A new energy automobile interior part production process data analysis method and system

Abstract

This application discloses a data analysis method and system for the production process of interior parts for new energy vehicles, relating to the field of data analysis technology for the production process of interior parts for new energy vehicles. It aims to solve the technical problem of accurate diagnosis of production anomalies and quality traceability during the production process of interior parts for new energy vehicles. The method includes: updating the dynamic mapping relationship between the position of a single semi-finished interior part and the position of the pallet grid based on position transfer event information; using the single-piece real identity identifier of the single semi-finished interior part in the single-piece flow sequence as an index; dynamically retrieving the injection molding process parameter set uniquely corresponding to the currently processed single semi-finished interior part from the stored historical data; and performing logical consistency verification on the combination of injection molding process parameters and coating process parameters in the single-piece cross-process data set based on preset multiple verification rules that respectively characterize the logical association logic between injection molding and coating processes in different physical dimensions, thereby identifying suspected mis-binding data with logical conflicts.

Description

Technical Field

[0001] This application relates to the field of data analysis technology for the production process of interior parts for new energy vehicles, and in particular to a method and system for data analysis of the production process of interior parts for new energy vehicles. Background Technology

[0002] In the entire lifecycle production process of interior parts for new energy vehicles, in order to achieve production anomaly diagnosis and quality traceability, it is usually necessary to collect equipment operation data, process parameters, and quality inspection data, and establish a stable correspondence between these multi-source heterogeneous data. However, in the actual production process, after the injection molding process, interior parts usually enter a buffer area for cooling and waiting, and then are transferred according to the rhythm of subsequent wrapping and other stations.

[0003] During this process, when products originally arranged in the injection molding sequence are split, merged, or repositioned due to reasons such as waiting for subsequent processes, pallet capacity limitations, or local rework, the physical order of the products changes accordingly. However, existing process records often still follow the original batch or pallet order, resulting in a discrepancy between the physical flow order of the same product and the recorded flow order. This deviation causes the injection molding equipment status record, which should belong to one interior part, to be incorrectly linked to the subsequent wrapping process parameters and test results of another interior part, creating a misalignment between equipment status, process execution, and quality performance.

[0004] Because existing conventional methods typically rely on fixed batch or pallet serial numbers for cross-process data linking, it is difficult to dynamically perceive and update the true attribution of records when individual components undergo cycle time adjustments and position rearrangements. This easily leads to incorrect tracing of problems occurring during wrapping or inspection to the wrong injection molding state. Therefore, in scenarios where semi-finished products enter the buffer after injection molding and undergo pallet splitting and rearrangement, when the order of semi-finished products changes due to cycle time adjustments, subsequent process records are still bound according to the original acceptance order, resulting in mismatches between the injection molding equipment status and the wrapping and inspection results. This severely affects the accuracy of cross-process anomaly diagnosis and source quality traceability for interior parts of new energy vehicles. Summary of the Invention

[0005] This application provides a data analysis method and system for the production process of interior parts for new energy vehicles, aiming to solve the technical problem that, during the production process of interior parts for new energy vehicles, the repositioning of semi-finished products in the buffer area leads to inconsistencies between the physical flow order and the recorded flow order, which in turn causes cross-process data misbinding and seriously affects the accuracy of production anomaly diagnosis and quality traceability.

[0006] Firstly, to address the aforementioned technical problems, this invention provides a data analysis method for the production process of interior parts for new energy vehicles. This method includes: responding to a position reshuffling of individual semi-finished interior parts within a buffer area; acquiring position transfer event information corresponding to the picking and placing actions of individual semi-finished interior parts between pallets using a perception network integrating vision and radio frequency identification; updating the dynamic mapping relationship between the individual semi-finished interior parts and the pallet grid position based on the position transfer event information; and, upon receiving an outbound signal, extracting the individual true identity identifiers of all individual semi-finished interior parts in the current outbound pallet; generating a single-piece flow sequence containing the latest physical sorting according to a preset grid physical traversal order; and using the individual true identity identifiers of the individual semi-finished interior parts in the single-piece flow sequence... Using the identity identifier as an index, the injection molding process parameter set that uniquely corresponds to the current semi-finished interior part to be processed is dynamically retrieved from the stored historical data. After the current semi-finished interior part to be processed completes the covering process, the covering process parameter set generated in real time at the covering station is associated and fused with the retrieved injection molding process parameter set to generate a single-piece-level cross-process dataset. Based on preset multiple verification rules that respectively represent the logical association between different physical dimensions between the injection molding and covering processes, the combination of injection molding process parameters and covering process parameters in the single-piece-level cross-process dataset is logically consistent. Suspected mis-binding data with logical conflicts is identified, and the suspected mis-binding data is repositioned and corrected in identity, outputting a clean full lifecycle data archive.

[0007] Secondly, this application provides a data analysis system for the production process of interior parts for new energy vehicles. The system includes: an extraction unit, used to respond to a repositioning of individual semi-finished interior parts within a buffer area; acquire position transfer event information corresponding to the picking and placing actions of individual semi-finished interior parts between pallets through a perception network integrating vision and radio frequency identification; update the dynamic mapping relationship between the individual semi-finished interior parts and the pallet grid position based on the position transfer event information; and, upon receiving an outbound signal, extract the individual true identity identifiers of all individual semi-finished interior parts in the current outbound pallet, and generate a single-piece flow sequence containing the latest physical sorting according to a preset grid physical traversal order; and a fusion unit, used to use the individual true identities of the individual semi-finished interior parts in the single-piece flow sequence... The system uses an index to dynamically retrieve the injection molding process parameter set that uniquely corresponds to the current semi-finished interior part to be processed from stored historical data. After the current semi-finished interior part to be processed completes the wrapping process, the system associates and merges the wrapping process parameter set generated in real time at the wrapping station with the retrieved injection molding process parameter set to generate a single-part cross-process dataset. The identification unit is used to perform logical consistency verification on the combination of injection molding process parameters and wrapping process parameters in the single-part cross-process dataset based on preset multiple verification rules that respectively represent the logical association logic between injection molding and wrapping processes. It identifies suspected mis-binding data with logical conflicts, repositions and corrects the identity of the suspected mis-binding data, and outputs a clean full lifecycle data archive.

[0008] This application has at least the following beneficial effects: The method of this application can effectively overcome the limitations of existing technologies that rely on fixed batch or pallet serial numbers for data concatenation, and solves the problem that when the sequence of semi-finished products changes due to cycle time adjustments, subsequent process records are still bound according to the original acceptance sequence, leading to mismatches between the injection molding equipment status and the coating and inspection results. Through the technical solution of this application, single-piece-level precise traceability and anomaly diagnosis of production process data for new energy vehicle interior parts can be achieved, significantly improving the accuracy and reliability of data analysis, and providing strong technical support for the quality management and production optimization of new energy vehicle interior parts. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating a data analysis method for the production process of interior parts for new energy vehicles provided in this application. Figure 2 This is a flowchart illustrating another data analysis method for the production process of new energy vehicle interior parts provided in this application; Figure 3 This is a schematic diagram of the structure of a data analysis system for the production process of interior parts for new energy vehicles provided in this application. Detailed Implementation

[0010] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0011] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0012] In the production process of interior parts for new energy vehicles, to diagnose production anomalies and trace quality, it is typically necessary to collect equipment operation data, process parameters, and quality inspection data, and establish stable correspondences between these multi-source heterogeneous data. However, in actual production processes, after the injection molding process, interior parts often enter a buffer area for cooling and waiting, and are then transferred according to the cycle time of subsequent wrapping and other stations. During this process, when products originally placed in the injection molding order are split, merged, or repositioned due to waiting for subsequent processes, pallet capacity limitations, or local rework, the physical order of the products will change. However, existing process records often still accept them according to the original batch order or pallet order, resulting in a discrepancy between the physical flow order of the same product and the recorded flow order. This deviation causes the injection molding equipment status record that should belong to one interior part to be incorrectly continued to be bound to the subsequent wrapping process parameters and inspection results of another interior part, thus creating a misalignment between equipment status, process execution, and quality performance. Because existing conventional methods typically rely on fixed batch or pallet serial numbers for cross-process data linking, it is difficult to dynamically perceive and update the true attribution of records when individual components undergo cycle time adjustments and position rearrangements. This easily leads to incorrect tracing of problems occurring during wrapping or inspection to the wrong injection molding state. Therefore, in scenarios where semi-finished products enter the buffer after injection molding and undergo pallet splitting and rearrangement, when the order of semi-finished products changes due to cycle time adjustments, subsequent process records are still bound according to the original acceptance order, resulting in mismatches between the injection molding equipment status and the wrapping and inspection results. This severely affects the accuracy of cross-process anomaly diagnosis and source quality traceability for interior parts of new energy vehicles.

