A method, apparatus and storage medium for cross-process material timing repair
By constructing a production time-series data table and using a multi-dimensional similarity formula to repair time-series misaligned abnormal data, the problem of protocol adaptation and time-series alignment of cross-process data in steel bar and wire production was solved, achieving high-precision full-process data association and refined management.
Patent Information
- Application Number
- CN202611043762.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-14
- Publication Date
- 2026-11-10
AI Technical Summary
Existing technologies cannot effectively solve the problems of protocol adaptation, timing alignment, data association, and reliable transmission under strong interference in cross-process data during steel bar and wire production, making it difficult to meet the demand for refined analysis of production data.
By constructing a production timeline data table, obtaining transmission delay time and calculating deviation values, and using a multi-dimensional weighted similarity formula to repair timeline misalignment anomalies, combined with the MPT three-dimensional mapping table and statistical distribution detection, high-precision timeline alignment of materials across processes can be achieved.
It achieves high-precision time alignment of materials across processes, improves the refined management and control capabilities of the entire steel bar and wire rod process, reduces data packet loss rate, and improves the correlation and integrity of data.
Smart Images

Figure CN123068397A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial internet technology in the iron and steel metallurgy industry, specifically to a method, apparatus, and storage medium for cross-process material timing repair. Background Technology
[0002] Steel bar and wire rod manufacturing mainly includes three core processes: steelmaking, billet continuous casting, and bar and wire rod rolling. The production line is equipped with a large number of heterogeneous automated equipment, and the production data exhibits heterogeneous characteristics of multi-level, multi-protocol, and mixed sampling frequency.
[0003] Currently, the industry mainly adopts three types of data acquisition solutions for this scenario, all of which have significant technical shortcomings. For example, adding physical gateways to achieve standardized conversion of multiple protocols is problematic because older PLCs do not support native OPC UA, gateways are prone to single points of failure, and cross-process timing alignment is impossible. Existing technologies also configure independent gateways for data acquisition and uploading for different protocols, but their multi-gateway architecture is cumbersome to maintain; each gateway operates independently without a unified time base, resulting in second-level alignment errors in cross-process data, and the lack of a collaborative mechanism makes it impossible to perceive the cross-process flow status of materials. Integration through periodic extraction of data from upper-level business systems results in minute-level data delays, relies entirely on the accuracy of MES material association, and loses original high-frequency data from the PLC layer, retaining only aggregated statistical results, which cannot meet the needs of refined production analysis.
[0004] Existing technologies generally fall short of meeting the precise data acquisition requirements across the entire steelmaking process. Firstly, protocol compatibility is insufficient; the industry lacks a unified, purely software-based protocol adaptation solution that can be hot-loaded during runtime, while traditional hardware modifications are costly and inflexible. Secondly, there is a time base mismatch in mixed-frequency data; the frequency range between millisecond-level high-frequency process data and furnace secondary low-frequency quality data is as large as 10³~10⁻¹⁰. 7 First, the lack of a unified time base and divide-and-conquer storage strategy results in low data matching accuracy. Second, cross-process time-series drift is a significant problem. The billet process delay time is affected by production disturbances, with fluctuations reaching ±50% of the standard value. Existing technologies lack automatic repair mechanisms for statistical detection and similarity compensation, leading to misalignment of time-series data. Third, data reliability is poor under harsh operating conditions. Electric furnace arcs and rolling mill main motors generate strong electromagnetic interference, and the general acquisition channel lacks redundant protection, resulting in a monthly data loss rate of 3% to 5%.
[0005] In summary, existing data acquisition technologies cannot simultaneously achieve low-cost protocol adaptation, high-precision timing alignment, cross-process data correlation, and reliable transmission under strong interference, making it difficult to meet the production needs of refined management, intelligent optimization, and precise traceability throughout the entire steel bar and wire rod process. Summary of the Invention
[0006] The purpose of this invention is to provide a method, apparatus, and storage medium for cross-process material timing repair. This method achieves high-precision timing alignment of cross-process materials, improving the refined management and control of the entire steel bar and wire rod process.
[0007] To achieve the above objectives, embodiments of the present invention provide a method for cross-process material timing repair. The method includes: acquiring material identifiers, process codes, and timing information of a target steel plant to construct a production timing data table; acquiring the delay times of adjacent processes in the production timing data table; calculating the deviation value from normal operating conditions based on the delay time; identifying timing misalignment anomalies based on the deviation value and the delay time; constructing a maximum search time window based on process constraints, historical averages, and fluctuation coefficients to filter and match the production timing data table to obtain a candidate set; calculating the similarity of data in the candidate set using a multi-dimensional weighted similarity formula; and repairing the timing misalignment anomalies based on the similarity.
[0008] Optionally, the production time sequence data table is a three-dimensional mapping table, wherein the material identifier, process code, and time sequence information are the associated dimensions of the production time sequence data table, and the attribute fields of the production time sequence data table include transmission delay status and alignment status; the time sequence information includes process start time and process end time; the transmission delay status includes normal, abnormally short, and abnormally long; the alignment status includes automatic alignment, compensation repair, manual confirmation, and unresolved.
[0009] Optionally, the transmission delay time is Δti = t(i+1)_start ti_end, where Δti is the time it takes for the material to be transferred from the end of the i-th process to the start of the (i+1)-th process, t(i+1)_start is the start time of the (i+1)-th process, and ti_end is the end time of the i-th process.
[0010] Optionally, the deviation of the transmission delay time from the normal operating condition is: Z_score=|Δti μ| / σ, where Z_score is the deviation value, Δti is the transfer time of material from the end of the i-th process to the start of the i+1-th process, μ is the average transfer time of normal production in the factory's history, and σ is the fluctuation range of normal time.
