Self-adaptive data extraction method and device based on multi-level dynamic sampling

Abstract

The invention provides a self-adaptive data extraction method and device based on multi-level dynamic sampling, and relates to the technical field of data extraction analysis, and the key points of the technical scheme are that the method comprises the following steps: obtaining operation parameters and system resource use state information of a plurality of blast furnaces; the blast furnace monitoring parameters are divided into a high-frequency sampling layer, a conventional sampling layer and a low-frequency sampling layer according to the operation parameters; calculating a blast furnace abnormal state index and a resource utilization rate based on the operation parameters; executing cross-furnace data association analysis, and calculating a cross-furnace association degree; dynamically adjusting the sampling frequency of each sampling layer according to the abnormal state index, the resource utilization rate and the cross-furnace correlation degree; based on the resource use state information and the data importance, carrying out adaptive compression processing on the collected data; and converting the compressed heterogeneous data into a uniform data format. The adaptive data extraction method and device based on multi-level dynamic sampling provided by the invention have the advantage of improving the extraction efficiency and quality.

Description

Technical Field

[0001] The present application relates to the technical field of data extraction and analysis, and in particular to an adaptive data extraction method and device based on multi-level dynamic sampling. Background Art

[0002] A large steel plant is upgrading the health management of its blast furnace fleet. The fleet consists of multiple blast furnaces of different sizes and ages, which share some raw material supply and fuel systems. In order to achieve precise health management, the plant has installed a large number of sensors on each blast furnace to monitor key parameters such as temperature, pressure, and gas composition. However, due to the complexity of the blast furnace fleet, there are subtle interactions between different blast furnaces. For example, changes in the iron output of one blast furnace may affect the raw material distribution, which in turn affects the operating status of other blast furnaces.

[0003] Each blast furnace in a fleet may be in different stages of operation, such as overhaul, capacity increase, or normal production. This dynamically changing environment requires a data extraction system that can flexibly respond to the specific needs of different blast furnaces. Especially under extreme weather conditions, such as high temperatures in summer or severe cold in winter, the operating parameters of the entire blast furnace fleet will change significantly, requiring more sophisticated data collection and analysis.

[0004] What is more challenging is that different generations of blast furnaces in the blast furnace group use different generations of automation systems and sensor networks, and the data formats and transmission protocols are different. In this case, the factory hopes to implement an intelligent data extraction method that can not only dynamically adjust the data collection strategy according to the current status and surrounding environment of each blast furnace, but also coordinate the data collection of the entire blast furnace group to capture the mutual influence between blast furnaces.

[0005] However, the existing data acquisition system has the following problems:

[0006] The data collection frequency is fixed and cannot be adjusted dynamically according to the blast furnace status and system resources, resulting in a lack of sufficient data support at critical moments and a large amount of redundant data during non-critical periods.

[0007] The data formats of different blast furnaces are not unified, which increases the difficulty of data integration and analysis.

[0008] The lack of consideration of the mutual impact between blast furnaces makes it impossible to effectively capture and analyze the overall operating status of the furnace group.

[0009] The data compression method is fixed and cannot be adaptively adjusted according to data importance and system resource usage, which affects storage efficiency and data quality.

[0010] The lack of the ability to analyze long-term operating trends makes it difficult to support predictive maintenance and optimization decisions.

[0011] These problems seriously restrict the coordinated health management and optimized operation of the blast furnace group. For example, at certain critical moments, such as when the blast furnace experiences abnormal fluctuations, the fixed data collection frequency may not capture sufficiently detailed information, resulting in insufficient analysis of abnormal conditions. On the other hand, when the blast furnace is operating stably, too high a collection frequency will generate a large amount of redundant data, wasting storage resources and increasing the burden of data processing.

[0012] In addition, the lack of consideration of the mutual impact between blast furnaces makes it difficult for existing systems to identify and analyze the chain reactions caused by changes in one blast furnace. This can lead to missing important optimization opportunities or failing to consider the global impact when dealing with abnormal situations.

[0013] The problem of inconsistent data formats further exacerbates the difficulty of data integration and analysis. Data in different formats require additional processing steps for unified analysis, which not only increases the complexity of the system, but may also introduce errors in the data conversion process.

[0014] Fixed data compression methods also have limitations. When system resources are tight, important data may not be compressed effectively, affecting data quality; when resources are sufficient, non-critical data may be over-compressed, resulting in the loss of potentially valuable information.

[0015] Finally, the lack of long-term trend analysis capabilities means that the system cannot provide adequate support for predictive maintenance and long-term optimization decisions. This can result in missing early signs of equipment degradation or failing to identify long-term efficiency improvement opportunities.

[0016] In view of the above problems, the existing technology needs to be improved urgently. Summary of the invention

[0017] The purpose of this application is to provide an adaptive data extraction method and device based on multi-level dynamic sampling, which has the advantages of improving data extraction efficiency and quality.

[0018] The present application provides an adaptive data extraction method based on multi-level dynamic sampling, and the technical solution is as follows:

[0019] The method comprises the following steps: obtaining the operating parameters and system resource usage status information of a plurality of blast furnaces; dividing the blast furnace monitoring parameters into a high-frequency sampling layer, a conventional sampling layer and a low-frequency sampling layer according to the operating parameters; calculating the blast furnace abnormal state index and resource usage rate based on the operating parameters; performing cross-furnace data association analysis and calculating the cross-furnace correlation degree; dynamically adjusting the sampling frequency of each sampling layer according to the abnormal state index, resource usage rate and cross-furnace correlation degree; adaptively compressing the collected data based on the resource usage status information and data importance; and converting the compressed heterogeneous data into a unified data format.

[0020] Furthermore, the present application also proposes that the step of dynamically adjusting the sampling frequency of each sampling layer according to the abnormal state index, resource utilization rate and cross-furnace correlation includes: obtaining a preset baseline sampling frequency and time decay coefficient; predicting key time points based on historical data and current trends; calculating the time sensitivity function according to the predicted key time points and time decay coefficient; calculating the required sampling frequencies of the high-frequency sampling layer, conventional sampling layer and low-frequency sampling layer according to the abnormal state index, resource utilization rate, cross-furnace correlation, baseline sampling frequency and time sensitivity function; and adjusting the sampling frequencies of the high-frequency sampling layer, conventional sampling layer and low-frequency sampling layer respectively according to the required sampling frequencies.

[0021] Furthermore, the present application also proposes that the step of predicting key time points based on historical data and current trends includes: obtaining a preset time period and a parameter fluctuation threshold; obtaining historical data of blast furnace operating parameters within the preset time period; establishing a time series model based on the historical data; calculating the changing trend of blast furnace operating parameters based on the time series model; identifying the time interval in which parameter fluctuations exceed the parameter fluctuation threshold based on the changing trend and the parameter fluctuation threshold; and using the starting time point of the time interval as the key time point for prediction.

[0022] Furthermore, the present application also proposes that the step of calculating the required sampling frequencies of the high-frequency sampling layer, the conventional sampling layer and the low-frequency sampling layer according to the abnormal state index, resource utilization rate, cross-furnace correlation, baseline sampling frequency and time sensitivity function includes: obtaining a preset sampling frequency adjustment coefficient and a sampling layer frequency ratio; calculating the frequency adjustment factor according to the abnormal state index, resource utilization rate and cross-furnace correlation; calculating the required sampling frequency of the high-frequency sampling layer according to the baseline sampling frequency, frequency adjustment factor and time sensitivity function; calculating the required sampling frequencies of the conventional sampling layer and the low-frequency sampling layer according to the sampling layer frequency ratio and the required sampling frequency of the high-frequency sampling layer.

[0023] Furthermore, the present application also proposes that the step of identifying the time interval in which the parameter fluctuation exceeds the parameter fluctuation threshold according to the change trend and the parameter fluctuation threshold includes: obtaining a preset statistical time window; within the statistical time window, counting the maximum and minimum values ​​of the blast furnace operation parameters to determine the historical fluctuation range; calculating the dynamic parameter fluctuation threshold according to the historical fluctuation range and a preset fluctuation coefficient; comparing the change trend with the dynamic parameter fluctuation threshold; when the change trend exceeds the dynamic parameter fluctuation threshold, recording the corresponding time point; and determining the continuous time interval in which the parameter fluctuation exceeds the threshold according to the recorded time point sequence.

[0024] Furthermore, the present application also proposes that the step of adaptively compressing the collected data based on the resource usage status information and the data importance includes: obtaining a preset compression rate upper limit, compression rate lower limit and data importance evaluation standard; calculating the current system resource utilization based on the resource usage status information; performing importance evaluation on the collected data based on the data importance evaluation standard to obtain a data importance index; calculating a target compression rate based on the data importance index and the current system resource utilization; comparing the target compression rate with the compression rate upper limit and compression rate lower limit to obtain an actual compression rate; and compressing the collected data based on the actual compression rate to obtain compressed data.

[0025] Furthermore, the present application also proposes that the method also includes: deploying edge computing nodes in each blast furnace, and the edge computing nodes are used to perform data preprocessing, anomaly detection and preliminary analysis; using the edge computing nodes to perform real-time compression and feature extraction on the collected data; regularly calculating data integrity indicators and information entropy, evaluating the quality of the collected data to obtain data quality assessment results; and dynamically adjusting the sampling frequency and compression rate of each sampling layer based on the data quality assessment results.

