Medium plate quality tracing method and system based on pass quality genetic coefficient

Abstract

The application provides a plate quality tracing method and system based on pass quality genetic coefficient, and relates to the technical field of steel rolling quality detection. By collecting and preprocessing the pass process data and finished product quality detection data of plate rough rolling and finishing rolling, combining the reversible rolling time sequence and the flow characteristics of the rolled piece, a pass-to-pass quality genetic relationship matrix with rolling direction information is constructed. A variety of time sequence correlation quantitative algorithms are used to calculate the genetic coefficient of abnormal quality indicators on process parameters, support the general Deng's grey correlation degree and the special grey comprehensive correlation degree of the shape index, and calculate the genetic contribution rate from the finishing rolling to the rough rolling. The root pass and key process parameters are accurately located, and the quality abnormal evolution path diagram is generated. The application solves the problems of traditional tracing, such as fuzzy cause and effect, inaccurate positioning and poor working condition adaptability, and significantly improves the accuracy and real-time performance of plate quality tracing.

Description

Technical Field

[0001] This invention relates to the field of quality control technology in the metallurgical industry, and more specifically, to a method and system for tracing the quality of medium and heavy plates based on the quality inheritance coefficient of each pass. Background Technology

[0002] Medium and heavy plates (thickness of 4mm and above) are widely used in important fields such as shipbuilding, bridges, construction, pressure vessels, marine engineering, and machinery manufacturing, and are one of the basic raw materials for national economic construction.

[0003] Medium and heavy plate rolling typically employs a single-stand or multi-stand (roughing mill + finishing mill) reciprocating rolling process. A steel plate needs to undergo multiple passes on the same mill to achieve the required finished thickness. The number of rolling passes is usually 5 to 15. The material exhibits significant deformation and temperature gradients along the thickness direction, making the production process prone to fluctuations. With the continuous growth in demand for lightweight and high-performance high-end equipment, the quality requirements for medium and heavy plates are becoming increasingly stringent, with higher standards being set for strength uniformity, residual stress control, and plate shape accuracy. Simultaneously, in 2024-2025, the Ministry of Industry and Information Technology and other departments jointly issued the "Work Plan for Stabilizing Growth in the Steel Industry," explicitly promoting a significant improvement in the level of digital development in the steel industry and encouraging enterprises to utilize new-generation information technologies such as artificial intelligence and big data to optimize production processes and strengthen quality control. However, the complexity and variability of the medium and heavy plate rolling process pose severe challenges to quality control.

[0004] More complexly, quality defects in medium and heavy plate rolling exhibit significant inter-process inheritance effects and inter-pass transmission characteristics. Internal defects such as segregation, porosity, and inclusions generated in the upstream continuous casting process can be inherited, amplified, or transformed in subsequent rolling processes, directly affecting the final product. During rolling, the inheritance of quality anomalies between passes is equally significant. State variables such as thickness, temperature, and plate shape at the exit of the previous rolling pass serve as input variables for the next pass, and deviations are transmitted and accumulated step-by-step along the pass sequence, ultimately affecting the quality of the finished product.

[0005] Chinese patent application No. 201911065069.1 discloses a method and system for diagnosing wedge defects in hot-rolled strip steel based on gray comprehensive correlation. The method involves acquiring process parameter curves related to the wedge shape along the length of the hot-rolled strip steel production process and the wedge curve at the finishing mill exit, and preprocessing these curves. The preprocessed wedge curve at the finishing mill exit is used as a characteristic behavior sequence, and the process parameter curves are used as factor behavior sequences, undergoing dimensionless and positive / negative transformation processing. The Dunk's correlation coefficient and gray slope correlation coefficient between the processed characteristic behavior sequences and factor behavior sequences are calculated, and the calculation results are converted into Dunk's correlation degree and gray slope correlation degree, respectively. The converted Dunk's correlation degree and gray slope correlation degree are weighted and summed to obtain the gray comprehensive correlation degree, which is then compared with a set correlation degree threshold to obtain the diagnostic result. While this solution can quickly and automatically identify the causes of hot-rolled wedge defects, it is mainly geared towards hot continuous rolling mill finishing mills. Its core is to calculate the overall correlation between process parameters and the wedge shape at the finishing mill exit. It fails to break down and consider the contribution of each pass or stand to the wedge defect throughout the entire process, and cannot reflect the contribution rate of multiple factors within a certain pass or stand to the wedge shape at the final stand exit of the finishing mill. Consequently, it cannot clearly pinpoint the cause of the wedge shape.

[0006] Chinese patent application No. 202210613083.6 discloses a big data-driven method for backtracking anomalies in thickness control performance during rolling processes. Based on the similarity of reduction distribution, benchmark data is selected from historical production data of the same number of passes and plate types. Based on the selected benchmark data, the angle contribution value method is used to analyze passes with performance degradation, identifying a candidate set of abnormal loops. Based on the time series data of multiple candidate abnormal loops, the transfer entropy between loops is analyzed, and the transfer entropy value between each abnormal loop is calculated. A causal relationship diagram is established based on the calculated transfer entropy values, and the root cause of the anomaly is located based on the causal relationship diagram. This scheme and technology can be applied to the finishing rolling process of thick plates, but it relies on benchmark data selection and transfer entropy analysis, resulting in high computational complexity. Furthermore, it mainly focuses on the analysis of anomaly relationships between control system loops such as AGC, lacking direct modeling of the quality inheritance relationship between passes.

[0007] Chinese patent application No. 202510993301.7 discloses a method for tracing defects in hot-rolled steel coils based on continuously cast billet slices. This method involves collecting data from the continuous casting and hot rolling production process for online data governance; using a digital spatial twin of the online continuous casting process, employing a finite element method to slice the billet, recording slice number, length, width, thickness, and quality; mapping the continuous casting billet slices to hot-rolled steel coil slices across processes using a volumetric method, slicing the steel coils, and recording the defect data of each slice; and establishing spatiotemporal alignment between cross-process product production data and defect data for billet quality retrospective detection and prediction of localized steel coil defects. This scheme establishes a mapping relationship between continuous casting billet slice data and hot-rolled coil surface defect data, thereby supporting the prediction of hot-rolled steel coil defects based on continuous casting billet slice data and guiding improvements in continuous casting billet quality. However, its tracing granularity is at the process level, unable to pinpoint specific passes within the hot rolling process. For medium and heavy plate rolling, even if the problem is traced back to "a problem in the hot rolling process", it is impossible to further determine whether it is in the roughing or finishing stage, or which specific process parameters in which pass caused the quality problem.

