A Method and System for Vehicle Damage Quantification Based on Damage Propagation Maps

Abstract

This invention provides a method and system for quantifying vehicle damage based on a damage propagation graph, relating to the fields of data processing and knowledge graph technology. The method includes: constructing a damage propagation graph based on vehicle structural functional zoning and mechanical transmission logic; establishing a quantitative coding system for converting unstructured damage descriptions into numerical features; using this coding system to convert damage descriptions from multiple sources into numerical features, and resolving conflicts by using the damage propagation graph for logical consistency verification; and finally, classifying the damage based on the fused numerical features and an established hierarchical mechanism based on structural safety and repair reversibility. This application, by introducing a damage propagation graph for logical verification and deep reasoning, can significantly improve the standardization and accuracy of damage features and make damage assessment more closely reflect the actual safety status of the vehicle.

Description

Technical Field

[0001] This invention relates to the fields of data processing and knowledge graph technology, specifically to a method and system for quantifying vehicle damage based on damage propagation graphs. Background Technology

[0002] In fields such as vehicle assessment and insurance claims, automated and accurate processing of vehicle damage data is crucial. Existing technologies typically employ manual annotation or keyword matching to extract damage information from heterogeneous data sources such as repair work orders and inspection reports. In some structural engineering fields, such as bridge health monitoring, existing technologies have attempted to use knowledge graphs to fuse and evaluate multi-source data; however, these existing solutions still have significant shortcomings when applied to the field of vehicle damage.

[0003] First, there is a lack of a unified structured coding system. Damage data from different sources have varying formats and semantic ambiguity, making it difficult for downstream models to use them directly, resulting in high feature preprocessing costs and poor consistency.

[0004] Secondly, existing technologies typically address damage to individual components in isolation, neglecting the fact that a vehicle is a complex system, and that mechanical transmission and fault correlation exist between its internal components. For example, a frontal collision may trigger deformation of deep structural parts, and ignoring this cascading damage will lead to a serious misjudgment of the vehicle's safety status.

[0005] Furthermore, when integrating data from various sources such as different repair and testing organizations, descriptive conflicts often arise. For example, the same damage may be described as "sheet metal repair" or "cutting and replacing." A simple strategy of relying on a single source lacks an effective conflict resolution mechanism and can easily lead to data distortion in the final result.

[0006] Finally, the assessment criteria for damage severity are too simplistic, relying heavily on repair costs or the area of ​​visual damage. They fail to distinguish the essential differences between damage to body panels and structural components from the perspectives of structural safety and reversibility of repair, resulting in a low correlation between the assessment results and the vehicle's actual safety performance and residual value loss. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for quantifying vehicle damage based on damage propagation maps.

[0008] A method for quantifying vehicle damage based on damage propagation maps according to the present invention includes: Construct a corresponding damage propagation map based on the functional zoning and mechanical conduction logic of the vehicle structure; Obtain an unstructured damage description of the vehicle; Damage descriptions from multiple sources are transformed into numerical features through quantization encoding, and the numerical features are fused and conflict resolved. When there is a conflict among the numerical features, the logical consistency of the conflicting numerical features is checked using the damage propagation map to resolve the conflict. Furthermore, based on the numerical characteristics after fusion and conflict resolution, the damage is classified through an automatic damage severity classification mechanism based on structural safety and repair reversibility.

[0009] Furthermore, methods for constructing damage propagation maps include: The vehicle is divided into at least two functional areas from the following categories: body panel area, structural component area, chassis area, and electrical area. The core components within each functional area are used as nodes in the damage propagation map, and propagation edges are established between the nodes according to the mechanical conduction logic.

[0010] Furthermore, methods for integration and conflict resolution also include: Damage descriptions from multiple sources are aligned based on a common benchmark; Preset credibility weights for different categories of data sources; By combining the credibility weight and the results of the logical consistency verification, the final numerical features are determined to achieve the conflict resolution.

[0011] Furthermore, the method for determining conflicting numerical features is as follows: When the encoding difference of the aligned damage descriptions from different data sources in a preset dimension exceeds a preset threshold, the damage description is determined to be conflicting data.

