Blast furnace stokehole clay gun control system

Abstract

The invention relates to the technical field of blast furnace clay gun control, and discloses a blast furnace stokehole clay gun control system. The system comprises a data acquisition module, a state feature extraction module, a multi-dimensional difference analysis module, a risk area positioning module and a self-adaptive control strategy generation module. The data acquisition module acquires a clay gun pressure waveform, an iron notch temperature sequence and a stemming flow pulse in real time; the state feature extraction module extracts operation state features based on the operation data; the multi-dimensional difference analysis module performs multi-dimensional difference analysis on the state characteristics and a preset theoretical state to generate a state difference coefficient matrix; a risk area positioning module positions an operation risk area based on the matrix; and the adaptive control strategy generation module generates an adaptive control strategy according to the risk area. The system realizes comprehensive acquisition of multiple parameters of mud gun operation, accurate state analysis, accurate risk positioning and self-adaptive control, improves the mud gun operation stability, ensures continuous operation in front of a blast furnace, and adapts to modern blast furnace production requirements.

Description

Technical Field

[0001] This invention relates to the field of blast furnace mud gun control technology, specifically a blast furnace front mud gun control system. Background Technology

[0002] In the steelmaking process, the blast furnace mud gun is a crucial piece of equipment for ensuring taphole sealing and safe production. Its operational status directly affects the continuity and stability of blast furnace operations. Currently, most steel companies in China still rely on traditional manual operation or simple closed-loop control modes for blast furnace mud gun control, which has significant limitations. Manual operation mainly depends on the operator's experience and judgment, making it difficult to accurately control the dynamic parameters during mud gun operation in real time, such as fluctuations in mud gun pressure, changes in taphole temperature, and the stability of mud flow. If these subtle changes in parameters are not detected in time, they can easily lead to problems such as poor taphole sealing and abnormal mud consumption, resulting in serious consequences such as blast furnace shutdown and production reduction. As the steel industry moves towards intelligent and efficient operations, traditional control methods can no longer meet the demands of modern blast furnace production. While some existing mud gun control systems attempt to incorporate data acquisition and analysis functions, they fall short in the comprehensiveness and depth of data processing. For example, most systems only collect single parameters such as mud gun pressure or taphole temperature, lacking collaborative analysis of multiple parameters and failing to accurately reflect the overall operating status of the mud gun. Furthermore, in the status assessment and risk warning stages, fixed thresholds are often used for judgment, failing to dynamically adjust according to changes in blast furnace production conditions, resulting in low accuracy and timeliness of risk identification. In addition, existing systems rely heavily on preset fixed schemes when generating control strategies, lacking the ability to adaptively adjust to differences in actual operating conditions, making it difficult to achieve precise control of mud gun operation, thus affecting the stability and efficiency of blast furnace production. Summary of the Invention

[0003] The purpose of this invention is to provide a blast furnace front mud gun control system to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides a blast furnace front mud gun control system, the system comprising: Data acquisition module: Real-time acquisition of blast furnace mud gun operation data, including mud gun pressure waveform, taphole temperature sequence and mud flow pulse; Status feature extraction module: Extracts the operating status features of the mud gun based on the operation data; Multidimensional difference analysis module: Performs multidimensional difference analysis on the mud gun operating state characteristics and preset theoretical state to generate mud gun state difference coefficient matrix; Risk area location module: locates the risk area of ​​blast furnace front operation based on the mud gun state difference coefficient matrix; Adaptive control strategy generation module: Generates mud cannon adaptive control strategy based on the risk area.

[0005] Preferably, the data acquisition module specifically includes: Real-time operational data acquisition unit: synchronously captures the zero-crossing distortion rate and phase jitter parameters of the mud gun pressure waveform, the gradient change characteristics of the taphole temperature sequence, and the amplitude abrupt change gradient and time interval entropy value of the mud gun flow pulse through a sensor array deployed on the mud gun equipment. Historical data feature mining unit: performs multi-dimensional decomposition processing on the historical operation data of mud guns, including using an adaptive signal decomposition method to extract the energy ratio of steady-state and transient components of pressure waveforms, establishing a correlation matrix between flow pulses and mud gun thrust through load correlation analysis, and using a sequence alignment algorithm to align temperature sequence patterns under different operating conditions.

[0006] Preferably, the state feature extraction module specifically includes: Signal principal component channel construction unit: Constructs a pressure waveform principal component channel based on the signal decomposition method, retaining preset frequency band components; Signal auxiliary channel establishment unit: Establishes a flow pulse auxiliary channel based on frequency domain analysis method, and uses phase decoupling technology to decouple the preset frequency band; Feature fusion unit: It evaluates the orthogonality of the dual-channel signals of the main component channel of the pressure waveform and the auxiliary channel of the flow pulse by information entropy, dynamically adjusts the parameter weights between the channels to achieve feature fusion, and outputs the operating status features of the mud gun.

[0007] Preferably, the multidimensional difference analysis module specifically includes: Time-domain cumulative deviation calculation unit: The mud gun operating state characteristics and theoretical state are compared with a preset time window, the state sequence of asynchronous sampling is aligned by a sequence alignment algorithm, the cumulative deviation in each window is calculated, and a time-domain deviation vector is generated. Frequency domain energy shift detection unit: Decomposes the temperature sequence of mud gun operating state characteristics and theoretical state using frequency domain decomposition algorithm, calculates the energy spectral density ratio, extracts the energy shift index of each frequency band, and constructs frequency domain shift vector; Pulse sequence similarity evaluation unit: Based on the structure matching algorithm, the distance distribution of the flow pulse sequence of the mud gun operating state characteristics and theoretical state is matched, the phase synchronization error of the pulse amplitude change point is calculated, the divergence index of the pulse interval distribution is quantified, and the pulse similarity vector is generated. The difference coefficient matrix generation unit: concatenates the time-domain deviation vector, frequency-domain offset vector, and pulse similarity vector into tensors, and through normalization processing with feature importance weighting, outputs a third-order mud gun state difference coefficient matrix with the dimension of device number timestamp difference type.

