Metallurgical quality traceability method based on big data

Abstract

This invention relates to the field of supply chain management technology, specifically to a metallurgical quality traceability method based on big data. The method includes the following steps: extracting reverse segment indexes based on speed and concentration signal comparison; generating a path set by screening segments with consistent energy consumption and concentration; extracting high-risk furnace batches by combining power and hardness fluctuations; marking traceability entry points for critical performance samples; and establishing a multi-source data link mapping set. In this invention, a density distribution trend curve is formed by constructing a reverse matching relationship between the speed change trend and the direction of main alloy element concentration change. Path segment indexes are screened based on the consistency between concentration shift and energy consumption trend. High-risk batches are identified by comparing the time difference between smelting power fluctuations and ingot hardness distribution characteristics. A traceability entry list is formed using boundary crossing analysis of performance signals, achieving accurate identification and tracking mapping of key quality changes, and significantly improving the continuity and controllability of the quality response chain throughout the entire metallurgical process.

Description

Technical Field

[0001] This invention relates to the field of supply chain management technology, and in particular to a metallurgical quality traceability method based on big data. Background Technology

[0002] Supply chain management technology involves the coordination and optimization of the entire process from raw material procurement, product manufacturing, warehousing and logistics to product sales. Core aspects include material flow tracking, inventory control, procurement planning, production schedule management, and information transparency. It leverages information technology to improve supply chain collaboration efficiency and ensure information synchronization and optimal resource allocation across all stages. In practice, effective integration of data from each node in the supply chain is crucial to ensure product controllability and traceability from source to end, particularly in quality management and problem tracking. Traditional metallurgical quality traceability methods involve recording and tracing information such as raw material batches, smelting parameters, rolling data, and inspection records during the manufacturing process. This is achieved through manual input and decentralized storage of data at each stage using paper records, spreadsheets, or independent information systems. It primarily relies on operators entering data according to standard operating procedures and identifying and associating products based on static identifiers such as furnace number, batch number, and timestamps. This allows for the identification of relevant process data and responsible parties when quality issues arise. In traditional methods, key aspects include recording smelting process parameters such as feeding sequence, temperature control, and timing; registering deformation and temperature trajectories during the rolling process; and compiling test data on the chemical composition and mechanical properties of the final product. Most of this data is stored in the enterprise's local area system, lacking a unified platform for integration and dynamic data chain management capabilities, resulting in deficiencies in the completeness and consistency of traceability information.

[0003] In existing technologies, data recording relies on the standard operating procedures of operators for decentralized input and manual maintenance, lacking a dynamic mapping mechanism between data structures. This makes it impossible to effectively establish data correlation between rolling speed fluctuations and alloy composition changes. Especially when product performance abnormalities occur, static identifiers are difficult to reflect the changing trends of process data. After quality problems occur, the traceability path is unclear. For example, the potential correlation between smelting power fluctuations and sudden changes in finished product hardness is difficult to identify due to data fragmentation. At the same time, the phased data is distributed in various local systems for a long time, lacking unified temporal clue integration, which restricts the comprehensiveness of problem localization and the accuracy of traceability entry points. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a metallurgical quality traceability method based on big data.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a metallurgical quality traceability method based on big data, comprising the following steps:

[0006] S1: Based on the speed change signal sequence and the main alloy element concentration change signal sequence of the same rolling section, the time axis is divided and the direction is compared. The segment index with the opposite speed change trend and concentration change direction is extracted. The segment distribution density curve is established in time order, and the reverse cooperative distribution curve map is generated.

[0007] S2: Using the aforementioned reverse cooperative distribution curve spectrum, extract the unit energy consumption density value and main alloy element concentration sequence of the rolling process, perform trend consistency judgment, and obtain the composition offset driving path segment index set;

[0008] S3: Using the furnace number in the component offset driving path segment index set, extract the power input signal sequence of the smelting process and the hardness distribution array of the ingot cross section, calculate the hardness change difference between adjacent measuring points and extract the peak segment position, compare the time difference between the power peak index and the hardness fluctuation index, and generate a batch list of smelting intensity fluctuation risk.

[0009] S4: Based on the furnace number in the batch list of smelting strength fluctuation risk, extract the finished product hardness signal, yield ratio signal and toughness signal, calculate the interval distribution boundary value of the signal within the batch, determine whether there are samples that cross the extreme value interval, and obtain the finished product performance critical fluctuation traceability entry list.

[0010] As a further aspect of the present invention, the reverse cooperative distribution curve spectrum includes peak density of velocity change trend, abrupt change point of concentration change direction, distribution rate of segment with contradictory trend direction, and time series mapping node; the component offset driving path segment index set includes concentration offset direction judgment label, unit energy consumption density fluctuation point, trend consistency segment number, and energy consumption dominant path mark; the smelting intensity fluctuation risk batch list includes power input peak index, hardness fluctuation sensitive point, time difference screening range, and abnormal furnace identification mark; the finished product performance critical fluctuation traceability entry list includes performance signal extreme value boundary point, performance signal cross-segment sample number, furnace performance abnormal mark, and critical traceability entry number.

[0011] As a further aspect of the present invention, the step of obtaining the reverse cooperative distribution curve spectrum specifically includes:

[0012] S111: Based on the speed change signal sequence and the main alloy element concentration change signal sequence of the same rolling section, the two types of signals are divided into time axes, and the timestamp difference value and the corresponding signal change rate in the segment are extracted to obtain the time synchronization segment index table.