[0026] Furthermore, the present application also proposes that the method also includes: obtaining historical data and current trend information of blast furnace operation; based on the historical data and current trend information, using a time series analysis model to predict future key time points; according to the future key time points, adjusting the sampling frequencies of relevant parameters in the high-frequency sampling layer, the conventional sampling layer and the low-frequency sampling layer; the step of adjusting the sampling frequencies of relevant parameters in the high-frequency sampling layer, the conventional sampling layer and the low-frequency sampling layer according to the future key time points includes: obtaining a preset sampling frequency adjustment rule and a parameter set corresponding to the abnormal type; extracting time information and a predicted abnormal type from the future key time point; determining a time window in which the sampling frequency needs to be adjusted according to the time information; selecting relevant parameters from the high-frequency sampling layer, the conventional sampling layer and the low-frequency sampling layer based on the abnormal type and the parameter set corresponding to the abnormal type; calculating the target sampling frequency of the selected parameter according to the sampling frequency adjustment rule, in combination with the abnormal type and the time window; generating a sampling adjustment instruction including a parameter identifier, a time window and a target sampling frequency; and sending the sampling adjustment instruction to the corresponding data acquisition device to achieve dynamic adjustment of the sampling frequency of the relevant parameters.

[0027] Furthermore, the present application also proposes that it also includes: obtaining the baseline sampling frequency, abnormal state index, resource utilization rate and cross-furnace correlation of each blast furnace; obtaining the predicted key time point and data change rate; calculating the dynamic sampling frequency based on the following mathematical model: F(t) = base_f*[1+α*A(t)+β*R(t)+γ*C(t)]*exp(-λ*|t-tp|) wherein F(t) is the required sampling frequency at time t, base_f is the baseline sampling frequency, A(t) is the abnormal state index, R(t) is the resource utilization rate, C(t) is the cross-furnace correlation, α, β, γ are weight coefficients and α+β+γ=1, tp is the predicted key time point, λ is the time attenuation coefficient; based on data importance evaluation, calculate the data value V(t) and data quality evaluation function: Q(t) = 1-exp(-μ*D(t)) wherein D(t) is the number According to the change rate, μ is the quality sensitivity coefficient; according to the data value and data quality assessment results, calculate the adaptive compression rate: CR(t)=min{CRmax,max[CRmin,k1*V(t)+k2*(1-R(t))+k3*Q(t)]}, where CR(t) is the compression rate at time t, CRmax and CRmin are the upper and lower limits of the compression rate, and k1, k2, and k3 are adjustment coefficients; obtain the status indicators and weight coefficients of each blast furnace, and calculate the cross-furnace influence: I(t)=Σ(wi*Si(t))*[1-exp(-ρ*N(t))], where I(t) is the cross-furnace influence, Si(t) is the status indicator of the i-th blast furnace, wi is the blast furnace weight coefficient, N(t) is the number of related blast furnaces, and ρ is the synergy effect coefficient; according to the calculated sampling frequency, compression rate and cross-furnace influence, dynamically adjust the data collection and compression strategy.

[0028] Furthermore, the present application also proposes an adaptive data extraction device based on multi-level dynamic sampling, which includes: a parameter acquisition module, used to acquire the operating parameters and system resource usage status information of multiple blast furnaces; a parameter stratification module, used to divide the blast furnace monitoring parameters into a high-frequency sampling layer, a conventional sampling layer and a low-frequency sampling layer according to the operating parameters; an index calculation module, used to calculate the blast furnace abnormal state index and resource utilization rate based on the operating parameters; a correlation analysis module, used to perform cross-furnace data correlation analysis and calculate the cross-furnace correlation degree; a sampling adjustment module, used to dynamically adjust the sampling frequency of each sampling layer according to the abnormal state index, resource utilization rate and cross-furnace correlation degree; a data compression module, used to adaptively compress the collected data based on the resource usage status information and data importance; a format conversion module, used to convert the compressed heterogeneous data into a unified data format.

[0029] From the above, it can be seen that the present application provides an adaptive data extraction method and device based on multi-level dynamic sampling. The method solves the problem of data redundancy or insufficiency caused by fixed sampling frequency by dynamically adjusting the sampling frequency and adaptive compression processing, while taking into account the mutual influence between blast furnaces and realizing unified conversion of data formats. It has the advantages of being able to dynamically adjust the data acquisition frequency according to the blast furnace status and system resources, realize unified conversion of data formats, take into account the mutual influence between blast furnaces, support long-term trend analysis, and improve data acquisition efficiency and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A flow chart of an adaptive data extraction method based on multi-level dynamic sampling provided in this application.

[0031] Figure 2 A schematic diagram of an adaptive data extraction device based on multi-level dynamic sampling provided in the present application.

[0032] In the figure: 210, parameter acquisition module; 220, parameter stratification module; 230, indicator calculation module; 240, correlation analysis module; 250, sampling adjustment module; 260, data compression module; 270, format conversion module. DETAILED DESCRIPTION

[0033] The technical solutions in the present application will be clearly and completely described below in conjunction with the drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the present application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0034] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0035] In a furnace group consisting of multiple blast furnaces, achieving accurate health management and optimized operations faces complex technical challenges. The existing data acquisition system adopts fixed-frequency sampling and cannot be dynamically adjusted according to the blast furnace status and system resources, resulting in a lack of sufficient data support at critical moments and a large amount of redundant data during non-critical periods. In addition, the data formats of different blast furnaces are not unified, which increases the difficulty of data integration and analysis. The system also lacks consideration of the mutual influence between blast furnaces and cannot effectively capture and analyze the overall operating status of the furnace group. The data compression method is fixed and cannot be adaptively adjusted according to the importance of the data and the use of system resources, affecting storage efficiency and data quality. At the same time, the system lacks the ability to analyze long-term operating trends, making it difficult to support predictive maintenance and optimization decisions. These problems have seriously affected the effectiveness and accuracy of blast furnace group health management and limited the improvement of overall operational efficiency.

[0036] Specifically, these technical issues are particularly prominent in a large steel plant with multiple blast furnaces of different sizes and ages. For example, one of the blast furnaces is undergoing overhaul and requires more intensive data monitoring, while another blast furnace that has just completed its annual overhaul is in the stage of gradually increasing its production capacity, and the remaining blast furnaces are in normal production. In this dynamically changing environment, fixed-frequency data collection cannot meet the specific needs of different blast furnaces. Especially under extreme weather conditions, such as high temperatures in summer or severe cold in winter, the operating parameters of the entire blast furnace group will change significantly, requiring more sophisticated data collection and analysis. However, since the sampling frequency cannot be adjusted dynamically, the system cannot capture these key changes in time. In addition, since blast furnaces of different ages use different generations of automation systems and sensor networks, the data formats and transmission protocols are different, resulting in a lot of redundancy and inconsistency in the data integration process, which seriously affects the accuracy and efficiency of data analysis.

[0037] If these technical problems cannot be effectively solved, a series of serious technical consequences will occur. First, the lack of data at critical moments may lead to delayed identification of abnormal conditions of blast furnaces, increasing safety risks and the possibility of production interruptions. Second, the storage and processing of a large amount of redundant data will significantly increase system resource consumption and reduce overall operating efficiency. The difficulty in integrating data in different formats will hinder cross-furnace data analysis, making it impossible to fully utilize the synergy of the blast furnace group to optimize production. Fixed data compression methods may lead to a decline in the quality of important data and affect the accuracy of decision-making. The lack of long-term trend analysis capabilities will make predictive maintenance and optimization decisions difficult, increase the risk of equipment failure, and reduce production efficiency. The cumulative effect of these problems will seriously restrict the health management level and overall operational efficiency of the blast furnace group. Therefore, there is an urgent need for an adaptive data extraction method that can achieve comprehensive and real-time monitoring of multiple blast furnaces under limited system resources, while taking into account long-term trend analysis, to support the collaborative health management and optimized operational decisions of the blast furnace group.

[0038] In response to the technical challenges of achieving precise health management and optimized operation in a group of multiple blast furnaces, this application conducted in-depth problem analysis and solution exploration.

[0039] First of all, considering the operating status and demand differences of different blast furnaces, this application proposes a multi-level dynamic sampling idea. Specifically, the monitoring parameters can be divided into high-frequency sampling layer, conventional sampling layer and low-frequency sampling layer according to the operating parameters of the blast furnace. This layered strategy can better adapt to the specific needs of different blast furnaces. For example, a blast furnace under overhaul can allocate more resources for high-frequency sampling to ensure safety.

[0040] Secondly, in order to solve the problem of data redundancy or insufficiency caused by fixed sampling frequency, this application considers introducing a dynamic adjustment mechanism. By calculating the abnormal status index of blast furnaces, resource utilization rate and cross-furnace correlation, the status of each blast furnace and the resource situation of the entire system can be evaluated in real time. Based on these indicators, the system can dynamically adjust the sampling frequency of each sampling layer, increase the sampling frequency at critical moments, and appropriately reduce it during non-critical periods, thereby achieving efficient use of resources.

[0041] In terms of data processing, considering the limitation of system resources and the need for data storage efficiency, this application proposes the concept of adaptive compression processing. By evaluating the importance of data and the current resource usage status, the system can intelligently compress the collected data, effectively reducing the storage space while ensuring the quality of important data.

[0042] Finally, in order to solve the problem of different blast furnace data formats being inconsistent, this application proposes converting the compressed heterogeneous data into a unified data format. This step not only simplifies the subsequent data analysis process, but also lays the foundation for cross-furnace data correlation analysis.

[0043] Therefore, refer to Figure 1 , the present application proposes an adaptive data extraction method based on multi-level dynamic sampling, the method comprising the following steps:

[0044] S110, obtaining operating parameters and system resource usage status information of multiple blast furnaces;

[0045] S120, dividing the blast furnace monitoring parameters into a high-frequency sampling layer, a conventional sampling layer and a low-frequency sampling layer according to the operating parameters;

[0046] S130, calculating the blast furnace abnormal state index and resource utilization rate based on the operating parameters;

[0047] S140, performing cross-furnace data correlation analysis and calculating cross-furnace correlation;

[0048] S150, dynamically adjusting the sampling frequency of each sampling layer according to the abnormal status index, resource utilization rate and cross-furnace correlation;

[0049] S160, based on the resource usage status information and the importance of the data, adaptively compress the collected data;

[0050] S170: Convert the compressed heterogeneous data into a unified data format.

