A vehicle condition evaluation method

Abstract

This invention relates to the field of data assessment technology, and more particularly to a vehicle condition assessment method. The method includes the following steps: acquiring multi-source heterogeneous vehicle condition data of used cars; performing entity alignment and normalization processing on the multi-source heterogeneous vehicle condition data of used cars and constructing a vehicle component-fault association topology graph; extracting the node degree and edge association features corresponding to the vehicle component-fault association topology graph to generate a component-fault association strength matrix; performing spatiotemporal feature fusion and dimensionality reduction processing on the component-fault association strength matrix to output a vehicle condition health vector corresponding to the fused spatiotemporal features; constructing a multi-dimensional risk assessment function based on the vehicle condition health vector, outputting the vehicle condition risk level and component risk proportion; generating a visualization report based on the vehicle condition risk level and component risk proportion, including a vehicle condition radar chart, a fault time axis, and a residual value prediction curve, and pushing it to a web or mobile terminal. This invention can achieve a comprehensive assessment process covering complex vehicle conditions.

Description

[0001] Technical Field

[0002] This invention relates to the field of data evaluation technology, and in particular to a vehicle condition evaluation method. Background Technology

[0003] In the used car market, vehicle condition assessment is a core element determining the fairness of transactions and protecting consumer rights. Its accuracy directly impacts market trust and transaction efficiency. Currently, the used car market generally suffers from information asymmetry. Buyers struggle to fully grasp crucial information such as a vehicle's history of repairs, accidents, and operating conditions. Traditional vehicle condition assessments rely primarily on manual inspection and experience-based judgment, which is not only inefficient but also prone to inconsistencies in assessment results due to variations in the professional levels of inspectors, easily leading to transaction disputes.

[0004] Furthermore, Chinese patent CN117172420A discloses a vehicle condition assessment and testing system and method, including a data acquisition unit, a data aggregation unit, a calculation and evaluation unit, and a vehicle database. The vehicle database contains preset vehicle condition parameters, including at least detailed parameter statements and values ​​for the frame, exterior, and interior. Each parameter has a preset score. The data acquisition unit selectively inputs vehicle condition parameters and scores from the vehicle database via a terminal. The data aggregation unit stores the input vehicle condition parameter scores into the vehicle database. The calculation and evaluation unit calculates and evaluates the vehicle condition parameter values ​​in the database and outputs a vehicle condition report. This invention also discloses a vehicle condition assessment method. This invention enables a unified standard to be formed between various platforms and subsystems. Standardized output allows used car industry professionals to quickly and clearly identify vehicle conditions. This invention standardizes vehicle condition assessment by unifying parameter standards, thereby improving assessment efficiency. However, this solution relies on preset parameters and manual selective input, does not involve the automatic parsing and extraction of unstructured text data, cannot make full use of the hidden information in massive historical text data, and lacks the ability to integrate multi-source data, making it difficult to meet the comprehensive assessment needs of complex vehicle conditions. Summary of the Invention

[0005] To address the aforementioned technical problems in existing vehicle condition assessment processes, this invention provides a method that solves the issues of chaotic data formats and difficulty in reusing multi-source heterogeneous vehicle condition data through entity alignment and normalization, forming a standardized structured dataset. This overcomes the shortcomings of similar patents that cannot parse unstructured data and utilize massive amounts of historical information. Based on topology graphs, feature fusion, and risk assessment functions, it achieves component fault correlation analysis and accurate determination of vehicle health and risk levels. Combined with a visual report, the assessment results are presented intuitively, overcoming the limitations of manual experience-based judgment and reducing assessment discrepancies and transaction disputes. Simultaneously, residual value prediction and multi-terminal push functions provide comprehensive reference for both parties in the transaction, alleviating information asymmetry, improving market trust and transaction efficiency, thereby promoting the standardization and intelligent upgrading of used car condition assessment. The method includes the following steps:

[0006] Step S1: Obtain multi-source heterogeneous vehicle condition data of used cars, perform entity alignment and normalization processing on the multi-source heterogeneous vehicle condition data of used cars, and obtain a structured dataset of vehicle condition elements in a unified format.

[0007] Step S2: Construct a vehicle component-fault association topology graph based on the structured dataset of vehicle condition elements, extract the node degree and edge association features corresponding to the vehicle component-fault association topology graph, and generate a component fault association strength matrix;

[0008] Step S3: Perform spatiotemporal feature fusion and dimensionality reduction on the component fault correlation strength matrix, and output the vehicle health vector corresponding to the fused spatiotemporal features;

[0009] Step S4: Construct a multi-dimensional risk assessment function based on the vehicle health vector, and output the vehicle condition risk level and component risk ratio;

[0010] Step S5: Generate a visual report based on the vehicle condition risk level and component risk ratio, including a vehicle condition radar chart, a failure timeline, and a residual value prediction curve, and push it to the Web or mobile device.

[0011] This invention acquires multi-source heterogeneous vehicle condition data of used cars. Through entity alignment and normalization, it transforms scattered and inconsistently formatted vehicle condition data into a structured dataset with a unified format. This effectively solves the shortcomings of similar patents, such as the disordered and unreusable nature of multi-source data, and the inability to parse unstructured data or utilize massive amounts of historical hidden information. It achieves standardized integration of vehicle condition data, providing high-quality data support for subsequent assessment work and eliminating the tedious burden of manual data processing. Secondly, it constructs a scientific and systematic vehicle condition assessment system. Based on the structured dataset, it builds a vehicle component-fault association topology graph, extracts the association features of nodes and edges to generate a strength matrix, and then obtains a vehicle health vector through spatiotemporal feature fusion and dimensionality reduction. Finally, relying on a multi-dimensional risk assessment function, it outputs the vehicle condition risk level and component risk ratio. This process overcomes the limitations of manual experience-based judgment, comprehensively captures the association relationships and spatiotemporal variation characteristics of vehicle component faults, and can accurately identify potential problems under complex vehicle conditions. It solves the problems of traditional assessments being unable to comprehensively cover vehicle faults and having highly subjective assessment results, while also compensating for the shortcomings of similar patents in lacking multi-source data integration capabilities and being unable to handle complex vehicle conditions. Finally, a visual report is generated, including a vehicle condition radar chart, a fault timeline, and a residual value prediction curve, and pushed to the web or mobile devices. This intuitively presents the assessment results, effectively solving the problems of obscure and difficult-to-understand vehicle condition assessment results and poor information transmission. It allows both parties in the transaction to clearly and quickly grasp the true condition of the vehicle, alleviating information asymmetry and reducing transaction disputes caused by misperceptions about vehicle condition. Overall, the entire process of vehicle condition assessment, from data integration, feature extraction, risk analysis to result presentation, is automated and intelligent. This improves assessment efficiency while ensuring the consistency and objectivity of the assessment results, providing reliable vehicle condition references for used car transactions, enhancing market trust, and promoting the transformation of used car transactions towards transparency and intelligence.

[0012] Preferably, step S1 includes the following steps:

[0013] Step S11: Obtain multi-source heterogeneous vehicle condition data for used cars, including maintenance records, actual vehicle inspection values, and vehicle system operation logs;

[0014] Step S12: Perform entity extraction and relation extraction on the maintenance text records to obtain triplet data containing component name, fault type, and occurrence time;

[0015] Step S13: Filter outliers and normalize units on the actual vehicle inspection values ​​to obtain standardized inspection feature data;

[0016] Step S14: Perform timestamp alignment and event clustering on the vehicle system operation logs to obtain time-series operation event data;

[0017] Step S15: Align the triplet data, standardized detection feature data, and time-series running event data to generate a structured dataset of vehicle condition elements in a unified format.

