Steel production quality intelligent labeling and analysis method and system based on multi-process space-time mapping

Abstract

The invention provides an intelligent marking and analyzing method for steel production quality based on multi-process space-time mapping. The method comprises the following steps: acquiring a time sequence curve of characteristic parameters, material tracking data and material quality data of a target material in each process in production; according to the material tracking data, establishing a mapping model of the processing position, the process processing time and the cross-process processing position; standardized marking is conducted on the machining position and the machining time interval based on the mapping model; extracting corresponding time sequence curve features based on the annotation interval; analyzing the difference between the target material and the reference material based on feature comparison; and generating a diagnosis result according to the analysis result. The technical problems that in the prior art, the process curve feature extraction mode is single, cross-process data are difficult to associate, and the quality problem tracing efficiency is low are solved.

Description

Technical Field

[0001] This invention relates to the field of strip steel production quality control in the steel production industry, specifically to a method and system for intelligent labeling and analysis of steel production quality based on multi-process spatiotemporal mapping. Background Technology

[0002] In modern manufacturing, especially in continuous processing lines for strip steel and sheet metal, process monitoring and quality problem traceability are increasingly reliant on big data and intelligent technologies. Traditional methods for extracting feature values, such as mean, extreme values, and variance, from process curves based on rules or statistical thresholds to aid in quality problem analysis, often rely on rule-based or statistical threshold-based approaches. However, these methods have significant limitations when dealing with complex fluctuation patterns, cross-process data integration, and defect location and traceability.

[0003] Taking existing technology as an example, Chinese patent CN112085391A discloses a method for extracting feature values ​​from strip steel production process data based on process rules. This method mainly relies on statistical quantities in the time domain for feature analysis, such as average values ​​and extreme values, and is suitable for simple and stable production processes. However, this method has revealed the following shortcomings in practical applications: First, its feature extraction is limited to the time domain and lacks frequency domain analysis methods, making it difficult to capture deeper information such as high-frequency noise and fluctuation trends; second, it does not establish a systematic data alignment mechanism, resulting in a lack of correlation between the defect and multi-source data of the production process, making it impossible to achieve cross-process quality analysis; in addition, this method relies on manually set rules, lacks flexibility, and is difficult to adapt to complex and ever-changing on-site conditions.

[0004] Meanwhile, in practical quality analysis scenarios, manual annotation of defects and performance anomalies often suffers from high subjectivity, low efficiency, and difficulty in reproducing results. Especially in processing massive amounts of data, manual methods struggle to meet the demands of batch processing and automation. Faced with complex process paths and multidimensional process variables, existing methods are inadequate in areas such as visualization analysis, defect location accuracy, and inter-process information synchronization.

[0005] To address the aforementioned issues, this invention proposes an intelligent analysis scheme that combines manual and automatic annotation. It achieves automatic annotation triggering through an event-driven mechanism, and realizes structured analysis of complex process curves through multi-dimensional feature extraction such as frequency domain and volatility. At the same time, it combines data alignment and differential analysis strategies to effectively support cross-process quality analysis and problem traceability, significantly improving the utilization efficiency of manufacturing data and the systematization and automation level of analysis. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for intelligent labeling and analysis of steel production quality based on multi-process spatiotemporal mapping.

[0007] According to one aspect of the present invention, a method for intelligent labeling and analysis of steel production quality based on multi-process spatiotemporal mapping is provided, comprising: Step 1: acquiring time-series curves of characteristic parameters of target materials in each process of production, material tracking data, and material quality data; Step 2: establishing a mapping model between target processing position and target process processing time, as well as cross-process corresponding processing positions, based on material tracking data; Step 3: standardizing and labeling the intervals of target processing position and target process processing time based on the mapping model; Step 4: extracting features of the corresponding time-series curves based on the labeled intervals; Step 5: comparing and analyzing the differences between target materials and preset benchmark materials for the same target processing position based on the features of the time-series curves; Step 6: generating diagnostic results based on the analysis results.

[0008] Preferably, the material tracking data includes material identification information, time information, and material specification change information; the material quality data includes defect information, blockage information, detection point or sampling point information, and abnormal or intervention event information.