[0051] Among them, blast furnace operating parameters refer to various data indicators that reflect the operating status of the blast furnace, which can be specifically achieved by key parameters such as temperature, pressure, and gas composition.

[0052] The system resource usage status information refers to the usage of various hardware and software resources in the computer system, which can be specifically realized by indicators such as CPU usage, memory usage, and storage space usage.

[0053] Among them, the high-frequency sampling layer refers to a set of parameters that require a higher sampling frequency, which can be implemented by using key parameters that have a greater impact on the operating status of the blast furnace.

[0054] Among them, the conventional sampling layer refers to a parameter set using a general sampling frequency, which can be implemented by using parameters that have a certain impact on the blast furnace operating state but are not too sensitive.

[0055] Among them, the low-frequency sampling layer refers to a parameter set using a lower sampling frequency, which can be specifically implemented by using auxiliary parameters that have less impact on the operating status of the blast furnace.

[0056] Among them, the abnormal state index refers to a quantitative indicator that reflects whether the blast furnace operation is in an abnormal state, which can be specifically achieved by the degree to which the key parameters deviate from the normal range.

[0057] The resource utilization rate refers to the degree of occupation of system resources, which can be specifically achieved by the utilization percentage of various hardware and software resources.

[0058] Among them, the cross-furnace correlation refers to the degree of mutual influence between the operating conditions of different blast furnaces, which can be achieved by using the correlation analysis between the key parameters of multiple blast furnaces.

[0059] Among them, adaptive compression processing refers to dynamically adjusting the data compression method according to the importance of data and system resource conditions, which can be specifically achieved by using a variable compression algorithm.

[0060] The core innovation of this application is to propose an adaptive data extraction method based on multi-level dynamic sampling. This method divides the blast furnace monitoring parameters into different sampling layers, and dynamically adjusts the sampling frequency according to the blast furnace abnormal status indicators, resource utilization rate and cross-furnace correlation, so as to achieve comprehensive and real-time monitoring of multiple blast furnaces. At the same time, through adaptive compression processing and unified data format conversion, the problems of data storage efficiency and format inconsistency are effectively solved. This method can achieve accurate health management and optimized operation of blast furnace groups under limited system resources.

[0061] The working principle of this application can be described in detail as follows:

[0062] First, the system obtains the operating parameters and system resource usage status information of multiple blast furnaces. Operating parameters include key indicators such as temperature, pressure, and gas composition, which are collected in real time through various sensors. System resource usage status information includes CPU usage, memory occupancy, etc., reflecting the current system load.

[0063] Next, based on the acquired operating parameters, the system divides the blast furnace monitoring parameters into high-frequency sampling layer, conventional sampling layer and low-frequency sampling layer. This division is based on the degree of influence of the parameters on the blast furnace operation status. For example, the furnace top temperature may be classified into the high-frequency sampling layer, while the amount of auxiliary raw materials added may be classified into the low-frequency sampling layer.

[0064] Then, the system calculates the abnormal state index and resource utilization rate of the blast furnace based on the operating parameters. The abnormal state index can be calculated by comparing the deviation of key parameters from their normal operating range, while the resource utilization rate is directly derived from the system resource usage status information.

[0065] The system also performs cross-furnace data correlation analysis and calculates the cross-furnace correlation. This step quantifies the degree of mutual influence between blast furnaces by analyzing the correlation of key parameters between different blast furnaces. When there is a strong correlation between multiple blast furnaces, the abnormality of one blast furnace may be transmitted to affect other blast furnaces.

[0066] Based on the calculated abnormal status indicators, resource utilization and cross-furnace correlation, the system dynamically adjusts the sampling frequency of each sampling layer. For example, when the abnormal status indicator of a blast furnace increases, the system may increase the sampling frequency of the high-frequency sampling layer of the blast furnace to obtain more detailed data.

[0067] During the data collection process, the system performs adaptive compression on the collected data based on resource usage status information and data importance. Data with higher importance may use a lower compression rate to ensure data quality, while data with lower importance may use a higher compression rate to save storage space.

[0068] Finally, the system converts the compressed heterogeneous data into a unified data format to facilitate subsequent data analysis and processing.

[0069] The core of this working principle lies in its dynamic and adaptive nature. By adjusting the sampling frequency and compression strategy in real time, the system can maximize the value of data under limited resources and provide comprehensive and accurate data support for the health management of blast furnace groups.

[0070] As a preferred implementation, the specific embodiments of the present application can be described as follows:

[0071] In a steel plant with five blast furnaces, each blast furnace is equipped with a variety of sensors to monitor parameters such as temperature, pressure, and gas composition. The system first obtains the operating parameters of these blast furnaces, such as furnace top temperature, furnace belly pressure, CO content, etc., and also obtains system resource usage status information such as server CPU usage and memory usage.

[0072] Based on historical data analysis, the system divides the monitoring parameters into three layers: the high-frequency sampling layer includes key parameters such as furnace top temperature and furnace belly pressure; the conventional sampling layer includes secondary parameters such as supply air temperature and gas utilization rate; the low-frequency sampling layer includes auxiliary parameters such as raw material composition and auxiliary material addition amount.

[0073] The system calculates the abnormal status index of each blast furnace by comparing the deviation of the current parameter value with the normal operating range. For example, when the top temperature of a blast furnace exceeds the normal range by 10%, its abnormal status index may be set to 0.6. At the same time, the system calculates the current resource utilization rate, such as CPU utilization rate of 70% and memory utilization rate of 65%.

[0074] By analyzing the correlation between key parameters of different blast furnaces, the system calculates the cross-furnace correlation. For example, the similarity of the top temperature change trend between blast furnaces 1 and 2 is 0.8. According to the preset rules, the cross-furnace correlation between them may be judged to be 0.7.

[0075] Based on these calculation results, the system dynamically adjusts the sampling frequency of each sampling layer. For example, for a blast furnace with a high abnormal state index, the sampling frequency of its high-frequency sampling layer may be increased from the original 1 second / time to 0.5 seconds / time, while the low-frequency sampling layer may remain unchanged at 5 minutes / time.

[0076] During the data collection process, the system performs adaptive compression on the collected data based on the current 70% CPU usage and 65% memory usage, as well as the importance of the data. For example, the data in the high-frequency sampling layer may be compressed at a ratio of 5:1, while the data in the low-frequency sampling layer may be compressed at a ratio of 10:1.

[0077] Finally, the system converts the compressed heterogeneous data into a unified JSON format to facilitate subsequent data analysis and processing.

[0078] In this way, the system can achieve comprehensive and real-time monitoring of the five blast furnaces with limited resources, providing strong data support for the health management and optimized operation of the blast furnace group.

[0079] In some of the above-mentioned embodiments, during the implementation of the present application, there is still the problem of how to dynamically adjust the sampling frequency according to the blast furnace operating status and system resources.

[0080] In this regard, the present application further proposes that the method includes the following steps: obtaining a preset baseline sampling frequency and time attenuation coefficient; predicting key time points based on historical data and current trends; calculating the time sensitivity function according to the predicted key time points and time attenuation coefficient; calculating the required sampling frequencies of the high-frequency sampling layer, conventional sampling layer and low-frequency sampling layer according to abnormal status indicators, resource utilization rate, cross-furnace correlation, baseline sampling frequency and time sensitivity function; and adjusting the sampling frequencies of the high-frequency sampling layer, conventional sampling layer and low-frequency sampling layer respectively according to the required sampling frequencies.

[0081] This method achieves dynamic adjustment of the sampling frequency by introducing a time sensitivity function and multiple influencing factors, thereby solving the problem that the fixed sampling frequency cannot adapt to changes in the blast furnace's operating status. This dynamic adjustment mechanism can increase the sampling frequency at critical moments to capture important data, while reducing the sampling frequency during non-critical periods to reduce the generation of redundant data.

[0082] Specifically, the technical solution of the present application first obtains a preset baseline sampling frequency and a time decay coefficient. The baseline sampling frequency serves as a reference value for sampling frequency adjustment, while the time decay coefficient is used to control the rate of change of the sampling frequency over time. Next, based on historical data and current trends, key time points are predicted, which usually correspond to important changes in the operating status of the blast furnace. Then, based on the predicted key time points and the time decay coefficient, a time sensitivity function is calculated, which reflects the importance of data collection at different time points.

[0083] Furthermore, the present application considers multiple influencing factors to calculate the required sampling frequency. These factors include abnormal state index, resource utilization rate, cross-furnace correlation, baseline sampling frequency and time sensitivity function. The abnormal state index reflects whether the operation of the blast furnace deviates from the normal range, the resource utilization rate indicates the occupation of system resources, and the cross-furnace correlation reflects the mutual influence between different blast furnaces. By comprehensively considering these factors, the required sampling frequency of each sampling layer can be determined more accurately.

[0084] Finally, according to the calculated required sampling frequency, the actual sampling frequencies of the high-frequency sampling layer, the conventional sampling layer, and the low-frequency sampling layer are adjusted respectively. This hierarchical adjustment strategy can adopt different sampling strategies for parameters of different importance, which not only ensures the monitoring accuracy of key parameters, but also avoids oversampling of non-key parameters.

[0085] As a preferred implementation, the following specific steps can be used to implement the adjustment of the dynamic sampling frequency:

[0086] First, the baseline sampling frequency is set to 10 times / second and the time decay coefficient is set to 0.1. By analyzing the blast furnace operation data of the past 24 hours, a time series analysis model (such as the ARIMA model) is used to predict the key time points that may occur in the next 4 hours. It is assumed that a temperature anomaly may occur in 2 hours.

