Cold and hot rolled strip steel production quality supervision system based on multi-modal data analysis

Abstract

The invention relates to the technical field of steel production data processing and quality management and control, in particular to a cold and hot rolled strip steel production quality supervision system based on multi-modal data analysis, which comprises a production supervision center, a data acquisition module, a database construction module, a quality chain construction module, a quality supervision module and a display feedback module, through cooperative work of multiple modules, integration and association of multi-modal data of the whole cold and hot rolling strip steel production process are achieved, a complete quality chain is constructed, the problems that data are dispersed, tracing is difficult and optimization is insufficient in traditional quality supervision are solved, and on the basis of'production-process-quality 'whole link association of the quality chain, the quality of the whole cold and hot rolling strip steel production process is improved. According to the technical scheme, key process parameters are monitored in real time, early warning is triggered immediately when the key process parameters exceed the standard range, 'preventive management and control 'is achieved, namely, whole-process closed-loop supervision of'defect detection, source tracing, process optimization, real-time early warning and parameter adjustment' is formed, and the production quality stability and the process optimization efficiency are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of steel production data processing and quality control technology, and in particular to a quality supervision system for cold and hot rolled strip steel production based on multimodal data analysis. Background Technology

[0002] In the production process of hot and cold rolled strip steel, continuous processes such as hot rolling and cold rolling generate massive amounts of multi-dimensional data, covering production data (such as furnace operating parameters, mill speed, coiling temperature, etc.), process data (such as rolling force, reduction, cooling water volume, etc.), and quality data (such as strip thickness deviation, surface roughness, defect images, etc.). These data exhibit significant multimodal characteristics, including numerical, textual, and image types, and are strongly correlated with the quality of hot and cold rolled strip steel.

[0003] At present, although steel companies have accumulated a large amount of production data on cold and hot rolled strip steel, the multimodal data is stored in independent systems of each process, lacking a quality-centric link, which leads to a break in the production quality chain. Traditional data management methods focus on storing single data types and cannot establish a full-link relationship of "process parameters - production status - quality results". When strip steel quality problems occur, it is difficult to quickly trace the source process and key influencing factors, resulting in reduced production quality and increased management costs. For example, to address surface crack defects in cold-rolled strip steel, it is necessary to query data such as rolling temperature in the hot-rolled database, steel composition in the steelmaking database, and rolling force in the cold-rolled database. This tracing process is time-consuming and labor-intensive, and prone to data correlation deviations, which seriously affects the efficiency of solving quality problems and the accuracy of process optimization. Therefore, a solution is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide a quality supervision system for cold and hot rolled strip steel production based on multimodal data analysis, in order to solve the aforementioned technical defects.

[0005] The objective of this invention can be achieved through the following technical solution: a production quality supervision system for hot and cold rolled strip steel based on multimodal data analysis, comprising a production supervision center, a data acquisition module, a database construction module, a quality chain construction module, a quality supervision module, and a display feedback module; The data acquisition module collects multi-type, multi-modal data in real time from the cold and hot rolled strip steel production line through equipment and testing equipment deployed throughout the entire production and testing process, and sends the data to the production monitoring center for storage. The database construction module is used to preprocess multi-type and multi-modal data, and is responsible for building and storing the production line chain base database and the quality chain database; The quality chain construction module clarifies the material production process through the basic information of the quality chain. Based on the set positioning standardized value DW, it matches the quality data bar of the corresponding range in the quality dataset of each production line with the closest comprehensive data bar in the production line dataset, and integrates them to form complete quality chain information. The quality supervision module analyzes the constructed quality chain information and collected surface feature images of hot and cold rolled strip steel to identify the source factors and send them to the display feedback module for display.

[0006] Preferably, the process of building the production line chain database is as follows: S1: The production line chain-based database includes link sets and process datasets; S2: Construction of Link Sets: When new materials are rolled into cold and hot strip steel production lines, new data is created to record the flow and association information of materials in each production line; S3: Construction of process dataset: The process dataset includes process material tracking set, process data identifier set and process storage dataset. The process material tracking set stores the tracking information of materials through each process. The process data identifier set stores the unique identifier, data unit, and spatiotemporal compensation type of each recorded data; The process storage dataset is established separately according to the process flow, storing the production basic detection data, key process setting data, process feed length, and storage time point of the corresponding process.