[0013] S112: Call the time synchronization segment index table, extract the difference between the speed change trend and the concentration change direction in each segment, determine whether the sign product is negative, mark the segment index with opposite directions, filter the segments with continuity not less than the difference interval threshold, and obtain the segment index sequence with opposite directions.

[0014] S113: Based on the index sequence of the opposite-direction segments, the index distribution density within the time unit is statistically analyzed in chronological order, a segment joint structure of rolling behavior and component response is constructed, the segment cooperative dispersion is calculated, and a reverse cooperative distribution curve spectrum is generated.

[0015] As a further aspect of the present invention, the step of obtaining the component offset driving path segment index set specifically includes:

[0016] S211: Based on the reverse cooperative distribution curve spectrum, extract the unit energy consumption density value sequence and the main alloy element concentration sequence, construct the energy consumption and composition mapping pair sequence, perform data structure initialization and index labeling, and obtain the cooperative mapping sequence set;

[0017] S212: Call the cooperative mapping sequence set, combine the time series data of unit energy consumption density value and main alloy element concentration, divide the segments into equidistant segments according to the fixed window length, calculate and distinguish the trend direction within the segments, filter the segments whose unit energy consumption density change direction is consistent with the main alloy element concentration offset direction, and generate a consistent offset segment set.

[0018] S213: Based on the set of consistent offset segments, obtain the sequence of unit energy consumption change values ​​and the sequence of principal element concentration change values, calculate the offset driving sequence number value, perform path segment filtering operation, and obtain the set of component offset driving path segment indexes.

[0019] As a further aspect of the present invention, the offset driving sequence number value is determined by the following formula:

[0020] ;

[0021] in, Indicates the first The offset driving sequence value of each path segment Indicates the first Change in unit energy density of the segment Indicates the first Changes in the concentration of main alloying elements in the segment. Indicates the first Section 1 The stability coefficient of the trend direction at each time point Indicates the first The mean of the stability values ​​within the segment. Indicates the collaborative offset rate. Indicates the number of time points.

[0022] As a further aspect of the present invention, the step of obtaining the batch list of smelting intensity fluctuation risks specifically includes:

[0023] S311: Using the furnace number in the component offset driving path segment index set, obtain the power input signal sequence in the smelting process, extract the corresponding power signal and hardness data, and perform time alignment processing on the power and hardness data of the furnace to obtain a hardness distribution array.

[0024] S312: Based on the hardness distribution array, calculate the hardness variation difference between adjacent measuring points, and extract the hardness peak segment position of each data segment. Use the position of the peak segment as the dividing criterion to obtain the hardness fluctuation range.

[0025] S313: Based on the hardness fluctuation range, compare the time difference between the power input signal sequence and the fluctuation of the hardness fluctuation range, filter the furnace numbers whose time difference is within the set range, and summarize the filtered furnace numbers to generate a batch list of smelting intensity fluctuation risk.

[0026] As a further aspect of the present invention, the steps for obtaining the list of entry points for tracing critical fluctuations in finished product performance are as follows:

[0027] S411: Based on the furnace number in the batch list of smelting intensity fluctuation risk, obtain the finished product hardness signal, yield ratio signal and toughness signal corresponding to the furnace number, extract the corresponding signal data according to the furnace number, and obtain the finished product performance signal set.

[0028] S412: Based on the set of finished product performance signals, calculate the interval distribution boundary value of each signal within the batch. By evaluating the upper and lower limits of the signal interval, identify whether there are samples that cross the extreme value interval. If a sample crosses the extreme value interval, mark it and associate the furnace number with the corresponding performance signal to perform furnace number and performance signal association traceability, and obtain the finished product performance critical fluctuation traceability entry list.

[0029] As a further aspect of the present invention, the method further includes step S5:

[0030] S5: Through the critical fluctuation traceability entry list of finished product performance, extract the rolling reverse coordination segment, energy consumption trend segment and smelting intensity offset segment, establish time axis mapping relationship in sequence, integrate speed change, power input and component concentration with performance data points, assemble into a phased sequence link structure, and generate a metallurgical quality state mapping link set.

[0031] The metallurgical quality status mapping link set includes speed trend link units, power input associated segments, concentration response mapping nodes, and performance result connection relationships.

[0032] As a further aspect of the present invention, the step of obtaining the metallurgical quality state mapping link set specifically includes:

[0033] S511: Based on the critical fluctuation traceability entry list of finished product performance, extract the rolling reverse coordination segment, energy consumption trend segment and smelting intensity offset segment, and extract the differentiated segments in sequence to obtain a time series data set.

[0034] S512: Based on the time series data set, establish the mapping relationship between segments and time axis in sequence, and obtain the time series link structure by performing time series integration of velocity change, power input and component concentration data in each segment;

[0035] S513: Based on the time series link structure, integrate the multi-stage speed change, power input, component concentration and performance data points, assemble the data points into a component stage sequence link structure according to the stage, and generate a metallurgical quality state mapping link set.

[0036] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0037] In this invention, a density distribution trend curve is formed by constructing an inverse matching relationship between the speed change trend and the main alloy element concentration change direction. The path segment index is screened based on the consistency between concentration shift and energy consumption trend. High-risk batches are identified by combining the time difference comparison method of smelting power fluctuation and ingot hardness distribution characteristics. A traceability entry list is formed by using boundary crossing analysis of performance signals. The speed signal, power input, concentration change and performance data are integrated into a staged link through a time axis mapping structure. On the basis of realizing the serial integration of multi-source dynamic data, the correlation path between rolling behavior, smelting intensity and performance is constructed, realizing the accurate identification and tracking mapping of key quality changes, and significantly improving the continuity and controllability of the quality response chain in the entire metallurgical process. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the main steps of the present invention;

[0039] Figure 2 This is a flowchart illustrating the process of obtaining the reverse cooperative distribution curve spectrum of the present invention.

