A method for dynamically monitoring low-carbon components of vehicle fuel based on multi-dimensional feature coupling

Abstract

The present application relates to the field of dynamic monitoring of low-carbon components of vehicle fuel, and specifically discloses a method for dynamic monitoring of low-carbon components of vehicle fuel based on multi-dimensional feature coupling. The method comprises: obtaining engine identification and closed-loop configuration, constructing a combustion feature fingerprint library through standardization and coupling simulation; performing working condition data synchronization and key working condition matching, triggering tail gas sampling and obtaining samples through pre-operation; performing isotope analysis, detection standardization and multi-dimensional feature coupling on the samples to form an observation vector; performing deviation calculation and online calibration parameter generation, low-carbon component blending ratio inversion, carbon emission factor calling and fuel consumption integral calculation, and generating a closed-loop configuration to realize cycle optimization. Through the multi-dimensional feature coupling and closed-loop calibration mechanism, the present application realizes dynamic and accurate monitoring of low-carbon components of vehicle fuel, effectively improves the accuracy and reliability of the monitoring results, and provides technical support for real-time carbon footprint assessment.

Description

Technical Field

[0001] This invention relates to the field of dynamic monitoring of low-carbon components in vehicle fuels, and more particularly to a method for dynamic monitoring of low-carbon components in vehicle fuels based on multidimensional feature coupling. Background Technology

[0002] Deep decarbonization in the transportation sector, particularly in the commercial vehicle industry, has become an inevitable trend, with the application of low-carbon blended fuels such as methanol, biodiesel, and e-Fuel becoming increasingly widespread. However, in the process of this industry's development, how to scientifically and accurately determine the actual proportion of low-carbon components (such as carbon from biomass sources or carbon capture sources) in fuels has become a key bottleneck restricting the industry's development.

[0003] In existing technologies, the certification of low-carbon fuels mainly relies on supply chain auditing methods such as the "quality balance method," which is an indirect certification and cannot physically verify the fuel composition actually used in end vehicles. Although based on radioactive carbon isotopes ( 14 C) Standard detection methods can distinguish between biomass carbon and fossil carbon, but are currently mainly applied to the analysis of static liquid fuel samples. For commercial vehicles, especially heavy-duty commercial vehicles with dual-fuel injection systems, the fuel substitution rate of the engine changes dynamically under different speeds, loads, and transient operating conditions. Traditional static sampling and testing can only reflect the fuel ratio in the fuel tank and cannot truly reflect the proportion of low-carbon fuel actually participating in combustion under complex dynamic operating conditions. Existing emission assessment methods suffer from a disconnect between "actual measurement" and "simulation." Relying solely on bench testing is costly and time-consuming, and it is difficult to cover all operating conditions throughout the vehicle's entire life cycle; while relying solely on numerical simulation is low-cost, its predictive accuracy is often insufficient due to the lack of empirical data calibration under real combustion environments. Therefore, there is an urgent need for a precise monitoring method for low-carbon components that can organically combine microscopic combustion chemical reaction kinetics simulation with macroscopic emission physical measurement and can adapt to the dynamic operating characteristics of commercial vehicles.

[0004] Furthermore, in the field of dynamic monitoring of low-carbon components in automotive fuels, existing solutions for monitoring the operating conditions of on-board diagnostic systems and bus interface data typically rely on these operating conditions data for judgment. This is combined with exhaust gas sampling device triggering and sample pre-operation to obtain a set of physical properties for the sample. Isotope mass spectrometry analysis is then performed on this set of physical properties to form detection correlations. However, this approach suffers from limitations such as inconsistencies in the standardized definitions between the combustion feature fingerprint database and the sample physical property set, a lack of unified standardization constraints for multi-dimensional observation vector correlations, and discontinuous connections between operating condition indexing, theoretical value retrieval, and subsequent operations. Existing methods often rely on single detection sessions or segmented operation processes to associate operating condition tags, sampling start timestamps, and sampling end timestamps. Under the condition of parallel access to on-board diagnostic system operating condition data and bus interface operating condition data, issues such as the accumulation of unusable data due to missing version consistency and integrity checks, and the difficulty in forming theoretical-to-experimental comparisons due to failed candidate entry primary key set associations, make it difficult to achieve stable closed-loop configuration generation. For the joint operation of multidimensional observation vectors and combustion feature fingerprint databases, existing technologies generally lack a continuous constraint and recording mechanism from the result of the operating condition index positioning to the comparison between theory and actual measurement. They also lack the rules for the derivation of mass weight labels and the connection of session labels on the operation of deviation vector calculation and online calibration parameter set generation. It is difficult to form a consistent process of acquisition, synchronization, matching, retrieval, calculation and writing back in the operation scenario of dynamic monitoring of low carbon components of vehicle fuels. This leads to the problem of chain breakage and discontinuous recording in the closed-loop configuration generation link of low carbon component blending ratio inversion, fuel life cycle carbon emission factor database call and fuel consumption rate integral calculation operation. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method for dynamic monitoring of low-carbon components in vehicle fuels based on multidimensional feature coupling, comprising: Obtain the basic specifications and pre-set combustion model parameters of the target commercial vehicle engine, perform coupled combustion dynamics and fluid dynamics simulation based on the engine parameters, and construct a combustion feature fingerprint database. Based on the combustion feature fingerprint database, timestamp synchronization and key feature operating point set matching operations are performed, and exhaust gas sampling device triggering and sample pre-operation are executed to generate sample physical property set; Based on the set of physical properties of the sample, isotope mass spectrometry analysis, detection standardization and multidimensional feature coupling operation are performed to obtain a multidimensional observation vector; Based on the multidimensional observation vector, the system performs deviation vector calculation and online calibration parameter set generation operations, and performs low-carbon component blending ratio inversion, fuel life cycle carbon emission factor database retrieval, and fuel consumption rate integral calculation operations to generate a closed-loop configuration.

[0006] Furthermore, the process of implementing the standards also includes: The standardized operation includes a combination of operations to unify and constrain naming, units, types, value ranges, missing flags, and version flags. It also performs enumeration validity calibration of engine operating condition parameter ranges and sensor data calibration, endpoint coverage calibration and step size consistency calibration of mixing ratio gradient configuration, and conflict item determination and deduplication operations of key feature operating condition point set configuration.

[0007] Furthermore, the process of orchestrating coupled simulation tasks also includes: The coupled simulation task orchestration operation includes extracting the set of operating condition indexes and the set of mixing ratio indexes from the normalized configuration to generate a set of task coordinates, and generating a simulation task identifier, input boundary conditions, initial conditions, numerical grid version and solver version for each coordinate. At the same time, consistency calibration is performed between the chemical reaction kinetic model configuration and the three-dimensional computational fluid dynamics model configuration. The consistency calibration includes matching calibration between the fuel supply system type and fuel component entries, matching calibration between the post-operation configuration and product component entries, and matching calibration between geometric boundary entries and numerical grid version.

[0008] Furthermore, constructing a combustion feature fingerprint database also includes: The combustion feature fingerprint database contains a set of conventional gaseous pollutant concentrations and a set of theoretical carbon isotope distribution features, and is organized according to the target commercial vehicle engine identifier, operating condition index, blending ratio index, and model version label.

[0009] Furthermore, the process of synchronizing timestamps and matching key feature working point sets also includes: The timestamp synchronization operation includes performing integrity screening, clock offset estimation, and nearest neighbor pairing or window interpolation pairing on the vehicle diagnostic system operating condition data and bus interface operating condition data. It also includes performing key feature operating condition point set matching operation, which includes extracting real-time speed, load torque, intake air temperature, fuel injection pulse width, vehicle speed and gear from the operating condition data stream and performing sliding window aggregation to obtain window statistics. Then, it determines the operating condition label and trigger timestamp according to the trigger conditions in the key feature operating condition point set configuration, and maps them to the operating condition index of the combustion feature fingerprint database to obtain the fingerprint retrieval index and sampling trigger command.

[0010] Furthermore, the process of triggering the exhaust gas sampling device and pre-processing the sample also includes: The exhaust gas sampling device triggering and sample pre-operation includes sending a sampling trigger command to the exhaust gas sampling device and performing a full-process status readback, as well as performing encapsulation, sealing, purification and carbon dioxide separation preparation on the collected transient exhaust gas sample. The encapsulation includes sealing the sampling container and writing the sample number; the sealing includes writing a leak check record; the purification includes recording purification items to remove water vapor and particulate matter; and the carbon dioxide separation preparation includes introducing the sample into the separation unit inlet and recording the inlet pressure and inlet temperature.

[0011] Furthermore, the process of performing isotope mass spectrometry analysis, detection standardization, and multidimensional feature coupling operations also includes: The isotope mass spectrometry analysis operation includes performing injection, ionization, mass separation, and signal acquisition on the carbon dioxide separation products in the sample physical property set to obtain the raw isotope mass spectrometry signal, and extracting radiocarbon isotope abundance data and stable carbon isotope ratio data; the detection standardization operation includes performing unified system operations on radiocarbon isotope abundance data, stable carbon isotope ratio data, and sampling metadata, including outlier screening of signal intensity sequences, generation of quality scores for peak shape fitting residual information, generation of missing bitmaps for missing markers, and generation of calibration versions of calibration markers; the multidimensional feature coupling operation includes synchronizing the standardized detection results with the sampling metadata in the sample physical property set.

[0012] Furthermore, the process of performing deviation vector calculation and online calibration parameter set generation also includes: The deviation vector calculation and online calibration parameter set generation operation involves extracting measured and theoretical features from a multidimensional observation vector and a combustion feature fingerprint database to calculate the deviation vector. The deviation vector includes radioactive carbon isotope abundance deviation, stable carbon isotope ratio deviation, and conventional gaseous component concentration deviation. A multivariate statistical analysis model or a chemometric algorithm model is used to map the deviations to parameter update values ​​to generate an online calibration parameter set. The online calibration parameter set includes update values ​​of reaction rate constant parameter entries and boundary condition parameter entries, parameter source markers, parameter version markers, and update session markers.

[0013] Furthermore, the process of inverting the blending ratio of low-carbon components and accessing the fuel life cycle carbon emission factor database also includes: The low-carbon component blending ratio inversion operation includes, under the constraints of the online calibration parameter set, mapping the multidimensional observation vector to the blending ratio index in the combustion feature fingerprint database to output the blending ratio value, generating inversion confidence markers and inversion session markers, and performing a fuel life cycle carbon emission factor database retrieval operation. The fuel life cycle carbon emission factor database retrieval operation includes searching the fuel life cycle carbon emission factor database based on the blending ratio value and fuel type identifier to obtain carbon emission factor records.

[0014] Furthermore, the process of calculating fuel consumption rate integrals also includes: The fuel consumption rate integral calculation operation includes performing a window accumulation operation on the fuel consumption rate within the window corresponding to the sampling start time stamp to the sampling end time stamp to obtain the window fuel consumption amount, and coupling the window fuel consumption amount with carbon intensity and blending ratio to generate dynamic carbon footprint assessment data, generating a closed-loop configuration. The closed-loop configuration includes an online calibration parameter set, inversion records, dynamic carbon footprint assessment data, closed-loop update log, and anomaly records.

[0015] The key innovations of this invention include: (1) Link the multidimensional observation vector with the combustion feature fingerprint database, perform the operation condition index positioning and theoretical value retrieval, deviation vector calculation and online calibration parameter set generation operations, and generate closed-loop configuration.

[0016] (2) Based on the combustion feature fingerprint database, obtain the on-board diagnostic system operation condition data and bus interface operation condition data, perform timestamp synchronization and key feature operation point set matching, and link the exhaust gas sampling device to trigger and sample pre-operation to generate sample physical property set.

[0017] (3) Extract carbon dioxide from the physical property set of the sample and perform isotope mass spectrometry analysis and detection standardization and multidimensional feature coupling operation to obtain a multidimensional observation vector.