[0007] Preferably, the process of building the quality chain database is as follows: T1: The quality chain database includes basic information about the quality chain, production line datasets, and quality datasets; T2: The basic information of the quality chain is established at the end of the last process of the batch material. The preceding production records are identified by searching the link set in the production line chain base database, generating a unique identifier BatchId for the whole process and associating it with the material identifiers of each production line. T3: The production line dataset uses the unique identifier BatchId and the production line material identifier as a joint index to locate all production / process data of a batch of materials in a certain production line. At the same time, the production line dataset stores comprehensive data bars containing standardized location information. The comprehensive data bars are organized in a hierarchical format of data header information, process header information, and data identifier-data value. T4: The quality dataset uses a unique identifier BatchId and a production line material identifier as a joint index to store quality data containing standardized positioning information. The quality data is a normalized range identifier for the material segment corresponding to the quality data. The range identifier has a value between 0 and 1 at the beginning and end, with the end being greater than the beginning.

[0008] Preferably, the process of acquiring, storing, and analyzing the comprehensive data bars is as follows: Step 1: Determine the process flow and material head location of the batch of materials in the current production line through the link set; Step 2: Determine the start / end time of materials in each process using the process material tracking set; Step 3: Extract the production / process data for each process within its start / end time from the process storage dataset according to the set span; Step 4: Combine the compensation information of the process data identifier set to generate comprehensive data bars for the corresponding material segments, and finally write them into the production line dataset.

[0009] Preferably, the quality chain information acquisition process is as follows: Step 1: First, clarify the entire production process of the target material through the basic information of the quality chain, and determine the production lines and processes involved; Step 2: Set the positioning standardization value DW; Step 3: Match the quality data in each quality dataset where the minimum value (head) of the normalized range identifier is ≤ the location-standardized value DW, and the maximum value (tail) of the normalized range identifier is ≥ the location-standardized value DW; Step 4: Match the median value of the centralized normalized range identifier of each production line dataset with the comprehensive data bar that is closest to the positioning standardized value DW; Step 5: Integrate all matched comprehensive data bars and quality data bars to form complete quality chain information.

[0010] Preferably, the analysis process of the quality supervision module is as follows: Real-time acquisition of surface feature images of cold and hot rolled strip steel in the cold and hot rolled strip steel production line; inputting the surface feature images of cold and hot rolled strip steel into a pre-set defect recognition model to obtain the output defect recognition results, including the presence / absence of defects and the defect type; If a defective result is obtained based on the defect identification result, the normalized range identifier corresponding to the defect is determined from the quality dataset in the quality chain database by using the unique identifier BatchId + cold rolling production line material identifier through indexing. Based on the analysis in step three, the location standardization information DW corresponding to the normalized range identifier of the defect is obtained; Based on the defined location standardization information (DW), the associated production data and process parameters are retrieved through the quality chain construction module.

[0011] Preferably, the most closely related comprehensive data bar of the determined positioning standardized information DW is retrieved, and the influencing parameters corresponding to the preset threshold of parameter value deviation are filtered out based on the comprehensive data bar. Then, the influencing parameters are set as the source factors.

[0012] The beneficial effects of this invention are as follows: This invention effectively solves the problem of scattered multi-modal data across multiple production lines through data synchronization and association mechanisms, achieving quality-oriented integration of data throughout the entire process. At the same time, it integrates numerical, textual, and image-based multimodal data through a unified index of "unique identifier BatchId + production line material identifier," avoiding the cumbersome operation of cross-system queries.

[0013] This invention also relies on the full-link correlation of the "production-process-quality" quality chain to monitor key process parameters in real time. When parameters exceed the standard range, an early warning is triggered immediately, realizing "preventive control." That is, the quality chain can quickly locate the DW value corresponding to the defect and retrieve the associated full-process production data and process parameters with one click, eliminating the need for manual verification and significantly shortening the problem tracing time. This solves the problems of traditional tracing being time-consuming, labor-intensive, and prone to data correlation deviations. It forms a closed-loop supervision of the entire process of "defect detection-source tracing-process optimization-real-time early warning-parameter adjustment," avoiding the recurrence of similar quality problems and promoting the transformation of cold and hot rolled strip steel production from "post-event remediation" to "pre-event prevention." Attached Figure Description