[0011] Optionally, the step of identifying timing misalignment anomaly data based on the deviation value and the delay time includes: when the deviation value is not greater than the deviation threshold, the process corresponding to the delay time is normal; when the deviation value is greater than the deviation threshold and the delay time is less than the delay threshold, the process corresponding to the delay time is abnormally short; when the deviation value is greater than the deviation threshold and the delay time is not less than the delay threshold, the process corresponding to the delay time is abnormally long.
[0012] Optionally, the process constraints include: the candidate material and the target material belong to the same casting batch and have consecutive cutting sequence numbers; the sequence number position of the candidate coil in the rolling plan and the billet cutting sequence number of the corresponding casting batch satisfy a preset mapping relationship, so that when the coil mother billet information is missing, the corresponding billet can be deduced by back-calculating the rolling plan sequence number.
[0013] Optionally, the step of calculating the similarity of the data in the candidate set using a multi-dimensional weighted similarity formula and repairing the time-series misaligned abnormal data based on the similarity includes: extracting quality features of the candidate set, including C content, Mn content, temperature, diameter, and thickness; calculating the weighted standardized Euclidean distance similarity of the quality features; and when the weighted standardized Euclidean distance similarity is greater than a threshold similarity, then compensating and repairing the time-series misaligned abnormal data.
[0014] On the other hand, this application also proposes a cross-process material timing repair device, which includes: an acquisition module for acquiring material identification, process code, and timing information of a target steel plant, for constructing a production timing data table, and acquiring the delay time of adjacent processes in the production timing data table; a first processing module for calculating the deviation value of the delay time from normal operating conditions, and identifying timing misalignment anomaly data based on the deviation value and the delay time; a second processing module for constructing a maximum search time window based on process constraints, historical averages, and fluctuation coefficients, and for filtering and matching the production timing data table to obtain a candidate set; and a third processing module for calculating the similarity of data in the candidate set using a multi-dimensional weighted similarity formula, and repairing the timing misalignment anomaly data based on the similarity.
[0015] Optionally, the production time sequence data table is a three-dimensional mapping table, wherein the material identifier, process code, and time sequence information are the associated dimensions of the production time sequence data table, and the attribute fields of the production time sequence data table include transmission delay status and alignment status; the time sequence information includes process start time and process end time; the transmission delay status includes normal, abnormally short, and abnormally long; the alignment status includes automatic alignment, compensation repair, manual confirmation, and unresolved.
[0016] On the other hand, this application also proposes a machine-readable storage medium storing instructions for causing a machine to execute any of the cross-process material timing repair methods described above.
[0017] Through the above technical solution, this application constructs a production time sequence data table by using material identification, process code, and time sequence information of steel plants, and then obtains the transmission time. Through MPT three-dimensional mapping table, statistical distribution detection, and multi-dimensional similarity compensation, it improves the complete link coverage of cross-process material association, enables automatic alignment of mixed frequency data under a unified time reference, automatic compensation of time sequence drift of cross-process material association, and reduces data packet loss rate in strong electromagnetic interference environment. It achieves high-precision time sequence alignment of cross-process materials and improves the refined management and control of the entire process of steel bars and wire rods.
[0018] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description
[0019] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of a cross-process material timing repair method according to this application; Figure 2 This is a schematic diagram of the MPT three-dimensional mapping table structure model of this application; Figure 3 This is a schematic diagram of a cross-process material timing repair device according to this application.
[0020] Explanation of reference numerals in the attached figures 100 - Cross-process material timing repair device; 200 - Acquisition module; 300 - First processing module; 400 - Second processing module; 500 - Third processing module. Detailed Implementation
[0021] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.
[0022] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.
[0023] This invention provides a method for cross-process material timing repair, such as... Figure 1 As shown, the method includes: Step S101: Obtain the material identifier, process code, and timing information of the target steel plant to construct a production timing data table, and obtain the transfer time of adjacent processes in the production timing data table; Step S102: Calculate the deviation value from the normal operating condition based on the transmission delay time, and identify the timing misalignment abnormal data based on the deviation value and the transmission delay time. Step S103: Construct a maximum search time window based on process constraints, historical averages, and fluctuation coefficients to filter and match the production time series data table to obtain a candidate set; Step S104: Calculate the similarity of the data in the candidate set using a multi-dimensional weighted similarity formula, and repair the time-series misaligned abnormal data based on the similarity.
[0024] The aforementioned material identification refers to the unique identification information for all raw and auxiliary materials, intermediate billets, and finished steel products within the steel mill. This information is used for end-to-end traceability, warehousing, weighing, quality inspection, and shipment. It includes physical identification (such as steel coil labels, billet inkjet codes, steel pipe stamps, material tags, barcodes / RFID electronic tags, and stack tags) and electronic identification (such as unique material ID, material code, specification grade, furnace number, batch number, smelting number, weight, material, place of origin, warehousing time, and quality inspection level). The above process codes are standardized numbers for each production process in the entire steel production chain, used for production scheduling, time statistics, cost accounting, equipment inspection, quality defect statistics by process, and work order flow. The above time sequence information records the time sequence data of the entire life cycle of materials, completely restoring the time, dwell time, and sequence of each step from raw material to finished product of a piece of steel, including the start time and end time of the process.
