A low-quality industrial data-oriented steelmaking molten steel crane weighing system

Abstract

This invention discloses a steelmaking overhead crane weighing system for low-quality industrial data, belonging to the field of steel production technology. The system includes: a data acquisition and fusion module for collecting real-time crane data and production event data from the manufacturing execution system; a feature extraction and data governance module for identifying hoisting operation segments and performing time-series reconstruction and completion of production event data to generate a unified process time sequence table; and a core matching and calculation engine module, including a multi-condition dynamic matching engine and a multi-mode adaptive weight calculation engine, used respectively for multi-dimensional rule matching of hoisting segments and process nodes, and for calculating the weight of molten steel using robust mean or morphological recognition modes. This invention achieves highly robust and accurate molten steel weight tracking even in environments with missing, delayed, and noisy multi-source data, significantly improving the system's process adaptability and maintainability.

Description

Technical Field

[0001] This invention relates to the field of steel production technology, specifically to a steelmaking overhead crane weighing system for low-quality industrial data. Background Technology

[0002] In modern steel production, the weight data of molten steel is crucial for production scheduling, process control, and quality traceability. Currently, steel plants typically use overhead cranes equipped with load cells to weigh ladles in real time and upload the data to a Manufacturing Execution System (MES) for unified management. However, the data quality in actual production environments is generally low, mainly due to the following aspects: (1) The data sources are complex and inconsistent: There are problems such as time asynchrony, non-standard format and missing records among multiple sources of data, such as crane weighing data, MES performance data and process operation signals, which makes it difficult to directly match and utilize the data.

[0003] (2) Signal interference and abnormal fluctuations: During the hoisting process, the crane is affected by mechanical vibration, electromagnetic interference, sloshing of molten steel and other factors, resulting in noise and abnormal fluctuations in the weighing data, which directly affects the accuracy of weight calculation.

[0004] (3) Difficulty in matching process nodes: The time window for the transfer of steel ladles between key processes such as converter and continuous casting is narrow and the regions overlap. Traditional matching methods based on fixed thresholds or simple time windows are prone to mismatch or omission.

[0005] (4) Strong system dependence: Existing weighing systems are highly dependent on the integrity and accuracy of MES data. Once MES data is delayed, lost or incorrect, the system is difficult to correct or supplement on its own, resulting in a break in the data chain.

[0006] Furthermore, most common steelmaking weighing systems currently employ a method of directly acquiring crane weight signals and simply comparing them with MES time-series data. This approach has drawbacks including low matching accuracy, poor data robustness, rigid calculation models, and weak system scalability. Therefore, there is an urgent need for a system capable of achieving highly reliable and accurate molten steel weighing and tracking in low-quality data environments. Summary of the Invention

[0007] (a) Technical problems to be solved The technical problem to be solved by this invention is to propose a steelmaking molten steel crane weighing system for low-quality industrial data. Through multi-source data fusion, intelligent matching and adaptive calculation, the robustness and accuracy of the system in environments with incomplete and inaccurate data are improved.

[0008] (II) Technical Solution To solve the above-mentioned technical problems, the technical solution provided by the present invention includes: The data acquisition and fusion module is used to acquire real-time data of the crane from the crane weighing sensor and positioning device, and asynchronously acquire production event data from the manufacturing execution system. The production event data includes structured production performance data and equipment operation signal data. The feature extraction and data governance module, connected to the data acquisition and fusion module, is used to identify hoisting operation segments from real-time crane data and to perform time-series reconstruction and intelligent completion of production event data to generate a unified process time sequence table. The core matching and computation engine module, connected to the feature extraction and data governance module, includes: (1) Multi-condition dynamic matching engine, which is used to associate and match hoisting operation segments with process time nodes based on composite matching rules consisting of time tolerance interval, spatial location range, weight reasonable range and execution equipment information; (2) Multi-mode adaptive weight calculation engine, which is used to process the weight time series data in the hoisting segment using a robust mean calculation mode or a shape recognition calculation mode according to the matched process type, and extract the weight value of molten steel. The result feedback and configuration management module is connected to the core matching and calculation engine module. It is used to output the matching results and weight calculation results to the external production management system and provide a manual confirmation and supplementation interface. A configurable rule base, independent of the program logic of the above modules, stores the matching rules, signal mapping rules and calculation mode parameters in a structured manner, which is used to drive the multi-condition dynamic matching engine and the multi-mode adaptive weight calculation engine.

