Data-driven multi-furnace continuous smelting intelligent regulation and control system and method

Abstract

The invention relates to the technical field of smelting process parameter optimization, and discloses a data-driven multi-furnace continuous smelting intelligent regulation and control system and method, and the system comprises a data collection module, a machine learning module and an intelligent control module; the method comprises the following steps: acquiring operation component data of temperature, oxygen content and gas components in the oxygen-enriched side-blowing and top-blowing furnace; constructing a continuous smelting simulation model, inputting the real-time operation component data into the continuous smelting simulation model for analogue simulation, and adjusting process parameters according to the feedback of an analogue simulation result; according to the dynamic change rate in the learning production process, process parameters are adjusted in real time; a fuzzy rule base is defined to intelligently control multiple-input-multiple-output equipment, and a control strategy is dynamically adjusted through fuzzy reasoning. According to the invention, fine control in the smelting process is realized, the production efficiency is improved, the smelting time is shortened, the dependence on operators is reduced, and the labor intensity and the labor cost are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of smelting process parameter optimization, and in particular to a data-driven multi-furnace continuous smelting intelligent control system and method. Background Art

[0002] With the rapid development of the metallurgical industry, traditional smelting processes face many challenges in production efficiency, energy utilization and pollution control. The "side-blowing + top-blowing" multi-furnace continuous smelting technology has significant advantages such as strong raw material applicability, high smelting intensity, environmental friendliness, low labor intensity and low production cost, accounting for 60% of the new non-ferrous smelting capacity in the industry. Nowadays, several major domestic non-ferrous design institutes, such as Enfei, Ruilin, Changsha Design Institute, etc., are actively carrying out the design and development of "side-blowing + top-blowing" continuous smelting furnaces, and have built many production lines in China. At the same time, there are many 300,000 to 400,000 tons of non-ferrous smelting production lines built or under construction in China, such as Yunnan Copper Southwest Copper, Jinchuan Fangchenggang, Chifeng Jintong, Nanguo Copper, etc., indicating that this advanced smelting technology will play a more important role in the future non-ferrous smelting industry.

[0003] In the actual production process, the dynamic changes of parameters such as furnace temperature, gas composition, and molten pool state make the real-time optimization and regulation of process parameters the key to improving production efficiency, reducing energy consumption, and ensuring product quality. However, the existing smelting production mainly relies on the experience of operators and distributed control systems, resulting in instability and lag in the control method. Especially in the complex and dynamic multi-furnace continuous smelting system interconnected by chutes, it is difficult for traditional automatic control systems to implement intelligent analysis and rapid feedback for real-time working condition data changes, so as to effectively respond to the dynamic changes of process conditions. Therefore, there is an urgent need for a new method based on data-driven and intelligent control, which can realize the dynamic adjustment of process parameters according to the actual production situation in the furnace, thereby improving smelting efficiency, reducing energy consumption, and reducing pollution emissions.

[0004] Prior art 1, Chinese patent, patent number: 202411237580.6, relates to a two-furnace continuous smelting system for crude copper applied to the technical field related to copper smelting, which adopts a two-furnace process with oxygen-enriched double-side blowing smelting furnace and top-blowing converting furnace for continuous smelting of crude copper, cancels the double-side blowing slag making furnace, shortens the process flow, and improves the three-step crude copper smelting process to a two-step crude copper smelting process, thereby reducing the amount of water jacket, thereby effectively reducing fuel consumption, and also reducing heat loss; in addition, in the top-blowing converting furnace, by setting a reversing gas conversion unit, the traditional direct blowing method can be changed during the gas supply process, so that the air intake direction is radiated outward, and the overall radiation direction of the gas is not fixed, so that the airflow in the top-blowing converting furnace is turbulent. Although compared with the prior art, it can greatly improve the full oxidation mixing of oxygen and materials in the top-blowing converting furnace, thereby effectively accelerating the reaction speed, and can also reduce the gas supply time, so that the heat loss is reduced; however, there is no real-time monitoring and automatic adjustment of the furnace, and the stability of the composition and the quality of the product cannot be guaranteed.

[0005] Prior art 2, Chinese patent, patent number: 202410082502.7 discloses a bottom feeding tin smelting system, including a top-blowing lance, a cylindrical smelting furnace is provided at the lower end of the top-blowing lance, the top-blowing lance is perpendicular to the furnace body and extends into the bottom of the cylindrical smelting furnace along the center position, the outside of the cylindrical smelting furnace is penetrated by a furnace side sample rod, the furnace side sample rod forms an angle of ° with the wall surface, and extends toward the center line position of the cylindrical smelting furnace to below the slag line, and a top material inlet is provided on one side of the top of the cylindrical smelting furnace. The present invention can be used to screen tin-containing dry materials, so that materials with larger particle sizes are directly conveyed to the top-blown furnace molten pool for smelting through a belt, while finer tin-containing materials are sprayed from the bottom of the molten pool through a conveying pipe to reduce splashing or leakage losses during transportation. At the same time, mechanical inclusions or flying losses caused when the top-blown furnace is put into use are prevented. Although the smelting energy consumption and flue gas emissions can be reduced by avoiding the addition of moisture to fine-grained materials before entering the furnace, a more environmentally friendly and energy-saving smelting process can be achieved; however, there is no real-time monitoring and automatic adjustment of the furnace, resulting in high injection and fuel consumption.

[0006] Prior art three, Chinese patent, patent number: 202310468112.9, relates to a process for reducing the tin grade of tin smelting rotary kiln bag dust, belonging to the technical field of non-ferrous metal metallurgical equipment and methods. In the present invention, various mixed materials are put into the hopper and enter the rotary kiln through the feeding pipe. After high-temperature roasting in the rotary kiln, roasting slag containing arsenic ≤1.5% is produced for top-blown furnace and electric furnace treatment. The flue gas enters the electric dust collector from the kiln tail smoke pipe, and the tin and other valuable metals in the high-temperature flue gas are captured by the electric dust collector and returned for roasting. The high-temperature flue gas enters the bag through the environmental protection air duct, and is captured by the bag. The flue gas containing tin ≤1% and arsenic ≥73% can be collected and sent to the arsenic treatment system for harmless treatment. Although the real open circuit of arsenic is achieved, the impact of arsenic on the tin smelting system is greatly reduced, and the economic benefits of tin smelting are improved; however, there is no real-time monitoring and feedback adjustment in the furnace, resulting in high energy waste and low combustion rate.

[0007] At present, the existing technologies 1, 2 and 3 have the problem of real-time monitoring, automatic adjustment and feedback adjustment in the furnace. To solve the above problems, the present invention provides a data-driven multi-furnace continuous smelting intelligent control system and method. Summary of the invention

[0008] The main purpose of the present invention is to provide a data-driven multi-furnace continuous smelting intelligent control system and method to solve the problems of real-time monitoring, automatic adjustment and feedback adjustment in the furnace in the prior art.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A data-driven multi-furnace continuous smelting intelligent control system, comprising:

[0011] The data acquisition module is used to collect the operating composition data of the temperature, oxygen content and gas composition in the oxygen-enriched side-blowing furnace; perform preprocessing operations such as data cleaning, noise and outlier removal, and missing value filling on the collected operating composition data; and perform time series alignment on the preprocessed operating composition data to form an operating composition data set;

[0012] The machine learning module is used to retrieve historical production data and real-time collected operating composition data, build a continuous smelting simulation model, and divide the data set into a training set, a validation set, and a test set; use the training set to train the continuous smelting simulation model, and input the real-time collected operating composition data into the continuous smelting simulation model for simulation;

[0013] The intelligent control module is used to analyze the simulation results, compare the analysis results with the preset values, and adjust the process parameters according to the comparison results; according to the real-time production data feedback, adjust the process parameters for the data greater than the preset deviation value in the smelting process; and upload the adjusted parameters to the intelligent control cloud platform.