[0013] In view of the above problems, this application provides a data analysis method for the production process of interior parts for new energy vehicles. This application dynamically senses the positional changes of individual semi-finished interior parts in the buffer area and updates their mapping relationship with the pallet grid position in real time, thereby ensuring accurate acquisition of their latest physical sorting upon outbound shipment. Subsequently, using the individual part's real identity identifier as an index, the injection molding process parameter set and the overmolding process parameter set are associated and fused to form a single-part-level cross-process dataset. Finally, based on multiple verification rules, logical consistency checks are performed on the data to identify and correct suspected mis-binding data, ultimately outputting a clean, full-lifecycle data archive. This effectively solves the data mis-binding problem caused by physical position rearrangement in traditional methods, significantly improving the accuracy of data traceability and anomaly diagnosis.

[0014] The following specific embodiments will provide a detailed introduction and explanation of the data analysis method for the production process of new energy vehicle interior parts provided in this application.

[0015] Reference Figure 1 This application provides a method for analyzing production process data of interior parts for new energy vehicles, which may include the following steps: S101. In response to the repositioning of individual semi-finished interior parts in the buffer area, the system obtains the position transfer event information corresponding to the picking and placing actions of individual semi-finished interior parts between the pallets through a perception network that integrates vision and radio frequency identification. The system updates the dynamic mapping relationship between the individual semi-finished interior parts and the pallet grid position based on the position transfer event information. When the outbound signal is received, the system extracts the individual real identity identifier of all individual semi-finished interior parts in the current outbound pallet and generates a single-piece flow sequence containing the latest physical sorting according to the preset grid physical traversal order.

[0016] To better understand the technical solutions proposed in this application, it is necessary to explain some key terms and implementation environments involved.

[0017] "Interior component semi-finished product" refers to a single product entity that has undergone injection molding but has not yet completed all subsequent processing (such as overmolding) during the production of interior components for new energy vehicles. Each interior component semi-finished product has a unique "single-piece true identity identifier," which is used to uniquely identify and trace it throughout the entire production process.

[0018] "Sensing network" refers to a data acquisition system that integrates visual recognition technology (such as industrial cameras) and radio frequency identification (RFID) technology to monitor and record in real time the physical location and flow status of individual semi-finished interior parts on the production line.

[0019] A "tray" is a carrier used to carry and transfer individual semi-finished interior trim pieces. Its interior is usually divided into multiple "grid positions", and each grid position can hold one or more individual semi-finished interior trim pieces.

[0020] "Injection molding process parameter set" refers to the process data collected during the injection molding process related to the molding of individual semi-finished interior parts, such as temperature, pressure, and time.

[0021] "Coating process parameter set" refers to the process data collected during the coating process that are related to the surface treatment of individual semi-finished interior parts, such as plasma treatment parameters, adhesive spraying parameters, heating parameters, etc.

[0022] "Single-piece cross-process dataset" refers to a data set formed by linking and integrating the process parameter data generated by a single semi-finished interior part in different processes (such as injection molding and overmolding).

[0023] "Logical consistency verification" refers to using preset verification rules to determine whether there are combinations of process parameters in different processes that do not conform to the physical relationship logic, thereby identifying the possibility of data misbinding.

[0024] This method is typically implemented in an integrated production management system, which includes a data acquisition module, a data processing module, a data storage module, and a data analysis module, and interacts with injection molding machines, coating stations, buffer areas, and quality inspection equipment on the production line.

[0025] The method proposed in this application first responds to the repositioning of individual semi-finished interior parts within the buffer area. It acquires position transfer event information corresponding to the handling actions of these semi-finished interior parts between trays through a perception network that integrates vision and RFID. For example, the perception network can consist of an industrial camera deployed above the buffer area and an RFID reader mounted on the tray. When the semi-finished interior parts are handled between trays, the industrial camera captures a top-down view of the trays, and an image processing algorithm identifies the positional changes of the semi-finished interior parts. The RFID reader then reads the unique identifier of the tray. This information is summarized as position transfer event information, which can be recorded as, for example, "Interior part semi-finished product X moves from grid position 1 on tray A to grid position 2 on tray B."

[0026] Based on location transfer event information, the system updates the dynamic mapping relationship between individual interior trim parts and their positions on the pallet grid. For example, the system maintains a database storing the unique identifier of each individual interior trim part and its current pallet number and grid position. When a location transfer event occurs, the system modifies the corresponding mapping record in the database based on the event information.

[0027] Upon receiving an outbound signal, the system extracts the individual identification identifiers of all semi-finished interior trim pieces within the current outbound pallet and generates a sequence containing the latest physical order, following a preset grid physical traversal order. For example, the outbound signal can be triggered manually or automatically by automated equipment detecting the pallet entering the outbound area. The system queries the latest dynamic mapping relationship of the current outbound pallet, obtains the individual identification identifiers of all semi-finished interior trim pieces, and arranges these identifiers into a sequence according to a grid physical traversal order from left to right and top to bottom.

[0028] S102. Using the unique identifier of the interior semi-finished product in the single-piece flow sequence as an index, dynamically retrieve the injection molding process parameter set that uniquely corresponds to the current interior semi-finished product to be processed from the stored historical data. After the current interior semi-finished product to be processed completes the wrapping process, associate and merge the wrapping process parameter set generated in real time at the wrapping station with the retrieved injection molding process parameter set to generate a single-piece cross-process dataset.

[0029] Subsequently, using the unique identifier of each semi-finished interior part in the single-piece flow sequence as an index, the system dynamically retrieves the injection molding process parameter set that uniquely corresponds to the currently awaiting semi-finished interior part from the stored historical data. For example, when the wrapping station is ready to process a certain semi-finished interior part, it retrieves the unique identifier of that semi-finished interior part from the single-piece flow sequence. The system then uses this identifier as a query key to search and extract various process parameters generated during the injection molding process of that semi-finished interior part from the central database, such as melt temperature, injection pressure, and holding time.

[0030] After the current semi-finished interior trim part completes the wrapping process, the system will associate and merge the wrapping process parameter set generated in real time at the wrapping station with the retrieved injection molding process parameter set to generate a single-part-level cross-process dataset. For example, after completing the wrapping operation on the semi-finished interior trim part, the wrapping station will generate a series of wrapping process parameters in real time, such as plasma surface treatment power, adhesive spraying flow rate, and heating oven temperature. The system will then combine these wrapping process parameters with the previously retrieved injection molding process parameters of the same semi-finished interior trim part to form a single-part-level cross-process dataset containing process data for the entire process from injection molding to wrapping of the semi-finished interior trim part.

[0031] S103. Based on preset multiple verification rules that respectively characterize the logical association between different physical dimensions between injection molding and coating processes, perform logical consistency verification on the combination of injection molding process parameters and coating process parameters in the single-piece cross-process data set, identify suspected mis-binding data with logical conflicts, and perform position reorganization and identity correction on the suspected mis-binding data to output a clean full life cycle data archive.

[0032] For example, the verification rules may include: the influence of injection cooling time and mold temperature on the deformation of the coated product, the influence of injection pressure and holding pressure on the density of the product, and the influence of plasma surface treatment power and time on the bonding strength. The system will determine whether the actual combination of injection molding process parameters and coating process parameters conforms to these preset logical relationships for each interior part semi-finished product's single-piece cross-process dataset.

[0033] If a logical conflict exists, meaning the parameter combination exceeds the preset reasonable range, then suspected mis-binding data with logical conflicts will be identified. For example, if the injection cooling time of a single semi-finished interior part is too short, but the deformation test results after its covering show normal results, this may indicate a mis-binding between the injection molding data and the covering data.

[0034] The system repositions and corrects the identities of suspected mis-bound data, outputting clean, full-lifecycle data archives. For example, for data identified as potentially mis-bound, the system attempts to rematch injection molding and wrapping data within a certain range (such as between several adjacent semi-finished interior parts) to find the combination that best conforms to the logical consistency verification rules. Once the optimal combination is found, the system corrects the data relationships of these semi-finished interior parts and updates their full-lifecycle data archives, ensuring that the data archives of each semi-finished interior part are accurate and clean.