[0009] Preferably, step two, based on material tracking data, establishes a mapping model between the target processing location and the target process processing time, as well as the corresponding processing location across processes, including: Construction of an intra-process time-location mapping model: Based on the material's production start time, production end time, and material length, a linear mapping relationship between processing time and processing location within the same process is established, expressed by the following formula:

[0010] Where P represents the target processing position, L represents the material length, T represents the target process processing time, and T_start and T_end represent the production start time and production end time of the material in that process, respectively; Unified sampling frequency processing for multiple process parameters: For time series curves of multiple process parameters with different sampling frequencies in the same process, calculate the least common multiple of the number of sampling points of all parameters, and use this multiple as the target frequency to perform interpolation processing on the time series curves of different process parameters, so that all process parameters are unified to the same sampling points. Construction of cross-process position mapping model: Based on the flipping information, the materials in the upstream and downstream processes are aligned in the head and tail directions. Based on the specification change information of the materials between cross processes, the mapping relationship of the processing position across processes is established. The formula is expressed as:

[0011] Where P_up_corrected is the upstream process position after head-to-tail alignment correction, P_down represents the corresponding downstream process position obtained by mapping, and L_up and L_down represent the complete length of the material in the upstream and downstream processes, respectively.

[0012] Preferably, step three, based on the mapping model, standardizes and labels the intervals of the target processing location and the processing time of the target process, including: The system receives user input, which includes selected target quality information and one or more process parameters that need to be correlated and analyzed; the target quality information includes one or more of the following: defect type, defect location, blockade location, and sampling point location. Based on the target quality information, the corresponding target processing position range is automatically extracted; then, based on the established mapping model, the target processing position range is automatically mapped to the corresponding processing position range or processing time range of the process in which the associated process parameter is located, thus completing the initial automatic labeling of the time series data of the associated process parameter. The results of the automatic annotation are displayed through a graphical interface, and manual verification and adjustment are performed to correct annotation errors or supplement missing annotations. Record the final annotation results, establish the correspondence between the target quality information and the spatiotemporal intervals of the associated process parameters, and form standardized annotation data for subsequent feature extraction and quality analysis.

[0013] Preferably, the features include statistical features and frequency domain features. The statistical features include mean, variance, extreme values, kurtosis, and skewness. The frequency domain features include periodic fluctuations and energy distribution features.

[0014] Preferably, in step five, comparing and analyzing the differences between the target material and a preset benchmark material includes: For numerical features, the relative deviation calculation method is used to measure the numerical difference of feature values; for distribution features, the distribution difference is calculated based on the statistical divergence index; for waveform features, the dynamic time warping algorithm is used to align the feature curves in time series, and the similarity index is calculated based on the alignment to calculate the fluctuation difference; based on the analysis results of the difference, data visualization analysis graphics are generated, including distribution comparison charts, parameter heatmaps, trend overlay charts, and causal relationship charts.

[0015] According to another aspect of the present invention, a smart labeling and analysis system for steel production quality based on multi-process spatiotemporal mapping is provided, comprising: module M1: acquiring time-series curves of characteristic parameters of target materials in each process of production, material tracking data, and material quality data; module M2: establishing a mapping model between target processing position and target process processing time, as well as cross-process corresponding processing positions, based on material tracking data; module M3: standardizing and labeling the intervals of target processing position and target process processing time based on the mapping model; module M4: extracting features of the corresponding time-series curves based on the labeled intervals; module M5: comparing and analyzing the differences between target materials and preset benchmark materials for the same target processing position based on the features of the time-series curves; and module M6: generating diagnostic results based on the analysis results.

[0016] Preferably, the material tracking data includes material identification information, time information, and material specification change information; the material quality data includes defect information, blockage information, detection point or sampling point information, and abnormal or intervention event information.