[0014] As a further improvement of the present invention, the data acquisition module includes:

[0015] The real-time acquisition submodule is used to collect the operating composition data of the temperature, oxygen content and gas composition in the oxygen-enriched side-blown furnace; the collected operating composition data is input into the edge computing submodule 12 for data cleaning;

[0016] The edge computing submodule is used to remove noise and outliers from the collected operating component data, remove data points using the threshold method, complete missing data, and perform data analysis and feature extraction on the pre-processed operating component data.

[0017] The time series submodule is used to dynamically time-warp the running component data after feature extraction to obtain time series running component data; and to refine and decompose the time series running component data to obtain different feature levels.

[0018] As a further improvement of the present invention, the edge computing submodule includes:

[0019] Data cleaning unit is used to remove noise and outliers. If the collected data is x1, x2, ..., x n , use the threshold method to eliminate all l -μ|>kσ data points, if a data point meets the above conditions, then the point is regarded as an outlier and is removed;

[0020] The data completion unit is used to process missing data caused by process fluctuations and environmental changes. It processes the data by setting the system's high-order spline interpolation, prior distribution p(θ) and likelihood function p(D|θ) of the observed data to calculate the missing data value.

[0021] The dimension reduction extraction unit is used to perform data analysis and feature extraction on the completed data, centralize the data, calculate its covariance matrix to obtain its eigenvalues ​​and eigenvectors, and sort its feature data by importance.

[0022] As a further improvement of the present invention, the data completion unit includes:

[0023] The interpolation processing subunit is used to deal with the data missing problem in the data collection of the smelting process, and uses the high-order spline interpolation method to process the interpolation;

[0024] The data completion subunit is used to handle missing data due to process fluctuations and environmental changes. It sets the system's prior distribution p(θ) and defines the likelihood function p(D|θ) of the observed data to describe the probability of the data observed under given parameter conditions. The prior distribution is combined with the likelihood function using Bayes' theorem to obtain the posterior distribution p(θ|D).

[0025] The prediction mean subunit is used to calculate the predicted distribution p(y) of missing values ​​based on the posterior distribution. * ∣x * , D); confidence intervals are calculated based on the predicted means.

[0026] As a further improvement of the present invention, the dimension reduction extraction unit comprises:

[0027] The principal component dimensionality reduction subunit is used to set the running component data matrix to be X∈R m×n ; Where m is the number of samples and n is the number of features; perform central processing on the running component data and remove the mean of each column; calculate the co-matrix of the running component data and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors; select the eigenvectors corresponding to the first k largest eigenvalues ​​to form a dimensionality reduction matrix W∈R n×k , get the final running component data;

[0028] The independent component analysis subunit is used to collect the observed data signal represented as a linear combination of several independent components. Suppose the observed data X∈R m×n , obtain at least one separation matrix W, such that S = WX; where S is an independent component matrix, satisfying the statistical independence between each column;

[0029] The feature sorting subunit is used to evaluate the contribution of each feature on all tree splitting nodes through the random forest method of the tree model, and quantify the influence of each feature on the output of the tree model; sort the features according to their importance, quantify the influence of the features on the tree model, and eliminate unimportant features.

[0030] As a further improvement of the present invention, the time series submodule includes:

[0031] A dynamic time unit is used to compare different time series of running component data with each other, calculate according to the comparison results, and obtain at least one alignment path; calculate multiple alignment paths to obtain the optimal alignment path;

[0032] Multi-scale time unit, used to introduce the running component data of the optimal alignment path into the multi-scale function φ(t) and the detail function ψ(t); decompose the time series to obtain the dynamic change characteristics of different time scales;

[0033] The wavelet transform unit is used to refine and decompose the time series using discrete wavelet transform to obtain different feature levels and decompose the time series into approximation and detail parts of different scales.

[0034] As a further improvement of the present invention, the dynamic time unit comprises:

[0035] The sequence alignment subunit is used to align different time series of running component data X = [x1, x2, ..., x n ] and Y = [y1, y2, ..., y m ] to calculate and obtain at least one alignment path π = [(i1, j1), (i2, j2), ..., (i K , j K )], the total alignment cost of the two time series is minimized;

[0036] The optimal alignment path subunit 1312 is used to calculate the optimal path and construct the cumulative cost matrix D using a dynamic programming method. i，j , the minimum cumulative cost from the starting position (i, j) of the sequence;

[0037] The minimization subunit is used to obtain the optimal alignment path of sequences X and Y by minimizing the cumulative cost matrix and eliminating the nonlinear deviation of the timeline.

[0038] As a further improvement of the present invention, the machine learning module includes:

[0039] The data processing submodule is used to extract key indicators from historical production data and real-time collected operating component data; clean the collected data, remove noise and outliers, fill in missing values, and perform time series alignment;

[0040] The feature extraction submodule is used to extract meaningful features from the preprocessed data; select a machine learning model based on the characteristics of the continuous smelting process and build a continuous smelting simulation model; divide the data set into a training set, a validation set, and a test set, use the training set to train the model, and adjust the model parameters through the validation set;

[0041] The model verification submodule is used to input the real-time collected operating composition data into the trained continuous smelting simulation model, perform simulation, and generate prediction results.

[0042] As a further improvement of the present invention, the intelligent control module includes:

[0043] The forward feedback control submodule is used to adjust key process parameters according to the simulation results in the continuous smelting simulation model; adjust the deviations in the smelting process through real-time data feedback; and perform local feedback regulation through temperature control and oxygen content control;

[0044] The adaptive control submodule is used to learn the dynamic change rules in the production process through neural networks and dynamically adjust the process parameters according to the real-time process conditions; the adaptive algorithm is performed using the adaptive law based on the Lyapinov stability theory;

[0045] The fuzzy logic submodule is used to define the fuzzy rule base based on the complexity of multi-variable coupling in the smelting process, and to intelligently control the multi-input and multi-output equipment; the control strategy is dynamically adjusted through the fuzzy reasoning module;

[0046] The clustering fuzzy unit is used to determine the intelligent decision sequence based on the responsibility of multi-variable coupling in the smelting process; based on the preset similarity, the intelligent decision sequence is clustered and the decision fuzzification is performed for each cluster;

[0047] The fuzzy rule base unit is used to mine fuzzy logic rules based on logical necessity and logical commonality; there is a one-to-one correspondence between fuzzy logic rules and clusters; fuzzification and fuzzy decision analysis are performed on data;

[0048] The response unit is used to determine the fuzzy strategy according to the analysis results and perform fuzzification, determine the control strategy, obtain the process data of the smelting process output by the strategy based on the response terminal of the decision-making task, and perform technical architecture conversion and control response of the response terminal of the control strategy;

[0049] Determine a fuzzy processing method, wherein the fuzzy processing method includes data sharpening processing and data passivation processing; based on the fuzzy processing method, establish a bidirectional fuzzification block, wherein the bidirectional fuzzification block includes a fuzzy processing branch and a defuzzification branch; based on the fuzzy processing branch, perform fuzzy processing on the measurement series of each cluster;

[0050] Traverse the fuzzy logic rules, supervise the decision domain, and the decision domain corresponds to the fuzzy logic rules one by one; based on the decision frequency and decision operator requirements, integrate the decision domain and allocate resources, determine the fuzzy decision block, where the fuzzy decision block has extensibility and updateability, determine the entry and exit ports of the fuzzy processing branch and the output port of the defuzzification branch, and connect the fuzzy decision block in the middle;

[0051] A unified information database is established, wherein the unified information database is a conversion database of the process parameters of the oxygen-enriched side-blowing furnace and the unified process parameters; based on the response terminal, the unified information database is traversed to determine the data for adjusting the process parameters of the oxygen-enriched side-blowing furnace, and the data is converted to the control decision strategy to determine the terminal control strategy; the control data stream is obtained, and a bias decision analysis is performed on the control data stream, and the bias decision is free of the preset tolerance interval; the first bias decision is made based on the node control difference, the second bias decision is made based on the control trend bias, and the third bias decision is made based on the response delay. If the preset tolerance interval of the mapping is not met, feedback control management is performed.

