corresponding to the matched typical working conditions will be used to perform calculations for the NOx conversion characteristic analysis module, the urea level anomaly analysis module, and / or the OBD diagnostic system failure analysis module. If a match fails, the real-time valid operating condition data will be smoothed using an exponential moving average method.

[0008] According to a specific implementation of an embodiment of this application, the step of calculating the carrier failure index (CFI) through the NOx conversion characteristic analysis module includes: Based on the upstream and downstream NOx concentrations, the basic NOx conversion efficiency is calculated using the following expression: η NOx = 1 - (C out / C in ), Where, η NOx For the basic NOx conversion efficiency, C in C represents the average NOx concentration upstream. out This represents the average NOx concentration downstream. Based on the standard deviation of downstream NOx concentration and the mean of upstream NOx concentration, the NOx time-series fluctuation index is calculated, expressed as: NOx var = std(C out ) / mean(C in ), Among them, NOx var For NOx time series volatility index, std(C) out ) represents the standard deviation of downstream NOx concentration, mean(C in () represents the average upstream NOx concentration; The NOx conversion consistency index is calculated based on the standard deviation of NOx conversion efficiency, expressed as: NOx sta = 1 - std(η NOx,window) , Among them, NOx sta std(η) is the NOx conversion consistency index. NOx,window () represents the standard deviation of NOx conversion efficiency within multiple consecutive sliding time windows; The carrier failure index (CFI) is obtained by weighted summing of the NOx basic conversion efficiency, NOx time-series fluctuation index, and NOx conversion consistency index. A higher CFI value indicates more severe SCR carrier degradation. The expression for the carrier failure index CFI is: CFI = C1×(1-η NOx ) + C2×NOx var + C3×(1-NOx sta ), C1 + C2 + C3 = 1, Among them, C1, C2 and C3 are the first coefficient, the second coefficient and the third coefficient of the carrier failure index, respectively; The calculation of the liquid level failure index (LFI) through the urea liquid level anomaly analysis module includes: Calculate the liquid level stagnation characteristic L stuck The expression is: L stuck = |ΔL| / ΔMileage, Where ΔL is the change in urea level after traveling a certain distance, and ΔMileage is the corresponding travel distance. stuck When the value is approximately 0, it is determined that the liquid level sensor is stuck or the urea injection is blocked. Calculate the unreasonable liquid level jump characteristics L jump When the urea level changes by more than 5% in a single instance while not being filled, L jump =1, otherwise L jump =0; Calculate the low liquid level fault code characteristic L low_no_dtc When the urea level is <10% and the vehicle has been driven for >50km, but the OBD does not report any urea level-related fault codes, L low_no_dtc = 1, otherwise L low_no_dtc =0; Based on the liquid level stagnation feature L stuck Unreasonable liquid level jump characteristics L jump And the low liquid level without fault code feature L low_no_dtc The Liquid Level Failure Index (LFI) is obtained. The higher the LFI value, the higher the probability of urea system failure or cheating. The expression for the Liquid Level Failure Index (LFI) is: LFI = L1 × L stuck + L2×L jump + L3×L low_no_dtc , L1 + L2 + L3 = 1, Among them, L1, L2 and L3 are the first coefficient, the second coefficient and the third coefficient of the liquid level failure index, respectively. The calculation of the OBD Failure Index (OFI) through the failure analysis module of the OBD diagnostic system includes: When performing NOx abnormality fault code judgment, if the basic NOx conversion efficiency is consistently <50% but the OBD does not report SCR-related fault codes, I(no_DTC) = 1; otherwise, I(no_DTC) = 0. Here, I(no_DTC) is the NOx abnormality fault code characteristic. Execute monitoring readiness anomaly judgment when exhaust temperature > 240℃, vehicle continuous operating time > 30min, and OBD monitoring readiness status SCR. RDY When I(SCR) remains at 0, RDY =0) = 1, otherwise I(SCR) RDY =0)=0, where I(SCR RDY =0) indicates an abnormal monitoring readiness characteristic; When the fault counter is stuck, if there are obvious SCR failure characteristics but the OBD fault confirmation counter value remains unchanged, I(fault_cnt) is used for fault counter judgment. stuck ) = 1, otherwise I(fault_cnt stuck =0, where I(fault_cnt) stuck This indicates a fault counter is stuck. Based on the NOx anomaly without fault codes, the monitoring readiness anomaly, and the fault counter jamming characteristic, the OBD Failure Index (OFI) is calculated. When OFI ≥ 2, it is determined that the OBD diagnostic system itself has failed. The expression for the OBD Failure Index (OFI) is: OFI = I(no_DTC) + I(SCR RDY =0) + I(fault_cnt stuck ).

[0009] According to a specific implementation of an embodiment of this application, the step of predicting the carrier degradation trend includes: Construct and train an LSTM time series prediction model; The carrier failure index (CFI) is obtained over a historical period to form CFI time-series data. The CFI time series data is input into the trained LSTM time series prediction model. Through the forget gate, input gate and output gate of the hidden layer, the long-term trend of CFI changes is captured and the predicted CFI values ​​for multiple key time nodes in the future are output. Based on the NOx time series volatility index, a correction factor is constructed, and the CFI forecast value is corrected based on the correction factor to obtain the final corrected CFI forecast value. Different warning thresholds are preset. If the corrected CFI prediction value exceeds the warning threshold of the corresponding level, the warning operation corresponding to that level will be triggered, and the warning time node and the predicted degradation rate will be marked.

[0010] According to a specific implementation of an embodiment of this application, the output of any module is fed back to other related modules to adjust calculation parameters, identify faults, and trace failure chains. This includes dynamic correction of coefficients and bidirectional verification of the NOx conversion characteristic analysis module based on the output of the urea level anomaly analysis module and the OBD diagnostic system failure analysis module. The dynamic correction of the coefficients and the two-way verification include: Based on the liquid level failure index (LFI), the urea liquid level anomaly analysis module outputs a urea system reliability score, where the urea system reliability score = 1 - LFI. If CFI ≥ 0.3 and LFI > 0.5, then CFI is discounted and corrected. The correction formula is: CFI correction = CFI × 0.5 × urea system reliability score, in order to avoid misjudging abnormal urea supply as carrier deterioration. If CFI < 0.3, but LFI > 0.8, then increase the calculation weight of the first coefficient of the carrier failure index to investigate hidden carrier degradation. Simultaneously, the OBD monitoring readiness status output by the OBD diagnostic system failure analysis module is invoked. If the current operating conditions meet the effective working conditions of the SCR, but the SCR... RDY If the value is 0, the carrier degradation determination is paused, and the OBD diagnostic system failure analysis module is triggered to further check whether the OBD diagnostic logic has failed.

[0011] According to a specific implementation of an embodiment of this application, the output of any module is fed back to other related modules to adjust calculation parameters, identify faults, and trace failure chains. This includes identifying hidden faults within the urea injection system and OBD concealment in the urea level anomaly analysis module based on the output of the NOx conversion characteristic analysis module and the OBD diagnostic system failure analysis module. The identification of hidden faults and OBD concealment within the urea injection system includes: LFI and NOx var Perform a linkage analysis; if LFI < 0.3, but NOx var If the value is greater than 0.2, it is determined that there is a hidden fault inside the urea injection system. The calculation weight of the first coefficient of the liquid level failure index will be automatically increased to further investigate hidden faults. The LFI and SCR-related fault codes are checked for consistency. If LFI > 0.8, but the OBD diagnostic system failure analysis module has no corresponding SCR-related fault code and OFI < 2, it is marked as OBD concealment or masking, triggering the OBD diagnostic system failure analysis module to re-check the OBD diagnostic logic.

[0012] According to a specific implementation of an embodiment of this application, the output of any module is fed back to other related modules to adjust calculation parameters, identify faults, and trace the source of failure chains, including: If exhaust temperature T is detected exh If the data fluctuation is greater than 50℃ / min and lasts for ≥10s, a data anomaly signal is sent to the NOx conversion characteristic analysis module, the carrier degradation determination is paused, and the process continues until the exhaust temperature T... exh Recalculate CFI after the data stabilizes; Record the time sequence of anomalies in CFI, LFI, and OFI, and identify the propagation path of the failure chain through time-series correlation analysis based on the failure chain determination rules; Based on the current vehicle's real-time output of SCR-related fault code reporting frequency, fault confirmation counter changes, and OBD monitoring readiness status, the OBD diagnostic logic of the OBD diagnostic system failure analysis module is self-calibrated. The self-calibrated parameters are then synchronously fed back to the NOx conversion characteristic analysis module and the urea level anomaly analysis module, and the coefficients and LFI judgment thresholds in the CFI expression are adjusted.

[0013] According to a specific implementation of an embodiment of this application, the method further includes: Construct and train an isolated forest anomaly detection model; Extract the time-series data of urea liquid level and the corresponding driving mileage data within the first preset historical time period to construct the input dataset. Each sample includes the urea liquid level value, the urea liquid level change ΔL, the corresponding driving mileage ΔMileage, and the timestamp. The input dataset is fed into the trained Isolation Forest anomaly detection model. By randomly partitioning the feature space, anomaly scores are calculated for each sample, and anomalous samples are selected. Perform feature analysis on abnormal samples to determine the anomaly type; By combining the NOx conversion efficiency output by the NOx conversion feature analysis module, we can verify whether the anomaly type determination result is a misjudgment. If it is a misjudgment, we can adjust the threshold of the anomaly score of the isolated forest anomaly detection model. The frequency of abnormal liquid levels within a second preset time period is counted. The duration of the second preset time period is longer than that of the first preset time period. A health score of the urea system is constructed based on the frequency of abnormal liquid levels. If the health score is lower than the threshold, a soft failure warning is triggered.

[0014] According to a specific implementation of an embodiment of this application, the step of inputting CFI, LFI, and OFI into a multi-factor fusion decision module for fusion decision-making, outputting a comprehensive failure index SFI, and determining the failure level and root cause includes: The comprehensive failure index (SFI) is constructed as follows: SFI = α×CFI +β×LFI +γ×OFI, Where α, β and γ are the dynamic adaptive weights of CFI, LFI and OFI, respectively; If the judgment results of the NOx conversion feature analysis module and the urea liquid level anomaly analysis module are consistent, the values ​​of α and β will be increased; if the judgment results conflict, the corresponding weights will be decreased. When SFI < 0.3, the SCR system is normal; when 0.3 ≤ SFI < 0.5, the SCR system is under warning; when 0.5 ≤ SFI < 0.8, the SCR system is slightly faulty; when SFI ≥ 0.8, the SCR system is severely faulty. The root cause of failure is determined based on the proportions of CFI, LFI, and OFI. If CFI has the highest proportion, the root cause is SCR carrier degradation or failure. If LFI has the highest proportion, the root cause is urea system failure, liquid level sensor soft failure, or tampering. If OFI has the highest proportion, the root cause is OBD diagnostic system failure. If the difference between the proportions of CFI, LFI, and OFI is ≤10%, it is determined to be a multi-component coupled failure. The temporal changes of CFI, LFI, and OFI are synchronously correlated to identify the failure chain propagation path and core triggering components, and to locate the source of failure.