[0018] The following are its main beneficial effects: (1) After locating the working condition index and retrieving the theoretical value, the deviation vector calculation and online calibration parameter set generation operations are performed on the multidimensional observation vector and combustion feature fingerprint database, and the closed-loop configuration is output. This allows the closed-loop configuration to participate in the link connection of standardization and coupled simulation task arrangement, result extraction and structured storage operations in subsequent rounds, and enables the low-carbon component blending ratio inversion, fuel life cycle carbon emission factor database call and fuel consumption rate integral calculation operations to obtain the same operating input caliber under the closed-loop configuration.

[0019] (2) Compared with the segmented operation link formed solely based on the on-board diagnostic system operating condition data or the bus interface operating condition data, the exhaust gas sampling device is triggered and the sample pre-operation is completed by matching the timestamp synchronization with the key feature operating condition point set to form the sample physical property set. This ensures that the correspondence between the sample physical property set and the on-board diagnostic system operating condition data and the bus interface operating condition data remains consistent in subsequent isotope mass spectrometry analysis and detection standardization and multi-dimensional feature coupling operations.

[0020] (3) By extracting carbon dioxide from the sample physical property set, performing isotope mass spectrometry analysis and detection standardization and multidimensional feature coupling operation to form a multidimensional observation vector, the multidimensional observation vector can be directly entered into the operation link of working condition index positioning and theoretical value retrieval, deviation vector calculation and online calibration parameter set generation, and form a system that can be connected with the combustion feature fingerprint library, thereby supporting the continuous operation of the closed-loop configuration generation process. Attached Figure Description

[0021] Figure 1 A flowchart illustrating the overall process of a dynamic monitoring method for low-carbon components of vehicle fuel based on multidimensional feature coupling, provided in this application embodiment; Figure 2 A flowchart illustrating a method for dynamic monitoring of low-carbon components in vehicle fuel based on multidimensional feature coupling, provided for embodiments of this application; Figure 3 This is an architecture diagram of a method for dynamic monitoring of low-carbon components in vehicle fuel based on multidimensional feature coupling, provided in an embodiment of this application. Detailed Implementation

[0022] This invention provides a method for dynamic monitoring of low-carbon components in vehicle fuels based on multidimensional feature coupling. It aims to achieve accurate inversion of low-carbon components and dynamic carbon footprint calculation in commercial vehicle blended fuels by constructing a closed-loop technology of "virtual-real integration and model correction." The "virtual-real integration" is manifested in the interaction between a theoretical combustion feature fingerprint database constructed based on coupled simulation of chemical reaction kinetics and three-dimensional computational fluid dynamics, and the multidimensional observation vectors obtained through isotope mass spectrometry analysis triggered by real-time data from the vehicle diagnostic system and bus interface. The "model correction" is manifested in the online calibration of key parameters of the simulation model using the deviation vector between the measured multidimensional feature data and the theoretical values ​​in the fingerprint database, through multivariate statistical analysis or chemometric algorithms, generating an online calibration parameter set to improve model fidelity. The specific implementation process constitutes a closed-loop chain from fingerprint database construction (S100), dynamic sampling and detection (S200, S300) to model calibration, component inversion and carbon footprint accounting (S400). Among them, the determination of radioactive carbon isotope abundance provides an absolute physical benchmark for distinguishing between biomass source and fossil source carbon. Combined with stable carbon isotope ratios and conventional gaseous component concentrations, a multi-dimensional calibration system is formed. Finally, by inverting the real-time blending ratio and linking the fuel life cycle carbon emission factor with the real-time fuel consumption rate, a dynamic carbon footprint assessment of the entire process from "oil well to wheel" is achieved.

[0023] Figure 1This application provides an overall flowchart of a method for dynamic monitoring of low-carbon components in vehicle fuels based on multi-dimensional feature coupling. The complete systematic implementation process of this method can be summarized as a technical chain of five core functional units connected sequentially, with data interconnection and closed-loop optimization, as shown in the figure. This chain clearly embodies the logical closed loop from "theoretical simulation to build a benchmark," to "actual sampling to acquire data," then "precise detection to extract features," through "comparison and correction to achieve inversion," and finally "completing dynamic assessment of the carbon footprint throughout the entire life cycle." Each unit does not operate in isolation but is tightly coupled through strictly defined interfaces, data, and control logic. The output of the preceding unit constitutes the input of the subsequent unit, and the output of the final unit is fed back to the initial unit, forming a dynamic system that continuously self-optimizes and enhances model fidelity and monitoring accuracy. The following details the definition, function, specific implementation method, internal operation process, and connections and data flow between each unit.

[0024] Unit S1: Simulation. Definition and Function: This unit corresponds to step S100 in the patent document. Its core function is to construct a digital theoretical benchmark for the target commercial vehicle engine under specific operating conditions—namely, a "combustion feature fingerprint database." This unit employs the "virtual" part of the "virtual-real combination," namely, numerical simulation based on physicochemical principles. Its aim is to predict, through a high-fidelity calculation model, the chemical composition and isotopic characteristics of the combustion products (especially exhaust gases) of the engine under various design operating conditions, under different blending ratios of low-carbon components in different fuels, before any actual exhaust gas sampling and testing. This unit is the cornerstone of all subsequent comparison, correction, and inversion operations. The completeness, accuracy, and structure of its output fingerprint database directly determine the upper limit of the theoretical accuracy of the entire monitoring system.

[0025] Implementation and internal processes: The implementation of this unit is a highly automated, configurable, coupled simulation task orchestration and execution process.

[0026] 1. Input and Configuration Analysis: The unit's inputs are the "target commercial vehicle engine identifier" and the "closed-loop configuration." The engine identifier uniquely identifies the simulation object and includes key information such as manufacturer, model, displacement, and after-operation system. The closed-loop configuration contains all the control parameters for this simulation task, mainly including: Engine operating condition parameter range: Defines a discretized set of the target engine's operating range, such as speed (e.g., 800-2200 rpm, in 100 rpm intervals), torque (e.g., 100-1800 Nm, in 50 Nm intervals), intake air temperature, fuel injection strategy (e.g., main injection timing, pre-injection quantity), and generates a unique "operating condition index" for each operating condition combination.

[0027] Blending ratio gradient configuration: Defines the volume or mass percentage gradient of the low-carbon fuel under study (such as biodiesel, renewable synthetic fuel) in the blended fuel, for example from 0% (purified fossil diesel) to 100% (pure low-carbon fuel), with a step size of 5% or 10%, and generates a unique "blending ratio index" for each gradient.

[0028] Model configurations include "Chemical Reaction Kinetics Model Configuration" and "3D Computational Fluid Dynamics Model Configuration". The former defines detailed fuel oxidation reaction pathways, all chemical species involved (fuel molecules, intermediates, products), and reaction rate constants (Arrhenius parameters) for hundreds or thousands of elementary reactions; the latter defines an accurate geometric model of the engine combustion chamber, computational mesh, intake / exhaust boundary conditions (pressure, temperature, flow rate), turbulence model, spray fragmentation and evaporation model, etc.

[0029] Other configurations include fingerprint database storage path, simulation task scheduling strategy, and abnormal operation rules.

[0030] 2. Coupled Simulation Task Orchestration: Based on the above configuration, the system automatically plans tasks. The core operation is to combine the "operating condition index" and the "blending ratio index" using a Cartesian product to generate a large "task coordinate set". For each coordinate point in the set (e.g., operating condition index = 1024, representing a speed of 1500 rpm, torque of 800 Nm, and intake air temperature of 30 ℃; blending ratio index = 6, representing a low-carbon component blending ratio of 30%), the system creates an independent simulation subtask. During task creation, critical consistency calibrations are performed, such as: calibrating whether the fuel composition configured for this engine model is compatible with its fuel supply system type; whether the calibrated operating configuration (e.g., SCR catalyst) includes the corresponding catalytic reaction path in the reaction mechanism; and whether the calibration geometric model matches the specified computational grid version. After successful calibration, the system encapsulates the specific model parameters, boundary condition parameters, initial conditions, etc., into the input file for this subtask.

[0031] 3. Simulation Calculation Execution: The orchestrated simulation tasks are submitted to a high-performance computing cluster or local computing resources for execution. The calculation process is "coupled," typically employing a partitioned or sequential coupling strategy. For example, a 3D CFD model first calculates the transient evolution of the in-cylinder flow field, temperature field, and pressure field, and transmits information such as temperature, pressure, and component concentration in local regions to the chemical reaction kinetics solver. Based on these local conditions, the chemical reaction kinetics solver calculates detailed chemical reaction rates and component changes, and returns updated heat release rates and component concentrations to the CFD solver for subsequent flow field calculations. This bidirectional data exchange continues throughout the entire engine operating cycle until the calculation converges.

[0032] 4. Result Extraction and Fingerprint Database Construction: Each successful simulation subtask outputs a "raw result file" containing time series or cyclic average results. The result processing unit of this section extracts predefined "features" crucial for subsequent monitoring from these files. These mainly include two categories: Collection of common gaseous pollutant concentrations: such as the average mole fraction or mass concentration of nitrogen oxides, carbon monoxide, and unburned hydrocarbons.

[0033] The theoretical distribution characteristics of carbon isotopes: This is the key to this invention. The simulation model calculates the radioactive carbon isotopes of carbon dioxide in the exhaust gas by tracing the transport paths of carbon atoms from different sources in the fuel (such as "dead carbon" from fossil fuels and "modern carbon" from biomass) in the combustion reaction network. 14 C) Abundance (in pMC or The theoretical prediction of the ratio of the stable carbon isotope (13C / 12C) to the stable carbon isotope (δ13C).

[0034] All extracted features are strictly bound to the "operating condition index" and "blending ratio index" that generated them, and are also appended with a "model version tag." They are ultimately stored in a standardized database or file collection, known as a "combustion feature fingerprint database." Essentially, this is a multidimensional lookup table or high-dimensional response surface model, with primary keys of (engine identifier, operating condition index, blending ratio index, model version) and values ​​of the corresponding theoretical exhaust gas feature vectors.

[0035] Unit S2: Sampling. Definition and Function: This unit corresponds to step S200 in the patent document. Its core function is to automatically trigger and complete the collection and pre-processing of representative exhaust gas samples at appropriate times during actual vehicle operation. It is a key bridge connecting the "virtual" (simulation) and the "real" (actual measurement). Its goal is to capture transient exhaust gas generated under specific and representative engine operating conditions and materialize and stabilize it into a "sample physical property set" that can be analyzed by subsequent precision instruments. The intelligence of this unit is reflected in its ability to judge and trigger sampling in real time based on the "key characteristic operating conditions" preset by the simulation unit, ensuring that the operating conditions corresponding to the collected samples can find accurate or approximate theoretical references in the fingerprint database.

[0036] Implementation methods and internal processes: 1. Real-time Operating Condition Monitoring and Timestamp Synchronization: The unit continuously acquires real-time engine operating data from two main data streams: one is the internal parameters of the engine control unit (such as engine speed, torque, coolant temperature, fuel injection quantity, etc.) obtained through the on-board diagnostic system; the other is data from other relevant sensors or controllers (such as exhaust temperature, air-fuel ratio sensor signals, etc.) obtained through the vehicle bus interface. Since different data sources may have clock asynchrony issues, the unit first performs a "timestamp synchronization" operation. Through clock offset estimation and data interpolation / pairing algorithms, all data is unified onto a high-precision time reference axis, forming a continuous and synchronized "operating condition data stream."

[0037] 2. Key Feature Operating Condition Matching and Trigger Decision: The system analyzes the above operating condition data stream in real time and compares it with the "Key Feature Operating Condition Set Configuration" defined in the "Closed-Loop Configuration". These key points are not randomly selected, but are selected based on engineering knowledge and are sensitive to fuel characteristics or have typical representativeness, such as "High Load Climbing Condition" (high torque, medium speed), "Cold Start Condition" (low water temperature, idling speed), and "Steady-State High-Speed ​​Cruise Condition" (stable speed and torque). The matching algorithm calculates the similarity between the operating condition parameters (such as average speed, average torque, parameter fluctuation rate) within the current sliding time window and the preset key point features. When the similarity exceeds the preset threshold, the system determines that a certain key feature operating condition has been entered at the current moment, and then generates a "Sampling Trigger Command". This command includes the specific operating condition label of the trigger, the precise trigger timestamp, and the "Operating Condition Index" retrieved from the combustion feature fingerprint database that best matches the operating condition, as well as the corresponding theoretical feature set preview (for subsequent comparison).

