Motor vehicle inspection personnel operation completion credibility evaluation method based on knowledge graph

Abstract

The invention provides a motor vehicle inspection personnel operation completion credibility evaluation method based on a knowledge graph. According to the method, station videos, equipment output, OBD states, detection curves, inspection records and inspection standard data are obtained, standard steps, micro skills, step precedence relations and result corresponding relations are established, and action information, equipment response information, motor vehicle response information, record information and task situation information are extracted; generating execution sufficiency, sequence consistency, result correspondence and record consistency of each standard step by using a KAN operation evidence network, and associating the execution sufficiency, the sequence consistency, the result correspondence and the record consistency to a motor vehicle inspection capability knowledge graph for relation reasoning to form a step establishment result, a capability state, standard operation completion credibility and a post adaptation, retraining or rechecking conclusion; the method can be used for motor vehicle inspection personnel standard operation completion condition evaluation, and realizes closed evidence determination and source distinguishing of a standard step level.

Description

Technical Field

[0001] This invention relates to the field of motor vehicle inspection operation evaluation technology, and in particular to a method for evaluating the credibility of motor vehicle inspection personnel's work completion based on knowledge graphs. Background Technology

[0002] Motor vehicle inspection is a crucial operational link in ensuring road traffic safety, compliance with pollutant emission standards, and the standardized operation of inspection agencies. When performing inspection tasks, inspectors typically need to complete standard operations such as visual inspection, OBD inspection, and free acceleration testing at different workstations, according to inspection standards, and generate corresponding inspection records based on operational requirements, equipment usage requirements, record-keeping requirements, and result judgment requirements. With the informatization of the inspection process and the digitization of testing equipment, the inspection site can continuously generate multi-source data, including workstation videos, equipment outputs, OBD status, test curves, and inspection records. These data correspond to personnel actions, equipment changes, vehicle reactions, and result completion. Meanwhile, different vehicle types, fuel types, inspection task types, workstation types, equipment types, and task stages can affect the triggering conditions, result ranges, and recording methods of standard steps, resulting in evidence within the same inspection task exhibiting clear temporal sequence, object correspondence, and task context constraints. In practical management, inspection agencies not only need to determine whether a test result is abnormal, but also need to further determine whether the inspectors completed the work according to standard procedures, whether they possess the corresponding micro-skills, and whether the abnormality originated from personnel, equipment, vehicles, or the task context. This supports subsequent management activities such as personnel suitability, retraining, and review. Therefore, conducting credible evaluations of the standard operating procedures performed by vehicle inspectors has become a key fundamental issue in on-site quality control and competency management.

[0003] In existing technologies, the evaluation of motor vehicle inspectors' work typically combines manual spot checks, video playback, equipment log verification, and inspection record checking. Some solutions are based on inspection standards, setting checkpoints in the work process and performing rule matching based on action recognition results in the workstation video, status changes in equipment output, parameter changes in OBD status and detection curves, and field completion in inspection records. Other solutions employ behavior recognition, time series analysis, or comprehensive scoring methods to automate the analysis of inspectors' work processes and use the analysis results to determine whether the work was completed, whether the process was standardized, or whether the records were consistent. For the application of multi-source data, existing solutions generally extract corresponding features according to the data source and then output evaluation results using rule engines, threshold judgments, or comprehensive scoring models for daily supervision, internal assessment, and review of abnormal tasks.

[0004] However, in actual vehicle inspection scenarios, the aforementioned existing technologies often struggle to form stable and interpretable closed-loop evidence around standard procedures. Existing solutions often assess actions, equipment, vehicle responses, and records separately, or further compress these into a single evaluation result, making it difficult to maintain the continuous relationship between action implementation, step sequence, result correspondence, and record correspondence at the standard procedure level. Consequently, when an anomaly occurs in a certain step, it becomes difficult to determine whether the anomaly stems from insufficient operator skill, equipment factors, vehicle factors, or task context factors. This can easily affect the accurate identification of micro-skill capabilities and hinder the subsequent formation of differentiated treatment conclusions for suitability, retraining, or reassessment.

[0005] Therefore, a method for assessing the credibility of motor vehicle inspectors' work completion that can overcome the shortcomings of the existing technology is a problem that needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a knowledge graph-based method for assessing the credibility of motor vehicle inspection personnel's work completion. The core technical problem this application aims to solve is: in actual motor vehicle inspection scenarios, given the temporal coupling between personnel operation, equipment response, vehicle response, and record filling, and the difficulty in distinguishing the sources of anomalies, how to establish a closed evidence and source differentiation mechanism at the standard step level, relying solely on workstation videos, equipment output, OBD status, detection curves, inspection records, and inspection standard data, so that the results of the standard steps can be reliably determined, thereby supporting the credibility assessment of standard work completion and the formation of conclusions on suitability, retraining, or review.

[0007] The knowledge graph-based reliability assessment method for motor vehicle inspection personnel's job completion according to embodiments of the present invention includes:

[0008] S1. Acquire workstation video, equipment output, OBD status, detection curve, inspection record and inspection standard data. Based on the inspection standard data, determine the standard steps of the inspection task, corresponding micro-skills, the sequence of steps and the correspondence of results, and form the capability correspondence of inspection steps.

[0009] S2. Based on the capability correspondence of inspection steps, extract action information, equipment response information, vehicle response information, record information and task context information corresponding to each standard step from workstation video, equipment output, OBD status, detection curve and inspection record to form step process information;

[0010] S3. Input the process information into the KAN work evidence network. Within the standard steps defined by the capability correspondence of the verification steps, perform continuous mapping according to the order of action occurrence, step sequence, equipment response, vehicle response and record correspondence, so that the operation behavior, equipment change, vehicle response and record result of the same standard step form a closed chain. Calculate the execution sufficiency, sequence consistency, result correspondence and record consistency, and generate standard step closed evidence for each standard step.

[0011] S4. Link the standard procedure closure evidence to the relation items corresponding to the standard procedure and micro-skill in the knowledge graph of motor vehicle inspection capabilities, and generate source differentiation tags based on the equipment response information, motor vehicle response information and task context information in the procedure process information. Under the constraints of the step sequence relationship and the result correspondence relationship, perform relational reasoning on the standard procedure closure evidence and source differentiation tags to form the result of the procedure being valid.

[0012] S5. Determine the ability status based on the results of the steps and the correspondence between the standard steps and micro-skills, and determine the source type based on the source distinction mark;

[0013] S6. Calculate the reliability of standard operation completion based on the ability status, and form a job suitability conclusion, retraining conclusion, or review conclusion by combining the job suitability threshold and source type.

[0014] Optionally, S1 is as follows:

[0015] Acquire workstation videos, equipment outputs, OBD status, detection curves, inspection records, and inspection standard data corresponding to the same inspection task, and establish task correspondence relationships between workstation videos, equipment outputs, OBD status, detection curves, and inspection records based on the inspection task identifier;

[0016] The inspection standard data is analyzed step by step to extract the standard steps corresponding to the inspection task. Based on the operation requirements, equipment usage requirements, record filling requirements and result judgment requirements, the operation content, operation objects and corresponding results of each standard step are determined.

[0017] Based on the operation content and corresponding results of each standard step, determine the micro-skills corresponding to each standard step, establish the step sequence relationship between standard steps according to the operation sequence in the test standard data, and establish the result correspondence relationship between standard steps and corresponding results according to the corresponding results.

[0018] The standard steps, corresponding micro-skills, step sequence relationships, and result correspondences are linked according to the standard steps to form a capability correspondence for the verification steps. This capability correspondence is then used as a constraint on the information extraction path for the step process and the scope of evidence generation for the standard steps.

[0019] Optionally, S2 is as follows:

[0020] Based on the capability correspondence of the inspection steps, the standard steps are used as the benchmark for extracting process information. The operation content, operation objects, equipment usage requirements, vehicle response requirements and record filling requirements corresponding to each standard step are determined. The occurrence boundaries of each standard step are defined according to the sequence of steps, and the equipment response range, vehicle response range and record corresponding range corresponding to each standard step are defined according to the result correspondence.

[0021] The workstation video is analyzed according to standard steps to extract the occurrence, duration, coverage, number of times, sequence, matching of the target, workstation dwell time, and end of the action corresponding to the standard steps, forming action information that corresponds one-to-one with the standard steps.

[0022] The device output is parsed according to standard procedures, and the device startup status, measurement status, sampling status, code status, output change status and device completion status corresponding to the standard procedures are extracted to form device response information that corresponds one-to-one with the standard procedures. The device response information is then limited to the range of results corresponding to the standard procedures.

[0023] The OBD status and detection curve are analyzed according to standard procedures. The changes in rotational speed, smoke opacity, OBD connection status, OBD code status, component response status, and detection result changes corresponding to the standard procedures are extracted to form motor vehicle response information that corresponds one-to-one with the standard procedures. The motor vehicle response information is then limited to the range of results corresponding to the standard procedures.