[0015] Secondly, embodiments of this application also provide a multi-level coupling failure identification system for SCR based on remote OBD, used to execute the multi-level coupling failure identification method for SCR based on remote OBD as described in any embodiment of the first aspect, the system comprising: The data acquisition and operating condition filtering module is used to collect the mandatory data that must be transmitted by the vehicle remote OBD, and to filter the collected mandatory data for effective operating conditions, and to remove invalid data. The mandatory data includes exhaust temperature, upstream NOx concentration, downstream NOx concentration, urea level, engine speed, engine load, vehicle speed, SCR related fault codes, OBD monitoring ready status, fault confirmation counter, mileage and cumulative running time. The NOx conversion characteristic analysis module is used to calculate the carrier failure index (CFI) and predict the carrier degradation trend based on the effective operating data obtained after screening. The urea level anomaly analysis module is used to calculate the level failure index (LFI). The OBD diagnostic system failure analysis module is used to calculate the OBD failure index OFI. The multi-factor fusion decision module is used to make fusion decisions based on CFI, LFI and OFI, output the comprehensive failure index SFI and determine the failure level and root cause. The collaborative analysis module is used for communication and connection between modules to form a closed-loop feedback mechanism. The output results of any module are fed back to other relevant modules to adjust calculation parameters, identify faults, and trace the source of failure chains.

[0016] Beneficial effects: The SCR multi-level coupling failure identification method and system based on remote OBD in this application embodiment has the following beneficial effects: 1. Technical Aspect: Breaking through the dependence of existing SCR failure identification on detailed parameters, filling the technical gaps in the identification of OBD self-failure, sensor soft failure, and multi-component coupling failure; adding new algorithmic innovations such as dynamic weight, adaptive time window, soft failure early warning, module joint verification, and Bayesian-fuzzy fusion decision-making, solving the problems of poor working condition adaptability, easy misjudgment, inability to provide early warning, and independent and unrelated modules in existing algorithms, further improving the accuracy, real-time performance, foresight, and robustness of SCR failure identification; 2. Application level: No additional hardware is required. It can be directly deployed on the existing remote OBD monitoring platform, adapting to all China VI heavy-duty vehicles and multiple driving scenarios. It enables remote online diagnosis of SCR system failure, early warning, failure chain tracing and operation and maintenance suggestions, and builds a closed-loop management system of "identification-early warning-operation and maintenance". 3. Practical Value: Provides the ecological and environmental departments with reliable remote monitoring technology for heavy-duty vehicle emissions, and provides vehicle manufacturers and logistics fleets with accurate SCR system operation and maintenance early warning, reducing operation and maintenance costs, reducing the risk of exceeding emission standards, and helping to implement the National VI emission standards. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart of a multi-level coupling failure identification method for SCR based on remote OBD according to an embodiment of the present invention; Figure 2 This is another flowchart of a multi-level coupling failure identification method for SCR based on remote OBD according to an embodiment of the present invention. Detailed Implementation

[0019] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0020] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this application, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0022] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. The illustrations only show the components related to this application and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0023] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0024] Firstly, this application provides a multi-level coupled failure identification method for SCR based on remote OBD, applicable to remote online monitoring, failure diagnosis, and root cause localization of SCR systems in China VI heavy-duty vehicles. It can be directly deployed on a vehicle remote monitoring platform without requiring additional hardware. This method utilizes only the mandatory data transmitted via remote OBD for China VI vehicles to construct a four-level identification model: "operating condition screening - multi-module collaborative analysis - fusion decision-making - closed-loop feedback." Through five deeply coupled collaborative modules, it achieves accurate identification, root cause localization, early warning, and failure chain tracing of multiple types of failures in the SCR system. The core optimization direction is to strengthen the rigor and correlation of the calculation process of each module, highlight algorithmic innovation and technological breakthroughs, compensate for existing technical shortcomings, and focus on strengthening the joint analysis logic, calculation details, and implementation of innovative points between modules, ensuring that each calculation step is reproducible and each innovation is supported. See the appendix below. Figure 1 and Figure 2 Provide a detailed description.

[0025] In a first aspect, embodiments of this application provide a method for identifying multi-level coupling failures of SCRs based on remote OBD, referring to... Figure 1 include: The data acquisition and operating condition filtering module collects essential data for remote OBD transmission of the vehicle. This essential data is then filtered for valid operating conditions, eliminating invalid data. The essential data includes exhaust temperature (T). exh (before DOC), upstream NOx concentration (C) in ), downstream NOx concentration (C out ), urea level (%), engine speed (n), engine load (Load), vehicle speed (V), SCR-related fault codes (DTC), OBD monitoring readiness status (SCR) RDY ), fault confirmation counter, mileage, and cumulative running time; Based on the effective operating data obtained after screening, the carrier failure index (CFI) is calculated through the NOx conversion characteristic analysis module, and the carrier degradation trend is predicted. The liquid level failure index (LFI) is calculated using the urea liquid level anomaly analysis module. The OBD Failure Index (OFI) is calculated using the failure analysis module of the OBD diagnostic system. The CFI, LFI and OFI are input into the multi-factor fusion decision module for fusion decision, and the comprehensive failure index SFI is output to determine the failure level and root cause. The modules communicate with each other to form a closed-loop feedback mechanism, and the output of any module is fed back to other relevant modules to adjust calculation parameters, identify faults, and trace the source of failure chains.

[0026] In this embodiment, each module synchronizes data in real time through the vehicle's remote OBD transmission channel, ensuring that the calculation parameters can be dynamically adjusted according to the vehicle's operating status. This avoids the impact of fixed thresholds on the accuracy of failure identification under different operating scenarios. At the same time, the closed-loop feedback mechanism can verify the accuracy of each failure index calculation after tracing the root cause of the coupled failure, and continuously optimize the judgment accuracy of the fusion decision model.

[0027] The following is a detailed description of the methods implemented by the data acquisition and operating condition screening module. The effective operating condition screening specifically includes the following: Expiry data (cold start, idling, drastic fluctuations in operating conditions, data loss / mistransmission) was removed. The filtering criterion was: exhaust temperature T. exh The system operates within the following temperature ranges: >220℃ (effective operating temperature range of the SCR system), engine speed and load fluctuations both <12% for a duration ≥60s (stable operating condition), and vehicle speed >10km / h (non-idling). A lightweight anti-interference processing method (moving average filtering + DBSCAN outlier clustering and removal) is introduced to filter out instantaneous spikes and packet loss interference during OBD data transmission, ensuring the validity of subsequent analysis data. Preferably, a new operating condition labeling function is added, labeling the filtered effective operating conditions as high load (Load > 70%), medium load (30% ≤ Load ≤ 70%), and low load (Load < 30%), and simultaneously pushing this to all subsequent modules, providing a basis for dynamic weight adjustment and adaptive time window optimization. For example, if the duration of conditions meeting the above criteria (high load, low load, medium load) is greater than 60s, such as a continuous load greater than 70% per second for a duration of 120s, then the operating conditions within this 120s time period are marked as high load operating condition segments.

[0028] Furthermore, a working condition similarity matching algorithm is introduced, which overcomes the pain point of poor adaptability of existing working conditions. The method also includes: The system performs similarity matching between real-time valid operating condition data and typical operating conditions (covering multiple scenarios such as urban areas, highways, and mountainous areas) in a pre-set typical valid operating condition library based on the K-means clustering method, thereby solving the problem of misjudgment caused by data fluctuations under complex operating conditions. If the match is successful, the calculation parameters corresponding to the matched typical working conditions will be used to perform calculations for the NOx conversion characteristic analysis module, the urea level anomaly analysis module, and / or the OBD diagnostic system failure analysis module. If a match fails, the real-time valid operating condition data will be smoothed using an exponential moving average method.

[0029] Specifically, the similarity matching process employs normalization to eliminate dimensional interference, cosine similarity for precise matching, and EMA for smoothing and fallback, balancing accuracy and adaptability. This includes: Construction of a typical effective operating condition library: 1000 sets of effective operating condition data from different scenarios (urban, highway, mountainous) were collected, and engine load (Load) and exhaust temperature (T) were extracted. exh Four core features—engine speed (n), vehicle speed (V), and engine speed (V)—are used as input parameters for operating condition clustering. The K-means clustering algorithm (with k=3 clusters, corresponding to 3 typical scenarios) is used to obtain cluster centers for 3 typical operating conditions, constructing a typical and effective operating condition library. Example of cluster center parameters: High-speed operating condition (Load=75%, T...). exh =320℃, n=1800r / min, V=80km / h), urban driving conditions (Load=45%, T exh =260℃, n=1200r / min, V=30km / h), mountainous working conditions (Load=60%, T exh =290℃, n=1500r / min, V=50km / h); Extracting four core features of real-time operating conditions (Load, T) exh (x, n, V) undergoes min-max normalization to eliminate the influence of dimensions. The normalization formula is: x norm =(xx min ) / (x max -x min ), where x is the real-time feature value, x min x max These are the historical minimum and maximum values ​​of this feature (based on statistics from 1000 sets of typical working condition data, Load: 0-100%, T). exh :220-400℃, n: 800-2200r / min, V: 0-120km / h); Calculate the cosine similarity between the real-time normalized operating condition features and the three cluster centers in the typical effective operating condition database. The cosine similarity calculation formula is: cosθ=(A·B) / (|A|×|B|), where A is the real-time operating condition feature vector ([Load... norm ,T exhnorm , n norm V normB is the cluster center feature vector, and cosθ takes values ​​in the range [0,1]. The closer the cosθ value is to 1, the higher the similarity. Judgment Rules: The maximum cosine similarity value is used. If this value is ≥0.85, the match is considered successful, and algorithm parameters corresponding to typical working conditions (such as time window length and weight coefficients) are directly applied to reduce computational load. If the value is <0.85, the match is considered unsuccessful (belonging to an extreme working condition), and the exponential moving average (EMA) method is used for smoothing. The EMA calculation formula is: EMA t =α×x t +(1-α)×EMA t-1 Where α = 0.7 (a smoothing coefficient optimized through actual testing, balancing response speed and anti-interference capability), x t For real-time operating condition data, EMA t-1 The initial value of EMA is the average of the previous 5 sets of real-time data, which is the smoothed value of the previous time step, to ensure data stability under extreme conditions.

[0030] The role of the working condition similarity matching algorithm is to filter high-quality working condition data through similarity matching, and to combine it with EMA smoothing to process extreme working condition data, thereby improving the algorithm's adaptability to complex working conditions, reducing misjudgments caused by extreme data, and providing more accurate working condition basis for the adaptive adjustment of parameters in subsequent modules, thus solving the problem of poor data adaptability under complex and variable working conditions.

[0031] The following section provides a detailed introduction to the implementation method of the NOx conversion characteristic analysis module, referring to... Figure 2 Based on upstream and downstream NOx concentration data under effective operating conditions, this module innovatively extracts two core features: NOx temporal fluctuation and conversion consistency, and constructs a carrier failure index (CFI) to achieve accurate inversion of carrier deterioration status without relying on SCR inlet and outlet temperatures (solving the pain point of existing technologies relying on detailed parameters). At the same time, it establishes a two-way verification and parameter linkage mechanism with the urea liquid level anomaly analysis module and the OBD diagnostic system failure analysis module to avoid misjudgment by a single module.