[0087] Next, based on the predicted key time point and the time attenuation coefficient, the time sensitivity function S(t)=exp(-0.1*|t-tp|) is calculated, where t is the current time and tp is the predicted key time point.

[0088] Then, obtain the current abnormal state indicator (such as 0.8, indicating close to abnormal state), resource utilization rate (such as 0.6), and cross-furnace correlation (such as 0.7). Combine these factors with the baseline sampling frequency and time sensitivity function to calculate the required sampling frequency of the high-frequency sampling layer:

[0089] F_high=10*(1+0.8*0.8+0.6*0.6+0.7*0.7)*S(t)

[0090] Assuming that the calculated high-frequency sampling layer frequency is 18 times / second, according to the preset sampling layer frequency ratio (such as high frequency: normal: low frequency = 4:2:1), the required sampling frequencies of the normal sampling layer and the low-frequency sampling layer are calculated to be 9 times / second and 4.5 times / second respectively.

[0091] Finally, the calculated sampling frequency is applied to each sampling layer to achieve dynamic adjustment of the sampling frequency. For example, for the furnace temperature parameter in the high-frequency sampling layer, the sampling frequency is increased from the original 10 times / second to 18 times / second; for the pressure parameter in the conventional sampling layer, the sampling frequency is adjusted to 9 times / second; for the raw material composition parameter in the low-frequency sampling layer, the sampling frequency is set to 4.5 times / second.

[0092] Through this dynamic adjustment mechanism, the present application can adjust the sampling strategy in time when the blast furnace operation status changes. For example, 2 hours before predicting a possible temperature anomaly, the system will gradually increase the sampling frequency of related parameters to capture more detailed data. At the same time, for other relatively stable parameters, the sampling frequency can be appropriately reduced to optimize the use of system resources.

[0093] Compared with the prior art, the dynamic sampling method of the present application has significant advantages. The traditional fixed-frequency sampling method cannot adapt to the dynamic changes in the operating status of the blast furnace, and may miss important data at critical moments, or generate a large amount of redundant data during the stable period. The present application realizes the intelligent adjustment of the sampling frequency by comprehensively considering multiple influencing factors and time sensitivity. This not only improves the pertinence and efficiency of data collection, but also optimizes the use of system resources and greatly increases the probability of capturing abnormal events. At the same time, in non-critical periods, by reducing the sampling frequency, the data storage requirements can be reduced, thereby improving the operating efficiency of the entire system.

[0094] In some of the above embodiments, during the implementation of the present application, there is still the problem of how to accurately predict key time points to optimize the sampling strategy.

[0095] In this regard, the present application further proposes a method for predicting key time points based on historical data and current trends.

[0096] The method first obtains a preset time period and a parameter fluctuation threshold, which can be determined based on blast furnace operation experience and specific needs. For example, the preset time period can be set to 24 hours or 7 days, and the parameter fluctuation threshold can be set to ±5% of the normal operating parameter.

[0097] Next, obtain the historical data of blast furnace operation parameters within a preset time period. These data may include time series data of key parameters such as temperature, pressure, gas composition, etc. The frequency and accuracy of data collection should be sufficient to reflect the dynamic changes of blast furnace operation.

[0098] Based on the historical data obtained, a time series model is established. You can choose a suitable time series analysis method, such as the autoregressive integrated moving average model (ARIMA), exponential smoothing, or a more complex machine learning model such as the long short-term memory network (LSTM). The choice of model depends on the characteristics of the data and the needs of the forecast.

[0099] Using the established time series model, the changing trends of the blast furnace operating parameters are calculated. This step includes not only the prediction of parameter values, but also the calculation of the rate of change and acceleration to fully capture the dynamic characteristics of the parameters.

[0100] Based on the calculated change trend and the pre-set parameter fluctuation threshold, the time interval when the parameter fluctuation exceeds the threshold is identified. This step can be achieved by comparing the difference between the predicted value and the threshold, taking into account the duration and fluctuation amplitude.

[0101] Finally, the starting time points of the identified time intervals are used as the key time points for prediction. These key time points represent the moments when blast furnace operation may change significantly, requiring special attention and more intensive data collection.

[0102] In practical applications, this method can be further optimized. For example, multi-parameter joint analysis can be introduced to consider the correlation and interaction between different parameters. It can also be combined with an expert knowledge system to incorporate blast furnace operation experience into the prediction model to improve the accuracy and interpretability of the prediction.

[0103] In addition, the sliding window technology can be used to continuously update the historical data set so that the model can adapt to the long-term changes in the blast furnace operation status. At the same time, an adaptive learning mechanism can be introduced to continuously adjust and optimize the parameters of the prediction model based on the comparison between the prediction results and the actual situation.

[0104] In order to further improve the robustness of prediction, we can consider introducing multi-model integration technology. By combining multiple different types of prediction models, such as statistical models, machine learning models, and physical models, we can make full use of the advantages of various models and improve the overall performance of prediction.

[0105] In the specific implementation process, different warning levels can be set. For example, when the predicted parameter fluctuation is close to but not exceeding the threshold, the system can issue a mild warning to prompt operators to pay close attention; when the predicted parameter fluctuation will significantly exceed the threshold, the system will issue a high warning and recommend immediate adjustment of the sampling strategy or taking corresponding operational measures.

[0106] By implementing this method of predicting key time points based on historical data and current trends, the present application can more accurately capture the key moments of change in the blast furnace's operating status. This not only optimizes the data collection strategy, improves the quality and value of the data, but also provides timely decision support for blast furnace operations. Compared with the fixed frequency sampling method, this dynamic prediction and adjustment method can significantly reduce the pressure of data storage and transmission while ensuring data quality.

[0107] As a specific embodiment, the system sets the preset time period to 72 hours and the parameter fluctuation threshold to ±3%. The ARIMA model is combined with the random forest algorithm for time series analysis and prediction. The system updates the data every 15 minutes and uses the historical data of the last 72 hours for modeling and prediction.

[0108] During the prediction process, the system focuses on key parameters such as furnace top temperature, furnace belly temperature, air supply, hot air temperature, top pressure and slag iron ratio. By analyzing the historical data of these parameters, the system identifies a possible critical time point: it is expected that after 36 hours, the furnace belly temperature will rise by 5%, and the air supply will need to increase by 7% accordingly.

[0109] Based on this prediction, the system automatically adjusts the sampling strategy: starting 12 hours before the predicted critical time point, the sampling frequency of the furnace temperature and air supply volume is increased from once every 5 minutes to once every 1 minute, and continues until 6 hours after the critical time point. At the same time, the system also increases the sampling frequency of related parameters such as hot air temperature and top pressure.

[0110] This predictive sampling strategy adjustment enables operators to monitor changes in the blast furnace status more finely and to detect and handle potential abnormal situations in a timely manner.

[0111] Compared with the prior art, the method of the present application has significant advantages. The traditional fixed frequency sampling method cannot adapt to the dynamic changes of the blast furnace operation status, often lacks sufficient data support at critical moments, and generates a large amount of redundant data in non-critical periods. The method of the present application, through intelligent prediction and dynamic adjustment, not only ensures the data density at critical moments, but also avoids data redundancy, greatly improving the efficiency and value of data collection. In addition, the method of the present application also takes into account multi-parameter linkage analysis, which can more comprehensively reflect the operation status of the blast furnace and provide a more reliable basis for operational decisions.

[0112] In some of the above-mentioned embodiments, during the implementation of the present application, there is still a problem of how to accurately calculate the required sampling frequencies of the high-frequency sampling layer, the conventional sampling layer and the low-frequency sampling layer.

[0113] In this regard, the present application further proposes a method for calculating the sampling frequencies required for the high-frequency sampling layer, the conventional sampling layer and the low-frequency sampling layer.

[0114] The technical solution of the present application realizes accurate calculation and dynamic adjustment of sampling frequencies of different sampling layers by introducing sampling frequency adjustment coefficient and sampling layer frequency ratio, combining abnormal state index, resource utilization rate and cross-furnace correlation. This method can better adapt to the actual situation of blast furnace operation and improve the efficiency and accuracy of data collection.

[0115] Specifically, the technical solution of the present application first obtains the preset sampling frequency adjustment coefficient and sampling layer frequency ratio. The sampling frequency adjustment coefficient is used to control the adjustment amplitude of the sampling frequency, while the sampling layer frequency ratio defines the frequency relationship between different sampling layers. Then, the frequency adjustment factor is calculated based on the abnormal state index, resource utilization rate and cross-furnace correlation. This frequency adjustment factor reflects the degree of influence of the current blast furnace operation status on the sampling frequency.

[0116] Then, the application calculates the required sampling frequency of the high-frequency sampling layer using the reference sampling frequency, the frequency adjustment factor and the time sensitivity function. The time sensitivity function takes into account the distance between the sampling time point and the predicted key time point, so that more intensive sampling can be performed near the key time point. Finally, according to the sampling layer frequency ratio and the required sampling frequency of the high-frequency sampling layer, the required sampling frequencies of the conventional sampling layer and the low-frequency sampling layer are calculated.

[0117] The advantage of this calculation method is that it takes into account multiple influencing factors, including abnormal conditions of the blast furnace, system resource usage, cross-furnace correlation, and time sensitivity. By combining these factors, the optimal sampling frequency for each sampling layer can be determined more accurately. For example, when an abnormal condition of the blast furnace is detected, the frequency adjustment factor will increase, thereby increasing the sampling frequency; when system resources are tight, the sampling frequency may be appropriately reduced to avoid excessive consumption of resources.

[0118] In addition, the frequency ratio of the sampling layer is used to calculate the frequencies of the conventional sampling layer and the low-frequency sampling layer, ensuring a reasonable frequency relationship between different sampling layers. This method ensures that the high-frequency sampling layer can capture the rapid changes of key parameters and avoids the data collection of the low-frequency sampling layer being too sparse.