[0024] The inspection records are parsed according to the standard steps, and the triggering conditions, field completeness, filling order, result filling, object correspondence and step correspondence corresponding to the standard steps are extracted to form record information that corresponds one-to-one with the standard steps. Combined with the vehicle type, fuel type, inspection task type, workstation type, equipment type and task stage, task context information corresponding to the standard steps is formed.

[0025] Action information, equipment response information, vehicle response information, record information, and task context information are associated and merged according to the same standard steps. This ensures that personnel operations, equipment changes, vehicle reactions, and record entries within the same standard steps are under the same verification semantics, forming step process information that allows the KAN operation evidence network to generate standard step closure evidence according to the standard steps.

[0026] Optionally, S3 specifically refers to:

[0027] Based on the capability correspondence of the inspection steps, the process information of the steps is allocated according to a single standard step, the input range of the current standard step is determined, the process information of different standard steps is not merged into the input, and the sequence boundary between the current standard step and the preceding and subsequent standard steps is defined according to the step sequence relationship. The corresponding boundary of equipment response information, motor vehicle response information and record information is defined according to the result correspondence relationship.

[0028] The action information and task context information corresponding to the current standard steps are input into the action evidence layer of the KAN task evidence network. KAN neurons are used to map the input side functions of each input dimension, and then node summation is used to form action evidence, so that the action information is transformed into operation implementation evidence within the scope of the current standard steps, and the basis for calculating execution sufficiency is established.

[0029] The action evidence and equipment response information are input into the equipment association layer of the KAN work evidence network. Continuous mapping is performed within the current standard steps to form equipment association evidence that represents the correspondence between action implementation and equipment changes. The order of actions is constrained by the sequence of steps to establish the basis for calculating the degree of sequence consistency.

[0030] The equipment-related evidence and vehicle response information are input into the vehicle-related layer of the KAN operation evidence network. Continuous mapping is performed within the current standard procedure range to form vehicle-related evidence that characterizes the correspondence between equipment changes and vehicle responses. The calculation basis for the result correspondence degree is established by combining the result correspondence.

[0031] Motor vehicle-related evidence and record information are input into the record closure layer of the KAN operation evidence network, and continuous mapping is performed within the current standard procedure range to form a closure relationship quantity that represents the closure relationship between record filling and the implementation of preceding actions, equipment changes and motor vehicle reactions, and to establish the basis for calculating record consistency.

[0032] The closed relation quantity is input into the output layer of the KAN operation evidence network, and the execution sufficiency, sequence consistency, result correspondence and record consistency are output one by one, so that the four outputs correspond to the action implementation, step sequence, result correspondence and record correspondence of the current standard step, respectively.

[0033] The sufficiency of execution, the consistency of sequence, the correspondence of results, and the consistency of records are combined to form the standard procedure closure evidence for the current standard procedure. Among them, the action information must not bypass the equipment response information and the vehicle response information to form the standard procedure closure evidence, and the record information is based on the vehicle-related evidence to form the standard procedure closure evidence. The standard procedure closure evidence for each standard procedure is generated separately.

[0034] Optionally, S4 specifically refers to:

[0035] Based on the standard steps, corresponding micro-skills, step sequence relationships, and result correspondence relationships, the relationship items corresponding to the current standard steps are determined in the knowledge graph of motor vehicle inspection capabilities. The relationship items include execution relationship items that represent the execution requirements of the standard steps, step sequence relationship items that represent the order requirements of the standard steps, result correspondence relationship items that represent the result requirements of the standard steps, and record correspondence relationship items that represent the recording requirements of the standard steps.

[0036] The sufficiency of execution, consistency of sequence, correspondence of results, and consistency of records of the current standard procedures are respectively associated with execution relationship items, step sequence relationship items, result correspondence relationship items, and record correspondence relationship items, so that the closed evidence of the standard procedures is written into the knowledge graph of motor vehicle inspection capabilities one by one according to the relationship items, and the four pieces of evidence are not compressed into a single score.

[0037] A device source identification mark is generated based on the conformity of the correspondence between the device response information and the result within the specified range; a vehicle source identification mark is generated based on the conformity of the correspondence between the vehicle response information and the result within the specified range; and a task context source identification mark is generated based on the conformity of the task context information with the task context requirements corresponding to the current standard steps.

[0038] For the standard steps of the evidence written into the relation items, perform relational reasoning. First, determine whether the execution sufficiency supports the implementation of the action of the current standard step based on the execution relation item. Then, determine whether the sequence consistency supports the sequence of the current standard steps based on the step sequence relation item. Next, determine whether the result correspondence supports the result of the current standard step based on the result correspondence relation item. Finally, determine whether the record consistency supports the record of the current standard step based on the record correspondence relation item.

[0039] When the execution sufficiency, sequence consistency, result correspondence, or record consistency do not support the establishment of the corresponding relationship item, the equipment source differentiation mark, the motor vehicle source differentiation mark, and the task context source differentiation mark are introduced into the same relational reasoning chain to determine whether the closing breakpoint of the current standard step belongs to the equipment factor, the motor vehicle factor, or the task context factor. If the equipment source differentiation mark, the motor vehicle source differentiation mark, and the task context source differentiation mark are all not established, the closing breakpoint is determined to belong to the personnel factor.

[0040] Based on the relational reasoning results, the current standard step is established. The established step results include the standard step being established and the standard step being not established with source differentiation. The established step results are then consistently associated with the current standard step and the corresponding micro-skill.

[0041] Optional, S5 specifically includes:

[0042] The source type of each standard step is determined based on the equipment source differentiation mark, the motor vehicle source differentiation mark, and the task situation source differentiation mark. When the equipment source differentiation mark is true, the source type is determined to be equipment factor. When the motor vehicle source differentiation mark is true, the source type is determined to be motor vehicle factor. When the task situation source differentiation mark is true, the source type is determined to be task situation factor. When none of the aforementioned source differentiation marks are true and the standard step is not true, the source type is determined to be personnel factor.

[0043] Based on the correspondence between standard steps and micro-skills, the results and source types of each standard step are mapped to the corresponding micro-skills, forming a set of steps corresponding to each micro-skill;

[0044] The set of steps corresponding to each micro-skill is determined. When all the steps in the set of steps are the standard steps, the micro-skill is determined to be in a state where it has the ability.

[0045] When a standard step is not met in the set of corresponding steps and the source type is personnel factors, the micro-skill is determined to be a personnel-deficient capability state. When a standard step is not met in the set of corresponding steps and the source type is equipment factors, vehicle factors, or task situation factors, the micro-skill is determined to be a non-personnel-restricted capability state. The judgment results of each micro-skill are then aggregated into the capability state corresponding to the test task.

[0046] Optionally, step S6 specifically includes:

[0047] Based on the capability status, the standard steps corresponding to the capability status that is already available are determined as reliable completion steps, the standard steps corresponding to the capability status that is insufficient in personnel are determined as personnel deviation steps, and the standard steps corresponding to the capability status that is not limited by personnel are determined as limiting deviation steps.

[0048] Based on the correspondence between standard procedures and micro-skills, micro-skills are merged for reliable completion steps, personnel deviation steps, and limiting deviation steps to determine the completion status and deviation status of each micro-skill in the inspection task.

[0049] Based on the completion status and deviation of each micro-skill, and combined with the number of standard steps corresponding to each micro-skill, the reliability of the standard operation completion of the inspection task is calculated, so that the reliability of the standard operation completion reflects both the validity of the standard steps and the source differentiation.

[0050] The reliability of standard operation completion is compared with the job suitability threshold, and the conclusion is determined in combination with the source type. When the reliability of standard operation completion is not lower than the job suitability threshold, a job suitability conclusion is formed. When the reliability of standard operation completion is lower than the job suitability threshold and the source type is personnel factors, a retraining conclusion is formed. When the reliability of standard operation completion is lower than the job suitability threshold and the source type is equipment factors, vehicle factors, or task situation factors, a review conclusion is formed.

[0051] Optionally, the KAN operation evidence network determines the standard step type based on the operation content, equipment usage requirements, vehicle response requirements, and record filling requirements of the current standard step, and selects the evidence branch corresponding to the standard step type among the action evidence layer, equipment association layer, vehicle association layer, and record closure layer. Specifically, for manual visual inspection type standard steps, the visual evidence branch is selected based on the correspondence between action information and record information as the continuous mapping main path; for OBD inspection type standard steps, the OBD evidence branch is selected based on the sequential correspondence between action information, equipment response information, vehicle response information, and record information as the continuous mapping main path; and for free acceleration method detection type standard steps, the acceleration evidence branch is selected based on the sequential correspondence between action information, equipment response information, vehicle response information, and record information as the continuous mapping main path.