[0008] Given the shortcomings of existing medium and heavy plate quality traceability technologies, such as excessively coarse traceability granularity, poor process adaptability, lack of quantitative genetic mechanisms, limited coverage, and insufficient online practicality, there is an urgent need for a medium and heavy plate quality traceability solution that can overcome these deficiencies. Summary of the Invention

[0009] In view of the above-mentioned problems in existing medium and heavy plate quality traceability technologies, the purpose of this invention is to provide a medium and heavy plate quality traceability method and system based on the pass quality inheritance coefficient. Targeting the reversible multi-pass rolling characteristics of medium and heavy plates, this invention establishes a quantitative model of quality inheritance between medium and heavy plate rolling passes, advancing the traceability granularity from process and stand to pass and parameter, and realizing accurate, efficient, and visualized traceability of quality anomalies from finished product to specific pass and key parameters.

[0010] On one hand, the present invention provides a method for quality traceability of medium and heavy plates based on the genetic coefficient of pass quality, comprising: Collect process data for each pass of roughing and finishing rolling in a medium-thick plate rolling mill, as well as quality inspection data for finished steel plates, and preprocess the process data and the quality inspection data. Based on the temporal relationship and workpiece flow direction of reversible rolling of medium and heavy plates, a quality inheritance relationship matrix between passes is constructed using pre-processed process data and quality inspection data. The quality inheritance relationship matrix between passes uses the pass number as the row index and the quality index type as the column index. The matrix elements represent the inheritance influence relationship of each pass process parameter on each quality index, and the rolling direction information of each pass is marked in the quality inheritance relationship matrix between passes. A preset time-series correlation quantization method is adopted, using the time-series data sequence of abnormal quality indicators as the reference sequence and the time-series data sequence of process parameters of each historical pass as the comparison sequence, to calculate the inheritance coefficient of each pass process parameter on the abnormal quality indicators; and based on the inheritance coefficient, the constructed quality inheritance relationship matrix between passes is quantized to obtain the quantized inheritance coefficient matrix. When a quality abnormality is detected in the finished steel plate, based on the quantitative genetic coefficient matrix, the process traces backward from the last pass of the finishing mill to the first pass of the roughing mill. For the corresponding abnormal quality index, the genetic contribution rate of each process parameter to the abnormal quality index is calculated for each pass. The process is then sorted according to the size of the contribution rate to locate the root source pass with the largest contribution rate and the corresponding key process parameter. Based on the root cause pass with the largest contribution rate and the corresponding key process parameters, an evolution path diagram of quality anomalies along the pass is generated, and the key contribution parameters and heredity coefficient values ​​of each pass are marked on the evolution path diagram.

[0011] This method overcomes the limitations of traditional medium-thick plate quality traceability, which relies solely on single data statistics, cannot adapt to reversible rolling time sequence characteristics, and struggles to quantify the genetic influence of process passes. It constructs a quality genetic relationship matrix between passes with rolling direction annotations based on the time sequence relationship of reversible rolling in roughing and finishing of medium-thick plates and the flow law of the rolled piece. This accurately characterizes the implicit genetic association between multi-pass process parameters and quality indicators. Simultaneously, it introduces a time sequence association quantification method compatible with multiple algorithms, achieving precise quantification of the influence of pass process parameters on abnormal quality indicators, resulting in a standardized quantitative genetic coefficient matrix. This abandons the crude mode of traditional qualitative traceability and experience-based judgment. Based on this, a reverse traceability mechanism is established from the last pass of finishing to the first pass of roughing, accurately locating the root pass and key process parameters of quality anomalies, and visually outputting the evolution path of quality anomaly passes. This achieves intelligent source tracing of medium-thick plate rolling quality anomalies in a time-series, quantitative, and full-chain traceable manner, solving the technical defects of traditional traceability methods such as ambiguous causal relationships, inaccurate root cause location, and inability to reflect the cumulative genetic transmission of quality.

[0012] On the other hand, the present invention also provides a quality traceability system for medium and heavy plates based on the pass quality genetic coefficient, comprising: The data acquisition and preprocessing unit is used to acquire process data of each pass of roughing and finishing rolling in the medium and heavy plate rolling mill, as well as quality inspection data of finished steel plates, and to preprocess the process data and the quality inspection data. The genetic relationship matrix construction unit is used to construct an inter-pass quality genetic relationship matrix based on the temporal relationship and workpiece flow direction of reversible rolling of medium and heavy plates, using pre-processed process data and quality inspection data. The inter-pass quality genetic relationship matrix uses the pass number as the row index and the quality index type as the column index. The matrix elements represent the genetic influence relationship of each pass process parameter on each quality index, and the rolling direction information of each pass is marked in the inter-pass quality genetic relationship matrix. The genetic relationship matrix quantization unit is used to calculate the genetic coefficient of each process parameter on the abnormal quality index using a preset time-series correlation quantization method, with the time-series data sequence of abnormal quality index as the reference sequence and the time-series data sequence of process parameters of each historical pass as the comparison sequence; and to quantify the constructed quality genetic relationship matrix between passes based on the genetic coefficient to obtain the quantized genetic coefficient matrix. The reverse tracing unit is used to trace back from the last pass of finishing rolling to the first pass of roughing rolling when a quality abnormality is detected in the finished steel plate, based on the quantified genetic coefficient matrix. For the corresponding abnormal quality index, the genetic contribution rate of each process parameter to the abnormal quality index is calculated for each pass, and the pass is located according to the size of the contribution rate to pinpoint the root source pass with the largest contribution rate and the corresponding key process parameter. The visualization unit is used to generate an evolution path diagram of quality anomalies along the pass based on the root cause pass with the largest contribution rate and the corresponding key process parameters, and to mark the key contribution parameters and genetic coefficient values ​​of each pass on the evolution path diagram.

[0013] Compared with existing technologies, the medium-thick plate quality traceability method and system based on the pass quality genetic coefficient of the present invention, by constructing a quality genetic relationship matrix between passes and using grey relational analysis to quantify genetic contributions, refines the medium-thick plate quality traceability from the process level to the pass and parameter level, adapts to the characteristics of reversible rolling processes, and can simultaneously trace multiple types of quality anomalies such as thickness, crown, wedge shape, and plate-to-plate differences. It is computationally efficient and can be applied online, and can intuitively display the evolution path of quality anomalies and quantify the contribution rate of each pass. It provides reliable support for the optimization of medium-thick plate rolling processes, precise parameter adjustment, and stable quality control, and effectively improves the accuracy of quality traceability, on-site practicality, and process adaptability.