[0032] In one embodiment, calculating the carrier failure index (CFI) via the NOx conversion characteristic analysis module includes: (1) Based on the upstream and downstream NOx concentrations, the basic NOx conversion efficiency is calculated, and the expression is: η NOx = 1 - (C out / C in ), Where, η NOx For the basic NOx conversion efficiency, C in C represents the average NOx concentration upstream. out This represents the average NOx concentration downstream. (2) Based on the standard deviation of downstream NOx concentration and the mean of upstream NOx concentration, calculate the NOx time-series fluctuation index, expressed as: NOx var = std(C out ) / mean(C in ), Among them, NOx var For NOx time series volatility index, std(C) out ) represents the standard deviation of downstream NOx concentration, mean(C in The value represents the average upstream NOx concentration. When the carrier deteriorates, the fluctuation of downstream NOx concentration will increase significantly. (3) Based on the standard deviation of NOx conversion efficiency, calculate the NOx conversion consistency index, the expression of which is: NOx sta = 1 - std(η NOx,window) , Among them, NOx sta std(η) is the NOx conversion consistency index. NOx,window The standard deviation of NOx conversion efficiency over multiple consecutive sliding time windows is given. The length of the sliding time window is adaptively adjusted according to the degree of fluctuation in operating conditions, and the adjustment logic is as follows: Operating condition fluctuation judgment: Calculate the 30s sliding standard deviation of engine load under real-time operating conditions. If it is ≤5%, it is judged as a stable operating condition, and the window length is set to 60s; if it is >5%, it is judged as a fluctuating operating condition, and the window length is shortened to 30s. Trend fitting optimization: Introducing a linear trend fitting factor k trend For η within each time window NOx The data is linearly fitted, and instantaneous outliers deviating from the fitted line by ±2σ are removed, where σ is the standard deviation, to avoid consistency calculation bias caused by transient interference; the lower the conversion consistency (NOx)... sta The smaller the value, the more severe the carrier degradation. The linear trend fitting factor can accurately capture the gradual process of carrier degradation, making up for the shortcomings of existing methods that can only identify severe degradation. The NOx conversion consistency index is based on adaptive time windows and trend fitting, breaking through the limitations of fixed time windows; (4) The NOx basic conversion efficiency, NOx time-series fluctuation index, and NOx conversion consistency index are weighted and summed to obtain the carrier failure index (CFI). The expression for the carrier failure index (CFI) is: CFI = C1×(1-η NOx ) + C2×NOx var + C3×(1-NOx sta ), C1 + C2 + C3 = 1, Among them, C1, C2 and C3 are the first coefficient, the second coefficient and the third coefficient of carrier failure index, respectively. The larger the CFI value, the more severe the degradation of the SCR carrier. At the same time, combined with the time drift rate of CFI (weekly drift amount, monthly drift amount), the carrier degradation trend can be warned 7-15 days in advance.

[0033] In practice, the values ​​of C1, C2, and C3 are set according to the actual situation, for example, set to 0.4, 0.3, and 0.3 respectively. Then, the CFI calculation formula is: CFI = 0.4 × (1 - η) NOx ) + 0.3×NOx var + 0.3×(1-NOx sta ).

[0034] Furthermore, joint module verification is performed to address the pain point of misjudgment by a single module. The output of any module is fed back to other relevant modules to adjust calculation parameters, identify faults, and trace failure chains. This includes dynamic correction of coefficients in the NOx conversion characteristic analysis module and bidirectional verification based on the output of the urea level anomaly analysis module and the OBD diagnostic system failure analysis module. The dynamic correction of the coefficients and the two-way verification include: Based on the liquid level failure index (LFI), the urea liquid level anomaly analysis module outputs a urea system reliability score. The urea system reliability score = 1 - LFI, with a range of [0,1]. Correction logic is established as follows: If CFI ≥ 0.3 (indicating carrier degradation) and LFI > 0.5 (confidence < 0.5, the urea level anomaly analysis module determines the urea system is abnormal), then CFI will be discounted and corrected. The correction formula is: CFI correction = CFI × 0.5 × urea system confidence score, in order to avoid misjudging urea supply anomalies as carrier degradation. If CFI < 0.3 (carrier is normal), but LFI > 0.8 (urea level anomaly analysis module determines the urea system is severely abnormal), then the calculation weight of the first coefficient of the carrier failure index is increased to investigate hidden carrier degradation. For example, CFI = 0.5 × (1 - ηNOx) + 0.3 × NOxvar + 0.2 × (1 - NOxsta) is recalculated to investigate hidden carrier degradation. Simultaneously, the OBD monitoring readiness status (SCR) output by the OBD diagnostic system failure analysis module is invoked. RDY If the current operating conditions meet the effective operating conditions of SCR (T) exh >220℃, continuous operation ≥60s), but SCR RDY If the value is 0, the carrier degradation determination is paused, and the OBD diagnostic system failure analysis module is triggered to further verify whether the OBD diagnostic logic has failed, ensuring that the determination result is rigorous.

[0035] Furthermore, the carrier degradation trend prediction innovatively introduces an LSTM (Long Short-Term Memory) time-series prediction model, combined with operational condition label linkage, to achieve accurate prediction of carrier degradation trends (15-30 days in advance warning), providing sufficient preparation time for operation and maintenance, and overcoming the limitation of existing methods that can only make post-event judgments. This includes: Construct and train an LSTM time series prediction model; The carrier failure index (CFI) is obtained over a historical period to form CFI time-series data. The CFI time series data is input into the trained LSTM time series prediction model. Through the forget gate, input gate and output gate of the hidden layer, the long-term trend of CFI changes is captured and the predicted CFI values ​​for multiple key time nodes in the future are output. Based on the NOx time series volatility index, a correction factor is constructed, and the CFI forecast value is corrected based on the correction factor to obtain the final corrected CFI forecast value. Different warning thresholds are preset. If the corrected CFI prediction value exceeds the warning threshold of the corresponding level, the warning operation corresponding to that level will be triggered, and the warning time node and the predicted degradation rate will be marked.

[0036] The specific implementation includes the following steps: 1) Model building process: 1. Data preprocessing: Collect CFI time series data for the past 30 days (sampling frequency 1 time / hour) and corresponding operating condition labels (high / medium / low load), remove outliers (using the 3σ criterion, i.e., remove CFI data that deviates from the mean ± 3 times the standard deviation), and normalize the CFI data (min-max normalization, range [0,1]). 2. Dataset partitioning: The dataset is divided into training set, validation set, and test set in a 7:2:1 ratio. The training set is used for model training, the validation set is used for adjusting model parameters, and the test set is used for verifying model accuracy. 3. Model structure settings: Input layer neurons = 2 (CFI time series data, working condition label encoded values, working condition labels use one-hot encoding: high load [1,0,0], medium load [0,1,0], low load [0,0,1]), hidden layer has 2 layers, each layer has 32 neurons, the activation function is ReLU, output layer neurons = 1 (CFI predicted values ​​for the next 15-30 days), optimizer is Adam, learning rate = 0.001, number of iterations = 100 rounds, batch size = 32; 2) Prediction calculation steps (working condition label linkage + dynamic correction, improving prediction accuracy): 1. Input features: Input the CFI time series data of the past 30 days (sampling frequency 1 time / hour, a total of 720 groups) and the corresponding operating condition labels (high / medium / low load, using one-hot encoding: high load [1,0,0], medium load [0,1,0], low load [0,0,1]) into the LSTM model in time series order to ensure time series correlation; 2. Model Inference: By using the forget gate, input gate, and output gate of the hidden layer, the long-term trend of CFI (the gradual characteristics of carrier degradation) is captured, and the predicted CFI values ​​for the next 15 days, 20 days, and 30 days are output. 3. Precision correction (in conjunction with NOx) var Dynamic correction to avoid prediction bias): Combined with the NOx time-series fluctuation index (NOx) output by the NOx transformation characteristic analysis module. var ), construct the correction factor k = 1 - 0.2 × NOx var (NOx) var The larger the value, the more obvious the carrier fluctuation; the smaller the correction factor, the stronger the correction. Multiply the CFI prediction value by the correction factor to obtain the final corrected CFI prediction value. 4. Warning Trigger: Preset warning thresholds (0.5 for mild failure and 0.6 for moderate failure). If the corrected CFI prediction value is ≥0.5, an early warning is triggered. At the same time, the warning time node and the predicted degradation rate are marked, and a precise prompt "It is expected to reach the XX failure level in XX days" is output.

[0037] This embodiment predicts carrier degradation trends, upgrading carrier degradation from "early warning" to "trend prediction," breaking through the limitation of existing systems that can only warn of immediate drift, and providing more preparation time for operation and maintenance.

[0038] The urea level anomaly analysis module, based on urea level time-series data, innovatively extracts three major anomaly features: stagnation, jumps, and low levels without fault codes. It constructs a Level Failure Index (LFI) to accurately identify urea system failures, soft failures of level sensors, and cheating behaviors, without relying on urea injection volume or temperature data (addressing the pain point of existing technologies that depend on detailed parameters). Simultaneously, it establishes a correlation analysis and consistency verification mechanism with the NOx conversion feature analysis module and the OBD diagnostic system failure analysis module to improve identification accuracy. See below for details. Figure 2 The method implemented by the urea level anomaly analysis module is described in detail.

[0039] In one embodiment, calculating the liquid level failure index (LFI) using the urea liquid level anomaly analysis module includes: (1) Calculate the liquid level stagnation characteristic L stuck The expression is: L stuck= |ΔL| / ΔMileage, Where ΔL is the change in urea level after traveling a certain distance, and ΔMileage is the corresponding travel distance. stuck When the value is approximately 0 (i.e., the mileage increases but the liquid level remains basically unchanged), it is determined that the liquid level sensor is stuck or the urea injection is blocked. (2) Calculate the characteristics of unreasonable liquid level jumps L jump When the urea level changes by more than 5% in a single instance while not being filled, L jump = 1, otherwise L jump =0, used to identify soft failure of liquid level sensor and failure of OBD urea monitoring; (3) Calculate the low liquid level fault code characteristic L low_no_dtc When the urea level is <10% and the vehicle has been driven for >50km, but the OBD does not report urea level-related fault codes such as P207F, L low_no_dtc = 1, otherwise L low_no_dtc =0, used to identify OBD urea monitoring failure; (4) Based on the liquid level stagnation feature L stuck Unreasonable liquid level jump characteristics L jump And the low liquid level without fault code feature L low_no_dtc The liquid level failure index (LFI) is obtained, and its expression is: LFI = L1 × L stuck + L2×L jump + L3×L low_no_dtc , L1 + L2 + L3 = 1, L1, L2, and L3 are the first, second, and third coefficients of the liquid level failure index, respectively. The larger the LFI value, the higher the probability of urea system failure or cheating. At the same time, the liquid level time-series drift rate calculation (such as monthly liquid level drift) is introduced to provide early warning of the soft failure trend of the liquid level sensor 7-15 days in advance, so as to realize early warning of soft failure.