[0052] Optionally, the KAN operational evidence network employs different input side function mapping rules for different evidence branches. In the appearance evidence branch, the action evidence layer maps input side functions to the occurrence of actions, action coverage, matching of operating objects, and workstation dwell time; the record closure layer maps input side functions to the triggering of records, field completeness, result filling, and step correspondence. In the OBD evidence branch, the device association layer maps input side functions to the device startup, code reporting, and output change; the vehicle association layer maps input side functions to the OBD connection, OBD code reporting, and component response. In the acceleration evidence branch, the device association layer maps input side functions to the measurement, sampling, and output change; the vehicle association layer maps input side functions to the speed change, smoke opacity change, and detection result change; and the record closure layer maps input side functions to the result filling and step correspondence.

[0053] Optionally, when two or three of the equipment source differentiation marker, vehicle source differentiation marker, and task context source differentiation marker are simultaneously true, the source differentiation markers are introduced into the same relational reasoning chain for conflict resolution, and the attribution of the closure breakpoint is determined sequentially according to the order of equipment source differentiation marker, vehicle source differentiation marker, and task context source differentiation marker. Specifically, if the equipment response information does not support the establishment of the result corresponding relationship item, it is preferentially determined to be an equipment factor; if the equipment response information supports the establishment of the result corresponding relationship item but the vehicle response information does not support the establishment of the result corresponding relationship item, it is determined to be a vehicle factor; if both the equipment response information and the vehicle response information support the establishment of the result corresponding relationship item but the task context information does not support the task context requirements corresponding to the current standard step, it is determined to be a task context factor; and if none of the above determinations are true, it is determined to be a personnel factor.

[0054] The beneficial effects of this invention are:

[0055] (1) This invention proposes an improved KAN (Kinetic Anchorage) work evidence network method. In the specific task scenario of motor vehicle inspection, which has strict step constraints and result correspondence requirements, it no longer uses the existing technology of judging separately according to data sources or directly compressing multi-source features into a single score. Instead, it takes standard steps as independent units and, within the scope of the correspondence between inspection steps and capabilities, continuously maps action information, equipment response information, motor vehicle response information, and record information in the order of action occurrence, step sequence, equipment response, motor vehicle response, and record correspondence. It also selects corresponding evidence branches in combination with standard steps of manual visual inspection, standard steps of OBD inspection, and standard steps of free acceleration method detection, and outputs execution sufficiency, sequence consistency, result correspondence, and record consistency. This design establishes the standard step closure evidence on a traceable process chain, and can reliably determine whether the standard steps form a complete work closure under the conditions of available workstation videos, equipment outputs, OBD status, detection curves, and inspection records.

[0056] (2) This invention proposes a novel knowledge graph relational reasoning method for motor vehicle inspection capabilities. It associates execution sufficiency, sequence consistency, result correspondence, and record consistency with execution relation items, step sequence relation items, result correspondence relation items, and record correspondence relation items, respectively. It avoids compressing the four pieces of evidence into a single score and further generates equipment source differentiation markers, motor vehicle source differentiation markers, and task context source differentiation markers based on equipment response information, motor vehicle response information, and task context information. This completes closed-loop breakpoint attribution and conflict resolution within the same relational reasoning chain. Compared to existing technologies that struggle to distinguish anomaly sources and easily misjudge equipment or vehicle anomalies as personnel issues, this solution can differentiate between equipment factors, motor vehicle factors, task context factors, and personnel factors with boundaries when standard steps are not met, thereby enhancing the interpretability of the results of successful steps and the consistency of subsequent handling.

[0057] (3) This invention proposes a method for assessing the credibility of standard operation completion based on the correspondence between standard steps and micro-skills. It maps the results and source types of step establishment to micro-skills, forming states of existing capabilities, insufficient capabilities, and non-restricted capabilities. Based on this, credible completion steps, personnel deviation steps, and restrictive deviation steps are merged to calculate the credibility of standard operation completion, resulting in suitability conclusions, retraining conclusions, or review conclusions. This overall approach does not rely solely on the final inspection results or a single anomaly for personnel evaluation. Instead, it completes evidence closure, source differentiation, and capability mapping at the standard step level, ensuring consistency between the assessment of inspection personnel's capabilities and the on-site operation process. This facilitates a closed-loop connection in motor vehicle inspection operations, from process evidence collection and step determination to the output of management conclusions. Attached Figure Description

[0058] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0059] Figure 1 This is a flowchart of a knowledge graph-based method for evaluating the credibility of motor vehicle inspection personnel's work completion, as proposed in this invention. Detailed Implementation

[0060] In Example 1, reference Figure 1 A knowledge graph-based method for assessing the credibility of motor vehicle inspectors' work completion includes:

[0061] S1. Acquire workstation video, equipment output, OBD status, detection curve, inspection record and inspection standard data. Based on the inspection standard data, determine the standard steps of the inspection task, corresponding micro-skills, the sequence of steps and the correspondence of results, and form the capability correspondence of inspection steps.

[0062] S2. Based on the capability correspondence of inspection steps, extract action information, equipment response information, vehicle response information, record information and task context information corresponding to each standard step from workstation video, equipment output, OBD status, detection curve and inspection record to form step process information;

[0063] S3. Input the process information into the KAN work evidence network. Within the standard steps defined by the capability correspondence of the verification steps, perform continuous mapping according to the order of action occurrence, step sequence, equipment response, vehicle response and record correspondence, so that the operation behavior, equipment change, vehicle response and record result of the same standard step form a closed chain. Calculate the execution sufficiency, sequence consistency, result correspondence and record consistency, and generate standard step closed evidence for each standard step.

[0064] S4. Link the standard procedure closure evidence to the relation items corresponding to the standard procedure and micro-skill in the knowledge graph of motor vehicle inspection capabilities, and generate source differentiation tags based on the equipment response information, motor vehicle response information and task context information in the procedure process information. Under the constraints of the step sequence relationship and the result correspondence relationship, perform relational reasoning on the standard procedure closure evidence and source differentiation tags to form the result of the procedure being valid.

[0065] S5. Determine the ability status based on the results of the steps and the correspondence between the standard steps and micro-skills, and determine the source type based on the source distinction mark;

[0066] S6. Calculate the reliability of standard operation completion based on the ability status, and form a job suitability conclusion, retraining conclusion, or review conclusion by combining the job suitability threshold and source type;

[0067] In this embodiment, step S1 specifically includes:

[0068] The inspection task identifier is recorded as The timestamp is in seconds, and the serial number is... The value is taken from the natural number index range of the corresponding sampling sequence, and the workstation video is written as... The device output is written as The OBD status is written as The detection curve is written as The inspection record is written as The test standard data is written as ,in, , , , and These represent the number of sampled items in the corresponding sequence, respectively, for those already carrying the test task identifier. The sampling items, directly press Write the corresponding sequence, and include sampling items that lack a test task identifier. First, start with the inspection records. Extract task start time Then, the workstation type field must be consistent and satisfy the following conditions. Under the conditions, according to Perform task binding, among which This indicates the window corresponding to the task. After the task binding is completed, [the window will be configured]. , , , and Sort them in ascending order by timestamp, so that workstation videos, equipment outputs, OBD status, test curves, and inspection records for the same inspection task are under a unified task identifier. Establish task mapping relationships;

[0069] For test standard data When performing step parsing, first extract the standard step set. ,in Indicates the standard step number. This indicates the number of standard steps, and then for each standard step... Extraction Step Description Item ,in The operation details are derived from the analysis of job requirements and equipment usage requirements. The target of the operation is determined by parsing the inspected part, diagnostic interface, device connection object, or detection object. This indicates the corresponding result, obtained by parsing according to the record filling requirements and result judgment requirements. To ensure that different standard texts can be directly compared, the operation content is... Operation objects and corresponding results Standardization is performed according to a pre-defined glossary, converting synonymous descriptions into unified descriptive values. After standardization, each standard step has a unique operational content, a unique operational object, and a unique corresponding result, thus forming a step description item that can directly participate in relationship establishment. ;

[0070] Micro-skills are recorded as Used to characterize the completion of standard steps The required and independently assessable work competency units are micro-skills, which are not manually assigned labels but rather defined by step descriptions. Directly obtain, specifically, establish a sequence of micro-skill descriptions. ,in Indicates the micro-skill serial number. Indicates the number of micro-skills. Indicates the first Standardized description items corresponding to each micro-skill, and current standard steps. Step description item Perform a match when there exists At that time, the current standard steps will be mapped to When it does not exist At that time, new micro-skills were added. and order Then map the current standard steps to After adopting this matching process, standard steps with the same operation content, the same operation object and the same corresponding result are classified into the same micro-skill, and standard steps with any difference are classified into different micro-skills. Therefore, micro-skills have clear acquisition methods and judgment boundaries.

[0071] The order of steps is denoted as The corresponding relationship is denoted as First, start with the test standard data. Read each standard step The order of tasks ,in If it is a positive integer, then press Establish the sequence of steps Then, for each standard step Establish result correspondence items And form a result correspondence. The capability correspondence of the testing steps is denoted as: ,in , , indicating the standard procedure The direct sequence of steps. Representation and standard procedures The direct correlation between results, and thus the correlation between the capabilities of the testing steps. Provide a complete association structure for each standard step, corresponding micro-skill, step sequence, and result correspondence, organized by standard steps. When extracting information from the step process, use an item sequence. As an extraction path, it only revolves around the current standard steps. Organizational action information, equipment response information, vehicle response information, record information, and task context information are used to generate closed-loop evidence in accordance with standard procedures. Define the order boundary, with Define the boundaries of the results, thereby enabling individual standard steps to form an independent, computable, and graphable input organization basis.