[0025] The aforementioned production sequence data table is a three-dimensional mapping table. The material identifier, process code, and sequence information are the associated dimensions of the production sequence data table. The attribute fields of the production sequence data table include delay status and alignment status. The delay status includes normal, abnormally short, and abnormally long; the alignment status includes automatic alignment, compensation and repair, manual confirmation, and unresolved. The delay time is the delay time from the end of the i-th process to the start of the (i+1)-th process. The deviation value is the deviation of the delay time from the normal operating condition. The aforementioned time sequence misalignment anomaly data corresponds to the delay status, including normal process, abnormally short process, and abnormally long process.
[0026] The aforementioned process constraints include: the candidate material and the target material belong to the same casting batch and have consecutive cutting sequence numbers; the sequence number position of the candidate coil in the rolling plan and the billet cutting sequence number of the corresponding casting batch satisfy a preset mapping relationship, so that when the information of the coil mother billet is missing, the corresponding billet can be deduced from the rolling plan sequence number.
[0027] The historical average mentioned above refers to the arithmetic mean of the delay time between two adjacent processes under the three-dimensional grouping of "production line + steel grade + process path". It is calculated by taking the average of all completed production material records within the group over the past N months (N≥3), calculating the delay time, and then taking the mean. For example: for the group of No. 2 steelmaking - large billet, GCr15, LF → continuous casting, data from 150 heats over the past 3 months is taken, μ=1890 seconds. The fluctuation coefficient mentioned above is the historical standard deviation of the delay time under the same group dimension, σ=√[Σ(Δti μ)² / (n 1). σ and μ together constitute the normal fluctuation range [μ 3σ, μ+3σ], participate in the z-score calculation z=|Δti μ| / σ is used to construct the maximum search window Δmax=μ+5σ. For example: σ=285 seconds, normal range [1035s, 2745s]. The above maximum search time window is Δmax=μ+5σ, where μ is the historical average and σ is the fluctuation coefficient. The maximum search window Δmax =μ+5σ, where μ+3σ is the normal judgment boundary (covering approximately 99.73% of the distribution), and an additional 2σ redundancy is added to expand it to μ+5σ (covering approximately 99.9999% of the distribution) for the candidate material search window boundary under abnormal conditions. The 2σ redundancy covers factors such as sensor timestamp cumulative error, seasonal operating condition fluctuations, and short-term equipment adjustments. 5σ is the upper bound of the search window, and it will not be expanded further to avoid the decrease in matching accuracy and the increase in computational overhead caused by the expansion of the candidate set. In the SHORT case, the search window is [ti_end [5σ, ti_end + 5σ], in the case of LONG, the search window is [ti_end, ti_end + 5σ] (time cannot be reversed, only search backward).
[0028] The above screening and matching adopts a two-level progressive rule: Level 1 (initial screening): Candidates must have the same steel grade, specifications (consistent diameter and thickness), and the start time of their process must be within the μ+5σ search window. Level 2 (process rule constraint pruning): Continuous casting sequence continuity rule: within the same casting cycle, the continuous casting machine processes each heat in a fixed order, and the time interval between the start of continuous casting in adjacent heats must be within the normal casting cycle rhythm, without skipping or reversing the order; Rolling plan arrangement order rule: billets in the rolling plan are arranged sequentially according to specifications, and random skipping is not allowed. Rule execution order: all records within the search window, initial screening for the same steel grade, initial screening for the same specifications, continuous casting sequence continuity verification, rolling plan order verification, forming the final candidate set.
[0029] The above candidate set is a set of candidates that may be the same as the target material but were incorrectly recorded within the search time window after two levels of screening. For example, scenario: furnace number 24B01234 (GCr15, diameter 150mm), LF ends at 14:30:00, MES records continuous casting start at 17:45:00 (Δti=11700s, z=34.4, LONG abnormal), search window is [14:30:00, 14:53:45]. The initial candidate set (within the window + GCr15 + 150mm) contains 5 candidate records: C1(24B01235, 14:46:20), C2(24B01236, 14:48:35), C3(24B01237, 14:50:50), C4(24B01238, 14:53:10), and C5(24B01233, 14:43:55). Verification of the continuous casting sequence shows the sequence is →24B01233→24B01234→24B01235, with 24B01234 being followed by 24B01235. Since C2~C5 contradict the casting sequence, only C1 is retained after pruning. The final candidate set is {24B01235}.
[0030] The above multi-dimensional weighted similarity formula is sim=1 / (1+√(Σw_j(v_Mj)). v_Cj)² / σ_j²)), where sim is the similarity, taking values (0, 1]; M is the target material (the current material to be repaired); C is the candidate material; j is the quality feature dimension number (j=1, 2, ..., 5); v_Mj is the quality feature value of the target material in the j-th dimension; v_Cj is the quality feature value of the candidate material in the j-th dimension; σ_j is the historical standard deviation of the j-th dimension under the corresponding steel grade group; w_j is the process sensitivity weight of the j-th dimension (determined by the steel grade, Σw_j=1). Five quality feature dimensions: target C content (σ1 typical value 0.015), target Mn content (σ2 typical value 0.025), target temperature (σ3 typical value 8.5), product diameter (σ4 typical value 1.2), and product thickness (σ5 typical value 0.8).
[0031] The above-mentioned repair refers to the process where, when MPT detects a material delay time z>3, it searches for candidate materials and calculates similarity to find the most matching candidate. This candidate information is then used to automatically correct the broken material identification chain, restoring the traceability link to its integrity. When sim>0.85, it is automatically repaired as COMPENSATED; when sim≤0.85, it is marked as UNRESOLVED. For example: Furnace number 24B01234 (GCr15, diameter 150mm), LF ends at 14:30:00, MES mistakenly records continuous casting start at 17:45:00, Δti=11700s, z=34.42, and is judged as LONG. After screening within the search window [14:30:00, 14:53:45] and pruning according to process rules (see Section 4.3), the candidate set is {C1:24B01235, CC starts at 14:46:20}.