[0018] This invention comprehensively acquires three core vehicle condition data types: maintenance texts, actual vehicle inspection values, and vehicle system operation logs. Different processing methods are employed for each data type: entity and relationship extraction is performed on the maintenance texts to uncover hidden component fault associations; anomalies are filtered and units are standardized for actual vehicle inspection values ​​to ensure data reliability; and time alignment and event clustering are performed on the vehicle system operation logs to analyze temporal operational characteristics. Finally, entity alignment integrates the three data types to generate a structured dataset in a unified format. This standardization and integration of multi-source heterogeneous data fully unlocks the value of massive historical vehicle condition data, overcomes the limitations of manual data processing, and provides comprehensive, high-quality data support for subsequent topology map construction and risk assessment. This improves the comprehensiveness and objectivity of vehicle condition assessment and alleviates the pain point of information asymmetry in the used car transaction market.

[0019] Preferably, step S2 includes the following steps:

[0020] Step S21: Extract vehicle component entities and fault event entities based on the structured dataset of vehicle condition elements, and construct an initial vehicle component-fault association topology map;

[0021] Step S22: Count the number of fault events associated with each component node in the initial vehicle component-fault association topology graph, and calculate the node degree value;

[0022] Step S23: Based on the occurrence frequency and time interval of each fault event in the structured dataset of vehicle condition elements, calculate the association feature value of each association edge in the initial vehicle component-fault association topology graph;

[0023] Step S24: Integrate the node degree value and the edge association feature value to generate the component fault association strength matrix;

[0024] Step S25: Perform spectral normalization on the component fault correlation strength matrix to obtain the normalized component fault correlation strength matrix.

[0025] This invention constructs an initial component-fault association topology map based on a structured dataset, clearly presenting the correspondence between components and faults, breaking the limitations of traditional assessments that rely on isolated analysis of individual components. By statistically analyzing node degree values, the frequency of fault associations for each component is quantified; edge association feature values ​​are calculated by combining fault occurrence frequency and time intervals, accurately reflecting the closeness of the association between components and faults. Integrating these two types of features generates an association strength matrix, which is then spectral normalized to ensure the rationality and comparability of the matrix data, achieving a quantitative representation of component-fault associations. This step transforms abstract fault associations into computable and analyzable matrix data, capturing potential fault associations between components, providing core feature support for subsequent spatiotemporal feature fusion and health assessment, improving the reliability of fault identification and assessment under complex vehicle conditions, reducing assessment bias caused by missing component-related faults, and further alleviating the problem of information asymmetry in transactions.

[0026] Preferably, step S23 includes the following steps:

[0027] Step S231: Extract the occurrence time and associated component information of each fault event from the structured dataset of vehicle condition elements;

[0028] Step S232: Calculate the time interval between consecutive failure events associated with the same component, and generate a time interval sequence;

[0029] Step S233: Calculate the entropy value of the time interval sequence to obtain the time dispersion of the fault event;

[0030] Step S234: Combine the frequency of occurrence and time dispersion of fault events to calculate the association feature value of each associated edge in the initial vehicle component-fault association topology graph;

[0031] Step S235: Update the association feature values ​​of all associated edges to the initial vehicle component-fault association topology graph.

[0032] This invention focuses on the temporal dimension of fault events, extracting fault occurrence time and related component information. It calculates the time interval between consecutive faults of the same component and generates a sequence. The temporal dispersion is calculated using entropy values ​​to quantify the uniformity of the fault occurrence time distribution—higher temporal dispersion indicates more random fault occurrence and poorer component stability. The edge association feature values ​​are updated by combining fault occurrence frequency and temporal dispersion, ensuring that the feature values ​​not only reflect association frequency but also the temporal pattern of faults and component stability. This overcomes the shortcomings of the initial topology graph, which only focuses on frequency and ignores temporal features. This step enriches the dimensions of edge association features, improves the comprehensiveness and rationality of quantifying component fault association strength, and allows the topology graph to more accurately reflect the actual fault state of the vehicle. It provides more realistic feature support for subsequent matrix processing and health assessment, further enhancing the objectivity and reliability of vehicle condition assessment and helping to accurately identify potential risks under complex vehicle conditions.

[0033] Preferably, step S233 includes the following steps:

[0034] The time interval sequence is normalized to obtain a normalized time interval sequence;

[0035] Probability distribution is calculated based on normalized time interval sequences;

[0036] Calculate the entropy value of the time interval sequence based on the probability distribution;

[0037] By performing a nonlinear mapping on the entropy value, the temporal discreteness of the fault event can be obtained;

[0038] The time dispersion is stored in the attribute field of the associated edge in the initial vehicle component-fault association topology graph.

[0039] This invention first normalizes the time interval sequence to eliminate interference from differences in the numerical scale of time intervals between different components and faults, ensuring the comparability of the sequence data. Based on the normalized sequence, a probability distribution is calculated to clearly present the distribution characteristics of the time intervals. Entropy values ​​are used to quantify the dispersion of the time intervals, accurately reflecting the randomness of fault occurrence. Finally, the entropy values ​​are nonlinearly mapped to obtain the time dispersion that meets the assessment requirements, and stored in the topological graph's associated edge attributes, achieving standardized storage and reuse of time features. This step standardizes the calculation process of time dispersion, ensuring the stability and rationality of the calculation results. It allows time features to be effectively integrated into component fault correlation analysis, compensating for the shortcomings of traditional assessments that ignore fault temporal features, further improving the accuracy of component fault correlation strength quantification, providing more accurate time dimension feature support for subsequent vehicle health assessment and risk level determination, reducing assessment bias, and ensuring the consistency of vehicle condition assessment results.

[0040] Preferably, step S3 includes the following steps:

[0041] Step S31: Extract spatial correlation features from the normalized component fault correlation strength matrix to generate a spatial feature matrix;

[0042] Step S32: Combine the fault timestamp information in the structured dataset of vehicle condition elements, extract time-series correlation features, and generate a time-series feature matrix;

[0043] Step S33: Perform element-wise fusion of the spatial feature matrix and the temporal feature matrix to obtain the spatiotemporal fusion feature matrix;

[0044] Step S34: Use principal component analysis to reduce the dimensionality of the spatiotemporal fusion feature matrix and remove redundant features;

[0045] Step S35: Normalize the reduced feature matrix and output the vehicle health vector corresponding to the fused spatiotemporal features.

[0046] This invention first extracts spatial correlation features from the correlation strength matrix to capture the spatial correlation patterns between components and faults; then, it combines fault timestamps to extract temporal correlation features, analyzing the changing trends of vehicle condition over time, achieving a two-way fusion of spatial and temporal features. Principal component analysis is used to eliminate redundant features, reducing data dimensionality and computational complexity while retaining core features. The dimensionality-reduced data is normalized to ensure the rationality and comparability of the feature data, ultimately outputting a vehicle health vector. This step achieves multi-dimensional fusion and optimization of vehicle condition features, comprehensively capturing the spatial correlation and temporal change patterns of vehicle faults, providing more comprehensive and representative feature support for subsequent risk assessment, improving the objectivity and reliability of vehicle condition assessment, overcoming the limitations of single-dimensional analysis in manual assessment, and further alleviating the pain point of information asymmetry in used car transactions.