[0012] Furthermore, the quantitative coding adopts a three-dimensional coding system, which includes a damage type dimension, a repair method dimension, and an impact range dimension.

[0013] A vehicle damage quantification system based on damage propagation mapping provided by the present invention includes: Construct a corresponding damage propagation map based on the functional zoning and mechanical conduction logic of the vehicle structure; Obtain an unstructured damage description of the vehicle; Damage descriptions from multiple sources are transformed into numerical features through quantization encoding, and the numerical features are fused and conflict resolved. When there is a conflict among the numerical features, the logical consistency of the conflicting numerical features is checked using the damage propagation map to resolve the conflict. Furthermore, based on the numerical characteristics after fusion and conflict resolution, the damage is classified through an automatic damage severity classification mechanism based on structural safety and repair reversibility.

[0014] Furthermore, methods for constructing damage propagation maps include: The vehicle is divided into at least two functional areas from the following categories: body panel area, structural component area, chassis area, and electrical area. The core components within each functional area are used as nodes in the damage propagation map, and propagation edges are established between the nodes according to the mechanical conduction logic.

[0015] Furthermore, methods for integration and conflict resolution also include: Damage descriptions from multiple sources are aligned based on a common benchmark; Preset credibility weights for different categories of data sources; By combining the credibility weight and the results of the logical consistency verification, the final numerical features are determined to achieve the conflict resolution.

[0016] Furthermore, the method for determining conflicting numerical features is as follows: When the encoding difference of the aligned damage descriptions from different data sources in a preset dimension exceeds a preset threshold, the damage description is determined to be conflicting data.

[0017] Furthermore, the quantitative coding adopts a three-dimensional coding system, which includes a damage type dimension, a repair method dimension, and an impact range dimension.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. Improve feature standardization and uniformity: Through an innovative three-dimensional quantization coding system, multi-source, heterogeneous, and unstructured damage descriptions are automatically converted into standardized numerical feature vectors, solving the problem of inconsistent data formats and significantly improving the data processing efficiency and input consistency of downstream models.

[0019] 2. Achieve deep reasoning on damage correlation: By constructing a damage propagation map based on the functional zoning of vehicle structure, prior knowledge of vehicle mechanical transmission is introduced, breaking through the limitation of isolated assessment of single-point damage. It can accurately identify and quantify cascade damage caused by collision force transmission, effectively reducing the missed detection rate of key structural damage.

[0020] 3. Improve the accuracy of multi-source data fusion: By combining the credibility of data sources with the logical verification of damage propagation maps, an intelligent conflict resolution mechanism has been established. When faced with data contradictions from different sources, it can arbitrate based on the physical structure logic of the vehicle, which significantly improves the accuracy and consistency of the fused data features.

[0021] 4. Enhance the accuracy and objectivity of damage assessment: By establishing a multi-dimensional grading rule based on structural safety and repair reversibility, it replaces the traditional single assessment method that relies on repair costs or appearance, making the classification of damage levels closer to the actual safety status and value depreciation of the vehicle, and providing a more reliable quantitative basis for downstream applications such as residual value assessment and insurance loss assessment. Attached Figure Description

[0022] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the structure of a vehicle damage quantification system provided in an embodiment of this application; Figure 2 A flowchart illustrating a method for quantifying vehicle damage provided in an embodiment of this application; Figure 3 A partial structural schematic diagram of a damage propagation pattern provided in an embodiment of this application; Figure 4 This is a timing diagram of multi-source data fusion signaling interaction provided in an embodiment of this application. Detailed Implementation

[0023] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0024] Example 1 This embodiment provides a specific implementation of a vehicle damage quantification method and system based on damage propagation mapping. In this embodiment, the entire process, from receiving multi-source heterogeneous raw damage data to finally outputting structured and standardized vehicle damage features and damage levels, will be fully illustrated with reference to the accompanying drawings.

[0025] Reference Figure 1 This diagram illustrates a preferred structural design of a vehicle damage quantification system according to an embodiment of this application. The system can be deployed on a single server, a server cluster, or in a cloud computing environment. From a hardware perspective, the system includes one or more processors and a memory, wherein the memory stores computer program instructions. When the processor executes these instructions, the methods described in this application can be implemented. From a software functional perspective, the system includes a data input interface 10, a feature extraction engine 20, a damage propagation map database 30, and a data output interface 40.

