Differential intelligent traceability test comparative analysis method, device and medium

Abstract

This invention relates to a method, equipment, and medium for differentiated intelligent traceability of experimental comparative analysis, belonging to the field of automotive testing technology. Addressing the pain points of existing test data, such as complex formats, difficulties in template parsing, and the inability to intelligently trace differences, this invention first collects and structures test data from multiple platforms, vehicle models, and stages; then, it uses an adaptive parsing algorithm to locate header segmentation points and constructs a multi-branch tree of horizontal and vertical headers, enabling automatic parsing and incremental updates of test templates in different formats; subsequently, it supports multi-mode comparisons, including single-indicator, multi-indicator, and time-series curve comparisons, presenting differences visually and extracting key features; finally, it constructs a correlation model based on a Bayesian network, automatically locating the root causes of test differences and outputting the traceability results through difference feature extraction, influencing factor screening, and multi-dimensional correlation analysis. This invention is compatible with multiple templates, improves comparison efficiency, and achieves intelligent traceability of differences, providing efficient and reliable support for automotive test data analysis and R&D optimization.

Description

Technical Field

[0001] This invention relates to the field of automotive testing technology, and specifically to a comparative analysis method for differentiated intelligent traceability testing. Background Technology

[0002] With the rapid development and intensified competition in the automotive industry, automakers and R&D institutions face the need to process and analyze a large amount of test data. However, automotive test data management suffers from problems such as scattered and easily lost manual management, large and complex data volume, inconsistent data types and formats, cumbersome manual comparison operations, and a lack of data validity. In the field of automotive R&D, comparative test analysis is a core means of verifying the feasibility of solutions, optimizing product performance, and identifying test anomalies. A large amount of test data is generated in various test scenarios, and the template formats, data dimensions, and indicator definitions corresponding to different test types vary significantly.

[0003] Existing experimental comparative analysis methods suffer from numerous technical shortcomings, making it difficult to meet the needs for efficient and accurate experimental comparison and difference identification. Firstly, existing methods mostly only support the import of experimental data in fixed formats, failing to automatically parse experimental templates of different types and formats. Secondly, the comparative analysis of existing methods is often limited to single-index numerical comparisons or simple curve comparisons, unable to achieve in-depth fusion comparisons of multiple parameters, resulting in incomplete and inaccurate comparison results, making it difficult to uncover the underlying correlations in the experimental data. Thirdly, and most importantly, existing methods can only identify differences between experimental data, unable to intelligently trace the root causes of these differences. When significant differences are found in experimental results, experimenters must rely on their experience to systematically investigate multiple dimensions such as experimental conditions, sample parameters, and experimental procedures, resulting in long investigation cycles and low accuracy in tracing the root causes. This is especially problematic in complex experimental scenarios with multiple coupled factors, making it difficult to quickly locate the root causes of differences, severely impacting experimental progress and optimization efficiency.

[0004] In view of the shortcomings of the existing technologies, there is an urgent need for a test comparison analysis method that can automatically analyze automotive test templates, perform multi-dimensional in-depth comparisons, and intelligently trace the source of test differences, so as to solve the technical pain points of existing methods such as cumbersome operation, low comparison accuracy, and difficulty in tracing the source of differences. Summary of the Invention

[0005] The present invention proposes a comparative analysis method for differentiated intelligent traceability tests, specifically a visual comparative analysis method for parsing key parameters in automotive test forms and supporting the traceability of the reasons for differences, which can at least solve one of the technical problems in the background art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A comparative analysis method for differentiated intelligent traceability involves performing the following steps using computer equipment: S1. Collect test data from multiple platforms, projects, models, and stages reported by engineers, convert the test data into structured information, and complete the structured data entry into the database to provide a data foundation for subsequent analysis and comparison. S2. Filter invalid cells by calculating the confidence level of valid information, locate the horizontal and vertical header dividing points, construct a multi-level header multi-branch tree structure, associate and store the header with the experimental data, and support incremental updates of templates to adapt to template changes, so as to realize automatic parsing and data structured management of experimental templates of different formats. S3. Select single-index, multi-index, time-series curve, or image feature comparison mode to quantitatively compare structured experimental data, display the results in a visual form, automatically extract key features, and label significant difference indicators. S4. Extract the difference feature vector based on the significant difference index, screen four types of potential influencing factors: experimental conditions, equipment, samples, and operations, construct a correlation model based on the Bayesian network algorithm to conduct multi-dimensional correlation analysis, locate key influencing factors, and output the source tracing results including root causes and verification basis. S5 integrates template parsing, data comparison, and difference tracing results throughout the entire process to form an integrated analysis conclusion, which is used for automotive product performance evaluation, design optimization, and R&D decision-making.