[0072] In this embodiment, step S2 specifically includes:

[0073] In one implementation, the verification task identifier is denoted as The number of standard steps is denoted as The capability correspondence of the testing steps is denoted as: ,in, , Indicates the first A standard procedure, Representation and standard procedures Corresponding micro-skills Representation and standard procedures The sequence of related steps Representation and standard procedures The associated results correspondence, with the workstation video denoted as ,in, Indicates the first Each workstation video sampling timestamp Indicates the first The video sampling value of each workstation, as output by the device, is recorded as... The OBD status is recorded as The detection curve is denoted as The inspection record is recorded as follows Establish a unified timeline ,in, Indicates the start time of the inspection task. Indicates a fixed sampling interval. , This indicates the number of sampling points on a uniform timeline for any data source. ,in, This represents the timestamp of the data source. This represents the sampled values ​​from the data source, forming a discrete sequence using nearest neighbor binding. ,in, Indicates time axis index The corresponding discrete sampled values, when At that time, take And order ,when season , This represents the window that binds to the data source;

[0074] For each standard step The steps to establish the vector ,in, Indicates the operation content. Indicates the object being operated on. Indicates equipment usage requirements. Indicates the motor vehicle's response requirements. This indicates the requirements for filling out the record. Indicates the corresponding result, based on and Determine standard procedures on a unified timeline. The occurrence boundary And calculate using the following formula:

[0075] ;

[0076] In the formula, Indicate standard steps In the Step-matching values ​​for each sampling point This indicates the matching value of the operation content, based on the analysis results of the workstation video and the operation content. The consistency value is or , This represents the matching value of the operation object, based on the operation object identification result and the operation object. The consistency value is or , This indicates the matching value for device usage requirements, based on the device output fields and the device usage requirements. The consistency value is or , This indicates the matching value for the vehicle's response requirements, based on the OBD status and detection curve, and the vehicle's response requirements. The consistency value is or , This indicates the value that matches the record entry requirements, based on the verification of the record fields and the record entry requirements. The consistency value is or , to This represents the preset weight coefficients for the five categories of matching values. Each preset weight coefficient is a dimensionless value and satisfies the following conditions: , Indicate standard steps The step-matching threshold is a dimensionless value. This indicates the end of the preceding standard steps index. This indicates the starting index for subsequent standard steps. If no sampling point meets the conditions, then... and Set it to an invalid index, and set the components of action information, equipment response information, vehicle response information, and record information corresponding to the current standard procedure to [value missing]. , upon obtaining the boundary of occurrence Then, based on the correspondence of the results right Filter the internal sampling points and match the object fields with... Consistent and the result field is the same as Consistent device output sampling points are limited to the device response range. Connect object fields with Consistent and the result field is the same as Consistent OBD status sampling points and detection curve sampling points are defined as the vehicle response range. Connect object fields with Consistent and the result field is the same as Consistent inspection records are limited to sampling points within the corresponding range of records. ;

[0077] Discretized sequence of workstation video When the boundary occurs Internal input video parsing model ,in, This represents the parameters of the video parsing model, which are applied to each video parsing model. Output action labels and operation object labels, and form an action information vector based on the action labels and operation object labels. ,in, Indicates the situation in which the action occurs, when Memory exists with Take consistent action labels Otherwise take , Indicates the duration of an action, taking or taking. The number of consecutive sampling points corresponding to consistent action labels This indicates the action coverage status, taking the number of required action units and the operation content that have appeared. The ratio of the total number of required motion units, with the denominator being Time to take , This represents the number of actions, specifically the number of transitions where the action label changes from not occurring to occurring. This indicates the position of the action sequence, taking the index of the first matching action tag in the current standard step as its ranking position within the index of the first matching actions in all standard steps. This indicates the matching status of the operation object, taking the operation object label and... The ratio of the number of consistent sampling points to the number of sampling points where the action occurs, with the denominator being... Time to take , This indicates the workstation status, showing the personnel's position within the standard procedure. The number of sampling points in the specified workstation area. Indicates the end status of an action; when the action end label is in... Take when it appears before Otherwise take ;

[0078] The device outputs a discrete sequence. Within the device response range Parse by field to form a device response information vector. ,in, To indicate the device startup status, retrieve the status value from the Device Startup field. To indicate the measurement status, retrieve the status value from the measurement status field. To indicate the sampling status, retrieve the status value from the sampling status field. This indicates the status of the code, retrieving the status value of whether a valid code exists in the code field. To indicate the changes in output, the numerical output field is used. The difference between the last sampled value and the first sampled value. To indicate the device's completion status, the status value of the completion field is taken, and the OBD status is discretely sequenced. and detection curve discrete sequence Within the range of motor vehicle response Internal joint analysis to form a vehicle response information vector ,in, To indicate changes in rotational speed, the difference between the last sampled value and the first sampled value in the rotational speed field is used. To indicate changes in smoke opacity, the value is the difference between the last sampled value and the first sampled value in the smoke opacity field. This indicates the OBD connection status; the status value of the OBD connection field is retrieved. This indicates the OBD code status; the status value of the OBD code field is retrieved. Indicates the component's response status, taking into account the vehicle's response requirements. Consistent number of component status bits. To indicate the changes in test results, the difference between the last sampled value and the first sampled value in the test result field is taken, and the discrete sequence of test records is used. Record the corresponding range Internally, it is parsed by field to form a record information vector. ,in, This indicates the trigger status of the record, retrieving the status value of the record's trigger field. To indicate the completeness of fields, take the ratio of the number of required fields filled to the total number of required fields, with the denominator being... Time to take , This indicates the order of data entry, taking the position of the first record written to the index in the current standard step within the overall index of the first record written to the index in all standard steps. This indicates the result filling status; when the result field is consistent with... Take when consistent Otherwise take , This indicates the corresponding situation of the object, when the object field is... Take when consistent Otherwise take , This indicates the corresponding situation of the steps. When the step field is... Take when consistent Otherwise take ;

[0079] A task context information vector is formed based on the vehicle type, fuel type, inspection task type, workstation type, equipment type, and task stage in the inspection task. ,in, Indicates the vehicle type code. Indicates fuel type code, This indicates the code for the inspection task type. Indicates the workstation type code. Indicates the device type code. This represents the task stage encoding, where each encoding is converted to a numerical value from a fixed encoding table, and the step process information is recorded as follows. Step process information refers to information about a single standard step. The process description results are formed by units, and include standard steps. The corresponding action information, equipment response information, vehicle response information, recording information, and task context information are used to characterize the standard procedures. The correspondence between personnel operations, equipment changes, vehicle responses, and record keeping is used to vectorize motion information. Device response information vector Motor vehicle response information vector Record information vector and task context information vector Follow the same standard steps Perform association and merging to form a step-by-step process information vector. Step-by-step process information vector The input vector corresponds to a single standard step, consisting of 8 dimensions for action information, 6 dimensions for equipment response information, 6 dimensions for vehicle response information, 6 dimensions for record information, and 6 dimensions for task context information. The KAN work evidence network receives the step process information vectors one by one according to the standard step sequence number. The input layer receives 32-dimensional inputs, the action evidence layer receives 8-dimensional action information and 6-dimensional task context information, totaling 14-dimensional inputs, the device association layer receives the output of the action evidence layer and 6-dimensional device response information, the vehicle association layer receives the output of the device association layer and 6-dimensional vehicle response information, the record closure layer receives the output of the vehicle association layer and 6-dimensional record information, and the output layer outputs execution sufficiency, sequence consistency, result correspondence, and record consistency.

[0080] In this embodiment, step S3 specifically includes:

[0081] The inspection task identifier is recorded as The capability correspondence of the testing steps is denoted as: From the perspective of the correspondence between the testing steps and capabilities Read the current standard steps Step sequence Correspondence between results A single standard step The process information is recorded as follows ,in, Represents a 32-dimensional real vector space. The action information consists of eight components: action occurrence marker, action duration, action coverage, number of actions, action sequence position, target object matching quantity, workstation dwell time, and action completion status. The device response information consists of six components: device startup status, measurement status, sampling status, code reporting status, output change, and device completion status. This indicates the vehicle's response information, with six components in order: change in engine speed, change in smoke opacity, OBD connection status, OBD code status, component response status, and change in detection result. The information recorded consists of six components: record trigger status, field completeness, entry order, result entry status, object correspondence, and step correspondence. The task context information consists of six components: vehicle type, fuel type, inspection task type, workstation type, equipment type, and task stage. When calculating the current standard step, only the step process information corresponding to the current standard step is considered. The input to the KAN work evidence network will not include the process information corresponding to other standard steps in the current input, thus satisfying the requirements. The standard steps are denoted as the preceding standard steps, which satisfy... The standard steps are denoted as subsequent standard steps. A topological sort is performed on the current standard step and all standard steps that have a step order relationship with the current standard step to obtain the required order position of each standard step. The lower bound of the sequence is determined by the order of the preceding standard steps. Specifically, take the maximum value of the order position of the preceding standard steps and add it to the standard steps. If no preceding standard steps are available, take The upper bound of the sequence is determined by the order of the subsequent standard steps. Specifically, take the minimum value of the order of the subsequent standard steps and subtract it from the minimum value. If no subsequent standard steps exist, the total number of standard steps for the current testing task is taken, based on the result correspondence. Device response information Motor vehicle response information and record information Boundary constraints are applied, and only components whose object fields are consistent with the operation object and whose result fields are consistent with the corresponding result are allowed to enter the corresponding layer for calculation.