[0026] Data input interface 10 is configured to receive unstructured or semi-structured vehicle damage data from various external data sources. These data sources can include, but are not limited to, 4S dealership repair management systems, electronic reports from third-party testing agencies, handwritten or electronic work orders from ordinary repair shops, and insurance company survey records. The data format can be plain text, XML, JSON, or other formats.

[0027] The feature extraction engine 20 is the core processing unit of the entire system, and it can further include multiple cooperating modules. Specifically: the encoding conversion module 21 is responsible for parsing the input data, extracting key damage entities (such as damaged parts, repair operations, damage phenomena, etc.) through natural language processing techniques such as named entity recognition, and converting these text descriptions into standardized numerical features according to the quantitative encoding system detailed later; the data fusion module 22 is responsible for processing records from different data sources about the same damage event, performing event alignment and conflict determination, and, when a conflict occurs, performing logical consistency verification through interaction with the damage propagation graph database 30, ultimately generating unified and reliable damage features; and the severity grading module 23 analyzes the unified damage features output by the data fusion module 22 according to preset rules to automatically determine the severity level of each damage and the overall damage level of the vehicle.

[0028] The damage propagation graph database 30 is a pre-built knowledge base used to store damage propagation graphs for specific vehicle models or general vehicle models. This graph, in the form of a graph data structure, represents the physical connections, mechanical transmission paths, and fault association logic between various vehicle components. In one embodiment of this application, the database can be implemented using technologies such as graph databases to facilitate efficient complex graph queries and path analysis.

[0029] The data output interface 40 is configured to output the structured data processed by the feature extraction engine 20 in a predetermined format such as JSON or XML for direct use by downstream applications (such as vehicle valuation models, intelligent damage assessment engines, repair plan recommendation systems, etc.).

[0030] The following will combine Figure 2 The flowchart shown illustrates the specific steps of the vehicle damage quantification method in this embodiment.

[0031] Understandably, a damage propagation map needs to be pre-constructed and loaded before implementing this method. The map construction process may include the following steps: First, perform functional zoning of the vehicle structure. A complete vehicle is divided into at least two, preferably four, main functional areas based on the function and structural characteristics of its components: the body panel area, the structural component area, the chassis area, and the electrical area. The body panel area mainly includes non-load-bearing external components such as front and rear bumpers, door panels, fenders, and the hood; damage to these primarily affects the appearance. The structural component area includes the load-bearing frame of the vehicle body, such as front and rear longitudinal beams, A / B / C pillars, sill beams, and the radiator support frame; damage to these directly affects the vehicle's passive safety. The chassis area includes the suspension system, steering system, and braking system, which relate to the vehicle's handling and driving safety. The electrical area includes all the vehicle's sensors, actuators, control units, and wiring harnesses, which are essential for the normal operation of all vehicle functions.

[0032] Next, define the graph nodes. The core components within each functional area are used as the nodes of the graph. For example, such as... Figure 3 As shown, the "front bumper" can be defined as a node and marked as belonging to the body panel area; the "radiator frame" and "left longitudinal beam" can be defined as nodes and marked as belonging to the structural component area; and the "headlight wiring harness" can be defined as a node and marked as belonging to the electrical area. It should be noted that a typical compact car diagram may contain 60 to 100 core component nodes.

[0033] Based on this, propagation edges and configuration weights are constructed. According to the laws of mechanical transmission in vehicle design and common fault correlation logic, directed propagation edges are established between nodes, their direction indicating the typical transmission direction of damage energy or fault impact. For example, in a frontal collision, force is typically transmitted from the front bumper 110 to the radiator frame 120, and then from the radiator frame 120 to the left longitudinal beam 130. Correspondingly, directed edges can be established from the node "front bumper 110" to "radiator frame 120," and from "radiator frame 120" to "left longitudinal beam 130." Simultaneously, to more precisely describe the propagation relationship, a propagation weight coefficient between 0 and 1 is assigned to each propagation edge to represent the probability or degree of impact of damage being transmitted from the upstream node to the downstream node. Figure 3 As shown, the edge weight from the front bumper 110 to the radiator frame 120 is w=0.7, indicating that when the bumper is damaged, there is a high probability that the radiator frame will be affected; while the edge weight from the radiator frame 120 to the left longitudinal beam 130 is w=0.9, indicating that once the radiator frame deforms, the longitudinal beam behind it will be highly likely to be affected. These weighting coefficients can be determined by analyzing a large amount of historical maintenance data, collision simulation data, or based on the experience and knowledge of maintenance experts.