[0119] In practical applications, the technical solution of this application can be implemented as follows:

[0120] First, the preset sampling frequency adjustment coefficient is 0.5, and the sampling layer frequency ratio is set to high frequency: normal: low frequency = 4:2:1. Assume that the base sampling frequency is once per minute.

[0121] Next, based on the current blast furnace operation status, the abnormal state index is calculated to be 0.3, the resource utilization rate is 60%, and the cross-furnace correlation is 0.4. Using these data, the frequency adjustment factor is calculated:

[0122] Frequency adjustment factor = 1 + 0.5 * (0.3 + 0.6 + 0.4) = 1.65

[0123] Then, suppose that through time series analysis, it is predicted that a critical time point may occur in 2 hours, and there is still 1 hour to this time point. Use the exponential decay function to calculate the time sensitivity:

[0124] Time sensitivity = exp(-0.5*1)≈0.61

[0125] Calculate the required sampling frequency for the high-frequency sampling layer:

[0126] High frequency sampling frequency = 1*1.65*0.61≈1 time / minute

[0127] Finally, the sampling frequencies of other layers are calculated according to the sampling layer frequency ratio:

[0128] Conventional sampling frequency = 1*(2 / 4)≈0.5 times / minute

[0129] Low frequency sampling frequency = 1*(1 / 4)≈0.25 times / minute

[0130] Through this method, the present application realizes the accurate calculation and dynamic adjustment of the sampling frequency of different sampling layers. Compared with the method of fixed sampling frequency, the technical solution of the present application can better adapt to the dynamic changes of blast furnace operation, optimize the use of system resources while ensuring data quality. For example, when the blast furnace is in a stable state, the sampling frequency can be appropriately reduced to save resources; when an abnormality is detected or approaching a critical time point, the sampling frequency will be automatically increased to capture more detailed information.

[0131] In addition, the technical solution of this application also takes into account the cross-furnace correlation, which enables the data collection strategy of the entire blast furnace group to be coordinated and consistent. When the status change of a blast furnace may affect other blast furnaces, the sampling frequency of the relevant blast furnaces will also be adjusted accordingly, thereby more comprehensively capturing the overall operating status of the blast furnace group.

[0132] Compared with the prior art, the technical solution of the present application realizes the dynamic adjustment of the sampling frequency, avoiding the problem of data redundancy or insufficiency caused by the fixed sampling frequency. Secondly, by introducing multiple influencing factors and time sensitivity functions, the sampling strategy is made more accurate and flexible. Finally, the method of stratified sampling and frequency ratio is adopted to ensure high-frequency sampling of key parameters and avoid excessive consumption of system resources. These advantages enable the technical solution of the present application to realize comprehensive, real-time and efficient monitoring of multiple blast furnaces under limited system resources.

[0133] In some of the above-mentioned embodiments, during the implementation of the present application, there is still a problem of how to accurately identify the time interval in which the parameter fluctuation exceeds the threshold.

[0134] In this regard, the present application further proposes, including obtaining a preset statistical time window; within the statistical time window, counting the maximum and minimum values ​​of the blast furnace operating parameters to determine the historical fluctuation range; calculating the dynamic parameter fluctuation threshold based on the historical fluctuation range and the preset fluctuation coefficient; comparing the change trend with the dynamic parameter fluctuation threshold; when the change trend exceeds the dynamic parameter fluctuation threshold, recording the corresponding time point; and determining the continuous time interval in which the parameter fluctuation exceeds the threshold based on the recorded time point sequence.

[0135] The method proposed in this application can more flexibly and accurately identify the time interval of abnormal parameter fluctuation by introducing the concept of dynamic parameter fluctuation threshold. This method takes into account the historical fluctuation characteristics of blast furnace operation parameters, making the threshold setting more reasonable and adaptive.

[0136] Specifically, the present application first obtains a preset statistical time window. The size of this time window can be determined according to the characteristics of the blast furnace operation and the frequency of data collection, for example, it can be set to 24 hours or 7 days. Within this statistical time window, the system will count the maximum and minimum values ​​of the blast furnace operation parameters to determine the historical fluctuation range of the parameters.

[0137] Next, the present application introduces a preset fluctuation coefficient. This coefficient is used to adjust the historical fluctuation range to calculate the dynamic parameter fluctuation threshold. The selection of the fluctuation coefficient can be determined according to the stability of the blast furnace operation and the requirements for abnormal sensitivity. For example, if you want the system to be more sensitive to parameter fluctuations, you can choose a smaller fluctuation coefficient; conversely, if you want the system to be able to tolerate larger fluctuations, you can choose a larger fluctuation coefficient.

[0138] The calculation formula of the dynamic parameter fluctuation threshold can be expressed as:

[0139] Dynamic threshold = historical fluctuation range * fluctuation coefficient

[0140] After calculating the dynamic parameter fluctuation threshold, the present application compares the change trend of the blast furnace operation parameter with this threshold. The change trend can be represented by calculating the change rate or slope of the parameter within a certain time interval. When the change trend exceeds the dynamic parameter fluctuation threshold, the system will record the corresponding time point.

[0141] Finally, the present application determines the continuous time interval in which the parameter fluctuation exceeds the threshold value based on the recorded time point sequence. This step can be achieved by analyzing the intervals between the recorded time points. If the interval between two adjacent time points is less than a preset value (e.g., twice the data acquisition period), the two time points are considered to belong to the same continuous time interval.

[0142] Through this method, the present application can more accurately identify the time interval of abnormal fluctuations in blast furnace operating parameters, providing an important basis for subsequent data analysis and decision-making.

[0143] As a preferred implementation, the following specific examples may be considered:

[0144] Assume that the temperature parameters of a blast furnace are monitored, and the system sets the statistical time window to 24 hours and the fluctuation coefficient to 1.5. In the past 24 hours, the lowest value of the temperature parameter is 1100℃ and the highest value is 1200℃. Then, the historical fluctuation range is 100℃ (1200℃-1100℃).

[0145] The dynamic parameter fluctuation threshold is calculated as follows:

[0146] Dynamic threshold = 100°C * 1.5 = 150°C

[0147] Next, the system calculates the rate of change of the temperature parameter every 10 minutes. Assume that at a certain moment, the system detects that the temperature rises from 1150°C to 1320°C in 30 minutes, with a change of 170°C, exceeding the dynamic threshold of 150°C. The system will record this time point.

[0148] If the temperature continues to remain above the normal range for the next few 10-minute intervals, the system will record these time points. When the temperature begins to fall back to the normal range, the system can determine a continuous abnormal time interval.

[0149] For example, if the abnormal time point sequence recorded by the system is: 14:00, 14:10, 14:20, 14:30, then it can be determined that the 30-minute time period from 14:00 to 14:30 is the continuous time interval in which the parameter fluctuation exceeds the threshold.

[0150] This method has obvious advantages over the fixed threshold method. It can adapt to the dynamic changes of blast furnace operating parameters and avoid false alarms or missed alarms caused by improper threshold settings. At the same time, by introducing statistical time windows and fluctuation coefficients, the system can dynamically adjust the threshold according to the actual operating conditions of the blast furnace, improving the accuracy and flexibility of anomaly detection.

[0151] In addition, this approach can be combined with other data analysis techniques. For example, the identified abnormal time intervals can be correlated with other blast furnace parameters to find potential root causes of problems. Or, this information can be input into a predictive model to improve the accuracy of predictions about future blast furnace conditions.

[0152] In general, the method proposed in this application significantly improves the recognition accuracy of abnormal fluctuations in blast furnace operating parameters by introducing the concepts of dynamic parameter fluctuation thresholds and continuous time intervals. This not only helps to timely discover potential problems in blast furnace operation, but also provides important data support for the coordinated management and optimization decision-making of blast furnace groups.

[0153] In some of the above-mentioned embodiments, during the implementation of the present application, there is still the problem of how to adaptively compress the collected data according to the system resource status and data importance.

[0154] In this regard, the present application further proposes to perform adaptive compression processing on the collected data based on resource usage status information and data importance.

[0155] The method first obtains the preset compression rate upper limit, compression rate lower limit and data importance evaluation criteria. These preset parameters provide the basic framework and constraints for the subsequent adaptive compression processing. The compression rate upper limit and lower limit ensure that the compression process will not excessively lose data quality or occupy too many system resources. The data importance evaluation criteria provide a basis for distinguishing the value of different data.

[0156] Next, the current system resource utilization is calculated based on the resource usage status information. This step can be achieved by monitoring indicators such as CPU usage, memory usage, storage space, etc. For example, a weighted average method can be used to integrate various resource usage indicators into a unified resource utilization indicator.

[0157] Then, based on the data importance evaluation criteria, the importance of the collected data is evaluated to obtain the data importance index. The evaluation criteria can include multiple dimensions such as data change rate, abnormality, and correlation with key process parameters. By setting the weights of different dimensions, a comprehensive data importance index can be calculated.

[0158] Calculate the target compression rate based on the data importance index and the current system resource utilization. This step can use an adaptive algorithm, such as:

[0159] Target compression rate = basic compression rate + α*(1-data importance index) + β*system resource utilization

[0160] Among them, α and β are adjustment coefficients used to balance the impact of data importance and system resource status on the compression rate. The basic compression rate can be set to an initial value based on historical experience.

[0161] The target compression ratio is compared with the upper and lower compression ratio limits to obtain the actual compression ratio. This step ensures that the final compression ratio is within the preset reasonable range to avoid over-compression or under-compression.

