Continuous casting production process time series data segmentation method and device and computer equipment

Abstract

The invention relates to the technical field of data processing, and discloses a continuous casting production process time series data segmentation method and device and computer equipment, and the method comprises the steps: collecting time series data in the continuous casting production process, including the pulling speed, the stopper rod position, the crystallizer liquid level height and the like; and any kind of time sequence data collected in the preset data segmentation period is denoised to obtain denoised time sequence data. And grouping the reference de-noising time series data groups based on a preset unit time span, and determining a benchmarking group by calculating the dynamic time warping distance between each group and the reference de-noising time series data group corresponding to the plurality of preset key events. When the benchmarking group exists, the number covering the preset unit time span is determined according to the comparison of the preset duration of the key event corresponding to the benchmarking group and the preset unit time span, and then the actual data segment is determined and marked. By flexibly and reasonably segmenting the de-noising time sequence data according to different conditions, a more appropriate data unit can be provided for subsequent data analysis and processing.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and device for segmenting time series data of a continuous casting production process, and a computer device. Background Art

[0002] Continuous casting is the metallurgical process of solidifying liquid steel into billets through continuous cooling. As a core production process in the modern steel industry, it is a typical high-temperature process manufacturing field. As the key link between steelmaking and rolling, the stability of this process not only directly affects the geometric dimensional accuracy and internal metallurgical quality of the continuously cast billets, but also, through genetic effects, has a decisive influence on the yield rate, microstructure, and surface defects of the final steel. Therefore, establishing a real-time monitoring system and abnormality warning mechanism for the continuous casting process has important engineering practical value for improving the quality of steel products.

[0003] my country's steel industry is currently at a critical stage of transformation and upgrading towards intelligent manufacturing. With the advancement of digital infrastructure, including multi-source heterogeneous sensor networks, Industrial Internet of Things (IIoT) platforms, and Manufacturing Execution Systems (MES), the sampling frequency, dimensionality, and temporal continuity of process data available during production have been significantly improved. The data now covers multiple modalities, including process operating parameters, equipment operating status, and product quality indicators. However, while existing data collection systems have enabled the acquisition of massive amounts of industrial data, significant technical gaps remain in the deep mining and intelligent application of this data's value.

[0004] Currently, steel companies' production process optimization primarily relies on structured data sets provided by process control systems (PCS) and manufacturing execution systems, analyzed using traditional statistical process control (SPC) or shallow machine learning algorithms. Practice has proven that existing technical solutions have the following inherent flaws: Low data value density: For high-frequency sampling of multi-dimensional time-series data streams (including but not limited to key process variables such as casting speed, stopper opening, and mold vibration parameters), the existing system only implements basic trend display and static threshold alarm functions, failing to effectively establish a deep mapping relationship between process parameters and billet quality indicators; Insufficient model robustness: When traditional machine learning algorithms such as support vector machines (SVM) and random forests (RF) directly process raw time series signals, they suffer from low feature extraction efficiency, high noise sensitivity, and exponentially increasing computational complexity, making them unable to meet the millisecond-level real-time response requirements of industrial sites. Lack of process interpretability: Existing data analysis methods lack deep integration with continuous casting metallurgical mechanisms (including solidification heat transfer dynamics, molten steel flow control theory, etc.), making it difficult to convert analysis results into executable process adjustment strategies.

[0005] It is particularly important to note that key timing parameters such as casting speed, stopper position, and mold liquid level directly represent the stability of the continuous casting process. Specifically, unsteady fluctuations in casting speed are strongly correlated with the stability of the mold meniscus, potentially inducing defects such as mold slag entanglement and transverse surface cracks. Stopper positioning deviations can disrupt the mass flow balance from the ladle to the tundish, increasing the probability of centerline segregation and internal cracks in the strand.

[0006] However, the above-mentioned process time series data has typical characteristics such as high sampling frequency, significant non-stationary characteristics and low signal-to-noise ratio. Traditional time domain analysis methods have inherent limitations in feature extraction efficiency and engineering applicability, which seriously restricts the data-driven quality prediction and process optimization capabilities. Summary of the Invention

[0007] In view of this, the present application provides a method and device for segmenting time series data of a continuous casting production process, as well as computer equipment. By flexibly and reasonably segmenting denoised time series data according to different situations, more suitable data units can be provided for subsequent data analysis and processing.

[0008] According to one aspect of the present application, a method for segmenting time series data of a continuous casting production process is provided, the method comprising: Real-time collection of time series data generated during the continuous casting production process, wherein the time series data includes at least one of casting speed, stopper position, and mold liquid level; For any time series data collected within a preset data segmentation period, denoising the time series data to obtain denoised time series data; The denoised time series data are sequentially divided into multiple groups based on a preset unit time span. For any denoised time series data group, a benchmark denoised time series data group corresponding to the denoised time series data group is determined in the benchmark denoised time series data group based on the dynamic time warping distance between the benchmark denoised time series data group and the denoised time series data group corresponding to a plurality of preset key events in the continuous casting production process, wherein the preset key event includes at least one of nozzle replacement, pouring start, and pouring end, and the preset key event corresponds to a benchmark denoised time series data group and a preset duration; Determining the number of preset unit time spans that the denoised time series data set should actually cover based on a comparison between a preset duration of a preset key event corresponding to the benchmark denoised time series data set and a preset unit time span; Based on the position of the denoised time series data group in the denoised time series data and the number of preset unit time spans that should actually be covered, the actual denoised time series data segment corresponding to the denoised time series data group in the denoised time series data is determined, and the determined actual denoised time series data segment is marked as the preset key event corresponding to the benchmark denoised time series data group.