[0009] As an improvement, the feature extraction and data governance module includes: The intelligent identification unit for hoisting operations is used to identify complete hoisting operation segments from real-time crane data based on preset operating weight thresholds and minimum continuous operating time thresholds, and to extract their start and end times, start and end positions, and average weight characteristics. The process timing reconstruction and completion unit is used to maintain a unified process time sequence table. Its working logic includes: prioritizing the use of structured production performance data; when production performance data is missing, estimating approximate process times based on equipment operation signals and a predefined signal-time mapping rule base; when all the data is unavailable, inferring the estimated time of the missing process based on the production rhythm analysis model, and marking the source credibility level of all data.

[0010] As an improvement, the multi-condition dynamic matching engine adopts multi-dimensional composite matching rules. Each rule is defined by a combination of at least two conditions from the time tolerance interval, spatial location range, weight reasonable interval, and execution equipment constraints. The matching engine adopts a delayed start mechanism. After the relevant data stabilizes, it selects the segment whose timing is closest to the target process node from the candidate hoisting segments that meet all constraints for association.

[0011] As an improvement, the multi-mode adaptive weight calculation engine includes: The robust mean calculation mode is used to sequentially remove the first and last abnormal segments and filter the fluctuation points based on the statistical distribution in the weight time series data of the hoisting segment, and calculate the weighted or arithmetic mean in the stable operating range. The morphology recognition calculation mode is used to automatically identify key turning points of the morphology of hoisting segments with specific weight change patterns through time series analysis algorithms, calculate the weight value of each stable stage, and output multiple weight results.

[0012] As an improvement, the time-series analysis algorithm used in the morphology recognition calculation mode includes a slope change detection, statistical distribution mutation detection, or change point detection model, which is used to identify the "V-shaped" weight curve in the slag dumping process and calculate the weight before and after slag dumping respectively.

[0013] As an improvement, the configurable rule base is stored independently outside the system program logic, and includes condition parameters for matching rules, parameters for calculation modes, and signal-time mapping rules; administrators can adapt to different production lines or process changes by modifying the parameters in the configuration base without modifying the program code.

[0014] As an improvement, the result feedback and configuration management module also includes a manual confirmation and supplementation interface, which is used by operators to review the results automatically matched and calculated by the system and submit corrected data; the system prioritizes the adoption of data that has been manually confirmed or supplemented.

[0015] As an improvement, in the data acquisition and fusion module, the real-time vehicle data includes ladle weight data collected by the weighing sensor, spatial location data collected by the positioning device, and corresponding timestamps; the production event data includes structured production performance data and equipment operation signal data obtained from the manufacturing execution system, wherein the equipment operation signal data is real-time but the logical relationship is ambiguous.

[0016] As an improvement, a local processing database is also included, which is used to store the original data cache, feature fragments, reconstructed process sequence, matching relationships and calculation results in the internal processing of the system; the system interacts with an external production database, obtains data from the external production database and sends the results back.

[0017] (III) Beneficial Effects The advantages of this invention compared to the prior art are: (1) Through multi-source data fusion and intelligent completion mechanism (especially process sequence reconstruction and completion unit), effective results can still be output stably when MES data is missing or incorrect, ensuring that the production data chain is not broken.

[0018] (2) A multi-condition dynamic matching engine is adopted, which combines delayed start and optimal selection strategies. By utilizing multi-dimensional constraints such as time, space, weight, and equipment, high-precision association is achieved, effectively solving the problems of mismatch and missed match.