[0082] The operation content of the current standard procedure is coded as follows: The equipment usage requirement code is denoted as The code for motor vehicle response requirements is denoted as The record filling requirements are coded as follows: The standard step type is denoted as ,in, From current standard procedures Read from the operation rule field. , From the correspondence of results Read from the required fields. From current standard procedures Read from the record rule field, when Indicates manual visual inspection actions, This indicates that there are no equipment usage requirements. Indicates no motor vehicle response requirement and When indicating the requirements for manual record filling, Determined as a standard procedure for manual visual inspection, when Indicates OBD connection or OBD read action, Indicates the requirements for using OBD devices, Indicates the OBD status response requirements and When indicating the requirements for filling out OBD records, Determined as a standard procedure for OBD inspection, when Indicates free acceleration operation, Indicates the usage requirements of smoke opacity testing equipment, Indicates the requirements for changes in rotational speed or smoke opacity and When indicating the requirements for filling out the free acceleration method test record, This was determined to be the standard procedure for free acceleration testing.

[0083] KAN Work Evidence Network provides step-by-step information for each standard step. As input, data is passed sequentially through the action evidence layer, equipment association layer, vehicle association layer, record closure layer, and output layer. The input layer receives 32-dimensional input, and the action evidence layer receives... A 14-dimensional input is used, with six KAN neurons, to output a 6-dimensional action evidence vector. Device association layer receives The twelve-dimensional input is configured with six KAN neurons, and the output is a six-dimensional device-related evidence vector. Motor vehicle association layer receives The twelve-dimensional input is configured with six KAN neurons, and the output is a six-dimensional vector of evidence related to motor vehicles. Record closed layer reception A twelve-dimensional input is used, with four KAN neurons, to output a four-dimensional closed relation. The KAN operational evidence network follows standard procedure types between the action evidence layer, equipment association layer, vehicle association layer, and record closure layer. Choose the evidence branch without changing the input layer, the output dimensions of each layer, or the output layer structure. When performing standard procedures for manual visual inspection, select the visual evidence branch; the action evidence layer only applies to... , , and The input side function mapping is executed separately. The four components correspond to the action occurrence, action coverage, operation object matching, and workstation dwell time, respectively. The closure layer is recorded only for... , , and The input edge function mapping is executed separately, with the four components corresponding to the record triggering status, field completeness status, result filling status, and step correspondence status, respectively. The device association layer also includes device response information. The corresponding input edge function unit has a fixed output. Motor vehicle association layer and motor vehicle response information The corresponding input edge function unit has a fixed output. Only perform identity transfer on the upper-level evidence, making the continuous mapping main path a correspondence between action information and record information, when When performing standard OBD inspection procedures, select the OBD evidence branch and the device association layer. , and The input edge function mapping is executed separately, with the three components corresponding to the device startup status, code reporting status, and output change status, respectively. The vehicle association layer... , and The input side function mapping is executed separately, with the three components corresponding sequentially to OBD connection status, OBD code status, and component response status. This ensures that the continuous mapping main path follows a sequential correspondence between action information, device response information, vehicle response information, and record information. When performing standard procedures for free acceleration testing, select the accelerated evidence branch, and configure the device association layer. , and The input side function mapping is executed separately, with the three components corresponding to the measurement, sampling, and output changes, respectively. The motor vehicle association layer... , and The input side function mapping is performed separately, with the three components corresponding to changes in rotational speed, smoke opacity, and detection results, respectively. The closed layer is recorded. and Input side function mapping is performed separately, with the two components corresponding sequentially to the result filling status and the step correspondence status, making the continuous mapping main path a sequential correspondence between action information, equipment response information, vehicle response information, and record information. Each KAN neuron contains an input side function unit and a node summation unit. The input side function unit performs univariate piecewise mapping on a single input dimension, specifically reading the mapping values ​​of adjacent nodes according to the interval of the current input value and performing linear interpolation. The node summation unit sums the outputs of all activated input side function units of the neuron to form the output of the KAN neuron. The output layer... The four components are read out one by one to obtain the initial execution sufficiency readout value. Initial sequence consistency read value Initial result correspondence reading value Consistency reading with initial record ;

[0084] Read the initial execution sufficiency value, initial sequence consistency value, initial result correspondence value, initial record consistency value, and action order position. Fill in the order of appearance. lower bound of order and order upper bound Together, they are used to generate evidence of execution adequacy, sequence consistency, result correspondence, record consistency, and standard procedure closure, and are calculated using the following formula:

[0085] ;

[0086] In the formula, Indicates the standard step number. This represents the one-to-one readout value of the first output neuron in the output layer for the first closed relation quantity. This represents the one-to-one readout value of the second output neuron in the output layer for the second closed relation quantity. This represents the one-to-one readout value of the third output neuron in the output layer for the third closed relation quantity. This represents the one-to-one readout value of the fourth output neuron in the output layer for the fourth closed relation quantity. Indicates action information The fifth dimension of the action sequence position, Indicates recorded information The third dimension should be filled in according to the order of position. This indicates the lower bound of the allowed sequence of steps in the current standard procedure. This indicates the upper bound of the allowed sequence of steps in the current standard procedure. Indicates the adequacy of implementation. Indicates the degree of order consistency. Indicates the degree of correspondence of the results. Indicates the consistency of records. This indicates the standard procedure closure evidence representing the current standard procedure. This indicates the operation of taking the larger value. This indicates that the smaller value is taken in the calculation. All quantities involved in the calculation are dimensionless values. , , and The digit is a positive integer. , , , , , , and The range of values ​​is ,when At that time, , , and Set as ;

[0087] When using the above branching method, the standard steps for manual visual inspection, while maintaining the network hierarchy and output dimension, use the correspondence between action information and recorded information as the main continuous mapping path. The standard steps for OBD inspection use the sequential correspondence between action information, equipment response information, vehicle response information, and recorded information as the main continuous mapping path. The standard steps for free acceleration method detection also use the sequential correspondence between action information, equipment response information, vehicle response information, and recorded information as the main continuous mapping path. The step process information corresponding to each standard step is then mapped. Input the standard step sequence numbers into the KAN operation evidence network one by one to generate standard step closure evidence for each standard step. .

[0088] In this embodiment, step S4 specifically includes:

[0089] The inspection task identifier is recorded as The knowledge graph of motor vehicle inspection capabilities is denoted as ,in, Represents a set of nodes. Represents a set of relational items, denoted as in the current standard procedure. The corresponding micro-skill is recorded as The order of the steps is denoted as The corresponding relationship is denoted as The record filling requirements are recorded as follows: The standard procedure closure evidence is denoted as: ,in, Indicates the adequacy of implementation. Indicates the degree of order consistency. Indicates the degree of correspondence of the results. Indicates record consistency, according to standard procedures Corresponding micro-skills The combined encoding is used as the retrieval key from the set of relational items. In determining the set of relational items corresponding to the current standard step Among them, the execution relationship item is denoted as The order of steps is denoted as The corresponding terms are denoted as Record the corresponding relationship items as , , , , These represent the relational reasoning thresholds for the four relational terms. , , , These represent the support states of the four relation terms, and their values ​​belong to... ,Will , , and Write separately , , and The evidence points are not summed, averaged, or compressed into a single score;