[0008] Preferably, the frequency domain energy shift detection unit further includes: Sensitive frequency band extraction subunit: Extract the temperature sequence corresponding to the operating state characteristics and theoretical state of the mud gun, and use frequency band analysis technology to perform multi-scale frequency band analysis on each sequence to extract the energy distribution characteristics within the preset sensitive frequency band interval; Energy offset quantization sub-unit: Quantizes the energy density of the theoretical state and the mud gun operating state characteristics in the sensitive frequency band interval respectively, extracts the offset index of each frequency band based on the relative offset degree, and constructs the frequency domain offset vector.

[0009] Preferably, the risk area location module specifically includes: Blast furnace topology modeling unit: Construct a topology diagram of blast furnace node connection relationship based on the location information of mud gun equipment, mark the heat conduction parameters between each node, superimpose the reverse heat flow constraint conditions of the taphole area on the topology diagram, and generate a furnace front thermodynamic topology model including thermal resistance matrix and node admittance matrix. Risk propagation simulation unit: Maps the mud gun state difference coefficient matrix to the corresponding nodes of the blast furnace topology model, performs risk propagation simulation based on graph analysis network, including calculating the attenuation factor of risk heat flow according to the node heat conduction parameters, capturing cross-regional risk correlation characteristics through multi-head attention mechanism, and simulating the diffusion path of risk heat flow in the topology network using stochastic simulation method; Probability distribution generation unit: Statistically counts the frequency of risk heat flow occurrence in simulated propagation for each path, calculates the residence probability value of risk heat flow in combination with heat conduction parameters, generates a risk probability thermal distribution map covering the entire furnace, and marks the set of suspicious paths whose probability values ​​exceed the preset risk residence probability threshold; Physical region positioning unit: Performs spatial clustering analysis on the risk probability thermal distribution map, identifies risk probability clustering areas, and delineates the physical boundaries of risk operations based on the location of the mud gun equipment and the topological connection relationship with the blast furnace.

[0010] Preferably, the risk area location module further includes: Suspicious equipment identifier generation subunit: Based on the blast furnace nodes connected by the suspicious path set, the blast furnace nodes are bound to the mud gun equipment to form a suspicious equipment identifier set; Risk propagation main path generation sub-unit: In the randomized simulated risk diffusion process, the node paths and their order experienced in each round of propagation are recorded, the frequency of occurrence of each path is counted, and the path sequence with the highest cumulative frequency of occurrence is selected as the risk propagation main path sequence.

[0011] Preferably, the adaptive control strategy generation module specifically includes: High-frequency monitoring mode activation unit: When the abnormal probability value of a certain area in the risk probability thermal distribution map exceeds the preset risk residence probability threshold, a monitoring mode switching command is issued to the mud gun equipment to which the area belongs, including increasing the pressure and temperature sampling frequency to a multiple of the original frequency, simultaneously enabling the real-time temperature component tracking mode, capturing frequency band phase changes, and deploying transient event detectors on the edge side to record temperature transient waveform segments. Physical disturbance test execution unit: Apply multi-band physical disturbances to the adjacent node devices with the largest changes in risk probability gradient in the risk probability heat distribution map.

[0012] Preferably, the physical disturbance test execution unit further includes: Test signal injection subunit: Injects sweep frequency test signals through a controllable heat source; Theoretical response calculation subunit: Calculate the theoretical response thermal spectrum based on the furnace front thermodynamic topology model and thermal resistance matrix; Measured response acquisition subunit: Record the temperature of each node after the test signal is injected to obtain the measured response thermal spectrum; Anomaly offset determination subunit: Calculates the distance measurement between the theoretical response thermal spectrum and the measured response thermal spectrum, compares the distance measurement between the measured thermal spectrum and the theoretical thermal spectrum, calculates the percentage of abnormal node offset, and marks the abnormal node offset as a heat flow abnormal associated device when the percentage of abnormal node offset exceeds the second threshold.

[0013] Preferably, the state feature extraction module further includes a body motion event processing unit: Body movement candidate interval localization subunit: Performs energy gradient analysis on adjacent signal frames of the operation data, locates body movement candidate intervals, and generates body movement event markers; Body motion signal decoupling subunit: performs non-stationary feature decoupling on the marker signal of the body motion event, and separates the frequency band feature components that characterize the body motion event; Body motion intensity waveform generation subunit: Based on the motion trajectory discrimination criterion, it distinguishes between local body motion and global interference, and generates body motion intensity waveform by fusing the frequency band feature components.