[0052] To achieve the above object, the present invention also provides the following technical solutions:

[0053] A data-driven intelligent control method for continuous smelting of multiple furnaces, comprising:

[0054] Collect the operating composition data of the temperature, oxygen content and gas composition in the oxygen-enriched side-blown furnace; perform preprocessing operations such as data cleaning, noise and outlier removal, and missing value filling on the collected operating composition data; perform time series alignment on the preprocessed operating composition data to form an operating composition data set;

[0055] Retrieve historical production data and real-time collected operating composition data, build a continuous smelting simulation model, and divide the data set into a training set, a validation set, and a test set; use the training set to train the continuous smelting simulation model, and input the real-time collected operating composition data into the continuous smelting simulation model for simulation;

[0056] Analyze the simulation results, compare the analysis results with the preset values, and adjust the process parameters according to the comparison results; adjust the process parameters for data greater than the preset deviation value in the smelting process based on real-time production data feedback; upload the adjusted parameters to the intelligent control cloud platform.

[0057] The present invention realizes refined control of the smelting process through intelligent optimization of process parameters, thereby significantly improving production efficiency and reducing smelting time; through real-time monitoring and feedback adjustment, the air volume and heating intensity can be accurately controlled according to the temperature and reaction conditions in the furnace, reducing energy waste and improving combustion efficiency; through real-time monitoring and automatic adjustment of the components in the furnace, the stability of the molten pool composition is ensured, thereby improving the quality of the final product; the use of intelligent control reduces excessive blowing and fuel consumption, reduces exhaust gas emissions, and is more environmentally friendly; the present invention realizes automated control of the smelting process, reduces dependence on operators, and reduces labor intensity and labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1It is a functional module schematic diagram of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0059] Figure 2 A system schematic diagram of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0060] Figure 3 A functional module schematic diagram of a data acquisition module of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0061] Figure 4 This is a functional module schematic diagram of an edge computing submodule of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0062] Figure 5 It is a functional module schematic diagram of a data completion unit of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0063] Figure 6 It is a functional module schematic diagram of a dimensionality reduction extraction unit of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0064] Figure 7 It is a functional module schematic diagram of a time series submodule of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0065] Figure 8 It is a functional module schematic diagram of a dynamic time unit of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0066] Fig. 9 A schematic diagram of the functional modules of a machine learning module of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0067] Fig.10 It is a functional module schematic diagram of an intelligent control module of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0068] Fig.11 The specific principle of the intelligent control system of the data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0069] Fig.12 It is a specific schematic diagram of the reverse control module of the data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0070] Fig.13 The specific principle of the adaptive control module of the data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0071] Fig.14 It is a functional module schematic diagram of a fuzzy logic submodule of an embodiment of a data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0072] Fig.15 The specific principle of the fuzzy logic submodule of the data-driven multi-furnace continuous smelting intelligent control system of the present invention;

[0073] Fig.16 A schematic diagram of the steps of an embodiment of the data-driven multi-furnace continuous smelting intelligent control method of the present invention;

[0074] Fig.17 It is a structural schematic diagram of an embodiment of an electronic device of the present invention;

[0075] Fig.18 It is a schematic structural diagram of an embodiment of the storage medium of the present invention. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] The terms "first", "second" and "third" in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0078] Reference to an "embodiment" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0079] like Figure 1 As shown, this embodiment provides an embodiment of a data-driven multi-furnace continuous smelting intelligent control system. In this embodiment, the data-driven multi-furnace continuous smelting intelligent control system specifically includes:

[0080] Data acquisition module 1 is used to collect the operating composition data of temperature, oxygen content and gas composition in the oxygen-enriched side-blowing and top-blowing furnace; perform preprocessing operations such as data cleaning, noise and outlier removal, and missing value filling on the collected operating composition data; perform time series alignment on the preprocessed operating composition data to form an operating composition data set;

[0081] Machine learning module 2 is used to retrieve historical production data and real-time collected operating composition data, build a continuous smelting simulation model, and divide the data set into a training set, a validation set, and a test set; use the training set to train the continuous smelting simulation model, and input the real-time collected operating composition data into the continuous smelting simulation model for simulation;

[0082] The intelligent control module 3 is used to analyze the simulation results, compare the analysis results with the preset values, and adjust the process parameters according to the comparison results; according to the real-time production data feedback, adjust the process parameters for the data greater than the preset deviation value in the smelting process; and upload the adjusted parameters to the intelligent control cloud platform.

[0083] Preferably, in this embodiment, the data acquisition module 1 is responsible for collecting operating component data such as temperature, oxygen content and gas composition from the oxygen-enriched side-blown top-blown furnace; obtaining these data through sensors or other means, and converting them into digital signals for subsequent processing; having the characteristics of real-time, reliability and high efficiency, being able to read multiple different types of sensor data at the same time, and filtering, amplifying, separating and other processing of the signals to ensure the accuracy and consistency of the data. The machine learning module 2 constructs a continuous smelting simulation model by retrieving historical production data and real-time collected operating components, trains the model using a training set, and inputs real-time data into the model for simulation, thereby predicting and optimizing the smelting process; using the data processing tools of universities, pre-processing operations such as cleaning, missing value processing, outlier detection and processing, and data type conversion are performed on the data to improve the availability of data quality; the data set is divided into a training set, a validation set and a test set to provide support for model training and evaluation. The intelligent control module 3 is used to analyze the simulation results, compare the analysis results with the preset values, adjust the process parameters according to the comparison results, adjust the deviation values ​​in the smelting process according to the real-time data feedback, and upload the adjusted parameters to the intelligent control cloud platform; use fuzzy logic control to handle the complexity of multi-variable coupling in the smelting process, and realize intelligent control of multi-input and multi-output by defining a fuzzy rule database, and use fuzzy reasoning to dynamically adjust the control strategy, which can automatically identify and adapt to different industrial process parameters, realize more accurate and efficient control, and can monitor and adjust various parameters in the production process in real time to improve production efficiency and product quality. (Refer to the attached schematic for specific schematic diagram) Figure 2 )

[0084] Further, if Figure 3 As shown, the data acquisition module 1 specifically includes:

[0085] The real-time acquisition submodule 11 is used to collect the operating composition data of the temperature, oxygen content and gas composition in the oxygen-enriched side-blowing and top-blowing furnace; the collected operating composition data is input into the edge computing submodule 12 for data cleaning;

[0086] The edge computing submodule 12 is used to remove noise and outliers from the collected operating component data, and remove data points using a threshold method; to complete missing data; and to perform data analysis and feature extraction on the preprocessed operating component data;

[0087] The time series submodule 13 is used to dynamically time-warp the running component data after feature extraction to obtain time series running component data; and to refine and decompose the time series running component data to obtain different feature levels.

[0088] Preferably, the data acquisition submodule 11 of this embodiment is responsible for collecting the operating data such as temperature, oxygen content and gas composition in the oxygen-enriched side-blowing and top-blowing furnace, ensuring the comprehensiveness and accuracy of the data, transmitting the operating data of the equipment to the system in real time through sensors and data acquisition devices, and supporting multiple outgoing protocols and compatible with different types of detection equipment. The edge computing submodule 12 preprocesses the collected operating component data, including removing noise and outliers, data completion, data analysis and feature extraction, so as to improve data quality and availability; using the threshold method to remove data points, using the interpolation method or the intelligent filling method to process missing values, and using data analysis and feature extraction technology to provide clean and high-quality data for subsequent time series analysis. The time series submodule 13 dynamically time-warps the operating component data after feature extraction to obtain time series operating component data, and further refines and decomposes it to obtain different feature levels, thereby supporting in-depth data analysis and decision-making, and using time series analysis technology to dynamically warp and refine the data to reveal potential rules and trends in the data. This embodiment achieves efficient monitoring and optimization of complex industrial processes through real-time data acquisition, edge computing preprocessing and time series analysis.