[0019] (3) The multi-mode adaptive weight calculation engine effectively resists noise interference through the robust mean calculation mode, and outputs multiple weight values ​​for special processes such as slag dumping through the morphological recognition calculation mode, accurately describing the process and improving data availability.

[0020] (4) The configurable rule base separates business logic from program code. By modifying the configuration, it can quickly adapt to changes in production lines, equipment changes or process adjustments, reduce operation and maintenance costs, and achieve "one-time development, multiple deployments".

[0021] (5) Provide manual confirmation and supplementary data entry interfaces to ensure the reliability of data in extreme cases, combine manual experience with automatic processing, and improve the system's usability and success rate of implementation. Attached Figure Description

[0022] Figure 1 A schematic diagram of the overall architecture and data flow of the system of this invention. Detailed Implementation

[0023] The invention will now be described in further detail with reference to specific embodiments, but this should not be construed as limiting the scope of the subject matter of the invention to the following embodiments.

[0024] like Figure 1 As shown, the system of this invention includes a data acquisition and fusion module, a feature extraction and data governance module, a core matching and computing engine module, a result feedback and configuration management module, as well as a configurable rule base and a local processing database independent of the program logic. An external production database (such as an MES database) serves as the data source and the destination for result output.

[0025] The specific functions of each module are as follows: Data Acquisition and Fusion Module: Communicates with the crane PLC (Programmable Logic Controller) and MES database via industrial Ethernet to acquire real-time data (weight, position, timestamp) from the crane's weighing sensors and positioning devices. It also asynchronously acquires production performance data (such as steel tapping records and continuous casting start records) and equipment operation signals (such as ladle car arrival signals) from the MES. The data is stored in a local processing database.

[0026] Feature extraction and data governance module: Reads raw data from local database, identifies complete hoisting operation segments (based on weight threshold and duration threshold) through hoisting operation intelligent recognition unit, and extracts features; maintains a unified process time sequence table through process time sequence reconstruction and completion unit, completes missing data according to priority (actual data first, signal estimation second, rhythm reasoning as a fallback), and marks the credibility.

[0027] Core matching and calculation engine module: The multi-condition dynamic matching engine matches hoisting segments with process nodes based on multi-dimensional rules in the configurable rule base (delayed start, selecting the closest); the multi-mode adaptive weight calculation engine calculates the weight of molten steel based on the original weight sequence of the matched segments using robust mean or morphological recognition mode according to the process type.

[0028] The results feedback and configuration management module outputs the matching relationships and calculation results to the MES database and provides a human-computer interaction interface for operators to review, confirm or supplement data. Manually entered data has the highest priority.

[0029] Configurable rule base: Stored in XML or JSON format, containing matching rule parameters (time window, spatial region coordinates, weight threshold, device ID), signal mapping rules (such as "signal X = continuous casting start"), and calculation mode parameters (rejection ratio, stable interval definition, etc.). Administrators can edit the files to achieve system adaptation.

[0030] Local processing database: Relational databases or time-series databases are used to store intermediate data such as raw data, feature fragments, process sequence, and matching results.

[0031] Example

[0032] Taking a typical production heat of "converter tapping - ladle refining - continuous casting" as an example, the workflow of the system of the present invention is explained as follows: (1) Data acquisition: The system runs on a timed basis. The data acquisition and fusion module obtains the "tap" event record for this heat from the MES, but finds that the corresponding "continuous casting start" event record is missing. At the same time, the crane real-time data acquisition unit continuously collects the crane's weight and position signals.

[0033] (2) Feature extraction and data completion: The intelligent identification unit for hoisting operations in the feature extraction and data governance module identifies a hoisting segment A (lifting weight greater than the empty ladle threshold and duration meeting the requirements) that is completed near the converter area and meets the valid operation conditions from the crane data. The process timing reconstruction and completion unit detects the missing "continuous casting start" event, and then automatically queries the equipment operation signal. According to the rule in the configurable rule base that "signal X (such as the ladle car arriving at the continuous casting turret signal) corresponds to the arrival of the continuous casting ladle", it calculates the approximate continuous casting start time T_c and marks it as "signal calculation result" and stores it in the unified process time sequence table.