[0038] 3. Exhaust Gas Sampling Device Triggering and Sample Acquisition: The sampling trigger command is sent in real time to the vehicle-mounted or bench-integrated "exhaust gas sampling device." This device typically includes a fast-acting valve, a volumetric sampler, a sample bag, or an adsorption tube. The system precisely controls the sampling start and end times to ensure that the collected exhaust gas is generated during the steady-state or quasi-steady-state phase of the target operating condition, avoiding interference from the transition process. Simultaneously, the system monitors the status of the sampling device (such as valve opening / closing feedback, sampling flow rate, and sample bag pressure) and records all operational metadata, forming "sampling metadata."

[0039] 4. Sample Pre-processing: The raw exhaust gas sample may contain interfering substances such as moisture and particulate matter, which can affect instrument performance and accuracy if directly subjected to isotope analysis. Therefore, pre-processing steps are usually included after sampling, such as: removing moisture through condensation or permeation drying tubes; removing soot through a particulate filter; and, if necessary, selectively separating, purifying, and transferring the carbon dioxide component in the sample to a dedicated storage container or injection bottle for subsequent mass spectrometry analysis. All pre-processing steps, parameters, and results (such as post-purification humidity and particulate matter concentration) are recorded and attached to the sample metadata.

[0040] 5. Generating a Set of Sample Physical Properties: Finally, the physical sample, after pre-processing, packaging, and affixing a unique identifier, is digitally bound to its complete "sampling metadata" (including operating condition information, trigger information, sampling process information, and pre-processing information) to jointly constitute a "set of sample physical properties." This serves as the carrier connecting the physical sample and the digital world and is then transmitted to the next detection unit.

[0041] Unit S3: Detection. Definition and Function: This unit corresponds to step S300 in the patent document. Its core function is to perform precise, multi-dimensional physicochemical analysis on the "sample physical property set" collected by unit S2, extract core feature quantities that can be compared with simulated theoretical values, and organize them into a standardized "multi-dimensional observation vector." This unit is the source of measured data for realizing "multi-dimensional feature coupling," and its detection accuracy directly determines the quality of the measured data of the entire inversion system. In particular, for radioactive carbon isotopes ( 14 The high-precision measurement of C) provides a benchmark with absolute physical significance for distinguishing fossil carbon from biomass carbon in this method.

[0042] Implementation methods and internal processes: 1. Isotope mass spectrometry analysis: This is the core detection step in this unit. Accelerator mass spectrometry or high-precision gas isotope ratio mass spectrometry is used to analyze carbon dioxide in the pre-processed sample.

[0043] Radiocarbon isotopes ( 14 C) Analysis: The carbon in the sample is converted into graphite or directly introduced into the accelerator mass spectrometer as CO2. The instrument measures... 14 C and 12 C (or 13 C) The ratio of ion current intensity is used to accurately determine the concentration of ions in the sample. 14 The atomic abundance of C. Due to the presence of fossil fuels 14 Carbon has decayed almost entirely (“dead carbon”), while in modern biomass... 14The carbon content is basically in equilibrium with the atmosphere, therefore this measurement (usually expressed as "percent modern carbon," pMC) is the "gold standard" and "true anchor" for quantifying the proportion of biomass-derived carbon in fuels. The measurement process requires background correction, isotope fractionation correction, etc.

[0044] Stable carbon isotopes ( 13 C) Analysis: Isotope ratio mass spectrometry is typically used to measure the concentration of [unspecified substances] in the sample. 13 CO2 and 12 The ratio of CO2 ion current intensity is used to obtain δ 13 C-value. Biofuels and fossil fuels from different feedstock pathways have specific δ values. 13 The C-value range can be used as "fingerprint" information to aid in identification and calibration.

[0045] The analysis process generates raw mass spectrometry signals, intensity data, peak shape information, etc., from which the system extracts "radioactive carbon isotope abundance data" and "stable carbon isotope ratio data".

[0046] 2. Routine gaseous component concentration detection: Analyzers such as Fourier transform infrared spectroscopy, non-dispersive infrared spectroscopy, and chemiluminescence are typically used to analyze the same or parallel samples, measuring the concentrations of common pollutants such as nitrogen oxides, carbon monoxide, and total hydrocarbons. These data reflect the overall combustion state, including combustion efficiency and temperature field.

[0047] 3. Standardized Testing Procedures: Raw test data from different instruments may vary in format, units, and precision. This step involves standardized procedures: Data cleaning: Smoothing, denoising, and screening for outliers (such as spikes caused by cosmic rays) in mass spectrometry signals.

[0048] Calibration and conversion: Convert all test data to unified international standard units and apply instrument calibration curves and blank correction.

[0049] Quality assessment: Generate a "quality score" for each detection result, such as goodness of fit based on mass spectrum peak shape, signal stability, signal-to-noise ratio, etc. Simultaneously, generate explicit "missing data markers" and "calibration version markers" for any missing or invalid data points.

[0050] 4. Multidimensional Feature Coupling: Standardized isotope data and conventional gas concentration data are correlated and encapsulated with the "sampling metadata" (especially precise operating condition information) transmitted from the S2 unit and bound to the sample. All information is integrated into a unified "multidimensional observation vector." This observation vector represents a complete digital profile of the "measured state" and will be sent to the next unit for comparison with the "theoretical state."

[0051] Unit S4: Correction / Inversion. Definition and Function: This unit corresponds to step S400 in the patent document and is the core and "brain" of the entire method. It performs the dual functions of "model correction" and "component inversion." Its function is to use the difference between the "multidimensional observation vector" (measured value) generated by unit S3 and the "combustion feature fingerprint database" (theoretical value) constructed by unit S1 to correct the parameters of the simulation model (correction), and based on the corrected high-fidelity model, solve for the most likely blending ratio of low-carbon fuel components that leads to the current measured exhaust gas characteristics (inversion). This is the key to achieving a closed loop of "combining virtual and real, correcting virtual with real, and measuring real with virtual."

[0052] Implementation methods and internal processes: 1. Operating Condition Index Location and Theoretical Value Retrieval: First, based on the precise operating condition information (such as speed and torque) contained in the "multidimensional observation vector," a fast search or interpolation operation is performed in the combustion feature fingerprint database to locate the "operating condition index" that best matches it. Then, under this index, the set of theoretical feature vectors {V_theory(r)} corresponding to all different "blending ratio indices" is retrieved, where r represents different blending ratios.

[0053] 2. Deviation Vector Calculation: The measured vector V_obs is compared element-by-element with each theoretical vector V_theory(r) in the fingerprint database to calculate the deviation. The deviation includes not only the scalar difference but, more importantly, constitutes a "deviation vector." This vector contains information from multiple dimensions, such as the deviation in radioactive carbon isotope abundance, the deviation in stable carbon isotope ratios, and the deviations in the concentrations of various conventional gases. Deviations in different dimensions may be sensitive to different model parameters.

[0054] 3. Online Calibration Parameter Set Generation (Correction): The system utilizes multivariate statistical analysis models or chemometric algorithm models (e.g., partial least squares regression, artificial neural networks, support vector machines, etc.) to establish a mapping relationship between the "deviation vector" ΔV and the "key parameter correction amount" ΔP of the simulation model. These key parameters may include pre-exponential factors or activation energies of reactions sensitive to isotope fractionation effects in chemical reaction kinetic models, or turbulence model constants affecting mixing rates in CFD models. The algorithm finds a set of parameter correction amounts ΔP by learning from historical data or through physics-based sensitivity analysis, ensuring that the theoretical value V_theory'(r) calculated using the corrected parameters approximates the current V_obs as closely as possible. This set of ΔP is encapsulated as an "online calibration parameter set" and tagged with version and session information. This process, known as "model correction," enables the simulation model to adapt to individual differences, aging conditions, or unmodeled physical processes of specific engines, continuously approximating the real physical world.

[0055] 4. Low-carbon component blending ratio inversion: After obtaining the calibrated model parameters (or directly using the bias model), an inversion solution is performed. The core is to find a blending ratio r* such that the calibrated theoretical eigenvector V_theory'(r*) is closest to the measured vector V_obs under a certain metric (such as weighted Euclidean distance or Mahalanobis distance considering measurement errors in each dimension). This can be transformed into an optimization problem. The system will output this optimal blending ratio estimate r*, along with an "inversion confidence score" (e.g., based on residual size, discrimination against neighboring proportional solutions, etc.). The r* obtained at this point is the result after model correction, theoretically more accurate than the result obtained using the original uncorrected model.

[0056] 5. Data encapsulation and output preparation: This unit encapsulates key results such as the generated "online calibration parameter set" and "the inverted mixing ratio value r* and its confidence level" in preparation for passing them to the next unit, and also prepares for closed-loop feedback.

[0057] Unit S5: Life Cycle Assessment. Definition and Function: This unit is an extension of step S400 in the patent document and the final value realization stage. Its function is to calculate the greenhouse gas emission intensity of the vehicle during operation based on the real-time, dynamic blending ratio of low-carbon components obtained from the inversion of unit S4, combined with the vehicle's real-time energy consumption data, thus realizing dynamic accounting of the carbon footprint from "fuel tank to wheel" and even "oil well to wheel". It links the micro-level fuel component monitoring results with the macro-level environmental impact assessment, providing direct quantitative data support for carbon trading, environmental regulation, and corporate carbon accounting.

[0058] Implementation methods and internal processes: 1. Fuel Consumption Data Acquisition and Integration: Real-time reading of instantaneous fuel consumption rate data (e.g., grams per second) from the vehicle bus or on-board diagnostic system. For the operating condition represented by the S2 unit sampling (a time window from before sampling trigger to after sampling end), the fuel consumption rate within this time window is integrated to calculate the "total fuel consumed during this characteristic operating condition".

[0059] 2. Carbon Emission Factor Database Access: The system embeds or connects to a "Fuel Life Cycle Carbon Emission Factor Database." This database contains carbon emission intensity data for various fuels (such as fossil diesel, rapeseed oil biodiesel, waste cooking oil biodiesel, and electro-synthetic diesel) throughout their complete life cycle, typically expressed in grams of carbon dioxide equivalent per megajoule or grams of carbon dioxide equivalent per liter. The database must cover emissions from the entire process, including raw material planting / extraction, processing, transportation, and fuel production.

[0060] 3. Dynamic Carbon Footprint Calculation: For the inversely derived blending ratio r* (e.g., 30% biodiesel by volume), the system retrieves the corresponding fuel's carbon emission factor from the database. Then, based on the blending ratio of the mixed fuels, the system calculates the comprehensive carbon emission factor for that blend. Finally, by multiplying the total amount of fuel consumed within the window period by this comprehensive carbon emission factor, the system obtains the "carbon footprint" of the vehicle directly generated by fuel combustion and its upstream processes during that operating period.

[0061] 4. Data Aggregation and Output: The unit ultimately generates "Dynamic Carbon Footprint Assessment Data," which may include: time period, mileage, total fuel consumption, blending ratio, comprehensive carbon emission factor, total carbon footprint, and carbon footprint per unit mileage. This data can be displayed in real time or stored and used to generate reports periodically.

[0062] Inter-unit connections and closed-loop feedback: The five units are linearly connected through strictly defined data and control flow, forming an open-loop data operation pipeline: S1->S2->S3->S4->S5. However, one of the core innovations of this invention lies in establishing a powerful "closed-loop feedback" channel between S4 and S1.

[0063] Feedback path: The "online calibration parameter set" generated by the S4 unit is encapsulated in the "closed-loop configuration" and fed back to the input of the S1 unit.