[0089] Further, if Figure 4 As shown, the edge computing submodule 12 specifically includes:

[0090] The data cleaning unit 121 is used to remove noise and outliers. If the collected data is x1, x2, ..., x n , use the threshold method to eliminate all i -μ|>kσ data points, if a data point meets the above conditions, then the point is regarded as an outlier and is removed;

[0091] Among them, μ is the data mean; σ is the standard deviation; k is a constant with a value of 2, which indicates the multiple of the outlier point deviating from the mean;

[0092] The data completion unit 122 is used to process missing data due to process fluctuations and environmental changes, and processes the data by setting the system's high-order spline interpolation, prior distribution p(θ) and likelihood function p(D|θ) of the observed data to calculate the missing data value;

[0093] The dimension reduction extraction unit 123 is used to perform data analysis and feature extraction on the completed data, centralize the data, calculate its covariance matrix to obtain its eigenvalues ​​and eigenvectors, and sort its feature data by importance.

[0094] Preferably, the data cleaning unit 121 in this embodiment removes noise and outliers to improve data quality; the threshold method is used to eliminate outliers. A threshold is defined by calculating the mean and standard deviation of the data, and data points that deviate from the mean by more than a set multiple (such as 2 standard deviations) are regarded as outliers and eliminated. The data completion unit 122 processes data missing due to process fluctuations and environmental changes to ensure data integrity; the data is processed through high-order spline interpolation, prior distribution, and the likelihood function of the observed data to calculate the missing values. The dimensionality reduction and extraction unit 123 performs data analysis and feature extraction on the completed data, simplifies the data structure, and improves the analysis efficiency; first, the data is centralized, the covariance matrix is calculated, its eigenvalues and eigenvectors are obtained, and the feature data is sorted by importance; this embodiment is implemented through principal component analysis (PCA), which can effectively extract the main change directions of the data, reduce the data dimension, and retain key information at the same time. This embodiment ensures the quality improvement of the data in the preprocessing stage and provides high-quality data support for subsequent data analysis and modeling.

[0095] Further, as Figure 5 shown, the data completion unit 122 specifically includes:

[0096] The interpolation processing subunit 1221 is used to process data missing problems that occur in the data acquisition during the smelting process, and processes the interpolation using the high-order spline interpolation method;

[0097] Among them, the form of the high-order spline interpolation is as follows:

[0098] S i (x) = a i + b i *(x - x i ) + c i ·(x - x i ) 2 + d i ·(x - x i ) 3

[0099] Among them, the coefficients a i , b i , c i , d i are determined by natural boundary conditions and smoothness conditions, that is, S'(x n ) = 0 and S”(x n ) = 0;

[0100] The data completion subunit 1222 is used to handle data missing due to process fluctuations and environmental changes, set the prior distribution p(θ) of the system, and define the likelihood function p(D∣θ) of the observed data, which describes the probability of the observed data under given parameters; use Bayes' theorem to combine the prior distribution and the likelihood function to obtain the posterior distribution p(θ∣D).

[0101] Among them, the probability of the observed data under given parameters is as follows:

[0102]

[0103] Among them, y i is the observed value, f(x i , θ) is the prediction of the model for the data point x i , and σ is the noise standard deviation.

[0104] The prediction mean subunit 1223 is used to calculate the prediction distribution p(y * ∣x * , D) of the missing value according to the posterior distribution; calculate the confidence interval according to the predicted mean.

[0105] Among them, the confidence interval calculated according to the predicted mean is as follows:

[0106] p(y * |x * , D) = ∫p(y * |x * , θ)p(θ|D)dθ′

[0107] Among them, y * is the prediction result of the missing value, and x * is the position of the point to be interpolated.

[0108] Preferably, the interpolation processing subunit 1221 of this embodiment uses a high-order spline interpolation method to interpolate data. High-order spline interpolation processes the data in segments, uses low-order polynomials (usually quadratic or cubic) to fit the data points, and ensures the continuity and smoothness of the polynomials between the segments. The advantage of this method is that it can smoothly capture the trend of complex data, but the computational complexity is high, especially for large data sets. In addition, cubic spline interpolation can improve the interpolation effect and ensure smooth transitions between adjacent splines by matching the slope and depression of each interpolation segment on the internal data points. The data completion subunit 1222 uses a Bayesian method to deal with missing data caused by process fluctuations and environmental changes. The method first sets the system's prior distribution p(θ) and defines the likelihood function p(D|θ) of the observed data to describe the probability of the data observed under given parameter conditions; the prior distribution is combined with the likelihood function using Bayes' theorem to obtain the posterior distribution p(θ|D), thereby calculating the predicted distribution p(y*|x*,D) of the missing value. The preprocessing mean subunit 1223 calculates the predicted distribution p(y*|x*, D) of the missing value according to the posterior distribution, and calculates the confidence interval according to the predicted mean. This embodiment ensures the reliability and accuracy of the prediction results, so that the system can make reasonable decisions in the face of uncertainty.

[0109] Further, if Figure 6 As shown, the dimension reduction extraction unit 123 specifically includes:

[0110] The principal component dimension reduction subunit 1231 is used to set the running component data matrix to be X∈R m×n ; Where m is the number of samples and n is the number of features; perform central processing on the running component data and remove the mean of each column; calculate the co-matrix of the running component data and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors; select the eigenvectors corresponding to the first k largest eigenvalues ​​to form a dimensionality reduction matrix W∈R n×k , get the final running component data;

[0111] Among them, removing the mean of each column is as follows:

[0112] X centered =X-μ

[0113] Among them, μ is the mean vector of each column feature;

[0114] Among them, the co-matrix for calculating the running component data is as follows:

[0115]

[0116] Among them, the eigenvalues ​​and eigenvectors are obtained as follows:

[0117] ∑=PΛP T

[0118] Among them, Λ is the eigenvalue matrix and P is the eigenvector matrix;

[0119] Among them, the final running component data is obtained as follows:

[0120] Z = X centered W

[0121] The independent component analysis subunit 1232 is used to collect the observed data signal, which is expressed as a linear combination of several independent components. Let the observed data X ∈ R m×n , and obtain at least one separation matrix W such that S = WX; where S is the independent component matrix, and each column has statistical independence;

[0122] The feature ranking subunit 1233 is used to evaluate the contribution of each feature at all tree splitting nodes through the random forest method of the tree model, and quantitatively determine the influence degree of each feature on the output of the tree model;

[0123] The importance of feature j is defined as follows:

[0124]

[0125] Among them, T is the total number of trees, St is the set of all nodes in the t-th tree, and Δi(s) is the change in information gain before and after node s is split. is the exponential function, indicating that node s is split using feature j;

[0126] According to the feature importance ranking, quantify the influence of features on the tree model, and eliminate unimportant features.

[0127] Preferably, in this embodiment, the principal component dimension reduction subunit 1231 removes the mean of each column so that the mean of each column of the data matrix is ​​zero, thereby eliminating the offset in the data; the covariance matrix is ​​calculated for the centralized data, and the covariance matrix describes the correlation between different features in the data and is the basis of principal component analysis; the covariance matrix is ​​subjected to eigenvalue decomposition to obtain eigenvalues ​​and eigenvectors; the eigenvalue represents the amount of variance explained by the principal component, and the eigenvector indicates the direction of the principal component; the eigenvectors corresponding to the first k largest eigenvalues ​​are selected according to the size of the eigenvalue, forming the base layer of the new low-dimensional space, thereby realizing dimensionality reduction, and projecting the original data into the new space formed by the selected eigenvectors to obtain the data after dimensionality reduction. The independent component analysis subunit 1232 represents the observed data as a linear combination of several independent components, and converts the observed data into an independent component matrix S through a separation matrix W, wherein each column has statistical independence, and at least one separation matrix W is obtained through an independent component analysis algorithm, so that the observed data can be represented as a linear combination of independent components. The feature sorting subunit 1233 evaluates the contribution of each feature on all splitting nodes through the random forest method, and quantifies the influence of each feature on the output of the tree model; defines the importance of feature j by calculating the change in information gain before and after the node split, and uses an exponential function to represent the importance of node s using feature j for splitting; sorts the features according to their importance, and eliminates unimportant features that have little impact on the model, thereby optimizing model performance.