[0007] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0008] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0009] As described above, this invention provides a differentiated intelligent traceability method for comparative analysis of test data, addressing the problems of difficult parsing of automotive test data and difficulty in quickly locating the causes of differences. This method addresses the pain points of existing test data, such as complex formats, difficult template parsing, low comparison accuracy, and inability to intelligently trace differences. First, it collects and structures test data from multiple platforms, vehicle models, and stages. Then, it uses an adaptive parsing algorithm to locate header segmentation points and constructs a multi-branch tree of horizontal and vertical headers, enabling automatic parsing and incremental updates of test templates in different formats. Subsequently, it supports multi-mode comparisons, including single-indicator, multi-indicator, and time-series curve comparisons, presenting differences visually and extracting key features. Finally, it constructs a correlation model based on a Bayesian network, automatically locating the root causes of test differences and outputting the traceability results through difference feature extraction, influencing factor screening, and multi-dimensional correlation analysis. This invention is compatible with multiple templates, improves comparison efficiency, and achieves intelligent traceability of differences, providing efficient and reliable support for automotive test data analysis and R&D optimization.

[0010] Compared with existing experimental comparative analysis methods, the present invention has the following significant advantages: 1. It solves the problem of multiple test templates for various test types. Through an adaptive parsing algorithm, it eliminates the need for manual template formatting and data conversion, significantly reducing manual operation costs, minimizing human error, and improving test data processing efficiency. It also supports adaptive template updates, adapting to various test scenarios and demonstrating strong versatility. This helps users intuitively analyze and compare data results, efficiently evaluate product performance under different conditions, and identify differences and potential problems between products, thereby providing decision-making support for product improvement and R&D.

[0011] 2. It achieves deep integration and precise comparison of experimental data, breaking the limitations of single data format and single indicator comparison. By parsing complex forms, it centrally and uniformly manages experimental data, outputs comprehensive and quantitative comparison results, accurately identifies significant difference indicators, provides reliable data support for experimental analysis, and improves the scientificity and accuracy of experimental comparison.

[0012] 3. The core innovation lies in the application of the intelligent source tracing algorithm for experimental differences. It breaks through the limitation of existing methods that can only identify differences but cannot trace the root cause. Through the fully automated source tracing process of "difference feature extraction - influencing factor screening - multi-dimensional correlation analysis - root cause location", it can quickly locate the core root cause of differences, significantly shorten the difference investigation cycle, reduce the reliance on the experience of experimental personnel, and improve the efficiency and quality of experimental optimization.

[0013] 4. An integrated experimental comparison and analysis system has been built, which integrates the entire process functions such as data acquisition, template parsing, comparative analysis, and difference tracing. It is easy to operate, has a smooth process, and can quickly shorten the time for data search and analysis, thereby improving the overall efficiency of experimental comparison and analysis.

[0014] 5. By using methods such as data analysis and intelligent traceability, the utilization rate of test data and the level of intelligence in test analysis are improved, the number of invalid tests is reduced, and test costs are lowered; this helps automakers efficiently consolidate their testing capabilities and provides strong support for product development scheme optimization and product performance improvement. Attached Figure Description

[0015] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0017] The purpose of this invention is to integrate and process data to visually display the differences and trends in performance parameters through comparative lists and curves. This helps automakers and R&D engineers better understand and compare performance parameters from different test data. By comparing performance parameters from different test data, the performance of different models in various aspects, such as fuel efficiency, acceleration, braking performance, noise and vibration, can be evaluated, which helps identify performance differences and improve product design. Furthermore, this invention overcomes the shortcomings of existing test comparison analysis methods, such as the inability to identify multiple test templates, insufficient comparison accuracy, and difficulty in tracing test differences. It provides a test comparison analysis method for automotive test template parsing and differentiated intelligent tracing, with the specific objectives as follows: 1. Enables automatic identification and parsing of different test templates for different test types, eliminating the need for manual template formatting and data standardization, reducing manual operation costs, minimizing human error, and improving test data processing efficiency; 2. Achieve deep integration and precise comparison of experimental data, breaking the limitations of single data format and single indicator comparison, and outputting comprehensive and quantitative comparison results; 3. Develop an intelligent algorithm for tracing the source of experimental differences to achieve automatic identification and root cause analysis of experimental differences, quickly locate the core causes of differences, shorten the difference investigation cycle, and optimize work efficiency; 4. Construct an integrated experimental comparative analysis method that integrates functions such as template parsing, indicator extraction, comparative analysis, and difference tracing to adapt to various experimental scenarios and improve the versatility and practicality of the method.