[0034] (3) Intelligent matching: The multi-condition dynamic matching engine in the core matching and calculation engine module loads the corresponding rules for matching.

[0035] For the "Converter Empty Ladle" node, the loading rules are as follows: the time window is a certain period before tapping (e.g., from 10 minutes before tapping to the tapping time), the region is the converter area (geographic coordinate range), the weight range is the typical weight range of an empty ladle (e.g., 30-50 tons), and the crane equipment is a specified group (e.g., converter straddle crane). The engine searches among all identified hoisting segments and finds that segment A fully meets the conditions, and its lowering time is closest to the tapping time. Therefore, segment A is associated with the "Empty Ladle" node of that heat.

[0036] For the "continuous casting full ladle" node, within a time window before time T_c (e.g., within 15 minutes before T_c), the engine starts the corresponding rules to match and find the segment B in the continuous casting area that has a weight greater than the full ladle threshold (e.g., 200 tons) and is completed by continuous casting across the specified crane, and associates it.

[0037] (4) Weight calculation: The multi-mode adaptive weight calculation engine is triggered.

[0038] For segment A associated with "empty packets", a robust mean calculation mode is adopted: after removing the first and last 5% of abnormal segments, filtering out fluctuation points based on the 3σ principle, the arithmetic mean is calculated in the middle stable operating range to obtain the weight W1 of the empty packet.

[0039] For another segment C related to "continuous casting slag dumping" (assuming its weight curve is in a downward trend), the system automatically calls the shape recognition calculation mode, and identifies the start and end points of the slag dumping process (i.e. the start and end of the rapid descent) by detecting the slope change, and calculates the weight W2 of the stable section before slag dumping and the weight W3 of the stable section after slag dumping respectively.

[0040] (5) Result Output: The result feedback and configuration management module encapsulates the final matching relationship of the batch and the calculated weights W1, W2, W3, etc., and sends them back to the designated interface of the MES system to complete the weight data tracking loop for the batch. If the operator finds any abnormalities in the interface (such as matching errors or weight abnormalities), they can submit corrected data through the manual confirmation and supplementation interface. The system will give priority to adopting manual data.

[0041] In this embodiment, the weight of the ladle refining does not need to be calculated repeatedly because its weight entering the station equals the weight of the ladle leaving the converter, and its weight leaving the station equals the weight of the ladle leaving the continuous casting station. The system automatically associates these weights through a matching relationship. In practical use, Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents. In short, if those skilled in the art are inspired by these claims and design similar structural methods and embodiments without departing from the inventive spirit of the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A steelmaking overhead crane weighing system for low-quality industrial data, characterized in that, include: The data acquisition and fusion module is used to acquire real-time data of the crane from the crane weighing sensor and positioning device, and asynchronously acquire production event data from the manufacturing execution system. The production event data includes structured production performance data and equipment operation signal data. The feature extraction and data governance module, connected to the data acquisition and fusion module, is used to identify hoisting operation segments from real-time crane data and to perform time-series reconstruction and intelligent completion of production event data to generate a unified process time sequence table. The core matching and computation engine module, connected to the feature extraction and data governance module, includes: (1) Multi-condition dynamic matching engine, which is used to associate and match hoisting operation segments with process time nodes based on composite matching rules consisting of time tolerance interval, spatial location range, weight reasonable range and execution equipment information; (2) Multi-mode adaptive weight calculation engine, which is used to process the weight time series data in the hoisting segment using a robust mean calculation mode or a shape recognition calculation mode according to the matched process type, and extract the weight value of molten steel. The result feedback and configuration management module is connected to the core matching and calculation engine module. It is used to output the matching results and weight calculation results to the external production management system and provide a manual confirmation and supplementation interface. A configurable rule base, independent of the program logic of the above modules, stores the matching rules, signal mapping rules and calculation mode parameters in a structured manner, which is used to drive the multi-condition dynamic matching engine and the multi-mode adaptive weight calculation engine.