[0032] Calculate the 5-dimensional weighted similarity between the target material M and candidate C1:
[0033] sim = 1 / (1 + √0.004) = 1 / (1 + 0.0632) = 0.941, sim = 0.941 > 0.85, automatic repair successful. Confidence level = 0.941 × (1 0.05 × log21) = 0.941.
[0034] CC.time_start was corrected from 17:45:00 in the original MES record to 14:45:00, CC.time_end was synchronously corrected to 15:12:00, and aligned_status was set to COMPENSATED. The traceback link was restored to complete. It was verified that the MES operator mistakenly recorded 14:45 as 17:45.
[0035] Through the above technical solution, this application constructs a production time sequence data table by using material identification, process code, and time sequence information of steel plants, and then obtains the transmission time. Through MPT three-dimensional mapping table, statistical distribution detection, and multi-dimensional similarity compensation, it improves the complete link coverage of cross-process material association, enables automatic alignment of mixed frequency data under a unified time reference, automatic compensation of time sequence drift of cross-process material association, and reduces data packet loss rate in strong electromagnetic interference environment. It achieves high-precision time sequence alignment of cross-process materials and improves the refined management and control of the entire process of steel bars and wire rods.
[0036] In one embodiment, the production time sequence data table is a three-dimensional mapping table, wherein the material identifier, process code, and time sequence information are the associated dimensions of the production time sequence data table, and the attribute fields of the production time sequence data table include transmission delay status and alignment status; the time sequence information includes process start time and process end time; the transmission delay status includes normal, abnormally short, and abnormally long; the alignment status includes automatic alignment, compensation repair, manual confirmation, and unresolved.
[0037] like Figure 2 As shown in Table 1, the MPT (Material-Process-Time) three-dimensional mapping table is a 14-field data structure, with material identifier as the row, process path as the column, and timestamp as the value.
[0038] Table 1:
[0039] As shown in Table 2, the propagation decision for row 3 (LF→CC): MES original record CC.time_start = 17:45:00, Δt3 = 11700s, z = |11700 1890| / 285 = 34.42, z>3 and Δt>μ+3σ (11700>2745), so it is determined to be a LONG anomaly.
[0040] After the system determines LONG, it filters candidate materials within the search window [14:30:00, 14:53:45]. After pruning according to process rules, candidate C1 (24B01235) is retained. The 5-dimensional weighted similarity sim=0.941>0.85 is calculated, and automatic repair is successful. CC.time_start is corrected from 17:45:00 to 14:45:00, CC.time_end is simultaneously corrected to 15:12:00, and aligned_status is set to COMPENSATED. The original detection value before correction is retained in the table, and the corrected alignment link (Δti = 900s) is used for subsequent quality traceability.
[0041] Table 2:
[0042] The transition relationship between the above four states is as follows: AUTO indicates normal automatic alignment; when z>3 and similarity sim>0.85, it transitions to COMPENSATED (compensation and repair successful); when z>3 but sim ≤ 0.85, it transitions to UNRESOLVED (automatic repair failed, awaiting manual confirmation); after manual confirmation, it transitions to MANUAL.
[0043] The MPT table construction relies on three external data sources: MES process specifications provide material identification, steel type, production line, and process path; PLC / SCADA real-time data provides the actual start and end times of each process; and a historical statistics database provides μ and σ grouped in three dimensions (updated monthly).
[0044] In one embodiment, the transmission delay time is Δti = t(i+1)_start ti_end, where Δti is the time it takes for the material to be transferred from the end of the i-th process to the start of the (i+1)-th process, t(i+1)_start is the start time of the (i+1)-th process, and ti_end is the end time of the i-th process.
[0045] For example, the MPT search hit all four process time windows. Delay time calculation: Δt1(BOF→LF) = 408 seconds (NORMAL, z = 0.35), Δt2(LF→CC) = 950 seconds (NORMAL, z = 0.25), Δt3(CC→HR) = 8970 seconds (LONG, z = 5.97).
[0046] In one embodiment, the deviation of the transmission time from the normal operating condition is: Z_score=|Δti μ| / σ, where Z_score is the deviation value, Δti is the transfer time of material from the end of the i-th process to the start of the i+1-th process, μ is the average transfer time of normal production in the factory's history, and σ is the fluctuation range of normal time.
[0047] For example, for normal materials: furnace number 24B01205, Δti = 1850s, z = |1850s| 1890 / 285 = 0.14, z ≤ 3, judged NORMAL. The deviation of this furnace from the mean is only 0.14 standard deviations, which is within the normal fluctuation range. Normal material (near the boundary): Furnace number 24B01203, Δti = 1780s, z = |1780| 1890 / 285 = 0.39, z ≤ 3, judged NORMAL. 1780s still falls within the normal range [1035s, 2745s]. Abnormally long (LONG): Furnace number 24B01234, MES original record CC.time_start = 17:45:00, Δti = 11700s, z = |11700| 1890| / 285 = 34.42, z>3 and Δti>μ+3σ, therefore LONG is determined. Investigation revealed that the operator mistakenly recorded 14:45 as 17:45. Abnormally short (SHORT): Assuming a certain furnace LF cycle ends at 14:30:00, and the MES erroneously records CC starting at 14:32:00, Δti = 120s, z =|120 1890| / 285 = 6.21, z>3 and Δti<μ 3σ, determine SHORT. Physically, it is impossible to complete the hoisting and preparation in two minutes; it may be due to a sensor timestamp jump.