[0018] In summary, the purpose of the differentiated intelligent traceability test comparison analysis method is to assist automotive companies in better managing, understanding and utilizing test data. By comparing parameters with those of competitors, intelligent analysis can trace the causes, enabling the formulation of R&D strategies, optimization of product performance, and improvement of design, thereby efficiently consolidating capabilities and enhancing product competitiveness.

[0019] like Figure 1 As shown in this embodiment, the experimental comparative analysis method for differentiated intelligent traceability includes the following steps: S1: Test data entry and collection; collect test data from multiple platforms, projects, models, and stages reported by engineers, convert the test data into structured information and complete the structured data entry into the database to provide a data foundation for subsequent analysis and comparison; S2: Adaptive parsing of multi-type test templates; filtering invalid cells by calculating the confidence level of valid information, locating the horizontal and vertical header split points, constructing a multi-level header multi-branch tree structure, storing the headers in association with the test data, and supporting incremental updates of templates to adapt to template changes, realizing automatic parsing and structured data management of different types of test templates; S3: Multi-dimensional comparative analysis of experimental data; Select single-index, multi-index, time-series curve or image feature comparison mode to quantitatively compare structured experimental data, display the results in a visual form, automatically extract key features and label significant difference indicators; S4: Intelligent source tracing of experimental differences; extracting difference feature vectors based on significant difference indicators, screening four categories of potential influencing factors: experimental conditions, equipment, samples, and operations, constructing a correlation model based on Bayesian network algorithm to conduct multi-dimensional correlation analysis, locating key influencing factors, and outputting source tracing results including root causes and verification evidence; S5: Results Integration and Application Output; Integrate template parsing, data comparison, and difference tracing results throughout the entire process to form an integrated analysis conclusion for use in automotive product performance evaluation, design optimization, and R&D decision-making.

[0020] The following is a detailed explanation: 1. Data entry and collection To collect various test data submitted by engineers from different platforms, projects, vehicle models, and vehicle model stages, engineers need to submit test results after completing the tests. Then, based on different types of templates, the data is parsed and converted into structured information. This completes the collection and storage of test data, enabling structured data entry and effectively ensuring the accuracy, completeness, and relevance of the test data. This provides a reliable data foundation for subsequent data fusion and comparative analysis.

[0021] 2. Parsing different types of templates To address the structural differences between different template types (such as different header row numbers, different data starting positions, and different distributions of merged cells), an adaptive parsing algorithm was designed to automatically extract experimental index parameters.

[0022] Specifically, it includes the following steps: S21. Valid cell filtering: Calculate the confidence level of valid information in cells, filter out invalid information such as blanks and explanatory text, and retain valid indicator and numerical cells; S22. Locating the dividing point between the horizontal and vertical headers: Traverse the table rows and determine the header boundary row according to the column index change rules to distinguish between the horizontal and vertical headers.

[0023] S23. Construction of a multi-level header multi-branch tree: Traverse the headers of the horizontal and vertical axes from top to bottom / from left to right, locate the parent node of each level, merge adjacent identical content, and generate a multi-branch tree structure.

[0024] S24. Structured storage of table headers: Store multi-branch tree nodes in the database, persist hierarchical relationships and table header types, and establish a standardized storage structure.

[0025] S25. The test data is stored in the database in association with the table header; the data cells are located using the IDs of the leaf nodes of the horizontal and vertical table headers to achieve a strong association between the "table header and data" storage.

[0026] S26. Form query redraw: Extract table headers and data from the database according to the query conditions, restore the original form structure and display it.

[0027] S27. Template incremental adaptive update: Compare the header nodes of the old and new templates, add / modify nodes incrementally, keep the historical data association unchanged, and be compatible with template changes.