[0064] Closed-loop effect: In the next cycle, when the S1 unit performs a simulation task again (perhaps for a new engine, or periodically updating the existing fingerprint database), it will use this revised and more accurate "online calibration parameter set" to update the parameters of its chemical reaction kinetics or CFD model. This means that the simulation model is no longer static, but can continuously learn and self-correct using actual measurement data, allowing its theoretical predictive ability to continuously improve as the system operates. This "model evolution" capability ensures that even if engine performance degrades with use, or fuel properties change slightly, the accuracy of the entire monitoring system remains at a high level.

[0065] Permissible Process: The entire process allows for iterative execution. It enables continuous sampling, detection, correction, inversion, and evaluation at different operational stages of the same vehicle, thereby obtaining the usage of low-carbon components and carbon footprint trends over time. Furthermore, calibration parameters learned from one vehicle, after verification, can be applied to other vehicles of the same model, achieving knowledge transfer and accelerating the model calibration process for new vehicles.

[0066] In summary, these five units constitute a complete, closed-loop, and self-evolving technical system, encompassing theoretical modeling, intelligent sampling, precise detection, intelligent inversion correction, and final environmental assessment. S1 provides the theoretical benchmark, S2 and S3 are responsible for acquiring high-quality measured data, S4 is the core algorithm and optimization hub, S5 realizes the final application value, and the feedback from S4 to S1 endows the system with the intelligence of continuous learning and adaptation. Together, they achieve high-precision, dynamic, and online monitoring and carbon accounting of low-carbon components in vehicle fuels.

[0067] In one embodiment, reference is made to Figure 2 This is a flowchart illustrating a method for dynamic monitoring of low-carbon components in vehicle fuel based on multidimensional feature coupling, provided by an embodiment of the present invention. The process may include at least steps S100-S400: S100: Obtain the basic specifications and pre-set combustion model parameters of the target commercial vehicle engine, perform coupled combustion dynamics and fluid dynamics simulation based on the engine parameters, and construct a combustion feature fingerprint library; S200: Based on the combustion feature fingerprint database, perform timestamp synchronization and key feature operating point set matching operations, and execute exhaust gas sampling device triggering and sample pre-operation to generate sample physical property set; S300: Based on the set of physical properties of the sample, perform isotope mass spectrometry analysis, detection standardization and multidimensional feature coupling operation to obtain a multidimensional observation vector; S400, based on multi-dimensional observation vectors, performs deviation vector calculation and online calibration parameter set generation operations, and performs low-carbon component blending ratio inversion, fuel life cycle carbon emission factor database retrieval, and fuel consumption rate integral calculation operations to generate closed-loop configuration.

[0068] In S100, the basic specifications and preset combustion model parameters of the target commercial vehicle engine are obtained, and combustion dynamics and fluid dynamics coupled simulation is performed based on the engine parameters to build a combustion feature fingerprint library. Specifically, the target commercial vehicle engine identifier is a set of combined identifiers used to uniquely locate the engine object. It includes the manufacturer identifier, engine model, displacement, emission stage, post-operation configuration, fuel supply system type and calibration version, and can be further extended to include the vehicle platform identifier and engine controller identifier. The closed-loop configuration is a set of configurations that runs through the S100 to S400 operation process. Its initial inputs include the engine operating condition parameter range, blending ratio gradient configuration, key characteristic operating point set configuration, chemical reaction kinetic model configuration, three-dimensional computational fluid dynamics model configuration, online calibration parameter set, fingerprint database storage mode configuration and anomaly record configuration. During closed-loop operation, it is written back and iterated through the updated online calibration parameter set generated in step S400.

[0069] Specifically, the input sources for this step are the target commercial vehicle engine identifier and closed-loop configuration. The target commercial vehicle engine identifier is defined as a set of combined identifiers used to uniquely locate the engine object, including the manufacturer identifier, engine model, displacement, emission stage, post-operation configuration, fuel supply system type, and calibration version, and may further include the vehicle platform identifier and engine controller identifier. The closed-loop configuration is defined as a set of operational configurations spanning S100 to S400, including the engine operating parameter range, blending ratio gradient configuration, key characteristic operating point set configuration, chemical reaction kinetic model configuration, three-dimensional computational fluid dynamics model configuration, online calibration parameter set, fingerprint database storage mode configuration, and anomaly record configuration. At the start of operation, the target commercial vehicle engine identifier is connected to the object registration unit, generating an engine object index and writing it to the operation session record; the closed-loop configuration is connected to the configuration parsing unit, entering the standardized operation chain. The standard is defined as a combination of operations that unify and constrain naming, units, types, value ranges, missing value markers, and version markers. During the operation, enumeration validity calibration and sensor data calibration are performed on the engine operating condition parameter range. The engine operating condition parameter range is defined as a discretized set of the target engine's operable operating conditions, including at least real-time speed, load torque, intake air temperature, and injection strategy, and an operating condition index is generated for each condition. Endpoint coverage calibration and step size consistency calibration are performed on the blending ratio gradient configuration. The blending ratio gradient configuration is defined as a discretized sequence of low-carbon component blending ratios, and a blending ratio index is generated for each blending ratio level. Conflict item determination and deduplication operations are performed on the key feature operating condition point set configuration. The key feature operating condition point set configuration is defined as a set of operating condition labels used for subsequent sampling triggers, including high-load ramp conditions, cold start conditions, and steady-state high-speed cruise conditions, and an operating condition label is generated for each type of label. If missing values, inconsistent units, or out-of-bounds values ​​are found, an anomaly record is generated and written to the recording medium pointed to by the anomaly record configuration. Simultaneously, an input snapshot is retained, and after revision, a standardized configuration is obtained, and the process proceeds to the next stage.

[0070] Furthermore, coupled simulation task orchestration is performed based on the standardized configuration. Coupled simulation task orchestration is defined as a combined operation that binds, assembles, schedules, and traceably places the chemical reaction kinetics model configuration and the three-dimensional computational fluid dynamics model configuration under a unified operating condition index and blending ratio index. In specific implementation, the operating condition index set and the blending ratio index set are extracted from the standardized configuration, combined and generated to form a task coordinate set, and a simulation task identifier, input boundary conditions, initial conditions, numerical mesh version, and solver version are generated for each coordinate. The chemical reaction kinetics model configuration is defined as a set of reaction mechanism entries and reaction rate constant parameter entries. The reaction mechanism entries include fuel component entries, reaction path entries, and product component entries, and the reaction rate constant parameter entries include parameter values ​​and parameter version tags. The three-dimensional computational fluid dynamics model configuration is defined as a set of geometric boundary entries and boundary condition parameter entries. The geometric boundary entries include combustion chamber boundary entries, intake boundary entries, and exhaust boundary entries, and the boundary condition parameter entries include pressure, temperature, and flow rate and their version tags. During the task assembly phase, two types of model configurations are written into the simulation task description, and consistency calibration is performed. Consistency calibration includes matching calibration between fuel supply system type and fuel component entries, matching calibration between post-operation configuration and product component entries, and matching calibration between geometric boundary entries and numerical grid versions. Tasks that pass calibration are added to the task scheduling queue and a coupled simulation task set is generated. Tasks that fail calibration generate task blocking flags and are written to the exception log. During operation, the compute nodes pull the coupled simulation task set from the task scheduling queue and perform numerical calculations, generating raw result files. When computation is interrupted, convergence anomalies occur, or file calibration fails, the reason for failure, the failure stage, and the number of retries are recorded. Retrying or skipping is performed according to the exception log configuration, while the input snapshot and task identifier of the failed task are retained for subsequent linkage.

[0071] In the results processing phase, the original results file undergoes results extraction and structured storage operations. Results extraction is defined as extracting a standard set from the simulation results and synchronizing it. Structured storage is defined as writing fingerprint entries into the database or file storage medium pointed to by the fingerprint database storage mode configuration according to a fixed primary key and version system. Specifically, the set of common gaseous pollutant concentrations and the set of theoretical carbon isotope distribution characteristics are extracted from the original results file. The set of common gaseous pollutant concentrations is defined as a combination of nitrogen oxide concentration, carbon monoxide concentration, and unburned hydrocarbon concentration, and is bound to the operating condition index and blending ratio index. The set of theoretical carbon isotope distribution characteristics is defined as a combination of isotope distribution characterization and its corresponding operating condition index, blending ratio index, and model version tag. After extraction, a type consistency operation is performed, a dimension verification operation is performed, and missing items are written with missing tag rules, generating fingerprint entries. Subsequently, the fingerprint entries are written into the storage medium according to the primary key composed of the target commercial vehicle engine identifier, operating condition index, blending ratio index, and model version tag, constructing a combustion feature fingerprint database. At the end of this step, the output is named "Combustion Feature Fingerprint Library". Its subsequent input position is the combustion feature fingerprint library of S200. It is used by S200 to access the on-board diagnostic system operating condition data and bus interface operating condition data based on the combustion feature fingerprint library, and to perform key feature operating condition point set matching and exhaust gas sampling device triggering. At the same time, after the combustion feature fingerprint library performs operating condition indexing and theoretical value retrieval with the multi-dimensional observation vector in S400 and generates an online calibration parameter set, the online calibration parameter set is written back to the closed-loop configuration and used as one of the input sources of S100 in the next cycle, realizing cross-step connection with S200, S300 and S400.

[0072] In summary, the technical effects of this step are as follows: by extracting the coupled simulation results driven by standardized configuration into a unified system and storing them in a structured manner as a combustion feature fingerprint library, a benchmark input that can be directly called upon in subsequent sampling detection and online calibration links is formed, and the closed-loop configuration is updated in the cyclic iteration.

[0073] S200: Based on the combustion feature fingerprint database, perform timestamp synchronization and key feature operating point set matching operations, and execute exhaust gas sampling device triggering and sample pre-operation to generate sample physical property set; Specifically, the input sources for this step include at least the combustion feature fingerprint database output from the previous main step S100 and the real-time data channel accessed from the vehicle side. The combustion feature fingerprint database is defined as a set of fingerprint entries organized according to the target commercial vehicle engine identifier, operating condition index, blending ratio index, and model version tag. The fingerprint entries include a set of conventional gaseous pollutant concentrations, a set of carbon isotope theoretical distribution characteristics, and operating condition labels, etc. The real-time data channel accessed from the vehicle side is used to provide operating condition data of the on-board diagnostic system and the bus interface. The operating condition data of the on-board diagnostic system is defined as a set of engine operating states from the on-board diagnostic system, including at least real-time speed, load, coolant temperature, intake air temperature, throttle opening, injection pulse width, and fault codes. The operating condition data of the bus interface is defined as a set of control and sensing data from the vehicle control network, including at least torque request, actual load torque, accelerator pedal opening, gear, vehicle speed, subsequent operation status, and message count. At the start of operation, the data access unit establishes subscription relationships with the on-board diagnostic system operating condition data and the bus interface operating condition data, sets the sampling period and cache window, and generates an access timestamp and data source mark in each data record. When any data source experiences a short-term interruption, the completion strategy is executed according to the cache window. The completion strategy includes maintaining the previous valid value, inserting a missing mark, and recording the interruption duration. The abnormal situation is written into the abnormal record for subsequent tracing.

[0074] Further, after data access is completed, the timestamp synchronization operation link is entered. Timestamp synchronization is defined as the operation process of mapping records from different data sources under different transmission delays and different sampling periods to a unified time axis and forming a computable operating condition data stream. In specific implementation, firstly, integrity screening is performed on the operating condition data of the vehicle diagnostic system and the operating condition data of the bus interface. The screening includes missing data determination, value out-of-bounds determination, jump anomaly determination, and message count wrap-around determination. Records that pass the screening enter the synchronization buffer and are sorted according to the access timestamp. Then, clock offset estimation is performed on the access timestamps of the two types of data sources. The clock offset estimation generates an offset based on the combination relationship between message count and sampling period, and the offset is written into the synchronization parameter record. On this basis, the operating condition data of the vehicle diagnostic system and the operating condition data of the bus interface are paired according to a unified time axis. The pairing rules include nearest neighbor pairing and window interpolation pairing, and records that cannot be paired are marked as unsynchronized. After timestamp synchronization is completed, an operating condition data stream is obtained. This data stream is defined as a fused sequence along a unified time axis, containing at least real-time engine speed, load torque, intake air temperature, injection pulse width, gear position, vehicle speed, subsequent operating status, and synchronization quality markers. It may further include fault codes and interruption durations. This operating condition data stream is subsequently used for matching key feature operating condition point sets. Simultaneously, its operating condition labels maintain the same semantic meaning as the operating condition labels in the combustion feature fingerprint database. These operating condition labels are defined as categorized annotations of operating states, including high-load climbing conditions, cold start conditions, and steady-state high-speed cruising conditions.