[0040] In practice, the specific values ​​of L1, L2, and L3 can be adjusted according to the actual situation. For example, in this embodiment, they are set to 0.4, 0.3, and 0.3 respectively, i.e., LFI = 0.4 × L stuck + 0.3×L jump + 0.3×L low_no_dtc .

[0041] Furthermore, the joint analysis of the urea level anomaly analysis module with other modules is described in detail. The output results of any module are fed back to other relevant modules to adjust calculation parameters, identify faults, and trace failure chains. This includes identifying hidden faults within the urea injection system and OBD concealment within the urea level anomaly analysis module based on the output results of the NOx conversion characteristic analysis module and the OBD diagnostic system failure analysis module. The identification of hidden faults and OBD concealment within the urea injection system includes: Linking with the NOx Conversion Characteristic Analysis Module: Linking the LFI with the NOx time-series fluctuation index (NOx) output by the NOx Conversion Characteristic Analysis Module. var Perform a linkage analysis; if LFI < 0.3, but NOx var If the value is greater than 0.2, it is determined that there is a hidden fault within the urea injection system (such as internal leakage or poor atomization). In this case, the weight of the first coefficient of the liquid level failure index is automatically increased, and LFI is recalculated as 0.5 × L. stuck + 0.3×L jump + 0.2×L low_no_dtc Further investigate hidden failures; Verification with the OBD diagnostic system failure analysis module: The LFI is checked for consistency with the SCR-related fault codes output by the OBD diagnostic system failure analysis module. If LFI > 0.8 (serious failure), but the OBD diagnostic system failure analysis module has no corresponding SCR-related fault codes and OFI < 2, it is marked as OBD concealment or masking, triggering the OBD diagnostic system failure analysis module to re-check the OBD diagnostic logic, forming a cross-module constraint.

[0042] Furthermore, a urea level anomaly tracing algorithm is introduced to overcome the limitations of existing algorithms that can only identify anomalies but cannot trace their origins. This algorithm innovatively incorporates the Isolation Forest anomaly detection algorithm, combined with reverse verification of NOx conversion efficiency, to achieve accurate tracing of level anomalies (time point, anomaly type), and simultaneously outputs a health score, providing precise data for operation and maintenance. The specific calculation method and process are as follows (enhancing algorithm parameters and calculation reproducibility): Construct and train an isolated forest anomaly detection model; Extract the time-series data of urea liquid level and the corresponding driving mileage data within the first preset historical time period to construct the input dataset. Each sample includes the urea liquid level value, the urea liquid level change ΔL, the corresponding driving mileage ΔMileage, and the timestamp. The input dataset is fed into the trained Isolation Forest anomaly detection model. By randomly partitioning the feature space, anomaly scores are calculated for each sample, and anomalous samples are selected. Perform feature analysis on abnormal samples to determine the anomaly type; By combining the NOx conversion efficiency output by the NOx conversion feature analysis module, we can verify whether the anomaly type determination result is a misjudgment. If it is a misjudgment, we can adjust the threshold of the anomaly score of the isolated forest anomaly detection model. The frequency of abnormal liquid levels within a second preset time period is counted. The duration of the second preset time period is longer than that of the first preset time period. A health score of the urea system is constructed based on the frequency of abnormal liquid levels. If the health score is lower than the threshold, a soft failure warning is triggered.

[0043] In practice, the urea level anomaly tracing algorithm includes the following steps: 1) Algorithm parameter settings: In the isolated forest model, the number of decision trees = 100, the number of samples per decision tree = 256, and the anomaly score threshold = 0.7 (a score ≥ 0.7 is considered anomaly, and a score < 0.7 is considered normal). 2) Anomaly tracing calculation process (multi-feature linkage tracing + reverse verification to ensure tracing accuracy): 1. Data Input: Extract the time-series data of urea liquid level over the past 7 days (sampling frequency 1 time / 10min, a total of 1008 sets) and the corresponding driving mileage data to construct the input dataset. Each sample includes: urea liquid level value, urea liquid level change ΔL, driving mileage change ΔMileage, and timestamp to ensure comprehensive sample features. 2. Model Training and Anomaly Detection: Input the input dataset into the Isolation Forest model. The model calculates the anomaly score for each sample by randomly dividing the feature space (score range [0,1]) and filters out abnormal samples (score ≥ 0.7, a threshold optimized by actual testing, taking into account both false negatives and false positives). 3. Anomaly Type Localization (Precise Classification Based on Feature Rules): Perform feature analysis on anomaly samples and determine the anomaly type based on the following rules: a) If ΔL≈0 (≤0.1%) and ΔMileage>50km, it is judged as "urea injection shielding" (cheating behavior); b) If, under non-filling conditions, a single sudden change in ΔL > 5% and there is no change in mileage (ΔMileage = 0), it is determined to be "liquid level sensor drift" (soft failure). c) If a single mutation in ΔL is greater than 10% and the corresponding time point is when the vehicle is stationary (V=0), it is determined to be "abnormal urea filling" (illegal filling or leakage). 4. Anomaly Time Node Location: Extract the timestamp of the anomaly sample to determine the specific date and time of the anomaly (accurate to the minute) and generate an anomaly source tracing report; 5. Result Verification (combined with reverse verification of NOx conversion efficiency to eliminate false positives): Call the NOx conversion efficiency (η) output by the NOx conversion feature analysis module. NOxIf the anomaly type is "urea injection shielding", then η NOx The percentage should be less than 50%; otherwise, it is considered a misjudgment, and the abnormal scoring threshold (±0.05) is readjusted. 6. Health score: Calculate the frequency of abnormal liquid levels over the past 30 days to construct a health score for the urea system. Formula: Health score = 100 - (abnormal frequency × 5). The score range is 0-100 (0 is the worst and 100 is the best). If the health score is <80, a soft failure warning is triggered.

[0044] The role of the urea level anomaly tracing algorithm is to not only identify obvious anomalies such as level stagnation and jumps, but also to accurately locate the specific time point and type of anomaly. Combined with reverse verification of NOx conversion efficiency, it solves the pain point of existing algorithms that can only identify anomalies but cannot locate the source of the anomaly. At the same time, it can intuitively reflect the overall status of the urea system through health scores.

[0045] The OBD diagnostic system failure analysis module constructs the OBD Failure Index (OFI) based on abnormal NOx conversion characteristics, OBD monitoring readiness status, and changes in the fault confirmation counter. This is the first time that remote identification of OBD diagnostic system failures (missed reports, monitoring in readiness issues, diagnostic logic freezes) has been achieved (filling a gap in existing technology). Simultaneously, as a global verification module, it inversely constrains the judgment results of the NOx conversion characteristic analysis module and the urea level anomaly analysis module, synchronously achieving OBD diagnostic logic self-calibration and ensuring the long-term stability of the overall identification system. See below for details. Figure 2 A detailed description of the failure analysis module of the OBD diagnostic system is provided.

[0046] In one embodiment, calculating the OBD Failure Index (OFI) through the OBD diagnostic system failure analysis module includes: (1) Perform NOx abnormality without fault code judgment. When the NOx basic conversion efficiency is continuously <50%, but the OBD does not report SCR related fault codes such as P20EE and P204F, I(no_DTC) = 1, otherwise I(no_DTC) = 0. Wherein, I(no_DTC) is the NOx abnormality without fault code characteristic. (2) Execute monitoring readiness anomaly judgment when exhaust temperature > 240℃, vehicle continuous running time > 30min and OBD monitoring readiness status SCR RDY When I(SCR) remains at 0, RDY =0) = 1, otherwise I(SCR) RDY =0)=0 indicates that OBD monitoring is not enabled or is disabled, where I(SCR) RDY =0) indicates an abnormal monitoring readiness characteristic; (3) Execute fault counter stuck judgment. When there are obvious SCR failure characteristics, but the OBD fault confirmation counter value remains unchanged, I(fault_cnt) stuck ) = 1, otherwise I(fault_cnt stuck )=0 indicates that the OBD diagnostic logic is stuck, where I(fault_cnt) stuck This indicates a fault counter is stuck. (4) Based on the NOx abnormality without fault code characteristics, the monitoring readiness abnormality characteristics, and the fault counter stuck characteristics, calculate the OBD failure index OFI. When OFI ≥ 2, it is determined that the OBD diagnostic system itself has failed. The expression for the OBD failure index OFI is: OFI = I(no_DTC) + I(SCR RDY =0) + I(fault_cnt stuck ).

[0047] Furthermore, the OBD diagnostic system's failure analysis module utilizes reverse constraints and failure chain tracing, based on temporal correlation and failure chain localization, overcoming the current limitation of failing to identify coupled failures. Specifically, the output of any module is fed back to other related modules to adjust calculation parameters, identify faults, and trace failure chains, including: 1) Reverse constraint: If the exhaust temperature T is detected exh If the data fluctuation is greater than 50℃ / min and lasts for ≥10s, a data anomaly signal is sent to the NOx conversion characteristic analysis module, the carrier degradation determination is paused, and the process continues until the exhaust temperature T... exh Once the data has stabilized, the CFI is recalculated to avoid misjudgments caused by abnormal data. 2) Failure chain tracing: Record the time sequence of anomalies in CFI, LFI and OFI, and identify the failure chain propagation path through time sequence correlation analysis based on failure chain determination rules; 3) OBD Diagnostic Logic Self-calibration Algorithm: Based on the current vehicle's real-time output of SCR-related fault code reporting frequency, fault confirmation counter changes, and OBD monitoring readiness status, the OBD diagnostic logic self-calibrates the OBD diagnostic system failure analysis module. The self-calibrated parameters are then synchronously fed back to the NOx conversion feature analysis module and the urea level anomaly analysis module to adjust the coefficients in the CFI expression and the LFI judgment threshold.

[0048] In specific implementation, the failure chain determination rules include: 1. OFI increases first (OBD fails first) → CFI / LFI increases later (carrier / urea fails later), which is judged as "OBD diagnosis failure → systemic missed diagnosis → component failure"; 2. CFI increases first (carrier deteriorates first) → OFI increases later (OBD fails later), which is judged as "severe carrier deterioration → OBD diagnostic logic collapse → missed fault". 3. The LFI increases first (urea fails first) → CFI increases later (carrier deteriorates later), which is judged as "urea failure → abnormal NOx conversion → accelerated carrier deterioration"; at the same time, the core triggering components of the failure chain are marked to provide accurate direction for operation and maintenance.