[0035] The data analysis method for the production process of new energy vehicle interior parts proposed in this application aims to solve the problem of data misbinding caused by the physical repositioning of individual semi-finished interior parts in the buffer area during traditional production processes. Traditional methods usually rely on fixed batch or pallet serial numbers for cross-process data linkage. When the cycle time and position of an individual entity are adjusted, it is difficult to dynamically perceive and update the true attribution of the records, resulting in mismatches between the injection molding equipment status and the wrapping and inspection results. This seriously affects the accuracy of cross-process anomaly diagnosis and source quality traceability.

[0036] Specifically, the steps described above for generating a single-item flow sequence containing the latest physical order can be further refined into the following process.

[0037] According to the above method, generating a single-item flow sequence containing the latest physical order includes: S201. When the injection molding machine completes the mold opening cycle, based on the current injection molding equipment number, mold cavity number, production work order number and system timestamp, generate a unique identification identifier for each semi-finished interior part that corresponds to the finished product. Establish an initial mapping relationship between the tray number carrying the semi-finished interior part, the grid position inside the tray and the unique identification identifier.

[0038] S202. Through the perception network deployed in the buffer area, capture the picking and placing actions of single semi-finished interior parts when they are split, folded, or swapped between pallets. Transform the picking and placing actions into a position transfer matrix containing the source pallet number, source grid coordinates, target pallet number, and target grid coordinates, and update the dynamic mapping relationship in real time based on the position transfer matrix.

[0039] S203. Upon receiving the outbound signal, lock the latest dynamic mapping relationship of the current outbound pallet, extract all the real identity identifiers of the bound single items according to the preset grid physical traversal order, and generate the single item flow sequence.

[0040] Specifically, when an interior trim component rolls off the injection molding machine, the system assigns it a unique, individual identification identifier. This identifier is generated based on specific information from the injection molding equipment, such as the current equipment number, mold cavity number, production order number, and a precise system timestamp, ensuring the traceability of each semi-finished product throughout the production process. Simultaneously, this individual identification identifier establishes an initial mapping relationship with the pallet number carrying the semi-finished product and its specific grid position within the pallet, laying the foundation for subsequent tracking and circulation.

[0041] The sensing network deployed in the buffer area primarily functions to continuously monitor the movement of individual semi-finished interior trim parts between different pallets. When semi-finished parts undergo actions such as splitting, merging, or repositioning, the sensing network accurately captures these events. These actions are then transformed into a position transfer matrix, which details the source pallet and grid coordinates from which the semi-finished part was retrieved, and the target pallet and grid coordinates to which it was placed. Based on this position transfer matrix, the system can update the dynamic mapping relationship between individual semi-finished interior trim parts and pallet grid positions in real time, ensuring that the data always reflects the actual physical location of the semi-finished parts.

[0042] In practical applications, upon receiving an outbound signal, the system immediately locks the latest dynamic mapping relationship of the pallet awaiting outbound shipment. Subsequently, the system extracts all semi-finished interior parts with bound individual identification tags within the pallet according to a preset grid physical traversal order. In this way, a single-piece flow sequence containing the latest physical order can be generated, which accurately reflects the actual arrangement order of the semi-finished products upon outbound shipment.

[0043] Through the above technical solution, this application enables refined management of the complex flow process of individual semi-finished interior parts for new energy vehicles within the buffer area. Specifically, by generating a unique, authentic identifier for each part at the source and establishing an initial mapping, combined with the real-time capture and dynamic updating of location transfer events by a sensing network, the accuracy of data-physical consistency is greatly improved. Thus, upon outbound shipment, a single-part flow sequence containing the latest physical order can be generated. This not only provides a highly reliable single-part-level data foundation for subsequent cross-process data association but also effectively avoids data chaos problems such as mis-binding and omissions caused by changes in the location of semi-finished products, significantly improving the accuracy and reliability of production process data analysis.

[0044] In some of the above embodiments, methods have been proposed for generating individual true identities and individual flow sequences for semi-finished interior parts. However, in order to achieve subsequent cross-process data association and logical consistency verification at the individual part level, it is necessary to obtain and store the injection molding process parameter set that uniquely corresponds to each semi-finished interior part.

[0045] In this regard, this application further proposes that after generating the single-item authentic identity identifier, the method also includes: Simultaneously collect multi-source heterogeneous data during the injection molding process. The multi-source heterogeneous data includes real-time temperature of each heating section of the melt barrel, screw speed, peak injection pressure and holding pressure curve, inlet and outlet cooling water temperature and flow rate, and molding cycle time. The multi-source heterogeneous data is aligned with time series and features are extracted to form the injection molding process parameter set. The feature extraction includes converting the continuous injection pressure time series into injection peak pressure, pressure integral area and pressure rise slope. The injection molding process parameter set is bound and stored with the individual part's real identity identifier, as part of the stored historical data.

[0046] Specifically, the multi-source heterogeneous data refers to various types and formats of production process data acquired in real time from injection molding equipment and auxiliary equipment during the injection molding process. This data covers key process parameters affecting product quality. For example, the real-time temperature of each heating section of the melt cylinder reflects the plastic's melting state; screw speed affects plasticizing efficiency; the peak injection pressure and holding pressure curves directly relate to the product's filling and holding pressure effects; the inlet and outlet temperatures and flow rates of cooling water affect the product's cooling rate; and the molding cycle time reflects production efficiency. Time series alignment and feature extraction are crucial steps in data preprocessing for this multi-source heterogeneous data. Time series alignment aims to eliminate differences in data acquisition frequencies and timestamps from different sensors, ensuring consistency across all data points in the time dimension. Feature extraction involves extracting representative and physically meaningful indicators from the raw, continuous process data. For example, for a continuous injection pressure time series, converting it into discrete features such as peak injection pressure, pressure integral area, and pressure ramp-up slope can more effectively characterize the pressure change characteristics during injection molding. These features are of significant value for subsequent quality analysis and process optimization. The injection molding process parameter set, after the aforementioned processing, is a collection of key parameters that comprehensively and accurately describe the injection molding process of a single interior trim semi-finished product. This parameter set is then bound to the unique, individual identifier of each interior trim semi-finished product and stored in a central database or distributed storage system, thus forming an important component of the stored historical data. This binding storage method ensures that each interior trim semi-finished product has a traceable process data file uniquely associated with its identity from the injection molding stage.

[0047] The aforementioned technical solution enables refined, single-piece-level data collection and management of injection molding process data in the production of interior parts for new energy vehicles. This not only solves the problem of poor correlation between process data and specific products in traditional production models, but also improves data usability and analytical value through time-series alignment and feature extraction of multi-source heterogeneous data. Consequently, it provides reliable, high-quality source data support for subsequent cross-process data association, logical consistency verification, and the final construction of a full lifecycle data archive, greatly enhancing the traceability of the production process and the accuracy of data analysis, and providing a solid data foundation for product quality control and process optimization.

[0048] Specifically, the above-mentioned perception network deployed in the buffer area captures the picking and placing actions of individual semi-finished interior parts when they are split, folded, or swapped between trays, and converts the picking and placing actions into a position transfer matrix containing the source tray number, source grid coordinates, target tray number, and target grid coordinates, which can be implemented in the following way.

[0049] The process includes: reading and confirming the unique physical number of the transfer pallet using radio frequency identification (RFID) technology; acquiring a top-view image stream of the pallet using a high-resolution industrial camera, using image processing algorithms to locate the pallet edge reference points to establish a two-dimensional coordinate system inside the pallet, and using a preset target detection model to identify the occupancy status of individual semi-finished interior parts at each grid position within the pallet; when the monitored grid position status changes from occupied to empty, recording the source grid coordinates and timestamp of the picking action; when the monitored grid position status changes from empty to occupied, recording the target grid coordinates and timestamp of the placing action; and summarizing the continuously captured picking and placing action event streams to generate the position transfer matrix.

[0050] Specifically, when individual semi-finished interior components circulate within the buffer area, the transfer pallets carrying these semi-finished products are first identified using RFID technology to obtain and confirm the unique physical number of each pallet. RFID technology ensures the accuracy of pallet identification, providing a reliable foundation for subsequent location tracking.