[0009] According to another aspect of the present application, a continuous casting production process time series data segmentation device is provided, the continuous casting production process time series data segmentation device comprising: a data acquisition module, configured to collect time series data generated during the continuous casting process in real time, wherein the time series data includes at least one of casting speed, stopper position, and mold liquid level; A data denoising module is used to perform denoising processing on any time series data collected within a preset data segmentation period to obtain denoised time series data; A data benchmarking module is used to divide the denoised time series data into multiple groups based on a preset unit time span, and for any denoised time series data group, determine a benchmark benchmark denoised time series data group corresponding to the denoised time series data group based on the dynamic time warping distance between the benchmark denoised time series data group and the denoised time series data group corresponding to multiple preset key events in the continuous casting production process, wherein the preset key event includes at least one of nozzle replacement, casting start and casting end, and the preset key event corresponds to a benchmark denoised time series data group and a preset duration; a segmentation strategy determination module, configured to determine the number of preset unit time spans that the denoised time series data set should actually cover based on a comparison between a preset duration of a preset key event corresponding to the benchmark denoised time series data set and a preset unit time span; A data segmentation module is used to determine the actual denoised time series data segment corresponding to the denoised time series data group in the denoised time series data based on the position of the denoised time series data group in the denoised time series data and the number of preset unit time spans that should actually be covered, and mark the determined actual denoised time series data segment as the preset key event corresponding to the benchmark denoised time series data group.

[0010] According to another aspect of the present application, a computer device is provided, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein when the processor executes the program, the above-mentioned method for segmenting time series data of a continuous casting production process is implemented.

[0011] By means of the above technical solution, the present application provides a method and device for segmenting time series data of a continuous casting production process, as well as computer equipment, which collect time series data of the continuous casting production process, covering casting speed, stopper rod position and crystallizer liquid level, etc. Any time series data collected within the preset data segmentation period is denoised to obtain denoised time series data. It is grouped based on the preset unit time span, and the benchmarking group is determined by calculating the dynamic time regularization distance between each group and the benchmark denoised time series data group corresponding to a plurality of preset key events. When there is a benchmarking group, the number of preset unit time spans is determined by comparing the preset duration of the corresponding key event with the preset unit time span, and then the actual data segment is determined and marked. By flexibly and reasonably segmenting the denoised time series data according to different situations, more suitable data units can be provided for subsequent data analysis and processing.

[0012] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A schematic flow chart of a method for segmenting time series data of a continuous casting production process provided by an embodiment of the present application is shown; Figure 2 A schematic diagram of a process of an adaptive wavelet denoising method provided in an embodiment of the present application is shown; Figure 3 The present invention provides a schematic structural diagram of a time series data segmentation device for a continuous casting production process provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0015] In this embodiment, a method for segmenting time series data of a continuous casting production process is provided. Figure 1 As shown, the method includes: Step 101: collecting time series data generated during the continuous casting production process in real time, wherein the time series data includes at least one of casting speed, stopper position, and mold liquid level.

[0016] In the above embodiments of the present application, time series data such as the casting speed, stopper position, and mold liquid level during the continuous casting production process are collected in real time, which can be specifically carried out by relying on the sensor network and Industrial Internet of Things (IIoT) platform widely deployed at the continuous casting production site.

[0017] Specifically, the sensor network acts like the "nerve sensor endings" of continuous casting production, distributed throughout key locations along the production line. For example, in the casting speed control area, a high-precision casting speed sensor is installed that accurately measures the casting speed of the continuous casting billet at a specific frequency (e.g., ten or a hundred times per second), converting the physical casting speed information into an electrical signal in real time. For the stopper rod position, a corresponding position sensor continuously monitors its elevation and descending status, accurately capturing subtle changes in its position. The mold level sensor constantly monitors fluctuations in the mold's liquid level, providing timely feedback on the liquid level.

[0018] The Industrial Internet of Things (IIoT) platform acts as a data aggregation hub. Various sensor data is quickly and reliably transmitted to the IIoT platform via wired or wireless communication methods, such as Industrial Ethernet and 5G wireless communications. This platform boasts powerful data reception and processing capabilities, enabling preliminary collation and verification of massive amounts of data from numerous sensors to ensure data accuracy and integrity.

[0019] In particular, the IIoT platform can adaptively segment raw time series data based on the cyclical fluctuations of continuous casting production. For example, within a complete continuous casting cycle, the casting speed, stopper position, and mold level vary uniquely at different stages. Based on these inherent patterns, the platform accurately segments continuous time series data into independently meaningful segments.

[0020] Finally, the processed and segmented data is stored in a structured database on the local computing device. This standardized data storage format facilitates efficient subsequent querying, analysis, and mining of the data, providing solid data support for monitoring, optimization, and fault diagnosis of the continuous casting production process.

[0021] Step 102 : For any time series data collected within a preset data segmentation period, denoising is performed on the time series data to obtain denoised time series data.

[0022] Next, for any time series data collected within a preset data segmentation period (e.g., 50 hours), the aforementioned time series data is denoised to obtain denoised time series data, in preparation for further processing of subsequent data.

[0023] Alternatively, as Figure 2 As shown, in step 102, the time series data is subjected to denoising processing to obtain denoised time series data, which specifically includes: Step 1021 : adaptively screening the optimal wavelet basis and optimal decomposition level corresponding to the time series data.

[0024] Step 1022 : Based on the selected optimal wavelet basis and optimal decomposition level, the time series data is subjected to wavelet decomposition, followed by threshold processing, and reconstructed signal denoising processing to obtain denoised time series data corresponding to the time series data.