[0014] To achieve the foregoing and related objectives, one or more aspects of the invention include the features which will be described in detail below and specifically pointed out in the claims. The following description and accompanying drawings illustrate certain exemplary aspects of the invention. However, these aspects indicate only a few of the various ways in which the principles of the invention can be used. Furthermore, the invention is intended to include all such aspects and their equivalents. Attached Figure Description

[0015] Other objects and results of the invention will become more apparent and readily understood with reference to the following description taken in conjunction with the accompanying drawings and the contents of the claims, and with a more complete understanding of the invention. In the drawings: Figure 1 This is a schematic diagram of the process for a medium-thick plate quality traceability method based on the pass quality genetic coefficient according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the genetic evolution and root cause tracing of thickness deviation in medium-thick plate rolling according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of a medium-thick plate quality traceability system based on the pass quality genetic coefficient according to an embodiment of the present invention. Detailed Implementation

[0016] In the following description, numerous specific details are set forth for illustrative purposes and to provide a thorough understanding of one or more embodiments. However, it will be apparent that these embodiments may also be implemented without these specific details. In other instances, well-known structures and devices are shown in block diagram form for ease of description of one or more embodiments.

[0017] To better illustrate the technical solution of this invention, the following explanations are provided for some abbreviations and key terms involved in this invention: Medium and heavy plates: generally refers to steel plates with a thickness of ≥4mm, which are produced by reversible rolling in multiple passes and are used in shipbuilding, bridges, pressure vessels and other fields.

[0018] Rolling passes: The process of a plate being rolled down once on a rolling mill. A medium-thick plate typically requires 5–15 passes of reciprocating rolling.

[0019] Reversible rolling: The rolling mill alternates between forward and reverse rolling, and the steel plate passes back and forth through the rolls. It is commonly used in both roughing and finishing rolling stages.

[0020] Quality inheritance: The phenomenon in which process deviations or quality conditions in previous processes are transmitted, accumulated, and affect the quality of subsequent processes and the finished product.

[0021] Quality heritability coefficient: A numerical value that quantifies the influence of preceding process parameters on subsequent quality indicators; the larger the value, the stronger the heritability effect.

[0022] Quality inheritance matrix between passes: A two-dimensional matrix with passes as rows and quality indicators as columns, used to characterize the genetic association between process parameters of each pass and quality indicators.

[0023] Reference sequence: The time / space sequence of abnormal quality indicators (such as the distribution of finished product thickness deviation along the length) is used as a traceability benchmark.

[0024] Comparison Series: The time / space sequence of process parameters for each historical pass (such as the rolling force sequence of a certain pass), used for correlation analysis with the reference series.

[0025] Grey relational analysis: a method based on grey system theory, used to analyze the degree of correlation between sequences. This scheme is used to calculate the heritability coefficient of trace parameters on quality.

[0026] Deng's correlation coefficient: a basic indicator of grey relational analysis, which measures the degree of similarity in numerical magnitude between sequences.

[0027] Grey slope correlation coefficient: measures the similarity of sequences in terms of trend, and is suitable for trend-based indicator analysis such as plate shape.

[0028] Grey Comprehensive Relationship: A weighted fusion of Deng's correlation and grey slope correlation, taking into account both numerical and trend matching, used for complex indicators such as plate convexity and wedge shape.

[0029] Genetic contribution rate: The proportion of single-path / single-parameter genetic coefficient to the total genetic coefficient, used for ordination and location of key root causes.

[0030] Traceability at the pass level: This is the granularity of tracing quality issues down to the specific rolling pass (rather than just the roughing / finishing stages).

[0031] Evolutionary path diagram: A visual chart showing the accumulation and transmission of quality deviations from the first to the last pass, with key contribution parameters and heritability coefficients marked.

[0032] To address the problems of excessively coarse traceability granularity, poor process adaptability, lack of quantitative genetic mechanisms, and limited coverage in existing traceability systems, this invention proposes a quality traceability method and system for medium and heavy plates based on pass quality genetic coefficients. By establishing a quantitative model of quality inheritance between medium and heavy plate rolling passes, the traceability granularity is advanced from process and stand to pass and parameter, achieving accurate, efficient, and visualized traceability of quality anomalies from finished product to specific pass and key parameters. The specific embodiments of this invention will be described in detail below with reference to the accompanying drawings.

[0033] Figure 1 The flowchart of the medium-thick plate quality traceability method based on the pass quality genetic coefficient of the present invention is illustrated by way of example.

[0034] like Figure 1 As shown in the figure, the medium-thick plate quality traceability method based on the pass quality inheritance coefficient provided in this embodiment mainly includes the following steps: S110: Collect process data for each pass of roughing and finishing rolling in the medium and heavy plate rolling mill, as well as quality inspection data of the finished steel plates, and preprocess the process data and the quality inspection data.

[0035] This step is used for data acquisition and preprocessing. For each pass of the roughing and finishing rolling mills in medium and heavy plate mills, process data and post-rolled steel plate quality data are collected for each pass. Specifically, as an example, process data includes, but is not limited to: rolling force, roll gap, rolling speed, inlet temperature, outlet temperature, inlet thickness, outlet thickness, reduction rate, bending roll force, and roll gap inclination value. Quality data includes, but is not limited to: finished product thickness profile, shape profile, and steel plate crown. In addition, rolling identification data, such as steel plate grade, steel type, slab size, and finished product size, also needs to be collected.

[0036] Since the rows (passes), columns (quality indicators), and elements (genetic relationships) in the matrix all come from the collected process data and finished product quality data, and the unprocessed data is raw, messy, missing, abnormal, and out of sequence, it is impossible to directly map the correspondence between passes and quality indicators, and the matrix cannot be constructed at all. Therefore, it is necessary to preprocess the collected process data and quality inspection data to perform format alignment, time sequence alignment, and direction alignment for matrix construction. Among them, time sequence alignment sorts the data of each pass according to the rolling sequence to match the flow sequence of the rolled piece and ensure the correct row order of the matrix; direction alignment unifies the direction of forward and reverse pass data for reversible rolling to avoid matrix relationship misalignment; and dimension alignment unifies the number of sampling points and data format so that subsequent gray relational calculations (filling matrix elements) can be performed normally.

[0037] Only by collecting and properly processing the process and quality data for each pass can we establish a mapping framework between passes and quality indicators, thereby constructing a quantifiable matrix of quality inheritance relationships between passes.

[0038] The preprocessing in this invention may include at least one of missing value filling, outlier removal, timing alignment, rolling direction alignment, and dimensionless processing.