[0051] Meanwhile, a high-resolution industrial camera is deployed above the buffer area to continuously acquire a top-down image stream of the pallet. This image stream is then fed into an image processing algorithm that accurately locates the pallet's edge reference points and establishes a two-dimensional coordinate system inside the pallet. Within this coordinate system, the grid positions inside the pallet can be accurately described. Furthermore, a pre-defined target detection model is used to analyze the images to identify the occupancy status of individual semi-finished interior components at each grid position within the pallet, i.e., to determine whether a semi-finished product currently exists at a given grid position.

[0052] When the system detects that the state of a monitored grid location changes from occupied to empty, this is determined to be a retrieval action, and the source grid coordinates and the timestamp of the retrieval action are recorded. Conversely, when the system detects that the state of a grid location changes from empty to occupied, this is determined to be a placement action, and the target grid coordinates and the timestamp of the placement action are recorded.

[0053] Finally, these continuously captured pick-up and place-up event streams are summarized. Through logical correlation and time series analysis of these events, a position transfer matrix containing the source pallet number, source grid coordinates, target pallet number, and target grid coordinates can be generated. This matrix accurately records all movement trajectories of individual interior trim semi-finished products between pallets or within the grid positions of the pallets.

[0054] By integrating RFID and high-resolution industrial vision technologies, the above technical solution enables refined and real-time capture of position transfer events for individual semi-finished interior parts within the buffer area. Compared to relying solely on a single sensing method or manual recording, this solution significantly improves the accuracy and automation level of position transfer event information acquisition, effectively avoiding data omissions or errors caused by human error or blind spots. Consequently, the generated position transfer matrix, containing the source pallet number, source grid coordinates, target pallet number, and target grid coordinates, is more accurate and complete. This provides a solid data foundation for the real-time updating of the dynamic mapping relationship between individual semi-finished interior parts and pallet grid positions, thereby ensuring the integrity and reliability of the entire production process data chain.

[0055] Specifically, in the above method, the steps regarding using the unique identifier of each semi-finished interior part in the single-piece flow sequence as an index to dynamically retrieve the injection molding process parameter set that uniquely corresponds to the currently processed semi-finished interior part from the stored historical data, and after the current semi-finished interior part completes the wrapping process, associating and fusing the wrapping process parameter set generated in real time at the wrapping station with the retrieved injection molding process parameter set to generate a single-piece level cross-process dataset can be further refined into the following process.

[0056] The process involves using the unique identifier of each semi-finished interior trim piece in the single-piece flow sequence as an index to dynamically retrieve the injection molding process parameter set that uniquely corresponds to the currently processed semi-finished interior trim piece from stored historical data. After the currently processed semi-finished interior trim piece completes the covering process, the covering process parameter set generated in real time at the covering station is associated and fused with the retrieved injection molding process parameter set to generate a single-piece level cross-process dataset, including: Upon receiving the trigger signal for the part retrieval action at the wrapping station, the true identity identifier of the interior part corresponding to the currently retrieved interior part semi-finished product is extracted sequentially according to the arrangement order of the interior part semi-finished products in the single-part flow sequence. Using the extracted real identity identifier of the single piece as the query key, a data retrieval request is initiated to the central database to obtain the injection molding process parameter set that uniquely corresponds to the single semi-finished interior part to be processed. After the current interior trim semi-finished product is wrapped, the wrapping process parameter set generated in real time by the wrapping station controller is collected. The wrapping process parameter set includes plasma surface treatment power and time, glue spraying robot trajectory coordinates and coating flow rate, heating oven temperature curve and vacuum forming machine negative pressure value and holding time. The coating process parameter set is spliced ​​and merged with the retrieved injection molding process parameter set to generate the single-piece level cross-process dataset.

[0057] Specifically, when the wrapping station is ready to process a single semi-finished interior trim component, the system receives a pick-up action trigger signal. This signal indicates that a single semi-finished interior trim component is about to enter the wrapping process. Upon receiving this signal, the system accurately identifies and extracts the unique identifier of the currently picked-up semi-finished interior trim component according to the pre-generated component flow sequence and the physical arrangement of the semi-finished components. This unique identifier is a unique code for the semi-finished component throughout the entire production process, ensuring the accuracy of its identity.

[0058] Using the extracted individual component's unique identifier as the query key, the system sends a data retrieval request to the central database. The central database stores historical data, including the injection molding process parameter sets generated during the injection molding process for each interior trim component's semi-finished product. Through this query key, the system can accurately and dynamically retrieve the injection molding process parameter set uniquely corresponding to the currently awaiting interior trim component's semi-finished product from massive amounts of historical data. This injection molding process parameter set contains all key production data for the semi-finished product during the injection molding stage, such as the real-time temperature of each heating section of the melt cylinder, screw speed, peak injection pressure, and holding pressure curves.

[0059] In practical applications, after the wrapping operation of a single semi-finished interior component is completed, the wrapping station controller generates and collects a series of wrapping process parameters in real time. These parameters record key process data in detail, including plasma surface treatment power and time (characterizing the effectiveness of surface activation), the trajectory coordinates and coating flow rate of the adhesive spraying robot (reflecting the uniformity and accuracy of adhesive application), the temperature profile of the heating oven (recording temperature changes during baking and curing), and the negative pressure value and holding time of the vacuum forming machine (reflecting key parameters of the wrapping process). These parameters are crucial for evaluating the wrapping quality.

[0060] Furthermore, the collected overmolding process parameters will be combined and merged with the injection molding process parameters previously retrieved from the central database. This merging operation aims to logically correlate data generated from different processes (injection molding and overmolding) of the same interior trim semi-finished product, thereby generating a complete, single-part-level cross-process dataset. This dataset will contain detailed process information from injection molding to overmolding, laying the foundation for subsequent data analysis and quality traceability.

[0061] The aforementioned technical solution enables refined, piece-level correlation of data on individual semi-finished products during the injection molding and overmolding processes of new energy vehicle interior parts production. This not only significantly improves the accuracy and completeness of data collection, ensuring that each interior part semi-finished product has a dedicated data file covering its entire injection molding and overmolding process, but also provides a solid data foundation for subsequent quality traceability, process optimization, and problem diagnosis based on full lifecycle data. Especially when quality issues arise, the process parameters of specific batches, workstations, and even individual parts can be quickly located, thereby significantly improving the transparency and controllability of the production process and effectively reducing production risks and rework costs.

[0062] In some of the embodiments described above in this application, a scheme is proposed to link and merge the injection molding process parameter set and the overmolding process parameter set to generate a single-piece-level cross-process dataset. However, in actual production, relying solely on process parameter data may not fully reflect the final quality status of the product, thus limiting the completeness of product lifecycle traceability and the ability to accurately locate potential quality problems. If the final quality inspection results are not included in the data archive, it is difficult to effectively assess the impact of process parameters on product quality, and it may also lead to insufficient accuracy in logical consistency checks.

[0063] In this regard, this application further proposes, in addition to the above method, the following: The quality inspection station collects quality inspection data for individual semi-finished interior parts that have been covered. Before acquiring the data, the true identity of the individual semi-finished interior parts to be inspected is confirmed by scanning a code or comparing visual textures. Using the confirmed real identity as an index, the quality inspection data is appended and associated with a single-piece cross-process dataset that already contains injection molding process parameters and coating process parameters, forming a historical archive covering data from the entire stages of injection molding, coating, and inspection. The quality inspection data includes at least one of the following: appearance defect type code, defect severity score, key dimension deviation value, surface defect coordinates and area, and serves as one of the input data for the subsequent logical consistency verification.

[0064] Specifically, after the wrapping process is completed, each semi-finished interior trim component is sent to a quality inspection station for various quality checks. Before quality inspection, the true identity of the semi-finished interior trim component to be inspected needs to be confirmed through barcode scanning or visual texture comparison. Barcode scanning can be achieved by reading the QR code or barcode attached to the semi-finished interior trim component or its carrier, ensuring consistency with the true identity of the individual component during the production process. Visual texture comparison uses a high-resolution camera to capture the unique texture features of the surface of the semi-finished interior trim component and matches them with known texture features stored in a database to confirm its identity.

[0065] Once the true identity of a single semi-finished interior component is confirmed, the quality inspection data collected at the quality inspection station will be indexed by this confirmed identity and added to the existing single-piece-level cross-process dataset for that semi-finished interior component. This dataset already contains injection molding and overmolding process parameters. By adding quality inspection data, a historical archive covering the entire process from injection molding to overmolding and inspection is formed. This quality inspection data can include various types, such as appearance defect type codes to identify specific defects like scratches, color differences, and bubbles; defect severity scores to quantify the severity of defects; key dimension deviation values ​​to record deviations between key geometric dimensions and design standards; and surface defect coordinates and area to accurately describe the location and size of defects. This quality inspection data not only completes the product's full lifecycle archive, but more importantly, it serves as crucial input data for subsequent logical consistency checks, providing a basis for more accurately identifying potentially mis-attached data.