[0014] Compared with the prior art, the beneficial effects of the present invention are: In the parameter acquisition phase, the data acquisition module can collect various operational data in real time, including mud gun pressure waveforms, taphole temperature sequences, and mud flow pulses. Compared to existing systems that only collect single parameters, this achieves comprehensive coverage of key mud gun operation parameters, providing a rich data foundation for accurate subsequent analysis of the mud gun's operating status. By acquiring continuous pressure waveforms, temperature sequences, and flow pulse data, the parameter change trends during mud gun operation can be captured more precisely, avoiding misjudgments caused by incomplete acquisition of a single parameter. The state feature extraction module extracts the operating state features of the mud gun based on the collected multi-type operational data, enabling the extraction of key information reflecting the essence of mud gun operation from complex raw data. Compared to the existing system's simple threshold judgment of raw data, extracting state features can more accurately reflect the actual operating status of the mud gun. For example, by analyzing the fluctuation frequency and amplitude characteristics of the pressure waveform, it is possible to determine whether the mud gun propulsion mechanism is operating normally; by analyzing the changing trend characteristics of the taphole temperature sequence, it is possible to determine the heat dissipation and sealing conditions of the taphole area, providing a reliable basis for subsequent state analysis and risk identification. The multidimensional difference analysis module performs a multidimensional difference analysis between the extracted mud gun operating state characteristics and the preset theoretical state, generating a mud gun state difference coefficient matrix, breaking through the single-dimensional difference comparison mode of existing systems. By analyzing differences from multiple dimensions such as pressure, temperature, and flow rate, it can comprehensively consider the synergistic effects between various parameters, avoiding the one-sidedness of difference judgments caused by single-dimensional analysis. The generated difference coefficient matrix can quantify the degree of state difference under different dimensions, clearly presenting the deviation of the mud gun operating state from the ideal state, providing accurate data support for subsequent risk area location.

[0015] The risk area location module locates risk areas in blast furnace operations based on a mud gun state difference coefficient matrix, changing the existing system's reliance on fixed thresholds for risk assessment. By combining quantitative data from the multidimensional difference coefficient matrix, it can accurately identify specific risk areas during mud gun operation. For example, abnormal differences in pressure and flow rates can pinpoint risk areas such as nozzle blockage or insufficient propulsion pressure; abnormal differences in temperature and pressure can pinpoint risk areas such as poor taphole sealing or abnormal mud solidification. This transforms risk identification from fuzzy judgment to precise location, helping operators to promptly grasp the specific location and cause of risks.

[0016] The adaptive control strategy generation module generates adaptive control strategies for the mud gun based on the identified risk areas. Compared to existing systems that use fixed control schemes, this module can dynamically adjust control parameters and operating procedures according to different risk areas and conditions. For example, when a risk area of ​​nozzle blockage is identified, the mud gun propulsion pressure and flow parameters can be automatically adjusted to clear the nozzle; when a risk area of ​​poor taphole sealing is identified, the mud formula ratio or propulsion speed can be automatically adjusted to enhance the taphole sealing effect. This adaptive control strategy generation method allows the mud gun control scheme to match the actual operating conditions, avoiding the problem of poor control performance caused by fixed control schemes failing to adapt to changes in operating conditions. This improves the stability of mud gun operation, ensures continuous operation of the blast furnace, and reduces blast furnace shutdowns and production reductions due to mud gun malfunctions. Attached Figure Description

[0017] Figure 1 This is a timing diagram of the blast furnace front mud gun control system described in this invention; Figure 2 This is a flowchart illustrating the working principle of the data acquisition module. Figure 3 This is a flowchart illustrating the working principle of the multidimensional difference analysis module. Figure 4 A flowchart illustrating the working principle of the adaptive control strategy generation module. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 The present invention provides a blast furnace front mud gun control system, the system comprising: a data acquisition module, a state feature extraction module, a multidimensional difference analysis module, a risk area positioning module, and an adaptive control strategy generation module.

[0020] The data acquisition module collects real-time operational data of the blast furnace mud gun, including mud gun pressure waveforms, taphole temperature sequences, and mud flow pulses. The state feature extraction module extracts the mud gun's operational state characteristics based on the operational data. The multidimensional difference analysis module performs multidimensional difference analysis between the mud gun's operational state characteristics and preset theoretical states, generating a mud gun state difference coefficient matrix. The risk area location module locates the operational risk areas in front of the blast furnace based on the mud gun state difference coefficient matrix. The adaptive control strategy generation module generates an adaptive control strategy for the mud gun based on the risk areas.

[0021] Example 1: See Figure 2This paper describes the specific implementation of the data acquisition module and the state feature extraction module in the blast furnace mud gun control system. The data acquisition module achieves real-time capture of operational data through a sensor array deployed at key parts of the mud gun equipment. This sensor array includes a high-frequency dynamic pressure sensor, an infrared temperature sensor, and an electromagnetic flowmeter. The high-frequency dynamic pressure sensor is installed on the inlet and outlet oil lines of the mud gun hydraulic cylinder, recording oil pressure changes at a sampling frequency of 5,000 times per second to generate pressure waveform data. This sensor has high anti-interference capability and can effectively filter noise from the strong electromagnetic environment at the blast furnace site. The data processing unit calculates the zero-crossing distortion rate of the waveform in real time, i.e., the degree of abnormal frequency change when the pressure signal crosses zero, and analyzes the phase jitter parameters of the waveform to reflect the stability of the pressure output. The infrared temperature sensor is arranged in an array on the refractory material surface around the taphole, collecting temperature data at a sampling frequency of 100 times per second to form a temperature sequence. A dedicated algorithm calculates the temperature difference between adjacent sampling points in real time, extracts gradient change features, and identifies areas of sudden temperature rise or fall. The electromagnetic flowmeter is installed in the mud conveying pipeline to monitor the mud flow state and record flow pulse signals. The signal processing unit analyzes the amplitude gradient of the pulse, i.e. the rate of change of the peak values ​​of adjacent pulses, and calculates the time interval entropy value to characterize the regularity or randomness of pulse occurrence.

[0022] The historical data feature mining unit performs multi-dimensional decomposition processing on the historical operation data of mud guns stored in the database. This unit employs an adaptive signal decomposition method to decompose historical pressure waveform data into multiple intrinsic mode functions (EMFs), calculates the energy value of each function, and obtains the energy ratio of steady-state and transient components. Through load correlation analysis, a correlation matrix between flow pulse characteristics and mud gun thrust parameters is established, revealing the thrust variation patterns under different flow patterns. Using a sequence alignment algorithm, temperature sequences under different production conditions are aligned to eliminate data bias caused by asynchronous sampling times and extract comparable temperature change patterns.