[0048] Judgment rules: |z|≤3 is marked as NORMAL, no processing required; z>3 and Δti>μ+3σ is marked as LONG, triggering similarity repair; z>3 and Δti<μ The 3σ threshold is set to SHORT to trigger similarity repair; when there is no μ / σ reference, it is set to UNKNOWN and incrementally updated after manual verification. The 3σ threshold covers approximately 99.73% of samples under a normal distribution.
[0049] In one embodiment, identifying timing misalignment anomalies based on the deviation value and the delay time includes: when the deviation value is not greater than a deviation threshold, the process corresponding to the delay time is normal; when the deviation value is greater than the deviation threshold and the delay time is less than the delay threshold, the process corresponding to the delay time is abnormally short; when the deviation value is greater than the deviation threshold and the delay time is not less than the delay threshold, the process corresponding to the delay time is abnormally long.
[0050] For example, when the deviation value is ≤3, the process corresponding to the delay time is normal (marked as NORMAL); when the deviation value is >3 and the delay time is <μ... When the deviation is 3σ, the process corresponding to the delay time is abnormally short (marked as SHORT). When the deviation value is >3 and the delay time is >μ+3σ, the process corresponding to the delay time is abnormally long (marked as LONG). The statistical parameters are calculated by taking the data from the most recent N months in three-dimensional grouping of production line, steel grade and process path. The UNKNOWN records are updated incrementally after manual confirmation of μ and σ.
[0051] This method uses z-score standardization to eliminate dimensional differences between processes and steel grades, achieving a unified anomaly detection threshold. The delay times for different process paths vary significantly. Continuous casting delays are typically 1000–3000 seconds, while delays from the heating furnace to rolling can reach 5000–8000 seconds. If an absolute threshold (e.g., "delay > 3600 seconds is an anomaly") is used, numerous false alarms will occur in the heating furnace to rolling section. z-score normalizes the delay time into a deviation value in standard deviation units, making the deviations between different groups comparable. |z| > 3 is uniformly used as the anomaly detection threshold, and the detection result is unaffected by process or steel grade.
[0052] For example: For a furnace in the LF→CC process, Δti = 3200s, μ = 1890, σ = 285, z = 4.60, it is judged as abnormal; For a furnace in the HR→Roll process, Δti = 4500s, μ = 5700, σ = 420, z = 2.86, it is judged as normal. The absolute deviations of the two furnaces are both on the order of ±1200~1300s, and they can be accurately distinguished after z-score standardization.
[0053] This method uses bidirectional anomaly identification via LONG and SHORT, covering two types of timing misalignment scenarios. LONG (excessively long) indicates a delay significantly greater than normal, typically caused by an error in time input by the MES operator; SHORT (excessively short) indicates a delay significantly less than normal, typically caused by sensor timestamp jumps or PLC clock drift. Traditional methods only use a unidirectional upper bound for a rough judgment and cannot identify SHORT anomalies. This method uses μ... The 3σ lower bound enables bidirectional coverage. Incremental updates are supported through the UNKNOWN status, allowing the system to operate even with insufficient sample size and continuously improving decision coverage as production data accumulates. New steel grades or new production lines are initially marked as UNKNOWN. Once the number of manually confirmed records reaches the minimum statistical sample size (N≥30), the system automatically generates μ and σ, and UNKNOWN transitions to the normal decision pipeline.
[0054] In one embodiment, the process constraints include: the candidate material and the target material belong to the same casting batch and have consecutive cutting sequence numbers; the sequence number position of the candidate coil in the rolling plan and the billet cutting sequence number of the corresponding casting batch satisfy a preset mapping relationship, so that when the coil mother billet information is missing, the corresponding billet can be deduced by the rolling plan sequence number.
[0055] Specifically, the process constraint requires that the candidate material and the target material belong to the same casting batch and have consecutive cutting sequence numbers; the rolling plan coil arrangement order rule requires that the sequence number position of the candidate coil in the rolling plan and the billet cutting sequence number of the corresponding casting batch satisfy a preset mapping relationship, so that when the coil mother billet information is missing, the corresponding billet can be deduced from the rolling plan sequence number.
[0056] In one embodiment, the step of calculating the similarity of the data in the candidate set using a multi-dimensional weighted similarity formula and repairing the time-series misaligned abnormal data based on the similarity includes: extracting quality features of the candidate set, including C content, Mn content, temperature, diameter, and thickness; calculating the weighted standardized Euclidean distance similarity of the quality features; and when the weighted standardized Euclidean distance similarity is greater than a threshold similarity, then compensating and repairing the time-series misaligned abnormal data.
[0057] This method extracts quality features from the candidate set, including C content, Mn content, temperature, diameter, and thickness; calculates the weighted normalized Euclidean distance similarity (sim) of these quality features, and compensates for and repairs time-series misaligned abnormal data when sim > 0.85. Through joint constraints of five quality features, the risk of mismatches in a single dimension is significantly reduced. If only a single dimension (such as matching only the same diameter) is used for repair, a unique correct match cannot be determined among materials of the same specification from the same production line. This application employs joint constraints of five quality dimensions; with each additional dimension, the probability of false matches within the candidate set decreases exponentially. The five-dimensional joint constraint effectively distinguishes different batches of materials of the same specification from the same production line.