[0162] Finally, according to the actual compression rate, the collected data is compressed to obtain compressed data. Compression can use a variety of algorithms, such as lossless compression algorithms (such as Huffman coding, LZW algorithm) or lossy compression algorithms (such as wavelet transform, principal component analysis). When selecting a suitable compression algorithm, factors such as data type, compression efficiency, and decompression speed need to be considered.

[0163] Through this adaptive compression processing method, the present application can effectively reduce the resource consumption of data storage and transmission while ensuring data quality. When system resources are tight, the compression rate can be appropriately increased; when processing important data, the compression rate can be reduced to retain more detailed information. This flexible strategy enables the data acquisition system to better adapt to the complex and changeable operating environment of the blast furnace group.

[0164] Assume that in a blast furnace group monitoring system, the upper limit of the compression rate is set to 90% and the lower limit is set to 10%. The data importance evaluation criteria include three dimensions: data change rate (weight 0.4), abnormality (weight 0.3) and correlation with key process parameters (weight 0.3).

[0165] At some point, the system detects:

[0166] The current CPU usage is 75%, memory usage is 80%, and storage space usage is 60%.

[0167] For the data of temperature sensor T1:

[0168] The data change rate is 0.8 (large change);

[0169] The abnormality level is 0.6 (slight abnormality);

[0170] The correlation with key process parameters is 0.9 (highly correlated).

[0171] First, calculate the system resource utilization:

[0172] System resource utilization = (75% + 80% + 60%) / 3 = 71.67%;

[0173] Then, calculate the data importance index:

[0174] Data importance index = 0.8*0.4+0.6*0.3+0.9*0.3=0.77;

[0175] Assuming the base compression rate is 50%, α = 0.3, β = 0.2, the target compression rate is calculated as follows:

[0176] Target compression rate = 50% + 0.3*(1-0.77) + 0.2*71.67% = 61.42%;

[0177] Comparing the target compression ratio with the upper and lower limits, the actual compression ratio is 61.42%.

[0178] Finally, the system selects an appropriate compression algorithm (such as wavelet transform) to perform 61.42% compression processing on the data of the temperature sensor T1.

[0179] In this way, the application can dynamically adjust the compression strategy according to the current system resource status and data importance. For temperature data with higher importance, a relatively low compression rate is used to retain more detailed information. At the same time, considering the current high system resource utilization rate, the compression rate is appropriately increased to reduce the system burden. This balanced strategy not only ensures the quality of key data, but also avoids excessive consumption of system resources.

[0180] Compared with the prior art, the present application can adjust the compression strategy according to the real-time system resource status and data characteristics, instead of adopting a fixed compression method. Through multi-dimensional data importance assessment, differentiated processing of different types of data is achieved, avoiding excessive compression of important data. By incorporating system resource utilization into the compression rate calculation, the compression rate can be appropriately increased when the system load is high, effectively preventing system overload. By setting the upper and lower limits of the compression rate and various adjustment parameters, the system administrator can flexibly adjust the compression strategy according to the specific application scenario. This method can adapt to different types of data and compression algorithms, and has good versatility and scalability.

[0181] Through this adaptive compression processing method, the present application effectively solves the problem of how to balance data quality and system resource consumption in a complex and changeable blast furnace group monitoring environment, and provides strong support for the intelligent health management of the blast furnace group.

[0182] In some of the above-mentioned embodiments, during the implementation of the present application, there are still problems such as low efficiency in data collection and processing and difficulty in timely responding to dynamic changes of the blast furnace group.

[0183] In this regard, the present application further proposes to deploy edge computing nodes in each blast furnace to perform data preprocessing, anomaly detection and preliminary analysis; use edge computing nodes to perform real-time compression and feature extraction on the collected data; regularly calculate data integrity indicators and information entropy, evaluate the quality of the collected data to obtain data quality assessment results; and dynamically adjust the sampling frequency and compression rate of each sampling layer based on the data quality assessment results.

[0184] This technical solution realizes localized data processing and analysis by deploying edge computing nodes in each blast furnace, reducing the amount of data transmission and improving the system response speed. The edge computing nodes perform data preprocessing, anomaly detection and preliminary analysis, which can quickly identify potential problems and provide support for subsequent in-depth analysis.

[0185] Using edge computing nodes for real-time compression and feature extraction can reduce data storage and transmission requirements while retaining key information. This approach can effectively balance data quality and system resource usage.

[0186] Regularly calculating data integrity indicators and information entropy can objectively evaluate the quality of collected data. Data integrity indicators reflect the continuity and reliability of data, while information entropy measures the information content of data. These two indicators together form the basis for data quality assessment.

[0187] The sampling frequency and compression rate are dynamically adjusted according to the data quality assessment results, realizing the adaptive optimization of the data collection strategy. When the data quality is low, the system may increase the sampling frequency or reduce the compression rate to obtain more information; conversely, when the data quality is good, the sampling frequency can be appropriately reduced or the compression rate can be increased to save resources.

[0188] In specific implementation, edge computing nodes can use industrial-grade embedded computers equipped with appropriate processors, memory and storage devices. These nodes are connected to the central control system via industrial Ethernet or wireless networks. The software running on the edge computing nodes includes data acquisition modules, preprocessing modules, anomaly detection modules and compression modules.

[0189] Data preprocessing may include operations such as denoising, filtering, and standardization. Anomaly detection can use statistical methods or machine learning algorithms, such as moving average, CUSUM algorithm, or isolation forest. Feature extraction can use time domain analysis, frequency domain analysis, or wavelet transform to extract key features such as mean, variance, peak frequency, etc.

[0190] Data compression can select appropriate algorithms according to the data type, such as differential encoding for slowly changing parameters such as temperature, and wavelet transform compression for fast changing parameters such as pressure. The compression rate can be dynamically adjusted according to the importance of the data and system resources, for example, the compression rate can be set in the range of 20% to 80%.

[0191] Data integrity indicators can be evaluated by calculating indicators such as data missing rate, outlier ratio, etc. Information entropy can be calculated using Shannon entropy or its variants. These indicators can be calculated once every hour or every shift to form a data quality assessment report.

[0192] Based on the data quality assessment results, the system can formulate dynamic adjustment strategies. For example, when the data integrity is less than 90%, the sampling frequency of the corresponding parameters can be increased; when the information entropy is lower than the threshold, the compression rate can be reduced to retain more information. The specific adjustment range can be determined by preset rules or machine learning algorithms.

[0193] By deploying edge computing nodes and implementing the above strategies, this application can significantly improve the efficiency of data collection and processing.

[0194] Dynamically adjusting the sampling strategy enables the system to increase the sampling frequency at critical moments (such as when the blast furnace status changes suddenly) to capture more detailed information, while appropriately reducing the sampling frequency during stable operation to save resources.

[0195] In addition, the anomaly detection function of the edge computing node can issue an alarm at the early stage of the problem, discovering potential problems 20-30 minutes in advance on average, leaving more response time for operators. This not only improves the safety of blast furnace operation, but also provides strong support for preventive maintenance.

[0196] Compared with the traditional centralized data processing method, the distributed edge computing architecture of this application has obvious advantages. The traditional method requires all data to be transferred to the central server for processing, which is prone to network congestion and processing delays. However, this application greatly reduces the burden on the central system and improves overall efficiency by performing preliminary processing and analysis at the edge node.

[0197] In addition, the dynamic adjustment mechanism of the present application is more flexible and adaptable than the fixed parameter system. It can automatically optimize the data collection and processing strategy according to the actual situation, maximizing resource utilization efficiency while ensuring data quality. This intelligent management method not only improves the reliability of the system, but also reduces the need for manual intervention and reduces the risk of operational errors.

[0198] In general, this application effectively solves the efficiency problems in data collection and processing of blast furnace groups through edge computing and dynamic adjustment strategies, and provides strong technical support for the intelligent management and optimized operation of blast furnace groups.

[0199] In some of the above-mentioned embodiments, during the implementation of the present application, there is still a problem of being unable to respond to changes in the operating status of the blast furnace in a timely manner.

[0200] In this regard, the present application further proposes that the method includes obtaining historical data and current trend information of blast furnace operation, using a time series analysis model to predict future key time points based on this information, and adjusting the sampling frequency of relevant parameters in the high-frequency sampling layer, conventional sampling layer and low-frequency sampling layer according to the predicted key time points.

[0201] This method optimizes the data collection strategy by predicting future key time points, improving the pertinence and efficiency of data collection. By dynamically adjusting the sampling frequency, this application can obtain more detailed data at critical moments while reducing data redundancy during non-critical periods, thereby achieving effective use of resources.

[0202] Specifically, the method of the present application first obtains the preset sampling frequency adjustment rules and parameter sets corresponding to the abnormal types. These rules and parameter sets provide the basis for the subsequent sampling frequency adjustment. Next, the time information and the predicted abnormal type are extracted from the predicted future key time points. This step enables the system to prepare for specific time points and possible abnormal situations.

[0203] Based on the extracted time information, the time window in which the sampling frequency needs to be adjusted is determined. This time window setting ensures that the sampling frequency adjustment is targeted and only performed during the necessary time period. Based on the predicted anomaly type and the parameter set corresponding to the anomaly type, relevant parameters are selected from the high-frequency sampling layer, the regular sampling layer, and the low-frequency sampling layer. This selective parameter adjustment ensures efficient use of system resources, and only the sampling frequency of the parameters that may be affected is adjusted.

[0204] Then, according to the sampling frequency adjustment rules, combined with the anomaly type and time window, the target sampling frequency of the selected parameter is calculated. This calculation takes into account multiple factors to ensure that the adjustment of the sampling frequency is comprehensive and reasonable. Based on the calculation results, sampling adjustment instructions containing parameter identification, time window and target sampling frequency are generated. These instructions contain all the necessary information required to perform the sampling frequency adjustment.