[0039] Specifically, missing value imputation is used to complete missing data caused by equipment sampling fluctuations and signal interruptions during the rolling process, ensuring the integrity of the data sequence; outlier removal is used to filter out abnormal noise data caused by equipment failure and external interference, avoiding extreme data interference with the accuracy of correlation quantization; time sequence alignment is used to strictly sort the process data and quality inspection data of each pass according to the rolling sequence, accurately matching the actual flow time sequence of the rolled piece, ensuring the accuracy of the row index order of the quality genetic relationship matrix between subsequent passes; rolling direction alignment is used to unify the data arrangement rules of different rolling directions for the alternating forward and reverse rolling characteristics of reversible rolling of medium and heavy plates, eliminating the problem of disordered data arrangement of forward and reverse passes, and avoiding the misalignment and failure of the genetic correlation relationship of passes; dimensionless processing is used to unify the data dimension of different types and magnitudes of process parameters and quality indicators, eliminate the difference in physical dimensions, adapt to the operation requirements of various subsequent time sequence correlation quantization algorithms, and ensure the accuracy and effectiveness of the genetic coefficient calculation results.

[0040] S120: Based on the temporal relationship and workpiece flow direction of reversible rolling of medium-thick plates, a quality inheritance matrix between passes is constructed using pre-processed process data and quality inspection data.

[0041] This step is used to construct an inter-pass quality inheritance matrix, serving as a framework for mapping the quality indicators of each pass to the process parameters of preceding and following passes. The inter-pass quality inheritance matrix uses the pass number as the row index and the quality indicator type as the column index. Matrix elements represent the inheritance relationship between each process parameter of that pass and each quality indicator. Due to the reversible rolling of medium-thick plates, the inheritance matrix also needs to indicate the rolling direction of each pass to account for the impact of direction changes on genetic transmission during tracing.

[0042] S130: Using a preset time-series correlation quantization method, the time-series data sequence of abnormal quality indicators is used as a reference sequence, and the time-series data sequence of process parameters of each historical pass is used as a comparison sequence to calculate the genetic coefficient of each pass process parameter on the abnormal quality indicators; and the quality genetic relationship matrix between passes is quantized based on the genetic coefficient to obtain the quantized genetic coefficient matrix.

[0043] This step is used to calculate and optimize the genetic coefficient. The preset time-series correlation quantification method can be at least one of the following: grey relational analysis, Pearson linear correlation analysis, Spearman rank correlation analysis, partial least squares regression analysis, and transfer entropy analysis.

[0044] This invention supports the combined use of multiple time-series correlation quantification methods. For example, two or more algorithms can be used in parallel to calculate multiple sets of genetic coefficients. These coefficients can then be weighted, fused, averaged, or filtered by confidence level to obtain a fused comprehensive genetic coefficient, further improving the accuracy of root cause localization. Alternatively, different algorithms can be adaptively switched based on the characteristics of the pass data to achieve segmented hybrid calculation. This hybrid calculation mode still relies on a unified calculation framework of "quality anomaly index sequence as reference sequence and pass process parameter sequence as comparison sequence".

[0045] Specifically, as an example, the grey relational analysis method is used to calculate the genetic influence coefficient between the current abnormal indicators and the process parameters of each historical pass, including: S131: Using the time series data sequence of abnormal quality indicators as a reference sequence and the time series data sequence of process parameters of each historical pass as a comparison sequence, the reference sequence and the comparison sequence are respectively processed to be dimensionless to obtain the standardized reference sequence and the standardized comparison sequence.

[0046] For example, using the data sequence of quality anomaly indicators as a reference sequence.

[0047] For example, finished product thickness deviation The values ​​at each sampling point along the length direction are used as a comparison sequence for the process parameter data of each historical pass.

[0048] Where n represents the total number of valid sampling points in a single time series data sequence. Specifically, it refers to the total number of valid data sampling points retained after preprocessing in a single-pass process parameter time series and a single abnormal quality index time series. It is used to unify the data length of all sequences, ensure the consistency of the dimension of point-by-point difference calculation, and is the basic dimension parameter for time series correlation quantization calculation.

[0049] i is a combination identifier of the pass number and process parameter, which can also be understood as the process parameter sequence index number, i=1,2,3...m (m is the total number of process parameter categories to be calculated). Specifically, it refers to the process parameter comparison sequence used to distinguish different historical rolling passes and different types of process parameters. Each i corresponds to a set of independent pass process parameter time series data, realizing the one-to-one association calculation of multi-parameter, multi-pass process sequences and abnormal quality index sequences.

[0050] Because rolling process parameters and finished product quality indicators have significant differences in physical units and numerical magnitudes (rolling force in kN, temperature in °C, thickness in mm, and thickness deviation only a few tenths of a millimeter), directly using the original numerical values ​​for calculation will lead to deviations in the correlation coefficient results. Therefore, before performing grey relational coefficient calculations, it is necessary to perform dimensionless conversion on the reference series and all comparison series. Dimensionless conversion can eliminate dimensional interference, remove the influence of physical units, and provide a basis for horizontal comparison of different types of parameters such as temperature, rolling force, and thickness; it smooths out differences in numerical magnitudes, preventing large numerical indicators (such as rolling forces of thousands of Newtons) from naturally suppressing small numerical indicators (micrometer-level plate shape deviations), thus avoiding distortion of the correlation coefficient calculation results; and it unifies the data value range, mapping all data to similar numerical intervals, ensuring the fairness and effectiveness of the grey relational coefficient and genetic coefficient calculation results.

[0051] In one specific embodiment of the present invention, the data sequences are processed in a dimensionless manner by initialization or averaging, and each time series data sequence is mapped to a unified numerical range to eliminate the calculation error caused by the dimensions and magnitude, and to ensure that the correlation between each parameter and the quality anomaly index is calculated fairly and reliably.

[0052] S132: Calculate the absolute difference point by point based on the standardized reference sequence and the standardized comparison sequence, and solve for the two-level minimum difference and the two-level maximum difference based on all the absolute differences.

[0053] S133: Based on the minimum difference and maximum difference between the two levels, and combined with the preset resolution coefficient, calculate the correlation coefficient corresponding to each sampling point.

[0054] Calculate the correlation coefficient: First, calculate the minimum difference and the maximum difference between the two levels. , That is, the minimum and maximum values ​​of the absolute differences among all time-series sampling points and all process parameter sequences:

[0055] Where k is the index number of the time-series sampling point, k=1,2,3...n. Specifically, it refers to the position number of each sampling point in a single time-series data sequence, used to traverse all data points of the time-series sequence, calculating the absolute difference between the standardized reference sequence and the standardized comparison sequence point by point. It is the core traversal parameter for solving the correlation coefficient. Correspondingly, This is the original data of the kth time-series sampling point, which is a reference series of abnormal quality indicators. This is the raw data of the k-th time-series sampling point for the i-th set of process parameter comparison sequences.