[0047] Preferably, step S4 includes the following steps:

[0048] Step S41: Extract the health sub-vectors of each vehicle component and the global health features of the vehicle based on the vehicle health vector corresponding to the fused spatiotemporal features;

[0049] Step S42: Construct a multi-dimensional risk assessment function, and clarify the assessment dimensions and the calculation logic of each dimension;

[0050] Step S43: Input the component health subvector into the multi-dimensional risk assessment function to calculate the risk contribution characteristics of each component;

[0051] Step S44: Based on the risk contribution characteristics of each component, the risk percentage of each component is statistically calculated;

[0052] Step S45: Combine the vehicle's global health characteristics with the component risk percentage to output the vehicle condition risk level, and store the vehicle condition risk level and component risk percentage in the evaluation result dataset.

[0053] This invention extracts component health sub-vectors and global health features from the overall health vector to achieve hierarchical assessment of individual components and the vehicle's condition, breaking away from the limitations of traditional "one-size-fits-all" assessments. By constructing a multi-dimensional risk assessment function, the assessment dimensions and calculation logic are clearly defined, achieving standardization and normalization of risk assessment and avoiding the problem of inconsistent manual assessment standards. The component health sub-vectors are input into the assessment function to calculate the risk contribution characteristics of each component, quantifying the impact of a single component on the overall vehicle condition and statistically obtaining the component risk percentage. Combining the global health features and component risk percentages, the vehicle condition risk level is output, achieving dual quantification of risk level and component impact percentage. This step enables refined and standardized assessment of vehicle condition risk, allowing both parties in a transaction to clearly understand the overall vehicle risk level and the risk impact of each component, reducing transaction disputes caused by ambiguous risk perception, enhancing the credibility of vehicle condition assessments, and further promoting the standardized development of the used car market.

[0054] Preferably, step S5 includes the following steps:

[0055] Step S51: Based on the vehicle condition risk level and component risk ratio, extract the component health sub-vector and fault timestamp information from the structured dataset of vehicle condition elements to generate a radar chart dimensional feature set;

[0056] Step S52: Associate the corresponding fault events and maintenance records based on the fault timestamp information to generate a fault timeline node dataset;

[0057] Step S53: Combine the vehicle health vector with the vehicle's basic attribute data to construct a residual value change trend calculation model and generate residual value prediction curve data;

[0058] Step S54: Import the radar chart dimensional feature set, fault time axis node dataset, and residual value prediction curve data into the visualization rendering engine to generate a visualization report containing vehicle condition radar chart, fault time axis, and residual value prediction curve.

[0059] Step S55: Perform format adaptation processing on the visualization report and push it to the Web or mobile terminal.

[0060] This invention extracts relevant features based on risk level and component risk ratio to generate a radar chart feature set, intuitively presenting vehicle condition performance across various dimensions. It combines fault timestamps and maintenance records to generate a fault timeline, clearly outlining the vehicle's fault history and repair trajectory. Furthermore, it integrates health vectors and basic vehicle attributes to construct a residual value prediction model, generating a residual value prediction curve to provide a reference for transaction pricing. A visualization rendering engine generates a report containing radar charts, fault timelines, and residual value prediction curves, enriching the report's presentation and improving readability. The report is formatted and pushed to web or mobile devices for convenient querying and transmission of assessment results. This process visualizes, intuitively presents, and facilitates the assessment results, allowing both parties to quickly and clearly understand the vehicle's true condition, fault history, and residual value trends. This effectively alleviates information asymmetry, improves transaction efficiency, reduces disputes, and meets the comprehensive vehicle condition information needs of both parties in used car transactions.

[0061] Preferably, step S51 includes the following steps:

[0062] Step S511: Extract the risk assessment results of each assessment dimension from the vehicle condition risk level, and associate the component risk contribution characteristics in the component risk proportion;

[0063] Step S512: Based on the component risk contribution features, match the component health sub-vectors in the vehicle condition element structured dataset to obtain the corresponding health feature values ​​for each dimension;

[0064] Step S513: Generate the scale mapping rules for each dimension of the radar chart based on the distribution of health feature values;

[0065] Step S514: Associate the health feature values ​​with the scale mapping rules to generate a radar chart dimensional feature set;

[0066] Step S515: Redundant features are removed from the radar chart dimensional feature set, and core dimensional features are retained for visualization rendering.

[0067] This invention extracts risk assessment results from each evaluation dimension from the risk level, correlates component risk contribution characteristics, and achieves precise alignment between risk results and radar chart dimensions. It matches component health sub-vectors to obtain health feature values ​​for each dimension, ensuring the authenticity and relevance of the radar chart data. Based on the distribution of health feature values, it generates radar chart scale mapping rules to achieve a reasonable fit between the scale and feature values, avoiding visualization biases caused by unreasonable scale settings. By associating feature values ​​with scale rules, it generates a radar chart dimension feature set, eliminating redundant features and retaining core dimensions to ensure the radar chart is concise, clear, and representative. This step achieves standardized and refined generation of vehicle condition radar charts, enabling them to accurately and intuitively present vehicle condition performance across various dimensions. This helps both parties in a transaction quickly identify vehicle strengths and weaknesses, improves the readability and practicality of the assessment report, further alleviates information asymmetry, and provides an intuitive reference for transaction decisions.

[0068] Preferably, step S53 includes the following steps:

[0069] Step S531: Extract global health features and component degradation features from the vehicle health vector, and associate them with vehicle age and brand / model information in the vehicle's basic attribute data;

[0070] Step S532: Calculate the component degradation rate based on component degradation characteristics, and generate the vehicle condition health decay coefficient by combining global health characteristics;

[0071] Step S533: Based on the vehicle's age and brand / model information, match the market depreciation data for similar vehicles;

[0072] Step S534: Integrate the vehicle condition health degradation coefficient with market depreciation data to construct a residual value change trend calculation model;

[0073] Step S535: Generate residual value prediction curve data through the residual value change trend calculation model for displaying residual value trends in the visualization report.

[0074] This invention extracts global health features and component degradation features from the health vector, and associates them with basic attributes such as vehicle age, brand, and model. This achieves an organic combination of vehicle condition details and basic vehicle information, breaking the limitation of traditional residual value prediction that only focuses on basic attributes and ignores actual vehicle condition. By calculating the component degradation rate and the vehicle health decay coefficient, the deterioration pattern of vehicle condition is quantified. Matching market depreciation data of similar vehicles with the vehicle condition decay coefficient, a residual value prediction model is constructed, achieving a two-way adaptation between residual value prediction and actual vehicle condition and market patterns. The model generates residual value prediction curve data, clearly showing the trend of vehicle residual value changes over time, providing a scientific and objective reference for used car transaction pricing. This step realizes the scientific and refined nature of residual value prediction, avoids the subjectivity and blindness of manual pricing, reduces transaction disputes caused by unreasonable pricing, improves the fairness and efficiency of used car transactions, further improves the vehicle condition assessment system, and meets the core needs of both parties in the transaction for residual value reference.

[0075] It has the following beneficial effects:

[0076] (1) By comprehensively acquiring various types of vehicle condition data, such as maintenance records, actual vehicle inspection values, and vehicle operation logs, and adopting differentiated processing methods for different data characteristics, comprehensive analysis and extraction of unstructured text, numerical data, and time-series data are achieved. Through entity alignment and normalization, the scattered and disordered data are transformed into a structured dataset with a unified format, fully releasing the hidden information in the massive historical vehicle condition data and overcoming the limitations and subjectivity of manual data processing. This provides comprehensive and high-quality data support for subsequent topology map construction, feature extraction, and risk assessment, ensuring the comprehensiveness and objectivity of vehicle condition assessment, effectively alleviating the pain point of information asymmetry in used car transactions, and laying a solid foundation for the standardization and intelligentization of the entire assessment process.