[0058] This method also employs a weighting mechanism to automatically adapt different steel grades to varying process sensitivity distributions. The weight vector w_j is determined by the normalization of the steel grade's process sensitivity (Σw_j=1), with different weight configurations automatically switching between different steel grades. For example, GCr15 bearing steel has the highest sensitivity to C content (w_1=0.35), while Q355D low-alloy steel has relatively high sensitivity to Mn content and temperature (w_2=0.25, w_3=0.25). The same formula adapts to different steel grades without requiring separate parameter configurations. Normalization eliminates dimensional differences between quality characteristics. With carbon content on the order of 0.01 and temperature on the order of 1500, the temperature term completely dominates when directly calculating the Euclidean distance. After dividing by the historical standard deviation σ_j of each dimension, all dimensions are normalized to comparable relative deviation scales, ensuring that the contribution of each dimension is determined by its actual deviation and process weights, rather than by dimensional magnitude. The sim>0.85 threshold balances the success rate of automatic repair and the risk of false repair. Setting the threshold too low increases the false positive rate, while setting it too high requires manual handling of numerous anomalies. A threshold of 0.85 is considered the optimal threshold, indicating a high degree of agreement across at least four of the five dimensions. Items below the threshold are marked as UNRESOLVED and require manual review. The confidence score is calculated as: confidence = sim × (1 / 2) * ... 0.05 × log2|candidates|) provides a priority ranking basis for the review.
[0059] In one embodiment, this embodiment adopts a bottom-up three-level progressive design, with the levels as follows: a unified protocol adaptation layer, a hybrid frequency buffer pool, and a process timing alignment engine. The overall data flow follows a unidirectional progressive mode. Each level performs its own function: the first level realizes unified access to multi-source data, the second level completes differentiated caching and division, and the third level realizes global timing data alignment, adapting to industrial production data acquisition and processing scenarios.
[0060] The aforementioned unified protocol adaptation layer is used to shield the differences between various industrial protocols, build standardized data access capabilities, and achieve unified data collection from multiple devices and protocols. Specifically, this method defines a universal standard interface to unify the entire lifecycle operation of all industrial protocols, including four core capabilities: connection establishment, batch data reading, periodic data subscription, and connection termination. Through unified protocol adaptation at the software layer, it eliminates the need for hardware procurement and maintenance investment for more than five independent protocol gateways.
[0061] The aforementioned hybrid frequency buffer pool is used to standardize timestamp processing and hierarchically cache industrial data from unified access, distinguishing between high-frequency real-time data and low-frequency batch data to achieve efficient data caching and reliable persistence. After protocol parsing is completed and the data is structured to generate TagValue objects, a 64-bit nanosecond-level unified timestamp is immediately generated. A multi-level time synchronization degradation strategy is adopted, specifically including a high-frequency queue (HFQ) and a low-frequency queue (LFQ). This method, through NTP unified timestamps (deviation <1ms) + dual network card redundancy (switching <50ms) + circular buffer (60-second data patching), reduces packet loss rate and acquisition latency in environments with strong electromagnetic interference.
[0062] The aforementioned process sequence alignment engine, based on underlying unified protocol data and hierarchical cached data, completes the global association and alignment of process, material, and time sequence data throughout the entire process. It is the core data governance module at the top level of the architecture. For example... Figure 2 As shown, this method improves the complete link coverage of cross-process material association by using MPT three-dimensional mapping table + statistical distribution detection (μ±3σ) + multi-dimensional similarity compensation.
[0063] The aforementioned process sequence alignment engine uses the unique material identifier × process code × sequence information as the three-dimensional core association dimension, containing a total of 14 attribute fields, including material identifier, process start / end time, transfer time, transfer status, and alignment status.
[0064] Timing alignment algorithm flow: MPT table retrieval → delay time calculation (Δti = ti + 1_start) ti_end) → Statistical distribution comparison (z-score=|Δti) μ| / σ)→Abnormal trigger compensation sub-process→Output complete alignment link.
[0065] Compensation sub-process: Time window search → Process rule constraints (continuous casting same casting cycle constraint, rolling same plan constraint) → Candidate set non-empty judgment → Weighted similarity calculation (sim=1 / (1+√(Σw_j(v_Mj)) v_Cj)² / σ_j²))), select the candidate repair association with the highest similarity and >0.85.
[0066] This application configures the data acquisition nodes with dual gigabit physical network ports, using the Linux kernel bonding driver in Active-Backup mode (mode=1). The switchover sequence is as follows: T0 failure occurs → T0+30ms MII detection confirmation → T0+45ms eth1 takes over the Active role → T0+50ms traffic recovery. The total switchover window is approximately 20ms, and the HFQ ring buffer retains 60 seconds of history, far exceeding the switchover window, ensuring complete data recovery.
[0067] In one embodiment, an industrial edge data acquisition server is deployed in the converter station control room. Data sources include: the converter body PLC (S7 ISO-on-TCP protocol) acquiring 50 data points such as oxygen lance position, bottom-blown argon flow rate, furnace pressure, and converter tilt angle, with a 100ms subscription cycle; the secondary lance temperature and oxygen determination system (Modbus TCP protocol, event-driven acquisition, data stored in the LFQ persistent queue); and the converter flue gas analyzer (Siemens ULTRAMAT 23), provided via CP443-1 Advanced with an OPC UAServer, subscription mode, publishing interval = 1000ms. The data flow is as follows: protocol parsing → NTP timestamp tagging → splitter (HIGH data enters the HFQ ring buffer / LOW data enters the LFQ persistent queue) → batch upload to the central server via the network.
[0068] In one embodiment, the continuous casting PLC adopts an S7-414H redundant system. The drive module has built-in S7-400H redundancy management logic: upon startup, it simultaneously initiates connections to Rack 0 and Rack 1 → reads SZL 0x71 every 500ms heartbeat to confirm the Active status → determines a fault after two consecutive timeouts (>1 second) → automatically switches to the Standby rack → replenishes data during the switchover period from the HFQ ring buffer. The collected data includes 80 data points such as tundish temperature, crystallizer level (4 channels, 50ms), casting speed, cooling water flow / temperature difference, and electromagnetic stirring current / frequency.