[0017] Preferably, module M3 includes: submodule M1: locating the target processing position interval based on the input interval selection information or sampling position, blockade position, and defect position of the target processing position; locating the cross-process corresponding processing position interval based on the mapping model between the target processing position and the corresponding processing position across processes; locating the corresponding target process processing time interval based on the mapping model between the target processing position and the target process processing time; and tracking the corresponding time series data and material quality information; submodule M2: locating the target process processing time interval based on the input interval selection information or manual intervention time and process abnormality time of the target process processing time; locating the corresponding target processing position interval based on the mapping model between the target processing position and the target process processing time; locating the cross-process corresponding processing position interval based on the mapping model between the target processing position and the corresponding processing position across processes; and tracking the corresponding time series data and material quality information.

[0018] Preferably, the features include statistical features and frequency domain features. The statistical features include mean, variance, extreme values, kurtosis, and skewness. The frequency domain features include periodic fluctuations and energy distribution.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. This application establishes a conversion mechanism between processing time and processing location, and further constructs a cross-process location mapping model, achieving unified integration and precise alignment of production data from various processes, forming a complete data chain spanning time, space, and quality information. This technical solution eliminates the drawbacks of data fragmentation between different processes, providing a stable and reliable foundation for subsequent quality traceability, anomaly diagnosis, and process optimization.

[0020] 2. This application introduces a standardized annotation mechanism that enables automatic or manual annotation within time or material location intervals, and synchronously expands annotation information across processes, achieving accurate mapping and efficient correlation between quality events and process data. This mechanism significantly improves the efficiency and accuracy of anomaly tracing and location, providing an effective means for rapid identification and root cause analysis of quality problems. Attached Figure Description

[0021] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart of an intelligent annotation and analysis method for steel production quality based on multi-process spatiotemporal mapping. Detailed Implementation

[0022] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0023] For ease of understanding, the following explanations are provided for the terms or concepts involved in this application: (1) Target processing position refers to the specific position of the material in the target process, used to identify the specific processing section of the material in a certain process.

[0024] (2) The target process processing time refers to the time interval during which the material is processed in a specific process, including the start time and end time of processing.

[0025] (3) The mapping model between the target processing position and the target process processing time refers to a mathematical model that establishes a one-to-one correspondence between spatial coordinates and time axis, used to describe the dynamic mapping relationship between the "position-time" of materials in the process. Through this model, the corresponding processing time of any process can be calculated based on any processing position, or the processing position can be determined based on the processing time of any process, thus achieving unified spatiotemporal calibration.

[0026] (4) The mapping model of corresponding processing positions across processes refers to the model that describes the correspondence between material processing positions in different processes, and is used to establish spatial continuity mapping in multi-process production flow. This model comprehensively considers material tracking number, process connection logic and specification change information to achieve accurate matching and positioning of the same physical part in each process.

[0027] (5) Material tracking data refers to structured data used to identify the identity of materials and their flow path between processes, including material identification information, time information, specification change information, etc. This data provides a spatiotemporal reference for establishing a mapping model.

[0028] (6) Material quality data refers to information data that records the changes in the state of materials during the production process, including defects, blockages, detection points, sampling points, abnormal events and human intervention information, which are used to assist in the analysis of the causes and effects of quality changes.

[0029] (7) Time series curves refer to the continuous recording curves of process parameters (such as temperature, pressure, current, etc.) changing over time. They reflect the dynamic characteristics of the production process and are the main data form for feature extraction and difference analysis.

[0030] (8) Statistical characteristics refer to the indicators extracted from time series data that describe the overall distribution and variation characteristics of the data, including mean, variance, extreme values, kurtosis, skewness, etc., which are used to quantitatively characterize the stability and volatility of the process.

[0031] (9) Frequency domain features refer to the periodicity and energy distribution features extracted from time-series signals through spectrum analysis. They are used to reflect the frequency structure and energy concentration of process fluctuations and are often obtained through Fast Fourier Transform (FFT) or wavelet analysis.

[0032] (10) Dynamic Time Warping (DTW) algorithm refers to an algorithm used to compare the similarity of two time-series curves. It achieves alignment and matching of time-series signals by nonlinearly stretching or compressing the time axis, so as to eliminate the influence of sampling time difference on waveform similarity calculation.