[0049] The OBD diagnostic logic self-calibration algorithm addresses the pain point of existing OBD diagnostic logic being fixed and unable to adapt to aging vehicles. This algorithm innovatively constructs an OBD diagnostic logic benchmark model and dynamic threshold library, achieving self-calibration through real-time deviation comparison, and synchronously linking upstream and downstream modules to adjust parameters, thereby improving the long-term operational stability of the system. The specific calculation method and process are as follows (strengthening the linkage between calibration logic and parameters): 1) Construction of benchmark model and threshold library: 1. Select OBD data from 100 normal vehicles (without SCR-related faults) and collect data under normal operating conditions (T... exh Frequency of DTC reporting, change in fault counter, and SCR (for conditions exceeding 220℃, load of 30%-70%, and continuous operation for ≥60s) RDY State, construct a benchmark dataset; 2. Calculation of baseline parameters: a) DTC reporting baseline threshold: Under normal operating conditions, the frequency of SCR-related DTC reporting is ≤ 1 time / 100h; b) Fault counter baseline threshold: Under normal operating conditions, the change in the fault counter is ≥ 0.1 / 10min (normally changing with operating conditions); c) SCR RDY Reference threshold: Under normal operating conditions, SCR RDY The duration of =1 accounts for ≥95%; the above parameters are compiled into a benchmark threshold library; 2) Self-calibration calculation process (deviation triggering + collaborative calibration to achieve system-wide parameter linkage): 1. Real-time data acquisition: Acquires real-time output data from the vehicle's OBD (DTC reporting frequency, fault counter changes, SCR). RDY (Status), calculate the average value over the past hour to ensure data representativeness; 2. Deviation Calculation: Calculate the relative deviation between the real-time parameter and the benchmark threshold. The deviation calculation formula is: Deviation = |Real-time value - Benchmark value| / Benchmark value × 100%, which accurately quantifies the degree of deviation. 3. Calibration trigger judgment: If the deviation of any parameter is greater than 15% for a sustained period of ≥30 min to avoid instantaneous deviation triggering false calibration, the self-calibration mechanism will be triggered. 4. Calibration execution (precise calibration for different scenarios, adapted to vehicle aging characteristics): a) If the frequency deviation of DTC reporting exceeds the standard (e.g., frequent reporting when there is no fault or failure to report when there is a fault), adjust the judgment time for NOx abnormality without fault code (from 30min to 25min) to improve the sensitivity of DTC reporting. b) If the deviation of the fault counter exceeds the standard (such as counter jamming or abnormal fluctuation), adjust the judgment threshold of fault counter jamming (from "constantly unchanged" to "change < 0.05 / 10min"). c) If SCR RDY Excessive deviation in operating conditions (e.g., SCR under normal operating conditions) RDY (Continued at 0), correct SCR RDY The temperature threshold for determining =0 has been adjusted from 240℃ to 230℃. 5. Feedback Adjustment (Cross-module Collaborative Calibration to Ensure System Consistency): The calibrated parameters are synchronously fed back to the NOx Conversion Characteristic Analysis Module and the Urea Liquid Level Anomaly Analysis Module. The CFI correction coefficient of the NOx Conversion Characteristic Analysis Module (±0.1) and the LFI judgment threshold of the Urea Liquid Level Anomaly Analysis Module (±0.05) are adjusted to achieve linkage of parameters across the entire system and avoid judgment deviations caused by calibration of a single module.

[0050] The role of the OBD diagnostic logic self-calibration algorithm is to achieve collaborative calibration between the OBD diagnostic logic and upstream and downstream modules, improve the long-term stability of the overall recognition system, and solve the problem that the existing OBD diagnostic logic is fixed and cannot adapt to parameter drift after vehicle aging.

[0051] The multi-factor fusion decision module innovatively constructs a comprehensive failure index (SFI) based on the carrier failure index (CFI), liquid level failure index (LFI), and OBD failure index (OFI). It combines dynamic weight adaptation, Bayesian-fuzzy fusion, and DS evidence theory for triple optimization to achieve accurate determination of the failure level and root cause localization of the SCR system (breaking through the limitations of existing fixed weight and hard threshold determination). The method implemented by the multi-factor fusion decision module is described in detail below.

[0052] In one embodiment, the step of inputting CFI, LFI, and OFI into a multi-factor fusion decision module for fusion decision-making, outputting a comprehensive failure index SFI, and determining the failure level and root cause includes: The comprehensive failure index (SFI) is constructed as follows: SFI = α×CFI +β×LFI +γ×OFI, Where α, β and γ are the dynamic adaptive weights of CFI, LFI and OFI, respectively; If the judgment results of the NOx conversion feature analysis module and the urea liquid level anomaly analysis module are consistent, the values ​​of α and β will be increased; if the judgment results conflict, the corresponding weights will be decreased. When SFI < 0.3, the SCR system is normal; when 0.3 ≤ SFI < 0.5, the SCR system is under warning (minor abnormality, requiring attention); when 0.5 ≤ SFI < 0.8, the SCR system is slightly faulty (minor fault of a single component); when SFI ≥ 0.8, the SCR system is severely faulty (serious fault of a single component or coupling failure of multiple components). The root cause of failure is determined based on the proportions of CFI, LFI, and OFI. If CFI has the highest proportion, the root cause is SCR carrier degradation or failure. If LFI has the highest proportion, the root cause is urea system failure, liquid level sensor soft failure, or tampering. If OFI has the highest proportion, the root cause is OBD diagnostic system failure. If the difference between the proportions of CFI, LFI, and OFI is ≤10%, it is determined to be a multi-component coupled failure. The temporal changes of CFI, LFI, and OFI are synchronously correlated to identify the failure chain propagation path and core triggering components, and to locate the source of failure.

[0053] In practical implementation, a Comprehensive Failure Index (SFI) is constructed (with dynamic adaptive weights to overcome the limitations of fixed weights): SFI = α×CFI + β×LFI + γ×OFI, where α, β, and γ are dynamic adaptive weights. Based on a Bayesian network, the index is automatically calibrated and adjusted according to the vehicle's real-time operating conditions (engine speed, load, exhaust temperature) and the joint analysis results between modules (consistency / conflict), ensuring optimal recognition accuracy under different operating conditions and scenarios. Specific calibration logic (enhancing reproducibility): 1. Operating condition adaptation calibration: High load (Load > 70%), high exhaust temperature (T) exh At temperatures >300℃, the carrier degradation effect is significant, with α=0.6, β=0.2, and γ=0.1; at low loads (Load < 30%) and low exhaust temperatures (220℃ < T), the degradation effect is also significant. exh At ≤260℃, the urea system is significantly affected, with α=0.4, β=0.4, and γ=0.2; under moderate operating conditions, α=0.5, β=0.3, and γ=0.2. 2. Module Consistency Calibration: If the NOx conversion feature analysis module and the urea level anomaly analysis module make the same judgment (both are abnormal or both are normal), then the corresponding weight is increased by 0.1 (if both are abnormal, α and β are each increased by 0.1, and γ is decreased by 0.2); if the judgment results conflict, then the corresponding weight is decreased by 0.1, and the OFI weight is increased by 0.1 (if CFI is abnormal and LFI is normal, α is decreased by 0.1, and γ is increased by 0.1). By adjusting the weights, the module conflict is balanced, and the accuracy of decision-making is improved.

[0054] Further optimization (fuzzy logic reasoning, overcoming the limitations of hard threshold judgment): Fuzzy logic reasoning (FIS) is introduced, and the SFI is input into the fuzzy logic system. Four fuzzy sets are defined: "normal, slight, moderate, and severe". The joint fuzzy information between modules (such as the fuzzy intersection of "high CFI and high LFI") is processed through triangular membership functions. The specific membership function settings are as follows: Normal (SFI < 0.3): membership function y = 1 - 3.33 × SFI; Slight (0.3 ≤ SFI < 0.5): y = 5 × SFI - 1.5 (rising segment), y = 3.5 - 7 × SFI (falling segment); Moderate (0.5 ≤ SFI < 0.8): y = 2 × SFI - 1 (rising segment), y = 4 - 5 × SFI (falling segment); Severe (SFI ≥ 0.8): y = 5 × SFI - 4. Through membership calculation, the "one-size-fits-all" problem of hard threshold judgment is avoided, and the decision-making error tolerance is improved, which is especially suitable for boundary value scenarios.

[0055] Furthermore, a new multi-source data fusion and noise reduction algorithm is added (addressing the pain points of data conflict between multiple modules and misjudgment by a single module): It innovatively introduces DS evidence theory, using the output results (CFI, LFI, OFI) of the NOx conversion feature analysis module, urea level anomaly analysis module, and OBD diagnostic system failure analysis module as independent evidence sources. Through confidence allocation, synthesis, and calibration, it filters out interference from misjudgments by a single module, improving decision accuracy. The specific calculation method and process are as follows: 1. Introduce evidence source credibility calculation (dynamic correction factor, combined with actual vehicle condition to calibrate credibility): a) Define evidence sources: Evidence source A = NOx conversion feature analysis module (CFI), evidence source B = urea level anomaly analysis module (LFI), evidence source C = OBD diagnostic system failure analysis module (OFI), corresponding to the judgment results of the three core components of carrier, urea and OBD respectively; b) Initial confidence level allocation: Based on the measured recognition accuracy of each module (NOx conversion feature analysis module accuracy 93%, urea level anomaly analysis module accuracy 95%, OBD diagnostic system failure analysis module accuracy 97%), the initial confidence levels are allocated as follows: m(A) = 0.93, m(B) = 0.95, m(C) = 0.97, and the uncertainty m(Ω) = 1 - ((m(A) + m(B) + m(C)) / 3 = 0.05 (Ω is the total set, representing the uncertain state). c) Dynamic credibility correction (combining vehicle operating status to improve credibility adaptability): Combining the vehicle's cumulative operating time (t), mileage (M), and historical fault records (F), a dynamic credibility correction factor k is constructed, with the formula: k=1-0.01×(t / 10000)+0.005×(F / 10), where t is in hours and F is the number of SCR-related faults in the past year; the corrected credibility is: m'(A)=m(A)×k, m'(B)=m(B)×k, m'(C)=m(C)×k, and the corrected uncertainty is m'(Ω)=1-((m'(A)+m'(B)+m'(C)) / 3, ensuring more accurate credibility calibration for older vehicles and vehicles with frequent faults; 2. Evidence synthesis and fusion decision-making process (step-by-step synthesis + weighted linkage to achieve noise reduction and decision optimization): a) Evidence synthesis rules: Using the DS evidence synthesis formula, first synthesize evidence source A (carrier) and B (urea) to obtain synthesized evidence m1, then synthesize m1 with evidence source C (OBD) to obtain the final fused evidence m_final; Synthesis formula (clearly define the calculation logic to ensure reproducibility): m(X)=[m1(X)×m2(X)+m1(X)×m2(Ω)+m1(Ω)×m2(X)] / [1-m1(X)×m2(Y)-m1(Y)×m2(X)], where X and Y are mutually exclusive evidence (such as A being abnormal and B being normal), and the denominator is used to avoid synthesis conflicts; b) Fusion result calibration and weight linkage (fusion results are directly associated with dynamic weights to improve decision consistency): The fused evidence is mapped to the weight adjustment of CFI, LFI, and OFI. If m_final(A) is the largest (the carrier evidence is the most credible), the CFI weight is increased by 0.1; if m_final(B) is the largest (the urea evidence is the most credible), the LFI weight is increased by 0.1; if m_final(C) is the largest (the OBD evidence is the most credible), the OFI weight is increased by 0.1. c) Noise reduction effect: By fusing multiple evidences, the interference of misjudgment by a single module is filtered out (such as a single misjudgment of CFI abnormality by the NOx conversion feature analysis module, which can be offset by normal evidence from the urea level abnormality analysis module and the OBD diagnostic system failure analysis module, thus reducing the false alarm rate). It is especially suitable for multi-module data conflict scenarios under complex working conditions.