[0028] The following is a detailed explanation: (1) The test template contains a large amount of invalid information (such as blank cells and explanatory text), so it is necessary to first filter out the cells containing valid indicator parameters. By calculating the confidence score of the valid information of each cell's row and col (row is the row index, col is the column index), it is determined whether it is an indicator or numerical cell to be extracted. The confidence score calculation formula is as follows: P = αf1 + βf2 + γf3 In the formula, α, β, and γ are weighting coefficients, and satisfy α + β + γ = 1 (empirical values ​​α = 0.4, β = 0.3, and γ = 0.3). f1 represents the matching score between the cell text and the standard indicator library (1 for a successful match, 0.5 for a partial match, and 0 for no match). f2 is the cell's position score relative to the header and data area (1 if the cell is below the header and within the data area, and 0 if it is far from the data area). F3 represents the probability that the cell contains a valid numerical value (pure numerical value or numerical value with a unit of 1, and pure text value of 0). Set a confidence threshold τ (empirical value τ=0.5). When P>τ, the cell is determined to be a valid information cell and included in the subsequent parsing range; otherwise, it is determined to be an invalid cell and filtered out.

[0029] After parsing the filtered valid cell data, the number of header rows and the starting position of data differ among different types of test templates. Using fixed row and column indices to locate the data area will lead to parsing failure. Therefore, a standardized preprocessing and structured data entry technology was designed. The core process includes horizontal and vertical header partitioning, multi-branch tree construction, related data entry, and query redrawing. The parsing rules are clarified through quantitative formulas to ensure stable parsing of different structure templates. The specific implementation is as follows.

[0030] a. Positioning of horizontal and vertical table header dividing points Traverse the form from top to bottom, using the column number of the last cell with a value in each row as the criterion. The first time the column number of the current row ends is less than that of the previous row is the dividing point between the horizontal and vertical headers.

[0031] Split point determination formula: i0=min{i|L(i) <L(i 1), i≥2} L(i) is the column index of the last non-empty cell in the i-th row. i0 is the row containing the split point. Rules Explanation: Line 1 - i0 Row 1: Horizontal axis header Rows i0 through the last row: Header of the y-axis b. Construction of the multi-way tree for the horizontal axis header The template may contain more than one row and one column header for the horizontal and vertical axes, with multiple levels of headers. Therefore, the horizontal axis headers are traversed from top to bottom to locate the level and parent node, and adjacent identical content is merged to form a multi-branch tree structure.

[0032] Let the content of cell (x,y) be C(x,y), the parent node be P(x,y), and the tree node be Node(x,y).

[0033] Parent node location:

[0034] Merge adjacent identical content:

[0035] The multi-branch tree structure of the vertical axis header is logically symmetrical with that of the horizontal axis, traversing from left to right and merging adjacent identical content in the same column.

[0036] Parent node location:

[0037] Merge adjacent identical content:

[0038] c. Structured header data import Store the horizontal and vertical header multi-branch tree nodes in the database to persist the hierarchical relationship: DB(Node)=(Node_ID,Content,Parent_ID,Type) In the formula: DB(Node) is the database storage structure of the table header node; Node_ID is the unique number of the node; Content is the text content of the table header; Parent_ID is the parent node number, which represents the hierarchical relationship of the table header; Type is the table header type number, which is used to distinguish between the horizontal axis table header and the vertical axis table header; where Type identifies the table header type: 1 = horizontal axis table header, 2 = vertical axis table header.

[0039] d. Experimental data is linked to the table headers and stored in the database. Each data cell is located by the leaf node ID on the horizontal axis and the leaf node ID on the vertical axis: DB(Data)=(Data_ID,Value,IDx,IDy) to achieve a strong association between "table header and data", providing a basis for query redrawing.

[0040] e. Form redrawing based on query conditions Based on the query criteria Cond, retrieve the matching headers and data from the database, and redraw the cells according to the original structure:

[0041] In the formula This indicates the first element in the redrawn form. line, number The content displayed in the column cells; The row index in the redraw form is used to locate the vertical position of the cell in the redraw form; The column index in the redraw form is used to locate the horizontal position of the cell within the redraw form; : These are the query conditions set; When the cell is a header node and meets the query conditions. When this happens, the cell displays the text content of that header node; When the cell is a data cell and its associated horizontal axis header node identifier... Vertical axis header node identifier When all query conditions are matched, the cell displays the corresponding test data value; The unique identifier of the leaf node in the header corresponding to the horizontal axis of the data cell; The unique identifier of the leaf node in the header of the corresponding y-axis of the data cell; : This indicates that the cell is blank when there is no matching content.