[0205] Finally, the sampling adjustment instruction is sent to the corresponding data acquisition device to achieve dynamic adjustment of the sampling frequency of the relevant parameters. In this way, the application can flexibly adjust the data acquisition strategy according to the predicted future situation and improve the accuracy and efficiency of data acquisition.

[0206] The method of the present application can be further refined in practical applications. For example, when predicting future key time points, a variety of time series analysis models, such as ARIMA, LSTM, etc., can be used to select the most suitable model according to the characteristics of historical data. When calculating the target sampling frequency, a weight coefficient can be introduced to assign different importance to different influencing factors.

[0207] The specific implementation is as follows:

[0208] Assume that there is a blast furnace A in a blast furnace group that is about to undergo an overhaul. Based on historical data and current trends, the time series analysis model predicts that the temperature of blast furnace A may fluctuate abnormally in the next 72 hours. After the system obtains this prediction result, it performs the following steps:

[0209] Extract time information: next 72 hours;

[0210] Predicted anomaly types: abnormal temperature fluctuations;

[0211] Determine the adjustment time window: 72 hours from the current time;

[0212] Select relevant parameters: temperature sensor, pressure sensor, gas composition sensor;

[0213] Calculate the target sampling frequency:

[0214] Temperature sensor: increased from 5 minutes / time to 1 minute / time;

[0215] Pressure sensor: increased from 10 minutes / time to 3 minutes / time;

[0216] Gas composition sensor: from 15 minutes / time to 5 minutes / time;

[0217] Generate sampling adjustment instructions:

[0218] {Parameters: temperature sensor, time window: 72 hours, target frequency: 1 minute / time};

[0219] {Parameters: pressure sensor, time window: 72 hours, target frequency: 3 minutes / time};

[0220] {Parameters: gas composition sensor, time window: 72 hours, target frequency: 5 minutes / time};

[0221] Send the command to the corresponding data acquisition device.

[0222] Through this dynamic adjustment, the application can obtain more intensive and detailed data during the expected abnormal period, providing more reliable support for subsequent analysis and decision-making. At the same time, because the adjustment is targeted, it avoids oversampling of other irrelevant parameters, thereby saving system resources.

[0223] Compared with the prior art, this application predicts future key time points through time series analysis, achieving forward-looking adjustment of data collection rather than passive response. Selecting relevant parameters according to the predicted anomaly type ensures the accuracy and pertinence of the sampling frequency adjustment. The dynamic adjustment mechanism can flexibly adjust the sampling strategy according to different time windows and anomaly types. By selectively adjusting the sampling frequency, unnecessary data redundancy is avoided and the utilization efficiency of system resources is improved.

[0224] In some of the above-mentioned embodiments, during the implementation of the present application, there is still a technical problem of how to realize dynamic sampling, adaptive compression and cross-furnace impact analysis of blast furnace group data.

[0225] In this regard, the present application further proposes a technical solution for obtaining the benchmark sampling frequency, abnormal status indicators, resource utilization rate and cross-furnace correlation of each blast furnace; obtaining the predicted key time points and data change rates; calculating the dynamic sampling frequency, data quality evaluation function, adaptive compression rate and cross-furnace influence based on a mathematical model; and dynamically adjusting the data collection and compression strategy according to the calculated sampling frequency, compression rate and cross-furnace influence.

[0226] The technical solution proposed in this application introduces multiple mathematical models to realize dynamic sampling, adaptive compression and cross-furnace impact analysis of blast furnace group data. Through the dynamic sampling frequency model F(t), factors such as the baseline sampling frequency, abnormal state indicators, resource utilization rate and cross-furnace correlation are considered, and the time decay function is introduced so that the sampling frequency can be dynamically adjusted according to the blast furnace status and the key time points of the prediction. The data quality evaluation function Q(t) evaluates the data quality based on the data change rate, providing a basis for subsequent data compression. The adaptive compression rate model CR(t) comprehensively considers the data value, resource utilization rate and data quality, and realizes the dynamic adjustment of the compression rate. The cross-furnace impact model I(t) quantifies the mutual influence between blast furnaces by considering the status indicators, weight coefficients and number of related blast furnaces of multiple blast furnaces.

[0227] Specifically, the calculation formula of the dynamic sampling frequency model F(t) is:

[0228] F(t)=base_f*[1+α*A(t)+β*R(t)+γ*C(t)]*exp(-λ*|t-tp|);

[0229] Among them, base_f is the base sampling frequency, A(t) is the abnormal state indicator, R(t) is the resource utilization rate, and C(t) is the cross-furnace correlation. α, β, γ are weight coefficients, and α+β+γ=1, which is used to balance the influence of different factors. tp is the key time point of prediction, and λ is the time decay coefficient. This model allows the system to dynamically adjust the sampling frequency according to the current state of the blast furnace and the predicted key time point. For example, when the abnormal state indicator A(t) is high, the sampling frequency will increase accordingly to capture more detailed information.

[0230] The calculation formula of the data quality assessment function Q(t) is:

[0231] Q(t)=1-exp(-μ*D(t));

[0232] Where D(t) is the data change rate and μ is the quality sensitivity coefficient. This function can evaluate the quality of the data and provide a basis for subsequent compression processing. When the data change rate is large, the Q(t) value will be close to 1, indicating that the data quality is high and higher compression accuracy is required.

[0233] The calculation formula of the adaptive compression rate model CR(t) is:

[0234] CR(t)=min{CRmax,max[CRmin,k1*V(t)+k2*(1-R(t))+k3*Q(t)]};

[0235] Among them, CRmax and CRmin are the upper and lower limits of the compression rate, V(t) is the data value, R(t) is the resource utilization rate, Q(t) is the data quality assessment result, and k1, k2, and k3 are adjustment coefficients. This model can dynamically adjust the compression rate according to the importance of the data, system resource status, and data quality, thereby improving storage efficiency while ensuring the quality of important data.

[0236] The calculation formula of the cross-furnace influence model I(t) is:

[0237] I(t)=Σ(wi*Si(t))*[1-exp(-ρ*N(t))];

[0238] Among them, Si(t) is the status index of the i-th blast furnace, wi is the blast furnace weight coefficient, N(t) is the number of related blast furnaces, and ρ is the synergy effect coefficient. This model can quantify the mutual influence between blast furnaces and provide a basis for capturing and analyzing the overall operating status of the furnace group.

[0239] These models work together to achieve dynamic adjustment of data collection and compression strategies. The system can increase the sampling frequency at critical moments to capture more detailed information; reduce the sampling frequency during non-critical periods to reduce redundant data; adjust the compression rate according to the importance of data and system resource conditions, balance data quality and storage efficiency; and consider the mutual influence between blast furnaces to support the overall optimization management of the furnace group.

[0240] In actual application, the system first obtains the benchmark sampling frequency, abnormal state index, resource utilization rate and cross-furnace correlation of each blast furnace. For example, the benchmark sampling frequency of a blast furnace may be set to 1 time per minute, the abnormal state index is 0.2 (indicating a slight abnormality), the resource utilization rate is 60%, and the cross-furnace correlation is 0.3. At the same time, the system predicts the key time points that may occur in the next 4 hours through time series analysis, such as predicting that temperature anomalies may occur in 2 hours.

[0241] Next, the system uses the dynamic sampling frequency model F(t) to calculate the required sampling frequency. Assume that the calculation results show that the sampling frequency needs to be increased to 3 times per minute near the predicted temperature anomaly time point. The system will adjust the sampling frequency of the data acquisition device accordingly.

[0242] At the same time, the system continuously monitors the data change rate and uses the data quality evaluation function Q(t) to evaluate the data quality. For example, if the data change rate at a certain moment is large, Q(t) may reach 0.8, indicating that the data quality is high.

[0243] Based on the data quality assessment results and current resource usage, the system uses the adaptive compression rate model CR(t) to calculate the appropriate compression rate. Assuming the calculated result is 0.7, it means that the data will be compressed to 70% of the original size.

[0244] In addition, the system also analyzes the mutual influence between blast furnaces through the cross-furnace influence model I(t). For example, if the calculated cross-furnace influence is 0.5, it means that there is a strong cross-furnace influence, and the system may adjust the sampling strategy of the relevant blast furnaces accordingly.

[0245] In this way, the application can dynamically adjust the data collection and compression strategy according to the real-time status and prediction results of the blast furnace group, which not only ensures the quality of key data, but also improves storage efficiency, while also capturing the mutual influence between blast furnaces.

[0246] Compared with the prior art, the technical solution of the present application has the following advantages: First, through the dynamic sampling frequency model, the problem that the fixed sampling frequency cannot adapt to the dynamic changes of the blast furnace is solved. Secondly, the adaptive compression rate model can flexibly adjust the compression strategy according to the importance of the data and the system resource status, thereby improving the storage efficiency. Thirdly, the introduction of the cross-furnace influence model enables the system to capture and analyze the overall operating status of the blast furnace group, providing more comprehensive information support for optimizing decisions. Finally, the combined use of these models realizes the intelligence and adaptability of data collection, compression and analysis, and can better cope with changes in the operating status of the blast furnace and extreme environmental conditions.

[0247] Second, refer to Figure 2 The present application further proposes an adaptive data extraction device based on multi-level dynamic sampling, the device comprising:

[0248] The parameter acquisition module 210 is used to obtain the operating parameters and system resource usage status information of multiple blast furnaces;

[0249] A parameter stratification module 220, for dividing the blast furnace monitoring parameters into a high-frequency sampling layer, a conventional sampling layer and a low-frequency sampling layer according to the operating parameters;

[0250] An index calculation module 230, for calculating an abnormal state index of the blast furnace and a resource utilization rate based on operating parameters;

[0251] The correlation analysis module 240 is used to perform cross-furnace data correlation analysis and calculate the cross-furnace correlation degree;

[0252] The sampling adjustment module 250 is used to dynamically adjust the sampling frequency of each sampling layer according to the abnormal state index, resource utilization rate and cross-furnace correlation;

[0253] A data compression module 260, configured to perform adaptive compression processing on the collected data based on resource usage status information and data importance;

[0254] The format conversion module 270 is used to convert the compressed heterogeneous data into a unified data format.