[0014] The invention will now be further described with reference to the accompanying drawings; Figure 1 This is a flowchart of the system of the present invention; Figure 2 This is a partial reference diagram of Embodiment 1 of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments; Example 1: Please refer to Figures 1 to 2As shown, the present invention is a production quality supervision system for hot and cold rolled strip steel based on multimodal data analysis, including a production supervision center, a data acquisition module, a database construction module, a quality chain construction module, a quality supervision module, and a display feedback module. The production supervision center has a bidirectional communication connection with the data acquisition module, the database construction module, and the quality chain construction module, a unidirectional communication connection with the quality supervision module, and a unidirectional communication connection with the display feedback module. The data acquisition module collects multi-type and multi-modal data in real time from the cold and hot rolled strip steel production line through equipment and testing equipment deployed throughout the entire production process and sends it to the production monitoring center for storage. The multi-type and multi-modal data includes production data, process data, quality data, etc. The production data includes equipment operating data such as heating furnace operating parameters, rolling mill speed, and coiling temperature; the process data includes process setting and execution data such as rolling force, reduction, and cooling water volume; the quality data includes numerical detection data such as strip thickness deviation and surface roughness, as well as non-numerical data such as strip surface defect images and quality inspection text records. The acquisition process supports a combination of periodic acquisition and triggered acquisition to ensure data integrity and timeliness. The database construction module is used to preprocess multi-type and multimodal data (including cleaning, enhancement, etc.), and is responsible for building and storing the production line chain base database and the quality chain database, providing a data foundation for the construction of the quality chain; The process of building the production line chain base database is as follows: S1: The production line chain-based database includes link sets and process datasets; S2: Construction of Link Set: When new materials are put into the cold and hot rolling strip steel production line, new data is created to record the flow association information of materials in each production line. The new data structure includes the current production line material identifier, the current production line production process identifier, the same batch material identifier of the previous production line, the previous production line production process identifier, the current production line material basic information, and the current production line material process flow. The current production line material basic information includes the total length of the material before it is put into the line and the cutting head and tail information of this production line. The process flow is stored in a format of globally unique process code concatenated with delimiters. For example, “J01||J02||J03||...”, where J01 is the code for the heating process, J02 is the code for the roughing process, etc. S3: Construction of process dataset: The process dataset includes process material tracking set, process data identifier set and process storage dataset. The process material tracking set stores the tracking information of materials through each process (at least including production line material identifier, process code, etc.), for example: "Material M001, process J01, entry time T1, exit time T2". The process data identifier set stores the unique identifier, data unit, spatiotemporal compensation type, etc. of each record data, for example: "Process J01, data identifier D01, unit ℃, compensation type time, compensation value 5s"; The process storage dataset is established separately according to the process flow, storing the corresponding process's basic production inspection data, key process setting data, process feed length, storage time point, etc. The process dataset is stored using a periodic recording method, and it also supports deleting redundant data generated by the completed quality chain according to a set period to avoid insufficient storage space. The process of building the quality chain database is as follows: T1: The quality chain database includes basic information about the quality chain, production line datasets, and quality datasets; T2: The basic information of the quality chain is established at the end of the last process of the batch material. The preceding production records are identified by searching the link set in the production line chain base