[0128] Further, if Figure 7 As shown, the time series submodule 13 specifically includes:

[0129] The dynamic time unit 131 is used to compare different time series of the running component data with each other, calculate according to the comparison result, and obtain at least one alignment path; calculate multiple alignment paths to obtain the optimal alignment path;

[0130] The multi-scale time unit 132 is used to introduce the running component data of the optimal alignment path into the multi-scale function φ(t) and the detail function ψ(t); decompose the time series to obtain dynamic change characteristics of different time scales;

[0131] Among them, the time series is decomposed as follows:

[0132]

[0133] Among them, the time series is X(t), Aj(t) represents the approximation component at the jth scale, which contains low-frequency information; D j (t) represents the detail component at the jth scale, which contains high-frequency information;

[0134] The wavelet transform unit 133 is used to refine and decompose the time series using discrete wavelet transform to obtain different feature levels and decompose the time series into approximation and detail parts of different scales;

[0135] Among them, the discrete wavelet changes in the following form:

[0136]

[0137] Among them, ψ j，k (t) is the wavelet basis function, and the mother wavelet (t) is generated by scaling and unique operations;

[0138] Among them, the mother wavelet (t) is generated by scaling and unique operations as follows:

[0139] ψ j，k (t) = 2 -j / 2 ψ(2 -j tk).

[0140] Preferably, the dynamic time unit 131 in this embodiment is used to compare different time series of operating component data with each other, calculate according to the comparison results, and obtain at least one alignment path; calculate multiple alignment paths to obtain the optimal alignment path; compare different time series of operating component data with each other, calculate according to the comparison results, and obtain at least one alignment path; align different time series data to ensure that the corresponding relationship on the time axis is consistent, and improve the accuracy and consistency of time series data by calculating multiple alignment paths and selecting the optimal path; the multi-scale time unit 132 introduces the operating component data of the optimal alignment path into the multi-scale function and the detail function, decomposes the time series, and obtains the dynamic change characteristics of different time scales. The multi-scale decomposition method of this embodiment can capture the characteristics of the time series at different scales, including low-frequency and high-frequency information; the multi-scale time unit 132 decomposes the time series into approximation components and detail components through wavelet transform, revealing the dynamic change characteristics of the time series at different time scales. The wavelet transform unit 133 uses discrete wavelet transform to refine the time series, obtain different feature levels, and decompose the time series into approximation and detail parts of different scales; discrete wavelet transform gradually decomposes the time series into low-frequency parts and high-frequency parts, thereby extracting features at different scales.

[0141] Further, if Figure 8 As shown, the dynamic time unit 131 specifically includes:

[0142] The sequence alignment subunit 1311 is used to align different time series of running component data X=[x1, x2, ..., x n ] and U = [y1, y2, ..., y m] to calculate and obtain at least one alignment path π = [(i1, j1), (i2, j2), ..., (i K , j K )], the total alignment cost of the two time series is minimized;

[0143] The optimal alignment path subunit 1312 is used to calculate the optimal path and construct the cumulative cost matrix D using a dynamic programming method. i，j , the minimum cumulative cost from the starting position (i, j) of the sequence;

[0144] Among them, the calculation of the optimal distance is defined as follows:

[0145]

[0146] in, It's time point and The local distance between them, K is the length of their path;

[0147] Among them, the minimum cumulative cost from the starting position (i, j) of the sequence is as follows:

[0148] D i，j =d(x i ,y j )+min{D i-1，i , D i，i-1 , D i-1,i-1}

[0149] The boundary conditions are as follows:

[0150] D 0，0 =0,D i，0 =D 0，j =∞(when i, j>0);

[0151] The minimization subunit 1313 is used to obtain the best alignment path of the sequences X and Y by minimizing the cumulative cost matrix, thereby eliminating the nonlinear deviation of the timeline.

[0152] Preferably, in this embodiment, the sequence alignment subunit 1311 is used to calculate the different time series of the running Chen Fen data, and obtain at least one alignment path, which is the minimum total alignment cost of the two time series, find the best match between the different sequence lengths or the existence of time distortion, construct the cumulative cost matrix through the dynamic programming method, calculate the local distance between each time point, and use these local distances to determine the global optimal alignment path. The optimal alignment path subunit 1312 is responsible for calculating the optimal path, using the dynamic programming method to construct the cumulative cost matrix, starting from the starting position of the sequence, find the path with the minimum cumulative cost, and even in the case of time distortion, find the best matching path. Through the dynamic programming algorithm, the cumulative distance matrix is ​​constructed from the upper left corner to the lower right corner to calculate the minimum cumulative distance to each store, and find the optimal path by backtracking. The minimization subunit 1313 obtains the optimal alignment path of the sequence X and Y by minimizing the matrix of the cumulative cost, eliminating the nonlinear deviation of the timeline; using the dynamic programming technology, backtracking from the upper right corner of the matrix, selects the path with the minimum cumulative cost, and ensures that the path meets the boundary conditions, continuity and monotonicity constraints.

[0153] Furthermore, if Fig. 9 As shown, the machine learning module 2 specifically includes:

[0154] The data processing submodule 21 is used to extract key indicators such as temperature, oxygen content, gas composition, etc. from historical production data and real-time collected operating component data; clean the collected data, remove noise and outliers, fill in missing values, and perform time series alignment;

[0155] The feature extraction submodule 22 is used to extract meaningful features from the preprocessed data, such as temperature change rate, oxygen content fluctuation, etc., to enhance the prediction ability of the model; select a machine learning model according to the characteristics of the continuous smelting process, and build a continuous smelting simulation model; divide the data set into a training set, a validation set, and a test set, use the training set to train the model, and adjust the model parameters through the validation set;

[0156] The model verification submodule 23 is used to input the real-time collected operating composition data into the trained continuous smelting simulation model, perform simulation, and generate prediction results; perform detailed analysis on the simulation results to evaluate the accuracy and stability of the model.

[0157] Preferably, the data processing submodule 21 of this embodiment ensures the quality and consistency of the input data by removing noise and outliers and completing missing data; ensures that the data at different time points can accurately correspond to each other, and provides a reliable basis for feature extraction and model training. Significance: High-quality data is the basis for the accuracy of machine learning models. The data processing submodule ensures the reliability of input data and provides a solid foundation for analysis and prediction; by processing outliers and missing data, it reduces interference factors in the model training process and enhances the stability and robustness of the model. The feature extraction submodule 22 extracts meaningful features from the original data, such as temperature change rate, oxygen content fluctuation, etc. These features can better reflect the dynamic changes of the smelting process; according to the characteristics of the smelting process, select a suitable machine learning model, and construct a continuous smelting simulation model to ensure that the model can accurately simulate the actual production process; divide the data set into a training set, a validation set, and a test set, train the model through the training set, and adjust the model parameters through the validation set to ensure that the model performs consistently on different data sets. Significance: Through feature extraction, the model can better capture the key changes in the smelting process, thereby improving the accuracy and reliability of the prediction; through reasonable data set division and model training, the model's performance on different data sets is ensured to be consistent, thereby improving the generalization ability and practical application value of the model. The model verification submodule 23 inputs the real-time collected operating component data into the trained model, performs simulation, and generates prediction results; the simulation results are analyzed in detail to evaluate the accuracy and stability of the model to ensure the reliability of the model in practical applications. Significance: Through simulation and result analysis, the effectiveness of the model in the actual production environment is verified to ensure that the model can accurately predict changes in the smelting process; through continuous verification and evaluation, the deficiencies of the model are discovered, and continuous improvements are made to improve the performance and application effect of the model.

[0158] In summary, the various submodules of this embodiment work together to improve the intelligence level of the continuous smelting process. The data processing submodule ensures the quality and consistency of the data, the feature extraction submodule improves the prediction ability of the model, and the model verification submodule ensures the reliability and effectiveness of the model in practical applications. Together, they promote the intelligence and automation of the smelting process and improve production efficiency and product quality.