[0034] After the completed map is stored in the damage propagation map database 30, the damage data received online can be processed, as shown in steps S101 to S108.

[0035] Step S101: Receive Data. The system receives three sets of original damage data for a vehicle (with a unique vehicle identification number) through data input interface 10: Data Source A (4S shop work order, preset confidence weight 0.9): "Front-end collision, front bumper sheet metal and painting, radiator frame correction, left longitudinal beam sheet metal repair." Data Source B (third-party inspection report, preset confidence weight 0.7): "Left front longitudinal beam deformation repair, front bumper replacement." Data Source C (ordinary repair shop work order, preset confidence weight 0.5): "Front bumper scratch polishing."

[0036] Step S102: Encoding Conversion. The encoding conversion module 21 processes each of the above text descriptions, converting them into standardized three-dimensional numerical codes [T, R, S]. The first dimension is the damage type code T, used to identify the root cause of the damage. For example, T=1 represents collision damage, T=2 represents water blister damage, etc. Each type can be associated with a base influence coefficient T_base. In this example, all damage originates from a collision, so T=1, T_base=1.0. The second dimension is the repair method code R, which reflects the degree of impact of the repair operation on the vehicle's structural integrity and the reversibility of the repair. A higher code value indicates more severe damage. An exemplary encoding rule is: R=1 (non-destructive repair, such as polishing); R=2 (sheet metal repair / correction); R=3 (part replacement); R=4 (cutting and welding). The third dimension is the influence range code S, which reflects the spatial range of the damage and the structural levels affected. A higher code value indicates a wider and deeper influence range. An example coding rule is: S=1 (single overlay); S=2 (multiple overlays); S=3 (single structural component); S=4 (multiple structural components); S=5 (cross-functional area influence).

[0037] According to this encoding system, the output of encoding conversion module 21 is as follows: For data source A: "Front bumper sheet metal painting" is converted to [1, 2, 1]; "Radiator frame correction" is converted to [1, 2, 3]; "Left longitudinal beam sheet metal repair" is converted to [1, 2, 3]. For data source B: "Left front longitudinal beam deformation repair" is converted to [1, 2, 3]; "Front bumper replacement" is converted to [1, 3, 1]. For data source C: "Front bumper scratch polishing" is converted to [1, 1, 1].

[0038] Step S103: Event Alignment. The data fusion module 22 aligns these records based on information such as vehicle identification number and accident occurrence time. In this example, it is found that the recording times of data sources A and B are similar and describe the same frontal collision accident; while the recording time of data source C is earlier, and the damage description differs greatly from the severity of A and B. Therefore, it is determined to be an independent historical minor maintenance event and will not participate in this fusion.

[0039] Step S104: Conflict Determination. The data fusion module 22 compares the codes of the aligned data sources A and B. For the "left longitudinal beam," both codes are [1, 2, 3], indicating no conflict. For the "front bumper," data source A's code is [1, 2, 1], and data source B's code is [1, 3, 1], showing a difference in the repair method dimension R (2 vs 3). The system can determine whether a conflict exists based on preset rules. For example, a conflict threshold can be set; when the difference in the code values ​​of a key dimension (such as R) is greater than or equal to 2, it is determined to be a strong conflict. In this example, the difference is |3-2|=1, which is less than the threshold of 2, therefore it is determined to be a minor conflict.

[0040] Step S106: Conflict Resolution. For minor conflicts, the system employs a confidence-weighted strategy for resolution. Since the confidence of data source A (0.9) is higher than that of data source B (0.7), the system accepts the description from data source A. Ultimately, the unified feature for the "front bumper" is determined to be [1, 2, 1]. Simultaneously, the system also includes the damage information [1, 2, 3] of the "water tank frame" and the damage information [1, 2, 3] of the "left longitudinal beam" in the unified damage list.