2. The steelmaking molten steel overhead crane weighing system for low-quality industrial data as described in claim 1, characterized in that, The feature extraction and data governance module includes: The intelligent identification unit for hoisting operations is used to identify complete hoisting operation segments from real-time crane data based on preset operating weight thresholds and minimum continuous operating time thresholds, and to extract their start and end times, start and end positions, and average weight characteristics. The process timing reconstruction and completion unit is used to maintain a unified process time sequence table. Its working logic includes: prioritizing the use of structured production performance data; when production performance data is missing, estimating approximate process times based on equipment operation signals and a predefined signal-time mapping rule base; when all the data is unavailable, inferring the estimated time of the missing process based on the production rhythm analysis model, and marking the source credibility level of all data.

3. The steelmaking molten steel overhead crane weighing system for low-quality industrial data as described in claim 1, characterized in that, The multi-condition dynamic matching engine adopts multi-dimensional composite matching rules. Each rule is defined by a combination of at least two conditions from the time tolerance interval, spatial location range, reasonable weight range, and execution equipment constraints. The matching engine adopts a delayed start mechanism. After the relevant data stabilizes, it selects the segment whose timing is closest to the target process node from the candidate hoisting segments that meet all constraints and associates them.

4. The steelmaking molten steel overhead crane weighing system for low-quality industrial data as described in claim 1, characterized in that, The multi-mode adaptive weight calculation engine includes: The robust mean calculation mode is used to sequentially remove the first and last abnormal segments and filter the fluctuation points based on the statistical distribution in the weight time series data of the hoisting segment, and calculate the weighted or arithmetic mean in the stable operating range. The morphology recognition calculation mode is used to automatically identify key turning points of the morphology of hoisting segments with specific weight change patterns through time series analysis algorithms, calculate the weight value of each stable stage, and output multiple weight results.

5. A steelmaking molten steel overhead crane weighing system for low-quality industrial data as described in claim 4, characterized in that, The temporal analysis algorithm used in the morphological recognition calculation mode includes a slope change detection, statistical distribution mutation detection, or change point detection model, which is used to identify the "V-shaped" weight curve in the slag dumping process and calculate the weight before and after slag dumping respectively.

6. The steelmaking molten steel overhead crane weighing system for low-quality industrial data as described in claim 1, characterized in that, The configurable rule base is stored independently outside the system program logic and includes conditional parameters for matching rules, parameters for calculation modes, and signal-time mapping rules. Administrators can adapt to changes in different production lines or processes by modifying the parameters in the configuration base without modifying the program code.

7. A steelmaking molten steel overhead crane weighing system for low-quality industrial data as described in claim 1, characterized in that, The result feedback and configuration management module also includes a manual confirmation and supplementation interface, which is used by operators to review the results automatically matched and calculated by the system and submit corrected data; the system prioritizes the adoption of data that has been manually confirmed or supplemented.

8. A steelmaking molten steel overhead crane weighing system for low-quality industrial data according to claim 1, characterized in that, In the data acquisition and fusion module, the real-time vehicle data includes ladle weight data collected by the weighing sensor, spatial location data collected by the positioning device, and corresponding timestamps; the production event data includes structured production performance data and equipment operation signal data obtained from the manufacturing execution system. The equipment operation signal data is real-time but has an ambiguous logical relationship.

9. A steelmaking molten steel overhead crane weighing system for low-quality industrial data as described in claim 1, characterized in that, It also includes a local processing database for storing raw data caches, feature fragments, reconstructed process sequences, matching relationships, and calculation results during the internal processing of the system; the system interacts with an external production database to obtain data from the external production database and send the results back.

Images