[0023] The signal principal component channel construction unit of the state feature extraction module processes the real-time acquired pressure waveform. This unit uses a signal decomposition method to decompose the pressure waveform into components of different frequency bands. Based on prior knowledge, it retains preset frequency band components closely related to the mud gun's operating state while eliminating high-frequency noise and low-frequency interference. The constructed principal component channel can highlight key information reflecting the operating state of the mud gun's hydraulic system. The signal auxiliary channel establishment unit performs frequency domain analysis on the flow pulse signal, using a fast Fourier transform to convert the time-domain signal into a frequency-domain representation. Phase decoupling technology is applied to decouple preset frequency bands, separating mutually coupled frequency components to obtain purer frequency domain features. The feature fusion unit evaluates the orthogonality of the dual-channel signals of the pressure waveform principal component channel and the flow pulse auxiliary channel by calculating the information entropy value, determining the complementarity of the information provided by the two channels. Based on the entropy calculation results, the parameter weights of the two channels in the feature fusion process are dynamically adjusted to achieve optimal feature fusion, ultimately outputting a feature vector that comprehensively reflects the mud gun's operating state.

[0024] The motion event processing unit performs energy gradient analysis on adjacent signal frames of the operational data, calculating the rate of energy change between consecutive signal frames. When the energy gradient exceeds a set threshold, it is identified as a candidate motion interval, and a motion event marker is generated. Non-stationary features are decoupled from the marker signal segment, and adaptive filtering techniques are used to separate specific frequency band feature components characterizing the motion event. Based on motion trajectory discrimination criteria, the distribution characteristics of the feature components in the time and frequency domains are analyzed to distinguish between local motion and global interference. Local motion typically manifests as a brief energy surge in a specific frequency band, while global interference presents as a sustained energy change over a wide frequency band. Finally, the feature components of each frequency band are fused to generate a motion intensity waveform, which quantitatively describes the intensity and temporal distribution characteristics of the motion interference.

[0025] The data acquisition module and the state feature extraction module work together to ensure accurate and comprehensive acquisition of mud gun operating status characteristics under various working conditions. The careful arrangement of the sensor array and the optimized design of the signal processing algorithm enable the system to effectively cope with various interference factors in the complex environment of the blast furnace. In-depth mining of historical data provides an important reference benchmark for real-time feature extraction, while motion event processing enhances the system's robustness under abnormal operating conditions. The dynamic weight adjustment mechanism in the feature fusion process ensures that the final output state features include both pressure information and flow characteristics, comprehensively reflecting the actual operating status of the mud gun.

[0026] Example 2: See Figure 3This paper describes a multi-dimensional difference analysis module in the blast furnace mud gun control system. Through multi-dimensional comparative analysis, it accurately compares the real-time extracted mud gun operating status characteristics with preset theoretical states, generating a coefficient matrix that comprehensively reflects the differences in equipment states. The time-domain cumulative deviation calculation unit performs a sliding comparison between the mud gun operating status characteristic sequence and the theoretical state sequence within a fixed time window. Since the actual sampling frequency and theoretical sampling may be asynchronous, this unit uses a sequence alignment algorithm to accurately align the two sets of asynchronously sampled state sequences along the time axis. During the alignment process, the algorithm finds the optimal matching path to establish a correspondence between the actual sampling points and the theoretical sampling points, eliminating comparison errors caused by sampling time deviations. Within each aligned time window, the cumulative deviation of corresponding data points in the two sets of sequences is calculated, including various metrics such as absolute difference, relative error, and root mean square error. These deviations are standardized to form a time-domain deviation vector, which fully records the degree of deviation of the mud gun operating state from the theoretical state in the time dimension.

[0027] The frequency domain energy shift detection unit performs frequency domain decomposition processing on the temperature sequence of the mud gun's operating state characteristics and theoretical state. This unit uses a frequency domain decomposition algorithm to convert the time-domain temperature signal to the frequency domain space and calculates the energy spectral density value of each frequency band. By comparing the energy distribution differences between the real-time state and the theoretical state in each frequency band, the energy spectral density ratio is calculated, and the energy shift index of each frequency band is extracted. The sensitive frequency band extraction subunit determines the specific frequency band intervals most sensitive to changes in the mud gun's operating state based on prior knowledge, and uses frequency band analysis technology to perform multi-scale frequency band analysis on the temperature sequence to extract the energy distribution characteristics within these sensitive frequency bands. The energy shift quantification subunit calculates the energy density values ​​of the theoretical state and the real-time state within the sensitive frequency band intervals, and calculates the shift index of each frequency band based on the relative shift degree. These shift indices, after normalization, constitute a frequency domain shift vector, reflecting the abnormal state of the mud gun system in the frequency dimension.

[0028] The pulse sequence similarity assessment unit processes the flow pulse sequences of mud guns based on their operational characteristics and theoretical conditions. This unit, based on a structure matching algorithm, analyzes the degree of similarity between the two sets of pulse sequences in their overall morphology. The algorithm first identifies key feature points in the pulse sequence, including the pulse start point, peak point, and end point, and then calculates the distribution distance of these feature points along the time axis. By measuring the phase synchronization error at abrupt changes in pulse amplitude, the difference in time synchronization between the real-time pulse sequence and the theoretical sequence is evaluated. Simultaneously, this unit quantifies the divergence index of the pulse interval distribution to analyze the regular changes in pulse occurrence. After comprehensive calculation of these similarity indices, a pulse similarity vector is generated, comprehensively describing the anomalous characteristics of the flow pulse sequence.