[0075] In the key feature operating condition set matching stage, the matching unit extracts operating condition features for judgment from the operating condition data stream and performs rule-based judgment and index mapping. Specifically, firstly, sliding window aggregation is performed on real-time speed, load torque, intake air temperature, injection pulse width, vehicle speed, and gear position to obtain window statistics. The window statistics at least include the mean, fluctuation amplitude, and duration. Then, the window statistics are judged according to the trigger conditions in the key feature operating condition set configuration. The trigger conditions are defined as a set of conditions used to identify high-load climbing conditions, cold start conditions, and steady-state high-speed cruising conditions, including threshold conditions, duration conditions, and state combination conditions. When the judgment is successful, an operating condition label and a trigger timestamp are generated. Further, the operating condition label and trigger timestamp are mapped to the operating condition index of the combustion feature fingerprint database to obtain a fingerprint retrieval index. The fingerprint retrieval index is defined as an index set used to locate similar operating condition entries in the combustion feature fingerprint database, and a sampling trigger command is generated at the same time. The sampling trigger instruction is defined as a set of controls used to drive the exhaust gas sampling device to perform transient sampling. It includes at least a trigger timestamp, operating condition label, sampling duration, sampling channel identifier, sample number pre-assignment, and safety interlock mark. The safety interlock mark is defined as a constraint used to restrict sampling under specific fault codes or subsequent operation states. When a fault code triggers the interlock condition or the subsequent operation state is in a prohibited sampling state, the sampling trigger instruction is marked as blocked and written into the exception record.

[0076] During the exhaust gas sampling device triggering and sample pre-operation phase, the control unit sends the sampling trigger command to the exhaust gas sampling device and performs a full-process status readback. The exhaust gas sampling device is defined as a hardware unit with a sampling channel, sampling valve group, flow regulation component, and sampling container interface. Its sampling channel identifier is used to select the sampling path, the sampling duration is used to control the sampling window, and the sample number is pre-assigned to bind the sample identifier. After triggering, the exhaust gas sampling device returns the sampling start timestamp, sampling end timestamp, actual flow rate, and sampling status. When the sampling status indicates sampling failure or the actual flow rate exceeds the allowable range, a sampling failure reason is generated and written to the exception record. Simultaneously, the sampling is marked as invalid, and the operating condition data stream continues to be monitored. Upon successful sampling, the sample pre-operation operation link is entered. Sample pre-operation is defined as the process of encapsulating, sealing, purifying, and preparing the collected transient exhaust gas sample for carbon dioxide separation. Encapsulation includes sealing the sampling container and writing the sample number; sealing includes writing a leak check record; purification includes recording purification entries for removing water vapor and particulate matter; and carbon dioxide separation preparation includes introducing the sample into the separation unit inlet and recording the inlet pressure and inlet temperature. After sample pre-processing, a sample physical property set is generated. This set is defined as a chemical product binding the exhaust gas sample with sampling metadata. The sampling metadata includes at least the sample number, sampling start timestamp, sampling end timestamp, operating condition label, sampling channel identifier, synchronization quality marker, and safety interlock marker. At the end of this step, an output named "Sample Physical Property Set" is output. Its subsequent input location is the Sample Physical Property Set of S300. S300 extracts carbon dioxide from this set and performs isotope mass spectrometry analysis, detection standardization, and multi-dimensional feature coupling operations. Simultaneously, the operating condition label and operating condition index, along with the combustion feature fingerprint database in S400, participate in operating condition index positioning and theoretical value retrieval. This establishes a cross-step connection and cyclic write-back relationship between the combustion feature fingerprint database of S100, the sample physical property set of S200, the multi-dimensional observation vector of S300, and the closed-loop configuration of S400.

[0077] In summary, the technical effects of this step are as follows: By synchronizing the operating condition data of the on-board diagnostic system with the operating condition data of the bus interface on a unified time axis and matching the set of key characteristic operating condition points, the exhaust gas sampling device is driven to generate a set of sample physical properties bound to the operating condition information, providing traceable input for subsequent multi-dimensional detection and online calibration links.

[0078] S300: Based on the set of physical properties of the sample, perform isotope mass spectrometry analysis, detection standardization and multidimensional feature coupling operation to obtain a multidimensional observation vector; Specifically, the input source for this step is the set of sample physical attributes output from the previous main step S200. The set of sample physical attributes is defined as the chemical product formed by binding the exhaust gas sample with sampling metadata. The exhaust gas sample represents the transient exhaust gas sample within the sampling container. The sampling metadata includes at least the sample number, sampling start timestamp, sampling end timestamp, operating condition label, sampling channel identifier, synchronization quality flag, and safety interlock flag. At the start of this step, the sample receiving unit reads the sample number and queries the corresponding sampling metadata, writes the sample number into the detection session record, performs consistency calibration on the sampling start timestamp and sampling end timestamp, performs legality calibration on the operating condition label and sampling channel identifier, and performs threshold determination on the synchronization quality flag. When there is a missing flag, a reversed timestamp, or the safety interlock flag is in a blocked state, a detection blocking flag is generated and written into the anomaly record. Simultaneously, the sample number is retained in the pending verification queue and will not subsequently enter the isotope mass spectrometry analysis link.

[0079] Further, in the carbon dioxide extraction stage, the separation unit introduces the exhaust gas sample into the pre-operation channel and performs water vapor removal, particulate matter filtration, and background gas purging. Water vapor removal refers to reducing the interference of moisture in the sample on subsequent detection through adsorption material channels; particulate matter filtration refers to retaining solid particles through a microporous membrane; and background gas purging refers to cleaning the residual gas in the separation channel and recording a blank mark. After completing the pre-operation, the separation unit performs carbon dioxide separation on the sample gas. This carbon dioxide separation operation includes a selective adsorption-release process or a membrane separation process, generating a carbon dioxide separation product at the outlet. Simultaneously, the separation inlet pressure, separation inlet temperature, separation duration, and channel status are recorded. When the channel status indicates separation failure, leakage, or pressure exceeding the limit, a separation failure reason is generated and written into the anomaly record. The carbon dioxide separation product is also marked as invalid and will not be included in subsequent isotope mass spectrometry analysis.

[0080] During the isotope mass spectrometry analysis phase, the analysis unit receives valid carbon dioxide separation products and performs sample introduction, ionization, mass separation, and signal acquisition operations to obtain the raw isotope mass spectrometry signal. The isotope mass spectrometry analysis is defined as the process of measuring the carbon isotope-related signals in the carbon dioxide separation products and generating calculable detection data. The raw isotope mass spectrometry signal includes at least a signal intensity sequence, acquisition time sequence, instrument status, and calibration markers. At the start of each detection session, the analysis unit reads the calibration markers, loads the standard sample record, and performs baseline drift correction and peak shape fitting operations. Simultaneously, it performs a stability check on the instrument status. If the stability check fails, an instrument anomaly marker is generated and written to the anomaly record, and the raw isotope mass spectrometry signal for that session is marked as failing. For sessions that pass the stability assessment, the analysis unit extracts radiocarbon isotope abundance data and stable carbon isotope ratio data from the raw isotope mass spectrometry signal. Radiocarbon isotope abundance data is defined as a detection that characterizes the proportion of radiocarbon isotope signals, and stable carbon isotope ratio data is defined as a detection that characterizes the ratio relationship of stable carbon isotope signals. Both are bound to the sample number and written to the detection result cache.

[0081] Entering the standardization phase, the standardization unit performs unified system operations on the radiocarbon isotope abundance data, stable carbon isotope ratio data, and the aforementioned sampling metadata. The standardization is defined as a unified process for naming the detection results, specifying the unit of measurement, value range, missing data markers, and version markers. Specifically, the standardization unit uses the sample number as the primary key, writes the radiocarbon isotope abundance data and stable carbon isotope ratio data into the standardized detection set, and maps the sampling start timestamp, sampling end timestamp, operating condition label, sampling channel identifier, and synchronization quality marker to the same record. Simultaneously, it performs anomaly screening on the signal intensity sequence, generates a quality score from the peak shape fitting residual information, generates a missing bitmap from the missing data markers, and generates a calibration version from the calibration markers. When the quality score is below a threshold or the missing bitmap covers a key detection, a detection failure marker is generated and written to the anomaly record, which is then retained in the pending review queue. When the detection passes, a standardized detection result is generated and enters the multi-dimensional feature coupling operation.

[0082] In the multidimensional feature coupling operation phase, the coupling unit synchronizes the standardized detection results with the sampling metadata in the sample physical property set and performs cross-consistency checks and combination encapsulation operations. The multidimensional feature coupling operation is defined as combining radioactive carbon isotope abundance, stable carbon isotope ratio, and conventional gaseous component concentration within the same session to form an observation vector for subsequent operating condition indexing and theoretical value retrieval. Specifically, the coupling unit reads the operating condition label, sampling channel identifier, and synchronization quality mark from the sample physical property set and associates them with the standardized detection results according to sample number. When sample number mismatch, timestamp window non-overlap, or synchronization quality mark is in a low-quality state, an association failure reason is generated and written into an exception record, and no observation vector is output. For records with successful association, the coupling unit writes the radioactive carbon isotope abundance, stable carbon isotope ratio, and conventional gaseous component concentration into the same data object and appends the operating condition label, sampling start timestamp, sampling end timestamp, model version placeholder, and detection version mark to obtain the multidimensional observation vector. The multidimensional observation vector is defined as a set of elements used to characterize the position of a single sampling detection in the multidimensional detection space, and forms a traceable primary key with the same product number.

[0083] At the end of this step, the output is named Multidimensional Observation Vector. Its subsequent input position is the Multidimensional Observation Vector of S400. S400 uses the Multidimensional Observation Vector to perform working condition indexing and theoretical value retrieval with the combustion feature fingerprint database and generate an online calibration parameter set. Then, after S400 outputs the closed-loop configuration, it is sent back to S100 as part of the closed-loop configuration to enter the next round of fingerprint database construction link.

[0084] In summary, the technical effects of this step are as follows: by performing carbon dioxide extraction, isotope mass spectrometry analysis, detection standardization, and multidimensional feature coupling operations on the sample's physical property set, a multidimensional observation vector bound to the same sampling metadata is formed, providing a unified input for subsequent online calibration and inversion calculation links.

[0085] S400, based on multi-dimensional observation vectors, performs deviation vector calculation and online calibration parameter set generation operations, and performs low-carbon component blending ratio inversion, fuel life cycle carbon emission factor database call, and fuel consumption rate integral calculation operations to generate closed-loop configuration; Specifically, the input sources for this step include the multidimensional observation vector output from the previous main step S300 and the combustion feature fingerprint database output from the preceding main step S100 and continuously maintained, and further incorporate the fuel consumption rate and its timestamp related to vehicle operation. The multidimensional observation vector is defined as a set bound to the sample number, containing at least radioactive carbon isotope abundance, stable carbon isotope ratio, common gaseous component concentration, operating condition label, sampling start timestamp, sampling end timestamp, detection version marker, and synchronization quality marker. The combustion feature fingerprint database is defined as a set of fingerprint entries organized according to the target commercial vehicle engine identifier, operating condition index, blending ratio index, and model version marker. Each fingerprint entry contains at least a set of common gaseous pollutant concentrations, a set of theoretical carbon isotope distribution characteristics, operating condition label, boundary condition parameter entries, and reaction rate constant parameter entries. The fuel consumption rate is defined as the vehicle's fuel consumption per unit time, and its source can be a consumption rate record calculated from the engine controller's injection-related data, or a consumption rate record generated by the fuel flow meter. The timestamp is used for synchronization with the sampling time window. At the start of this step, the verification unit reads the sample number and establishes a running session record. It performs version consistency and integrity checks on the multidimensional observation vector and combustion feature fingerprint database, and performs missing marker and abnormal jump checks on the fuel consumption rate. When a key missing, version marker conflict, or synchronization quality marker is in a low quality state occurs, an abnormal record is generated and written into the abnormal record configuration. At the same time, the sample number is written into the queue to be reviewed and is not entered into the subsequent working condition indexing and positioning link.