[0025] In the above embodiments of the present application, the casting speed and stopper position data in the continuous casting process are usually subject to noise interference, such as equipment vibration, sensor error, or instability of molten steel flow. Traditional wavelet denoising methods rely on fixed wavelet bases (such as Daubechies-4) and decomposition levels (such as level=2), which cannot adapt to the noise characteristics of different castings or process stages, and may lead to insufficient denoising or signal distortion. To this end, the "adaptive wavelet denoising method" of the above embodiments of the present application, that is, by automatically selecting the optimal wavelet base and the optimal decomposition level, can optimize the denoising effect of the continuous casting process data. Its core steps are as follows: 1. Construct a set of candidate wavelet basis: Construct a set of multiple wavelet bases. This set of wavelet bases can be represented by W, such as W = {db4, sym5, coif2}. These wavelet bases exhibit good locality and smoothness in time series data processing, making them suitable for use in continuous casting processes. In W, db4, also known as the Daubechies 4 wavelet, is one of a series of wavelet bases with compact support and orthogonality proposed by renowned wavelet analysis scholar Ingrid Daubechies. The "4" represents specific parameter-related features, such as the order of its vanishing moment. sym5, also known as the Symlet 5 wavelet, is an improvement on the Daubechies wavelet system, exhibiting near-symmetry. The "5" also refers to the characteristic parameters of the wavelet. coif2, also known as the Coiflet 2 wavelet, is also known as the Coiflet 2 wavelet. The Coiflet wavelet has unique properties such as a higher number of vanishing moments, and "2" is used to distinguish the wavelet basis under different parameter settings in this wavelet system.

[0026] 2. Multi-level decomposition: For the input time series data, for each wavelet basis, the maximum decomposition level is determined with the help of a wavelet calculator, and wavelet decomposition is performed on each level to obtain approximate coefficients and detail coefficients.

[0027] 3. Adaptive threshold processing: Soft thresholding is applied to the detail coefficients. The threshold is calculated based on the signal length and noise characteristics, specifically the standard deviation of the detail coefficient multiplied by the logarithm of the signal length. The soft thresholding formula is: the absolute value of the detail coefficient is compared with the threshold, the larger value is taken, and then the sign is assigned based on whether the original detail coefficient is positive or negative.

[0028] 4. Denoising effect evaluation and optimal parameter selection: For each combination of wavelet basis and decomposition level, reconstruct the denoised signal. Calculate the residual variance between the denoised signal and the original signal. Select the combination with the smallest residual variance as the optimal wavelet basis and decomposition level, and return the corresponding denoised signal. If the signal length does not meet the requirements, return the original signal directly.

[0029] 5. Signal reconstruction and denoising: Based on the selected optimal wavelet basis and optimal decomposition level, the time series data is subjected to wavelet decomposition and threshold processing, and then the signal is reconstructed to obtain the denoised time series data (denoised time series data).

[0030] To this end, through the above steps, the optimal wavelet basis and decomposition level can be automatically selected according to the noise characteristics of the time series data in the continuous casting process, effectively removing noise while retaining the key features of the signal. Compared with the traditional wavelet denoising method with fixed parameters, it has higher robustness and adaptability, and can significantly improve the accuracy of subsequent analysis.

[0031] Step 103: Divide the denoised time series data into multiple groups based on a preset unit time span. For any denoised time series data group, determine a benchmark denoised time series data group corresponding to the denoised time series data group in the benchmark denoised time series data group based on the dynamic time warping distance between the benchmark denoised time series data group corresponding to a plurality of preset key events in the continuous casting production process and the denoised time series data group, wherein the preset key event includes at least one of nozzle replacement, pouring start and pouring end, and the preset key event corresponds to a benchmark denoised time series data group and a preset duration.

[0032] Next, during the continuous casting process, key events such as nozzle replacement, pouring start, and pouring end have specific impacts on time series data (such as casting speed, stopper position, and mold liquid level). By calculating the dynamic time warping distance between the denoised time series data set and the benchmark denoised time series data set corresponding to each preset key event, we can accurately find the most similar benchmark denoised time series data set, thereby precisely identifying the key event corresponding to the current data segment and helping to timely grasp important nodes in the production process.

[0033] Denoised time series data is grouped into preset unit time spans (e.g., one hour) and the most similar key event data groups (benchmarked against the baseline denoised time series data groups) are determined based on the dynamic time warping distance. This approach fully considers the duration of key events. By combining the preset duration of each key event, the actual coverage of the data grouping can be more reasonably determined, making the data segmentation more consistent with the key event process in actual production. This avoids the fragmentation or incompleteness of key event data that may result from fixed-length segmentation, thereby improving the quality and practicality of data segmentation.

[0034] By accurately identifying key events and properly segmenting data, we can more clearly observe and analyze the changes in time series data before and after each key event. For example, during a nozzle replacement event, we can analyze the changing patterns of data such as casting speed and stopper position in detail, identifying potential problems or optimization points. This provides more valuable information for monitoring and analyzing the production process, helping to promptly detect potential production anomalies and take appropriate measures.

[0035] Accurately identifying key events and rationally segmenting data allows for deeper analysis and evaluation of the continuous casting process. By comparing and studying time series data from different key events, we can summarize production experience and optimize process parameters and production plans. For example, based on data characteristics at the start and end of a pour, we can rationally schedule raw material supply and production rhythm, improve production efficiency and product quality, and reduce production costs.

[0036] The above embodiment of the present application can also be integrated into the automated control system of continuous casting production to achieve automatic identification of key events and automatic segmentation of data, thereby reducing manual intervention, improving the automation and intelligent level of production, making the production process more stable and reliable, and reducing the workload of operators.

[0037] In particular, regarding the preset duration of nozzle replacement, pouring start and pouring end, for example: For nozzle replacement: For small continuous casting machines, if the equipment is relatively small and operating space is limited, but the overall process is simple, the nozzle replacement operation can be relatively quick, with a preset duration of 10 or 20 minutes. For example, on some small continuous casting equipment used in experiments, workers can quickly complete nozzle removal, installation, and commissioning of the new nozzle after becoming familiar with the operating procedures.

[0038] For large continuous casting machines, which have complex structures, nozzle replacement may involve additional auxiliary operations, such as partial equipment shutdown and the implementation of safety precautions. The preset duration can be 30, 60 minutes, or even longer. For example, nozzle replacement on a large slab continuous casting machine requires strict adherence to operating procedures to ensure safe and stable operation, which can be time-consuming.