[0042] Adaptive Template Updates: To handle form template changes, incremental template update logic has been added, supporting adaptive expansion of form templates without requiring secondary code modifications. This ensures historical data compatibility and improves the parsing and adaptation capabilities for new experimental templates. The incremental update logic for a multi-branch tree header is designed. Let the original header node set be `Told`, and the new template header node set be `Tnew`. The comparison rule for nodes under the same parent node `Pid` is as follows:

[0043] In the formula, C represents the node content, and Told(Pid) is the set of all node contents under the parent node Pid in the original table header.

[0044] After traversing the new template header from top to bottom and processing it according to the above rules, the newly added set of nodes satisfies the following: Tadd={Nodenew|Nodenew∈Told} Tadd represents adding a new leaf node or a new subtree. The parent node of all leaf nodes in the new subtree is the newly generated parent node ID, and the original header coordinate ID of the historical data remains unchanged, without affecting the correlation of the data already in the database.

[0045] Type identifies the header type (1 = x-axis, 2 = y-axis). During incremental updates, only the node record corresponding to Tadd is added / modified.

[0046] Experimental data is linked to the database: DB(Data)=(Data_ID,Value,IDx,IDy). The data is strongly linked to the table header by using the IDs of the leaf nodes in the horizontal and vertical headers. The IDx and IDy of historical data remain unchanged.

[0047] Query Redraw: Based on the query condition Cond, restore the form structure according to the formula, and automatically incorporate newly added header nodes into the redraw logic.

[0048]

[0049] 3. Data Comparison and Analysis The method, based on the collected experimental dataset, enables precise comparative analysis of multiple sets of experimental data, outputting quantitative comparison results, specifically including: (1) Comparison mode selection: Supports users to select multiple comparison modes such as single indicator comparison, multi-indicator comprehensive comparison, time series curve comparison, and image feature comparison. The comparison threshold and comparison weight can be customized. Through the comparison mode selection, users can independently select key parameters for targeted comparison, focus on core performance indicators, and accurately filter samples / prototypes from the same platform, the same project, the same model, and the same model stage, ensuring that the selected comparison objects have the same source and comparability, and ensuring the scientificity and effectiveness of the comparative analysis.

[0050] (2) Visualization of Comparison Results: Automatically retrieves test result data for each vehicle model corresponding to the test items, and presents the comparison results of test data for multiple vehicle models in a visual form. Displays the full-dimensional test results comparison of multiple samples / vehicles under the same item, covering multiple repeated test data and multi-index dimension test data. It can automatically extract key feature parameters (such as peak value, valley value, change cycle, fluctuation amplitude, etc.) from the visualized curve, and mark significant differences, so that users can quickly view the differences in test data, help test personnel to more clearly understand the test phenomena and processes represented by the time series curve, accurately discover the deficiencies in the product design process, and then propose targeted improvement suggestions.

[0051] 4. Differentiated Intelligent Traceability Module The differentiated intelligent traceability module is the core of this invention. In multi-factor coupled scenarios, it is difficult to accurately locate the root cause of differences. Experimental differences are often the result of the combined effects of multiple influencing factors (experimental conditions, equipment, samples, operations). There are coupling relationships between these factors, making it difficult to distinguish the degree of influence of a single factor on the difference index. Furthermore, the association rules of influencing factors differ in different experimental scenarios.

[0052] This invention, based on significant difference indicators identified by a comparative analysis module, utilizes an intelligent source tracing algorithm for experimental differences to achieve automatic identification and multi-dimensional analysis of the root causes of differences. The specific technical solution is as follows: (1) Core logic of the source tracing algorithm: Taking the significant difference index as the starting point, a Bayesian network algorithm is used to construct a correlation model of "difference index - influencing factors - root cause". Combined with experimental data, experimental template information, and source tracing rule base, the root cause of the difference is located through multi-dimensional feature matching and causal relationship analysis. The specific steps include: Step 1: Difference Feature Extraction: The significant difference index identified by the comparison analysis module is used to extract its multi-dimensional features and construct a difference feature vector to provide a foundation for subsequent association analysis. Let the significant difference index be Id, and the extracted feature vector is defined as: F(Id)=[fnum,ftime,frel].