[0066] Through the aforementioned technical solution, this application significantly improves the comprehensiveness and accuracy of data analysis in the production process of interior parts for new energy vehicles. Firstly, by incorporating quality inspection data into a single-piece, cross-process dataset, it achieves full lifecycle data traceability for the product, from injection molding and overmolding to final inspection, greatly enhancing the integrity of the data archive. Secondly, this quality inspection data serves as input for subsequent logical consistency checks, enabling the check rules to make judgments based on more comprehensive information, thereby improving the accuracy of identifying potentially mislabeled data and reducing the possibility of misjudgments. Furthermore, the complete historical archive provides a solid data foundation for subsequent product quality analysis, process optimization, and defect tracing, helping enterprises to quickly respond to quality issues, continuously improve production processes, and ultimately enhance product quality and production efficiency.

[0067] In some preferred embodiments, it is assumed that after a semi-finished interior trim part completes the injection molding and overmolding processes, its injection molding process parameter set (e.g., peak injection pressure, cooling time) and overmolding process parameter set (e.g., plasma surface treatment power, oven temperature profile) have been associated and stored through its true identity identifier. When the semi-finished interior trim part is sent to the quality inspection station, its true identity identifier is first confirmed by scanning the QR code printed on its surface. Subsequently, the quality inspection equipment performs appearance inspection and dimensional measurement. For example, the inspection results show that the semi-finished interior trim part has a scratch (appearance defect type code "A01"), a severity score of "3", and a critical dimension deviation of "+0.5mm". This quality inspection data, including the "A01" defect type code, severity score of 3, critical dimension deviation of +0.5mm, and the surface coordinates and area of ​​the scratch, will be added to its existing single-piece-level cross-process data set, indexed by the true identity identifier of the semi-finished interior trim part. Thus, each semi-finished interior component possesses a complete historical archive covering the entire process from injection molding and overmolding to inspection. In subsequent logical consistency checks, the system can utilize this quality inspection data to determine, for example, whether insufficient injection molding cooling time led to dimensional deviations, or whether improper handling during overmolding resulted in scratches, thereby more accurately identifying potential errors or process anomalies in the data association.

[0068] Specifically, the above-mentioned multiple verification rules are defined in detail, and the process of performing logical consistency verification on the combination of injection molding process parameters and overmolding process parameters in a single-piece cross-process dataset is also explained in detail.

[0069] The multiple verification rules include a first verification rule, a second verification rule, and a third verification rule; The first verification rule is used to determine the thermal history correlation deviation between the cooling time and mold temperature of the injection molding process and the product deformation detection results after heating in the subsequent coating process. The second verification rule is used to determine the pressure-mass transfer deviation between the peak injection pressure and the holding pressure integral value of the injection process and the final product weight or density test result; The third verification rule is used to determine the surface treatment correlation deviation between the plasma surface treatment power and time and the adhesive spraying flow rate of the coating process and the final product bonding strength test result. The logical consistency verification of the combination of injection molding process parameters and coating process parameters in the single-piece cross-process dataset includes: for each interior part single semi-finished product dataset, determining whether the actual combination of its injection molding process parameters and coating process parameters has the thermal history correlation deviation, the pressure-mass transfer deviation, or the surface treatment correlation deviation. If the deviation exceeds the preset range, the dataset is marked as the suspected mis-binding data.

[0070] The aforementioned multiple verification rules aim to evaluate the data logical consistency between the two core processes of injection molding and overmolding from different physical dimensions and process correlation perspectives. Specifically, the first verification rule focuses on the impact of thermal history on product performance. In the injection molding process, cooling time and mold temperature directly determine the internal stress distribution and material crystallization state of a single semi-finished interior part. These factors significantly affect the deformation behavior of the product when heated in the subsequent overmolding process. Therefore, by analyzing the cooling time and mold temperature in the injection molding process and comparing them with the deformation detection results of the overmolded product, it is possible to determine whether there are abnormal thermal history correlation deviations. For example, insufficient injection cooling or excessively high mold temperature may cause excessive deformation of the product during overmolding heating, which will be identified by this rule.

[0071] The second verification rule focuses on the impact of pressure during the injection molding process on the physical properties of the product. Peak injection pressure and the integral of holding pressure are key parameters for measuring the material's filling density and internal structural uniformity during injection molding. These pressure parameters have a direct physical correlation with the weight and density of the final product. By comparing the peak injection pressure and the integral of holding pressure during the injection molding process with the weight or density test results of the final product, it is possible to determine if there is a pressure-mass transfer deviation. For example, insufficient injection pressure or improper holding time may lead to voids or uneven density within the product, thus affecting its weight and density, and thus being identified by this rule.

[0072] The third verification rule focuses on the surface treatment and adhesive performance of the coating process. The power and time of plasma surface treatment, as well as the adhesive spraying flow rate, are important process parameters affecting the adhesive strength between the coating material and the substrate. Plasma treatment improves the surface activity of the material, while the adhesive spraying flow rate determines the uniformity and thickness of the adhesive layer. By correlating these coating process parameters with the adhesive strength test results of the final product, it is possible to determine whether there are any surface treatment-related deviations. For example, insufficient plasma treatment or uneven adhesive spraying may lead to substandard adhesive strength, which would be identified by this rule.

[0073] During logical consistency verification, for each interior trim component semi-finished product, the actual combination of its injection molding process parameters and overlay process parameters in the single-piece-level cross-process dataset is input into the aforementioned verification rules. The system will determine whether this combination has thermal history correlation deviations, pressure-mass transfer deviations, or surface treatment correlation deviations. If the calculated result of any deviation exceeds the preset allowable range, the dataset will be marked by the system as suspected misbinding data, indicating that there may be a mismatch or correlation error between its injection molding and overlay data.

[0074] Through the aforementioned technical solution, this application significantly improves the accuracy and reliability of single-piece-level cross-process data association in the production of interior parts for new energy vehicles. Specifically, by pre-setting and applying multiple verification rules based on physical dimension association logic, the system no longer relies solely on simple identity matching but can deeply analyze the inherent logical relationship between process parameters in the two key processes of injection molding and overmolding. This allows the system to accurately identify potentially misbound data even in cases of identity confusion or minor deviations in data collection, by judging key physical associations such as thermal history, pressure-mass transfer, and surface treatment. This multi-dimensional, in-depth logical consistency verification mechanism effectively avoids the difficulties in quality traceability and low production efficiency caused by data misbinding, laying a solid foundation for subsequent data correction and the clean output of full lifecycle data archives, thereby ensuring the authenticity and usability of production process data.

[0075] In some of the embodiments described above in this application, although multiple verification rules can effectively identify suspected misbound data with logical conflicts, simply marking this suspected data cannot directly solve the data misbound problem. This may impair the accuracy of subsequent data analysis and tracing, and it also fails to fully achieve the goal of outputting a clean, full-lifecycle data archive. Therefore, this application further proposes a scheme for automatically repositioning and correcting the identity of suspected misbound data.

[0076] After marking the suspected mis-bound data, the method further includes: Extract the target interior component semi-finished product marked as suspected of mis-binding, and a predetermined number of interior component semi-finished products that are adjacent to it in the single-piece circulation sequence, to form a suspected single-piece set; Retrieve the injection molding process parameter set, the corresponding overlay process parameter set, and the quality inspection data for each semi-finished interior component in the suspected component set, and iterate and rearrange all possible matching combinations between the injection molding process parameter set and the overlay and inspection data. Substitute each matching combination back into the first, second, and third verification rules to calculate the conflict penalty score that represents the degree of logical conflict. If, among all matching combinations, there is a unique combination whose conflict penalty score is lower than the preset safety tolerance threshold, then the inter-process data association of each interior component semi-finished product in the suspected component set is re-established according to that combination, and its full life cycle data file is updated.

[0077] Specifically, when the system identifies suspected mis-binding data with logical conflicts, it first extracts the suspected target interior part semi-finished product to perform precise correction. Simultaneously, to account for the possibility of mis-binding in adjacent batches or sequences, it also extracts a preset number of adjacent interior part semi-finished products in the individual part flow sequence, collectively forming a suspected set of individual parts. This preset number can be configured based on actual production experience and common patterns of data mis-binding; for example, it can be set to one or two before and after, forming a localized data correction range.