[0040] Figure 3 This is a flowchart illustrating the process of obtaining the component offset-driven path segment index set of the present invention.

[0041] Figure 4 This is a flowchart illustrating the process of obtaining the batch list of smelting intensity fluctuation risk according to the present invention.

[0042] Figure 5 This is a flowchart illustrating the process of obtaining the entry list for tracing critical fluctuations in the performance of the finished product according to the present invention.

[0043] Figure 6 This is a flowchart illustrating the process of obtaining the metallurgical quality state mapping link set of this invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0046] Please see Figure 1 The metallurgical quality traceability method based on big data includes the following steps:

[0047] S1: Based on the speed change signal sequence and the main alloy element concentration change signal sequence of the same rolling section, the time axis is divided and the direction is compared. The index of the section with the opposite speed change trend and concentration change direction is extracted. The section distribution density curve is established in time order. The correlation structure between rolling behavior and composition response is constructed, and the reverse cooperative distribution curve spectrum is generated.

[0048] S2: Using the reverse cooperative distribution curve spectrum, extract the unit energy consumption density value and main alloy element concentration sequence of the rolling process, judge the trend consistency and screen the segments with the energy consumption trend and concentration shift direction, mark the path segment index, and obtain the composition shift driven path segment index set.

[0049] S3: Using the furnace number in the component offset drive path segment index set, extract the power input signal sequence of the smelting process and the hardness distribution array of the ingot cross section, calculate the hardness change difference between adjacent measuring points and extract the peak segment position, compare the time difference between the power peak index and the hardness fluctuation index, filter the furnace numbers that meet the set interval, and generate a batch list of smelting intensity fluctuation risk.

[0050] S4: Based on the furnace number in the batch list of smelting strength fluctuation risk, extract the finished product hardness signal, yield ratio signal and toughness signal, calculate the interval distribution boundary value of the signal within the batch, determine whether there are samples that cross the extreme value interval, mark the correlation information between the furnace number and the corresponding performance signal, and obtain the finished product performance critical fluctuation traceability entry list.

[0051] S5: By tracing the entry list of critical fluctuations in finished product performance, extract the rolling reverse coordination segment, energy consumption trend segment and smelting intensity offset segment, establish time axis mapping relationship in sequence, integrate speed change, power input and component concentration with performance data points, assemble them into a phased sequence link structure, and generate a metallurgical quality state mapping link set.

[0052] The reverse cooperative distribution curve spectrum includes peak density of velocity change trend, abrupt change points of concentration change direction, distribution rate of segments with contradictory trend directions, and time series mapping nodes. The component offset driving path segment index set includes concentration offset direction judgment label, unit energy consumption density fluctuation point, trend consistency segment number, and energy consumption dominant path mark. The smelting intensity fluctuation risk batch list includes power input peak index, hardness fluctuation sensitive point, time difference screening range, and abnormal furnace identification mark. The finished product performance critical fluctuation traceability entry list includes performance signal extreme boundary point, performance signal cross-segment sample number, furnace performance abnormal mark, and critical traceability entry number. The metallurgical quality status mapping link set includes velocity trend link unit, power input associated segment, concentration response mapping node, and performance result connection relationship.

[0053] Please see Figure 2 The specific steps for obtaining the reverse cooperative distribution curve spectrum are as follows:

[0054] S111: Based on the speed change signal sequence and the main alloy element concentration change signal sequence of the same rolling section, the two types of signals are divided into time axes, and the timestamp difference value and the corresponding signal change rate in the segment are extracted to obtain the time synchronization segment index table.

[0055] The two types of signals were processed by dividing them into time axes. By collecting time-series data from the speed sensor of the cold rolling mill, the speed change rate of each rolling section was recorded, and a time series was formed in seconds. Time series of changes in alloy element content were synchronously acquired from online spectral analysis equipment. The unit is mass percentage, defining a unified start time point for the two signals. Time alignment was performed with a step size of 0.5 seconds. Based on this, the difference in velocity change between each segment of the velocity change signal was calculated. and the differences in alloy element variations To determine whether two signals have the same trend of change at the same time step, the product is determined by their opposite direction of change. At that time, the time index of the recorded moment is the synchronous change segment index, as shown in the following example: between rolling segments 1 and 5, the speed changes as follows: The alloy concentration changes as , can be obtained , Therefore, segments 1 and 2 are synchronous segments (with the same direction of change), and segments 3 and 4 are also synchronous segments (with the same direction of change), thus obtaining the time synchronization segment index table.

[0056] S112: Call the time-synchronized segment index table, extract the difference between the speed change trend and the concentration change direction in each segment, determine whether the sign product is negative, mark the segment index with opposite directions, filter the segments with continuity not less than the difference interval threshold, and obtain the segment index sequence with opposite directions.