[0090] Device response information is recorded as The six components represent, in order, the device startup status, measurement status, sampling status, code reporting status, output change, and device completion status. The vehicle response information is denoted as... The six components represent, in order, the change in rotational speed, the change in smoke opacity, the OBD connection status, the OBD code status, the component response status, and the change in detection results. The task context information is denoted as... The six components represent, in order, vehicle type, fuel type, inspection task type, workstation type, equipment type, and task stage. The equipment source is distinguished by the following marker: The vehicle origin differentiation mark is recorded as follows: The task context source is distinguished by the following markers: From the correspondence of results Read device response template ,in, , , , and For state coding, and These are the lower bound and the upper bound of the output change, respectively. , , , and Compare each item with the corresponding state code, and... and interval The comparison shows that there exists any instance that does not conform to the season. All are in accordance with the season ,when When, it indicates that the device response information supports the corresponding relationship item, and when When this occurs, it indicates that the device response information does not support the establishment of the corresponding result relationship item. (This is based on the result relationship.) Read vehicle response template ,in, and This represents the boundary of the range of rotational speed variation. and This represents the boundary of the range of smoke opacity changes. , , For state coding, and To define the boundaries of the range of changes in the detection results, , and Compare with the corresponding intervals respectively, and , and Compared with the corresponding state code, there exists any instance that does not conform to the timetable. All are in accordance with the season ,when When, it indicates that the corresponding relationship item of the motor vehicle response information support result is established, when When this occurs, it indicates that the vehicle response information does not support the establishment of the corresponding relationship item in the result, and the task scenario requires the vector to be denoted as... The six components represent, in order, the requirements for vehicle type, fuel type, inspection task type, workstation type, equipment type, and task stage. and Dimension-by-dimensional comparison reveals that any encoding does not conform to the time order. All are in accordance with the season ,when When, it indicates that the task context information supports the task context requirements corresponding to the current standard steps. When this occurs, it indicates that the task context information does not support the task context requirements corresponding to the current standard steps;

[0091] Will and Comparison, satisfaction season Not in season ,Will and Comparison, satisfaction and season Other values ​​are allowed ,Will and Comparison, satisfaction and season Other values ​​are allowed ,Will and Comparison, satisfaction and season Other values ​​are allowed In relational reasoning, the order of steps is considered first. Read from And confirm that the sequence constraints of the current standard steps have been changed by... and If satisfied, then in the result correspondence item Read from And confirm that the result constraints of the current standard procedure have been changed by... and Satisfaction, when , and There are two or three that are simultaneously At that time, instead of judging the causes independently, we will... , and Conflict resolution is achieved by introducing the same relational reasoning chain. A fixed order is followed in the conflict resolution process: first, it checks whether the device response information supports the validity of the corresponding relational item in the result; then... At that time, directly attributing the closed break point to equipment factors is only... At that time, continue to check whether the vehicle response information supports the establishment of the corresponding relationship item in the result. At that time, attributing the closed breakpoint to motor vehicle factors only applies to... and At that time, continue to check whether the task context information supports the task context requirements corresponding to the current standard steps. At that time, the closure breakpoint is attributed to task context factors, only when... , and Only when the closed breakpoint is attributed to human factors, when the equipment source differentiation mark, the motor vehicle source differentiation mark, and the task context source differentiation mark are all present, or when any two of them are present, the closed breakpoint is determined in the order of the equipment source differentiation mark, the motor vehicle source differentiation mark, and the task context source differentiation mark.

[0092] Based on the above-mentioned sequence constraints, result constraints, and conflict resolution rules, the following formula is used to generate the result of the steps being valid. :

[0093] ;

[0094] In the formula, This indicates the result of the step being true, in numerical form. This indicates that the standard procedure is valid, and the numerical value... This indicates that the standard procedure involving equipment factors is invalid, and the numerical value... The standard procedure involving motor vehicle factors is invalid; the numerical value... This indicates that the standard steps involving task context factors are invalid, and the numerical values... This indicates that the standard procedure involving human factors is invalid; in this formula, the term... Corresponding device priority determination rules, item The corresponding motor vehicle determination rule under the condition that the corresponding relationship item of the equipment response information support result is met, item The corresponding task scenario determination rule is based on the condition that both the device response information and the vehicle response information support the establishment of a corresponding relationship between the results. In accordance with the personnel determination rules when none of the aforementioned determinations are met, Compared with current standard procedures and corresponding micro-skills Maintain consistent associations when writing into the knowledge graph of motor vehicle inspection capabilities. .

[0095] In this embodiment, step S5 specifically includes:

[0096] The inspection task identifier is recorded as The total number of standard steps is denoted as The standard steps are denoted as ,in , with standard procedures The corresponding micro-skills are denoted as The result of the current standard procedure being true is denoted as ,in numerical values This indicates that the standard procedure is valid, and the numerical value... This indicates that the standard procedure involving equipment factors is invalid, and the numerical value... The standard procedure involving motor vehicle factors is invalid; the numerical value... This indicates that the standard steps involving task context factors are invalid, and the numerical values... The standard procedure involving human factors is invalid; the equipment origin differentiation marker is recorded as follows. The vehicle origin differentiation mark is recorded as follows: The task context source is distinguished by the following markers: The source type is denoted as , where the numerical This indicates that the standard procedure is valid, and the numerical value... Indicates equipment factors, numerical values Indicates motor vehicle factors, numerical values Represents task context factors, numerical values Indicates personnel factors;

[0097] For each standard step Determine the source type one by one ,when At that time, directly Set as ,when At that time, the three source distinction tags are read according to a fixed priority order. At that time, Set as ,when and At that time, Set as ,when , and At that time, Set as ,when , , At that time, Set as When this determination method is adopted, equipment factors, vehicle factors, task context factors, and personnel factors are all obtained from directly readable discrete labels, without introducing new higher-order quantities;

[0098] All Deduplication is performed according to standard steps to form a micro-skill sequence. ,in Indicates inspection task The number of different micro-skills Indicates the first Each micro-skill, targeting each micro-skill Establish a standard procedure index set Then Arranged in the order of standard steps as follows: ,in Micro-skills The corresponding number of standard steps, Micro-skills The corresponding number A standard step index is used to map the step's success result and source type to the corresponding micro-skill, thus forming micro-skills. The steps to establish a set ,in , Steps to establish a set Each element consists of a step-based result and a source type, which can fully represent the success status and source attribution of the micro-skill across all corresponding standard steps.

[0099] Micro skills The ability status is denoted as , where the numerical Indicates a state where the ability is already possessed; the numerical value is... This indicates a shortage of personnel, and the numerical value represents this. Represents a non-personnel-restricted capability state, and the set that holds true for the steps. Perform sequence determination, when All At that time, Set as When the step is true, the set There exists Continue checking the corresponding source type; when at least one exists... At that time, Set as When it does not exist And there exists at least one At that time, Set as ,when and When they occur simultaneously, priority will be given to Set as This is used to clearly indicate that there is already a shortage of personnel for this micro-skill;

[0100] All micro-skill ability states are arranged according to the micro-skill sequence. The sequential aggregation forms the capability state vector corresponding to the testing task. Capability state vector The Each component With micro-skills One-to-one correspondence is used to characterize the existing capability status, insufficient personnel capability status, and non-personnel-limited capability status of each micro-skill in the current testing task.

[0101] In this embodiment, step S6 specifically includes:

[0102] The inspection task identifier is recorded as The total number of standard steps is denoted as The standard steps are denoted as ,in , with standard procedures The corresponding micro-skills are denoted as , will all Deduplication is performed according to standard steps to form a micro-skill sequence. ,in This represents the number of micro-skills, and the ability state vector of each micro-skill is denoted as... ,in numerical values Indicates a state where the ability is already possessed; the numerical value is... This indicates a shortage of personnel, and the numerical value represents this. Indicates the non-personnel-restricted capability state for each standard step. Find the satisfying Micro-skill serial number ,when At that time, the standard steps When determined as a trusted completion step, At that time, the standard steps The step was identified as a personnel deviation. At that time, the standard steps The steps that are identified as limiting deviations are thus used to form a set of reliable completion steps. Personnel Deviation Step Set and the set of limiting deviation steps ;

[0103] For each micro-skill Establish a standard procedure index set And arranged in the order of standard steps as follows ,in Micro-skills The corresponding number of standard steps will be located in And belongs to the set of trusted completion steps. Standard steps merged , will be located And belongs to the set of personnel deviation steps Standard steps merged , will be located And belongs to the set of limiting deviation steps Standard steps merged ,Will The number of steps is denoted as ,Will The number of steps is denoted as ,Will The number of steps is denoted as ,when At that time, micro-skills The completion status is recorded as completed, when When, the deviation is recorded as personnel deviation, when and When this happens, the deviation is recorded as the limit deviation;

[0104] The reliability of standard operation completion is denoted as The range of values ​​is The reduction factor is denoted as ,in In one implementation, take For each micro-skill Calculate the number of valid completed steps. When the completion status is "completed", the number of valid completed steps is taken as... When the deviation is due to personnel error, the number of valid completed steps is taken as follows: When the deviation is a limited deviation, the number of valid completed steps is taken as follows: Sum the number of valid completion steps for all micro-skills, then divide the sum by the total number of standard steps. To obtain the reliability of the standard operation completion of the inspection task. This calculation process uses the standard number of steps simultaneously. And deviation situations, therefore, reliable completion steps, personnel deviation steps, and limit deviation steps are related to Their contributions differed;

[0105] The job suitability threshold is denoted as ,in The source type of the step is denoted as ,in Indicates equipment factors. Indicates motor vehicle factors, Indicates task context factors, Indicating personnel factors, the conclusion is denoted as ,in This indicates a suitability assessment. This indicates the conclusion of the retraining. Indicates the review conclusion, and In comparison, when Not less than At that time, Set as ,when Below And there exists at least one standard step that satisfies At that time, Set as ,when Below And there is no such thing as satisfying The standard steps, and at least one standard step satisfies At that time, Set as .