[0039] For pouring start: For single-strand continuous casting machines, the start of a pour primarily involves pre-casting preparations, such as docking the ladle, preheating the tundish, and pouring molten steel. If the equipment and processes are adequately prepared, the default duration can be 15 or 30 minutes. For example, some simple billet casting lines can quickly complete all operations required for the start of a pour after the molten steel arrives.

[0040] Multi-strand continuous casting machines, which have multiple strands simultaneously casting, require coordinated operation of the strands at the start of a casting run, ensuring molten steel supply and temperature control for each strand. This run can be preset for 30 or 60 minutes. Large multi-strand slab casters require comprehensive inspection and commissioning to ensure that each strand can start casting normally, which can be quite time-consuming.

[0041] For the end of pouring: At the end of a normal production run, if the casting is completed as planned, for a small continuous casting machine, it may only be necessary to process the remaining molten steel and perform a simple equipment cleaning and inspection. The preset duration can be 20 or 30 minutes. For example, in some small-scale alloy steel continuous casting operations, after completing the scheduled casting volume, the tail billet is quickly processed and the equipment is cleaned.

[0042] For abnormal casting terminations, such as those caused by equipment failure or molten steel quality issues, more detailed equipment inspection, troubleshooting, and quality analysis are required. For large continuous casting machines, the preset duration can be one or two hours, or even longer. For example, if a serious fault such as mold leakage occurs during continuous casting, a comprehensive inspection of the equipment damage and the development of a repair plan are required, significantly increasing the duration of the casting termination.

[0043] Optionally, in step 103, based on the dynamic time warping distance between the benchmark denoised time series data group corresponding to a plurality of preset key events in the continuous casting production process and the denoised time series data group, before determining the benchmark denoised time series data group corresponding to the denoised time series data group in the benchmark denoised time series data group, the method further includes: Step 106: Select a benchmark denoised time series data set corresponding to any preset key event, and calculate the dynamic time warping distance between the denoised time series data set and the selected benchmark denoised time series data set based on the dynamic time warping distance calculation formula, wherein the dynamic time warping distance calculation formula is: , Represents a measure of the denoised time series data set and the selected benchmark denoised time series data set The dynamic time warping distance of the similarity between To align the paths, Used to describe how to and Align and match the data points in until a and The alignment with the smallest difference, Indicates alignment path The data point pairs in , i is The data point index in , j is The index of the data point in , Used to measure Middle data points and Middle data points The degree of difference between Represents that by all possible alignment paths Search to find The smallest path, as and The dynamic time warping distance between them.

[0044] In the above embodiment of the present application, taking the key event "casting start" as an example, it is necessary to select the denoised time series data such as casting speed, stopper position or mold liquid level recorded at the start of a normal casting from the historical production data as the baseline denoised time series data set corresponding to "casting start". Assuming that casting speed data is selected as the analysis object, the baseline denoised time series data set is a series of casting speed values ​​recorded at the start of a normal casting in the past that have been denoised, for example: , in, is the number of reference denoised time series data in the reference denoised time series data group.

[0045] For example, when monitoring the continuous casting process in real time, we get a period of denoised casting speed time series data. , n is the number of casting speed time series data. If it is necessary to determine whether this data corresponds to the "pouring start" event, it is necessary to perform DTW (Dynamic Time Warping) calculation on it and the previously selected "pouring start" benchmark denoised time series data group.

[0046] Assumptions , .

[0047] First, list all possible alignment paths ,For example: , wait.

[0048] For each alignment path, calculate .by For example: , then calculate ,for , that is .

[0049] Perform the above calculations on all possible alignment paths and find The smallest path, the value corresponding to the path is .

[0050] To this end, the dynamic time warping distance between the denoised time series data set to be analyzed and the benchmark denoised time series data set can be obtained, thereby judging the similarity between the two and further identifying the key events in the continuous casting production process.

[0051] Step 104 : determining the number of preset unit time spans that the denoised time series data set should actually cover based on a comparison between the preset durations of the preset key events corresponding to the benchmark denoised time series data set and the preset unit time spans.

[0052] Next, in continuous casting production, key events such as nozzle replacement, start of a pouring cycle, and end of a pouring cycle all have specific durations. By comparing the preset duration with the preset unit time span to determine the coverage quantity, we can more accurately define the scope of key events in time series data. For example, a nozzle replacement event may involve a complex series of operational steps, and its actual duration may span multiple preset unit time spans. Accurately calculating the coverage quantity fully captures the period of time during which the event affects the time series data, avoiding incomplete analysis of event characteristics due to data truncation.

[0053] Preset unit time spans (e.g., 1 hour) are the basic unit for data segmentation, but the duration of key events often varies. Determining the actual coverage number based on comparative relationships can help data segmentation more closely align with the actual development of key events. For example, for a starting event, the changes in the time series data associated with it may unfold gradually over a specific time period. Properly determining the coverage number ensures that data segments containing the complete characteristics of the event are accurately segmented, improving the quality of data segmentation and providing a more reasonable data foundation for subsequent data analysis.

[0054] Accurately determining the number of preset unit time spans that should be included in a denoised time series data set allows for more precise analysis of the impact of key events on time series data. For example, during a nozzle change, the mold level may fluctuate due to factors such as changes in molten steel flow. A data segment that fully encompasses the period affected by the event can more accurately reflect the pattern of mold level changes, allowing for more accurate analysis of the relationship between events and data changes, providing a reliable basis for optimizing production processes.

[0055] Real-time monitoring of the occurrence and impact of key events is crucial in the continuous casting process. By determining the coverage quantity, the location and scope of key events in time series data can be promptly and accurately identified, allowing production monitors to more intuitively observe the changes in various parameters at the time of the event. For example, for the end-of-casting event, monitors can promptly identify abnormal data changes related to the end of the casting based on accurately divided data segments and take appropriate measures to ensure stable production operations.