[0057] For each synchronization segment, calculate the difference between the velocity change trend and the concentration change direction. The difference is calculated as follows: Let the velocity difference be... The concentration difference is The trend difference term is defined as follows: Then judge the difference item. The sign of the product is used to determine its value. The product is less than 0 to identify whether the paragraphs are in opposite directions. If the product is less than 0, it means that the trend of change is opposite, and the index is recorded as the index of the opposite direction paragraph. After filtering, it is judged whether the opposite direction relationship is still satisfied between two consecutive paragraphs of not less than 0. That is, the sliding window filters out two or more adjacent paragraph groups that meet the conditions. The filtering result is the list of opposite direction paragraph indexes. For example, in the change of speed and concentration, the speed change of paragraph 2 is 0.5 and the concentration change is 0.03%, the product is positive, so it is not recorded. The speed change of paragraph 3 is -0.1 and the concentration change is -0.01%, the product is positive, so it is not recorded. However, if a paragraph has a speed change of 0.3 and a concentration change of -0.02%, the product is negative, so it is recorded. If the subsequent paragraphs continue to meet the condition that the product is less than 0, it is recorded as a continuous opposite paragraph sequence, and the opposite direction section index sequence is obtained.

[0058] S113: Based on the index sequence of opposite-direction segments, the index distribution density within the time unit is statistically analyzed in chronological order to construct a segment joint structure of rolling behavior and component response, using the formula:

[0059] ;

[0060] Calculate the paragraph co-dispersion degree and generate the inverse co-dispersion curve spectrum;

[0061] in, Represents the coherence dispersion of paragraphs. Representing the The rate of change of the segment velocity signal Representing the Rate of change of segment concentration signal Representing the Segment start timestamp, For the first The start timestamp of the segment. Representing the The duration of the segment Number of paragraphs;

[0062] Formula calculation logic: Extract the rate of change of velocity for each segment sequentially. Concentration change rate The start time of the current segment Previous start time and the duration of the current segment. Multiply the rate of change of concentration by the rate of change of concentration. As a measure of signal linkage strength, it calculates the sum of the squares of the time span and the duration. This indicates the degree of expansion in the time dimension. Dividing the two quantities mentioned above and taking the square root yields the co-dispersion factor for each segment. Summing these factors for each segment gives the overall segment co-dispersion factor. The measurement system is constructed with the structure of "cooperative change as numerator and time span as denominator", which makes segments with the same direction of change but different time densities comparable, and ensures that the calculated values ​​can truly reflect the degree of coupling of each segment under multivariable signals.

[0063] Segment coordinating dispersion is a statistic used to measure the strength of synchronicity between velocity and concentration changes in different segments. It is obtained by summing the square root of the ratio between the signal linkage strength and the time extension of each segment. It reflects the characteristics of coordinated changes of multiple signal sources in time. The larger the value, the more intense the coordinated fluctuation between signals, and the smaller the value, the more dispersed or inconsistent the change trend.

[0064] The parameters are explained as follows:

[0065] : No. The rate of change of velocity of the segment, in m / s;

[0066] : No. The concentration change rate of the segment, in %

[0067] : No. Segment start timestamp, in seconds;

[0068] : The start timestamp of the previous segment;

[0069] : No. The duration of the segment, in seconds;

[0070] The statistical window is divided according to time order, with each time unit divided into a fixed-length window, such as 1 minute or 300 sampling points. The distribution density of segment indexes appearing within the time period is statistically analyzed, and a joint structure of segments is constructed.

[0071] The following are actual data examples:

[0072] Table 1: Calculation Parameter Table

[0073] ;

[0074] As shown in Table 1, substitute the above parameters into the calculation:

[0075] Section 1:

[0076] ;

[0077] ;

[0078] Section 2:

[0079] ;

[0080] ;

[0081] Section 3:

[0082] ;

[0083] ;

[0084] Substitute into the formula to calculate:

[0085] ;

[0086] The results show that the paragraph co-dispersion is 0.023104, which can be used as a reference value for the joint distribution characteristics of the statistical paragraph group behavior.

[0087] The advantage of the formula is that by using the product of velocity change and concentration change as a synergistic factor, and introducing a joint discrete influence factor of time span and segment duration, it can accurately characterize the coupling characteristics of segment group behavior under multi-source signals.

[0088] Please see Figure 3 The specific steps for obtaining the component offset-driven path segment index set are as follows:

[0089] S211: Based on the reverse cooperative distribution curve spectrum, extract the unit energy consumption density value sequence and the main alloy element concentration sequence, construct the energy consumption and composition mapping pair sequence, perform data structure initialization and index labeling, and obtain the cooperative mapping sequence set;

[0090] The unit energy consumption density value sequence and the main alloying element concentration sequence were extracted. The unit energy consumption density sequence was obtained by recording the electrical power consumed in a single segment of aluminum alloy rolling (in kW) and dividing it by the corresponding weight of the produced material (in kg). One data point was generated every second to obtain the energy consumption density time series. Main alloying element concentration sequence The mass percentage value is obtained every 2 seconds using an online spectrometer. The sampling rate is uniformly set to a 2-second interval within the intersection time window of the two sequences. A dual-channel array structure for energy consumption density and alloy concentration, based on the timestamp, is then established, as shown in the image. ; The data is divided into non-overlapping segments with a window length of 10 seconds. The average and standard deviation of the contained values ​​in each segment are taken as segment features. Then, the energy density feature values ​​are normalized, with the normalization range set to [0, 1]. The minimum-maximum normalization operation is used, and the minimum value of the i-th segment is set to... The maximum value is The original value is The normalized value is Then, the paragraph numbers are sorted from smallest to largest according to the normalization value, and the paragraph numbers under the same normalization energy consumption density are marked to obtain the co-mapping sequence set.

[0091] S212: Call the co-mapping sequence set, combine the time series data of unit energy consumption density value and main alloy element concentration, divide the segments into equidistant segments according to the fixed window length, calculate and determine the trend direction within the segments, filter the segments whose unit energy consumption density change direction is consistent with the main alloy element concentration offset direction, and generate a set of consistent offset segments.