[0078] Subsequently, for each semi-finished interior component in the suspected component set, the system retrieves its corresponding injection molding process parameter set, overlay process parameter set, and quality inspection data from stored historical data. Based on this, to identify the correct relationships, the system iterates through and rearranges all possible matching combinations between these retrieved injection molding process parameter sets, overlay process parameter sets, and quality inspection data. For example, if the suspected component set contains N semi-finished interior components, there may be N! matching combinations between their injection molding process parameter sets, overlay process parameter sets, and quality inspection data.

[0079] Furthermore, for each matching combination obtained through the traversal, the system re-evaluates it using the first, second, and third verification rules mentioned above. Unlike simply determining whether it exceeds a preset range, a conflict penalty score representing the degree of logical conflict is calculated here. This score aims to quantify the severity of logical inconsistencies between injection molding, coating, and detection data in the current combination; a lower score indicates higher logical consistency in the data association.

[0080] Ultimately, among all possible matching combinations, the system will search for the combination with the lowest conflict penalty score. If there exists a unique combination whose conflict penalty score is not only the lowest but also below a preset safety tolerance threshold, then this combination is considered a correct association. In this case, the system will re-establish the inter-process data associations of each interior component semi-finished product within the suspected component set according to this unique combination and update its full lifecycle data archive, thereby completing the automatic correction of the mis-bound suspected data.

[0081] Through the aforementioned technical solution, this application elevates data analysis from simple error identification to automatic correction, significantly improving the accuracy and completeness of data archives in the production process of new energy vehicle interior parts. This solution reduces the need for manual intervention, lowers the time cost of data correction, and reduces potential human error. Furthermore, by introducing a conflict penalty score and a unique combination method judgment mechanism, the reliability and high confidence of the automatic correction results are ensured, providing a cleaner and more trustworthy full lifecycle data archive for subsequent quality traceability, process optimization, and production decisions, thereby improving the intelligence level and data management efficiency of the entire production process.

[0082] In some preferred embodiments, it is assumed that on the production line, due to momentary errors in pallet transfer or sensor readings, misalignment occurs between the injection molding process parameter sets (IA, IB, IC) of three consecutively produced interior trim semi-finished products A, B, and C and the subsequent overlay process parameter sets (PA, PB, PC) and quality inspection data (QA, QB, QC). For example, the initial associations recorded by the system might be (IA-PA-QA), (IB-PB-QB), and (IC-PC-QC). However, during logical consistency checks, a significant logical conflict is found between the injection molding process parameters, overlay process parameters, and quality inspection data of dataset B, and it is marked as suspected misaligned data.

[0083] According to this implementation, the system extracts adjacent semi-finished interior trim parts (e.g., A and C) in the single-part flow sequence, centered on part B, thus forming a suspected part set {A, B, C}. Next, the system retrieves all injection molding process parameter sets {IA, IB, IC}, overmolding process parameter sets {PA, PB, PC}, and quality inspection data {QA, QB, QC} for these three semi-finished interior trim parts. Then, the system iterates through all possible matching combinations. For example, in addition to the original (IA-PA-QA, IB-PB-QB, IC-PC-QC), it also tries all 3! = 6 permutations such as (IA-PB-QB, IB-PA-QA, IC-PC-QC) and (IA-PC-QC, IB-PB-QB, IC-PA-QA). For each combination, the system re-substitutes the first, second, and third verification rules to calculate its conflict penalty score. Suppose that calculations show the combination (IA-PA-QA, IB-PC-QC, IC-PB-QB) has the lowest conflict penalty score, significantly lower than the preset safety tolerance threshold, and is the only combination among all that meets this condition. In this case, the system will determine that the original data for B and C has been misbound, and will re-establish the data associations for these three interior trim parts according to (IA-PA-QA, IB-PC-QC, IC-PB-QB), updating their full lifecycle data archives, thereby achieving automatic correction of misbound data.

[0084] In some embodiments described above, this application proposes a scheme to reposition and correct the identity of mis-bound suspected data by calculating conflict penalty scores. However, in actual production environments, due to the complex positional changes of individual semi-finished interior parts within the buffer area, or the low recognition confidence of the perception network at certain times, multiple matching combinations may have very similar conflict penalty scores during data reassembly, making it difficult to determine a unique and reliable correction scheme. If this problem is not addressed, the system may perform erroneous automatic corrections, or attempt corrections even when the data corruption exceeds the correction capability, thereby introducing new data errors and affecting the accuracy and reliability of the entire lifecycle data archive. To address this, this application further proposes a method to improve the robustness and reliability of data correction by introducing multi-dimensional offset costs and evaluating the convergence of correction schemes after calculating conflict penalty scores.

[0085] After calculating the conflict penalty score, the method further includes: Obtain the visual perception confidence decay curve corresponding to the fuzzy recognition period when the suspected single item set changes position in the buffer area, and the two-dimensional spatial adjacency matrix of the interior part semi-finished product in the tray; The time dimension offset cost is determined based on the visual perception confidence decay curve, and the spatial dimension offset cost is determined based on the degree of deviation of the adjacency relationship between individual semi-finished interior parts in the two-dimensional spatial adjacency matrix. The conflict penalty score, the time dimension offset cost, and the spatial dimension offset cost are weighted and summed to obtain the comprehensive offset cost. Calculate the difference between the optimal and suboptimal values ​​of the overall offset cost among all matching combinations. If the difference is less than the preset convergence isolation threshold, it is determined that the current data disorder level exceeds the automatic correction capability, automatic correction is abandoned, and a downgrade isolation process is performed on the dataset corresponding to the suspected single item set. The downgrade isolation process includes marking the dataset as untrusted and storing it in hard isolation, while outputting warning information containing abnormal batches, time periods, and cache locations.

[0086] Specifically, the visual perception confidence decay curve refers to the curve showing how the confidence of the visual perception system in target recognition and tracking changes over time during the fuzzy recognition period when the position of a single semi-finished interior part changes. This curve reflects the degree to which the accuracy and reliability of visual recognition decrease within a specific time period due to factors such as occlusion, changes in lighting, or rapid movement. Its purpose is to quantify the uncertainty of perception data and provide a basis for subsequent corrective decisions.

[0087] The two-dimensional spatial adjacency matrix of the individual semi-finished interior trim parts within the tray can be understood as a matrix describing the physical proximity relationships between the individual semi-finished interior trim parts at each grid position within the tray. This matrix characterizes the relative position and layout of each individual semi-finished interior trim part within the tray by recording the connection relationships between each individual semi-finished interior trim part and its adjacent individual semi-finished interior trim parts. Its purpose is to capture the physical spatial correlation of the individual semi-finished interior trim parts, assisting in determining the rationality of position rearrangement.

[0088] In practical applications, the time dimension offset cost is determined based on the visual perception confidence decay curve and is used to measure the potential deviation cost of associating individual semi-finished interior parts data over time. For example, when the visual perception confidence is low, the time dimension offset cost will increase accordingly, indicating a higher risk of data misbinding during that period. The spatial dimension offset cost is determined based on the degree of deviation in the adjacency relationships between individual semi-finished interior parts in the two-dimensional spatial adjacency matrix and is used to measure the potential deviation cost of associating individual semi-finished interior parts data in the spatial dimension. For example, if the corrected matching combination causes a significant change in the physical adjacency relationships of individual semi-finished interior parts within the tray, the spatial dimension offset cost will increase.

[0089] The comprehensive offset cost is obtained by weighting and summing the conflict penalty score, the time dimension offset cost, and the spatial dimension offset cost. The weights can be adjusted according to the actual application scenario and data importance to balance the impact of logical conflicts, temporal uncertainty, and spatial variations on the reliability of data correction. The aim is to provide a more comprehensive and robust metric for evaluating the rationality of each matching combination.

[0090] The convergence isolation threshold is a preset value used to determine whether the difference between the optimal and suboptimal values ​​of the overall offset cost among all matching combinations is large enough. If the difference is less than the convergence isolation threshold, it indicates that there are multiple "optimal" or "suboptimal" correction schemes, and the differences between them are not significant. In this case, the reliability of automatic correction is low. Its purpose is to avoid blind correction when the correction scheme is unclear or the data is highly chaotic.