[0255] By dynamically adjusting the sampling frequency and adaptive compression processing, the problem of data redundancy or insufficiency caused by fixed sampling frequency is solved. At the same time, the mutual influence between blast furnaces is taken into consideration, and the unified conversion of data format is realized. It has the advantages of being able to dynamically adjust the data collection frequency according to the blast furnace status and system resources, realize unified conversion of data format, consider the mutual influence between blast furnaces, support long-term trend analysis, and improve data collection efficiency and quality.

[0256] In addition, in some preferred embodiments, an adaptive data extraction device based on multi-level dynamic sampling proposed in the present application can perform any step of the above method.

[0257] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An adaptive data extraction method based on multi-level dynamic sampling, characterized in that: The method comprises the following steps: Obtain the operating parameters and system resource usage status information of multiple blast furnaces; Dividing the blast furnace monitoring parameters into a high-frequency sampling layer, a conventional sampling layer and a low-frequency sampling layer according to the operating parameters; Calculate the blast furnace abnormal state index and resource utilization rate based on the operating parameters; Perform cross-furnace data correlation analysis and calculate cross-furnace correlation; Dynamically adjust the sampling frequency of each sampling layer according to the abnormal status index, resource utilization rate and cross-furnace correlation; Based on the resource usage status information and data importance, adaptively compressing the collected data; Convert the compressed heterogeneous data into a unified data format.

2. The adaptive data extraction method based on multi-level dynamic sampling according to claim 1, characterized in that: The step of dynamically adjusting the sampling frequency of each sampling layer according to the abnormal state index, resource utilization rate and cross-furnace correlation includes: Obtaining a preset reference sampling frequency and time attenuation coefficient; Predict key time points based on historical data and current trends; Calculating a time sensitivity function according to the predicted key time points and the time decay coefficient; Calculate the required sampling frequencies of the high-frequency sampling layer, the conventional sampling layer, and the low-frequency sampling layer according to the abnormal state index, the resource utilization rate, the cross-furnace correlation, the baseline sampling frequency, and the time sensitivity function; According to the required sampling frequency, the sampling frequencies of the high-frequency sampling layer, the normal sampling layer and the low-frequency sampling layer are adjusted respectively.

3. The adaptive data extraction method based on multi-level dynamic sampling according to claim 2 is characterized in that: The steps of predicting key time points based on historical data and current trends include: Get the preset time period and parameter fluctuation threshold; Acquire historical data of blast furnace operation parameters within the preset time period; Building a time series model based on the historical data; Based on the time series model, calculating the change trend of blast furnace operation parameters; According to the change trend and the parameter fluctuation threshold, identifying a time interval during which the parameter fluctuation exceeds the parameter fluctuation threshold; The starting time point of the time interval is used as the key time point for prediction.

4. The adaptive data extraction method based on multi-level dynamic sampling according to claim 2 is characterized in that: The step of calculating the required sampling frequencies of the high-frequency sampling layer, the conventional sampling layer and the low-frequency sampling layer according to the abnormal state index, the resource utilization rate, the cross-furnace correlation, the reference sampling frequency and the time sensitivity function comprises: Obtain the preset sampling frequency adjustment coefficient and sampling layer frequency ratio; Calculating a frequency adjustment factor according to the abnormal state index, resource utilization rate and cross-furnace correlation; Calculating the required sampling frequency of the high-frequency sampling layer according to the reference sampling frequency, the frequency adjustment factor and the time sensitivity function; The required sampling frequencies of the normal sampling layer and the low-frequency sampling layer are calculated according to the sampling layer frequency ratio and the required sampling frequency of the high-frequency sampling layer.

5. The adaptive data extraction method based on multi-level dynamic sampling according to claim 3 is characterized in that: The step of identifying a time interval during which the parameter fluctuation exceeds the parameter fluctuation threshold according to the change trend and the parameter fluctuation threshold comprises: Get the preset statistical time window; In the statistical time window, the maximum and minimum values ​​of the blast furnace operating parameters are counted to determine the historical fluctuation range; Calculating a dynamic parameter fluctuation threshold value according to the historical fluctuation range and a preset fluctuation coefficient; comparing the change trend with the dynamic parameter fluctuation threshold; When the change trend exceeds the dynamic parameter fluctuation threshold, record the corresponding time point; Based on the recorded time point sequence, determine the continuous time intervals in which the parameter fluctuation exceeds the threshold.

6. The adaptive data extraction method based on multi-level dynamic sampling according to claim 1, characterized in that: The step of adaptively compressing the collected data based on the resource usage status information and the data importance includes: Obtain the preset compression rate upper limit, compression rate lower limit and data importance evaluation standard; Calculating the current system resource utilization rate according to the resource usage status information; Based on the data importance evaluation standard, the collected data is evaluated for importance to obtain a data importance index; Calculating a target compression ratio based on the data importance index and current system resource utilization; Compare the target compression rate with the compression rate upper limit and the compression rate lower limit to obtain an actual compression rate; According to the actual compression rate, the collected data is compressed to obtain compressed data.

7. The adaptive data extraction method based on multi-level dynamic sampling according to claim 1, characterized in that: The method further comprises: Deploy edge computing nodes at each blast furnace to perform data preprocessing, anomaly detection and preliminary analysis; Using the edge computing node to perform real-time compression and feature extraction on the collected data; Regularly calculate data integrity indicators and information entropy to evaluate the quality of collected data and obtain data quality assessment results; The sampling frequency and compression rate of each sampling layer are dynamically adjusted according to the data quality evaluation result.

8. The adaptive data extraction method based on multi-level dynamic sampling according to claim 1, characterized in that: The method further comprises: Obtain historical data and current trend information of blast furnace operation; Based on the historical data and current trend information, use a time series analysis model to predict future key time points; According to the future key time point, adjusting the sampling frequencies of relevant parameters in the high-frequency sampling layer, the conventional sampling layer and the low-frequency sampling layer; The step of adjusting the sampling frequencies of relevant parameters in the high-frequency sampling layer, the conventional sampling layer and the low-frequency sampling layer according to the future key time point comprises: Obtain the preset sampling frequency adjustment rules and parameter sets corresponding to the exception types; extracting time information and predicted anomaly types from the future key time points; Determine the time window in which the sampling frequency needs to be adjusted according to the time information; Based on the abnormal type and the parameter set corresponding to the abnormal type, selecting relevant parameters from the high-frequency sampling layer, the regular sampling layer and the low-frequency sampling layer; Calculate the target sampling frequency of the selected parameter according to the sampling frequency adjustment rule in combination with the anomaly type and the time window; generating a sampling adjustment instruction including a parameter identifier, a time window, and a target sampling frequency; The sampling adjustment instruction is sent to the corresponding data acquisition device to realize dynamic adjustment of the sampling frequency of relevant parameters.

9. The adaptive data extraction method based on multi-level dynamic sampling according to claim 1, characterized in that: Also includes: Obtain the benchmark sampling frequency, abnormal status indicators, resource utilization rate and cross-furnace correlation of each blast furnace; Obtain the key time points and data change rates for prediction; The dynamic sampling frequency is calculated based on the following mathematical model: F(t)=base_f*[1+α*A(t)+β*R(t)+γ*C(t)]*exp(-λ*|t-tp|), where F(t) is the required sampling frequency at time t, base_f is the base sampling frequency, A(t) is the abnormal state indicator, R(t) is the resource utilization rate, C(t) is the cross-furnace correlation, α, β, γ are weight coefficients and α+β+γ=1, tp is the key time point for prediction, and λ is the time decay coefficient; Based on the data importance assessment, the data value V(t) and data quality assessment function are calculated: Q(t)=1-exp(-μ*D(t)) Where D(t) is the data change rate, μ is the quality sensitivity coefficient; According to the data value and data quality evaluation results, the adaptive compression rate is calculated: CR(t)=min{CRmax,max[CRmin,k1*V(t)+k2*(1-R(t))+k3*Q(t)]} Among them, CR(t) is the compression rate at time t, CRmax and CRmin are the upper and lower limits of the compression rate, and k1, k2, and k3 are adjustment coefficients; Obtain the status index and weight coefficient of each blast furnace and calculate the cross-furnace influence: I(t)=Σ(wi*Si(t))*[1-exp(-ρ*N(t))] Among them, I(t) is the cross-furnace influence, Si(t) is the status index of the i-th blast furnace, wi is the blast furnace weight coefficient, N(t) is the number of related blast furnaces, and ρ is the synergy effect coefficient; The data collection and compression strategies are dynamically adjusted based on the calculated sampling frequency, compression rate and cross-furnace influence.

10. An adaptive data extraction device based on multi-level dynamic sampling, characterized in that: The device includes: Parameter acquisition module, used to obtain the operating parameters of multiple blast furnaces and system resource usage status information; A parameter stratification module, used to divide the blast furnace monitoring parameters into a high-frequency sampling layer, a conventional sampling layer and a low-frequency sampling layer according to the operating parameters; An index calculation module, used for calculating the abnormal state index and resource utilization rate of the blast furnace based on the operating parameters; The correlation analysis module is used to perform cross-furnace data correlation analysis and calculate the cross-furnace correlation degree; A sampling adjustment module, used to dynamically adjust the sampling frequency of each sampling layer according to the abnormal state index, resource utilization rate and cross-furnace correlation; A data compression module, used for adaptively compressing the collected data based on the resource usage status information and data importance; The format conversion module is used to convert the compressed heterogeneous data into a unified data format.

Images