[0033] (11) Relative deviation refers to an index used to measure the difference in numerical characteristics. It reflects the degree of deviation between the target material and the reference material at the numerical level by calculating the relative change ratio between the characteristic values ​​of the target material and the reference material.

[0034] (12) Statistical divergence index refers to a statistical measure used to measure the degree of difference between two distributions, such as Kullback-Leibler divergence (KL divergence) or Jensen-Shannon divergence (JS divergence), which is used to quantify the inconsistency between the target feature distribution and the benchmark distribution.

[0035] (13) Standardized annotation refers to the process of uniformly encoding and normalizing time and space intervals based on the mapping model, so that the spatiotemporal information of different processes and different data sources can be compared, indexed and calculated in the same coordinate system.

[0036] (14) Reference material refers to the reference material selected during the production process that has stable quality and whose characteristic parameters meet the process standards. It is used as a comparison benchmark for subsequent difference analysis to identify abnormal deviations of the target material.

[0037] (15) Data visualization analysis graphics refer to the output that presents the analysis results in a graphical form, including distribution comparison charts, parameter heat maps, trend overlay charts, causal relationship charts, etc., which are used to intuitively show the differences in features and the patterns of quality deviations.

[0038] This application provides a method for intelligent labeling and analysis of steel production quality based on multi-process spatiotemporal mapping. This method acquires multi-process time-series curves, material tracking data, and material quality data to establish intra-process processing time-to-process location mapping models and cross-process location mapping models, achieving integrated alignment of time, location, and quality information. Based on this, target processing locations or time intervals are standardized and labeled, statistical and frequency domain features are extracted, and differences are calculated with benchmark materials. Differences are quantified using methods such as relative deviation, statistical divergence, and DTW. The results are then combined with visualization analysis to generate final quality diagnostic conclusions, enabling accurate traceability and rapid analysis of multi-process quality problems.

[0039] The following is combined with Figure 1 The method is explained in detail below: Step 1: This step involves acquiring time-series curves of characteristic parameters, material tracking data, and material quality data for each process stage of the target material's production. The time-series curves include data reflecting dynamic changes in the process, such as temperature, pressure, flow rate, current, and voltage. Material tracking data includes material identification information, time information, and material specification change information. Material quality data includes defect information, blockage information, detection point or sampling point information, and information on abnormal or intervention events. The technical principle of this step lies in achieving unified acquisition and fusion of multi-source data through a production process information system, ensuring temporal and spatial consistency of the data, and providing a foundation for subsequent spatiotemporal mapping.

[0040] Step Two: Based on material tracking data, establish a mapping model between target processing location, target process processing time, and corresponding processing location across processes.

[0041] The specific construction method includes the following three core parts: Construction of an intra-process time-location mapping model: This model aims to solve the correspondence between material processing time and its physical location on the production line within the same process. Its core technology is to use the overall production time of the material in the process and the material length to perform a linear conversion.

[0042] Formula and steps for construction:

[0043] Parameter description: P: Target processing position to be determined (unit: meters), calculated from the head of the material.

[0044] L: The complete length of the material in this process (unit: meters), from material tracking data.

[0045] T: The given processing time for the target operation (in seconds).

[0046] T_start: The start time of material production in this process (in seconds), derived from material tracking data.

[0047] T_end: The end time of material production in this process (in seconds), derived from material tracking data.

[0048] Model Application: Through this linear model, the system can accurately convert the time T of any process parameter event into a specific location P on the material, laying the foundation for unified labeling of time domain data and spatial domain data.

[0049] Unified sampling frequency processing for multiple process parameters: This processing aims to solve the problem of inconsistent sampling frequencies and inability to directly align data points for multiple process parameters within the same process due to different sensor configurations.

[0050] Sub-step 1: Obtain the number of original sampling points for the time-series curves of multiple process parameters (such as temperature, rolling force, speed, etc.) that need to be included in the analysis.

[0051] Sub-step 2: Calculate the least common multiple of the number of these sampling points, which will be used as the target number of data points after unified interpolation.

[0052] Sub-step 3: Based on the number of target data points, generate a new, uniformly distributed time series T_new within the total processing time interval [T_start, T_end] of the material.