[0056] Accurate data segmentation and key event analysis provide strong support for production decision-making. Understanding the duration and impact of key events allows production managers to rationally plan production and optimize resource allocation. For example, based on the number and impact period of nozzle replacement events, backup nozzles and related tools can be prepared in advance to reduce replacement time and improve production efficiency. Alternatively, the molten steel supply rhythm can be adjusted based on the start and end times of pouring operations to avoid production interruptions or resource waste.

[0057] Optionally, in step 104, the number of preset unit time spans that the denoised time series data set should actually cover is determined based on a comparison between a preset duration of a preset key event corresponding to the benchmark denoised time series data set and a preset unit time span, specifically including: Step 1041: If the preset duration of the preset key event corresponding to the determined benchmark denoised time series data set is less than or equal to the preset unit time span, then the number of preset unit time spans that the denoised time series data set should actually cover is 1; Step 1042: If the preset duration of the preset key event corresponding to the determined benchmark denoised time series data group is greater than the preset unit time span, the preset duration is converted into a multiple of the preset unit time span and rounded up to obtain the number of preset unit time spans that the denoised time series data group should actually cover.

[0058] In the above embodiments of the present application, there are two cases: Case 1: The preset duration is less than or equal to the preset unit time span: For example, the preset unit time span is set to 1 hour. In case 1, the preset duration of the preset key event corresponding to the determined benchmark denoised time series data group is less than or equal to 1 hour, for example, the preset duration is 30 minutes. Since the duration of the key event does not exceed the preset 1-hour unit time span, from the perspective of data segmentation and covering the complete impact of the key event, a data segment of the preset unit time span (1 hour) is sufficient to include the impact of the key event on the time series data. Therefore, at this time, the number of preset unit time spans that the aforementioned denoised time series data group should actually cover is 1. In other words, during data segmentation processing, based on the location of the current denoised time series data group, taking a data segment of 1 hour in length can fully capture the data features related to the key event.

[0059] Case 2: The preset duration is greater than the preset unit time span: Similarly, the preset unit time span is 1 hour. If the preset duration of the preset key event corresponding to the determined benchmark denoised time series data group is greater than 1 hour, for example, the preset duration is 1 hour and 20 minutes.

[0060] First, convert the preset duration of 1 hour and 20 minutes into multiples of 1 hour. After conversion, 1 hour and 20 minutes is approximately 1.33 times the preset unit time span. Then round up, because even if the remaining time is less than a complete preset unit time span, it is also part of the duration of the key event. In order to ensure that the data segment can fully cover the impact of the key event, it needs to be rounded up. 1.33 is rounded up to get 2. Therefore, the number of preset unit time spans that the aforementioned denoised time series data group should actually cover is 2. This means that when segmenting the data, a 2-hour data segment is required to fully include the impact of the key event on the time series data, so that accurate analysis and processing can be performed later.

[0061] Optionally, in step 104, based on the dynamic time warping distance between the benchmark denoised time series data groups corresponding to a plurality of preset key events in the continuous casting production process and the denoised time series data group, determining a benchmark denoised time series data group corresponding to the denoised time series data group in the benchmark denoised time series data group, specifically includes: Step 1043, if , then the selected benchmark denoised time series data group is the benchmark denoised time series data group that is most similar to the denoised time series data group, where For judgment and The preset similarity threshold of the similarity between is the preset threshold coefficient, is the set of dynamic time warping distances calculated historically, express The maximum value in .

[0062] In the above embodiment of the present application, if the dynamic time warping distance , then it is considered that there is a denoised time series data set The most similar benchmark denoised time series data set .

[0063] In particular, if the dynamic time warping distance Greater than or equal to the preset similarity threshold δ (i.e. ), it means that the current denoised time series data group With the benchmark denoised time series data set Not similar. In this case, you can continue to search for other possible benchmark data sets (benchmark benchmark denoised time series data sets) in the benchmark denoised time series data set set to determine whether there is a benchmark benchmark denoised time series data set that is more similar to the current denoised time series data set. Specifically, you can take the following steps: 1. Traverse all benchmark data groups (benchmark denoised time series data groups): Traverse all benchmark denoised time series data groups corresponding to various preset key events in the continuous casting production process.

[0064] 2. Calculate the dynamic time warping distance: Calculate the DTW distance between the current denoised time series data set and each benchmark denoised time series data set.

[0065] 3. Comparison and threshold: Compare the calculated DTW distance with the preset similarity threshold δ.

[0066] 4. Determine the benchmark: If there is any benchmark denoised time series data set whose DTW distance with the current denoised time series data set is less than δ, it is determined as the benchmark denoised time series data set; if the DTW distance of all benchmark data sets is greater than or equal to δ, it may be necessary to review the threshold setting or consider other factors to determine the benchmark (that is, re-determine δ).

[0067] To this end, when the DTW is greater than or equal to the preset similarity threshold δ, the search can continue until a more suitable benchmark denoised time series data set is found.

[0068] Step 105: Based on the position of the denoised time series data group in the denoised time series data and the number of preset unit time spans that should actually be covered, determine the actual denoised time series data segment corresponding to the denoised time series data group in the denoised time series data, and mark the determined actual denoised time series data segment as the preset key event corresponding to the benchmark denoised time series data group.

[0069] In the above embodiment of the present application, for example, in a continuous casting production process, the preset unit time span is 1 hour. The denoised time series data is casting speed data recorded in chronological order, with a total length of 5 hours. Each hour is used as a basic data grouping unit (preset unit time span), and a total of 5 basic data groups (denoised time series data groups) are divided.