[0092] The data structure retrieves the number of each pair of coordinating segments. Combining the time data of the unit energy density sequence and the main alloying element concentration sequence, a fixed-length sliding window is selected and divided into equally spaced segments. The window length is set to 20 seconds, and each window includes at least 5 segments. The trend direction is calculated for each segment sequence within the window. The trend direction is determined by comparing the first and last values ​​of the time series within the segment. If the last value is greater than the first value, the trend is defined as upward; otherwise, it is defined as downward. For example, the energy density sub-segment... Comparison of first and last values If the trend direction is determined to be upward, the same determination is made for the alloy element concentration segment. When the trend direction is consistent with the energy consumption trend, it is considered a segment with consistent offset direction. The segment numbers that meet the above conditions are summarized to form a consistent offset segment set. To ensure the stability of the judgment, a trend difference ratio index is introduced. ,when The trend direction is valid when the (empirical benchmark value) is reached; otherwise, it is considered an invalid segment and is removed. For example, if the concentration change in a segment is [0.19%, 0.1903%, 0.1906%], the trend difference is 0.0006. If it is less than 0.005, it is removed. Only the set of segments with clear trends and consistent directions is retained to generate a set of consistent offset segments.

[0093] S213: Based on the set of consistent offset segments, obtain the sequence of unit energy consumption change values ​​and the sequence of principal element concentration change values, using the following formula:

[0094] ;

[0095] Calculate the offset-driven sequence number value, perform path segment filtering operation, and obtain the component offset-driven path segment index set;

[0096] in, Indicates the first The offset driving sequence value of each path segment Indicates the first Change in unit energy density of the segment Indicates the first Changes in the concentration of main alloying elements in the segment. Indicates the first Section 1 The stability coefficient of the trend direction at each time point Indicates the first The mean of the stability values ​​within the segment. Indicates the collaborative offset rate. Indicates the number of time points;

[0097] Formula calculation logic: By calculating the current path segment's... Change in energy consumption per unit at location Changes in the concentration of main alloying elements The difference is used to obtain the offset difference as a driving factor. Trend stability value at each time point with the mean Find the absolute value of the deviation. This is used to characterize the range of trend fluctuations within the segment. Multiplying the two parts yields a joint term reflecting the difference in change and the amplitude of fluctuation. Then, the segment's co-offset rate is used as the final term. Add 1 as the denominator and take the square root to form the normalization coefficient, thus obtaining the offset driving strength of the path segment. In the calculation process, each indicator has physical meaning: the difference reflects the degree of coordination and consistency, the deviation reflects the internal fluctuation characteristics, the normalization factor eliminates the influence of different segment lengths, and the whole calculation process is a combination of linear product and square normalization structure to ensure that the offset driving value considers both local differences and global coordination.

[0098] The offset drive index is a numerical indicator used to quantify the strength of the driving force generated by a certain path segment in the unit energy consumption and main alloy concentration change. It is calculated by normalizing the product of the cooperative offset difference and the trend fluctuation amplitude. It reflects the dominant role of the segment in the offset process. The larger the value, the stronger the influence in path formation, and the smaller the value, the weaker the cooperative offset effect.

[0099] Parameter meaning:

[0100] : The change in unit energy consumption density in segment a, in kW / kg;

[0101] : Change in main alloy concentration in segment a, in %

[0102] : Trend direction stability coefficient at time point j in segment a, unit: none;

[0103] : The mean stability coefficient of segment a;

[0104] : Represents the co-offset rate of segment a (the percentage of segments with consistent trends within the statistical interval multiplied by 100);

[0105] Substitute the following actual data into the formula to perform the calculation:

[0106] Table 2: Path Segment Offset Driving Parameters

[0107] ;

[0108] The calculation process based on the formula is as follows:

[0109] Difference item: ;

[0110] Deviations and terms: ;

[0111] Partial product of the numerator: ;

[0112] Denominator part: ;

[0113] Substitute into the formula to calculate:

[0114] ;

[0115] The results show that the offset driving value IH of segment 1 is 0.000252, which indicates that this segment has a low offset driving force in terms of energy consumption and alloy concentration offset.

[0116] The advantage of the formula is that by multiplying the difference between the two core variables with the cumulative term of the trend deviation, and introducing the cooperative offset rate as a normalization term, it can accurately characterize the driving force of different path segments in cooperative evolution.

[0117] Please see Figure 4 The specific steps for obtaining the batch list of smelting intensity fluctuation risk are as follows:

[0118] S311: Using the furnace number in the component offset driving path segment index set, obtain the power input signal sequence in the smelting process, extract the corresponding power signal and hardness data, and perform time alignment processing on the power and hardness data of the furnace to obtain the hardness distribution array.

[0119] The system retrieves the path segment index set information corresponding to each heat from the production database. Combined with the smelting batch number of each heat, the number is input into the composition offset recording device to obtain the composition change record corresponding to the path segment. Then, according to the number of each heat, the corresponding power input signal sequence is retrieved. The signal is generally stored at a second-level sampling frequency and can be directly exported from the smelting process control equipment. For different heats, the power curve shows different fluctuation characteristics. The system retrieves the hardness distribution data of the ingot cross section related to the heat. The data comes from the point test in the quality inspection process. The point spacing is 10mm, and the points are arranged radially from the center outwards. After extracting the power and hardness data corresponding to the heat, the power signal needs to be preprocessed. Through linear matching or interpolation on the time axis, the power data points and hardness test points are made consistent in the time dimension. Taking a certain ingot as an example, the power is in a stable rising range at a certain stage in the smelting process, and there is also a significant change in the corresponding hardness distribution. Therefore, after aligning the data, a hardness distribution array is obtained.