[0053] Sub-step four: For each original time series curve, based on its original time-parameter value sequence (T_original, Y_original), use a suitable interpolation algorithm to calculate its parameter value Y_new on the new time series T_new.

[0054] Interpolation method selection: Linear interpolation: Simple to calculate and efficient, suitable for scenarios where data changes gradually and smoothness requirements are not high. Its formula is:

[0055] Among them, Y_a is the parameter value at time point T_a (the parameter value of the previous sampling point is known), and Y_b is the parameter value at time point T_b (the parameter value of the next sampling point is known).

[0056] Spline interpolation (such as cubic splines): can generate smooth interpolation curves, better preserve the continuous trend of the original data, and is suitable for data with strong fluctuations and requiring high precision to maintain waveform characteristics.

[0057] Polynomial interpolation: can also be considered for other specific needs.

[0058] The system can automatically select the most suitable interpolation method based on preset rules or user configuration, so as to maximize the restoration of the true variation law of process parameters while ensuring calculation efficiency.

[0059] Through the above interpolation process, it is ensured that all process parameters have values ​​at the same point in time, thus achieving strict synchronization of multi-source heterogeneous time-series data.

[0060] Construction of cross-process location mapping model: This model aims to address the issue of the positional correspondence of a piece of physical material in different processing steps after undergoing multiple processes. The modeling process needs to comprehensively consider material flow direction (flipping) and length changes.

[0061] The formula and steps are as follows: Head-to-tail alignment correction. First, query the "material flipping information" in the material tracking data. If the material flips head-to-tail when flowing from the upstream process to the downstream process, the position of the upstream process needs to be reversed before position mapping. If it does not flip, it remains unchanged.

[0062] Proportional mapping. After directional alignment, the position is proportionally converted based on the length change of the material between adjacent processes. The core principle is that the proportion of the same physical point on the material within the upstream process length L_up should be equal to its proportion within the downstream process length L_down. Therefore, the formula for calculating P_down is as follows:

[0063] Parameter description: P_up_corrected: The position of the upstream process after head-to-tail alignment correction.

[0064] P_down: The location of the downstream process obtained through mapping.

[0065] L_up and L_down: represent the complete length (in meters) of the material in the upstream and downstream processes, respectively, derived from material tracking data.

[0066] Through the construction and processing of the above three types of models, the system realizes the unification of spatiotemporal data within the process, the synchronization of multi-parameter data, and the continuous traceability of spatial data across processes, providing an accurate spatiotemporal benchmark for subsequent standardized labeling and full-process quality analysis.

[0067] Step 3: Based on the mapping model, the intervals of target processing location and target process processing time are standardized and labeled. The specific process is as follows: Annotation Information Input and Association: The system receives annotation instructions from the user. The user first selects the target quality information requiring in-depth analysis from the material status data, such as defect type, specific defect location, a blocked area, or a sampling point. Simultaneously, the user needs to specify one or more process parameters (such as temperature, rolling force, liquid level fluctuation, etc.) that they wish to associate with the analysis. This step clarifies the object and objective of the analysis.

[0068] Automatic mapping and preliminary annotation: Based on the input target quality information, the system automatically locates its corresponding physical location range (e.g., the start and end coordinates of the defect on the steel coil). Then, the system calls the mapping model built in step two to automatically map this quality location range to the corresponding processing location range or processing time range of the process specified by the user for the associated process parameters. This completes the preliminary automatic annotation of the time-series data of the associated process parameters.

[0069] Manual verification and adjustment: The system visualizes the results of automatic annotation to users through a graphical interface. Users can intuitively review the annotated ranges and perform manual verification to correct annotation errors or supplement missing annotations, ensuring the accuracy and completeness of the annotations.

[0070] Annotation Results Recording and Application: The system stores the final annotation results in a structured manner in the database, forming standardized annotation data. This establishes a precise and traceable link from "quality events" to "process data," which is directly used for subsequent feature extraction and quality analysis.