[0070] Next, determine the position of the denoised time series data set. Suppose during the analysis process, it is determined that the third basic data set, that is, the denoised time series data set (i.e., the data set from the second to the third hour), is most similar to the baseline denoised time series data set corresponding to the preset key event of the water outlet replacement. Then, the position of this denoised time series data set in the denoised time series data set is the third position.

[0071] Next, determine the number of preset unit time spans that should actually be covered. For example, the preset duration of the preset key event of water outlet replacement is 1 hour and 20 minutes. Since the preset duration is longer than the preset unit time span (1 hour), the preset duration is converted into a multiple of the preset unit time span and rounded up. 1 hour and 20 minutes is approximately 1.33 times 1 hour after conversion. Rounding up to get the number of preset unit time spans that should actually be covered is 2. According to the position of the denoised time series data group (the third position) and the number of preset unit time spans that should actually be covered (2), because 2 preset unit time spans are to be covered and the data group starts at the second hour, the actual denoised time series data segment starts from the second hour and covers the data from the second hour to the fourth hour (the second to third hours are the original data group, plus the next adjacent 1 hour data to meet the requirement of covering 2 unit time spans).

[0072] Next, we mark the preset key events, labeling the actual denoised time series data segment from the second to the fourth hour as the preset key event of nozzle replacement. This allows us to clearly identify the data segment corresponding to the nozzle replacement event in subsequent analysis of continuous casting production time series data, and further analyze its impact on data such as casting speed, providing a basis for monitoring and optimizing the production process.

[0073] Through the above steps, the actual data segments corresponding to key events (actual denoised time series data segments) can be accurately determined from the denoised time series data and marked, which helps to deeply analyze the relationship between various events and data changes in the continuous casting production process.

[0074] Optionally, the continuous casting production process time series data segmentation method further includes: Step 107 : When no benchmark denoised time series data group is determined, the divided denoised time series data group is directly used as the actual denoised time series data segment.

[0075] In the above embodiment of the present application, in the continuous casting production process, the time series data is continuous and has an internal logical relationship. When no benchmark denoised time series data group is determined, the divided denoised time series data group is directly used as the actual denoised time series data segment, avoiding data truncation or segmentation that may be caused by forced matching of key events. For example, the change of casting speed data in different time periods is continuous. If the data group is divided in order to match a certain key event, the change trend and integrity of the data may be destroyed. Directly using the divided data group can completely retain the characteristics of the data in a specific time period, providing a more realistic and reliable basis for subsequent data analysis.

[0076] If the actual data segments are determined by matching key events without a benchmark denoised time series data set, the complexity and computational effort of data processing will increase. Directly using the divided data sets as the actual data segments (actual denoised time series data segments) reduces additional computational steps and judgment logic, making the data processing process more concise and efficient. For example, there is no need to calculate and compare dynamic time warping distances to determine similarity with key events, saving computing resources and time costs.

[0077] Most time series data in continuous casting production may not be directly related to pre-defined key events. Using these segmented data groups directly as actual data segments facilitates conventional data analysis, such as statistical mean, variance, and trend analysis. For example, using segmented data for mold level data directly allows for easier analysis of fluctuations over different time periods, identifying anomalies or patterns in the data without considering the impact of key events, allowing analysis to focus more on the data's inherent characteristics.

[0078] The continuous casting production process is complex and diverse, potentially subject to a variety of unforeseen circumstances and events. Using segmented data groups directly as actual data segments (actual denoised time series data segments) allows for adapting to this diverse production situation. Even if no predefined key events are matched, data can be effectively segmented and analyzed, enabling timely identification of potential problems or changes in the production process. For example, if temporary, unforeseen fluctuations or anomalies occur during production, directly using segmented data segments can better capture this information, providing a reference for production adjustments and optimization.

[0079] Directly determining the actual data segments based on the segmentation method better maintains the data's chronological order. This is crucial for analyzing the temporal relationships of data during continuous casting, such as the relationship between casting speed changes over time and product quality. Disrupting the chronological order of data to match key events can lead to biased analysis results. Directly using segmented data ensures the data's temporal integrity, making analysis more accurate and reliable.

[0080] Optionally, the continuous casting production process time series data segmentation method further includes: Step 108 : performing dimension reduction and symbolic mapping processing on the denoised time series data in the actual denoised time series data segment to obtain character string data converted based on the actual denoised time series data segment.

[0081] In the above-described embodiments of the present application, the denoised time series data in the actual denoised time series data segment typically contains a large number of data points, occupying a large amount of storage resources. By mapping the high-dimensional time series data into a low-dimensional discrete symbol space through dimensionality reduction processing, the amount of data can be significantly reduced. For example, an actual denoised time series data segment that originally required the storage of tens of thousands of data points may only require the storage of dozens or even fewer symbols after processing, significantly saving storage space and reducing storage costs.

[0082] Optionally, in step 108, dimensionality reduction and symbolic mapping processing are performed on the denoised time series data in the actual denoised time series data segment, specifically including: Step 1081 , performing dimensionality reduction processing on the denoised time series data in the actual denoised time series data segment using a segmented aggregation approximation method until the denoised time series data in the actual denoised time series data segment is mapped to a discrete symbol space.

[0083] In the above embodiment of the present application, for example, in the continuous casting production process, an actual denoised time series data segment is determined through the previous steps. The data segment records the casting speed data within a certain time period, which is 3 hours long. The data acquisition frequency can be once per minute. Then the number of data points contained in this data segment is 3×60=180. Suppose the sequence composed of these data points is .

[0084] First, decide how many segments you want to divide the data into. For example, you might decide to divide the 180 data points into 18 segments, with each segment containing 180 ÷ 18 = 10 data points.

[0085] For each paragraph, calculate the mean of the data points. Through calculation, we get the mean sequence of 18 paragraphs , thus completing the initial dimensionality reduction, reducing the 180-dimensional data to 18 dimensions.