[0120] S312: Based on the hardness distribution array, calculate the hardness variation difference between adjacent measuring points, extract the hardness peak segment position of each data segment, and use the position of the peak segment as the dividing criterion to obtain the hardness fluctuation range.

[0121] The hardness distribution array needs to be analyzed point by point, using the change in hardness difference between points as the criterion. The difference between every two adjacent points is calculated, and the trend of the difference is used to preliminarily determine which data segment belongs to the fluctuation area. If a segment shows a large continuous change in the difference, it can be considered that there is obvious fluctuation, and the local extreme value area, namely the hardness peak segment, is identified. The peak segment reflects the location where the material hardness suddenly changes. After selecting the peak segment, its location needs to be extracted. The location can be marked by the measurement point number or the physical distance on the actual ingot structure. Taking a batch of ingots as an example, the hardness in the central area rises continuously and then drops rapidly. This is identified as a hardness peak segment. Using the peak segment as the boundary, the entire hardness data is divided into multiple sub-intervals. Each sub-interval represents a relatively stable or drastically changing hardness area. If a fluctuation interval is formed between the 2nd and 4th measurement points, and the hardness values ​​of the 4th and 5th points are similar, it can be regarded as a non-fluctuation interval, thus obtaining the hardness fluctuation interval.

[0122] S313: Based on the hardness fluctuation range, compare the time difference between the power input signal sequence and the fluctuation of the hardness fluctuation range, filter the furnace number with the time difference within the set range, and summarize the filtered furnace number to generate a batch list of smelting intensity fluctuation risk.

[0123] The starting and ending points of each fluctuation range in time are traced back and compared with the power input signal sequence in the smelting process in the time dimension. The physical location corresponding to the hardness fluctuation range is converted into a smelting time point. Based on the process flow and the test point layout sequence, the corresponding power input period is estimated. The fluctuation degree of the power signal curve is analyzed within the period, paying particular attention to whether there is a sudden change in power value in a short period of time. If the rate of change of power exceeds the statistical average within a certain period of time and the duration is within the set range, it is determined to be a power fluctuation point. The timestamp of the power fluctuation point is compared with the center time of the hardness fluctuation range. If the time difference is within the preset comparison range, it is considered that there is a synchronous relationship between the two, and the furnace is marked as a risk furnace. Taking a real case as an example, the ingot with furnace number 210316A has a hardness peak at 1900 seconds and a power sudden change is detected at 1870 seconds, with a time difference of 30 seconds. If the threshold is set to 60 seconds, the furnace is included in the risk batch list. After completing the screening of furnaces, the furnace numbers that meet the conditions are uniformly archived as smelting intensity fluctuation risk batches, generating a smelting intensity fluctuation risk batch list.

[0124] Please see Figure 5 The specific steps for obtaining the list of entry points for tracing critical fluctuations in finished product performance are as follows:

[0125] S411: Based on the furnace number in the batch list of smelting strength fluctuation risk, obtain the finished product hardness signal, yield ratio signal and toughness signal corresponding to the furnace number, extract the corresponding signal data according to the furnace number, and obtain the finished product performance signal set.

[0126] The system sequentially retrieves the finished product performance signal data corresponding to each heat number from the finished product performance database. The data includes three dimensions: hardness signal, yield ratio signal, and toughness signal. During the operation, the heat numbers are first traversed, and each number is entered into the database query. The corresponding quality inspection report or online test data record is retrieved, and the raw signal data of the three performance indicators is extracted. The signal data is presented as a sequence that changes with the test point location or test time. Taking 210320C as an example, five test positions are set up during the test, corresponding to the hardness value, yield ratio value, and impact toughness value, respectively. The actual measurement records at each point form three sets of signal sequences. After completing the batch traversal and extracting the signals corresponding to the target heat, the signals are organized into a signal set with a unified format. Structurally, each heat number is bound to the corresponding three performance signals and standardized. The organized signal set can serve as the input basis for subsequent performance trend analysis within the batch, ensuring that each heat sample has searchability and complete performance attribute comparison records in the subsequent processing, thus forming a finished product performance signal set.

[0127] S412: Based on the set of finished product performance signals, calculate the interval distribution boundary value of each signal within the batch. By evaluating the upper and lower limits of the signal interval, identify whether there are samples that cross the extreme value interval. If a sample crosses the extreme value interval, mark it and associate the furnace number with the corresponding performance signal. Perform furnace number and performance signal association and traceability to obtain the finished product performance critical fluctuation traceability entry list.

[0128] Each performance data point in the signal set needs to undergo batch-specific distribution range analysis. Based on the overall distribution of each type of signal within the batch range, upper and lower boundary values ​​are extracted. The maximum and minimum values ​​of the hardness values ​​in the hardness signal are selected as boundaries to form the performance value range. The values ​​of each sample in the three signal dimensions are compared one by one to determine whether they are near the boundary or have crossed the extreme value range. If the toughness signal of a certain batch exceeds the maximum toughness value of the entire batch or is lower than the minimum value, it is considered a sample that has crossed the extreme value range and is marked. After marking, the sample number is associated with the corresponding abnormal signal item through a data table to establish a binding traceability structure between the batch number and the corresponding performance signal. If the yield ratio signal value of batch number 210325B is 0.98, while the batch yield ratio distribution range is 0.82~0.95, the value is determined to be an abnormal point that has crossed the extreme value range. Such samples will be selected and recorded in the traceability entry list to provide a traceable abnormal entry path for process adjustment, resulting in the finished product performance critical fluctuation traceability entry list.