database, and a unique identifier BatchId (such as the hot-rolled material identifier corresponding to the cold-rolled material identifier) ​​is generated for the whole process, and associated with the material identifiers of each production line. If there is no preceding production record (i.e. the material is the initial process material, such as the billet after steelmaking), then add a new record in the quality chain basic information set (composed of multiple quality chain basic information): fill the current production line material identifier into the "link information field", and at the same time create a unique identifier BatchId for the entire process of this batch of materials (used for cross-production line data association). If there are previous production records, the link set is continuously searched upwards (e.g., from cold rolling → hot rolling → steelmaking) until there are no more previous records. Then, the record corresponding to the previous material is found in the quality chain basic information set, and the material identifier of the current production line is added to the "link information field" to realize the connection of material identifiers throughout the entire process. T3: The production line dataset uses a unique identifier BatchId and the production line material identifier as a joint index (e.g., BatchId=B001+Hot-rolled material identifier=M001) to ensure that all production / process data of a certain batch of materials in a certain production line can be accurately located, avoiding data confusion. Meanwhile, the production line dataset stores integrated data bars containing standardized positioning information. The integrated data bars are organized in a hierarchical format of data header information, process header information, and data identifier-data value. The data header information represents the standardized positioning information (the standardized positioning information is the ratio of the middle position of the material segment corresponding to the data bar to the total length of the material (including the head and tail). This value is less than or equal to 1, approaching 0 at the head of the material and approaching 1 at the tail of the material). For example, the data header information "positioning=0.3" means that the middle position of the corresponding material segment is 30% of the total length. Process header information: Identified by a "globally unique process code" (such as heating process code J01, rough rolling process code J02), clearly indicating the process to which the data belongs; Data Identifier - Data Value: Stores specific production / process data. The "Data Identifier" is a unique identifier for data within a process (e.g., D01 represents the furnace temperature). The "Data Value" supports multimodal formats (numerical type such as "1200℃", text type such as "Equipment Normal"). Multiple sets of data are separated by the "Data Separator [-]" (e.g., D01@1200℃-D02@3000kN). Among them, the acquisition and storage of integrated data bars: Step 1: Determine the process flow of the batch of materials in the current production line (e.g., J01→J02→J03) and "material head positioning" (determine the starting position of the material based on the cutting head information) through the link set. Step 2: Determine the start / end time of materials in each process through the process material tracking set (e.g., process J01: start at T1 - end at T2). Step 3: Extract the production / process data of each process within the start / end time from the process storage dataset according to the set span (which can be consistent with the storage period of the process dataset, such as 10 seconds / time); Step 4: Combine the compensation information of the process data identifier set (such as time compensation of 5 seconds and distance compensation of 2 meters) to correct the spatiotemporal deviation of the data, generate the comprehensive data bar of the corresponding material segment, and finally write it into the production line dataset; T4: The quality dataset uses a unique identifier BatchId and a production line material identifier as a joint index to store quality data containing standardized location information. The quality data is the normalized range identifier of the material segment corresponding to the quality data. The range identifier has a value between 0 and 1 at the beginning and end, and the end is greater than the beginning. In summary, the quality chain database is a key data storage carrier for realizing "integration and association of multimodal data across the entire process with quality as the core". Its core function is to receive the basic data from the production line chain base database, and through standardized data structure design and storage logic, establish a direct association between production / process data and quality data, providing structured data support for subsequent quality chain construction, quality traceability and process optimization.