[0056] The role of the newly added multi-source data fusion and noise reduction algorithm is to further improve the accuracy of failure level determination and root cause location, and to solve the decision-making bias problem caused by multi-module data conflicts and misjudgment of a single module.

[0057] Furthermore, for root cause identification, a combination of proportion determination and failure chain tracing is adopted to overcome the pain point of existing methods that cannot locate coupled failure sources. Specifically, this includes: 1. Single failure root cause determination: The root cause is determined based on the proportion of each index, specifically: CFI has the highest proportion (>40%): the root cause is SCR carrier deterioration / failure; LFI has the highest proportion (>30%): the root cause is urea system failure, liquid level sensor soft failure or cheating; OFI has the highest proportion (>20%): the root cause is OBD diagnostic system failure itself. 2. Root Cause Location of Coupled Failures: If the proportions of the three are close (difference ≤ 10%), it is determined to be a multi-component coupled failure. The temporal change sequence of CFI, LFI, and OFI is synchronously correlated to identify the propagation path of the failure chain and the core triggering component (e.g., "carrier degradation → OBD false alarm → urea injection shielding"). The source of failure is located, and a complete root cause report of "core triggering component + propagation path" is output to provide accurate guidance for operation and maintenance. a) CFI accounts for the highest proportion (>40%): The root cause of failure is SCR carrier degradation / failure; b) LFI accounts for the highest proportion (>30%): The root cause of failure is urea system failure, soft failure of liquid level sensor or cheating; c) OFI accounts for the highest proportion (>20%): The root cause of failure is the failure of the OBD diagnostic system itself; d) The proportions of the three are close: the root cause of failure is the coupling failure of multiple components (failure chain). At the same time, by associating the temporal change sequence of CFI, LFI, and OFI, the propagation path of the failure chain and the core triggering component are identified, and the source of failure is located.

[0058] This application presents a multi-level coupled failure identification method for SCR based on remote OBD. Firstly, it achieves multi-level coupled identification, breaking down and analyzing failures layer by layer, from carrier degradation and soft failures of the urea system to OBD diagnostic failures themselves. This addresses the limitation of existing technologies, which can only identify single-component failures and cannot identify coupled failures of multiple components. Secondly, it innovates a scenario-specific dynamic self-calibration mechanism to adapt to the parameter drift characteristics after vehicle aging, solving the problem of declining accuracy of traditional fixed diagnostic logic as vehicle age increases, effectively improving long-term identification stability. Thirdly, through dynamic weight adaptation, Bayesian-fuzzy fusion, and DS evidence theory triple optimization, it overcomes the limitations of traditional fixed weights and hard threshold judgments, effectively reducing the misjudgment rate of single modules and improving identification accuracy under complex operating conditions. Fourthly, it enables failure chain tracing of coupled failures, locating the core triggering failure source and providing precise guidance for the operation and maintenance of the post-processing system.

[0059] Secondly, embodiments of this application also provide a multi-level coupling failure identification system for SCR based on remote OBD, used to execute the multi-level coupling failure identification method for SCR based on remote OBD as described in any embodiment of the first aspect, the system comprising: The data acquisition and operating condition filtering module is used to collect the mandatory data that must be transmitted by the vehicle remote OBD, and to filter the collected mandatory data for effective operating conditions, and to remove invalid data. The mandatory data includes exhaust temperature, upstream NOx concentration, downstream NOx concentration, urea level, engine speed, engine load, vehicle speed, SCR related fault codes, OBD monitoring ready status, fault confirmation counter, mileage and cumulative running time. The NOx conversion characteristic analysis module is used to calculate the carrier failure index (CFI) and predict the carrier degradation trend based on the effective operating data obtained after screening. The urea level anomaly analysis module is used to calculate the level failure index (LFI). The OBD diagnostic system failure analysis module is used to calculate the OBD failure index OFI. The multi-factor fusion decision module is used to make fusion decisions based on CFI, LFI and OFI, output the comprehensive failure index SFI and determine the failure level and root cause. The collaborative analysis module is used for communication and connection between modules to form a closed-loop feedback mechanism. The output results of any module are fed back to other relevant modules to adjust calculation parameters, identify faults, and trace the source of failure chains.

[0060] The algorithm in this application is deployed on a remote OBD monitoring platform for China VI heavy-duty vehicles. The platform has the functions of remote OBD data acquisition, storage, real-time calculation, and result output. The data source is the mandatory data transmitted by the vehicle's OBD terminal according to national standards, without the need for any additional sensors or hardware, and is compatible with all China VI heavy-duty vehicles. It also supports multi-scenario adaptation and can automatically adjust the algorithm parameters (time window length, weight coefficient) according to different driving scenarios such as urban roads, highways, and mountain roads to ensure high-precision recognition in various scenarios. The platform supports functions such as failure warning push, failure chain graph display, and operation and maintenance suggestion output to realize closed-loop management of "identification-early warning-operation and maintenance".

[0061] The method of this application is described below with a specific example, taking a certain China VI heavy-duty truck as an example. The specific implementation steps are as follows: (1) Data collection The remote OBD monitoring platform collects the exhaust temperature (T) of the truck in real time. exh ), upstream NOx concentration (C in ), downstream NOx concentration (C out Urea level (L), engine speed (n), load, vehicle speed (V), SCR-related DTCs, SCR RDY The system collects data on fault confirmation counters, mileage, and cumulative running time, with a sampling frequency of 1 time per second.

[0062] (2) Working condition screening Filter out T exh =Valid operating data at 250℃, speed fluctuation of 8%, load fluctuation of 10%, continuous operation for 70 seconds, and vehicle speed of 15km / h, excluding cold start (T exh Invalid data such as 80℃ and idling speed (vehicle speed = 0) were included. At the same time, the K-means operating condition similarity matching algorithm was introduced to calculate the cosine similarity between the real-time operating condition data and the preset "typical valid operating condition library". The similarity was 0.88 (≥0.85), and the match was successful. No EMA smoothing was required. The operating condition was marked as a medium load condition (Load=55%) and pushed to the subsequent modules.

[0063] (3) Carrier degradation identification Calculate C under effective operating conditions in The average value is 800 ppm, C out The average value is 300 ppm, η NOx =1-300 / 800=0.625; C out The standard deviation is 120 ppm for NOx. var =120 / 800=0.15; Due to stable operating conditions, the sliding time window is automatically set to 60s, and the η of the three consecutive time windows... NOx The values ​​were 0.625, 0.61, and 0.59, respectively, with a standard deviation of 0.0175. (NOx) sta =1-0.0175=0.9825; CFI=0.4×(1-0.625)+0.3×0.15+0.3×(1-0.9825)=0.15+0.045+0.00525=0.20025, the carrier is judged to be normal; at the same time, the LSTM carrier degradation trend prediction model is called, and the specific calculation process is as follows: 1) Input data: CFI time series data for the past 30 days, corresponding operating condition labels (mainly medium load, accounting for 70%), 2 sets of abnormal CFI data were removed (3σ criterion, deviation from the mean ± 3 times the standard deviation), and the CFI data were normalized by min-max, with the normalized range being [0.15, 0.22]. 2) Model inference: Input the normalized input data into the trained LSTM model (2 hidden layers, 32 neurons, learning rate 0.001), and output the CFI prediction values ​​for the next 15 days as 0.21, 20 days as 0.215, and 30 days as 0.22; 3) Precision correction: Current NOx var =0.15, correction factor k=1-0.2×0.15=0.97, the corrected CFI prediction values ​​are 0.2037, 0.20855 and 0.2134 respectively, none of which reach the mild failure threshold (0.5), so no warning needs to be triggered.

[0064] (4) Urea system identification The truck traveled continuously for 100km, during which the urea level dropped from 30% to 28%, ΔL=2%, ΔMileage=100km, L stuck =0.02 / 100=0.0002≈0; No level jump in non-filling state (L) jump =0); liquid level always >10% (L low_no_dtc =0); LFI=0.4×0.0002+0.3×0+0.3×0≈0.00008, the urea system is judged to be normal; at the same time, the isolated forest algorithm is introduced to trace the source of anomalies in the liquid level time series data. The specific calculation process is as follows: 1) Input data: Time series data of urea liquid level over the past 7 days (1008 sets in total, once every 10 minutes), and corresponding driving mileage data, to construct a sample set; 2) Anomaly detection: Input the sample set into the isolated forest model (100 decision trees, 256 samples, anomaly score threshold of 0.7), calculate the anomaly score of all samples, and if all scores are <0.7, there are no abnormal samples. 3) Anomaly location: The anomaly time point is empty, and the anomaly type is unknown; 4) Health score: The frequency of abnormal liquid level in the past 30 days is 0, and the health score = 100 - (0 × 5) = 100 points, further verifying that there is no abnormality in the urea system.

[0065] (5) Identification of OBD diagnostic failure η NOx =0.625>50%(I(no_DTC)=0);T exh =250℃, run for 40 minutes, SCR RDY =1(I(SCR RDY =0)=0); The fault confirmation counter changes normally with the fluctuation of operating conditions (I(fault_cnt) stuck OFI = 0); OFI = 0+0+0 = 0, indicating the OBD diagnostic system is functioning normally; simultaneously, the OBD diagnostic logic self-calibration algorithm is activated. The specific calculation process is as follows: 1) Real-time parameter acquisition: Statistical analysis of OBD output parameters over the past hour, including DTC reporting frequency = 0 times / 100h, fault counter change = 0.12 / 10min, and SCR... RDY =1. The duration of =1 accounts for 100%; 2) Deviation calculation: Compare with the baseline threshold library (DTC reporting ≤ 1 time / 100h, fault counter change ≥ 0.1 / 10min, SCR... RDY(Percentage ≥ 95%), calculate the deviation: DTC deviation = |0 - 1| / 1 × 100% = 100% (but not lasting 30 minutes, calibration is not triggered), fault counter deviation = |0.12 - 0.1| / 0.1 × 100% = 20%, SCR RDY Deviation = |100% - 95%| / 95% × 100% ≈ 5.26%; 3) Calibration judgment: The fault counter deviation is 20% > 15%, but the duration is only 10 minutes within 1 hour, which does not reach 30 minutes, so self-calibration is not triggered; 4) Feedback: The calibration results are fed back to the NOx conversion characteristic analysis module and the urea level anomaly analysis module, maintaining the original judgment parameters unchanged.