[0029] The difference coefficient matrix generation unit performs tensor concatenation operations on the time-domain deviation vector, frequency-domain offset vector, and pulse similarity vector. During concatenation, the dimensionality of each vector is maintained, forming a multi-dimensional data tensor. This unit employs a feature importance weighting mechanism, assigning different weight coefficients based on the importance of each feature dimension to the mud gun state assessment. The weight allocation is based on historical data analysis results, assigning higher weights to feature dimensions that more sensitively reflect equipment state changes. After weighting, a normalization operation is performed to eliminate dimensional differences between dimensions, ultimately outputting a third-order mud gun state difference coefficient matrix with dimensions corresponding to equipment number and timestamp difference types. The row dimension of this matrix corresponds to different mud gun equipment numbers, the column dimension records timestamp information, and the depth dimension contains specific values ​​for various difference types, forming a complete multi-dimensional state difference description system.

[0030] Throughout the implementation process, the multidimensional difference analysis module achieved a comprehensive assessment of the mud gun's operating status through integrated analysis across three dimensions: time domain, frequency domain, and pulse sequence. Time domain analysis captured abnormal trends in state parameters over time, frequency domain analysis revealed subtle changes in the system's frequency characteristics, and pulse sequence analysis focused on the accuracy of flow control. This multi-faceted analysis method can promptly identify potential anomalies in the mud gun equipment, providing accurate data support for subsequent risk area location. The algorithms used in each analysis unit were optimized to ensure both analytical accuracy and real-time performance, guaranteeing the system's timely response to changes in the mud gun's status.

[0031] The structural design of the difference coefficient matrix fully considers the needs of subsequent processing, containing detailed state difference information while maintaining data regularity, facilitating further analysis by the risk area location module. Each element in the matrix represents the degree of state difference of a specific piece of equipment at a specific point in time, and this data forms the basis for mud gun equipment condition monitoring. By continuously updating and maintaining this difference coefficient matrix, the system can establish a historical archive of the mud gun equipment's operating status.

[0032] Example 3: See Figure 4This paper details the risk area location module in the blast furnace mud gun control system. Based on the mud gun state difference coefficient matrix and combined with the physical structural characteristics of the blast furnace, it accurately locates physical areas with operational risks through thermodynamic modeling and risk propagation simulation. The blast furnace topology modeling unit first constructs a topology diagram of the blast furnace node connections. This topology diagram uses the installation position of the mud gun equipment as the reference node, and determines the spatial connection relationships between each node according to the blast furnace body structure drawings. Each node represents a specific location on the blast furnace, and the lines between nodes represent the actual heat conduction paths. The heat conduction parameters between each node are marked on the topology diagram, including the material thermal conductivity, contact area, and equivalent thermal resistance. Specifically, a reverse heat flow constraint condition is superimposed in the taphole region. This condition reflects the reverse influence of molten iron flow on the conventional heat conduction path through a heat flow direction correction factor. The modeling process generates two core data structures: a thermal resistance matrix recording the heat flow per unit temperature difference between nodes, and a node admittance matrix describing the heat absorption capacity of each node. The topology model comprehensively characterizes the thermodynamic properties of the blast furnace front region.

[0033] The risk propagation simulation unit maps the mud gun state difference coefficient matrix to the corresponding nodes in the blast furnace topology model. The mapping process establishes a correspondence between equipment numbers and topology nodes, associating the difference coefficient of each mud gun with a specific topology node. Risk propagation is then performed based on a graph analysis network, which transforms the topology graph into a weighted directed graph structure. During the simulation, the attenuation factor of the risk heat flow is calculated based on the node's thermal conduction parameters; this attenuation factor is inversely proportional to the thermal resistance between nodes. A multi-head attention mechanism is used to capture cross-regional risk association characteristics. This mechanism calculates the risk association weight between any two nodes, with the weight depending on the thermal conduction efficiency and spatial distance between the nodes. A stochastic simulation method is used to simulate the diffusion path of the risk heat flow in the topology network: starting from the initial node with an excessive difference coefficient, the next hop node is randomly selected based on the thermal conduction probability distribution, and the propagation path is recorded. Each simulation generates a complete propagation trajectory from the risk source to the peripheral nodes.

[0034] The probability distribution generation unit statistically analyzes the frequency of risky heat flux occurrences for each path in multiple random simulations. The residence probability of the risky heat flux is then calculated based on heat conduction parameters; this value is determined by both the path occurrence frequency and the path's heat conduction efficiency. The residence probability is calculated using the following formula:

[0035] in: Representing a path The probability of risk residence. For the total number of simulation rounds, It is a path In the The occurrence indicator function in the round simulation (1 if it occurs, 0 otherwise). For the first The weighting coefficients of the round simulation (determined by the initial node difference coefficients). It is the number of nodes contained in the path. For the first Thermal conduction correction factor for each node Indicates the first The admittance coefficient of each node. This formula quantifies the tendency of risky heat flow to accumulate along a specific path. A risk probability thermal distribution map covering the entire furnace is generated, presented in the form of a three-dimensional heat map, with different color depths representing risk probability values ​​for different regions. Suspicious paths whose probability values ​​exceed a preset risk retention probability threshold are marked; these paths constitute high-risk transmission channels.

[0036] The physical region positioning unit performs spatial clustering analysis on the risk probability heat map. A density clustering algorithm is used to identify risk probability clusters, automatically discovering continuous blocks with significantly higher probability values ​​than surrounding areas. Based on the connection between the mud gun equipment location and the blast furnace topology, the physical boundary of the risk operation is delineated with the center of the cluster as the center, combined with the heat conduction attenuation radius. The boundary range considers the impact radius of equipment operation to ensure coverage of all potentially affected equipment locations. The suspicious equipment identification generation subunit establishes a mapping relationship between nodes and mud gun equipment based on the blast furnace nodes connected by the suspicious path set. Equipment IDs matching the node coordinates are queried from the equipment location database to form a suspicious equipment identification set. This set is dynamically updated to reflect the list of equipment associated with the current risk status in real time.