[0056] Then calculate the reference sequence. Comparison of sequences Grey relational coefficient corresponding to the kth sampling point :

[0057] in, The resolution coefficient takes values ​​in the range (0,1], and is usually set to... It is used to adjust the distribution range of the correlation coefficient and improve the data discrimination.

[0058] S134: The mean value of the correlation coefficients of all sampling points is used to obtain the Deng's grey correlation degree, and the Deng's grey correlation degree is used as the genetic coefficient of the corresponding process parameter.

[0059] Comparison Sequence The correlation coefficients at each sampling point are averaged to obtain the Deng's grey correlation degree with the reference series. :

[0060] This correlation This refers to the heritability coefficient of the process parameters for the corresponding pass (group i) on the quality index. Finally, the calculated correlation values ​​are... As genetic coefficients, they are filled into the constructed inter-track quality genetic relationship matrix to form a quantified genetic coefficient matrix.

[0061] Furthermore, when the abnormal quality index is plate convexity deviation or wedge deviation, the genetic coefficient is calculated comprehensively by combining the gray slope correlation degree. That is, for the calculation of the genetic coefficient of plate shape indices such as convexity in medium-thick plate rolling, the gray comprehensive correlation degree method can be further used to calculate the Dönbach correlation degree separately. Correlation with gray slope Then, the overall correlation degree is obtained by weighted summation. in, The weighting coefficient is used to weight and fuse Deng's gray relational degree and gray slope relational degree, adapt to the calculation of plate convexity and wedge deviation index, and reflect the hereditary effect of the entry convexity of the rolled piece on the exit convexity.

[0062] S140: When a quality abnormality is detected in the finished steel plate, based on the quantified genetic coefficient matrix, trace back from the last pass of finishing rolling to the first pass of roughing rolling, calculate the genetic contribution rate of each process parameter to the abnormal quality index for each pass, sort them according to the size of the contribution rate, and locate the root source pass with the largest contribution rate and the corresponding key process parameter.

[0063] This step is used to achieve root cause localization at the pass / end level. When an abnormality in the quality of the finished steel plate is detected, such as excessive thickness, excessive convexity, excessive wedge shape, or excessive difference between plates, the genetic coefficient matrix is ​​traced backward to calculate the genetic contribution rate of each process parameter to the abnormal index for each pass / end. The parameters are then sorted from high to low contribution rate to locate the pass / end with the largest contribution rate and the corresponding key process parameter.

[0064] S150: Based on the root cause pass with the largest contribution rate and the corresponding key process parameters, generate an evolution path diagram of quality anomalies along the pass, and mark the key contribution parameters and genetic coefficient values ​​of each pass on the evolution path diagram.

[0065] This step is a visualization and traceability display step. With the track number as the vertical axis and the quality index deviation value as the horizontal axis, an evolution path diagram of quality anomalies along the track is output, which intuitively shows the cumulative transmission process of each track, and marks the key contribution parameters and heredity coefficient values ​​of each track on the evolution path diagram.

[0066] Furthermore, this invention can also support simultaneous tracing and coupled anomaly analysis of multiple quality indicators. Based on the time series sequence of a single abnormal quality indicator, it can simultaneously collect time series deviation data of multiple quality indicators such as thickness deviation, convexity deviation, wedge deviation, and same-plate difference as multiple reference sequences as needed. , ,..., Construct a three-dimensional genetic coefficient matrix of pass number × process parameters × quality index. Where N is the total number of passes in the rolling of medium and heavy plates, P is the total number of process parameters involved in the modeling, the superscript number corresponds to different quality index types, and Q is the total number of quality index types involved in this source analysis. By comparing the genetic contribution of the same pass to different quality indicators, the common root pass of multi-indicator coupling anomalies is identified, and a multi-quality index joint evolution path diagram is generated.

[0067] Specifically, for each set of quality indicator reference sequences, time-series comparison sequences for each rolling pass and various process parameters are matched. The time-series correlation quantification method is used to calculate the corresponding genetic coefficients for each set, ultimately constructing... Each element of the three-dimensional genetic coefficient matrix uniquely corresponds to a specific pass, process parameter, and the degree of quantitative genetic influence on a specific quality indicator, enabling structured and quantitative storage of the coupled relationships among multiple passes, parameters, and quality indicators.

[0068] Based on the constructed three-dimensional genetic coefficient matrix, root cause analysis of coupled anomalies in multiple quality indicators can be achieved. For coupled anomalies in finished steel plates where two or more types of quality deviations occur simultaneously, by comparing the genetic coefficients and genetic contribution rates of different quality indicators under the same rolling pass and process parameters, common process parameters and common passes that contribute significantly to multiple anomalous quality indicators are identified. This accurately identifies the common root cause passes and core influencing parameters of coupled anomalies in multiple indicators, avoiding the problems of scattered root causes and unidentifiable coupling mechanisms caused by independent tracing of single indicators. Furthermore, based on the quantitative results of the genetic contributions of multiple indicators, a joint evolution path diagram of multiple quality indicators is generated, simultaneously displaying the cumulative transmission law, evolution trend, and parameter contribution differences of different quality indicator deviations in each rolling pass. This intuitively presents the entire process of the generation, accumulation, and transmission of coupled anomalies in multiple quality indicators, enabling global tracing and mechanism visualization analysis of multi-dimensional quality anomalies in medium and heavy plates.

[0069] Furthermore, in another embodiment of the present invention, the above method further includes a classification and management step of the genetic coefficient matrix to achieve differentiated modeling and precise adaptation of quality inheritance laws under different rolling conditions and different steel plate categories. Specifically, based on the actual production characteristics of medium and heavy plates, the production conditions are finely classified according to steel grade and finished product thickness range. Medium and heavy plates of different steel grades and thickness ranges have significant differences in rolling deformation resistance, temperature drop characteristics, pass process matching laws, and quality inheritance mechanisms, making a unified genetic coefficient matrix model unsuitable. Therefore, the present invention classifies production samples according to preset steel grade classification standards and thickness range thresholds. For each sub-category, data collection and preprocessing, sequence construction, construction of quality inheritance relationship matrix between passes, genetic coefficient quantification, and quantification of quality inheritance relationship matrix between passes are completed independently. Dedicated quantitative genetic coefficient matrices for each category are constructed in batches and summarized to form a multi-category hierarchical genetic coefficient matrix library.