[0129] Please see Figure 6 The specific steps for obtaining the metallurgical quality state mapping link set are as follows:

[0130] S511: Based on the critical fluctuation traceability entry list of finished product performance, extract the rolling reverse coordination segment, energy consumption trend segment and smelting intensity offset segment, and extract the differentiated segments in sequence to obtain a time series data set.

[0131] The rolling reverse coordination segment, energy consumption trend segment, and smelting intensity offset segment information corresponding to each heat are extracted sequentially from the production process data. Each heat is used as an index item, and horizontal correlation matching is performed on data tables from different sources. When extracting the rolling reverse coordination segment, it is necessary to lock the rolling time period corresponding to the heat, and then extract the speed command, roll gap adjustment information, and tension feedback data within the forward and reverse alternating control interval based on the rolling equipment operation record to restore the flow path characteristics of raw materials during the rolling process. The energy consumption trend segment is obtained based on the energy consumption record logs of each section of the production line where the heat is located, extracting the energy consumption change trends of the melting, refining, rolling, and heat treatment stages, forming an energy consumption fluctuation sequence in minutes. During the extraction of the smelting intensity offset segment, it is necessary to review the strength control records of the heat at the main alloy element control points in the smelting record, and obtain the corresponding power input and time node markers as strength offset references. In this way, multiple time-series segments corresponding to each heat can be constructed from the original metallurgical process big data to obtain a time series data set.

[0132] S512: Based on the time series data set, the mapping relationship between segments and the time axis is established sequentially. By integrating the velocity change, power input and component concentration data in each segment in a time series, the time series link structure is obtained.

[0133] A clear timeline correspondence needs to be established for each segment in the set to ensure that subsequent analysis can accurately locate the temporal position of each data point. During operation, each data segment is sorted in order according to the collection timestamp, and the start and end times are confirmed as the starting and ending points of the segment's mapping on the timeline. Taking the rolling reverse collaborative segment as an example, the system will divide several sub-intervals according to the recorded forward and reverse switching command time points, and map the sub-intervals to the unified timeline of the entire process. The speed change data, power input data, and component concentration data in each segment are integrated, and different types of data points are merged in chronological order to form a complete ternary time series structure. In a certain heat 210329D, the forward rolling duration is 180 seconds and the reverse rolling duration is 160 seconds. Several speed change data points are collected in these two stages and recorded synchronously with the power input and component concentration data. This processing method can ensure that the dimensional data in the segment are strictly synchronized in time, providing a consistent time reference basis for cross-segment analysis and state modeling, and ensuring that there is no time misalignment between data in different segments, thus obtaining a time series link structure.

[0134] S513: Based on the time series link structure, integrate multi-stage velocity change, power input, component concentration and performance data points, assemble the data points into a component stage sequence link structure according to the stage, and generate a metallurgical quality state mapping link set.

[0135] This approach achieves data linkage and integration across multiple segments over time. It progressively aggregates speed change data, power input data, component concentration data, and product performance signals related to each stage, and combines them into a stage sequence link structure according to the order of each stage. In practice, the furnace number is used as the primary key to classify associated data segments into stages, such as smelting, rolling, heat treatment, and testing, and the start and end times of each stage are clearly defined. Based on this classification, the speed, power, and component information within each stage are sequentially assembled to form a time-progressing sequence. Corresponding performance testing data, such as hardness and toughness, are appended at the end of each stage as stage output status identifiers. In furnace 210329D, the smelting stage lasts 400 seconds, the rolling stage lasts 300 seconds, and the heat treatment stage lasts 240 seconds. Core data points are extracted from each segment and arranged sequentially on the time axis, generating a continuous information flow link from initial to final state. The link structures constructed for each furnace are then summarized to form a metallurgical quality state mapping link set.

[0136] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A metallurgical quality traceability method based on big data, characterized in that, Includes the following steps: S1: Based on the signal sequence of the rolling speed change segment and the signal sequence of the main alloy element concentration change segment in the same furnace, the time axis is divided and the direction is compared. The segment index with the opposite trend of speed change and concentration change is extracted. The segment distribution density curve is established in time order, and the reverse cooperative distribution curve spectrum is generated. S2: Using the aforementioned reverse cooperative distribution curve spectrum, extract the unit energy consumption density value and main alloy element concentration sequence of the rolling process, perform trend consistency judgment, and obtain the composition offset driving path segment index set. The specific steps are as follows: S211: Based on the reverse cooperative distribution curve spectrum, extract the unit energy consumption density value sequence and the main alloy element concentration sequence, construct the energy consumption and composition mapping pair sequence, perform data structure initialization and index labeling, and obtain the cooperative mapping sequence set; S212: Call the cooperative mapping sequence set, combine the time series data of unit energy consumption density value and main alloy element concentration, divide the segments into equidistant segments according to the fixed window length, calculate and distinguish the trend direction within the segments, filter the segments whose unit energy consumption density change direction is consistent with the main alloy element concentration offset direction, and generate a consistent offset segment set. S213: Based on the set of consistent offset segments, obtain the sequence of unit energy consumption change values ​​and the sequence of principal element concentration change values, calculate the offset driving sequence number value, perform path segment filtering operation, and obtain the set of component offset driving path segment indexes. The offset drive sequence number value is calculated using the following formula: ； in, Indicates the first The offset driving sequence value of each path segment Indicates the first Change in unit energy density of the segment Indicates the first Changes in the concentration of main alloying elements in the segment. Indicates the first Section 1 The stability coefficient of the trend direction at each time point Indicates the first The mean of the stability values ​​within the segment. Indicates the collaborative offset rate. Indicates the number of time points; S3: Using the furnace number in the component offset driving path segment index set, extract the power input signal sequence of the smelting process and the hardness distribution array of the ingot cross section, calculate the hardness change difference between adjacent measuring points and extract the peak segment position, compare the time difference between the power peak index and the hardness fluctuation index, and generate a batch list of smelting intensity fluctuation risk. S4: Based on the furnace number in the batch list of smelting strength fluctuation risk, extract the finished product hardness signal, yield ratio signal and toughness signal, calculate the interval distribution boundary value of the signal within the batch, determine whether there are samples that cross the extreme value interval, and obtain the finished product performance critical fluctuation traceability entry list.