[0077] (2) By clearly extracting vehicle components and fault event entities based on structured datasets, a vehicle component-fault association topology graph is constructed, intuitively presenting the correspondence between components and faults, breaking the limitations of traditional single component analysis in assessment. By extracting node degree and edge association features, the fault association frequency of each component and the degree of association between components and faults are quantified, generating a component-fault association strength matrix, realizing the computability and analyzability of fault association relationships. Spectral normalization of the matrix further ensures the rationality and comparability of the data, making the component fault association features more valuable for reference. This step captures the potential fault associations between vehicle components, improves the comprehensiveness of fault identification under complex vehicle conditions, provides core feature support for subsequent spatiotemporal feature fusion and health assessment, reduces assessment bias caused by missing component-related faults, and further improves the reliability of vehicle condition assessment.

[0078] (3) By extracting spatial correlation features and temporal correlation features respectively, a two-way fusion of the spatial distribution pattern of component failures and the temporal trend of vehicle condition changes is achieved, comprehensively capturing the dynamic and static features of vehicle failures and avoiding the limitations of single-dimensional feature analysis. Principal component analysis is used to eliminate redundant features, reduce data dimensionality and computational complexity, while retaining core features to ensure a balance between assessment efficiency and assessment quality. The normalization process after dimensionality reduction further standardizes the data format and outputs a vehicle condition health vector that integrates spatiotemporal features, transforming complex vehicle condition features into concise and representative quantitative indicators. This step achieves optimized integration of vehicle condition features, enabling the health vector to comprehensively and objectively reflect the actual condition of the vehicle, providing more targeted and representative feature support for subsequent risk assessment, and promoting the upgrading of vehicle condition assessment towards refinement and intelligence.

[0079] (4) By extracting component health sub-vectors and overall vehicle health features from the vehicle health vector, a hierarchical assessment of individual components and the overall vehicle condition is achieved, breaking the limitations of the traditional "one-size-fits-all" assessment. By constructing a multi-dimensional risk assessment function, the assessment dimensions and calculation logic are clarified, achieving standardization and normalization of risk assessment and avoiding inconsistencies in assessment results due to differences in the professional level of inspectors. By quantifying the risk contribution characteristics of each component, statistically analyzing the risk proportion of each component, and combining the overall health features, the vehicle condition risk level is output, making the impact of overall vehicle risk and component risk clearly identifiable. This step achieves a refined and standardized assessment of vehicle condition risk, providing clear risk references for both parties in the transaction, reducing transaction disputes caused by ambiguous risk perception, and enhancing the credibility of vehicle condition assessment.

[0080] (5) By integrating relevant vehicle condition data based on vehicle condition risk level and component risk ratio, various forms of visualization content such as vehicle condition radar chart, fault time axis, and residual value prediction curve are generated. These visually present the vehicle condition performance, fault history trajectory, and residual value change trend in various dimensions, enriching the report presentation format and improving report readability. The format adaptation and multi-terminal push of the visualization report enable convenient querying and transmission of assessment results, allowing both parties to the transaction to quickly and clearly grasp the true condition of the vehicle, fault history, and residual value trend, and understand the assessment results without professional knowledge. This effectively alleviates the core dilemma of information asymmetry in used car transactions, provides a scientific reference for transaction pricing, reduces transaction disputes caused by misperceptions of vehicle condition, improves transaction efficiency and market trust, and promotes the upgrading of used car transactions towards transparency and standardization. Attached Figure Description

[0081] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0082] Figure 1 This is a schematic diagram of the steps in the vehicle condition assessment method of the present invention;

[0083] Figure 2 for Figure 1 A detailed flowchart of step S1. Detailed Implementation

[0084] The present invention will be further described below with reference to the accompanying drawings and embodiments, but this should not be construed as limiting the present invention.

[0085] To achieve the above objectives, please refer to Figure 1 Embodiment 1 of the present invention provides a vehicle condition assessment method, including the following steps:

[0086] Step S1: Obtain multi-source heterogeneous vehicle condition data of used cars, perform entity alignment and normalization processing on the multi-source heterogeneous vehicle condition data of used cars, and obtain a structured dataset of vehicle condition elements in a unified format.

[0087] In this embodiment of the invention, multi-source data acquisition tools are used to comprehensively acquire multi-source heterogeneous vehicle condition data of used cars. The collected data covers three core types of data: maintenance text records, actual vehicle inspection values, and vehicle system operation logs. Maintenance text records include repair items, replaced parts, and fault descriptions. Actual vehicle inspection values ​​include performance parameters and wear levels of various components. Vehicle system operation logs include information such as the operating status of various vehicle systems, fault alarms, and runtime. After collection, text processing tools are used to extract entities from the maintenance text records, such as component names and fault types. Data preprocessing tools are used to filter outliers in the actual vehicle inspection values, removing invalid data caused by inspection errors. Log processing tools are used to align the timestamps of the vehicle system operation logs to ensure time consistency. An entity alignment tool is used to link and integrate relevant data of the same vehicle and the same component from different sources, using vehicle identification codes and component names as core identifiers. A normalization tool is used to standardize all data to eliminate differences in units and magnitudes. A data processing tool is used to integrate all processed data according to a preset format to obtain a structured dataset of vehicle condition elements in a unified format, providing standardized data support for subsequent evaluation operations.

[0088] Step S2: Construct a vehicle component-fault association topology graph based on the structured dataset of vehicle condition elements, extract the node degree and edge association features corresponding to the vehicle component-fault association topology graph, and generate a component fault association strength matrix;

[0089] In this embodiment of the invention, a data extraction tool is used to extract all vehicle component entities and fault event entities from a structured dataset of vehicle condition elements, clarifying the fault event corresponding to each component entity. A vehicle component-fault association topology graph is constructed using a topology graph construction tool, with vehicle components as nodes and the association between components and fault events as edges. A node degree calculation tool is used to count the number of fault events associated with each component node in the topology graph, calculating the node degree value of each node, which reflects the frequency of component-related faults. An edge feature extraction tool is used to extract the association features of each association edge in the topology graph, and combined with information such as the frequency of fault event occurrence and time interval, the association feature value of each association edge is calculated. A matrix generation tool is used to integrate the node degree values ​​of all component nodes and the association feature values ​​of each association edge, arranging them according to the correspondence between nodes and association edges to generate a component-fault association strength matrix. Each element in the matrix precisely corresponds to the relevant data on the association strength between components and fault events.

[0090] Step S3: Perform spatiotemporal feature fusion and dimensionality reduction on the component fault correlation strength matrix, and output the vehicle health vector corresponding to the fused spatiotemporal features;

[0091] In this embodiment of the invention, spatial correlation features between vehicle components are extracted from the component fault correlation strength matrix using a feature extraction tool. These features are then organized according to a preset structure to form a spatial feature matrix, clearly presenting the correlation patterns between components. Combining the fault timestamp information in the structured dataset of vehicle condition elements, a temporal feature extraction tool is used to extract the temporal correlation features of fault occurrence, which are then organized according to the time dimension to form a temporal feature matrix. A matrix fusion tool is used to perform element-level fusion of the spatial feature matrix and the temporal feature matrix, integrating corresponding elements according to pre-designed rules to obtain a spatiotemporal fusion feature matrix, achieving a comprehensive combination of spatial and temporal features. Principal component analysis is used to reduce the dimensionality of the spatiotemporal fusion feature matrix, identifying and eliminating redundant features and highly correlated repetitive features, while retaining core features. A normalization tool is used to standardize the dimensionality-reduced feature matrix, and a vector generation tool is used to convert the processed feature matrix into a vehicle condition health vector corresponding to the fused spatiotemporal features for subsequent risk assessment.