[0086] Further, the process proceeds to the operational condition index positioning and theoretical value retrieval link. The operational condition index positioning is defined as the operation process of mapping a multi-dimensional observation vector to a candidate set of operational condition indexes and blending ratio indexes in a combustion feature fingerprint database. Specifically, the index positioning unit extracts operational condition labels, sampling start timestamps, sampling end timestamps, and the concentration of conventional gaseous components from the multi-dimensional observation vector to form a retrieval key. This retrieval key is then matched against the operational condition labels in the combustion feature fingerprint database. After successful matching, the index positioning unit filters candidate entries in the combustion feature fingerprint database based on the operational condition index and performs time window constraint judgment on the candidate entries. The time window constraint judgment is based on the operational condition window record associated with the sampling start timestamp and sampling end timestamp, generating candidate operational condition window markers. Subsequently, the index positioning unit performs feature synchronization operations on the set of conventional gaseous pollutant concentrations of the candidate entries and the conventional gaseous component concentrations. Feature synchronization operations include name mapping, unit consistency, missing bitmap synchronization, and dimensional verification. Upon completion of synchronization, the operational condition index positioning result is generated. The operating condition index location result includes at least a candidate operating condition index set, a candidate blending ratio index set, a model version tag set, and a candidate entry primary key set. Entering the theoretical value retrieval stage, the retrieval unit, based on the operating condition index location result, extracts the carbon isotope theoretical distribution feature set and the conventional gaseous pollutant concentration set corresponding to the candidate entry primary key set from the combustion feature fingerprint database, generating a theoretical feature set. This theoretical feature set is then synchronized twice with the aforementioned multi-dimensional observation vector within the same system to obtain a theoretical-measured comparison. This theoretical-measured comparison includes at least theoretical features, measured features, operating condition index, blending ratio index, and model version tag. If, during the retrieval process, a candidate entry is empty, a primary key mismatch occurs, or synchronization fails, a retrieval failure reason is generated and written into the exception record, and the sample number is marked as unusable.

[0087] In the deviation vector calculation and online calibration parameter set generation phase, the calculation unit performs difference generation, deviation vector encapsulation, and parameter update record writing on the theoretical-measured comparison. The deviation vector calculation is defined as the process of combining the differences between measured and theoretical features according to preset dimensions into a deviation vector in a unified order. The deviation vector includes radioactive carbon isotope abundance deviation, stable carbon isotope ratio deviation, conventional gaseous component concentration deviation, and their corresponding mass weight markers. These mass weight markers are derived from a synchronization mass marker, a detection version marker, and a missing bitmap. Specifically, the calculation unit first performs outlier screening on each deviation, including boundary threshold determination, jump determination, and consistency determination, and writes the deviations that pass the screening into the deviation vector. Then, the deviation vector is bound to the operating condition index, blending ratio index, and model version marker to generate a deviation vector record. Subsequently, the online calibration parameter set generation operation is performed. The online calibration parameter set is defined as a set of parameters formed after updating the reaction rate constant parameter entries and boundary condition parameter entries in the combustion feature fingerprint database, including parameter update values, parameter source markers, parameter version markers, and update session markers. In practice, the calibration unit reads the deviation vector record and loads the configuration items of the multivariate statistical analysis model or the chemometrics algorithm model. The multivariate statistical analysis model is defined as a statistical mapping model that maps multidimensional deviations to parameter update values, while the chemometrics algorithm model is defined as an algorithm model that fits the relationship between deviations and parameter space within the sample space. Within the running session, the calibration unit performs input calibration and determines the upper limit of the number of iterations for the model configuration items, and generates calibration quality markers from the fitting residual information. Based on the calibration quality markers, the calibration unit selects a writing strategy and writes the parameter update values ​​to the update records of the reaction rate constant parameter entry and the boundary condition parameter entry, respectively, while simultaneously writing the update session marker, thus realizing the generation of the online calibration parameter set. If the calibration quality marker is in a failing state, the calibration unit writes the calibration failure reason and retains the original parameter version marker, without outputting the updated parameter version. After the online calibration parameter set is generated, it is synchronously written to the closed-loop update log. The closed-loop update log includes the sample number, operating condition index, model version marker, and update session marker, used to facilitate subsequent write-backs.

[0088] Further, the process involves the inversion of low-carbon component blending ratios, access to the fuel lifecycle carbon emission factor database, and calculation of fuel consumption rate integrals. The low-carbon component blending ratio inversion is defined as the process of mapping multidimensional observation vectors to blending ratio indices and outputting corresponding blending ratio values ​​under the constraints of an online calibration parameter set. Specifically, the inversion unit accesses the online calibration parameter set and the operating condition index positioning results. First, it performs recalculation or re-retrieval operations on the fingerprint entries corresponding to candidate blending ratio indices in the combustion feature fingerprint database. These recalculation or re-retrieval operations are used to generate a theoretical feature set consistent with the online calibration parameter set. Subsequently, the inversion unit performs matching determination on the multidimensional observation vectors and candidate theoretical feature sets. The matching determination includes consistency checks, distance metric determination under mass weight label constraints, and candidate ranking. When the candidate ranking satisfies the uniqueness determination, the blending ratio value is output, and simultaneously, an inversion confidence flag and an inversion session flag are generated and written into the inversion record. Entering the fuel lifecycle carbon emission factor database retrieval phase, the factor unit searches the fuel lifecycle carbon emission factor database based on the blending ratio value and fuel type identifier. This database is defined as a set of factor records organized by fuel source, production process path, transportation mode, and carbon intensity. After retrieval, carbon emission factor records are obtained, and carbon intensity is bound to the inversion record within the same time window. Entering the fuel consumption rate integral calculation phase, the integral unit performs a window accumulation operation on the fuel consumption rate within the window corresponding to the sampling start timestamp to the sampling end timestamp. The window accumulation operation includes time synchronization, a completion strategy under missing bitmap constraints, and writing of the accumulated record to obtain the window fuel consumption. Subsequently, the window fuel consumption is coupled with carbon intensity and blending ratio values ​​within the same session to generate dynamic carbon footprint assessment data, and an assessment session marker and data version marker are written. If the fuel consumption rate is missing a critical interval within the window, the integral unit generates an integration unavailable marker and writes it to the anomaly record, while retaining the window fuel consumption for the available interval for subsequent backtracking.

[0089] At the end of this step, the online calibration parameter set, inversion records, dynamic carbon footprint assessment data, closed-loop update log, and anomaly records are encapsulated and written back to generate a closed-loop configuration. This closed-loop configuration is recorded as the output product in this step, named "Closed-Loop Configuration," and serves as the input source for the next run. Its subsequent input location is the closed-loop configuration of S100, enabling S100 to load the updated online calibration parameter set and model version marker during the specification and coupled simulation task orchestration phase. Simultaneously, the key feature operating point set configuration and anomaly records in the closed-loop configuration participate in the generation of sampling trigger commands and the writing back of sample pre-operation records in S200 during subsequent runs, thus forming a traceable link with the multi-dimensional observation vectors of S300, creating a continuous session.

[0090] In one embodiment, in the deviation vector calculation and online calibration parameter set generation process of step S400, assuming that the target commercial vehicle engine is operating under steady-state high-speed cruise conditions, the multidimensional observation vector obtained through steps S200-S300 shows the following key deviations: based on the deviation of radioactive carbon isotope abundance and the deviation of conventional components, the reaction rate constant is corrected by the Arrhenius equation. Radiocarbon isotope abundance bias (The measured value is higher than the theoretical value) Stable carbon isotope ratio deviation Nitrogen oxide concentration deviation carbon monoxide concentration deviation Mapping process based on physical / chemical laws 1. Fundamental Principles and Mathematical Models In chemical reaction kinetics, the rate constants of key reaction steps obey the Arrhenius equation: in: : No. The rate constant of each elementary reaction; Pre-exponential factor; :activation energy; Universal gas constant; : Reaction temperature; Meanwhile, the boundary condition parameters in computational fluid dynamics include: Intake temperature ; Turbulence model constants ; Wall thermal boundary condition coefficient ; 2. Algorithm operation and training process; A multivariate statistical analysis model (using partial least squares regression, PLS-R) was trained using a historical calibration dataset to establish a bias vector. With parameter correction amount Mapping relationship: The training dataset contains Groups of samples, each group containing: Deviation vector: Parameter correction amount: The PLS-R model establishes the latent variable space by maximizing the covariance, ultimately obtaining the mapping relationship: in: Weight matrix (dimension: ) : Residual vector The specific coefficients are obtained through training, for example: 3. Calculation of specific parameter corrections For the currently observed deviation vector: Substituting the values ​​into the trained model yields the parameter corrections: 4. Parameter updates and interpretation of physical meaning; The updated key parameters are: New activation energy: ; New pre-finger factor: ; New intake air temperature: ; Analysis of physical significance: A positive value indicates that the actual biomass carbon ratio is higher than theoretically predicted, and the combustion reaction pathway needs to be adjusted. Positive combination Negative values ​​indicate that the local temperature field distribution needs to be corrected. Increased activation energy The change in reaction energy barrier due to the increase in biomass components; Intake temperature fine-tuning It reflects the difference between actual intake conditions and theoretical settings; 5. Coordinated adjustment of boundary conditions; Simultaneously, based on the deviation pattern identification of the intake system error, the CFD boundary conditions are updated: Turbulence model constant correction: ; Adjustment of wall heat transfer coefficient: ; The final generated online calibration parameter set includes: Reaction kinetic parameters: ; Boundary condition parameters: ; Metadata: Parameter version tag, calibration session identifier, confidence score.

[0091] In summary, the technical effects of this step are as follows: by using multi-dimensional observation vectors to perform working condition indexing and positioning, theoretical value retrieval, deviation vector calculation and online calibration parameter set generation with the combustion feature fingerprint database, and linking the inversion and verification links to generate a closed-loop configuration, a closed-loop input that can be used for the next round of fingerprint database construction and sampling detection linkage is formed.

[0092] Figure 3 This application provides an architecture diagram of a method for dynamic monitoring of low-carbon components in vehicle fuels based on multi-dimensional feature coupling. The fully systematized implementation of this method can be structured into a three-tiered system consisting of a data acquisition layer, a data operation layer, and an application layer. Each layer contains specific functional units, which are vertically integrated and horizontally coordinated through strictly defined interfaces, data, and control logic, forming an organic whole from raw data perception to advanced intelligent decision-making, possessing closed-loop self-optimization capabilities. This architecture not only embodies the physical carrier and logical mapping of the method steps (S100-S400), but also clearly reveals the data flow, control flow, and feedback mechanism. The following, in conjunction with the illustrated architecture, elaborates on the definition, core function, specific implementation method, internal operation process, and inter-unit connections and interactions of each layer and unit.

[0093] I. Data Acquisition Layer (Input End); The data acquisition layer is the interface through which the system directly interacts with the physical world (the target engine and exhaust gas). It is responsible for acquiring all raw data and physical samples, and its output forms the basis for data operations in the upper layers. This layer emphasizes real-time performance, synchronization, and fidelity to ensure that the data used in subsequent analyses accurately reflects the transient operating state of the engine and the actual chemical properties of the exhaust gas.

[0094] 1. Onboard Signal Acquisition Unit. Definition and Function: This unit is the "nerve ending" of the system for real-time sensing of the engine's operating status. Its core function is to continuously and synchronously acquire various real-time operating parameters generated by the vehicle during operation, providing crucial time-series data for determining sampling timing, defining the operating condition boundaries for simulation and analysis, and ultimately calculating the carbon footprint. It plays a bridging role between the digital system and the physical vehicle.