[0037] In the random simulation process, the risk transmission main path generation subunit records the node paths and their order traversed in each round of transmission. A path frequency statistics table is established to record the frequency of occurrence of each node sequence. Using the maximum frequency screening method, the path sequence with the highest cumulative frequency is selected as the risk transmission main path sequence. This sequence reveals the most probable path of risk transmission, providing a basis for targeted prevention and control.

[0038] The entire implementation process transforms equipment state differences into heat flow propagation risks through thermodynamic modeling, utilizes stochastic simulation to capture uncertainties in complex systems, and ultimately outputs risk area location results with clear physical meaning. The heat conduction parameters of the topology model are derived from blast furnace design data and thermal test data, ensuring consistency between the model and the actual physical process. The multi-round design of the risk propagation simulation effectively reduces the impact of random factors, and the weighting mechanism in probability distribution generation highlights the identification of high-risk paths. Spatial clustering analysis employs an adaptive radius algorithm, automatically adjusting the clustering range based on the risk probability distribution density to ensure the accuracy of boundary delineation. The set of suspicious equipment identifiers and the main path sequence together constitute a dual chain of evidence for risk location, providing reliable input for subsequent control strategies.

[0039] Example 4: When the anomaly probability value of a certain area in the risk probability heat map exceeds the preset risk residence probability threshold, the high-frequency monitoring mode activation unit immediately issues a monitoring mode switching command to the mud gun equipment belonging to that area. The command includes three upgrade operations: the sampling frequency of the pressure sensor is increased from the baseline value of 200Hz to 800Hz, and the sampling frequency of the temperature sensor is increased from 50Hz to 200Hz; the real-time temperature component tracking mode is activated simultaneously, which separates the temperature fluctuation components in a specific frequency band (0.5-5Hz) through a bandpass filter bank to capture phase change events in real time; a transient event detector is deployed on the edge computing node of the equipment, configured to automatically trigger when the temperature change rate exceeds 15℃ / second, record the temperature waveform segment 5 seconds before and after the event, and add a millisecond-level timestamp. All upgraded data streams are transmitted to the central analysis system through a dedicated channel.

[0040] The physical disturbance test execution unit performs diagnostic tests on the adjacent node equipment group with the largest change in risk probability gradient. Taking the mud gun equipment group numbered G07-G09 as an example, its risk probability gradient reaches 0.35 / m. The test process first injects a swept-frequency test signal through a controllable heat source installed on the heat conduction path of the equipment. The signal frequency changes in 0.5Hz steps within the range of 0.1-10Hz, with each frequency point lasting for 30 seconds, and the amplitude controlled within 120% of the normal thermal fluctuation range. The theoretical response calculation subunit calculates the theoretical temperature response of each node under the test signal based on the thermal resistance matrix of the furnace front thermodynamic topology model, generating a theoretical response thermal spectrum table containing the frequency-amplitude-phase relationship, as shown in Table 1.

[0041] Table 1: Theoretical response thermal spectrum of nodes G07-G09.

[0042] During the test signal injection, the measured response acquisition subunit records the temperature data of each node at a sampling frequency of 10Hz. Measured response thermal spectrum data is generated by extracting data from the same frequency points. The abnormal offset determination subunit calculates the distance measure between the theoretical and measured thermal spectra: first, it calculates the relative amplitude error and absolute phase deviation for each frequency point, and then sums them by frequency weight to obtain the comprehensive distance measure. Taking node G08 at 2Hz as an example, the measured amplitude is 4.1℃ (theoretical value 4.6℃), and the phase is 0.49rad (theoretical value 0.45rad). The calculated amplitude offset percentage is -10.9%, and the phase deviation is 0.04rad. When the amplitude offset percentage of a node exceeds 25% or the phase deviation exceeds 0.1rad at more than three frequency points, it is marked as a heat flow abnormality associated device.

[0043] During testing, the system monitored the equipment's operating status in real time. When a 2Hz test signal was applied to equipment G07, an abnormal harmonic component was detected in the main component channel of the pressure waveform, triggering a collaborative analysis of the flow pulse auxiliary channel. The dual-channel feature fusion results showed an increase in the flow pulse interval entropy during the temperature response lag; this correlation was recorded as diagnostic auxiliary evidence. All test data and diagnostic results were automatically archived to the equipment's historical database for subsequent analysis and comparison.

[0044] For devices marked as associated with abnormal heat flow, the system generates a secondary control strategy: extending the high-frequency monitoring mode duration to three times the normal cycle; increasing the frequency of physical disturbance tests to twice per shift; and activating the device linkage monitoring mechanism, automatically expanding the monitoring range to associated devices within a 3-meter topological distance when an abnormal device triggers an alarm. The control strategy is distributed to the device controller in real time via the industrial IoT platform, and the strategy validity period is dynamically set to 4-72 hours based on the risk level.

[0045] Example 5: The motion candidate interval localization subunit divides the real-time acquired operational data into continuous signal frames, with each frame duration set to 200 milliseconds. Energy gradient analysis is performed on adjacent signal frames to calculate the inter-frame energy change rate. When the energy change rate exceeds a dynamic threshold for three consecutive frames, it is determined as the starting point of the motion candidate interval; when the energy change rate returns to the normal range and remains for more than five frames, it is determined as the end point of the interval. The localization result generates a motion event marker containing a start timestamp, duration, and maximum energy gradient. This marker is recorded in the form of a time interval, with additional energy intensity level classification information.