[0092] Step S4: Construct a multi-dimensional risk assessment function based on the vehicle health vector, and output the vehicle condition risk level and component risk ratio;

[0093] In this embodiment of the invention, a vector parsing tool is used to extract the health sub-vectors corresponding to each vehicle component and the global health feature of the vehicle from the vehicle health vector. The health sub-vectors correspond to the health status of individual components, and the global health feature reflects the overall health level of the vehicle. A function construction tool is used to build a multi-dimensional risk assessment function, clearly defining the assessment dimensions to cover component failure risk, overall degradation risk, etc., while also clarifying the calculation logic of each assessment dimension to ensure that the assessment process is quantifiable. A feature input tool is used to input the health sub-vectors of each component one by one into the function. The function operates according to a preset calculation logic and outputs the risk contribution feature of each component, reflecting the degree of influence of a single component on the overall risk of the vehicle. A statistical tool is used to calculate the sum of the risk contribution features of all components and calculate the proportion of each component's risk contribution feature to the sum, obtaining the component risk percentage. A risk level determination tool is used to combine the global health feature and the component risk percentage to determine and output the vehicle condition risk level according to a preset standard, clarifying the overall risk level of the vehicle.

[0094] Step S5: Generate a visual report based on the vehicle condition risk level and component risk ratio, including a vehicle condition radar chart, a failure timeline, and a residual value prediction curve, and push it to the Web or mobile device.

[0095] In this embodiment of the invention, a data extraction tool is used to extract component health sub-vectors and fault timestamp information from a structured dataset of vehicle condition elements based on vehicle condition risk level and component risk ratio. This data is then organized to generate a radar chart dimensional feature set for radar chart visualization. Based on the fault timestamp information, corresponding fault events and maintenance records are associated, and a fault timeline node dataset is generated in chronological order, with each node corresponding to a fault and maintenance detail. Combining the vehicle condition health vector and basic vehicle attribute data, a model building tool is used to construct a residual value change trend calculation model. The model generates residual value prediction curve data, presenting the trend of vehicle residual value changes over time. The three datasets are input into a visualization rendering engine to generate a visualization report containing a vehicle condition radar chart, a fault timeline, and a residual value prediction curve. The radar chart displays the health of each component, the timeline presents the fault repair process, and the curve presents residual value changes. A format adaptation tool is used to adjust the report format to meet the display requirements of web and mobile devices. The report is then pushed to the corresponding terminals for relevant personnel to view.

[0096] Furthermore, such as Figure 2 As shown, Figure 2 for Figure 1 A detailed flowchart of step S1 is shown below. In this embodiment, step S1 includes the following steps:

[0097] Step S11: Obtain multi-source heterogeneous vehicle condition data for used cars, including maintenance records, actual vehicle inspection values, and vehicle system operation logs;

[0098] Step S12: Perform entity extraction and relation extraction on the maintenance text records to obtain triplet data containing component name, fault type, and occurrence time;

[0099] Step S13: Filter outliers and normalize units on the actual vehicle inspection values ​​to obtain standardized inspection feature data;

[0100] Step S14: Perform timestamp alignment and event clustering on the vehicle system operation logs to obtain time-series operation event data;

[0101] Step S15: Align the triplet data, standardized detection feature data, and time-series running event data to generate a structured dataset of vehicle condition elements in a unified format.

[0102] In this embodiment of the invention, multi-source data acquisition tools are used to comprehensively acquire multi-source heterogeneous vehicle condition data of used cars. The acquired data explicitly includes three core data categories: maintenance text records, actual vehicle inspection values, and vehicle system operation logs. Maintenance text records cover past repair items, replaced parts, and fault descriptions. Actual vehicle inspection values ​​include performance parameters and wear levels of various components. Vehicle system operation logs cover the operating status of various systems during vehicle operation, fault alarms, and runtime, ensuring that the acquired data comprehensively covers the core elements required for vehicle condition assessment. After acquisition, for the maintenance text records, text processing tools are used for entity extraction and relation extraction. Entity recognition algorithms extract component names and fault type entities from the text, and relation extraction algorithms uncover the relationships between them, forming triplet data containing component name, fault type, and occurrence time. Each triplet data corresponds to a specific component fault record. For actual vehicle inspection data, data preprocessing tools are used for outlier filtering and unit normalization. Outlier detection algorithms identify and remove abnormal data caused by detection errors. Normalization algorithms convert inspection values ​​of different magnitudes and units into standardized values, obtaining standardized inspection feature data and eliminating the impact of unit and magnitude differences. For vehicle operation logs, log processing tools are used for timestamp alignment and event clustering. Time synchronization algorithms align all log data to a unified time standard to ensure log time consistency. Event clustering algorithms group similar operational events and fault events together to obtain time-series operational event data, clearly presenting the temporal distribution patterns of various events during vehicle operation. Finally, entity alignment tools are used to align triplet data, standardized inspection feature data, and time-series operational event data. Using vehicle identification codes and component names as core alignment identifiers, related data from different sources on the same vehicle and the same component are linked and integrated, forming a unified formatted structured dataset of vehicle condition elements. This provides standardized and structured data support for subsequent vehicle condition assessment operations.

[0103] Furthermore, step S2 includes the following steps:

[0104] Step S21: Extract vehicle component entities and fault event entities based on the structured dataset of vehicle condition elements, and construct an initial vehicle component-fault association topology map;

[0105] Step S22: Count the number of fault events associated with each component node in the initial vehicle component-fault association topology graph, and calculate the node degree value;

[0106] Step S23: Based on the occurrence frequency and time interval of each fault event in the structured dataset of vehicle condition elements, calculate the association feature value of each association edge in the initial vehicle component-fault association topology graph;

[0107] Step S24: Integrate the node degree value and the edge association feature value to generate the component fault association strength matrix;

[0108] Step S25: Perform spectral normalization on the component fault correlation strength matrix to obtain the normalized component fault correlation strength matrix.

[0109] In this embodiment of the invention, a data extraction tool is used to extract all vehicle component entities and fault event entities from a structured dataset of vehicle condition elements. The fault event entity corresponding to each component entity is identified. An initial vehicle component-fault association topology graph is constructed using a topology graph construction tool. Vehicle components are used as nodes in the topology graph, and the association between components and fault events is used as the association edges. Each node corresponds to a specific vehicle component, and each association edge corresponds to the association between a component and a fault event, clearly presenting the association logic between components and fault events. After the topology graph is constructed, a node degree value calculation tool is used to count the number of fault events associated with each component node in the initial vehicle component-fault association topology graph. Based on the statistical results, the node degree value of each component node is calculated. The node degree value directly reflects the frequency of fault events associated with the corresponding component; the higher the node degree value, the more times the component fails and the higher the probability of failure. Subsequently, an eigenvalue calculation tool was used to calculate the association eigenvalues ​​for each association edge in the initial vehicle component-fault association topology graph based on the occurrence frequency and time interval of each fault event in the structured dataset of vehicle condition elements. The occurrence frequency reflects how often the fault event occurs on the corresponding component, and the time interval reflects the density of the fault event. Combining the association eigenvalues ​​calculated by both methods accurately reflects the association strength of each association edge. A matrix generation tool was used to integrate the node degree values ​​of all component nodes with the association eigenvalues ​​of each association edge, constructing a component fault association strength matrix according to the correspondence between component nodes and association edges. Each element in the matrix corresponds to the association strength data between a component node and an association edge. Finally, a spectral normalization tool was used to perform spectral normalization on the component fault association strength matrix. The normalization algorithm eliminated the differences in data magnitude in the matrix, resulting in a normalized component fault association strength matrix, ensuring that the matrix data can be directly used for subsequent vehicle condition assessment calculations.