[0095] Implementation method and internal process: This unit communicates directly with the vehicle's on-board diagnostic system interface and controller area network bus through a hardware interface.

[0096] Data Sources and Acquisition: Data is primarily acquired through two parallel channels: first, via the OBD interface, real-time parameters from the engine control unit are read, such as engine speed, torque demand, fuel injection quantity, intake manifold pressure, and coolant temperature; second, via the CAN bus interface, signals from other relevant control units (such as the transmission control unit and emission control unit) or sensors are acquired, such as vehicle speed, exhaust temperature, and air-fuel ratio sensor signals. These signals together constitute a multi-dimensional vector describing the engine's transient operating point.

[0097] Conditional Triggering: This unit embeds a logic judgment unit that continuously compares the real-time collected operating condition data stream (such as speed-torque pairs) with the "key feature operating condition point set configuration" (such as speed-torque boundary conditions for high-load ramp-up, cold start, etc.) issued by the data operation layer. Once the real-time data enters the "capture window" of the preset operating condition point, this unit immediately generates a high-priority "trigger signal". This signal is not only a Boolean value, but also encapsulates a precise trigger timestamp, the triggered operating condition label, and a snapshot of all relevant operating condition parameters at this moment.

[0098] Synchronization Locking: To ensure data time consistency, the unit integrates a high-precision clock source and performs hardware timestamping and software-level clock offset correction and synchronization operations on different data streams from OBD and CAN, forming a unified "time-condition" data stream with millisecond-level synchronization accuracy. This synchronized data stream is packaged together with the trigger signal to form a "condition boundary data packet," which is then uploaded to the data operation layer.

[0099] Permitted Process: Allows deployment in various scenarios, including actual vehicle road operation, chassis dynamometer testing, and engine bench testing. Allows dynamic loading or switching of different "key feature operating condition point set configurations" according to different test specifications or analysis requirements. Allows recording complete operating condition data before and after the triggered event for subsequent analysis of the stability and representativeness of the operating conditions.

[0100] 2. Physical Sampling and Detection Unit. Definition and Function: This unit is the "sensory system" for obtaining the "truth" about the physicochemical properties of exhaust gases. Its core function is to automatically and accurately capture transient exhaust gas samples under specific operating conditions after receiving a trigger signal from the on-board signal acquisition unit, and then use high-precision analytical instruments (especially isotope mass spectrometers) to perform in-depth analysis, obtaining measured values ​​of multi-dimensional chemical characteristics, including the abundance of radioactive carbon isotopes. Its output data serves as the "gold standard" for calibrating and correcting theoretical models.

[0101] Implementation method and internal process: This unit consists of two parts: an automated sampling system and a precision analytical instrument interface.

[0102] Sampler Execution: Upon receiving a trigger signal, the control unit drives a fast-acting solenoid valve to momentarily divert a portion of the exhaust gas from the main stream of the engine exhaust pipe into the "constant-volume sampling system" or a dedicated "sampling bag / canister." Sampling duration and flow rate are precisely controlled by the program to ensure that the collected sample represents the steady-state or quasi-steady-state emissions at the trigger moment. The entire sampling process is monitored, and its status (such as valve opening / closing, flow rate, pressure, and temperature) is recorded, forming "sampling metadata."

[0103] Sample preprocessing: Raw exhaust gas samples typically require preprocessing to ensure analytical quality. The process may include: removing moisture via condensation or a Nafion tube dryer; removing soot and particulate matter via a particulate filter; and, if necessary, introducing the sample into an online or offline "carbon dioxide separation and purification unit" to selectively separate and purify the CO2 components in the exhaust gas using low-temperature adsorption, chemical absorption, or gas chromatography, converting them into a form suitable for mass spectrometry injection.

[0104] Isotope Detector Interface: The purified CO2 sample is fed into an accelerator mass spectrometer or a high-precision stable isotope ratio mass spectrometer. AMS is used for ultra-high precision measurement of radioactive carbon isotope ratios to obtain the "percentage of modern carbon" value, which is the absolute benchmark for quantifying the carbon content of biomass. IRMS is used for accurate measurement of stable carbon isotope ratios, providing auxiliary fuel source fingerprint information. The control and data acquisition of the detection instruments are integrated with the system through a standardized interface, automatically acquiring raw spectral data, peak intensities, and ratio results, and performing necessary background correction, standard correction, and isotope fractionation correction.

[0105] Data Encapsulation: Finally, the original pMC values, δ13C values, and conventional gas concentrations obtained from the bypass analyzer, along with sampling metadata, instrument analysis conditions, and calibration parameters, are encapsulated to form a standardized "measured calibration data package." The core of this package is the measured portion of the "multidimensional observation vector" described in step S300. This data package, representing the digitization of the "true chemical state," is uploaded to the data manipulation layer.

[0106] Permitted processes: Allows for various sampling strategies, such as instantaneous sampling, proportional sampling, or integral calculus sampling. Allows integration with various mass spectrometers or analyzers via standard protocols. Allows for the insertion of quality control samples into the detection process and automatically evaluates the precision and accuracy of the detection data, generating data quality labels.

[0107] II. Data Operation Layer (Core End); The data operation layer is the system's "intelligent hub" and "algorithm engine," responsible for executing all complex calculations, analyses, comparisons, and decisions. It receives raw data from the acquisition layer, utilizes built-in models and algorithms to transform data into information, and generates key outputs such as corrected models and inversion results.

[0108] 1. Combustion Simulation Unit. Definition and Function: This unit is the core carrier of system knowledge and the generator of theoretical benchmarks. Its core function is to construct and maintain a digital "combustion characteristic fingerprint database" for a specific target engine. This fingerprint database, through high-fidelity combustion simulation, pre-calculates the theoretical chemical and isotopic characteristics of engine exhaust gas under different operating conditions and fuel blending ratios. It provides "prior knowledge" and a benchmark for comparison for the entire method, and is the embodiment of the "virtual" part in the "virtual-real combination".

[0109] Implementation method and internal process: This unit is the core executor of step S100.

[0110] It contains a pre-built fingerprint database: the unit internally stores and manages a "combustion characteristic fingerprint database," which is a standardized database. Its construction process is as described in S100: based on the detailed technical specifications of the target engine, a three-dimensional CFD model is coupled with a detailed chemical reaction kinetic model, and extensive simulation calculations are performed within a full parameter space spanned by "operating condition index" and "blending ratio index." The calculation results are structured and stored, forming the fingerprint database.

[0111] Parameter Calibration and Adjustment: The key innovation of this unit lies in its non-static nature. It receives "corrected model parameters" from the model correction unit. During initial construction or periodic updates of the fingerprint database, the unit does not always use the default model parameters. Instead, it can load these corrected parameters (such as updated reaction kinetic pre-exponential factors, turbulence model constants, etc.) learned online and driven by measured data, and rerun or locally correct the simulation task accordingly. This generates a "calibrated" combustion characteristic fingerprint database that better matches the current state of a specific engine. This enables the theoretical benchmark to be adaptive and evolutionary.

[0112] Providing theoretical baseline values: When receiving a query request from an upper layer (such as a component inversion unit) containing specific "operating condition indexes" and candidate "blending ratios," this unit can quickly retrieve the corresponding theoretical exhaust gas feature vector from the fingerprint database or through interpolation calculation. This vector is the theoretical value that "should" be observed under that assumption.

[0113] Permitted processes: Allows for different combinations of CFD and chemical reaction mechanism models. Allows the fingerprint database to be deployed as a local database, a cloud database, or a hybrid format. Allows for updating the fingerprint database via "incremental update" or "full reconstruction" upon receiving new calibration parameters. Allows for the management of simulation models with different fidelity to balance computational efficiency and accuracy requirements.

[0114] 2. Model correction unit; Definition and Function: This unit is the "algorithm core" and "referee" of the entire system. Its core function is to manipulate the difference between "real" and "virtual" data and make fair judgments and adjustments. It receives "measured data" from the physical sampling and detection unit and "theoretical benchmark values" provided by the combustion simulation unit, calculates the deviation through intelligent algorithms, and decides how to adjust the parameters of the simulation model to reduce this deviation. It ensures that the system's theoretical model can continuously approximate the physical processes of the real world and is the core decision-making unit for achieving closed-loop optimization.

[0115] Implementation method and internal process: This unit is the core of the "deviation vector calculation and online calibration parameter set generation operation" in step S400.

[0116] The core of the algorithm is a built-in multivariate statistical analysis or machine learning algorithm, such as partial least squares regression, Gaussian process regression, or neural networks. Its input is a "bias vector," which is the multidimensional difference between the measured feature vector and the theoretical feature vector in the current fingerprint database under the corresponding working conditions and different mixing ratios.

[0117] Receive measured data and adjust simulation parameters: the algorithm analyzes the pattern of the deviation vector. For example, if 14 The deviation of C exhibits a systematic positive shift across all test conditions, while the deviation of NOx is strongly correlated with the test conditions. The algorithm learns this pattern and maps it to adjustments to key parameters of the underlying simulation model. For example, it might determine that the rate constant of a key reaction in the reaction mechanism needs adjustment, or that parameters characterizing mixing efficiency in the CFD model need adjustment. Ultimately, it outputs a specific, quantified set of "online calibration parameters."

[0118] Referee role: This unit does not directly output fuel ratios; its core responsibility is to assess the "predictive accuracy" of the current theoretical model (fingerprint database) and "improve its refereeing level" by adjusting model parameters. It ensures that the predictions of this "theoretical prophet"—the combustion simulation unit—become increasingly accurate.

[0119] Allowed processes: Multiple calibration algorithms can be integrated, and algorithm selection or fusion can be based on historical performance. Calibration trigger conditions and frequencies can be set, such as initiating a calibration after accumulating a certain amount of new measured data. Uncertainty quantification of calibration results and evaluation of calibration effectiveness are also possible.

[0120] 3. Component inversion unit; Definition and Function: This unit is the system's "solver," directly addressing end-user needs. Its core function is to utilize a high-fidelity "combustion characteristic fingerprint database," optimized by the model correction unit, combined with the latest measured exhaust gas data, to reverse-engineer the blending ratio of low-carbon fuel components most likely to produce the measured exhaust gas characteristic. It is a crucial step in transforming complex, multi-dimensional detection data into intuitive and usable engineering solutions (fuel blending ratios).

[0121] Implementation method and internal process: This unit is the core of the "low carbon component blending ratio inversion operation" in step S400.

[0122] Using the modified model: When performing inversion calculations, this unit calls the combustion simulation unit service that has been "calibrated" by the model correction unit, that is, it uses the latest and more accurate "calibrated fingerprint database" as the theoretical basis for its reverse lookup.

[0123] Output fuel blending: The inversion process is typically transformed into an optimization problem. The unit compares the measured multidimensional observation vectors with theoretical vectors corresponding to different blending ratios under the same operating conditions in a calibrated fingerprint database. By calculating the distances between vectors (such as weighted Euclidean distance or Mahalanobis distance), the theoretical vector that best matches the measured vector is found; the blending ratio corresponding to this theoretical vector is the inversion result. More advanced algorithms may directly construct a surrogate model or inverse model from the feature space to the blending ratio.

[0124] Output results: The unit not only outputs a single mixing ratio value, but also outputs relevant confidence indicators (such as inversion residuals, distance from suboptimal solutions), operating conditions on which the inversion is based, model version used, etc., which are all packaged into a unified "inversion result package".

[0125] Permitted Process: Allows the use of different inversion algorithms, such as nearest neighbor search, interpolation, and surrogate model-based optimization. Allows outputting a probability distribution of the mixing ratio instead of a single point value. Allows triggering alarms or requesting supplementary sampling when the inversion confidence is too low.

[0126] III. Application Layer (Output End); The application layer is the ultimate embodiment of the system's value. It is responsible for transforming the core results (mixing ratio) generated by the data operation layer into business deliverables that end users can directly use—carbon footprint reports—and providing a human-computer interaction interface.