[0017] Example 2: The quality chain construction module is the core module for realizing the full-link association of "production data - process data - quality data". Its core function is to integrate multi-modal data of the entire process based on the production line chain base database and quality chain database built in the early stage, and form a complete production quality chain with quality as the core through standardized positioning and matching logic, so as to provide structured data link support for subsequent quality traceability and process optimization. The quality chain construction module clarifies the material production process through the basic information of the quality chain. Based on the set positioning standardized value DW, it matches the quality data bar of the corresponding range in the quality dataset of each production line with the closest comprehensive data bar in the production line dataset, and integrates them to form complete quality chain information. Quality chain information acquisition process: Step 1: First, clarify the entire production process of the target material through the basic information of the quality chain, and determine the production lines and processes involved; Step 2: Set the positioning standardization value DW; Step 3: Match the quality data in each quality dataset where the minimum value (head) of the normalized range identifier is ≤ the location-standardized value DW, and the maximum value (tail) of the normalized range identifier is ≥ the location-standardized value DW; Step 4: Match the median value of the centralized normalized range identifier of each production line dataset with the comprehensive data bar that is closest to the positioning standardized value DW; Step 5: Integrate all matched comprehensive data bars and quality data bars to form complete quality chain information, and achieve accurate correlation between production data and quality data; The quality chain information is sent to the production monitoring center for storage; For example, the matched "comprehensive data bars" (including basic production data such as mill speed and key process data such as rolling force) and "quality data bars" (including numerical quality data such as thickness deviation, image quality data such as defect images, and text quality records) are associated hierarchically according to the "production line → process → material section"; The integrated data must retain the original identification information of each data item (such as unique identifier BatchId, production line material identifier, process code, and data identifier code) to ensure that each data item can be traced back to the specific production line, process, and collection time, forming a one-to-one correspondence of "material segment location → production process parameters → quality results". For example, "DW=0.3 (30% position of material segment) → hot rolling process rolling force 3000kN, coiling temperature 1150℃ → cold rolling process thickness deviation 0.02mm, no surface defects"; The quality supervision module analyzes the constructed quality chain information and collected surface feature images of hot and cold rolled strip steel to identify the source factors and send them to the display feedback module for display. The specific analysis process is as follows: Real-time acquisition of surface feature images of cold and hot rolled strip steel in the cold and hot rolled strip steel production line; inputting the surface feature images of cold and hot rolled strip steel into a pre-set defect recognition model to obtain the output defect recognition results, including the presence / absence of defects, defect type, etc. If a defect is identified based on the defect identification result, the normalized range identifier corresponding to the defect is determined from the quality dataset in the quality chain database by using the unique identifier BatchId + cold rolling production line material identifier through indexing (e.g., a range identifier of 0.3-0.5 indicates that the defect is located in a material segment of 30%-50% of the total length of the cold rolled strip). Based on step three: matching the median value of the normalized range identifier of each production line dataset with the comprehensive data bar setting that is closest to the location standardization value DW, the location standardization information DW of the normalized range identifier corresponding to the defect is obtained. This value serves as a unified benchmark for subsequent association of production / process data of the entire process, ensuring data spatiotemporal alignment. Based on the defined standardized location information (DW), the associated production data and process parameters are retrieved through the quality chain construction module; By using the unique identifier BatchId in the quality chain basic information set, the entire production chain of this batch of materials (such as steelmaking → hot rolling → cold rolling) is identified, and the production line range from which data needs to be retrieved is determined. Retrieve the most closely related comprehensive data bar of the determined positioning standardization information DW, and filter out the influencing parameters corresponding to the preset threshold of parameter value deviation based on the comprehensive data bar. Then set the influencing parameters as the source factors. The display feedback module is used to respond to the source factors and display them immediately, that is, to monitor data changes in the production process in real time, and combine them with quality chain information to issue early warnings for abnormal parameters that may cause quality problems, thereby achieving precise control of the production process and solving the problems of "time-consuming cross-database queries and data correlation deviations" in traditional traceability. In summary, the data synchronization and association mechanism effectively solves the problem of scattered multi-modal data across multiple production lines, achieving quality-oriented integration of data throughout the entire process. Simultaneously, through a unified index of "unique identifier BatchId + production line material identifier," it integrates numerical, textual, and image-based multimodal data, avoiding cumbersome cross-system queries. Furthermore, relying on the full-link association of the quality chain—"production-process-quality"—it monitors key process parameters in real time, triggering immediate warnings when they exceed standard ranges, achieving "preventive control." This means quickly locating the DW value corresponding to a defect through the quality chain, and retrieving associated full-process production data and process parameters with a single click, eliminating the need for manual verification and significantly shortening problem tracing time. This solves the problems of traditional tracing being time-consuming, labor-intensive, and prone to data association deviations, thus forming a closed-loop supervision of the entire process—"defect detection-source tracing-process optimization-real-time warning-parameter adjustment"—preventing the recurrence of similar quality problems and driving the transformation of cold and hot rolled strip steel production from "post-event remediation" to "pre-event prevention."