[0046] The body motion signal decoupling subunit performs non-stationary feature separation processing on the original signal within the marked interval. An adaptive filter bank is used to process the pressure waveform, temperature sequence, and flow pulse signal separately. The filter bank contains six parallel channels with a center frequency covering the range of 0.5Hz to 50Hz, and the bandwidth is dynamically adjusted according to the signal characteristics. Each channel outputs characteristic components of the corresponding frequency band, and coupling components unrelated to the mud gun's operating state are eliminated through cross-correlation analysis. The separation process focuses on two types of features: low-frequency, high-amplitude mechanical vibration components and mid-frequency transient impact response components. The decoupled signal retains the frequency band characteristic components that characterize the essence of the body motion event, while eliminating normal signal components related to the equipment state.

[0047] The motion intensity waveform generation subunit processes the decoupled feature components based on a motion trajectory discrimination criterion. This criterion distinguishes interference types by analyzing the distribution characteristics of the feature components in the time-frequency domain: local motion is characterized by a sudden increase in energy in a specific frequency band, typically lasting less than 300 milliseconds, with a bandwidth of less than 5 Hz; global interference presents as wideband energy changes, lasting more than 1 second, accompanied by multi-band phase synchronous drift. A spatiotemporal correlation index is calculated for each feature component; components with an index below 0.3 are classified as local motion, and those above 0.7 are classified as global interference. The amplitude information and frequency domain characteristics of all components are fused to generate a motion intensity waveform with a time resolution of 10 milliseconds. This waveform is represented in three-dimensional vector form, containing the intensity value, dominant frequency band, and interference type encoding.

[0048] The generated body motion intensity waveform is input to the feature fusion unit for weight adjustment calculation. During feature fusion, the system monitors the changing trend of the body motion intensity waveform in real time. When the waveform shows continuous high-intensity global disturbance, the weight coefficient of the principal component channel of the pressure waveform is automatically reduced; when a sudden local body motion is detected, the data update of the auxiliary channel of the flow pulse is temporarily frozen to avoid transient interference affecting the stability of feature extraction. After the body motion event ends, the system performs data integrity verification and interpolates and reconstructs the state features of the disturbed period.

[0049] This unit establishes a bidirectional communication mechanism with the main component channel construction unit. When the main component channel detects abnormal frequency band energy distribution, it triggers the motion event processing unit to initiate a specific analysis; conversely, when the motion event processing unit detects a specific frequency band interference pattern, it sends a frequency band shielding suggestion to the main component channel. It shares frequency domain analysis resources with the auxiliary channel construction unit and reuses the Fast Fourier Transform processor, reducing hardware resource overhead.

[0050] The metadata record for bodily movement events includes timestamps, durations, maximum intensity values, dominant frequency band ranges, and interference type codes. This data is stored synchronously with state features, forming a feature database marked with interference. In the subsequent multidimensional difference analysis phase, the system refers to the bodily movement marking information to perform confidence-weighted processing on the difference coefficients, reducing the influence of data from the interference period on the analysis results.

[0051] The entire processing adopts a pipelined architecture, with signal frame processing latency controlled within 50 milliseconds, meeting the timing requirements of the blast furnace real-time control system. The decoupling algorithm uses a sliding window update mechanism, refreshing filter parameters every 10 milliseconds to adapt to the dynamic characteristics of non-stationary interference. The intensity waveform generation module has a built-in self-test function, periodically performing zero-input tests and standard signal injection tests to verify the reliability of each processing stage.

[0052] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A blast furnace front mud gun control system, characterized in that, include: Data acquisition module: Real-time acquisition of blast furnace mud gun operation data, including mud gun pressure waveform, taphole temperature sequence and mud flow pulse; Status feature extraction module: Extracts the operating status features of the mud gun based on the operation data; Multidimensional difference analysis module: Performs multidimensional difference analysis on the mud gun operating state characteristics and preset theoretical state to generate mud gun state difference coefficient matrix; Risk area location module: locates the risk area of ​​blast furnace front operation based on the mud gun state difference coefficient matrix; Adaptive control strategy generation module: Generates mud cannon adaptive control strategy based on the risk area.

2. The blast furnace front mud gun control system according to claim 1, characterized in that, The data acquisition module specifically includes: Real-time operational data acquisition unit: synchronously captures the zero-crossing distortion rate and phase jitter parameters of the mud gun pressure waveform, the gradient change characteristics of the taphole temperature sequence, and the amplitude abrupt change gradient and time interval entropy value of the mud gun flow pulse through a sensor array deployed on the mud gun equipment. Historical data feature mining unit: performs multi-dimensional decomposition processing on the historical operation data of mud guns, including using an adaptive signal decomposition method to extract the energy ratio of steady-state and transient components of pressure waveforms, establishing a correlation matrix between flow pulses and mud gun thrust through load correlation analysis, and using a sequence alignment algorithm to align temperature sequence patterns under different operating conditions.

3. A blast furnace front mud gun control system according to claim 2, characterized in that, The state feature extraction module specifically includes: Signal principal component channel construction unit: Constructs a pressure waveform principal component channel based on the signal decomposition method, retaining preset frequency band components; Signal auxiliary channel establishment unit: Establishes a flow pulse auxiliary channel based on frequency domain analysis method, and uses phase decoupling technology to decouple the preset frequency band; Feature fusion unit: It evaluates the orthogonality of the dual-channel signals of the main component channel of the pressure waveform and the auxiliary channel of the flow pulse by information entropy, dynamically adjusts the parameter weights between the channels to achieve feature fusion, and outputs the operating status features of the mud gun.