[0127] 1. LCA Accounting Unit. Definition and Function: This unit is the system's "environmental accountant," responsible for converting the chemical properties of fuel into environmental impact indicators. Its core function is to calculate the full life-cycle carbon footprint of vehicles based on real-time, dynamic fuel blending ratios obtained from the data operation layer, combined with real-time vehicle fuel consumption data, thus achieving a value closed loop from monitoring to assessment.

[0128] Implementation methods and internal processes: Calculate carbon footprint by calling factor library: The unit is embedded or connected to a "fuel life cycle carbon emission factor database", which stores greenhouse gas emission intensity data of various basic fuels (such as fossil diesel, biodiesel from different feedstock routes, electric fuels, etc.) from the "well to the wheel".

[0129] Real-time blending ratio × carbon emission factor × real-time fuel consumption: The unit obtains the real-time / time-period "blending ratio of low-carbon components in the fuel" from the component inversion unit and the "real-time fuel consumption rate" data stream corresponding to the time window from the on-board signal acquisition unit. Based on the blending ratio, the comprehensive carbon emission factor of the blended fuel is calculated. The fuel consumption rate is integrated over the time window to obtain the total fuel consumption.

[0130] Integral Accumulation (Time Axis): The carbon footprint for the given time period is calculated by multiplying the comprehensive carbon emission factor by the total fuel consumption. The system can accumulate the carbon footprint over consecutive trips or reporting periods along the time axis to obtain the cumulative emissions.

[0131] Output: Finally, the unit output carbon footprint assessment results include: time interval, driving mileage, total fuel consumption, average blending ratio, overall carbon intensity, total carbon emissions, carbon emissions per unit mileage, etc.

[0132] Allowed processes: Allows the use of LCA databases from different sources and with different accounting standards. Allows users to define accounting boundaries. Allows the generation of reports at different granularities, such as single trip reports, daily reports, monthly reports, and annual reports.

[0133] 2. Human-Computer Interface / Terminal. Definition and Function: This unit serves as a bridge between the system and the end user, responsible for presenting all process information, status, and results to the user in an intuitive and readable format, and providing necessary control and interaction entry points.

[0134] Implementation methods and internal processes: The report displays key information such as real-time inversion of fuel blending ratio changes, real-time / cumulative carbon footprint, carbon emission intensity comparison, and gap with preset targets in the form of graphs, charts, dashboards, and map tracks.

[0135] User-readable reports: Automatically generated text reports that can cover execution summaries, monitoring cycles, fuel usage and composition analysis, carbon emission accounting results, data quality descriptions, calibration history, etc., and can be exported to common document formats.

[0136] Permissions: Allows interaction through various means such as web interfaces, mobile applications, in-vehicle terminals, or API interfaces. Allows setting different data viewing and operation permissions for users with different roles.

[0137] IV. Connection methods and collaborative workflows between units The entire system operates on a data flow-centric model, with triggering and feedback mechanisms forming a tight collaborative system. 1. Forward Workflow (Monitoring and Accounting Flow): Trigger: The on-board signal acquisition unit monitors the operating conditions and, upon reaching the preset conditions, sends a trigger signal and the operating condition boundary to the physical sampling and detection unit.

[0138] Acquisition and Detection: The physical sampling and detection unit performs sampling and detection, generates measured calibration data, and sends it to the data operation layer.

[0139] Operation and Inversion: Within the data operation layer, the model correction unit can calibrate the combustion simulation unit based on new data. Subsequently, the composition inversion unit uses theoretical data from the (potentially calibrated) combustion simulation unit and new measured data to calculate and output the fuel blending ratio.

[0140] Calculation and Display: The proportions output by the component inversion unit and the fuel consumption data from the vehicle signal acquisition unit are input together into the LCA calculation unit at the application layer to calculate the carbon footprint. The final result is displayed on the human-machine interface / terminal.

[0141] 2. Closed-loop feedback flow (model optimization flow): This is the core innovation link of this system. The model correction unit collects a batch of measured calibration data uploaded by the physical sampling and detection unit periodically or irregularly, and calls the corresponding theoretical values ​​from the combustion simulation unit to calculate the deviation and run its core algorithm to generate a new and better set of model parameters (corrected model parameters).

[0142] This new set of parameters is passed to the combustion simulation unit. The combustion simulation unit uses the new parameters to recalculate the affected simulation cases or interpolate and correct the fingerprint database, thereby updating its internal "combustion feature fingerprint database".

[0143] Subsequently, all inversion requests from the component inversion unit will be based on this updated and more accurate fingerprint database. This forms an enhanced closed loop of "experimental measurement → model calibration → more accurate theory → more accurate inversion," enabling the system to continuously evolve with use and adapt to complex situations such as engine aging and fuel changes.

[0144] The system architecture of this invention, through a clear division of three levels and precise collaboration of six core units, engineers and modularizes the entire technology chain of "simulation modeling, intelligent sampling, precision detection, model correction, component inversion, and carbon footprint accounting." The data acquisition layer ensures the authenticity and synchronization of the data source; the data operation layer, acting as the intelligent hub, drives the "combustion simulation unit" (the predictor) to continuously learn and improve through the "model correction unit" (the referee and coach), thereby enabling the "component inversion unit" (the solver) to provide more accurate answers; the application layer then transforms the answers into final user value. The units are connected through standardized data packets and clear interface protocols. In particular, the parameter feedback loop from the model correction unit to the combustion simulation unit constitutes the soul of the system's continuous self-optimization capability, ensuring the long-term effectiveness, high accuracy, and strong robustness of the monitoring method.

Claims

1. A method for dynamic monitoring of low-carbon components in vehicle fuels based on multidimensional feature coupling, characterized in that, include: Obtain the basic specifications and pre-set combustion model parameters of the target commercial vehicle engine, perform coupled combustion dynamics and fluid dynamics simulation based on the engine parameters, and construct a combustion feature fingerprint database. Based on the combustion feature fingerprint database, timestamp synchronization and key feature operating point set matching operations are performed, and exhaust gas sampling device triggering and sample pre-operation are executed to generate sample physical property set; Based on the set of physical properties of the sample, isotope mass spectrometry analysis, detection standardization and multidimensional feature coupling operation are performed to obtain a multidimensional observation vector; Based on multidimensional observation vectors, the system performs deviation vector calculation and online calibration parameter set generation operations, and performs low-carbon component blending ratio inversion, fuel life cycle carbon emission factor database retrieval, and fuel consumption rate integral calculation operations to generate a closed-loop configuration.

2. The method according to claim 1, characterized in that, The process of implementing the standards also includes: The standardized operation includes a combination of operations to unify and constrain naming, units, types, value ranges, missing flags, and version flags. It also performs enumeration validity calibration of engine operating condition parameter ranges and sensor data calibration, endpoint coverage calibration and step size consistency calibration of mixing ratio gradient configuration, and conflict item determination and deduplication operations of key feature operating condition point set configuration.

3. The method according to claim 1, characterized in that, The process of orchestrating coupled simulation tasks also includes: The coupled simulation task orchestration operation includes extracting the set of operating condition indexes and the set of mixing ratio indexes from the normalized configuration to generate a set of task coordinates, and generating a simulation task identifier, input boundary conditions, initial conditions, numerical grid version and solver version for each coordinate. At the same time, consistency calibration is performed between the chemical reaction kinetic model configuration and the three-dimensional computational fluid dynamics model configuration. The consistency calibration includes matching calibration between the fuel supply system type and fuel component entries, matching calibration between the post-operation configuration and product component entries, and matching calibration between geometric boundary entries and numerical grid version.

4. The method according to claim 1, characterized in that, Building a combustion feature fingerprint database also includes: The combustion feature fingerprint database contains a set of conventional gaseous pollutant concentrations and a set of theoretical carbon isotope distribution features, and is organized according to the target commercial vehicle engine identifier, operating condition index, blending ratio index, and model version label.

5. The method according to claim 1, characterized in that, The process of synchronizing timestamps and matching key feature working point sets also includes: The timestamp synchronization operation includes performing integrity screening, clock offset estimation, and nearest neighbor pairing or window interpolation pairing on the vehicle diagnostic system operating condition data and bus interface operating condition data. It also includes performing key feature operating condition point set matching operation, which includes extracting real-time speed, load torque, intake air temperature, fuel injection pulse width, vehicle speed and gear from the operating condition data stream and performing sliding window aggregation to obtain window statistics. Then, it determines the operating condition label and trigger timestamp according to the trigger conditions in the key feature operating condition point set configuration, and maps them to the operating condition index of the combustion feature fingerprint database to obtain the fingerprint retrieval index and sampling trigger command.

6. The method according to claim 1, characterized in that, The process of triggering the exhaust gas sampling device and pre-processing the sample also includes: The exhaust gas sampling device triggering and sample pre-operation includes sending a sampling trigger command to the exhaust gas sampling device and performing a full-process status readback, as well as performing encapsulation, sealing, purification and carbon dioxide separation preparation on the collected transient exhaust gas sample. The encapsulation includes sealing the sampling container and writing the sample number; the sealing includes writing a leak check record; the purification includes recording purification items to remove water vapor and particulate matter; and the carbon dioxide separation preparation includes introducing the sample into the separation unit inlet and recording the inlet pressure and inlet temperature.

7. The method according to claim 1, characterized in that, The process of performing isotope mass spectrometry analysis, detection standardization, and multidimensional feature coupling also includes: The isotope mass spectrometry analysis operation includes performing injection, ionization, mass separation, and signal acquisition on the carbon dioxide separation products in the sample physical property set to obtain the raw isotope mass spectrometry signal, and extracting radiocarbon isotope abundance data and stable carbon isotope ratio data; the detection standardization operation includes performing unified system operations on radiocarbon isotope abundance data, stable carbon isotope ratio data, and sampling metadata, including outlier screening of signal intensity sequences, generation of quality scores for peak shape fitting residual information, generation of missing bitmaps for missing markers, and generation of calibration versions of calibration markers; the multidimensional feature coupling operation includes synchronizing the standardized detection results with the sampling metadata in the sample physical property set.

8. The method according to claim 1, characterized in that, The process of performing deviation vector calculation and online calibration parameter set generation also includes: The deviation vector calculation and online calibration parameter set generation operation involves extracting measured and theoretical features from a multidimensional observation vector and a combustion feature fingerprint database to calculate the deviation vector. The deviation vector includes radioactive carbon isotope abundance deviation, stable carbon isotope ratio deviation, and conventional gaseous component concentration deviation. A multivariate statistical analysis model or chemometric algorithm model is used to map the deviations to parameter update values ​​to generate an online calibration parameter set. The online calibration parameter set includes update values ​​for reaction rate constant parameter entries and boundary condition parameter entries. The reaction rate constant is corrected using the Arrhenius equation based on the radioactive carbon isotope abundance deviation and conventional component deviation.

9. The method according to claim 1, characterized in that, The process of inverting the blending ratio of low-carbon components and accessing the fuel life cycle carbon emission factor database also includes: The low-carbon component blending ratio inversion operation includes, under the constraints of the online calibration parameter set, mapping the multidimensional observation vector to the blending ratio index in the combustion feature fingerprint database to output the blending ratio value, generating inversion confidence markers and inversion session markers, and performing a fuel life cycle carbon emission factor database retrieval operation. The fuel life cycle carbon emission factor database retrieval operation includes searching the fuel life cycle carbon emission factor database based on the blending ratio value and fuel type identifier to obtain carbon emission factor records.

10. The method according to claim 1, characterized in that, The process of calculating fuel consumption rate credits also includes: The fuel consumption rate integral calculation operation includes performing a window accumulation operation on the fuel consumption rate within the window corresponding to the sampling start time stamp to the sampling end time stamp to obtain the window fuel consumption amount, and coupling the window fuel consumption amount with carbon intensity and blending ratio to generate dynamic carbon footprint assessment data, generating a closed-loop configuration. The closed-loop configuration includes an online calibration parameter set, inversion records, dynamic carbon footprint assessment data, closed-loop update log, and anomaly records.

Images