[0018] The threshold is set for comparative analysis of results to determine whether they are good or bad. The value of the threshold is determined by a combination of large-scale model analysis of sample data and human experience. It can also be adjusted appropriately based on seasonal or common-sense influencing factors.

[0019] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A quality supervision system for cold and hot rolled strip steel production based on multimodal data analysis, characterized in that, It includes a production monitoring center, a data acquisition module, a database construction module, a quality chain construction module, a quality monitoring module, and a display and feedback module; The data acquisition module collects multi-type, multi-modal data in real time from the cold and hot rolled strip steel production line through equipment and testing equipment deployed throughout the entire production and testing process, and sends the data to the production monitoring center for storage. The database construction module is used to preprocess multi-type and multi-modal data, and is responsible for building and storing the production line chain base database and the quality chain database; The quality chain construction module clarifies the material production process through the basic information of the quality chain. Based on the set positioning standardized value DW, it matches the quality data bar of the corresponding range in the quality dataset of each production line with the closest comprehensive data bar in the production line dataset, and integrates them to form complete quality chain information. The quality supervision module analyzes the constructed quality chain information and collected surface feature images of cold and hot rolled strip steel to identify the source factors and send them to the display feedback module for display.

2. The quality supervision system for cold and hot rolled strip steel production based on multimodal data analysis according to claim 1, characterized in that, The process of building the production line chain-based database is as follows: S1: The production line chain-based database includes link sets and process datasets; S2: Construction of Link Sets: When new materials are rolled into cold and hot strip steel production lines, new data is created to record the flow and association information of materials in each production line; S3: Construction of process dataset: The process dataset includes process material tracking set, process data identifier set and process storage dataset. The process material tracking set stores the tracking information of materials through each process. The process data identifier set stores the unique identifier, data unit, and spatiotemporal compensation type of each recorded data; The process storage dataset is established separately according to the process flow, storing the production basic detection data, key process setting data, process feed length, and storage time point of the corresponding process.

3. The quality supervision system for cold and hot rolled strip steel production based on multimodal data analysis according to claim 1, characterized in that, The process of building the quality chain database is as follows: T1: The quality chain database includes basic information about the quality chain, production line datasets, and quality datasets; T2: The basic information of the quality chain is established at the end of the last process of the batch material. The preceding production records are identified by searching the link set in the production line chain base database, generating a unique identifier BatchId for the whole process and associating it with the material identifiers of each production line. T3: The production line dataset uses the unique identifier BatchId and the production line material identifier as a joint index to locate all production / process data of a batch of materials in a certain production line. At the same time, the production line dataset stores comprehensive data bars containing standardized location information. The comprehensive data bars are organized in a hierarchical format of data header information, process header information, and data identifier-data value. T4: The quality dataset uses a unique identifier BatchId and a production line material identifier as a joint index to store quality data containing standardized positioning information. The quality data is a normalized range identifier for the material segment corresponding to the quality data. The range identifier has a value between 0 and 1 at the beginning and end, with the end being greater than the beginning.

4. The quality supervision system for cold and hot rolled strip steel production based on multimodal data analysis according to claim 3, characterized in that, The process of acquiring, storing, and analyzing the comprehensive data bars is as follows: Step 1: Determine the process flow and material head location of the batch of materials in the current production line through the link set; Step 2: Determine the start / end time of materials in each process using the process material tracking set; Step 3: Extract production / process data for each process within its start / end time from the process storage dataset according to the set span; Step 4: Combine the compensation information of the process data identifier set to generate comprehensive data bars for the corresponding material segments, and finally write them into the production line dataset.

5. The quality supervision system for cold and hot rolled strip steel production based on multimodal data analysis according to claim 1, characterized in that, The process for obtaining the quality chain information is as follows: Step 1: First, clarify the entire production process of the target material through the basic information of the quality chain, and determine the production lines and processes involved; Step 2: Set the positioning standardization value DW; Step 3: Match the quality data in each quality dataset where the minimum value (head) of the normalized range identifier is ≤ the location-standardized value DW, and the maximum value (tail) of the normalized range identifier is ≥ the location-standardized value DW; Step 4: Match the median value of the centralized normalized range identifier of each production line dataset with the comprehensive data bar that is closest to the positioning standardized value DW; Step 5: Integrate all matched comprehensive data bars and quality data bars to form complete quality chain information.

6. The quality supervision system for cold and hot rolled strip steel production based on multimodal data analysis according to claim 1, characterized in that, The analysis process of the quality supervision module is as follows: Real-time acquisition of surface feature images of cold and hot rolled strip steel in the cold and hot rolled strip steel production line; inputting the surface feature images of cold and hot rolled strip steel into a pre-set defect recognition model to obtain the output defect recognition results, including the presence / absence of defects and the defect type; If a defective result is obtained based on the defect identification result, the normalized range identifier corresponding to the defect is determined from the quality dataset in the quality chain database by using the unique identifier BatchId + cold rolling production line material identifier through indexing. Based on the analysis in step three, the location standardization information DW corresponding to the normalized range identifier of the defect is obtained; Based on the defined location standardization information (DW), the associated production data and process parameters are retrieved through the quality chain construction module.

7. The quality supervision system for cold and hot rolled strip steel production based on multimodal data analysis according to claim 6, characterized in that, Retrieve the most closely related comprehensive data bar from the established location standardization information DW, and based on the comprehensive data bar, filter out the influencing parameters corresponding to the preset threshold of parameter value deviation, and then set the influencing parameters as the source factors.

Images