[0066] (6) Integrated decision-making Introducing the DS evidence theory, CFI (0.20025), LFI (0.00008), and OFI (0) are considered as independent sources of evidence. The specific calculation process is as follows: 1) Initial confidence level allocation: m(A) = 0.93 (NOx conversion feature analysis module), m(B) = 0.95 (urea level anomaly analysis module), m(C) = 0.97 (OBD diagnostic system failure analysis module), uncertainty m(Ω) = 0.05; 2) Calculation of dynamic correction factor: Vehicle cumulative operating time t = 5000 hours, mileage M = 100000 km, historical fault records F = 0 times, k = 1 - 0.01 × (5000 / 10000) + 0.005 × (0 / 10) = 0.995; Corrected reliability: m'(A) = 0.93 × 0.995 ≈ 0.925, m'(B) = 0.95 × 0.9 95≈0.945, m'(C)=0.97×0.995≈0.965, uncertainty m'(Ω)=1-(0.925+0.945+0.965)=0.165 (Here, because the sum of confidence levels exceeds 1, it is re-normalized. After normalization, m'(A)=0.31, m'(B)=0.32, m'(C)=0.33, m'(Ω)=0.04). 3) Evidence synthesis: First synthesize A and B, obtaining m1(A)=0.48, m1(B)=0.50, m1(Ω)=0.02; then synthesize m1 and C, obtaining m_final(A)=0.37, m_final(B)=0.38, m_final(C)=0.23, m_final(Ω)=0.02; 4) Weight adjustment: m_final(B) is maximized, LFI weight is increased by 0.1 (from 0.3 to 0.4), CFI weight is increased by 0.4, and OFI weight is increased by 0.2; 5) Final calculation: SFI = 0.4 × 0.20025 + 0.4 × 0.00008 + 0.2 × 0 ≈ 0.0801, SFI < 0.3. Combining fuzzy logic reasoning (membership function calculation, the membership degree of SFI belonging to the "normal" fuzzy set is 0.92), the SCR system is determined to be normal, and the result "System normal, no maintenance required" is output.

[0067] Another example: When the truck's chassis ages, under effective operating conditions, C in The average value is 800 ppm, C out The average value is 600 ppm, η NOx =0.25; Cout standard deviation is 200ppm, NOx var =0.25; Due to the current high load condition (Load=75%), K-means load condition similarity matching is used for specific calculation: 1) Real-time operating characteristics: Load=75%, T exh =320℃, n=1800r / min, V=80km / h, after normalization it is [1,1,1,1]; 2) Cosine similarity calculation: Clustering centers with typical high-speed operating conditions (Load=75%, T) exh Similarity of (320℃, n=1800r / min, V=80km / h):

[0068] 3) Matching successful; the sliding time window automatically adjusts to 60 seconds, with 3 time windows η NOx The values ​​were 0.25, 0.22, and 0.18, respectively, with a standard deviation of 0.035. (NOx) sta=0.965; CFI=0.4×(1-0.25)+0.3×0.25+0.3×(1-0.965)=0.3+0.075+0.0105=0.3855; The LSTM trend prediction model is called, and the CFI time series data of the past 30 days (mean 0.35) and the high load condition label are input. The model outputs the predicted CFI value of 0.53 for the next 20 days. The correction factor k=1-0.2×0.25=0.95, and the corrected value is 0.5035, which reaches the mild failure threshold (0.5), triggering an early warning; At the same time, the data of the urea level anomaly analysis module is called, and LFI=0.0 (urea is normal). The NOx conversion feature analysis module and the urea level anomaly analysis module determine the result. Consistent with the previous results, the CFI weight was increased to 0.6; OFI=0; during the fusion decision, the data from the three modules were fused using the DS evidence theory, with a dynamic credibility correction factor k=0.98 (vehicle cumulative operation of 8000 hours, 1 failure). After correction, the credibility m'(A)=0.91, m'(B)=0.93, and m'(C)=0.95. After synthesis, m_final(A) was maximized, and the CFI weight remained at 0.6; SFI=0.6×0.3855+0.2×0+0.2×0≈0.2313. Combined with fuzzy logic reasoning, the SCR system was judged to issue a warning, the root cause of which was mild carrier degradation. The result was output: "The carrier is mildly degraded. Regular inspection is recommended. It is expected to reach a mild failure state in 20 days."

[0069] Another example (coupling failure scenario): A certain China VI heavy-duty truck, under effective operating conditions T exh =280℃, Load=60% (medium operating condition), K-means similarity matching shows a similarity of 0.86 with typical mountain operating conditions, indicating a successful match; C in The average value was 900 ppm, C out The average value is 720 ppm, η NOx =0.2, NOx var =0.22, NOx sta =0.95, CFI=0.4×(1-0.2)+0.3×0.22+0.3×(1-0.95)=0.32+0.066+0.015=0.401; The LSTM trend prediction model is called, inputting the CFI time series data of the past 30 days (mean 0.38) and the medium load condition label, predicting that the CFI will rise to 0.62 in the next 15 days, the correction factor k=1-0.2×0.22=0.956, the corrected value is 0.5927, triggering a medium failure warning; the urea level remains unchanged for 200km, and the source is traced using the isolated forest algorithm, specifically calculated as follows: ① Input the liquid level data of the past 7 days. There are 12 groups of samples with anomaly scores ≥0.7. All of them have ΔL≈0 and ΔMileage>50km. ② The anomaly type was determined to be "urea injection shielding", and the anomaly time point was located at 14:30 three days ago; ③ Health score: 6 abnormal occurrences in the past 30 days, health score = 100 - (6 × 5) = 70 points, adjusted based on NOx conversion efficiency η. NOx =0.2 < 50%, verifying the validity of the anomaly tracing results, the final health score was adjusted to 65; OBD did not report any SCR-related DTCs (I(no_DTC)=1), SCR RDY =0 for 40 minutes (I(SCR) RDY =0)=1), the fault counter remains unchanged (I(fault_cnt) stuck =1), OFI=3; The OBD self-calibration algorithm is activated, comparing the deviation between the actual OBD output and the baseline model: DTC reporting frequency deviation = |0-1| / 1×100%=100%, lasting 40min>30min, triggering self-calibration, correcting the judgment time for NOx abnormality without fault codes from 30min to 25min, and simultaneously feeding back to the NOx conversion feature analysis module and the urea level anomaly analysis module, adjusting the CFI correction coefficient of the NOx conversion feature analysis module from 0.5 to 0.4, and the LFI judgment threshold of the urea level anomaly analysis module from 0.5 to... 0.4; Module joint analysis: CFI anomaly occurs before OFI anomaly, and the failure chain is determined to be "moderate carrier degradation → OBD diagnostic logic jamming → missed fault"; During the fusion decision, due to the conflict between the NOx conversion feature analysis module and the OBD diagnostic system failure analysis module, the weight of CFI is reduced to 0.4 and the weight of OFI is increased to 0.3. The data is fused through DS evidence theory, and the dynamic credibility correction factor k=1.05 (vehicle cumulative operation 10,000 hours, 3 failures). After correction, the credibility m'(A)=0.97, m'(B)=0.99, and m'(C)=1.

[0070] The embodiments provided by the present invention have the following advantages: Technical aspects: Breaking through the dependence of existing SCR failure identification on detailed parameters, filling the technical gaps in the identification of OBD self-failure, sensor soft failure, and multi-component coupling failure; adding new algorithmic innovations such as dynamic weights, adaptive time windows, soft failure early warning, module joint verification, and Bayesian-fuzzy fusion decision-making, solving the problems of poor adaptability to working conditions, easy misjudgment, inability to provide early warning, and independent and unrelated modules in existing algorithms, further improving the accuracy, real-time performance, foresight, and robustness of SCR failure identification; Application level: No additional hardware is required. It can be directly deployed on the existing remote OBD monitoring platform, adapting to all China VI heavy-duty vehicles and multiple driving scenarios. It enables remote online diagnosis of SCR system failure, early warning, failure chain tracing and operation and maintenance suggestions, and builds a closed-loop management system of "identification-early warning-operation and maintenance". Practical value: It provides reliable remote monitoring technology for heavy-duty vehicle emissions to the ecological and environmental departments, and provides accurate SCR system operation and maintenance early warning for car manufacturers and logistics fleets, reducing operation and maintenance costs, reducing the risk of exceeding emission standards, and helping to implement the National VI emission standards.

[0071] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for identifying multi-level coupling failures in SCR based on remote OBD, characterized in that, include: The data acquisition and operating condition screening module collects the vehicle's remote OBD essential data, and performs effective operating condition screening on the collected essential data to remove invalid data. The essential data includes exhaust temperature, upstream NOx concentration, downstream NOx concentration, urea level, engine speed, engine load, vehicle speed, SCR related fault codes, OBD monitoring ready status, fault confirmation counter, mileage and cumulative running time. Based on the effective operating data obtained after screening, the carrier failure index (CFI) is calculated through the NOx conversion characteristic analysis module, and the carrier degradation trend is predicted. The liquid level failure index (LFI) is calculated using the urea liquid level anomaly analysis module. The OBD Failure Index (OFI) is calculated using the failure analysis module of the OBD diagnostic system. The CFI, LFI and OFI are input into the multi-factor fusion decision module for fusion decision, and the comprehensive failure index SFI is output to determine the failure level and root cause. The modules communicate with each other to form a closed-loop feedback mechanism, and the output of any module is fed back to other relevant modules to adjust calculation parameters, identify faults, and trace the source of failure chains.

2. The SCR multi-level coupling failure identification method based on remote OBD according to claim 1, characterized in that, The method further includes: The real-time effective operating condition data is matched with typical operating conditions in the preset typical effective operating condition library based on the K-means clustering method. If the match is successful, the calculation parameters corresponding to the matched typical working conditions will be used to perform calculations for the NOx conversion characteristic analysis module, the urea level anomaly analysis module, and / or the OBD diagnostic system failure analysis module. If a match fails, the real-time valid operating condition data will be smoothed using an exponential moving average method.