2. The metallurgical quality traceability method based on big data according to claim 1, characterized in that, The reverse cooperative distribution curve spectrum includes peak density of velocity change trend, abrupt change point of concentration change direction, distribution rate of segment with contradictory trend direction, and time series mapping node. The component offset driving path segment index set includes concentration offset direction judgment label, unit energy consumption density fluctuation point, trend consistency segment number, and energy consumption dominant path mark. The smelting intensity fluctuation risk batch list includes power input peak index, hardness fluctuation sensitive point, time difference screening range, and abnormal furnace identification mark. The finished product performance critical fluctuation traceability entry list includes performance signal extreme boundary point, performance signal cross-segment sample number, furnace performance abnormal mark, and critical traceability entry number.

3. The metallurgical quality traceability method based on big data according to claim 1, characterized in that, The specific steps for obtaining the reverse cooperative distribution curve spectrum are as follows: S111: Based on the speed change signal sequence and the main alloy element concentration change signal sequence of the same rolling section, the two types of signals are divided into time axes, and the timestamp difference value and the corresponding signal change rate in the segment are extracted to obtain the time synchronization segment index table. S112: Call the time synchronization segment index table, extract the difference between the speed change trend and the concentration change direction in each segment, determine whether the sign product is negative, mark the segment index with opposite directions, filter the segments with continuity not less than the difference interval threshold, and obtain the segment index sequence with opposite directions. S113: Based on the index sequence of the opposite-direction segments, the index distribution density within the time unit is statistically analyzed in chronological order, a segment joint structure of rolling behavior and component response is constructed, the segment cooperative dispersion is calculated, and a reverse cooperative distribution curve spectrum is generated.

4. The metallurgical quality traceability method based on big data according to claim 3, characterized in that, The specific steps for obtaining the batch list of smelting intensity fluctuation risk are as follows: S311: Using the furnace number in the component offset driving path segment index set, obtain the power input signal sequence in the smelting process, extract the corresponding power signal and hardness data, and perform time alignment processing on the power and hardness data of the furnace to obtain a hardness distribution array. S312: Based on the hardness distribution array, calculate the hardness variation difference between adjacent measuring points, and extract the hardness peak segment position of each data segment. Use the position of the peak segment as the dividing criterion to obtain the hardness fluctuation range. S313: Based on the hardness fluctuation range, compare the time difference between the power input signal sequence and the fluctuation of the hardness fluctuation range, filter the furnace numbers whose time difference is within the set range, and summarize the filtered furnace numbers to generate a batch list of smelting intensity fluctuation risk.

5. The metallurgical quality traceability method based on big data according to claim 4, characterized in that, The specific steps for obtaining the list of entry points for traceability of critical fluctuations in finished product performance are as follows: S411: Based on the furnace number in the batch list of smelting intensity fluctuation risk, obtain the finished product hardness signal, yield ratio signal and toughness signal corresponding to the furnace number, extract the corresponding signal data according to the furnace number, and obtain the finished product performance signal set. S412: Based on the set of finished product performance signals, calculate the interval distribution boundary value of each signal within the batch. By evaluating the upper and lower limits of the signal interval, identify whether there are samples that cross the extreme value interval. If a sample crosses the extreme value interval, mark it and associate the furnace number with the corresponding performance signal to perform furnace number and performance signal association traceability, and obtain the finished product performance critical fluctuation traceability entry list.

6. The metallurgical quality traceability method based on big data according to claim 1, characterized in that, The method further includes step S5: S5: Through the critical fluctuation traceability entry list of finished product performance, extract the rolling reverse coordination segment, energy consumption trend segment and smelting intensity offset segment, establish time axis mapping relationship in sequence, integrate speed change, power input and component concentration with performance data points, assemble into a phased sequence link structure, and generate a metallurgical quality state mapping link set. The metallurgical quality status mapping link set includes speed trend link units, power input associated segments, concentration response mapping nodes, and performance result connection relationships.

7. The metallurgical quality traceability method based on big data according to claim 6, characterized in that, The specific steps for obtaining the metallurgical quality state mapping link set are as follows: S511: Based on the critical fluctuation traceability entry list of finished product performance, extract the rolling reverse coordination segment, energy consumption trend segment and smelting intensity offset segment, and extract the differentiated segments in sequence to obtain a time series data set. S512: Based on the time series data set, establish the mapping relationship between segments and time axis in sequence, and obtain the time series link structure by performing time series integration of velocity change, power input and component concentration data in each segment; S513: Based on the time series link structure, integrate the multi-stage speed change, power input, component concentration and performance data points, assemble the data points into a component stage sequence link structure according to the stage, and generate a metallurgical quality state mapping link set.

Images