[0106] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0107] This invention addresses the technical problem of unreliable judgment on the completion of standard operations by motor vehicle inspectors through a closed-loop technical path involving standard procedure constraints, step process information extraction, continuous mapping of the KAN work evidence network, relational reasoning based on a knowledge graph of motor vehicle inspection capabilities, and source differentiation. First, it establishes standard procedures, corresponding micro-skills, step sequence relationships, and result correspondences based on inspection standard data, breaking down the inspection task into independently judgmentable standard procedure units, and defining the extraction boundaries for workstation videos, equipment outputs, OBD status, detection curves, and inspection records. Then, it extracts action information, equipment response information, motor vehicle response information, record information, and task context information around each individual standard procedure, enabling accurate assessment of personnel operations, equipment changes, and other factors. Motor vehicle response and record filling are under the same inspection semantics. Based on this, the KAN operation evidence network is used to continuously map the sequence from action implementation to record closure, generating execution adequacy, sequence consistency, result correspondence, and record consistency. These four pieces of evidence are then written into the motor vehicle inspection capability knowledge graph as standard procedure closure evidence. Further, by combining equipment source differentiation markers, motor vehicle source differentiation markers, and task context source differentiation markers, relational reasoning is performed to determine the standard procedure establishment result and source type. Finally, capability status, standard operation completion credibility, and job suitability conclusion, retraining conclusion, or review conclusion are formed. This transforms the inspection task from multi-source discrete data analysis into a closureable, interpretable, and deliverable operation credibility assessment process.

[0108] Compared to existing solutions that rely on a single data source for judgment, comprehensive scoring output, or evaluation based solely on the final recorded results, this invention makes targeted improvements to the algorithm structure and judgment mechanism to address the business characteristics of motor vehicle inspection sites, which include strong sequential steps, strong result correspondence, and uncertain sources of anomalies. First, standard steps are used as the smallest unit of analysis, avoiding the mixing of process information from different standard steps. Sequence and result boundaries are defined through step order and result correspondence, reducing cross-step interference on evaluation results from the outset. Second, the KAN operational evidence network does not use a uniform path to process all operations. Instead, it selects corresponding evidence branches based on manual visual inspection standard steps, OBD inspection standard steps, and free acceleration method detection standard steps. This ensures that the evidence formation logic under different inspection methods remains consistent with the actual operational mechanism, improving the interpretability of the closed-loop evidence of standard steps. Third, execution sufficiency, sequence consistency, result correspondence, and record consistency are not compressed into single scores. Instead, they are associated with execution relationship items, step order relationship items, result correspondence relationship items, and record correspondence relationship items, respectively. Conflict resolution of equipment factors, vehicle factors, task context factors, and personnel factors is completed within the same relational reasoning chain. This allows the reasons for the failure of standard steps to be distinguished within the business scenario, avoiding misjudging non-personnel-restricted situations as insufficient personnel capabilities, thus supporting subsequent capability mapping and management conclusion output.

Claims

1. A knowledge graph-based method for evaluating the credibility of motor vehicle inspection personnel's job completion, characterized in that, include: S1. Acquire workstation video, equipment output, OBD status, detection curve, inspection record and inspection standard data. Based on the inspection standard data, determine the standard steps of the inspection task, corresponding micro-skills, the sequence of steps and the correspondence of results, and form the capability correspondence of inspection steps. S2. Based on the capability correspondence of inspection steps, extract action information, equipment response information, vehicle response information, record information and task context information corresponding to each standard step from workstation video, equipment output, OBD status, detection curve and inspection record to form step process information; S3. Input the process information into the KAN work evidence network. Within the standard steps defined by the capability correspondence of the verification steps, perform continuous mapping according to the order of action occurrence, step sequence, equipment response, vehicle response and record correspondence, so that the operation behavior, equipment change, vehicle response and record result of the same standard step form a closed chain. Calculate the execution sufficiency, sequence consistency, result correspondence and record consistency, and generate standard step closed evidence for each standard step. S4. Link the standard procedure closure evidence to the relation items corresponding to the standard procedure and micro-skill in the knowledge graph of motor vehicle inspection capabilities, and generate source differentiation tags based on the equipment response information, motor vehicle response information and task context information in the procedure process information. Under the constraints of the step sequence relationship and the result correspondence relationship, perform relational reasoning on the standard procedure closure evidence and source differentiation tags to form the result of the procedure being valid. S5. Determine the ability status based on the results of the steps and the correspondence between the standard steps and micro-skills, and determine the source type based on the source distinction mark; S6. Calculate the reliability of standard operation completion based on the ability status, and form a job suitability conclusion, retraining conclusion, or review conclusion by combining the job suitability threshold and source type.

2. The method for assessing the credibility of motor vehicle inspection personnel's work completion based on knowledge graphs according to claim 1, characterized in that, S1 specifically refers to: Acquire workstation videos, equipment outputs, OBD status, detection curves, inspection records, and inspection standard data corresponding to the same inspection task, and establish task correspondence relationships between workstation videos, equipment outputs, OBD status, detection curves, and inspection records based on the inspection task identifier; The inspection standard data is analyzed step by step to extract the standard steps corresponding to the inspection task. Based on the operation requirements, equipment usage requirements, record filling requirements and result judgment requirements, the operation content, operation objects and corresponding results of each standard step are determined. Based on the operation content and corresponding results of each standard step, determine the micro-skills corresponding to each standard step, establish the step sequence relationship between standard steps according to the operation sequence in the test standard data, and establish the result correspondence relationship between standard steps and corresponding results according to the corresponding results. The standard steps, corresponding micro-skills, step sequence relationships, and result correspondences are linked according to the standard steps to form a capability correspondence for the verification steps. This capability correspondence is then used as a constraint on the information extraction path for the step process and the scope of evidence generation for the standard steps.

3. The method for assessing the credibility of motor vehicle inspection personnel's work completion based on knowledge graphs according to claim 1, characterized in that, S2 specifically refers to: Based on the capability correspondence of the inspection steps, the standard steps are used as the benchmark for extracting process information. The operation content, operation objects, equipment usage requirements, vehicle response requirements and record filling requirements corresponding to each standard step are determined. The occurrence boundaries of each standard step are defined according to the sequence of steps, and the equipment response range, vehicle response range and record corresponding range corresponding to each standard step are defined according to the result correspondence. The workstation video is analyzed according to standard steps to extract the occurrence, duration, coverage, number of times, sequence, matching of the target, workstation dwell time, and end of the action corresponding to the standard steps, forming action information that corresponds one-to-one with the standard steps. The device output is parsed according to standard procedures, and the device startup status, measurement status, sampling status, code status, output change status and device completion status corresponding to the standard procedures are extracted to form device response information that corresponds one-to-one with the standard procedures. The device response information is then limited to the range of results corresponding to the standard procedures. The OBD status and detection curve are analyzed according to standard procedures. The changes in rotational speed, smoke opacity, OBD connection status, OBD code status, component response status, and detection result changes corresponding to the standard procedures are extracted to form motor vehicle response information that corresponds one-to-one with the standard procedures. The motor vehicle response information is then limited to the range of results corresponding to the standard procedures. The inspection records are parsed according to the standard steps, and the triggering conditions, field completeness, filling order, result filling, object correspondence and step correspondence corresponding to the standard steps are extracted to form record information that corresponds one-to-one with the standard steps. Combined with the vehicle type, fuel type, inspection task type, workstation type, equipment type and task stage, task context information corresponding to the standard steps is formed. Action information, equipment response information, vehicle response information, record information, and task context information are associated and merged according to the same standard steps. This ensures that personnel operations, equipment changes, vehicle reactions, and record entries within the same standard steps are under the same verification semantics, forming step process information that allows the KAN operation evidence network to generate standard step closure evidence according to the standard steps.