4. A blast furnace front mud gun control system according to claim 1, characterized in that, The multidimensional difference analysis module specifically includes: Time-domain cumulative deviation calculation unit: The mud gun operating state characteristics and theoretical state are compared with a preset time window, the state sequence of asynchronous sampling is aligned by a sequence alignment algorithm, the cumulative deviation in each window is calculated, and a time-domain deviation vector is generated. Frequency domain energy shift detection unit: Decomposes the temperature sequence of mud gun operating state characteristics and theoretical state using frequency domain decomposition algorithm, calculates the energy spectral density ratio, extracts the energy shift index of each frequency band, and constructs frequency domain shift vector; Pulse sequence similarity evaluation unit: Based on the structure matching algorithm, the distance distribution of the flow pulse sequence of the mud gun operating state characteristics and theoretical state is matched, the phase synchronization error of the pulse amplitude change point is calculated, the divergence index of the pulse interval distribution is quantified, and the pulse similarity vector is generated. The difference coefficient matrix generation unit: concatenates the time-domain deviation vector, frequency-domain offset vector, and pulse similarity vector into tensors, and through normalization processing with feature importance weighting, outputs a third-order mud gun state difference coefficient matrix with the dimension of device number timestamp difference type.

5. A blast furnace front mud gun control system according to claim 4, characterized in that, The frequency domain energy shift detection unit further includes: Sensitive frequency band extraction subunit: Extract the temperature sequence corresponding to the operating state characteristics and theoretical state of the mud gun, and use frequency band analysis technology to perform multi-scale frequency band analysis on each sequence to extract the energy distribution characteristics within the preset sensitive frequency band interval; Energy offset quantization sub-unit: Quantizes the energy density of the theoretical state and the mud gun operating state characteristics in the sensitive frequency band interval respectively, extracts the offset index of each frequency band based on the relative offset degree, and constructs the frequency domain offset vector.

6. A blast furnace front mud gun control system according to claim 1, characterized in that, The risk area location module specifically includes: Blast furnace topology modeling unit: Construct a topology diagram of blast furnace node connection relationship based on the location information of mud gun equipment, mark the heat conduction parameters between each node, superimpose the reverse heat flow constraint conditions of the taphole area on the topology diagram, and generate a furnace front thermodynamic topology model including thermal resistance matrix and node admittance matrix. Risk propagation simulation unit: Maps the mud gun state difference coefficient matrix to the corresponding nodes of the blast furnace topology model, performs risk propagation simulation based on graph analysis network, including calculating the attenuation factor of risk heat flow according to the node heat conduction parameters, capturing cross-regional risk correlation characteristics through multi-head attention mechanism, and simulating the diffusion path of risk heat flow in the topology network using stochastic simulation method; Probability distribution generation unit: Statistically counts the frequency of risk heat flow occurrence in simulated propagation for each path, calculates the residence probability value of risk heat flow in combination with heat conduction parameters, generates a risk probability thermal distribution map covering the entire furnace, and marks the set of suspicious paths whose probability values ​​exceed the preset risk residence probability threshold; Physical region positioning unit: Performs spatial clustering analysis on the risk probability thermal distribution map, identifies risk probability clustering areas, and delineates the physical boundaries of risk operations based on the location of the mud gun equipment and the topological connection relationship with the blast furnace.

7. A blast furnace front mud gun control system according to claim 6, characterized in that, The risk area location module also includes: Suspicious equipment identifier generation subunit: Based on the blast furnace nodes connected by the suspicious path set, the blast furnace nodes are bound to the mud gun equipment to form a suspicious equipment identifier set; Risk propagation main path generation sub-unit: In the randomized simulated risk diffusion process, the node paths and their order experienced in each round of propagation are recorded, the frequency of occurrence of each path is counted, and the path sequence with the highest cumulative frequency of occurrence is selected as the risk propagation main path sequence.

8. A blast furnace front mud gun control system according to claim 1, characterized in that, The adaptive control strategy generation module specifically includes: High-frequency monitoring mode activation unit: When the abnormal probability value of a certain area in the risk probability thermal distribution map exceeds the preset risk residence probability threshold, a monitoring mode switching command is issued to the mud gun equipment to which the area belongs, including increasing the pressure and temperature sampling frequency to a multiple of the original frequency, simultaneously enabling the real-time temperature component tracking mode, capturing frequency band phase changes, and deploying transient event detectors on the edge side to record temperature transient waveform segments. Physical disturbance test execution unit: Apply multi-band physical disturbances to the adjacent node devices with the largest changes in risk probability gradient in the risk probability heat distribution map.

9. A blast furnace front mud gun control system according to claim 8, characterized in that, The physical disturbance test execution unit also includes: Test signal injection subunit: Injects sweep frequency test signals through a controllable heat source; Theoretical response calculation subunit: Calculate the theoretical response thermal spectrum based on the furnace front thermodynamic topology model and thermal resistance matrix; Measured response acquisition subunit: Record the temperature of each node after the test signal is injected to obtain the measured response thermal spectrum; Anomaly offset determination subunit: Calculates the distance measurement between the theoretical response thermal spectrum and the measured response thermal spectrum, compares the distance measurement between the measured thermal spectrum and the theoretical thermal spectrum, calculates the percentage of abnormal node offset, and marks the abnormal node offset as a heat flow abnormal associated device when the percentage of abnormal node offset exceeds the second threshold.

10. A blast furnace front mud gun control system according to claim 3, characterized in that, The state feature extraction module also includes a body motion event processing unit: Body movement candidate interval localization subunit: Performs energy gradient analysis on adjacent signal frames of the operation data, locates body movement candidate intervals, and generates body movement event markers; Body motion signal decoupling subunit: performs non-stationary feature decoupling on the marker signal of the body motion event, and separates the frequency band feature components that characterize the body motion event; Body motion intensity waveform generation subunit: Based on the motion trajectory discrimination criterion, it distinguishes between local body motion and global interference, and generates body motion intensity waveform by fusing the frequency band feature components.

Images