[0086] Next, set the symbol set. Pre-set a discrete symbol set and determine the mapping rule based on the data distribution. For example, divide the data range into 4 intervals, each interval corresponds to a symbol. According to the above mapping rule, map each element in the mean sequence Y to the corresponding symbol, and finally obtain a string data composed of symbols. For example, the string obtained after mapping is " ", this string is the string data converted based on the actual denoised time series data segment.

[0087] Through the above steps, the PAA (Piecewise Aggregate Approximation) method was used to reduce the dimension of the pulling speed data in the actual denoised time series data segment and map it to a discrete symbol space, obtaining string data that is convenient for subsequent analysis and processing.

[0088] By applying the technical solution of this embodiment, the complete processing flow for time series data from the continuous casting process is covered, from data acquisition, noise removal, key event identification, to data segmentation, forming an organic whole. Through meticulous data processing and analysis, anomalies in the production process can be promptly detected, production process parameters can be optimized, product quality and production efficiency can be improved, production costs can be reduced, and strong support can be provided for the stable operation and continuous improvement of continuous casting production.

[0089] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a continuous casting production process time series data segmentation device, such as Figure 3 As shown, the device includes: The data acquisition module 201 is used to collect time series data generated during the continuous casting production process in real time, wherein the time series data includes at least one of the casting speed, the stopper position and the crystallizer liquid level; The data denoising module 202 is configured to perform denoising processing on any time series data collected within a preset data segmentation period to obtain denoised time series data; The data benchmarking module 203 is configured to sequentially divide the denoised time series data into multiple groups based on a preset unit time span, and for any denoised time series data group, determine a benchmarked benchmark denoised time series data group corresponding to the denoised time series data group based on the dynamic time warping distance between the benchmark denoised time series data group and the denoised time series data group corresponding to a plurality of preset key events in the continuous casting production process, wherein the preset key events include at least one of nozzle replacement, pouring start, and pouring end, and the preset key events correspond to the benchmark denoised time series data group and a preset duration; A segmentation strategy determination module 204 is configured to determine the number of preset unit time spans that the denoised time series data set should actually cover based on a comparison between a preset duration of a preset key event corresponding to the benchmark denoised time series data set and a preset unit time span; The data segmentation module 205 is used to determine the actual denoised time series data segment corresponding to the denoised time series data group in the denoised time series data based on the position of the denoised time series data group in the denoised time series data and the number of preset unit time spans that should actually be covered, and mark the determined actual denoised time series data segment as the preset key event corresponding to the benchmark denoised time series data group.

[0090] It should be noted that for other corresponding descriptions of the functional units involved in the time series data segmentation device for continuous casting production process provided in the embodiment of the present application, please refer to Figures 1 to 2 The corresponding description in the method will not be repeated here.

[0091] Based on the above Figures 1 to 2 The method shown, and Figure 3 In order to achieve the above-mentioned purpose, the embodiment of the present application further provides a computer device, which can be a personal computer, a server, a network device, etc. The computer device includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to achieve the above-mentioned Figures 1 to 2 The segmentation method of time series data of continuous casting production process is shown.

[0092] Optionally, the computer device may also include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a Wi-Fi module, etc. The user interface may include a display, an input unit such as a keyboard, etc. Optional user interfaces may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a Bluetooth interface, a Wi-Fi interface), etc.

[0093] Those skilled in the art will understand that the computer device structure provided in this embodiment does not constitute a limitation on the computer device, and may include more or fewer components, or a combination of certain components, or different component arrangements.

[0094] The storage medium may also include an operating system and a network communication module. An operating system is a program that manages and stores the hardware and software resources of a computer device, supporting the execution of information processing programs and other software and / or programs. The network communication module facilitates communication between components within the storage medium, as well as with other hardware and software within the physical device.

[0095] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by means of software plus the necessary general hardware platform, or by hardware to collect time series data in the continuous casting production process, covering casting speed, stopper rod position and crystallizer liquid level, etc. Any time series data collected within the preset data segmentation period is denoised to obtain denoised time series data. It is grouped based on the preset unit time span, and the benchmarking group is determined by calculating the dynamic time regularization distance of each group with the benchmark denoised time series data group corresponding to a variety of preset key events. When there is a benchmarking group, the number of preset unit time spans is determined by comparing the preset duration of the corresponding key event with the preset unit time span, and then the actual data segment is determined and marked. By flexibly and reasonably segmenting the denoised time series data according to different situations, more suitable data units can be provided for subsequent data analysis and processing.

[0096] Those skilled in the art will understand that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required to implement the present application. Those skilled in the art will understand that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more devices different from the implementation scenario. The modules of the above-mentioned implementation scenario can be combined into one module, or can be further split into multiple sub-modules.

[0097] The serial numbers of the above application are for descriptive purposes only and do not represent the advantages or disadvantages of the implementation scenarios. The above disclosures are only a few specific implementation scenarios of the present application, but the present application is not limited thereto, and any changes that can be made by those skilled in the art should fall within the scope of protection of the present application.

Claims

1. A method for segmenting time series data of a continuous casting production process, characterized in that: The continuous casting production process time series data segmentation method includes: Real-time collection of time series data generated during the continuous casting production process, wherein the time series data includes at least one of casting speed, stopper position, and mold liquid level; For any time series data collected within a preset data segmentation period, denoising the time series data to obtain denoised time series data; The denoised time series data are sequentially divided into multiple groups based on a preset unit time span. For any denoised time series data group, a benchmark denoised time series data group corresponding to the denoised time series data group is determined in the benchmark denoised time series data group based on the dynamic time warping distance between the benchmark denoised time series data group and the denoised time series data group corresponding to a plurality of preset key events in the continuous casting production process, wherein the preset key event includes at least one of nozzle replacement, pouring start, and pouring end, and the preset key event corresponds to a benchmark denoised time series data group and a preset duration; Determining the number of preset unit time spans that the denoised time series data set should actually cover based on a comparison between a preset duration of a preset key event corresponding to the benchmark denoised time series data set and a preset unit time span; Based on the position of the denoised time series data group in the denoised time series data and the number of preset unit time spans that should actually be covered, the actual denoised time series data segment corresponding to the denoised time series data group in the denoised time series data is determined, and the determined actual denoised time series data segment is marked as the preset key event corresponding to the benchmark denoised time series data group.