[0041] Step S105: Logical Verification. As a preferred implementation, even without triggering a strong conflict, the system can still perform a logical consistency check on the fused damage combination. Specifically, the system queries the damage propagation graph database 30 to verify whether the damage chain of "front bumper" -> "water tank frame" -> "left longitudinal beam" exists and is reasonable. For example... Figure 3 As shown, the propagation path is a predefined path with high weight in the graph, so the logical verification passes, thereby enhancing the confidence of the fusion result.

[0042] Step S107: Severity Grading. The severity grading module 23 receives the unified damage feature list and grades it according to preset grading rules. The grading rules are mainly based on structural safety and repair reversibility, for example: Level I (Minor Damage): Only involves the body panel (S=1 or 2), and the repair method is non-destructive repair (R=1). Level II (General Damage): Only involves the body panel (S=1 or 2), but the repair method is sheet metal or replacement (R=2 or 3). Level III (Severe Damage): Involves structural components or chassis parts (S>=3), but the repair method is reversible (e.g., sheet metal correction with R=2). Level IV (Structural Damage): Involves structural components or chassis parts (S>=3), and the repair method is irreversible (e.g., cutting and welding with R=4).

[0043] Applying this rule, for the damage to the "front bumper" [1, 2, 1], since S=1 and R=2, it meets the definition of Class II (general damage); for the damage to the "radiator support" [1, 2, 3] and the damage to the "left longitudinal beam" [1, 2, 3], since S=3 and R=2, both meet the definition of Class III (serious damage). Accordingly, the final damage level of the entire vehicle is the highest level among all damage items, i.e., Class III (serious damage).

[0044] Step S108: Output Features. Finally, the system outputs a structured JSON report through data output interface 40. This report includes the component name, 3D code, corresponding damage level, and overall damage level of the entire vehicle for each damage item, and optionally includes the calculated overall damage score. This report can be directly used by downstream valuation models to accurately reflect that the vehicle has serious damage requiring attention, but has not yet reached the most severe structural damage level.

[0045] Example 2 As an optional implementation, this embodiment aims to highlight how the method proposed in this application utilizes damage propagation maps for intelligent logical arbitration when faced with severe conflicts in multi-source data, thereby ensuring the robustness and accuracy of the final output. The system architecture and basic process of this embodiment are the same as those of Embodiment 1, but the data processed and the internal decision-making logic are different.

[0046] In this embodiment, we assume a scenario of a rear-end collision.

[0047] Step S101: Receive Data. The system receives two data sets regarding the same rear-end collision: Data Source A (Insurance Company Survey Report, Credibility Weight 0.8): "Rear bumper replacement, trunk lid sheet metal work." Data Source B (Small Repair Shop Work Order, Credibility Weight 0.4): "Rear bumper replacement, rear longitudinal beam cutting and welding."

[0048] Steps S102 and S103: Encoding Conversion and Event Alignment. The system encodes and aligns the two sets of data, similar to the process in Example 1. The encoding results for data source A are: "Rear Bumper Replacement" -> [1, 3, 1]; "Trunking Lid Sheet Metal Repair" -> [1, 2, 1]. The encoding results for data source B are: "Rear Bumper Replacement" -> [1, 3, 1]; "Rear Longitudinal Beam Cutting and Welding" -> [1, 4, 3], where "Rear Longitudinal Beam" is a structural component (S=3), and "Cutting and Welding" is an irreversible repair (R=4).

[0049] Step S104: Conflict Determination. The data fusion module 22 compares the two sets of data. For "rear bumper," the descriptions are consistent, and there is no conflict. However, for "rear longitudinal beam," data source A has no record of this damage (which can be considered as a damage level of 0), while data source B records "cutting and welding" (R=4). Understandably, this difference is extremely large, and its encoding difference |4-0|=4 far exceeds the preset conflict threshold (e.g., 2), so the system determines this as a strong conflict.