[0110] Furthermore, step S23 includes the following steps:

[0111] Step S231: Extract the occurrence time and associated component information of each fault event from the structured dataset of vehicle condition elements;

[0112] Step S232: Calculate the time interval between consecutive failure events associated with the same component, and generate a time interval sequence;

[0113] Step S233: Calculate the entropy value of the time interval sequence to obtain the time dispersion of the fault event;

[0114] Step S234: Combine the frequency of occurrence and time dispersion of fault events to calculate the association feature value of each associated edge in the initial vehicle component-fault association topology graph;

[0115] Step S235: Update the association feature values ​​of all associated edges to the initial vehicle component-fault association topology graph.

[0116] In this embodiment of the invention, a data extraction tool is used to extract the occurrence time and associated component information of each fault event from a structured dataset of vehicle condition elements. The extracted information is then categorized and grouped according to associated components, and all fault events corresponding to the same component are grouped into the same data group. This ensures centralized management of fault event information for each component and avoids data confusion between different components. After grouping, a time interval calculation tool is used to calculate the time interval of consecutive fault events associated with the same component for each component group. This time interval is the difference between the occurrence time of the later fault event and the occurrence time of the previous fault event. The time intervals of all consecutive fault events are arranged in order of occurrence, generating a time interval sequence for each component. The time interval sequence clearly presents the time distribution characteristics of fault events occurring within the same component. An entropy calculation tool is used to calculate the entropy value of the time interval sequence for each component. The entropy value reflects the dispersion of the time interval sequence. A larger entropy value indicates a more dispersed time distribution of fault events occurring within the same component, while a smaller entropy value indicates a more concentrated time distribution of fault events occurring within the same component. The calculated entropy value is the time dispersion of the fault events. Subsequently, an association feature value update tool was used to recalculate the association feature values ​​of the corresponding edges in the initial vehicle component-fault association topology graph, combining the occurrence frequency of each fault event with the calculated time dispersion. Occurrence frequency reflects the frequency of fault events, while time dispersion reflects the temporal distribution characteristics of fault events. Combining these two factors provides a more accurate representation of the association strength between components and fault events. After the calculations were completed, the topology graph update tool was used to update the attributes of the corresponding edges in the initial vehicle component-fault association topology graph with the recalculated association feature values, replacing the original association feature values. This ensures that the association feature values ​​of the edges in the topology graph are accurate and comprehensive, providing more reliable support for subsequent vehicle condition assessments.

[0117] Furthermore, step S233 includes the following steps:

[0118] The time interval sequence is normalized to obtain a normalized time interval sequence;

[0119] Probability distribution is calculated based on normalized time interval sequences;

[0120] Calculate the entropy value of the time interval sequence based on the probability distribution;

[0121] By performing a nonlinear mapping on the entropy value, the temporal discreteness of the fault event can be obtained;

[0122] The time dispersion is stored in the attribute field of the associated edge in the initial vehicle component-fault association topology graph.

[0123] In this embodiment of the invention, a normalization tool is used to normalize the time interval sequence corresponding to each component. The normalization algorithm converts all values ​​in the time interval sequence into a unified standard range, eliminating the magnitude differences between time interval sequences of different components, resulting in a normalized time interval sequence and ensuring the accuracy of subsequent probability distribution calculations. After normalization, a probability distribution calculation tool is used to calculate the probability distribution based on the normalized time interval sequence. By counting the number of occurrences of each value in the normalized time interval sequence and combining it with the total sequence length, the probability of occurrence of each value is calculated, forming the probability distribution of the normalized time interval sequence, clearly presenting the probability of each time interval occurring. Based on the calculated probability distribution, an entropy calculation tool is used to calculate the entropy value of the time interval sequence. The entropy calculation process strictly relies on each probability value in the probability distribution. Through the entropy calculation formula, all probability values ​​are included in the calculation to obtain an entropy value that accurately reflects the dispersion of the time interval sequence. A nonlinear mapping tool is used to perform nonlinear mapping processing on the calculated entropy value. Through a preset nonlinear mapping algorithm, the entropy value is converted into a numerical range that meets the needs of vehicle condition assessment, obtaining the time dispersion of fault events. The time dispersion can be directly used to measure the temporal dispersion of fault events occurring for the same component. Finally, data storage tools are used to store the calculated time dispersion values ​​one by one into the attribute fields of the corresponding associated edges in the initial vehicle component-fault association topology graph. The attribute fields of each associated edge store the time dispersion values ​​of its associated fault events, ensuring the completeness of the attribute information of each associated edge in the topology graph. At the same time, the relevant information in the structured dataset of vehicle condition elements is updated to achieve synchronous data updates, providing accurate data support for subsequent component fault association strength analysis and comprehensive vehicle condition assessment.

[0124] Furthermore, step S3 includes the following steps:

[0125] Step S31: Extract spatial correlation features from the normalized component fault correlation strength matrix to generate a spatial feature matrix;

[0126] Step S32: Combine the fault timestamp information in the structured dataset of vehicle condition elements, extract time-series correlation features, and generate a time-series feature matrix;

[0127] Step S33: Perform element-wise fusion of the spatial feature matrix and the temporal feature matrix to obtain the spatiotemporal fusion feature matrix;

[0128] Step S34: Use principal component analysis to reduce the dimensionality of the spatiotemporal fusion feature matrix and remove redundant features;

[0129] Step S35: Normalize the reduced feature matrix and output the vehicle health vector corresponding to the fused spatiotemporal features.

[0130] In this embodiment of the invention, a feature extraction tool is used to extract spatial correlation features between all vehicle components from the normalized component fault correlation strength matrix. These spatial correlation features encompass the degree of correlation between component nodes and fault propagation correlation features. A matrix organization tool arranges the extracted spatial correlation features according to a preset structure to generate a spatial feature matrix, clearly presenting the spatial correlation patterns between components. A temporal feature extraction tool is used in conjunction with fault timestamp information from the structured dataset of vehicle condition elements to extract temporal correlation features of fault events. These temporal correlation features include the time sequence, interval patterns, and temporal correlation of fault occurrences, and are organized along the time dimension to form a temporal feature matrix. A matrix fusion tool is used to perform element-level fusion of the spatial feature matrix and the temporal feature matrix, integrating corresponding elements from the two matrices according to preset rules to obtain a spatiotemporal fusion feature matrix, achieving a comprehensive combination of spatial and temporal features. Principal component analysis is used to reduce the dimensionality of the spatiotemporal fusion feature matrix. A feature filtering algorithm identifies and removes redundant features and highly correlated repetitive features from the matrix, retaining the core features that accurately reflect the vehicle condition and simplifying the complexity of the feature matrix. The reduced feature matrix is ​​standardized using a normalization tool to eliminate differences in feature magnitude. The processed feature matrix is ​​then converted into a vehicle health vector corresponding to the fused spatiotemporal features using a vector generation tool. Each element in the vector corresponds to a core feature related to vehicle condition assessment, which is used for subsequent vehicle condition risk assessment.

[0131] Furthermore, step S4 includes the following steps:

[0132] Step S41: Extract the health sub-vectors of each vehicle component and the global health features of the vehicle based on the vehicle health vector corresponding to the fused spatiotemporal features;

[0133] Step S42: Construct a multi-dimensional risk assessment function, and clarify the assessment dimensions and the calculation logic of each dimension;

[0134] Step S43: Input the component health subvector into the multi-dimensional risk assessment function to calculate the risk contribution characteristics of each component;

[0135] Step S44: Based on the risk contribution characteristics of each component, the risk percentage of each component is statistically calculated;

[0136] Step S45: Combine the vehicle's global health characteristics with the component risk percentage to output the vehicle condition risk level, and store the vehicle condition risk level and component risk percentage in the evaluation result dataset.