[0071] Step Four: Based on the labeled intervals, features of the corresponding time-series curves are extracted. These features include statistical features and frequency domain features. Statistical features include mean, variance, extreme values, kurtosis, and skewness, used to describe the overall trend and fluctuation amplitude of the process data; frequency domain features include periodic fluctuations and energy distribution, used to reflect the frequency composition and energy concentration of the signal. This step, based on signal processing and feature engineering principles, uses discretization calculations and Fast Fourier Transform (FFT) to transform the time-series signal into a structured feature vector, providing a foundation for subsequent quantitative comparison.

[0072] Step 5: Based on the characteristics of time-series curves, this study compares and analyzes the differences between the target material and a preset reference material at the same target processing location. For numerical characteristics, relative deviation is used to measure numerical differences; for distributional characteristics, statistical divergence indices (such as KL divergence and JS divergence) are used to calculate distributional differences; for waveform characteristics, dynamic time warping (DTW) algorithm is used to achieve time-series alignment, and similarity indices are calculated based on the alignment. The analysis results are output in visual formats such as distribution comparison charts, parameter heatmaps, trend overlay charts, and causal relationship diagrams. This step is based on the principles of multidimensional feature difference analysis and time series homology measurement, and uses quantitative calculations to identify differences between the target material and the reference material in terms of process stability and abnormal fluctuations.

[0073] Step Six: Based on the analysis results, diagnostic results are generated. The system integrates differential characteristic information and process status parameters to form conclusions on anomaly diagnosis, deviation cause identification, and causal analysis of quality impact on the target material. This step is based on the principles of intelligent diagnosis and causal analysis, using a correlation analysis model to identify the inherent logical relationships between different characteristics, achieving automated diagnosis from data discrepancies to quality problems, and providing a basis for decision-making in production process optimization and quality control.

[0074] The present invention also provides an intelligent labeling and analysis system for steel production quality based on multi-process spatiotemporal mapping. The intelligent labeling and analysis system for steel production quality based on multi-process spatiotemporal mapping can be implemented by executing the process steps of the intelligent labeling and analysis method for steel production quality based on multi-process spatiotemporal mapping. That is, those skilled in the art can understand the intelligent labeling and analysis method for steel production quality based on multi-process spatiotemporal mapping as a preferred embodiment of the intelligent labeling and analysis system for steel production quality based on multi-process spatiotemporal mapping.

[0075] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0076] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for intelligent labeling and analysis of steel production quality based on multi-process spatiotemporal mapping, characterized in that, include: Step 1: Obtain the time-series curves of the characteristic parameters of the target material in each process of production, material tracking data, and material quality data; Step 2: Based on material tracking data, establish a mapping model between the target processing location and the processing time of the target process, as well as the corresponding processing location across processes; Step 3: Based on the mapping model, standardize and label the intervals of the target processing location and the processing time of the target process; Step 4: Extract the features of the corresponding time series curves based on the labeled intervals; Step 5: Based on the characteristics of the time series curve, for the same target processing position, compare and analyze the differences between the target material and the preset benchmark material; Step 6: Generate diagnostic results based on the analysis results.

2. The method according to claim 1, characterized in that, The material tracking data includes material identification information, material specification change information, material flipping information, production start time, production end time, and material length; The material quality data includes defect information, blockage information, detection point or sampling point information, and information on abnormal or intervention events.

3. The method according to claim 2, characterized in that, Step two involves establishing a mapping model based on material tracking data, connecting the target processing location with the target process processing time and the corresponding processing location across processes. This includes: Construction of an intra-process time-location mapping model: Based on the material's production start time, production end time, and material length, a linear mapping relationship between processing time and processing location within the same process is established, expressed by the following formula: Where P represents the target processing position, L represents the material length, T represents the target process processing time, and T_start and T_end represent the production start time and production end time of the material in this process, respectively; Unified sampling frequency processing for multiple process parameters: For time series curves of multiple process parameters with different sampling frequencies in the same process, calculate the least common multiple of the number of sampling points of all parameters, and use this multiple as the target frequency to perform interpolation processing on the time series curves of different process parameters, so that all process parameters are unified to the same sampling points. Construction of cross-process position mapping model: Based on the flipping information, the materials in the upstream and downstream processes are aligned in the head and tail directions. Based on the specification change information of the materials between cross processes, the mapping relationship of the processing position across processes is established. The formula is expressed as: Where P_up_corrected is the upstream process position after head-to-tail alignment correction, P_down represents the corresponding downstream process position obtained by mapping, and L_up and L_down represent the complete length of the material in the upstream and downstream processes, respectively.