[0050] Step S105: Logical verification based on damage propagation map. Due to the detection of a strong conflict, the data fusion module 22 initiates the logical verification process. (Refer to...) Figure 4 The diagram illustrates this interaction process in the form of a time sequence. The data fusion module 22, acting as the coordinator, initiates a logical consistency verification request to the damage propagation graph database 30. This request may include the following information: "In a rear-end collision, the upstream component 'rear bumper' is known to have been damaged by 'replacement' (R=3), and the 'trunk cover' is known to have been damaged by 'sheet metal' (R=2). Please verify the logical probability that the downstream component 'rear longitudinal beam' has suffered 'cutting and welding' damage (R=4)."

[0051] A query is performed on the damage propagation database 30. First, the database searches for a propagation path from the rear bumper or trunk lid to the rear longitudinal beam. The query results show that a propagation path does exist, but the propagation weight coefficient w is likely low (e.g., w=0.3). Physically, this means that a very severe collision would be required for the energy to penetrate the rear bumper, rear anti-collision beam, trunk floor, and other structures, ultimately necessitating the cutting and repair of the rear longitudinal beam.

[0052] Next, the data fusion module 22 uses the path and weight information returned by the graph to perform inference. It analyzes the damage level of upstream components: both highly reliable insurance company reports and less reliable repair shop work orders only mention "rear bumper replacement" (R=3) and "trunk cover sheet metal work" (R=2). These damages typically do not correspond to the collision intensity that would cause R=4 level damage to the rear longitudinal beam. A logically consistent scenario should be that if the rear longitudinal beam needs to be cut, then upstream components such as the rear bumper beam and trunk floor should also have records of severe deformation or replacement, but the existing data does not mention severe damage to these intermediate components.

[0053] Therefore, the damage propagation map database 30 (or the inference logic of the data fusion module 22) returns a result of "logic consistency check failed" or "low confidence".

[0054] Step S106: Conflict Resolution. Based on the result of the failed logical verification, the data fusion module 22 makes a decision to mark the record in data source B concerning "rear longitudinal beam cutting and welding" as "high risk, logical inconsistency, pending manual review", and temporarily exclude this record from the automatically generated unified features, or give it a very low confidence score.

[0055] Steps S107 and S108: Grading and Output. The severity grading module 23 grades the damage based on the resolved and confirmed reliable damage information (i.e., "rear bumper replacement" and "trunk cover sheet metal repair"). Both of these damages are body panel damages, therefore the overall vehicle damage level is determined to be Level II (general damage). However, in the final output JSON report, in addition to the Level II rating result, a special warning field will be attached, explicitly indicating that there is a logically questionable serious damage record regarding the "rear longitudinal beam," and recommending manual review or on-vehicle inspection.

[0056] Therefore, the technical solution provided in this application can effectively avoid serious misjudgment of vehicle condition due to a low-reliability but seriously erroneous data (which may be due to recording errors, insurance fraud, or over-repair), thereby ensuring the objectivity and reliability of the final evaluation result.

[0057] Example 3 To better illustrate the technical solution of this application, this embodiment aims to demonstrate the ability of this method to handle cross-functional area related damage, especially in complex scenarios where structural deformation causes electrical system failures.

[0058] In this embodiment, we assume a scenario of a vehicle experiencing a side collision.

[0059] Step S101: Receive data. The system receives a highly reliable 4S store repair work order (reliability 0.9): "Left B-pillar sheet metal correction, left front door replacement, after the accident it was found that the left front door window could not be raised or lowered, and the door wiring harness was repaired after inspection."

[0060] Step S102: Encoding Conversion and Graph Association. The encoding conversion module 21 processes the data. "Left B-pillar sheet metal correction" -> Component: B-pillar (structural component area), encoding [1, 2, 3]. "Left front door replacement" -> Component: Left front door (covering component area), encoding [1, 3, 1]. "Window cannot be raised or lowered", "Repair door wiring harness" -> Specifically, the system can use entity recognition and semantic understanding technology to associate the fault phenomenon "window cannot be raised or lowered" with the specific component "door wiring harness". In the graph, "door wiring harness" belongs to the node in the electrical area.