[0091] When the system determines that the current data corruption exceeds the capabilities of automatic correction, it will abandon automatic correction and perform degradation isolation processing on the dataset corresponding to the suspected individual data set. This degradation isolation processing includes marking the dataset as untrusted and storing it in hard isolation, i.e., separating it from the main data stream to prevent it from contaminating other data. Simultaneously, the system will output warning information including the abnormal batch, time period, and cache location, allowing for manual intervention for verification and handling. The purpose is to ensure the overall purity and reliability of the data archive and prevent the spread and use of erroneous data.

[0092] By calculating the combined offset cost of all possible matching combinations and further analyzing the difference between the optimal and suboptimal values, this application can determine the uniqueness and reliability of the data correction scheme. If the combined offset costs of the optimal and suboptimal schemes are very close, it indicates a high degree of data disorder and the existence of multiple seemingly reasonable correction schemes, in which case the risk of automatic correction is high. In this case, the system will not force correction but will choose to abandon automatic correction and perform degradation isolation processing on the relevant dataset. This mechanism effectively avoids erroneous automatic correction when the data is highly uncertain or disordered, thereby ensuring the accuracy and reliability of the data archive throughout its entire lifecycle.

[0093] In some preferred embodiments, a specific example is given below. Suppose that within the buffer area, three semi-finished interior parts A, B, and C on a tray undergo position rearrangement during a brief period of fuzzy recognition due to a temporary malfunction in the visual perception system. The original flow sequence record is ABC, but due to the position rearrangement, the system detects a new physical order of CAB upon outbound processing. At this point, according to the aforementioned verification rules, the system identifies a logical conflict in the datasets of A, B, and C and marks them as suspected mis-bound data.

[0094] Based on the above method, the system will extract the set of suspected individual parts {A, B, C} and retrieve their respective injection molding process parameter sets, overmolding process parameter sets, and quality inspection data. The system will iterate through all possible matching combinations, for example: 1. Injection A - Cover A - Detect A, Injection B - Cover B - Detect B, Injection C - Cover C - Detect C (Original Match).

[0095] 2. Injection A - Cover B - Detect B, Injection B - Cover C - Detect C, Injection C - Cover A - Detect A (a possible mis-binding correction).

[0096] 3. Injection A - Cover C - Detect C, Injection B - Cover A - Detect A, Injection C - Cover B - Detect B (Another possible mis-binding correction).

[0097] For each matching combination, the system calculates its conflict penalty score. Suppose that in a certain scenario, the conflict penalty scores of combination 2 and combination 3 are very close, for example, 10.2 and 10.5 respectively, making it difficult to determine the unique optimal correction scheme based solely on the conflict penalty score.

[0098] At this point, the solution of this application further obtains the visual perception confidence decay curve corresponding to the fuzzy recognition period during which the position change occurs, and determines the time dimension offset cost based on the curve. For example, if the physical flow path corresponding to combination method 2 occurs during a period of low visual confidence, its time dimension offset cost may be high. Simultaneously, the system obtains the two-dimensional spatial adjacency matrix of individual interior trim semi-finished products within the tray, and determines the spatial dimension offset cost based on the degree of deviation in adjacency relationships caused by the matching combination method. For example, if combination method 2 causes a drastic change in the physical adjacency relationship between individual interior trim semi-finished products A and B, while combination method 3 causes a smaller change, then the spatial dimension offset cost of combination method 2 will be higher.

[0099] The comprehensive offset cost is obtained by weighted summing of the conflict penalty score, the time dimension offset cost, and the spatial dimension offset cost. Assume that after weighted summation, the comprehensive offset cost of combination method 2 is 15.0, while the comprehensive offset cost of combination method 3 is 12.0. At this point, the difference between the optimal and second-best values ​​is 3.0. If the preset convergence isolation threshold is 2.0, since 3.0 is greater than 2.0, the system will determine that there is a clear optimal correction scheme (combination method 3), and will re-establish the inter-process data association according to this combination method, updating its full lifecycle data archive.

[0100] In some embodiments, this application proposes a data analysis system for the production process of interior parts for new energy vehicles, including: an extraction unit 301, used to respond to the repositioning of individual semi-finished interior parts in the buffer area, to obtain position transfer event information corresponding to the picking and placing actions of individual semi-finished interior parts between pallets through a perception network that integrates vision and radio frequency identification, to update the dynamic mapping relationship between individual semi-finished interior parts and pallet grid positions according to the position transfer event information, and when receiving the outbound signal, to extract the individual real identity identifier of all individual semi-finished interior parts in the current outbound pallet, and to generate a single-piece flow sequence containing the latest physical sorting according to the preset grid physical traversal order; The fusion unit 302 is used to dynamically retrieve the injection molding process parameter set that uniquely corresponds to the current semi-finished interior part in the single-piece flow sequence from the stored historical data, using the single-piece real identity identifier of the semi-finished interior part in the single-piece flow sequence as an index. After the current semi-finished interior part in the single-piece flow sequence completes the wrapping process, the fusion unit 302 associates and merges the wrapping process parameter set generated in real time at the wrapping station with the retrieved injection molding process parameter set to generate a single-piece cross-process dataset. The identification unit 303 is used to perform logical consistency verification on the combination of injection molding process parameters and coating process parameters in the single-piece cross-process data based on multiple preset verification rules that respectively characterize the logical relationship between injection molding and coating processes in different physical dimensions. It identifies suspected mis-binding data with logical conflicts, repositions and corrects the identity of the suspected mis-binding data, and outputs a clean full life cycle data archive.

[0101] This system aims to solve the problem of data misbinding caused by the physical repositioning of semi-finished products in the buffer area during the production of interior parts for new energy vehicles. The extraction unit senses and tracks the dynamic position changes of individual semi-finished interior parts between pallets in real time, ensuring that the individual part flow sequence obtained upon outbound is consistent with the actual physical order. Subsequently, the fusion unit, using the true identity of each individual part as the core, accurately correlates and merges the process parameters of different processes (such as injection molding and overmolding), constructing a comprehensive single-part-level cross-process dataset. Finally, the identification unit uses multiple verification rules to perform logical consistency checks on the data, proactively identifying and correcting potential misbinding data, thereby outputting a clean and reliable full lifecycle data archive, providing a solid data foundation for anomaly diagnosis and quality traceability in the production process.

[0102] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for analyzing production process data of interior parts for new energy vehicles, characterized in that, include: In response to the repositioning of individual semi-finished interior parts in the buffer area, the system obtains position transfer event information corresponding to the picking and placing actions of the individual semi-finished interior parts between the pallets through a perception network that integrates vision and radio frequency identification. The system updates the dynamic mapping relationship between the individual semi-finished interior parts and the pallet grid position based on the position transfer event information. When an outbound signal is received, the system extracts the individual real identity identifier of all individual semi-finished interior parts in the current outbound pallet and generates a single-piece flow sequence containing the latest physical sorting according to the preset grid physical traversal order. Using the unique identifier of the interior trim semi-finished product in the single-piece flow sequence as an index, the injection molding process parameter set that uniquely corresponds to the current interior trim semi-finished product to be processed is dynamically retrieved from the stored historical data. After the current interior trim semi-finished product to be processed completes the wrapping process, the wrapping process parameter set generated in real time at the wrapping station is associated and fused with the retrieved injection molding process parameter set to generate a single-piece cross-process dataset. Based on preset multiple verification rules that respectively characterize the logical association between different physical dimensions between injection molding and coating processes, the combination of injection molding process parameters and coating process parameters in the single-piece cross-process data is logically consistent. Suspected mis-binding data with logical conflicts is identified, and the suspected mis-binding data is repositioned and corrected to output a clean full lifecycle data archive.

2. The method according to claim 1, characterized in that, The generation of the single-item flow sequence containing the latest physical order includes: When the injection molding machine completes the mold opening cycle, a unique identification identifier is generated based on the current injection molding equipment number, mold cavity number, production work order number and system timestamp, which is uniquely associated with each semi-finished interior part. An initial mapping relationship is established between the tray number carrying the semi-finished interior part, the grid position inside the tray and the unique identification identifier. The perception network deployed in the buffer area captures the picking and placing actions of individual semi-finished interior parts when they are split, folded, or repositioned between pallets. The picking and placing actions are converted into a position transfer matrix containing the source pallet number, source grid coordinates, target pallet number, and target grid coordinates. The dynamic mapping relationship is updated in real time based on the position transfer matrix. Upon receiving the outbound signal, the latest dynamic mapping relationship of the current outbound pallet is locked, and all bound individual item real identity identifiers are extracted according to the preset grid physical traversal order to generate the individual item flow sequence.