4. The method according to claim 2, characterized in that, Step three involves standardizing and labeling the intervals between the target processing location and the target process processing time based on the mapping model, including: The system receives user input, which includes selected target quality information and one or more process parameters that need to be correlated and analyzed; the target quality information includes one or more of the following: defect name, defect location, blockade location, and sampling point location. Based on the target quality information, the corresponding target processing position range is automatically extracted; Based on the established mapping model, the target processing location range is automatically mapped to the corresponding processing location range or processing time range of the process in which the associated process parameter is located, thus completing the initial automatic labeling of the time series data of the associated process parameter. The results of the automatic annotation are displayed through a graphical interface. For annotation errors and omissions, manual verification and adjustment are performed to obtain corrected and supplementary data. Record the final annotation results, establish the correspondence between the target quality information and the spatiotemporal intervals of the associated process parameters, and form standardized annotation data.

5. The method according to claim 1, characterized in that, The features include statistical features and frequency domain features. The statistical features include mean, variance, extreme values, kurtosis, and skewness. The frequency domain features include periodic fluctuations and energy distribution.

6. The method according to claim 1 or 5, characterized in that, Step five involves comparing and analyzing the differences between the target material and a preset benchmark material, including: For numerical features, the relative deviation calculation method is used to measure the numerical difference of feature values; for distribution features, the distribution difference is calculated based on the statistical divergence index; for waveform features, the dynamic time warping algorithm is used to align the feature curves in time, and the similarity index is calculated based on the alignment, thereby calculating the fluctuation difference. Based on the analysis results of the differences, data visualization analysis graphics are generated, including distribution comparison charts, parameter heatmaps, trend overlay charts, and causal relationship charts.

7. A smart labeling and analysis system for steel production quality based on multi-process spatiotemporal mapping, characterized in that, include: Module M1: Acquires time-series curves of characteristic parameters of the target material at each process in production, material tracking data, and material quality data; Module M2: Based on material tracking data, establish a mapping model between the target processing location and the processing time of the target process, as well as the corresponding processing location across processes; Module M3: Based on the mapping model, standardizes and labels the intervals of target processing location and target process processing time; Module M4: Extracts features of the corresponding time series curves based on labeled intervals; Module M5: Based on the characteristics of the time-series curve, it compares and analyzes the differences between the target material and the preset benchmark material for the same target processing position; Module M6: Generates diagnostic results based on the analysis results.

8. The system according to claim 7, characterized in that, The material tracking data includes material identification information, time information, and material specification change information; The material quality data includes defect information, blockage information, detection point or sampling point information, and information on abnormal or intervention events.

9. The system according to claim 8, characterized in that, The module M3 includes: Submodule M1: Receives user input, which includes selected target quality information and one or more process parameters that need to be correlated and analyzed; the target quality information includes one or more of the following: defect type, defect location, blockade location, and sampling point location; Submodule M2: Based on the target quality information, automatically extract the corresponding target processing position range; then, based on the established mapping model, automatically map the target processing position range to the corresponding processing position range or processing time range of the process in which the associated process parameter is located, and complete the initial automatic labeling of the time series data of the associated process parameter; Submodule M3: Displays the results of the automatic annotation through a graphical interface, and allows for manual verification and adjustment to correct annotation errors or supplement missing annotations; Submodule M4: Records the final annotation results, establishes the correspondence between the target quality information and the spatiotemporal intervals of the associated process parameters, and forms standardized annotation data for subsequent feature extraction and quality analysis.

10. The system according to claim 7, characterized in that, The features include statistical features and frequency domain features. The statistical features include mean, variance, extreme values, kurtosis, and skewness. The frequency domain features include periodic fluctuations and energy distribution.

Images