[0061] Step S105: Propagation Path Analysis. The data fusion module 22 (or performs correlation analysis before grading) analyzes the relationships between these damages. It queries the damage propagation map database 30 for the correlation from "B-pillar" to "door wiring harness". Such cross-functional area propagation edges can be pre-constructed in the map because, in vehicle design, the door wiring harness needs to pass through the A-pillar or B-pillar to connect to the vehicle body. Therefore, there exists a propagation edge in the map pointing from "B-pillar" (structural area) to "door wiring harness" (electrical area), and this edge may have a high weight (e.g., w=0.8) because deformation of the B-pillar can easily squeeze, wear, or cut the internal wiring harness. The query results confirm a strong logical correlation between "B-pillar sheet metal correction" and "door wiring harness repair".

[0062] Step S107: Determination and Classification of Impact Scope. During severity classification, the severity classification module 23 identified that this damage event not only involved the "B-pillar" in the structural component area but also affected the "door wiring harness" in the electrical area. According to the rules of the impact scope coding S, when the damage crosses different functional areas, the value of S should be the highest level, i.e., S=5 (cross-functional area impact).

[0063] Therefore, for the damage to the "B-pillar," although the repair method is only "sheet metal straightening" (R=2), because it caused cross-zone effects, the system will consider the factor of S=5 when assessing its overall impact. According to the classification rules, any damage involving structural components (S>=3) and repaired by sheet metal straightening (R=2) already reaches Level III (severe damage). The confirmation of cross-zone effects further solidifies this rating and can clearly indicate the causal relationship between structural damage and electrical faults in the output report.

[0064] This embodiment demonstrates that the technical solution provided in this application can go beyond simple physical component damage assessment. It can accurately identify and quantify functional failures caused by structural damage through damage propagation mapping, reflecting a deep understanding of the correlation of vehicle damage, thereby avoiding the one-sidedness of the assessment.

[0065] Example 4 This embodiment aims to demonstrate the scalability and flexibility of the method of this application, including the use of finer-grained functional partitioning and the handling of non-collision damage types such as blister damage.

[0066] In this embodiment, the system employs an enhanced damage propagation map and an expanded set of coding rules. On one hand, the functional partitioning during the map construction phase is refined; for example, the "electrical zone" is further divided into "cockpit electrical sub-zone," "engine compartment electrical sub-zone," and "chassis electrical sub-zone," etc., to more accurately pinpoint the extent of damage impact. On the other hand, in the three-dimensional coding system, a new damage type code T=2 is defined for "water blister damage," and it is assigned a higher base impact coefficient than collision damage (e.g., T_base=1.5), because water blister damage typically causes systemic and irreversible damage to the electrical system and interior trim.

[0067] Imagine a scenario where a car is submerged in water.

[0068] Step S101: Receive data. The system receives a test report describing the following: "The vehicle was submerged in water, and the static water level reached the lower edge of the dashboard. Upon inspection, multiple wiring harness connectors in the engine compartment were found to be corroded by water and need to be replaced; multiple body control modules and comfort system control modules under the carpet in the driver's cabin were damaged by water and need to be replaced."

[0069] Steps S102 and S105: Coding and Analysis. The system first identifies the damage type as "water blister damage," therefore the T code for all damage items is 2. Subsequently, "Engine compartment wiring harness connector replacement" (repair method R=3) is located in the "Engine compartment electrical sub-area" through map mapping; "Cockpit control module replacement" (repair method R=3) is located in the "Cockpit electrical sub-area."

[0070] Step S107: Impact Scope Determination and Classification. System analysis revealed that the damage involved two different (sub)functional areas: the "Engine Compartment Electrical Sub-area" and the "Cockpit Electrical Sub-area." Therefore, the impact scope code S was determined to be 5 (cross-functional area impact).

[0071] At this point, a comprehensive damage score is calculated to quantify its severity. For example, the damage score for replacing a control module in the cockpit would be:

[0072] Understandably, this score is much higher than that for a single part replacement resulting from a typical collision (e.g., a front bumper replacement, which scores 1.0 × 3 × 1 = 3), thus accurately reflecting the severity of the blister damage.

[0073] During the classification process, the rule base of the severity classification module 23 can include special rules for flood-damaged vehicles. For example: "Any vehicle with damage type T=2 that involves the replacement of multiple core electrical control modules, regardless of the location of these modules, will be directly classified as Level IV (structural damage)." It should be noted that "structural damage" here is used in a broad sense, referring to damage that has caused fundamental and irreparable harm to the vehicle's core value and safety reliability.