[0159] Furthermore, if Fig.10 As shown, the intelligent control module 3 specifically includes:

[0160] The forward feedback control submodule 31 is used to adjust key process parameters according to the simulation results in the continuous smelting simulation model; adjust the deviation in the smelting process through real-time data feedback; and perform local feedback regulation through temperature control and oxygen content control;

[0161] The adaptive control submodule 32 is used to learn the dynamic change law in the production process through a neural network and dynamically adjust the process parameters according to the real-time process conditions; and use the adaptive law based on the Lyapinov stability theory to perform an adaptive algorithm;

[0162] The fuzzy logic submodule 33 is used to define a fuzzy rule base based on the complexity of multi-variable coupling in the smelting process, and to perform intelligent control on multi-input and multi-output devices; and to dynamically adjust the control strategy through the fuzzy reasoning module.

[0163] Preferably, the pre-feedback control submodule 31 of this embodiment adjusts the key process parameters through the simulation results in the continuous smelting simulation model, and uses real-time data feedback to adjust the deviations in the smelting process. In addition, it also performs local feedback regulation through temperature control and oxygen content control; this embodiment adjusts the process parameters through a pre-calculated model to reduce the deviations in actual operation. This method can improve the response speed and accuracy of the system and is suitable for industrial processes that require rapid adjustment. The adaptive control submodule 32 learns the dynamic change law in the production process through a neural network, and dynamically adjusts the process parameters according to real-time process conditions; an adaptive algorithm based on the adaptive law of Lyapunov stability theory is used to ensure that the system remains stable in the face of uncertainty and changes; adaptive control has strong flexibility and robustness, and can automatically adjust the controller parameters to cope with environmental changes or changes in system internal parameters; this control method is particularly suitable for complex dynamic systems in the metallurgical industry, and can significantly improve the stability and adaptability of the system. The fuzzy logic submodule 33 defines a fuzzy rule base based on the complexity of multi-variable coupling in the smelting process, and intelligently controls multi-input and multi-output devices. The control strategy is dynamically adjusted through the fuzzy reasoning module to cope with the uncertainty and nonlinear problems of the system. Through the fuzzy rule base and fuzzy reasoning mechanism, complex control problems are simplified into an easy-to-handle form, thereby achieving effective control of multivariable systems. (For the specific principles of intelligent control systems, please refer to the attached Fig.11 ; For the specific schematic diagram of the anti-control control module, please refer to the attached Fig.12 ; For the specific principle of the adaptive control module, please refer to the attached Fig.13 )

[0164] Furthermore, if Fig.14 As shown, the fuzzy logic submodule 33 specifically includes:

[0165] The clustering fuzzy unit 331 is used to determine the intelligent decision sequence based on the responsibility of multivariate coupling in the smelting process; cluster the intelligent decision sequence based on the preset similarity, and perform decision fuzzification for each cluster;

[0166] The fuzzy rule base unit 332 is used to mine fuzzy logic rules based on logic necessity and logic commonality; wherein there is a one-to-one correspondence between fuzzy logic rules and clusters; and to perform fuzzification and fuzzy decision analysis on data;

[0167] The response unit 333 is used to determine the fuzzy strategy according to the analysis results and perform fuzzification, determine the control strategy, obtain the process data of the smelting process output by the strategy based on the response terminal of the decision-making task, and perform technical architecture conversion and control response of the response terminal to the control strategy.

[0168] Preferably, the clustering fuzzy unit 331 of this embodiment determines a unique decision sequence based on the responsibility of multivariable coupling, and clusters the intelligent decision sequence based on a preset similarity; the data points are divided into several clusters, each cluster containing the most similar data points. This method allows the system to improve the accuracy and robustness of decisions by fuzzification when processing complex data. The fuzzy rule base unit 332 realizes fuzzification processing and fuzzy decision analysis of data by mining fuzzy logic rules; there is a one-to-one correspondence between fuzzy logic rules and clustering clusters, so that the system can generate corresponding control strategies according to different clustering clusters; the fuzzy logic rule base is used to automatically construct an algorithm to extract fuzzy rules from the data to achieve adaptive control of the system. The response unit 333 determines the fuzzy strategy and fuzzifies it according to the analysis results, and finally determines the control strategy; this embodiment obtains the process data of the smelting process, and converts the technical architecture and control response of the response terminal to the control strategy to ensure the efficient operation of the system; it combines fuzzy reasoning and defuzzification technology to convert fuzzy output into clear control quantities. (For specific schematic diagrams, please refer to the attached Fig.15 )

[0169] Furthermore, the clustering fuzzy unit 331 specifically includes:

[0170] Determine a fuzzy processing method, wherein the fuzzy processing method includes data sharpening processing and data passivation processing; based on the fuzzy processing method, establish a bidirectional fuzzification block, wherein the bidirectional fuzzification block includes a fuzzy processing branch and a defuzzification branch; based on the fuzzy processing branch, perform fuzzy processing on the measurement series of each cluster.

[0171] Preferably, the sharpening process of this embodiment makes the image clearer by enhancing the edges and details of the image; the passivation process suppresses the jump parts in the image through smoothing filtering to make the image smoother. The bidirectional fuzzification block realizes the encoding and decoding of image data through the steps of fuzzification and defuzzification, thereby effectively managing fuzziness and ambiguity. The fuzzification stage converts the image data from the grayscale plane to the member plane, and the defuzzification stage restores the clarity by modifying the member value. The original data is converted into a fuzzy set through a fuzzification function and processed using a fuzzy relationship. The defuzzification process obtains a clear numerical output by aggregating the output fuzzy set. Fuzzy clustering analysis divides the data set into multiple classes or clusters by minimizing the similarity between classes and maximizing the similarity within classes. The core of fuzzy clustering analysis is to construct a fuzzy matrix and determine the degree of membership, generate a fuzzy equivalence matrix through fuzzy operations and synthetic transformation, and finally classify according to the interception level.

[0172] Furthermore, the fuzzy rule unit 332 specifically includes:

[0173] Traverse the fuzzy logic rules and supervise the decision domain. The decision domain corresponds to the fuzzy logic rules one by one. Based on the decision frequency and decision operator requirements, integrate the decision domain and allocate resources to determine the fuzzy decision block, which has the characteristics of extensibility and updateability. Determine the input and output ports of the fuzzy processing branch and the output port of the defuzzification branch, and connect the fuzzy decision block in the middle.

[0174] Preferably, the present embodiment converts clear input variables into fuzzy variables so that the system can handle uncertainty and ambiguity. For example, in a fuzzy controller, fuzzification is used to determine the number of fuzzy domains covered by each fuzzy subset and the membership of each input variable after quantization factor processing. Different membership functions (such as triangular, Gaussian, etc.) are used to convert numerical inputs into fuzzy linguistic variables. This conversion allows the system to make decisions when faced with imprecise or incomplete information. The inference engine is responsible for reasoning according to the rules in the fuzzy rule base to generate fuzzy outputs; the present embodiment calculates the degree of match between the input variables and the rule premises and applies logical operations (such as "and" and "or") to derive conclusions. The fuzzy output generated by the inference engine is converted into a clear numerical output so that the actuator can actually operate according to these numerical values. The rule base stores all the fuzzy control rules required by the system, which define the relationship between input and output and guide the decision-making process of the system. By matching decision domains with fuzzy logic rules one by one, the system can flexibly handle various input situations and select appropriate control strategies based on different input combinations; the fuzzy decision block is scalable and updatable, which means that the system can dynamically adjust its rule base and parameter settings based on new requirements or data; enabling the fuzzy system to continuously improve and adapt to new application scenarios, thereby improving the overall performance and reliability of the system.

[0175] Furthermore, the response unit 333 specifically includes:

[0176] A unified information database is established, wherein the unified information database is a conversion database of the process parameters of the oxygen-enriched side-blowing and top-blowing furnace and the unified process parameters; based on the response terminal, the unified information database is traversed to determine the data for adjusting the process parameters of the oxygen-enriched side-blowing and top-blowing furnace, and the control decision strategy is converted to determine the terminal control strategy; the control data stream is obtained, and a bias decision analysis is performed on the control data stream, and the bias decision is free of the preset tolerance interval; the first bias decision is made based on the node control difference, the second bias decision is made based on the control trend bias, and the third bias decision is made based on the response delay. If the preset tolerance interval of the mapping is not satisfied, feedback control management is performed.