[0137] In this embodiment of the invention, a vector parsing tool is used to extract the health sub-vectors corresponding to each vehicle component and the global vehicle health feature from the vehicle health vector corresponding to the fused spatiotemporal features. The health sub-vectors correspond to the health status features of individual components, while the global health feature reflects the overall health level of the vehicle. A function construction tool is used to build a multi-dimensional risk assessment function, clarifying the core dimensions of risk assessment, covering component failure risk, overall degradation risk, and failure propagation risk, while also clarifying the specific calculation logic for each assessment dimension to ensure that the assessment process is quantifiable and traceable. A feature input tool is used to input the health sub-vectors of each component one by one into the multi-dimensional risk assessment function. The function performs calculations on the health sub-vectors according to the preset calculation logic and outputs the risk contribution features of each component. The risk contribution features reflect the degree of influence of a single component on the overall risk of the vehicle. A statistical tool is used to calculate the sum of the risk contribution features of all components based on the risk contribution features of each component, and calculate the proportion of the risk contribution features of each component to the sum to obtain the component risk proportion. A risk level determination tool is used to combine the global vehicle health features and the component risk proportions to determine and output the vehicle condition risk level according to the preset risk level standard, clarifying the overall risk level of the vehicle. The vehicle condition risk level is associated with the risk proportion of each component and stored in the assessment result dataset through data storage tools. The corresponding vehicle identification code, assessment time and other information are also linked to ensure that the assessment results are traceable and searchable.

[0138] Furthermore, step S5 includes the following steps:

[0139] Step S51: Based on the vehicle condition risk level and component risk ratio, extract the component health sub-vector and fault timestamp information from the structured dataset of vehicle condition elements to generate a radar chart dimensional feature set;

[0140] Step S52: Associate the corresponding fault events and maintenance records based on the fault timestamp information to generate a fault timeline node dataset;

[0141] Step S53: Combine the vehicle health vector with the vehicle's basic attribute data to construct a residual value change trend calculation model and generate residual value prediction curve data;

[0142] Step S54: Import the radar chart dimensional feature set, fault time axis node dataset, and residual value prediction curve data into the visualization rendering engine to generate a visualization report containing vehicle condition radar chart, fault time axis, and residual value prediction curve.

[0143] Step S55: Perform format adaptation processing on the visualization report and push it to the Web or mobile terminal.

[0144] In this embodiment of the invention, a data extraction tool is used to extract corresponding component health sub-vectors and fault timestamp information from the structured dataset of vehicle condition elements based on vehicle condition risk level and the risk proportion of each component. A feature processing tool is then used to classify the extracted information according to radar chart visualization requirements, generating a radar chart dimensional feature set and clarifying the feature data corresponding to each dimension of the radar chart. An association tool is used to associate the fault events and maintenance records in the structured dataset of vehicle condition elements based on the fault timestamp information. The specific content, maintenance method, and maintenance time of each fault event are organized chronologically to generate a fault timeline node dataset, with each node corresponding to one fault and maintenance record. A model building tool is used to combine the vehicle condition health vector and basic vehicle attribute data to build a residual value change trend calculation model. The model incorporates core factors such as the decay law of vehicle condition health and market depreciation characteristics. Through model calculation, residual value prediction curve data is generated, clearly presenting the trend of vehicle residual value changes over time. A visualization rendering tool is used to input the radar chart dimensional feature set, fault timeline node dataset, and residual value prediction curve data into the visualization rendering engine. The engine generates a visualization report containing a vehicle condition radar chart, fault timeline, and residual value prediction curve according to preset visualization rules. The radar chart presents the health status of each component, the timeline shows the fault repair process, and the curve shows the residual value changes. A format adaptation tool is used to adapt the visualization report to the desired format, adjusting the report resolution and layout to meet the display requirements of web and mobile devices. The adapted visualization report is then pushed to web or mobile devices via a push notification tool for relevant personnel to view.

[0145] Furthermore, step S51 includes the following steps:

[0146] Step S511: Extract the risk assessment results of each assessment dimension from the vehicle condition risk level, and associate the component risk contribution characteristics in the component risk proportion;

[0147] Step S512: Based on the component risk contribution features, match the component health sub-vectors in the vehicle condition element structured dataset to obtain the corresponding health feature values ​​for each dimension;

[0148] Step S513: Generate the scale mapping rules for each dimension of the radar chart based on the distribution of health feature values;

[0149] Step S514: Associate the health feature values ​​with the scale mapping rules to generate a radar chart dimensional feature set;

[0150] Step S515: Redundant features are removed from the radar chart dimensional feature set, and core dimensional features are retained for visualization rendering.

[0151] In this embodiment of the invention, a risk extraction tool is used to extract the risk judgment results corresponding to each assessment dimension from the vehicle condition risk level, clarifying the risk status of each assessment dimension. An association tool is used to associate the risk judgment results with the component risk contribution characteristics in the component risk proportion, establishing a correspondence between risk judgment and component risk contribution. A feature matching tool is used to match the component health sub-vectors in the vehicle condition element structured dataset based on the component risk contribution characteristics, extracting the health feature values ​​corresponding to each assessment dimension from the sub-vectors to ensure accurate correspondence between health feature values ​​and assessment dimensions. A scale mapping tool is used to set the scale range and scale interval for each dimension of the radar chart based on the distribution of health feature values ​​for each assessment dimension, generating scale mapping rules for each dimension of the radar chart to ensure accurate mapping of health feature values ​​to the radar chart scale. An association tool is used to associate the health feature values ​​of each assessment dimension with the corresponding scale mapping rules, categorizing and organizing them according to radar chart dimensions to generate a radar chart dimension feature set, clarifying the feature data and display specifications for each dimension. A redundancy removal tool is used to remove redundant features from the radar chart dimensional feature set. This identifies and removes feature data that is irrelevant to radar chart visualization and duplicates, while retaining the core dimensional features that can reflect the core health status of each dimension for subsequent visualization rendering, ensuring that the radar chart is clear and accurate.

[0152] Furthermore, step S53 includes the following steps:

[0153] Step S531: Extract global health features and component degradation features from the vehicle health vector, and associate them with vehicle age and brand / model information in the vehicle's basic attribute data;

[0154] Step S532: Calculate the component degradation rate based on component degradation characteristics, and generate the vehicle condition health decay coefficient by combining global health characteristics;

[0155] Step S533: Based on the vehicle's age and brand / model information, match the market depreciation data for similar vehicles;

[0156] Step S534: Integrate the vehicle condition health degradation coefficient with market depreciation data to construct a residual value change trend calculation model;

[0157] Step S535: Generate residual value prediction curve data through the residual value change trend calculation model for displaying residual value trends in the visualization report.

[0158] In this embodiment of the invention, a vector extraction tool is used to extract the overall vehicle health features and component degradation features from the vehicle health vector. The component degradation features reflect the wear and tear of components over time. An association tool is used to link the extracted features with the vehicle's age, brand, and model information from the vehicle's basic attribute data, integrating the vehicle condition and basic vehicle information. A rate calculation tool is used to calculate the degradation rate of each component based on the component degradation features, clarifying the degree of component degradation. Combined with the overall vehicle health features, a coefficient generation tool generates a vehicle health degradation coefficient, reflecting the degradation pattern of the overall vehicle health level. A data matching tool is used to retrieve market depreciation data for similar vehicles based on the vehicle's age and brand / model information, clarifying the depreciation ratio and residual value change patterns of similar vehicles at different ages. A model fusion tool is used to fuse the vehicle health degradation coefficient with the market depreciation data for similar vehicles, supplementing the core parameters of the model, building a residual value change trend calculation model, and clarifying the model's operational logic and parameter settings. The residual value change trend calculation model is started by using the model calculation tool. Vehicle-related feature data is input, and the model generates residual value prediction curve data according to the preset logic. The curve data contains the vehicle residual value at different time points, which is used to display the residual value trend in the visualization report and clearly present the future changes of vehicle residual value.