3. The SCR multi-level coupling failure identification method based on remote OBD according to claim 1, characterized in that, The calculation of the carrier failure index (CFI) through the NOx conversion characteristic analysis module includes: Based on the upstream and downstream NOx concentrations, the basic NOx conversion efficiency is calculated using the following expression: or NOx = 1 - (C out / C in ), Where, η NOx For the basic NOx conversion efficiency, C in C represents the average NOx concentration upstream. out This represents the average NOx concentration downstream. Based on the standard deviation of downstream NOx concentration and the mean of upstream NOx concentration, the NOx time-series fluctuation index is calculated, expressed as: NOx var = std(C out ) / mean(C in ), Among them, NOx var For NOx time series volatility index, std(C) out ) represents the standard deviation of downstream NOx concentration, mean(C in () represents the average upstream NOx concentration; The NOx conversion consistency index is calculated based on the standard deviation of NOx conversion efficiency, expressed as: NOx sta = 1 - std(η NOx,window) , Among them, NOx sta std(η) is the NOx conversion consistency index. NOx,window () represents the standard deviation of NOx conversion efficiency within multiple consecutive sliding time windows; The carrier failure index (CFI) is obtained by weighted summing of the NOx basic conversion efficiency, NOx time-series fluctuation index, and NOx conversion consistency index. A higher CFI value indicates more severe SCR carrier degradation. The expression for the carrier failure index CFI is: CFI = C1×(1-η NOx ) + C2×NOx var + C3×(1-NOx sta ), C1 + C2 + C3 = 1, Among them, C1, C2 and C3 are the first coefficient, the second coefficient and the third coefficient of the carrier failure index, respectively; The calculation of the liquid level failure index (LFI) through the urea liquid level anomaly analysis module includes: Calculate the liquid level stagnation characteristic L stuck The expression is: L stuck = |ΔL| / ΔMileage, Where ΔL is the change in urea level after traveling a certain distance, and ΔMileage is the corresponding travel distance. stuck When the value is approximately 0, it is determined that the liquid level sensor is stuck or the urea injection is blocked. Calculate the unreasonable liquid level jump characteristics L jump When the urea level changes by more than 5% in a single instance while not being filled, L jump = 1, otherwise L jump =0; Calculate the low liquid level fault code characteristic L low_no_dtc When the urea level is <10% and the vehicle has been driven for >50km, but the OBD does not report any urea level-related fault codes, L low_no_dtc = 1, otherwise L low_no_dtc =0; Based on the liquid level stagnation feature L stuck Unreasonable liquid level jump characteristics L jump And the low liquid level without fault code feature L low_no_dtc The Liquid Level Failure Index (LFI) is obtained. The higher the LFI value, the higher the probability of urea system failure or cheating. The expression for the Liquid Level Failure Index (LFI) is: LFI =L1×L stuck + L2×L jump + L3×L low_no_dtc , L1 + L2 + L3 = 1, Among them, L1, L2 and L3 are the first coefficient, the second coefficient and the third coefficient of the liquid level failure index, respectively. The calculation of the OBD Failure Index (OFI) through the failure analysis module of the OBD diagnostic system includes: When performing NOx abnormality fault code judgment, if the basic NOx conversion efficiency is consistently <50% but the OBD does not report SCR-related fault codes, I(no_DTC) = 1; otherwise, I(no_DTC) = 0. Here, I(no_DTC) is the NOx abnormality fault code characteristic. Execute monitoring readiness anomaly judgment when exhaust temperature > 240℃, vehicle continuous operating time > 30min, and OBD monitoring readiness status SCR. RDY When I(SCR) remains at 0, RDY =0) = 1, otherwise I(SCR) RDY =0)=0, where I(SCR RDY =0) indicates an abnormal monitoring readiness characteristic; When the fault counter is stuck, if there are obvious SCR failure characteristics but the OBD fault confirmation counter value remains unchanged, I(fault_cnt) is used for fault counter judgment. stuck ) = 1, otherwise I(fault_cnt stuck =0, where I(fault_cnt) stuck This indicates a fault counter is stuck. Based on the NOx anomaly without fault codes, the monitoring readiness anomaly, and the fault counter jamming characteristic, the OBD Failure Index (OFI) is calculated. When OFI ≥ 2, it is determined that the OBD diagnostic system itself has failed. The expression for the OBD Failure Index (OFI) is: OFI = I(no_DTC) + I(SCR RDY =0) + I(fault_cnt stuck )。 4. The SCR multi-level coupling failure identification method based on remote OBD according to claim 3, characterized in that, The process of predicting carrier degradation trends includes: Construct and train an LSTM time series prediction model; The carrier failure index (CFI) is obtained over a historical period to form CFI time-series data. The CFI time series data is input into the trained LSTM time series prediction model. Through the forget gate, input gate and output gate of the hidden layer, the long-term trend of CFI changes is captured and the predicted CFI values ​​for multiple key time nodes in the future are output. Based on the NOx time series volatility index, a correction factor is constructed, and the CFI forecast value is corrected based on the correction factor to obtain the final corrected CFI forecast value. Different warning thresholds are preset. If the corrected CFI prediction value exceeds the warning threshold of the corresponding level, the warning operation corresponding to that level will be triggered, and the warning time node and the predicted degradation rate will be marked.

5. The SCR multi-level coupling failure identification method based on remote OBD according to claim 3, characterized in that, The output of any module is fed back to other relevant modules to adjust calculation parameters, identify faults, and trace failure chains. This includes dynamic correction and bidirectional verification of coefficients in the NOx conversion characteristic analysis module based on the output of the urea level anomaly analysis module and the OBD diagnostic system failure analysis module. The dynamic correction of the coefficients and the two-way verification include: Based on the liquid level failure index (LFI), the urea liquid level anomaly analysis module outputs a urea system reliability score, where the urea system reliability score = 1 - LFI. If CFI ≥ 0.3 and LFI > 0.5, then CFI is discounted and corrected. The correction formula is: CFI correction = CFI × 0.5 × urea system reliability score, in order to avoid misjudging abnormal urea supply as carrier deterioration. If CFI < 0.3, but LFI > 0.8, then increase the calculation weight of the first coefficient of the carrier failure index to investigate hidden carrier degradation. Simultaneously, the OBD monitoring readiness status output by the OBD diagnostic system failure analysis module is invoked. If the current operating conditions meet the effective working conditions of the SCR, but the SCR... RDY If the value is 0, the carrier degradation determination is paused, and the OBD diagnostic system failure analysis module is triggered to further check whether the OBD diagnostic logic has failed.

6. The SCR multi-level coupling failure identification method based on remote OBD according to claim 3, characterized in that, The output of any module is fed back to other relevant modules to adjust calculation parameters, identify faults, and trace failure chains. This includes identifying hidden faults within the urea injection system and OBD concealment in the urea level anomaly analysis module based on the output of the NOx conversion characteristic analysis module and the OBD diagnostic system failure analysis module. The identification of hidden faults and OBD concealment within the urea injection system includes: LFI and NOx var Perform a linkage analysis; if LFI < 0.3, but NOx var If the value is greater than 0.2, it is determined that there is a hidden fault inside the urea injection system. The calculation weight of the first coefficient of the liquid level failure index will be automatically increased to further investigate hidden faults. The LFI and SCR-related fault codes are checked for consistency. If LFI > 0.8, but the OBD diagnostic system failure analysis module has no corresponding SCR-related fault code and OFI < 2, it is marked as OBD concealment or masking, triggering the OBD diagnostic system failure analysis module to re-check the OBD diagnostic logic.

7. The SCR multi-level coupling failure identification method based on remote OBD according to claim 3, characterized in that, The output of any module is fed back to other relevant modules to adjust calculation parameters, identify faults, and trace failure chains, including: If exhaust temperature T is detected exh If the data fluctuation is greater than 50℃ / min and lasts for ≥10s, a data anomaly signal is sent to the NOx conversion characteristic analysis module, the carrier degradation determination is paused, and the process continues until the exhaust temperature T... exh Recalculate CFI after the data stabilizes; Record the time sequence of anomalies in CFI, LFI, and OFI, and identify the propagation path of the failure chain through time-series correlation analysis based on the failure chain determination rules; Based on the current vehicle's real-time output of SCR-related fault code reporting frequency, fault confirmation counter changes, and OBD monitoring readiness status, the OBD diagnostic logic of the OBD diagnostic system failure analysis module is self-calibrated. The self-calibrated parameters are then synchronously fed back to the NOx conversion characteristic analysis module and the urea level anomaly analysis module, and the coefficients and LFI judgment thresholds in the CFI expression are adjusted.

8. The SCR multi-level coupling failure identification method based on remote OBD according to claim 3, characterized in that, The method further includes: Construct and train an isolated forest anomaly detection model; Extract the time-series data of urea liquid level and the corresponding driving mileage data within the first preset historical time period to construct the input dataset. Each sample includes the urea liquid level value, the urea liquid level change ΔL, the corresponding driving mileage ΔMileage, and the timestamp. The input dataset is fed into the trained Isolation Forest anomaly detection model. By randomly partitioning the feature space, anomaly scores are calculated for each sample, and anomalous samples are selected. Perform feature analysis on abnormal samples to determine the anomaly type; By combining the NOx conversion efficiency output by the NOx conversion feature analysis module, we can verify whether the anomaly type determination result is a misjudgment. If it is a misjudgment, we can adjust the threshold of the anomaly score of the isolated forest anomaly detection model. The frequency of abnormal liquid levels within a second preset time period is counted. The duration of the second preset time period is longer than that of the first preset time period. A health score of the urea system is constructed based on the frequency of abnormal liquid levels. If the health score is lower than the threshold, a soft failure warning is triggered.

9. The SCR multi-level coupling failure identification method based on remote OBD according to claim 3, characterized in that, The step of inputting CFI, LFI, and OFI into a multi-factor fusion decision module for fusion decision-making, outputting a comprehensive failure index SFI, and determining the failure level and root cause includes: The comprehensive failure index (SFI) is constructed as follows: SFI = α×CFI +β×LFI +γ×OFI, Where α, β and γ are the dynamic adaptive weights of CFI, LFI and OFI, respectively; If the judgment results of the NOx conversion feature analysis module and the urea liquid level anomaly analysis module are consistent, the values ​​of α and β will be increased; if the judgment results conflict, the corresponding weights will be decreased. When SFI < 0.3, the SCR system is normal; when 0.3 ≤ SFI < 0.5, the SCR system is under warning; when 0.5 ≤ SFI < 0.8, the SCR system is slightly faulty; when SFI ≥ 0.8, the SCR system is severely faulty. The root cause of failure is determined based on the proportions of CFI, LFI, and OFI. If CFI has the highest proportion, the root cause is SCR carrier degradation or failure. If LFI has the highest proportion, the root cause is urea system failure, liquid level sensor soft failure, or tampering. If OFI has the highest proportion, the root cause is OBD diagnostic system failure. If the difference between the proportions of CFI, LFI, and OFI is ≤10%, it is determined to be a multi-component coupled failure. The temporal changes of CFI, LFI, and OFI are synchronously correlated to identify the failure chain propagation path and core triggering components, and to locate the source of failure.

10. A multi-level coupled failure identification system for SCR based on remote OBD, used to execute the multi-level coupled failure identification method for SCR based on remote OBD as described in any one of claims 1-9, characterized in that, The system includes: The data acquisition and operating condition filtering module is used to collect the mandatory data that must be transmitted by the vehicle remote OBD, and to filter the collected mandatory data for effective operating conditions, and to remove invalid data. The mandatory data includes exhaust temperature, upstream NOx concentration, downstream NOx concentration, urea level, engine speed, engine load, vehicle speed, SCR related fault codes, OBD monitoring ready status, fault confirmation counter, mileage and cumulative running time. The NOx conversion characteristic analysis module is used to calculate the carrier failure index (CFI) and predict the carrier degradation trend based on the effective operating data obtained after screening. The urea level anomaly analysis module is used to calculate the level failure index (LFI). The OBD diagnostic system failure analysis module is used to calculate the OBD failure index OFI. The multi-factor fusion decision module is used to make fusion decisions based on CFI, LFI and OFI, output the comprehensive failure index SFI and determine the failure level and root cause. The collaborative analysis module is used for communication and connection between modules to form a closed-loop feedback mechanism. The output results of any module are fed back to other relevant modules to adjust calculation parameters, identify faults, and trace the source of failure chains.

Images