[0177] Preferably, this embodiment can accurately convert the process parameters of the oxygen-enriched side-blowing and top-blowing furnace into unified process parameters through a unified information library, thereby achieving more precise control. This helps to reduce operational errors and improve production efficiency and product quality; using the data in the unified information library, the process parameters of the oxygen-enriched side-blowing and top-blowing furnace can be optimized and adjusted, thereby improving smelting efficiency and product quality; through biased decision analysis and feedback control management, the system can quickly respond to changes in the production process to ensure the stability and continuity of the production process; the unified information library realizes data standardization and sharing between different devices and systems, avoids information islands, and improves data utilization efficiency. Biased decisions are made based on multi-dimensional factors such as node control difference, control trend bias, and response delay, which enhances the decision support capability of the system; through feedback control management, the system can adjust the process parameters in real time to ensure the stability of the production process and the consistency of product quality. This embodiment achieves precise control and optimization of the process parameters of the oxygen-enriched side-blowing and top-blowing furnace by establishing a unified information library and a decision conversion mechanism based on the response terminal, improves production efficiency and product quality, and enhances the responsiveness and stability of the system.

[0178] like Fig.16 As shown, this embodiment also provides an embodiment of a data-driven multi-furnace continuous smelting intelligent control method. In this embodiment, the data-driven multi-furnace continuous smelting intelligent control method is applied to the data-driven multi-furnace continuous smelting intelligent control system in the above embodiment. The data-driven multi-furnace continuous smelting intelligent control method specifically includes the following steps:

[0179] Step S1: collecting the operating composition data of the temperature, oxygen content and gas composition in the oxygen-enriched side-blowing and top-blowing furnace; performing preprocessing operations such as data cleaning, removing noise and outliers, and filling missing values ​​on the collected operating composition data; performing time series alignment on the preprocessed operating composition data to form an operating composition data set;

[0180] Step S2: retrieve historical production data and real-time collected operating composition data, build a continuous smelting simulation model, and divide the data set into a training set, a validation set, and a test set; use the training set to train the continuous smelting simulation model, and input the real-time collected operating composition data into the continuous smelting simulation model for simulation;

[0181] Step S3: Analyze the simulation results, compare the analysis results with the preset values, and adjust the process parameters according to the comparison results; adjust the process parameters for data greater than the preset deviation value in the smelting process according to real-time production data feedback; upload the adjusted parameters to the intelligent control cloud platform.

[0182] Preferably, in step S1 of this embodiment, preprocessing operations such as data cleaning, noise and outlier removal, and missing value filling are performed to ensure the accuracy and consistency of the data, thereby improving the reliability of subsequent analysis and model training; the preprocessed operating component data are time-series aligned to form a unified data set, which is helpful for the subsequent establishment and simulation of the continuous smelting simulation model; including operations such as removing duplicate data, screening required data, detecting and processing outliers, and filling missing values ​​to ensure the integrity and consistency of the data; through timestamp synchronization, resampling and filling technology, problems such as data loss and inconsistent sampling are solved to ensure the time continuity and accuracy of the data. In step S2, the continuous smelting simulation model is constructed using historical production data and real-time collected data, which can predict and optimize the key parameters in the smelting process and improve smelting efficiency and product quality; by dividing the data set into training set, validation set and test set, and using the training set to train the model, the performance and generalization ability of the model can be effectively evaluated; the data set is reasonably divided into training set, validation set and test set to ensure the effectiveness and generalization ability of model training; this embodiment uses numerical simulation and simulation technology to explore the gas-liquid flow characteristics in the molten pool, optimize the oxygen gun arrangement and flow rate, and improve the metallurgical reaction efficiency and molten pool stirring effect. In step S3, by comparing the simulation results with the preset values, the process parameters are adjusted to ensure the stability and efficiency of the smelting process; according to the real-time production data feedback, the deviations in the smelting process are adjusted in real time, and the process parameters are further optimized to improve product quality; the simulation results are compared and analyzed with the preset values ​​to identify the optimization direction of the process parameters; the adjusted parameters are uploaded through the intelligent control cloud platform, and feedback adjustments are made according to the real-time production data to achieve closed-loop control. The operation process of the oxygen-enriched side-blowing and top-blowing furnace has been fully supported by data and intelligent management. Data preprocessing improves data quality, the continuous smelting simulation model optimizes the smelting process, and the simulation result analysis and real-time adjustment ensure the optimal configuration of process parameters and the steady improvement of product quality. These technical means not only improve production efficiency, but also reduce production costs, which is in line with the trend of modern industrial intelligent development.

[0183] like Fig.17 As shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 8 includes a processor 81 and a memory 82 coupled to the processor 81.

[0184] The memory 82 stores program instructions for implementing a method for laying out process parameters for continuous smelting of multiple furnaces with side blowing and top blowing according to any of the above-mentioned embodiments.

[0185] The processor 81 is used to execute the program instructions stored in the memory 82 to layout the process parameter equipment of the side-blowing and top-blowing multi-furnace continuous smelting.

[0186] The processor 81 may also be referred to as a CPU (Central Processing Unit). The processor 81 may be an integrated circuit chip having signal processing capabilities. The processor 81 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0187] Further, Fig.18 This is a schematic diagram of the structure of a storage medium of an embodiment of the present application. The storage medium 9 of the embodiment of the present application stores program instructions 91 that can implement all the above methods, wherein the program instructions 91 can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0188] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0189] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units. The above is only an implementation mode of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present invention.

[0190] The specific implementation methods of the invention are described in detail above, but they are only examples, and the invention is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the invention, therefore, the equalization, modification, improvement, etc. made without departing from the spirit and principle of the invention should be included in the scope of the invention.

Claims

1. A data-driven multi-furnace continuous smelting intelligent control system, characterized in that: The data-driven multi-furnace continuous smelting intelligent control system includes: The data acquisition module is used to collect the operating composition data of the temperature, oxygen content and gas composition in the oxygen-enriched side-blowing furnace; perform data cleaning, remove noise and outliers, and fill in missing values ​​preprocessing operations on the collected operating composition data; perform time series alignment on the preprocessed operating composition data to form an operating composition data set; The machine learning module is used to retrieve historical production data and real-time collected operating composition data, build a continuous smelting simulation model, and divide the data set into a training set, a validation set, and a test set; use the training set to train the continuous smelting simulation model, and input the real-time collected operating composition data into the continuous smelting simulation model for simulation; The intelligent control module is used to analyze the simulation results, compare the analysis results with the preset values, and adjust the process parameters according to the comparison results; according to the real-time production data feedback, adjust the process parameters for the data greater than the preset deviation value in the smelting process; and upload the adjusted parameters to the intelligent control cloud platform.

2. The data-driven multi-furnace continuous smelting intelligent control system according to claim 1 is characterized in that: Data acquisition module, including: The real-time acquisition submodule is used to collect the operating composition data of the temperature, oxygen content and gas composition in the oxygen-enriched side-blown furnace; the collected operating composition data is input into the edge computing submodule 12 for data cleaning; The edge computing submodule is used to remove noise and outliers from the collected operating component data, remove data points using the threshold method, complete missing data, and perform data analysis and feature extraction on the pre-processed operating component data. The time series submodule is used to dynamically time-warp the running component data after feature extraction to obtain time series running component data; and to refine and decompose the time series running component data to obtain different feature levels.

3. The data-driven multi-furnace continuous smelting intelligent control system according to claim 2 is characterized in that: Edge computing submodule, including: Data cleaning unit is used to remove noise and outliers. If the collected data is x1, x2, ..., x n , use the threshold method to eliminate all i -μ|>kσ data points, if a data point meets the condition, the data point is regarded as an outlier and is removed; The data completion unit is used to process missing data caused by process fluctuations and environmental changes. It processes the data by setting the system's high-order spline interpolation, prior distribution p(θ) and likelihood function p(D|θ) of the observed data to calculate the missing data value. The dimension reduction extraction unit is used to perform data analysis and feature extraction on the completed data, centralize the data, calculate its covariance matrix to obtain its eigenvalues ​​and eigenvectors, and sort its feature data by importance.