[0070] When performing quality anomaly tracing tasks, the steel grade attributes and thickness specification range of the finished steel plate to be traced are first identified. Based on the category matching rules, a dedicated quantitative genetic coefficient matrix that completely corresponds to its category is retrieved from the preset genetic coefficient matrix library. Relying on this category matching matrix, subsequent genetic contribution rate calculation, root cause reverse tracing, and evolution path generation are completed. This avoids the problem of reduced traceability accuracy caused by cross-category model mismatch and greatly improves the adaptability and accuracy of quality traceability under different working conditions.

[0071] Furthermore, the classification and management steps of the genetic coefficient matrix can also include a periodic online update process for the genetic coefficient matrix based on a rolling window mechanism (such as a fixed-length rolling window or a preset-length rolling window). This enables the matrix model to be adaptively updated as production data iterates, ensuring that the genetic laws of quality are consistent with the real-time production status. A specific update logic example is as follows: a fixed sliding window size is preset, which represents the number of valid, latest production samples used for modeling. This size limits the effective data capacity for genetic coefficient matrix modeling and discards outdated historical production data.

[0072] During production, for each new qualified or defective finished steel plate of the current product category produced, the process data and quality inspection data for the entire series of passes corresponding to that steel plate are immediately collected and included in the current rolling window sample set. Simultaneously, the earliest historical steel plate sample data in the window is automatically removed, ensuring that the total number of effective samples in the rolling window remains constant. This enables the temporal iterative update of the sample pool.

[0073] After updating the rolling window samples, based on all the latest valid samples within the window, sequence modeling, dimensionless processing, and time-series correlation quantification calculations are re-executed. This updates the genetic coefficient values ​​of each process parameter for each quality indicator, and statistically solves for the mean distribution, numerical fluctuation range, and distribution characteristics of each genetic coefficient. Simultaneously, the matrix data and parameter distribution model in the corresponding product category's genetic coefficient matrix library are updated. This online iterative update via the rolling window allows the genetic coefficient matrix to continuously adapt to actual scenarios such as equipment state drift, process fine-tuning, and slow changes in production conditions, ensuring the real-time effectiveness of quality genetic quantification relationships and improving the stability and accuracy of quality traceability during long-term production.

[0074] The following will combine a more specific embodiment and Figure 2 The genetic evolution and root cause tracing diagram of the thickness deviation in medium and heavy plate rolling shown above provides a more detailed explanation of the medium and heavy plate quality tracing method based on the pass quality genetic coefficient of the present invention.

[0075] Taking a typical single-stand four-high reversible rolling mill production line for medium and heavy plates as an example, the billet, after exiting the heating furnace, undergoes multiple reversible rolling passes through the roughing and finishing stages to finally obtain the finished steel plate. The production line is equipped with a primary basic automation system and a secondary process control system, which can collect and store process data and finished product quality data for each pass in real time.

[0076] 1. Data Acquisition and Processing

[0077] The following data were collected for each rolled steel plate: 1) Material identification data: steel plate number, steel type, slab size, target finished product size, production time, total number of passes, number of roughing passes, number of finishing passes, and transfer identification.

[0078] 2) Process data for each pass: pass number, rolling stage, rolling direction, rolling force, roll gap, rolling speed, inlet temperature, outlet temperature, inlet thickness, outlet thickness, reduction rate, bending roll force, and rolling force deviation (drive side and operating side).

[0079] 3) Finished product inspection data: thickness curve, strip shape curve, final rolling temperature, width curve, and planar shape data.

[0080] 4) Data Preprocessing: The collected data is spliced ​​and outlier handling is performed. Material identification, pass / fail data, and finished product inspection data are spliced ​​using the steel plate number as the unique classification link standard. This ensures that each data entry has one and only one steel plate number. The pass / fail data is changed from multiple rows to a single row, and different passes / fail are transposed to different column attributes.

[0081] 2. Construction of the genetic coefficient matrix

[0082] Based on the thickness deviation of the finished product ( This serves as the core quality indicator. The steel plate is divided into n sampling points along its length, forming a reference sequence:

[0083] For each historical pass data, the data sequence of its key process parameters is used as a comparison series: First pass rolling force sequence: ; First-pass exit thickness sequence: ; Second rolling force sequence: ; Second exit thickness sequence: ; ...... Calculate each comparison sequence according to the method described in step S130. With reference sequence Grey relational degree The higher the correlation value, the greater the hereditary influence of the corresponding process parameters of that pass on the thickness deviation of the finished product.

[0084] 3. Root cause localization and tracing

[0085] When an abnormal thickness of the finished steel plate is detected, root cause localization should be performed according to the following steps: 1) Calculate the correlation between all historical process parameters and finished product thickness deviation, and sort them from largest to smallest correlation. 2) Select the top N parameters with the highest correlation as the candidate root cause set; 3) Combining the genetic coefficient matrix, trace backwards along the pass direction, from the last pass of finishing rolling to the first pass of roughing rolling, to determine the root pass with the highest genetic contribution rate and its corresponding rolling stage (finishing or roughing). Genetic contribution rate of a certain process parameter. It can be defined as the normalized value of its correlation degree:

[0086] in, Let be the heredity coefficient corresponding to the i-th comparison sequence, and m be the total number of comparison sequences.

[0087] 4) Output diagnostic results, for example: the correlation degree of rolling force deviation in the 5th pass of roughing is 0.85, with a contribution rate of 21% → the correlation degree of exit thickness deviation in the last pass of roughing is 0.78, with a contribution rate of 19% → the correlation degree of exit thickness deviation in the 3rd pass of finishing is 0.72, with a contribution rate of 18% → the finished product thickness is out of tolerance, and the key contributing pass is the 5th pass of roughing.

[0088] 4. Visual output

[0089] Generate as Figure 2 The diagram shows the evolution path of quality anomalies along the rolling passes: the horizontal axis represents the pass number (from the first pass of roughing to the last pass of finishing, R represents roughing and F represents finishing), and the vertical axis represents the thickness deviation value (mm). The cumulative transmission process of thickness deviation in each pass is displayed as a line graph, with key contribution parameters and their correlation values ​​marked on the graph. The roughing and finishing stages are distinguished by dashed lines in the evolution path diagram.

[0090] Compared with the prior art, the present invention has the following beneficial effects: 1. Significantly improved traceability accuracy: The smallest analytical unit for tracing the quality of medium and heavy plates is refined from the entire steel plate to the pass level, which can accurately pinpoint which specific pass or process parameter in the roughing or finishing rolling stage is the main contributor to quality anomalies.

[0091] 2. Adapted to the characteristics of medium and heavy plate processes: A genetic relationship matrix structure was designed for the characteristics of reversible rolling, multiple passes, and flexible rolling procedures of medium and heavy plates, which has good adaptability to various scenarios.