[0069] In one embodiment, taking furnace number 24B01234 (steel grade GCr15, specification 1500mm×230mm square billet) as an example. Process path P=[BOF, LF, CC, HR]. MPT search hit all 4 process time windows. Transfer time calculation: Δt1(BOF→LF)=408 seconds (NORMAL, z=0.35), Δt2(LF→CC)=950 seconds (NORMAL, z=0.25), Δt3(CC→HR)=8970 seconds (LONG, z=5.97). The CC→HR anomaly triggers the compensation subprocess, finding 3 candidates in the [13:15:30, 15:30:30] window. After similarity calculation C2 (sim=0.929>0.85), it was selected as the repair association. Afterwards, manual review confirmed that the MES input error caused the association break, and the algorithm successfully repaired it.
[0070] In one embodiment, the hot rolling production line includes a walking beam furnace → roughing mill (R1+R2) → finishing mill (F1~F7) → controlled cooling → cooling bed. Data sources include: a hot rolling main line PLC (S7-1500+ET200MP), 120 data points, a mixed cycle of 10ms / 100ms / 500ms / 1s; a multi-functional dimensional accuracy instrument (Zumbach / KelK), OPC UA interface, triggered after each coil is completed; and an eddy current / ultrasonic flaw detection system with TCP Socket + proprietary protocol. After the hot rolling furnace entry identification system detects the RFID tag, it automatically establishes a CC→RM association, which is consistent with the cross-validation results of the compensation and repair in Embodiment 3. The entire data chain for furnace number 24B01234 is completely closed.
[0071] This application discloses a method and system for heterogeneous data hybrid acquisition and time-series alignment across the entire steelmaking, refining, continuous casting, and rolling process, belonging to the field of industrial internet technology in the iron and steel metallurgy industry. In a multi-protocol unified adaptation layer, the lifecycle operations of various industrial protocols are encapsulated into a unified interface. Each protocol is implemented using pluggable driver plugins and supports hot loading. A hybrid frequency buffer pool consisting of an HFQ high-frequency circular buffer queue and an LFQ low-frequency persistent queue is constructed, and a 64-bit nanosecond-level unified timestamp is immediately applied to the data upon arrival using an NTP synchronous clock. An MPT three-dimensional mapping table (i.e., a production time-series data table) is established, calculating cross-process delay times and comparing them with historical statistical distributions (μ±3σ). Abnormal delays exceeding the normal range trigger time-series alignment compensation based on time window search, process rule constraints, and weighted Euclidean distance similarity calculation. This achieves unified acquisition of data from multiple heterogeneous protocols without increasing middleware hardware, reducing data packet loss rate, lowering acquisition latency, and improving the integrity of cross-process traceability links.
[0072] On the other hand, this application also proposes a cross-process material timing repair device, such as... Figure 3As shown, the cross-process material timing repair device 100 includes: an acquisition module 200, used to acquire the material identifier, process code, and timing information of the target steel plant, and to construct a production timing data table, and acquire the transmission delay time of adjacent processes in the production timing data table; a first processing module 300, used to calculate the deviation value of the transmission delay time from the normal operating conditions, and to identify timing misalignment abnormal data based on the deviation value and the transmission delay time; a second processing module 400, used to construct a maximum search time window based on process constraints, historical averages, and fluctuation coefficients, and to filter and match the production timing data table to obtain a candidate set; and a third processing module 500, used to calculate the similarity of the data in the candidate set using a multi-dimensional weighted similarity formula, and to repair the timing misalignment abnormal data based on the similarity.
[0073] In one embodiment, the production time sequence data table is a three-dimensional mapping table, wherein the material identifier, process code, and time sequence information are the associated dimensions of the production time sequence data table, and the attribute fields of the production time sequence data table include transmission delay status and alignment status; the time sequence information includes process start time and process end time; the transmission delay status includes normal, abnormally short, and abnormally long; the alignment status includes automatic alignment, compensation repair, manual confirmation, and unresolved.
[0074] Through the above technical solution, this application constructs a production time sequence data table by using material identification, process code, and time sequence information of steel plants, and then obtains the transmission time. Through MPT three-dimensional mapping table, statistical distribution detection, and multi-dimensional similarity compensation, it improves the complete link coverage of cross-process material association, enables automatic alignment of mixed frequency data under a unified time reference, automatic compensation of time sequence drift of cross-process material association, and reduces data packet loss rate in strong electromagnetic interference environment. It achieves high-precision time sequence alignment of cross-process materials and improves the refined management and control of the entire process of steel bars and wire rods.
[0075] The cross-process material timing repair device 100 includes a processor and a memory. The aforementioned acquisition module 200, first processing module 300, second processing module 400, and third processing module 500 are all stored in the memory as program units. The processor executes the aforementioned program units stored in the memory to realize the corresponding functions.
[0076] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and adjusting kernel parameters can improve the refined management and control of the entire steel bar and wire rod production process.
[0077] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.
[0078] This invention provides a storage medium storing a program that, when executed by a processor, implements the cross-process material timing repair method.
[0079] This invention provides a processor for running a program, wherein the program executes the cross-process material timing repair method during runtime.