[0053] In the formula, fnum is a numerical feature vector containing the difference value ΔV = |Vtest Vstandard | Difference Rate The trend is ftrend (1 for rising, -1 for falling, 0 for stable); ftime is the time-series feature vector, containing the time node t0 when the difference occurs and the duration Δt=tend. t0; frel represents the correlation characteristic, i.e., the degree of correlation between this difference index and other experimental indicators, calculated using the Pearson correlation coefficient:

[0054] Where Vd,i represents the i-th group of difference indicators, and Vo,i represents the i-th group of correlation indicators.

[0055] Step 2: Screening of Influencing Factors: Based on the experimental information obtained from the Excel template, screen out the set of potential influencing factors that may affect the difference index Id, F={F1,F2,F3,F4}, where: F1 is the experimental condition factor (temperature, humidity, etc.), F2 is the equipment factor (model, calibration accuracy, etc.), F3 is the sample factor, and F4 is the operational factor (process, parameter settings, etc.).

[0056] Define an influencing factor screening threshold λ. Using the factor association thresholds preset in the source tracing rule base, preliminary screening is conducted to identify potential factors related to ID. The screening formula is as follows: Fpot={Fk|Rel(Id,Fk)≥λ,k=1,2,3,4} In the formula, Rel(Id,Fk) is the initial correlation between the difference index and the k-th type of factor (derived from the preset value of the traceability rule base), and Fpot is the set of potential influencing factors after screening.

[0057] Step 3: Multi-dimensional Association Analysis: Using the Bayesian network algorithm, a model is constructed to link differential features with various potential influencing factors. This is combined with a knowledge graph and a source tracing rule base (with pre-defined association weights and causal relationships between different influencing factors and differential indicators) to calculate the association degree between each potential influencing factor and the differential indicator. Key influencing factors with association degrees higher than a preset threshold are initially screened. Let the set of Bayesian network nodes be N = {Id, Fpot1, Fpot2, ..., Fpotm} (where m is the number of potential influencing factors). The conditional probability between nodes represents the strength of the causal association. The core formula is the Bayesian posterior probability formula:

[0058] In the formula, P(Fpotk) is the prior probability of the k-th potential influencing factor (derived from the source rule base and historical experimental data); P(F(Id)|Fpotk) is the conditional probability of the current differential feature appearing under the influence of this influencing factor; P(F(Id)) is the prior probability of the differential feature vector, calculated using the law of total probability.

[0059] By setting a correlation threshold θ, factors with posterior probabilities P(Fpotk∣F(Id))>θ are selected as the key influencing factor set Fkey that affects the difference index, and multi-dimensional correlation analysis is completed.

[0060] Step 4: Source tracing result output: Based on the key influencing factor set Fkey, combined with experimental data verification and source tracing rule base, the source tracing result is automatically generated. The output format is: Result={Id,Fkey,E}.

[0061] In the formula, Id is the significant difference index, Fkey is the key influencing factor, and E is the verification basis, which includes Bayesian posterior probability, experimental data support, and traceability rule matching results. It clearly marks the root cause of the difference and the verification logic to ensure that the traceability results are traceable and verifiable.

[0062] Source tracing rule base: It presets association rules and confidence calculation standards for influencing factors corresponding to different test scenarios and test indicators. It also supports users to add and modify rules, continuously optimize association rules and confidence calculation models, and improve the accuracy of source tracing.

[0063] In this embodiment, Excel form templates and test data for three typical test scenarios (engine performance test, durability test, and powertrain test) are selected. Template preprocessing and data comparison are performed, and significant numerical differences are traced back to their sources. Compared to manual identification of key information + manual data integration and comparison + traditional source tracing methods, this invention takes an average of 5 minutes per test, while manual operation takes 35 minutes or more, resulting in an efficiency improvement of at least 86%. Furthermore, the accuracy of form parsing and the accuracy of accurately locating the causes of differences are both greater with this invention than with manual methods.

[0064] In summary, this invention effectively solves the technical pain points of existing technologies, such as low efficiency in parsing experimental forms, poor adaptability, and inaccurate difference tracing, by using structured form preprocessing and data entry technology and a significant difference index tracing algorithm, and significantly improves the standardization and automation level of experimental data processing.

[0065] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0066] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0067] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the differentiated intelligent traceability test comparison analysis methods in the above embodiments.