[0092] 3. Coverage of multiple quality indicators: It can simultaneously trace the root causes of multiple quality abnormalities such as thickness deviation, plate convexity, wedge shape, flatness, and same plate difference. Compared with existing methods that only focus on a single indicator, it has a wider range of applications.

[0093] 4. Quantitative genetic influence analysis: Through the genetic coefficient matrix, the contribution of each process parameter to the finished product quality index can be intuitively quantified, providing accurate data support for the optimization of rolling schedule and adjustment of process parameters.

[0094] 5. Low computational complexity and good interpretability: The grey relational method is used to calculate the genetic coefficient. The calculation speed is fast, it can be run online, and it can output the quality anomaly evolution path diagram and genetic contribution rate ranking along the track, providing on-site engineers with an intuitive and easy-to-understand root cause diagnosis report.

[0095] Corresponding to the above-mentioned method for tracing the quality of medium and heavy plates based on the genetic coefficient of track quality, the present invention also provides a system for tracing the quality of medium and heavy plates based on the genetic coefficient of track quality. Figure 3 The structure of a medium-thick plate quality traceability system based on the pass quality genetic coefficient according to an embodiment of the present invention is shown, as follows: Figure 3 As shown, the medium-thick plate quality traceability system 300 based on the genetic coefficient of track quality in this embodiment mainly includes a data acquisition and preprocessing unit 310, a genetic relationship matrix construction unit 320, a genetic relationship matrix quantification unit 330, a reverse traceability unit 340, and a visualization display unit 350.

[0096] The unit of this invention can also be called a module, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0097] The functions of each module / unit are as follows: The data acquisition and preprocessing unit 310 is used to acquire process data of each pass of roughing and finishing rolling in the medium and heavy plate rolling mill, as well as quality inspection data of finished steel plates, and to preprocess the process data and the quality inspection data. The genetic relationship matrix construction unit 320 is used to construct an inter-pass quality genetic relationship matrix based on the temporal relationship and workpiece flow direction of reversible rolling of medium-thick plates, using pre-processed process data and quality inspection data. The inter-pass quality genetic relationship matrix uses the pass number as the row index and the quality index type as the column index. The matrix elements represent the genetic influence relationship of each pass process parameter on each quality index, and the rolling direction information of each pass is marked in the inter-pass quality genetic relationship matrix. The genetic relationship matrix quantization unit 330 is used to calculate the genetic coefficient of each process parameter on the abnormal quality index using a preset time-series correlation quantization method, with the time-series data sequence of abnormal quality index as the reference sequence and the time-series data sequence of process parameters of each historical pass as the comparison sequence; and to quantify the constructed quality genetic relationship matrix between passes based on the genetic coefficient to obtain the quantized genetic coefficient matrix. The reverse tracing unit 340 is used to trace back from the last pass of finishing rolling to the first pass of rough rolling according to the quantitative genetic coefficient matrix when a quality abnormality is detected in the finished steel plate. For the corresponding abnormal quality index, the genetic contribution rate of each process parameter to the abnormal quality index is calculated for each pass, and the source pass with the largest contribution rate and the corresponding key process parameter are located according to the contribution rate. The visualization unit 350 is used to generate an evolution path diagram of quality anomalies along the pass based on the root cause pass with the largest contribution rate and the corresponding key process parameters, and to mark the key contribution parameters and genetic coefficient values ​​of each pass on the evolution path diagram.

[0098] The above-mentioned medium and heavy plate quality traceability system based on the genetic coefficient of the pass quality is an implementation system corresponding to the above-mentioned medium and heavy plate quality traceability method based on the genetic coefficient of the pass quality. Its specific execution steps can be referred to the specific implementation of the above-mentioned medium and heavy plate quality traceability method based on the genetic coefficient of the pass quality, and will not be described in detail here.

[0099] The above describes a method and system for tracing the quality of medium and heavy plates based on the genetic coefficient of pass quality according to an embodiment of the present invention. It can be determined that by utilizing the above-described method and system for tracing the quality of medium and heavy plates based on the genetic coefficient of pass quality, the quality genetic relationship matrix between passes, adapted to the reversible rolling sequence characteristics of medium and heavy plates, is constructed. Combined with multiple time-series correlation quantification algorithms, the genetic coefficients of process parameters on abnormal quality indicators are accurately solved, achieving a quantitative characterization of the influence relationship on rolling quality. This solves the problems of qualitative judgment, causal ambiguity, and low accuracy in traditional experience-based tracing. By classifying and building a library of genetic coefficient matrices according to steel type and specifications and matching and calling them, different rolling conditions and material characteristics are effectively adapted, avoiding traceability deviations caused by cross-category model mismatches. Combined with a rolling window online iterative update mechanism, the model can adapt to changes in production status such as equipment drift and process fine-tuning, ensuring the real-time effectiveness of the genetic quantification relationship. Meanwhile, relying on a multi-index three-dimensional genetic matrix, multiple quality anomalies are coupled and traced. By calculating the contribution rate of each pass in reverse order, the root pass and key process parameters are accurately located, and the evolution path of quality anomalies is visualized and output. This significantly improves the accuracy, real-time performance, and versatility of tracing quality anomalies in medium and heavy plates, and provides reliable data support for rolling process optimization and steady-state quality control.

[0100] In the several embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0101] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0102] The method and system for tracing the quality of medium and heavy plates based on the genetic coefficient of pass quality according to the present invention have been described above by way of example with reference to the accompanying drawings. However, those skilled in the art should understand that various improvements can be made to the method and system for tracing the quality of medium and heavy plates based on the genetic coefficient of pass quality proposed in the present invention without departing from the scope of the present invention.

Claims

1. A method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality, characterized in that, include: Collect process data for each pass of roughing and finishing rolling in a medium-thick plate rolling mill, as well as quality inspection data for finished steel plates, and preprocess the process data and the quality inspection data. Based on the temporal relationship and workpiece flow direction of reversible rolling of medium and heavy plates, a quality inheritance relationship matrix between passes is constructed using pre-processed process data and quality inspection data. The quality inheritance relationship matrix between passes uses the pass number as the row index and the quality index type as the column index. The matrix elements represent the inheritance influence relationship of each pass process parameter on each quality index, and the rolling direction information of each pass is marked in the quality inheritance relationship matrix between passes. A preset time-series correlation quantization method is adopted, using the time-series data sequence of abnormal quality indicators as the reference sequence and the time-series data sequence of process parameters of each historical pass as the comparison sequence, to calculate the inheritance coefficient of each pass process parameter on the abnormal quality indicators; and based on the inheritance coefficient, the constructed quality inheritance relationship matrix between passes is quantized to obtain the quantized inheritance coefficient matrix. When a quality abnormality is detected in the finished steel plate, based on the quantitative genetic coefficient matrix, the process traces backward from the last pass of the finishing mill to the first pass of the roughing mill. For the corresponding abnormal quality index, the genetic contribution rate of each process parameter to the abnormal quality index is calculated for each pass. The process is then sorted according to the size of the contribution rate to locate the root source pass with the largest contribution rate and the corresponding key process parameter. Based on the root cause pass with the largest contribution rate and the corresponding key process parameters, an evolution path diagram of quality anomalies along the pass is generated, and the key contribution parameters and heredity coefficient values ​​of each pass are marked on the evolution path diagram.

2. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 1, characterized in that, The process data includes at least one of the following: rolling force, roll gap, rolling speed, rolling direction, inlet temperature, outlet temperature, inlet thickness, outlet thickness, reduction rate, roll bending force, rolling force difference, and roll gap inclination value. The quality inspection data includes at least one of the following: finished product thickness curve, plate shape curve, final rolling temperature, width deviation, and planar shape data.

3. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 2, characterized in that, The preprocessing includes at least one of missing value filling, outlier removal, time sequence alignment, rolling direction alignment, and dimensionless processing.

4. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 3, characterized in that, The preset time-series correlation quantification method includes at least one of the following: grey relational analysis, Pearson linear correlation analysis, Spearman rank correlation analysis, partial least squares regression analysis, and transfer entropy analysis.

5. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 4, characterized in that, The method for calculating the genetic coefficient using the grey relational analysis method includes: Using the time series data sequence of abnormal quality indicators as a reference sequence and the time series data sequence of process parameters of each historical pass as a comparison sequence, the reference sequence and the comparison sequence are respectively processed to be dimensionless, and the standardized reference sequence and the standardized comparison sequence are obtained accordingly. The absolute difference is calculated point by point based on the standardized reference sequence and the standardized comparison sequence, and the two-level minimum difference and the two-level maximum difference are solved based on all the absolute differences. Based on the two-level minimum difference and the two-level maximum difference, and combined with the preset resolution coefficient, the correlation coefficient corresponding to each sampling point is calculated; The mean value of the correlation coefficients of all sampling points is used to obtain the Deng's grey correlation degree, and the Deng's grey correlation degree is used as the genetic coefficient of the corresponding process parameters.

6. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 5, characterized in that, When the abnormal quality index is plate convexity deviation or wedge deviation, the genetic coefficient is calculated by combining the gray slope correlation degree. These include: The Deng's grey relational degree and the grey slope relational degree were calculated respectively. The gray correlation degree and the gray slope correlation degree are weighted and summed according to the preset weight coefficients to obtain the gray comprehensive correlation degree, which is then used as the genetic coefficient of the corresponding process parameter.

7. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 5, characterized in that, The dimensionless processing employs an initialization method or a mean-based method to uniformly convert the values ​​of all sampling points in the reference and comparison sequences, thereby eliminating differences in data dimensions and numerical magnitudes.

8. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 1, characterized in that, The quality abnormalities of the finished steel plates include at least one of the following: excessive thickness, excessive convexity, excessive wedge shape, and excessive difference between plates of the same type.

9. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 1, characterized in that, The evolution path diagram uses the track number as the vertical axis and the quality index deviation value as the horizontal axis to show the cumulative transmission process of quality anomalies in each track.

10. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 1, characterized in that, It also includes the genetic coefficient matrix classification and management steps; including: Medium and heavy plate production categories are divided based on steel grade and finished product thickness range. Data collection and preprocessing, construction of quality inheritance relationship matrix between passes, calculation of inheritance coefficients and quantification of quality inheritance relationship matrix between passes are carried out independently according to the categories, forming exclusive quantitative inheritance coefficient matrix for each category. The exclusive genetic coefficient matrices corresponding to each category are compiled to construct a hierarchical multi-category genetic coefficient matrix library. During the traceability operation, based on the steel type and finished product thickness specifications of the steel plate to be traced, the corresponding category's exclusive quantitative genetic coefficient matrix is ​​matched from the genetic coefficient matrix library to carry out reverse root cause tracing.

11. The method for quality traceability of medium-thick plates based on the genetic coefficient of pass quality according to claim 10, characterized in that, A fixed-length rolling window mechanism is used to periodically update the category-specific quantified genetic coefficient matrix in the genetic coefficient matrix library. Set the window sample capacity. For each new piece of valid production data of medium-thick plate of the same category, the valid production data is incorporated into the corresponding category sample set. At the same time, the earliest set of historical samples within the window period is removed. The genetic coefficient is recalculated using the updated sample set and the exclusive quantitative genetic coefficient matrix of the corresponding category is refreshed.

12. A quality traceability system for medium-thick plates based on the genetic coefficient of pass quality, characterized in that, include: The data acquisition and preprocessing unit is used to acquire process data of each pass of roughing and finishing rolling in the medium and heavy plate rolling mill, as well as quality inspection data of finished steel plates, and to preprocess the process data and the quality inspection data. The genetic relationship matrix construction unit is used to construct an inter-pass quality genetic relationship matrix based on the temporal relationship and workpiece flow direction of reversible rolling of medium and heavy plates, using pre-processed process data and quality inspection data. The inter-pass quality genetic relationship matrix uses the pass number as the row index and the quality index type as the column index. The matrix elements represent the genetic influence relationship of each pass process parameter on each quality index, and the rolling direction information of each pass is marked in the inter-pass quality genetic relationship matrix. The genetic relationship matrix quantization unit is used to calculate the genetic coefficient of each process parameter on the abnormal quality index using a preset time-series correlation quantization method, with the time-series data sequence of abnormal quality index as the reference sequence and the time-series data sequence of process parameters of each historical pass as the comparison sequence; and to quantify the constructed quality genetic relationship matrix between passes based on the genetic coefficient to obtain the quantized genetic coefficient matrix. The reverse tracing unit is used to trace back from the last pass of finishing rolling to the first pass of roughing rolling when a quality abnormality is detected in the finished steel plate, based on the quantified genetic coefficient matrix. For the corresponding abnormal quality index, the genetic contribution rate of each process parameter to the abnormal quality index is calculated for each pass, and the pass is located according to the size of the contribution rate to pinpoint the root source pass with the largest contribution rate and the corresponding key process parameter. The visualization unit is used to generate an evolution path diagram of quality anomalies along the pass based on the root cause pass with the largest contribution rate and the corresponding key process parameters, and to mark the key contribution parameters and genetic coefficient values ​​of each pass on the evolution path diagram.

Images