[0080] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps: acquiring material identifiers, process codes, and timing information of a target steel plant to construct a production timing data table; acquiring the delay time of adjacent processes in the production timing data table; calculating the deviation value from normal operating conditions based on the delay time; identifying timing misalignment anomalies based on the deviation value and the delay time; constructing a maximum search time window based on process constraints, historical averages, and fluctuation coefficients to filter and match the production timing data table to obtain a candidate set; calculating the similarity of data in the candidate set using a multi-dimensional weighted similarity formula; and repairing the timing misalignment anomalies based on the similarity. The device described herein can be a server, PC, PAD, mobile phone, etc.
[0081] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program with the following method steps: acquiring material identifiers, process codes, and timing information of a target steel plant to construct a production timing data table; acquiring the delay time of adjacent processes in the production timing data table; calculating the deviation value from normal operating conditions based on the delay time; identifying timing misalignment anomaly data based on the deviation value and the delay time; constructing a maximum search time window based on process constraints, historical averages, and fluctuation coefficients to filter and match the production timing data table to obtain a candidate set; calculating the similarity of data in the candidate set using a multi-dimensional weighted similarity formula; and repairing the timing misalignment anomaly data based on the similarity.
[0082] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0083] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0084] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0085] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0086] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0087] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0088] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0089] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0090] The above are merely embodiments of this application and are not intended to limit the scope 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 the claims of this application.
Claims
1. A method for cross-process material timing repair, characterized in that, The method includes: Obtain the material identifier, process code, and timing information of the target steel plant to construct a production timing data table, and obtain the transfer time between adjacent processes in the production timing data table; The deviation value from the normal operating condition is calculated based on the transmission delay time, and the timing misalignment abnormal data is identified based on the deviation value and the transmission delay time. A maximum search time window is constructed based on process constraints, historical averages, and fluctuation coefficients to filter and match the production time series data table to obtain a candidate set. The similarity of the data in the candidate set is calculated using a multi-dimensional weighted similarity formula, and the temporal misalignment anomaly data is repaired based on the similarity.
2. The method according to claim 1, characterized in that, The production time sequence data table is a three-dimensional mapping table, wherein the material identifier, process code, and time sequence information are the associated dimensions of the production time sequence data table, and the attribute fields of the production time sequence data table include transmission status and alignment status. The timing information includes the process start time and the process end time; The transmission status includes normal, abnormally short, and abnormally long; The alignment status includes automatic alignment, compensation and repair, manual confirmation, and unresolved.
3. The method according to claim 1, characterized in that, The delay time is Δti = t(i+1)_start ti_end, Where Δti is the time it takes for the material to be transferred from the end of the i-th process to the start of the (i+1)-th process, t(i+1)_start is the start time of the (i+1)-th process, and ti_end is the end time of the i-th process.
4. The method according to claim 1, characterized in that, The deviation of the delay time from the normal operating condition is: Z_score=|Δti μ| / σ, Where Z_score is the deviation value, Δti is the transfer time of material from the end of the i-th process to the start of the i+1-th process, μ is the average transfer time of normal production in the factory's history, and σ is the fluctuation range of normal time.
5. The method according to claim 1, characterized in that, The step of identifying timing misalignment anomaly data based on the deviation value and transmission delay time includes: When the deviation value is not greater than the deviation threshold, the process corresponding to the delay time is normal. When the deviation value is greater than the deviation threshold and the delay time is less than the delay threshold, the process corresponding to the delay time is abnormally short. When the deviation value is greater than the deviation threshold and the delay time is not less than the delay threshold, the process corresponding to the delay time is abnormally long.
6. The method according to claim 1, characterized in that, The process constraints include: the candidate material and the target material belong to the same casting batch and the cutting sequence number is consecutive; the sequence number position of the candidate coil in the rolling plan and the billet cutting sequence number of the corresponding casting batch satisfy a preset mapping relationship, so that when the coil mother billet information is missing, the corresponding billet can be deduced by the rolling plan sequence number.
7. The method according to claim 1, characterized in that, The step of calculating the similarity of the data in the candidate set using a multi-dimensional weighted similarity formula, and repairing the temporally misaligned abnormal data based on the similarity, includes: The quality characteristics extracted from the candidate set include C content, Mn content, temperature, diameter, and thickness; Calculate the weighted standardized Euclidean distance similarity of the quality features. When the weighted standardized Euclidean distance similarity is greater than the threshold similarity, the time-series misaligned abnormal data is compensated and repaired.
8. A cross-process material timing repair device, characterized in that, The device includes: The acquisition module is used to acquire the material identifier, process code, and timing information of the target steel plant, and to construct a production timing data table, and to acquire the transmission time of adjacent processes in the production timing data table; The first processing module is used to calculate the deviation value of the transmission delay time from the normal working condition, and to identify the timing misalignment abnormal data based on the deviation value and the transmission delay time. The second processing module is used to construct a maximum search time window based on process constraints, historical averages and fluctuation coefficients, and to filter and match the production time series data table to obtain a candidate set; The third processing module is used to calculate the similarity of the data in the candidate set using a multi-dimensional weighted similarity formula, and to repair the time-series misaligned abnormal data based on the similarity.
9. The apparatus according to claim 8, characterized in that, The production time sequence data table is a three-dimensional mapping table, wherein the material identifier, process code, and time sequence information are the associated dimensions of the production time sequence data table, and the attribute fields of the production time sequence data table include transmission status and alignment status. The timing information includes the process start time and the process end time; The transmission status includes normal, abnormally short, and abnormally long; The alignment status includes automatic alignment, compensation and repair, manual confirmation, and unresolved.
10. A machine-readable storage medium storing instructions thereon, characterized in that, This instruction is used to cause the machine to perform any one of the cross-process material timing repair methods of claims 1-7 of this application.