4. The data-driven multi-furnace continuous smelting intelligent control system according to claim 3 is characterized in that: Data completion unit, including: The interpolation processing subunit is used to deal with the data missing problem in the data collection of the smelting process, and uses the high-order spline interpolation method to process the interpolation; The data completion subunit is used to handle missing data due to process fluctuations and environmental changes. It sets the system's prior distribution p(θ) and defines the likelihood function p(D|θ) of the observed data to describe the probability of the data observed under given parameter conditions. The prior distribution is combined with the likelihood function using Bayes' theorem to obtain the posterior distribution p(θ|D). The prediction mean subunit is used to calculate the predicted distribution p(y) of missing values ​​based on the posterior distribution. * ∣x * , D); confidence intervals are calculated based on the predicted means.

5. The data-driven multi-furnace continuous smelting intelligent control system according to claim 3 is characterized in that: Dimensionality reduction extraction unit, including: The principal component dimensionality reduction subunit is used to set the running component data matrix to be X∈R m×n ; Where m is the number of samples and n is the number of features; perform central processing on the running component data and remove the mean of each column; calculate the co-matrix of the running component data and perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors; select the eigenvectors corresponding to the first k largest eigenvalues ​​to form a dimensionality reduction matrix W∈R n×k , get the final running component data; The independent component analysis subunit is used to collect the observed data signal represented as a linear combination of several independent components. Suppose the observed data X∈R m×n , obtain at least one separation matrix W, such that S = WX; where S is an independent component matrix, satisfying the statistical independence between each column; The feature sorting subunit is used to evaluate the contribution of each feature on all tree splitting nodes through the random forest method of the tree model, and quantify the influence of each feature on the output of the tree model; sort the features according to their importance, quantify the influence of the features on the tree model, and eliminate unimportant features.

6. The data-driven multi-furnace continuous smelting intelligent control system according to claim 2 is characterized in that: Time series submodule, including: A dynamic time unit is used to compare different time series of running component data with each other, calculate according to the comparison results, and obtain at least one alignment path; calculate multiple alignment paths to obtain the optimal alignment path; Multi-scale time unit, used to introduce the running component data of the optimal alignment path into the multi-scale function φ(t) and the detail function ψ(t); decompose the time series to obtain the dynamic change characteristics of different time scales; The wavelet transform unit is used to refine and decompose the time series using discrete wavelet transform to obtain different feature levels and decompose the time series into approximation and detail parts of different scales.

7. The data-driven multi-furnace continuous smelting intelligent control system according to claim 6 is characterized in that: Dynamic time units, including: The sequence alignment subunit is used to align different time series of running component data X = [x1, x2, ..., x n ] and Y = [y1, y2, ..., y m ] to calculate and obtain at least one alignment path π = [(i1, j1), (i2, j2), ..., (i K , j K )], the total alignment cost of the two time series is minimized; The optimal alignment path subunit is used to calculate the optimal path and construct the cumulative cost matrix D using the dynamic programming method. ij , the minimum cumulative cost from the starting position (i, j) of the sequence; The minimization subunit is used to obtain the optimal alignment path of sequences X and Y by minimizing the cumulative cost matrix and eliminating the nonlinear deviation of the timeline.

8. The data-driven multi-furnace continuous smelting intelligent control system according to claim 1 is characterized in that: Machine Learning modules, including: The data processing submodule is used to extract key indicators from historical production data and real-time collected operating component data; clean the collected data, remove noise and outliers, fill in missing values, and perform time series alignment; The feature extraction submodule is used to extract meaningful features from the preprocessed data; select a machine learning model based on the characteristics of the continuous smelting process and build a continuous smelting simulation model; divide the data set into a training set, a validation set, and a test set, use the training set to train the model, and adjust the model parameters through the validation set; The model verification submodule is used to input the real-time collected operating composition data into the trained continuous smelting simulation model, perform simulation, and generate prediction results.

9. The data-driven multi-furnace continuous smelting intelligent control system according to claim 3 is characterized in that: Intelligent control module, including: The forward feedback control submodule is used to adjust key process parameters according to the simulation results in the continuous smelting simulation model; adjust the deviations in the smelting process through real-time data feedback; and perform local feedback regulation through temperature control and oxygen content control; The adaptive control submodule is used to learn the dynamic change rules in the production process through neural networks and dynamically adjust the process parameters according to the real-time process conditions; the adaptive algorithm is performed using the adaptive law based on the Lyapinov stability theory; The fuzzy logic submodule is used to define the fuzzy rule base based on the complexity of multi-variable coupling in the smelting process, and to intelligently control the multi-input and multi-output equipment; the control strategy is dynamically adjusted through the fuzzy reasoning module; The clustering fuzzy unit is used to determine the intelligent decision sequence based on the responsibility of multi-variable coupling in the smelting process; based on the preset similarity, the intelligent decision sequence is clustered and the decision fuzzification is performed for each cluster; The fuzzy rule base unit is used to mine fuzzy logic rules based on logical necessity and logical commonality; there is a one-to-one correspondence between fuzzy logic rules and clusters; fuzzification and fuzzy decision analysis are performed on data; The response unit is used to determine the fuzzy strategy according to the analysis results and perform fuzzification, determine the control strategy, obtain the process data of the smelting process output by the strategy based on the response terminal of the decision-making task, and perform technical architecture conversion and control response of the response terminal of the control strategy; Determine a fuzzy processing method, wherein the fuzzy processing method includes data sharpening processing and data passivation processing; based on the fuzzy processing method, establish a bidirectional fuzzification block, wherein the bidirectional fuzzification block includes a fuzzy processing branch and a defuzzification branch; based on the fuzzy processing branch, perform fuzzy processing on the measurement series of each cluster; Traverse the fuzzy logic rules, supervise the decision domain, and the decision domain corresponds to the fuzzy logic rules one by one; based on the decision frequency and decision operator requirements, integrate the decision domain and allocate resources, determine the fuzzy decision block, where the fuzzy decision block has extensibility and updateability, determine the entry and exit ports of the fuzzy processing branch and the output port of the defuzzification branch, and connect the fuzzy decision block in the middle; A unified information database is established, wherein the unified information database is a conversion database of the process parameters of the oxygen-enriched side-blowing furnace and the unified process parameters; based on the response terminal, the unified information database is traversed to determine the data for adjusting the process parameters of the oxygen-enriched side-blowing furnace, and the data is converted to the control decision strategy to determine the terminal control strategy; the control data stream is obtained, and a bias decision analysis is performed on the control data stream, and the bias decision is free of the preset tolerance interval; the first bias decision is made based on the node control difference, the second bias decision is made based on the control trend bias, and the third bias decision is made based on the response delay. If the preset tolerance interval of the mapping is not met, feedback control management is performed.

10. A data-driven intelligent control method for continuous smelting of multiple kilns, which is applied to the data-driven intelligent control system for continuous smelting of multiple kilns as claimed in any one of claims 1 to 9, characterized in that: The data-driven multi-furnace continuous smelting intelligent control method comprises: Collect the operating composition data of the temperature, oxygen content and gas composition in the oxygen-enriched side-blown furnace; perform preprocessing operations such as data cleaning, noise and outlier removal, and missing value filling on the collected operating composition data; perform time series alignment on the preprocessed operating composition data to form an operating composition data set; Retrieve historical production data and real-time collected operating composition data, build a continuous smelting simulation model, and divide the data set into a training set, a validation set, and a test set; use the training set to train the continuous smelting simulation model, and input the real-time collected operating composition data into the continuous smelting simulation model for simulation; Analyze the simulation results, compare the analysis results with the preset values, and adjust the process parameters according to the comparison results; adjust the process parameters for data greater than the preset deviation value in the smelting process based on real-time production data feedback; upload the adjusted parameters to the intelligent control cloud platform.

Images