[0068] It is understood that the systems, devices, and storage media provided in the embodiments of the present invention correspond to the methods provided in the embodiments of the present invention, and the explanations, examples, and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

[0069] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0070] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0071] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0072] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A comparative analysis method for differentiated intelligent traceability, characterized in that, Includes the following steps, S1. Collect test data from multiple platforms, projects, models, and stages reported by engineers, convert the test data into structured information, and complete the structured data entry into the database to provide a data foundation for subsequent analysis and comparison. S2. Filter invalid cells by calculating the confidence level of valid information, locate the horizontal and vertical header split points, construct a multi-level header multi-branch tree structure, associate and store the header with the experimental data, and support incremental updates of templates to adapt to template changes, so as to realize automatic parsing and data structured management of different types of experimental templates. S3. Select single-index, multi-index, time-series curve, or image feature comparison mode to quantitatively compare structured experimental data, display the results in a visual form, automatically extract key features, and label significant difference indicators. S4. Extract the difference feature vector based on the significant difference index, screen four types of potential influencing factors: experimental conditions, equipment, samples, and operations, construct a correlation model based on the Bayesian network algorithm to conduct multi-dimensional correlation analysis, locate key influencing factors, and output the source tracing results including root causes and verification basis. S5 integrates template parsing, data comparison, and difference tracing results throughout the entire process to form an integrated analysis conclusion, which is used for automotive product performance evaluation, design optimization, and R&D decision-making.

2. The experimental comparative analysis method for differentiated intelligent traceability according to claim 1, characterized in that: S2 specifically includes, S21. Valid Cell Filtering: Calculate the confidence level of valid information in cells, filter out invalid information such as blank spaces and explanatory text, and retain valid indicator and numerical cells. S22, Positioning of horizontal and vertical table header dividing points; Traverse the table rows of data, determine the header boundary row according to the column index change rules, and distinguish between the horizontal axis header and the vertical axis header; S23. Construction of a multi-level header multi-branch tree: Traverse the headers of the horizontal and vertical axes from top to bottom / from left to right, locate the parent node of the level, merge adjacent identical contents, and generate a multi-branch tree structure. S24. Structured header data is imported into the database; Store the multi-branch tree nodes in the database, persist the hierarchical relationship and table header type, and establish a standardized storage structure. S25. The test data is stored in the database in association with the table header; the data cells are located using the IDs of the leaf nodes of the horizontal and vertical table headers to achieve the association between the "table header and data" for storage. S26. Redraw the form query; Extract table headers and data from the database according to the query conditions, restore the original form structure and display it; S27. Template incremental adaptive update; Compare the header nodes of the old and new templates, add / modify nodes incrementally, maintain the historical data association, and ensure compatibility with template changes.

3. The experimental comparative analysis method for differentiated intelligent traceability according to claim 2, characterized in that: S21 specifically includes, By calculating the confidence score of each cell's row and column indices (row and col), we determine whether it is a metric or numerical cell to be extracted. The confidence score calculation formula is as follows: P = αf1 + βf2 + γf3 In the formula, α, β, and γ are weighting coefficients, and satisfy α+β+γ=1; f1 represents the matching score between the cell text and the standard indicator library: 1 for a successful match, 0.5 for a partial match, and 0 for no match. f2 is the cell's position score relative to the header and data area. It is 1 if the cell is below the header and within the data area, and 0 if it is far from the data area. F3 represents the probability that the cell contains a valid numerical value; 1 for pure numerical values ​​or numerical values ​​plus units, and 0 for pure text. Set a confidence threshold τ. When P>τ, the cell is determined to be a valid information cell and included in the subsequent parsing range; otherwise, it is determined to be an invalid cell and filtered out.

4. The experimental comparative analysis method for differentiated intelligent traceability according to claim 3, characterized in that: S22 specifically includes locating the horizontal and vertical table header dividing points, and the steps are as follows: Traverse the form from top to bottom, using the column number of the last cell with a value in each row as the criterion. The first time the column number of the current row ends is less than that of the previous row is the dividing point between the horizontal and vertical headers. Division point determination formula: i0 = min{i | L(i) <L(i 1), i ≥ 2} L(i) is the column index of the last non-empty cell in the i-th row; i0 is the row containing the split point; Rules Explanation: Line 1 - i0 Row 1: Horizontal axis header; Row i0 to the last row: Header of the vertical axis.