4. The method for assessing the credibility of motor vehicle inspection personnel's work completion based on knowledge graphs according to claim 1, characterized in that, S3 specifically refers to: Based on the capability correspondence of the inspection steps, the process information of the steps is allocated according to a single standard step, the input range of the current standard step is determined, the process information of different standard steps is not merged into the input, and the sequence boundary between the current standard step and the preceding and subsequent standard steps is defined according to the step sequence relationship. The corresponding boundary of equipment response information, motor vehicle response information and record information is defined according to the result correspondence relationship. The action information and task context information corresponding to the current standard steps are input into the action evidence layer of the KAN task evidence network. KAN neurons are used to map the input side functions of each input dimension, and then node summation is used to form action evidence, so that the action information is transformed into operation implementation evidence within the scope of the current standard steps, and the basis for calculating execution sufficiency is established. The action evidence and equipment response information are input into the equipment association layer of the KAN work evidence network. Continuous mapping is performed within the current standard steps to form equipment association evidence that represents the correspondence between action implementation and equipment changes. The order of actions is constrained by the sequence of steps to establish the basis for calculating the degree of sequence consistency. The equipment-related evidence and vehicle response information are input into the vehicle-related layer of the KAN operation evidence network. Continuous mapping is performed within the current standard procedure range to form vehicle-related evidence that characterizes the correspondence between equipment changes and vehicle responses. The calculation basis for the result correspondence degree is established by combining the result correspondence. Motor vehicle-related evidence and record information are input into the record closure layer of the KAN operation evidence network, and continuous mapping is performed within the current standard procedure range to form a closure relationship quantity that represents the closure relationship between record filling and the implementation of preceding actions, equipment changes and motor vehicle reactions, and to establish the basis for calculating record consistency. The closed relation quantity is input into the output layer of the KAN operation evidence network, and the execution sufficiency, sequence consistency, result correspondence and record consistency are output one by one, so that the four outputs correspond to the action implementation, step sequence, result correspondence and record correspondence of the current standard step, respectively. The sufficiency of execution, the consistency of sequence, the correspondence of results, and the consistency of records are combined to form the standard procedure closure evidence for the current standard procedure. Among them, the action information must not bypass the equipment response information and the vehicle response information to form the standard procedure closure evidence, and the record information is based on the vehicle-related evidence to form the standard procedure closure evidence. The standard procedure closure evidence for each standard procedure is generated separately.

5. The knowledge graph-based method for assessing the credibility of motor vehicle inspection personnel's work completion according to claim 1, characterized in that, S4 specifically refers to: Based on the standard steps, corresponding micro-skills, step sequence relationships, and result correspondence relationships, the relationship items corresponding to the current standard steps are determined in the knowledge graph of motor vehicle inspection capabilities. The relationship items include execution relationship items that represent the execution requirements of the standard steps, step sequence relationship items that represent the order requirements of the standard steps, result correspondence relationship items that represent the result requirements of the standard steps, and record correspondence relationship items that represent the recording requirements of the standard steps. The sufficiency of execution, consistency of sequence, correspondence of results, and consistency of records of the current standard procedures are respectively associated with execution relationship items, step sequence relationship items, result correspondence relationship items, and record correspondence relationship items, so that the closed evidence of the standard procedures is written into the knowledge graph of motor vehicle inspection capabilities one by one according to the relationship items, and the four pieces of evidence are not compressed into a single score. A device source identification mark is generated based on the conformity of the correspondence between the device response information and the result within the specified range; a vehicle source identification mark is generated based on the conformity of the correspondence between the vehicle response information and the result within the specified range; and a task context source identification mark is generated based on the conformity of the task context information with the task context requirements corresponding to the current standard steps. For the standard steps of the evidence written into the relation items, perform relational reasoning. First, determine whether the execution sufficiency supports the implementation of the action of the current standard step based on the execution relation item. Then, determine whether the sequence consistency supports the sequence of the current standard steps based on the step sequence relation item. Next, determine whether the result correspondence supports the result of the current standard step based on the result correspondence relation item. Finally, determine whether the record consistency supports the record of the current standard step based on the record correspondence relation item. When the execution sufficiency, sequence consistency, result correspondence, or record consistency do not support the establishment of the corresponding relationship item, the equipment source differentiation mark, the motor vehicle source differentiation mark, and the task context source differentiation mark are introduced into the same relational reasoning chain to determine whether the closing breakpoint of the current standard step belongs to the equipment factor, the motor vehicle factor, or the task context factor. If the equipment source differentiation mark, the motor vehicle source differentiation mark, and the task context source differentiation mark are all not established, the closing breakpoint is determined to belong to the personnel factor. Based on the relational reasoning results, the current standard step is established. The established step results include the standard step being established and the standard step being not established with source differentiation. The established step results are then consistently associated with the current standard step and the corresponding micro-skill.

6. The method for assessing the credibility of motor vehicle inspection personnel's work completion based on knowledge graphs according to claim 1, characterized in that, S5 specifically refers to: The source type of each standard step is determined based on the equipment source differentiation mark, the motor vehicle source differentiation mark, and the task situation source differentiation mark. When the equipment source differentiation mark is true, the source type is determined to be equipment factor. When the motor vehicle source differentiation mark is true, the source type is determined to be motor vehicle factor. When the task situation source differentiation mark is true, the source type is determined to be task situation factor. When none of the aforementioned source differentiation marks are true and the standard step is not true, the source type is determined to be personnel factor. Based on the correspondence between standard steps and micro-skills, the results and source types of each standard step are mapped to the corresponding micro-skills, forming a set of steps corresponding to each micro-skill; The set of steps corresponding to each micro-skill is determined. When all the steps in the set of steps are the standard steps, the micro-skill is determined to be in a state where it has the ability. When a standard step is not met in the set of corresponding steps and the source type is personnel factors, the micro-skill is determined to be a personnel-deficient capability state. When a standard step is not met in the set of corresponding steps and the source type is equipment factors, vehicle factors, or task situation factors, the micro-skill is determined to be a non-personnel-restricted capability state. The judgment results of each micro-skill are then aggregated into the capability state corresponding to the test task.

7. The method for assessing the credibility of motor vehicle inspection personnel's work completion based on knowledge graphs according to claim 1, characterized in that, Step S6 is as follows: Based on the capability status, the standard steps corresponding to the capability status that is already available are determined as reliable completion steps, the standard steps corresponding to the capability status that is insufficient in personnel are determined as personnel deviation steps, and the standard steps corresponding to the capability status that is not limited by personnel are determined as limiting deviation steps. Based on the correspondence between standard procedures and micro-skills, micro-skills are merged for reliable completion steps, personnel deviation steps, and limiting deviation steps to determine the completion status and deviation status of each micro-skill in the inspection task. Based on the completion status and deviation of each micro-skill, and combined with the number of standard steps corresponding to each micro-skill, the reliability of the standard operation completion of the inspection task is calculated, so that the reliability of the standard operation completion reflects both the validity of the standard steps and the source differentiation. The reliability of standard operation completion is compared with the job suitability threshold, and the conclusion is determined in combination with the source type. When the reliability of standard operation completion is not lower than the job suitability threshold, a job suitability conclusion is formed. When the reliability of standard operation completion is lower than the job suitability threshold and the source type is personnel factors, a retraining conclusion is formed. When the reliability of standard operation completion is lower than the job suitability threshold and the source type is equipment factors, vehicle factors, or task situation factors, a review conclusion is formed.

8. The knowledge graph-based method for evaluating the credibility of motor vehicle inspection personnel's work completion according to claim 4, characterized in that, The KAN operational evidence network determines the standard step type based on the operational content, equipment usage requirements, vehicle response requirements, and record-keeping requirements of the current standard steps. It then selects an evidence branch corresponding to the standard step type among the action evidence layer, equipment association layer, vehicle association layer, and record closure layer. Specifically, for manual visual inspection standard steps, the visual evidence branch uses the correspondence between action information and record information as the continuous mapping main path; for OBD inspection standard steps, the OBD evidence branch uses the sequential correspondence between action information, equipment response information, vehicle response information, and record information as the continuous mapping main path; and for free acceleration detection standard steps, the acceleration evidence branch uses the sequential correspondence between action information, equipment response information, vehicle response information, and record information as the continuous mapping main path.

9. The method for assessing the credibility of motor vehicle inspection personnel's work completion based on knowledge graphs according to claim 8, characterized in that, The KAN operational evidence network employs different input side function mapping rules for different evidence branches. In the appearance evidence branch, the action evidence layer maps input side functions to the occurrence of actions, action coverage, matching of operating objects, and workstation dwell time; the record closure layer maps input side functions to the triggering of records, field completeness, result filling, and step correspondence. In the OBD evidence branch, the device association layer maps input side functions to the device startup, code reporting, and output changes; the vehicle association layer maps input side functions to the OBD connection, OBD code reporting, and component response. In the acceleration evidence branch, the device association layer maps input side functions to the measurement, sampling, and output changes; the vehicle association layer maps input side functions to the speed change, smoke opacity change, and detection result change; and the record closure layer maps input side functions to the result filling and step correspondence.

10. The knowledge graph-based method for assessing the credibility of motor vehicle inspection personnel's work completion according to claim 5, characterized in that, When two or three of the equipment source differentiation markers, vehicle source differentiation markers, and task context source differentiation markers are simultaneously valid, the source differentiation markers are introduced into the same relational reasoning chain for conflict resolution. The attribution of the closed breakpoint is determined sequentially according to the order of equipment source differentiation markers, vehicle source differentiation markers, and task context source differentiation markers. Specifically, if the equipment response information does not support the establishment of the result corresponding relationship item, it is preferentially determined to be an equipment factor. If the equipment response information supports the establishment of the result corresponding relationship item but the vehicle response information does not support the establishment of the result corresponding relationship item, it is determined to be a vehicle factor. If both the equipment response information and the vehicle response information support the establishment of the result corresponding relationship item but the task context information does not support the task context requirements corresponding to the current standard step, it is determined to be a task context factor. If none of the above determinations are valid, it is determined to be a personnel factor.

Images