[0159] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0160] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A vehicle condition assessment method, characterized in that, Includes the following steps: Step S1: Obtain multi-source heterogeneous vehicle condition data of used cars, perform entity alignment and normalization processing on the multi-source heterogeneous vehicle condition data of used cars, and obtain a structured dataset of vehicle condition elements in a unified format. Step S2: Construct a vehicle component-fault association topology graph based on the structured dataset of vehicle condition elements, extract the node degree and edge association features corresponding to the vehicle component-fault association topology graph, and generate a component fault association strength matrix; Step S3: Perform spatiotemporal feature fusion and dimensionality reduction on the component fault correlation strength matrix, and output the vehicle health vector corresponding to the fused spatiotemporal features; Step S4: Construct a multi-dimensional risk assessment function based on the vehicle health vector, and output the vehicle condition risk level and component risk ratio; Step S5: Generate a visual report based on the vehicle condition risk level and component risk ratio, including a vehicle condition radar chart, a failure timeline, and a residual value prediction curve, and push it to the Web or mobile device.

2. The vehicle condition assessment method according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Obtain multi-source heterogeneous vehicle condition data for used cars, including maintenance records, actual vehicle inspection values, and vehicle system operation logs; Step S12: Perform entity extraction and relation extraction on the maintenance text records to obtain triplet data containing component name, fault type, and occurrence time; Step S13: Filter outliers and normalize units on the actual vehicle inspection values ​​to obtain standardized inspection feature data; Step S14: Perform timestamp alignment and event clustering on the vehicle system operation logs to obtain time-series operation event data; Step S15: Align the triplet data, standardized detection feature data, and time-series running event data to generate a structured dataset of vehicle condition elements in a unified format.

3. The vehicle condition assessment method according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Extract vehicle component entities and fault event entities based on the structured dataset of vehicle condition elements, and construct an initial vehicle component-fault association topology map; Step S22: Count the number of fault events associated with each component node in the initial vehicle component-fault association topology graph, and calculate the node degree value; Step S23: Based on the occurrence frequency and time interval of each fault event in the structured dataset of vehicle condition elements, calculate the association feature value of each association edge in the initial vehicle component-fault association topology graph; Step S24: Integrate the node degree value and the edge association feature value to generate the component fault association strength matrix; Step S25: Perform spectral normalization on the component fault correlation strength matrix to obtain the normalized component fault correlation strength matrix.

4. The vehicle condition assessment method according to claim 3, characterized in that, Step S23 includes the following steps: Step S231: Extract the occurrence time and associated component information of each fault event from the structured dataset of vehicle condition elements; Step S232: Calculate the time interval between consecutive failure events associated with the same component, and generate a time interval sequence; Step S233: Calculate the entropy value of the time interval sequence to obtain the time dispersion of the fault event; Step S234: Combine the frequency of occurrence and time dispersion of fault events to calculate the association feature value of each associated edge in the initial vehicle component-fault association topology graph; Step S235: Update the association feature values ​​of all associated edges to the initial vehicle component-fault association topology graph.

5. The vehicle condition assessment method according to claim 4, characterized in that, Step S233 includes the following steps: The time interval sequence is normalized to obtain a normalized time interval sequence; Probability distribution is calculated based on normalized time interval sequences; Calculate the entropy value of the time interval sequence based on the probability distribution; By performing a nonlinear mapping on the entropy value, the temporal discreteness of the fault event can be obtained; The time dispersion is stored in the attribute field of the associated edge in the initial vehicle component-fault association topology graph.

6. The vehicle condition assessment method according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Extract spatial correlation features from the normalized component fault correlation strength matrix to generate a spatial feature matrix; Step S32: Combine the fault timestamp information in the structured dataset of vehicle condition elements, extract time-series correlation features, and generate a time-series feature matrix; Step S33: Perform element-wise fusion of the spatial feature matrix and the temporal feature matrix to obtain the spatiotemporal fusion feature matrix; Step S34: Use principal component analysis to reduce the dimensionality of the spatiotemporal fusion feature matrix and remove redundant features; Step S35: Normalize the reduced feature matrix and output the vehicle health vector corresponding to the fused spatiotemporal features.

7. The vehicle condition assessment method according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Extract the health sub-vectors of each vehicle component and the global health features of the vehicle based on the vehicle health vector corresponding to the fused spatiotemporal features; Step S42: Construct a multi-dimensional risk assessment function, and clarify the assessment dimensions and the calculation logic of each dimension; Step S43: Input the component health sub-vector into the multi-dimensional risk assessment function to calculate the risk contribution characteristics of each component; Step S44: Based on the risk contribution characteristics of each component, the risk percentage of each component is statistically calculated; Step S45: Combine the vehicle's global health characteristics with the component risk percentage to output the vehicle condition risk level, and store the vehicle condition risk level and component risk percentage in the evaluation result dataset.

8. The vehicle condition assessment method according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: Based on the vehicle condition risk level and component risk ratio, extract the component health sub-vector and fault timestamp information from the structured dataset of vehicle condition elements to generate a radar chart dimensional feature set; Step S52: Associate the corresponding fault events and maintenance records based on the fault timestamp information to generate a fault timeline node dataset; Step S53: Combine the vehicle health vector with the vehicle's basic attribute data to construct a residual value change trend calculation model and generate residual value prediction curve data; Step S54: Import the radar chart dimensional feature set, fault time axis node dataset, and residual value prediction curve data into the visualization rendering engine to generate a visualization report containing vehicle condition radar chart, fault time axis, and residual value prediction curve. Step S55: Perform format adaptation processing on the visualization report and push it to the Web or mobile terminal.

9. The vehicle condition assessment method according to claim 8, characterized in that, Step S51 includes the following steps: Step S511: Extract the risk assessment results of each assessment dimension from the vehicle condition risk level, and associate the component risk contribution characteristics in the component risk proportion; Step S512: Based on the component risk contribution features, match the component health sub-vectors in the vehicle condition element structured dataset to obtain the corresponding health feature values ​​for each dimension; Step S513: Generate the scale mapping rules for each dimension of the radar chart based on the distribution of health feature values; Step S514: Associate the health feature values ​​with the scale mapping rules to generate a radar chart dimensional feature set; Step S515: Redundant features are removed from the radar chart dimensional feature set, and core dimensional features are retained for visualization rendering.

10. The vehicle condition assessment method according to claim 8, characterized in that, Step S53 includes the following steps: Step S531: Extract global health features and component degradation features from the vehicle health vector, and associate them with vehicle age and brand / model information in the vehicle's basic attribute data; Step S532: Calculate the component degradation rate based on component degradation characteristics, and generate the vehicle condition health degradation coefficient by combining global health characteristics; Step S533: Based on the vehicle's age and brand / model information, match the market depreciation data for similar vehicles; Step S534: Integrate the vehicle condition health degradation coefficient with market depreciation data to construct a residual value change trend calculation model; Step S535: Generate residual value prediction curve data through the residual value change trend calculation model for displaying residual value trends in the visualization report.

Images