2. The method for segmenting time series data of a continuous casting production process according to claim 1, characterized in that: The step of determining the number of preset unit time spans that the denoised time series data set should actually cover based on a comparison between a preset duration of a preset key event corresponding to the benchmark denoised time series data set and a preset unit time span comprises: If the preset duration of the preset key event corresponding to the determined benchmark denoised time series data set is less than or equal to the preset unit time span, then the number of preset unit time spans that the denoised time series data set should actually cover is 1; If the preset duration of the preset key event corresponding to the determined benchmark denoised time series data group is greater than the preset unit time span, the preset duration is converted into a multiple of the preset unit time span and rounded up to obtain the number of preset unit time spans that the denoised time series data group should actually cover.

3. The method for segmenting time series data of a continuous casting production process according to claim 1, characterized in that: The continuous casting production process time series data segmentation method further includes: When no benchmark denoised time series data group is determined, the divided denoised time series data group is directly used as the actual denoised time series data segment.

4. The method for segmenting time series data of a continuous casting production process according to claim 1, characterized in that: The denoising process is performed on the time series data to obtain denoised time series data, comprising: Adaptively screening the optimal wavelet basis and optimal decomposition level corresponding to the time series data; Based on the selected optimal wavelet basis and optimal decomposition level, the time series data is subjected to wavelet decomposition and then threshold processing, and then signal denoising processing is reconstructed to obtain denoised time series data corresponding to the time series data.

5. The method for segmenting time series data of a continuous casting production process according to claim 1, characterized in that: Based on the dynamic time warping distance between the benchmark denoised time series data group corresponding to a plurality of preset key events in the continuous casting production process and the denoised time series data group, before determining the benchmark denoised time series data group corresponding to the denoised time series data group in the benchmark denoised time series data group, the method for segmenting time series data in the continuous casting production process further includes: Select any one of the preset key events corresponding to the benchmark denoised time series data set, and calculate the dynamic time warping distance between the denoised time series data set and the selected benchmark denoised time series data set based on the dynamic time warping distance calculation formula, wherein the dynamic time warping distance calculation formula is: ， Represents a measure of the denoised time series data set and the selected benchmark denoised time series data set The dynamic time warping distance of the similarity between To align the paths, Used to describe how to and Align and match the data points in until a and The alignment with the smallest difference, Indicates alignment path The data point pairs in , i is The data point index in , j is The index of the data point in , Used to measure Middle data points and Middle data points The degree of difference between Represents that by all possible alignment paths Search to find The smallest path, as and The dynamic time warping distance between them.

6. The method for segmenting time series data of a continuous casting production process according to claim 5, characterized in that: The method of determining a benchmark denoised time series data group corresponding to the denoised time series data group in the benchmark denoised time series data group based on the dynamic time warping distance between the benchmark denoised time series data group and the denoised time series data group based on a plurality of preset key events in the continuous casting production process, comprises: like , then the selected benchmark denoised time series data group is the benchmark denoised time series data group that is most similar to the denoised time series data group, where For judgment and The preset similarity threshold of the similarity between is the preset threshold coefficient, is the set of dynamic time warping distances calculated historically, express The maximum value in .

7. The method for segmenting time series data of a continuous casting production process according to any one of claims 1 to 6, characterized in that: The continuous casting production process time series data segmentation method further includes: The denoised time series data in the actual denoised time series data segment is subjected to dimensionality reduction and symbolic mapping processing to obtain character string data converted based on the actual denoised time series data segment.

8. The method for segmenting time series data of a continuous casting production process according to claim 7, characterized in that: The dimensionality reduction and symbolic mapping processing of the denoised time series data in the actual denoised time series data segment includes: The denoised time series data in the actual denoised time series data segment is subjected to dimensionality reduction processing using a segmented aggregation approximation method until the denoised time series data in the actual denoised time series data segment is mapped to a discrete symbol space.

9. A time series data segmentation device for a continuous casting production process, characterized in that: The continuous casting production process time series data segmentation device includes: a data acquisition module, configured to collect time series data generated during the continuous casting process in real time, wherein the time series data includes at least one of casting speed, stopper position, and mold liquid level; A data denoising module is used to perform denoising processing on any time series data collected within a preset data segmentation period to obtain denoised time series data; A data benchmarking module is used to divide the denoised time series data into multiple groups based on a preset unit time span, and for any denoised time series data group, determine a benchmark benchmark denoised time series data group corresponding to the denoised time series data group based on the dynamic time warping distance between the benchmark denoised time series data group and the denoised time series data group corresponding to multiple preset key events in the continuous casting production process, wherein the preset key event includes at least one of nozzle replacement, casting start and casting end, and the preset key event corresponds to a benchmark denoised time series data group and a preset duration; a segmentation strategy determination module, configured to determine the number of preset unit time spans that the denoised time series data set should actually cover based on a comparison between a preset duration of a preset key event corresponding to the benchmark denoised time series data set and a preset unit time span; A data segmentation module is used to determine the actual denoised time series data segment corresponding to the denoised time series data group in the denoised time series data based on the position of the denoised time series data group in the denoised time series data and the number of preset unit time spans that should actually be covered, and mark the determined actual denoised time series data segment as the preset key event corresponding to the benchmark denoised time series data group.

10. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, wherein: When the processor executes the computer program, the method for segmenting time series data of a continuous casting production process according to any one of claims 1 to 8 is implemented.

Images