[0074] Therefore, the system ultimately classified the vehicle's damage level as Level IV and clearly stated in the output report that it was a severely flooded vehicle.

[0075] Therefore, the method framework proposed in this application has good scalability. By adjusting the granularity of the map and the coding and grading rules, it can flexibly adapt to the quantitative assessment needs of different damage types (such as collision, water immersion, fire, etc.) to achieve comprehensive and accurate vehicle damage assessment.

[0076] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for quantifying vehicle damage based on damage propagation maps, characterized in that, include: Construct a corresponding damage propagation map based on the functional zoning and mechanical conduction logic of the vehicle structure; Obtain an unstructured damage description of the vehicle; Damage descriptions from multiple sources are transformed into numerical features through quantization encoding, and the numerical features are fused and conflict resolved. When there is a conflict among the numerical features, the logical consistency of the conflicting numerical features is checked using the damage propagation map to resolve the conflict. Furthermore, based on the numerical characteristics after fusion and conflict resolution, the damage is classified through an automatic damage severity classification mechanism based on structural safety and repair reversibility.

2. The vehicle damage quantification method based on damage propagation mapping according to claim 1, characterized in that, Methods for constructing damage propagation maps include: The vehicle is divided into at least two functional areas from the following categories: body panel area, structural component area, chassis area, and electrical area. The core components within each functional area are used as nodes in the damage propagation map, and propagation edges are established between the nodes according to the mechanical conduction logic.

3. The vehicle damage quantification method based on damage propagation mapping according to claim 1, characterized in that, Other ways to integrate and resolve conflicts include: Damage descriptions from multiple sources are aligned based on a common benchmark; Preset credibility weights for different categories of data sources; By combining the credibility weight and the results of the logical consistency verification, the final numerical features are determined to achieve the conflict resolution.

4. The vehicle damage quantification method based on damage propagation mapping according to claim 1, characterized in that, The method for determining conflicting numerical features is as follows: When the encoding difference of the aligned damage descriptions from different data sources in a preset dimension exceeds a preset threshold, the damage description is determined to be conflicting data.

5. The vehicle damage quantification method based on damage propagation mapping according to claim 1, characterized in that, The quantitative coding adopts a three-dimensional coding system, which includes damage type dimension, repair method dimension, and impact range dimension.

6. A vehicle damage quantification system based on damage propagation maps, characterized in that, include: Construct a corresponding damage propagation map based on the functional zoning and mechanical conduction logic of the vehicle structure; Obtain an unstructured damage description of the vehicle; Damage descriptions from multiple sources are transformed into numerical features through quantization encoding, and the numerical features are fused and conflict resolved. When there is a conflict among the numerical features, the logical consistency of the conflicting numerical features is checked using the damage propagation map to resolve the conflict. Furthermore, based on the numerical characteristics after fusion and conflict resolution, the damage is classified through an automatic damage severity classification mechanism based on structural safety and repair reversibility.

7. The vehicle damage quantification system based on damage propagation mapping according to claim 6, characterized in that, Methods for constructing damage propagation maps include: The vehicle is divided into at least two functional areas from the following categories: body panel area, structural component area, chassis area, and electrical area. The core components within each functional area are used as nodes in the damage propagation map, and propagation edges are established between the nodes according to the mechanical conduction logic.

8. The vehicle damage quantification system based on damage propagation mapping according to claim 6, characterized in that, Other ways to integrate and resolve conflicts include: Damage descriptions from multiple sources are aligned based on a common benchmark; Preset credibility weights for different categories of data sources; By combining the credibility weight and the results of the logical consistency verification, the final numerical features are determined to achieve the conflict resolution.

9. The vehicle damage quantification system based on damage propagation mapping according to claim 6, characterized in that, The method for determining conflicting numerical features is as follows: When the encoding difference of the aligned damage descriptions from different data sources in a preset dimension exceeds a preset threshold, the damage description is determined to be conflicting data.

10. The vehicle damage quantification system based on damage propagation mapping according to claim 6, characterized in that, The quantitative coding adopts a three-dimensional coding system, which includes damage type dimension, repair method dimension, and impact range dimension.

Images