3. The method according to claim 2, characterized in that, After generating a unique, individual identification identifier corresponding to each semi-finished interior trim part produced off the production line, the method further includes: Simultaneously collect multi-source heterogeneous data during the injection molding process. The multi-source heterogeneous data includes real-time temperature of each heating section of the melt barrel, screw speed, peak injection pressure and holding pressure curve, inlet and outlet cooling water temperature and flow rate, and molding cycle time. The multi-source heterogeneous data is aligned with time series and features are extracted to form the injection molding process parameter set. The feature extraction includes converting the continuous injection pressure time series into injection peak pressure, pressure integral area and pressure rise slope. The injection molding process parameter set is bound and stored with the individual part's real identity identifier, as part of the stored historical data.

4. The method according to claim 2, characterized in that, The sensing network deployed in the buffer area captures the picking and placing actions of individual semi-finished interior parts when they are split, folded, or repositioned between pallets. These actions are then converted into a position transition matrix containing the source pallet number, source grid coordinates, target pallet number, and target grid coordinates. The unique physical number of the transfer tray is read and confirmed using radio frequency identification (RFID) technology. A high-resolution industrial camera is used to capture a top-view image stream of the pallet. Image processing algorithms are used to locate the reference points on the pallet edge to establish a two-dimensional coordinate system inside the pallet. A preset target detection model is used to identify the occupancy status of individual semi-finished interior parts at each grid position inside the pallet. When the monitored grid position status changes from occupied to empty, record the source grid coordinates and timestamp of the pick-up action; when the monitored grid position status changes from empty to occupied, record the target grid coordinates and timestamp of the place-down action. The continuously captured pick-up and put-down action event streams are aggregated to generate the position transition matrix.

5. The method according to claim 1, characterized in that, The process involves using the unique identifier of each semi-finished interior trim piece in the single-piece flow sequence as an index to dynamically retrieve the injection molding process parameter set that uniquely corresponds to the currently processed semi-finished interior trim piece from stored historical data. After the currently processed semi-finished interior trim piece completes the covering process, the covering process parameter set generated in real time at the covering station is associated and fused with the retrieved injection molding process parameter set to generate a single-piece level cross-process dataset, including: Upon receiving the trigger signal for the part retrieval action at the wrapping station, the true identity identifier of the interior part corresponding to the currently retrieved interior part semi-finished product is extracted sequentially according to the arrangement order of the interior part semi-finished products in the single-part flow sequence. Using the extracted real identity identifier of the single piece as the query key, a data retrieval request is initiated to the central database to obtain the injection molding process parameter set that uniquely corresponds to the single semi-finished interior part to be processed. After the current interior trim semi-finished product is wrapped, the wrapping process parameter set generated in real time by the wrapping station controller is collected. The wrapping process parameter set includes plasma surface treatment power and time, glue spraying robot trajectory coordinates and coating flow rate, heating oven temperature curve and vacuum forming machine negative pressure value and holding time. The coating process parameter set is spliced ​​and merged with the retrieved injection molding process parameter set to generate the single-piece level cross-process dataset.

6. The method according to claim 1 or 5, characterized in that, The method further includes: The quality inspection station collects quality inspection data for individual semi-finished interior parts that have been covered. Before acquiring the data, the true identity of the individual semi-finished interior parts to be inspected is confirmed by scanning a code or comparing visual textures. Using the confirmed real identity as an index, the quality inspection data is appended and associated with the single-piece cross-process dataset that already contains injection molding process parameters and coating process parameters, forming a historical archive covering data from the entire stages of injection molding, coating, and inspection; The quality inspection data includes at least one of the following: appearance defect type code, defect severity score, key dimension deviation value, surface defect coordinates and area, and serves as one of the input data for the subsequent logical consistency verification.

7. The method according to claim 1, characterized in that, The multiple verification rules include a first verification rule, a second verification rule, and a third verification rule; The first verification rule is used to determine the thermal history correlation deviation between the cooling time and mold temperature of the injection molding process and the product deformation detection results after heating in the subsequent coating process. The second verification rule is used to determine the pressure-mass transfer deviation between the peak injection pressure and the holding pressure integral value of the injection process and the final product weight or density test result; The third verification rule is used to determine the surface treatment correlation deviation between the plasma surface treatment power and time and the adhesive spraying flow rate of the coating process and the final product bonding strength test result. The logical consistency verification of the combination of injection molding process parameters and coating process parameters in the single-piece cross-process dataset includes: for each interior part single semi-finished product dataset, determining whether the actual combination of its injection molding process parameters and coating process parameters has the thermal history correlation deviation, the pressure-mass transfer deviation, or the surface treatment correlation deviation. If the deviation exceeds the preset range, the dataset is marked as the suspected mis-binding data.

8. The method according to claim 7, characterized in that, After marking the suspected mis-bound data, the method further includes: Extract the target interior component semi-finished product marked as suspected of mis-binding, and a predetermined number of interior component semi-finished products that are adjacent to it in the single-piece circulation sequence, to form a suspected single-piece set; Retrieve the injection molding process parameter set of each interior part semi-finished product in the suspected single part set, as well as the corresponding coating process parameter set and quality inspection data, and iterate and rearrange all possible matching combinations between the injection molding process parameter set and the coating and inspection data; Substitute each matching combination back into the first, second, and third verification rules to calculate the conflict penalty score that represents the degree of logical conflict. If, among all matching combinations, there is a unique combination whose conflict penalty score is lower than the preset safety tolerance threshold, then the inter-process data association of each interior component semi-finished product in the suspected component set is re-established according to that combination, and its full life cycle data file is updated.

9. The method according to claim 8, characterized in that, After calculating the conflict penalty score, the method further includes: Obtain the visual perception confidence decay curve corresponding to the fuzzy recognition period when the suspected single item set changes position in the buffer area, and the two-dimensional spatial adjacency matrix of the interior part semi-finished product in the tray; The time dimension offset cost is determined based on the visual perception confidence decay curve, and the spatial dimension offset cost is determined based on the degree of deviation of the adjacency relationship between individual semi-finished interior parts in the two-dimensional spatial adjacency matrix. The conflict penalty score, the time dimension offset cost, and the spatial dimension offset cost are weighted and summed to obtain the comprehensive offset cost. Calculate the difference between the optimal and suboptimal values ​​of the overall offset cost among all matching combinations. If the difference is less than the preset convergence isolation threshold, it is determined that the current data disorder level exceeds the automatic correction capability, automatic correction is abandoned, and a downgrade isolation process is performed on the dataset corresponding to the suspected single item set. The downgrade isolation process includes marking the dataset as untrusted and storing it in hard isolation, while outputting warning information containing abnormal batches, time periods, and cache locations.

10. A data analysis system for the production process of interior parts for new energy vehicles, characterized in that, include: The extraction unit is used to respond to the repositioning of individual semi-finished interior parts in the buffer area. It obtains the position transfer event information corresponding to the picking and placing actions of the individual semi-finished interior parts between the pallets through a perception network that integrates vision and radio frequency identification. It updates the dynamic mapping relationship between the individual semi-finished interior parts and the pallet grid position according to the position transfer event information. When the outbound signal is received, it extracts the individual real identity identifier of all individual semi-finished interior parts in the current outbound pallet and generates a single-piece flow sequence containing the latest physical sorting according to the preset grid physical traversal order. The fusion unit is used to dynamically retrieve the injection molding process parameter set that uniquely corresponds to the current semi-finished interior part in the single-piece flow sequence from the stored historical data, using the single-piece real identity identifier of the semi-finished interior part in the single-piece flow sequence as an index. After the current semi-finished interior part in the single-piece flow sequence completes the covering process, the fusion unit associates and merges the covering process parameter set generated in real time at the covering station with the retrieved injection molding process parameter set to generate a single-piece cross-process dataset. The identification unit is used to perform logical consistency verification on the combination of injection molding process parameters and coating process parameters in the single-piece cross-process data based on multiple preset verification rules that respectively characterize the logical association between different physical dimensions between injection molding and coating processes. It identifies suspected mis-binding data with logical conflicts, repositions and corrects the identity of the suspected mis-binding data, and outputs a clean full life cycle data archive.

Images