5. The experimental comparative analysis method for differentiated intelligent traceability according to claim 4, characterized in that: S24 specifically includes the structured data entry of table headers into the database, and the steps are as follows. Store the horizontal and vertical header multi-branch tree nodes in the database to persist the hierarchical relationship: DB(Node)=(Node_ID,Content,Parent_ID,Type) In the formula: DB(Node) is the database storage structure of the table header node; Node_ID is the unique number of the node; Content is the text content of the table header; Parent_ID is the parent node number, representing the table header hierarchy; Type is the header type number, used to distinguish between the horizontal axis header and the vertical axis header; The Type identifier indicates the header type: 1 = horizontal axis header, 2 = vertical axis header.

6. The experimental comparative analysis method for differentiated intelligent traceability according to claim 5, characterized in that: S27 includes template incremental adaptive updates: Design the incremental update logic for the multi-way tree header. Let the original header node set be Told, and the new template header node set be Tnew. The comparison rule for nodes under the same parent node Pid is as follows: In the formula, C represents the node content, and Told(P) id ) represents the parent node P in the original table header. id The collection of all node contents under; After traversing the new template header from top to bottom and processing according to the above rules, the newly added node set satisfies: Tadd={Nodenew|Nodenew∈Told} Among them, Tadd is a newly added leaf node or a newly added subtree. The parent node of all leaf nodes in the newly added subtree is the newly generated parent node ID, and the original table header coordinate ID bound to the historical data remains unchanged, without affecting the correlation of the data already entered into the database. Type identifies the header type, 1 = x-axis, 2 = y-axis. During incremental updates, only the node record corresponding to Tadd is added / modified. Experimental data is linked into the database: DB(Data)=(Data_ID,Value,ID) x ID y Strong associations between data and table headers are achieved through the IDs of the leaf nodes in the horizontal and vertical headers, and the IDs of historical data are also included. x ID y Remain unchanged; Query Redraw: Based on the query condition Cond, restore the form structure according to the formula, and automatically incorporate any newly added header nodes into the redraw logic. ； In the formula This indicates the first element in the redrawn form. line, number The content displayed in the column cells; The row index in the redraw form is used to locate the vertical position of the cell in the redraw form; The column index in the redraw form is used to locate the horizontal position of the cell within the redraw form; : These are the query conditions set; When the cell is a header node and meets the query conditions. When this happens, the cell displays the text content of that header node; When the cell is a data cell and its associated horizontal axis header node identifier... Vertical axis header node identifier When all query conditions are matched, the cell displays the corresponding test data value; The unique identifier of the leaf node in the header corresponding to the horizontal axis of the data cell; The unique identifier of the leaf node in the header of the corresponding y-axis of the data cell; : This indicates that the cell is blank when there is no matching content.

7. The experimental comparative analysis method for differentiated intelligent traceability according to claim 6, characterized in that: S3 specifically includes, Comparison mode selection: Supports users to select multiple comparison modes such as single indicator comparison, multi-indicator comprehensive comparison, time series curve comparison, and image feature comparison, and allows users to customize comparison thresholds and comparison weights; Visualization of comparison results: Automatically retrieves test result data for each vehicle model's corresponding test items and presents the comparison results of test data from multiple vehicle models in a visual format.

8. The experimental comparative analysis method for differentiated intelligent traceability according to claim 7, characterized in that: S4 specifically includes, Step 1: Extraction of differential features: The significant difference indicators identified by the comparison analysis module are extracted for their multi-dimensional features, and a differential feature vector is constructed to provide a foundation for subsequent correlation analysis. Step 2, Screening of Influencing Factors: Based on the experimental information obtained from the Excel template, screen out the set of potential influencing factors that may affect the difference index Id, F={F1,F2,F3,F4}, where: F1 is the experimental condition factor, F2 is the equipment factor, F3 is the sample factor, and F4 is the operational factor; Step 3: Multi-dimensional correlation analysis: Using the Bayesian network algorithm, a correlation model between differential features and various potential influencing factors is constructed. Combined with knowledge graphs and source tracing rule bases, the correlation degree between each potential influencing factor and the differential index is calculated, and key influencing factors with a correlation degree higher than the preset threshold are initially screened. Step 4: Source tracing result output: Based on the key influencing factor set Fkey, combined with experimental data verification and source tracing rule base, the source tracing result is automatically generated. The result output format is: Result={Id,Fkey,E}; In the formula, Id is the significant difference index, Fkey is the key influencing factor, and E is the verification basis, which includes Bayesian posterior probability, experimental data support, and traceability rule matching results. It clearly marks the root cause of the difference and the verification logic to ensure that the traceability results are traceable and verifiable.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 8.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the computer program is executed by the processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 8.

Images