Intelligent control system and method for multi-sensor fusion of iron concentrate titanium reduction process

Abstract

The present application belongs to the technical field of metallurgical process control, and discloses a multi-sensor fusion intelligent control system and method for iron concentrate powder titanium reduction process, aiming to solve the problem of real-time monitoring and inhibition of local carbon deposition in the traditional process. By obtaining the temperature distribution in the furnace, the gas composition distribution, the furnace wall vibration signal and the local pressure signal, a reaction thermodynamic constraint graph in the furnace is constructed, and accordingly the furnace area is divided into multiple reaction front subareas. In each subarea, the system constructs a carbon deposition trigger network based on vibration-pressure coupling, accurately calculates the carbon deposition trigger probability, and then generates a targeted adaptive gas injection strategy. By globally optimizing the gas injection parameters, the real-time inhibition of local carbon deposition and the effective separation of titanium in the iron concentrate powder titanium reduction process are realized. The present application reduces the agglomeration of iron concentrate powder, keeps the gas channel unobstructed, and improves the titanium recovery rate and product quality.

Description

Technical Field

[0001] This invention relates to the field of metallurgical process control technology, and more specifically, to an intelligent control system and method for titanium reduction in iron concentrate using multi-sensor fusion. Background Technology

[0002] Titanium, as a metallic material, possesses excellent properties such as low density, high strength, and corrosion resistance, and has wide applications in aerospace, chemical, and medical fields. The titanium reduction process from iron concentrate is a crucial metallurgical process for extracting titanium from ilmenite. It typically employs carbothermal reduction to reduce tetravalent titanium to a lower valence state under high-temperature conditions, achieving effective separation from iron. In the carbothermal reduction process, controlling the appropriate CO / CO2 ratio is crucial for maintaining the smooth progress of the reduction reaction. Early processes relied mainly on manual experience for control, roughly judging the furnace state by observing the flame color at the furnace mouth and measuring the outlet gas temperature. With the application of online gas analyzers, real-time monitoring of the CO / CO2 ratio has become possible, improving control accuracy. Subsequently, advanced control strategies based on model predictive control (MPC) have been applied in some large-scale production lines, but limitations in model accuracy and monitoring methods still cannot solve the problem of localized carbon deposition.

[0003] In the titanium reduction process of iron concentrate, localized carbon deposition (carbon precipitation) is a critical step restricting production efficiency and product quality. Existing technologies, employing single temperature monitoring or simple gas analysis, cannot capture the precursory characteristics and dynamic evolution of carbon deposition within micro-regions of the furnace. In actual production, when the reduction reaction reaches the mid-to-late stages, areas with an imbalanced CO / CO2 ratio often develop large amounts of carbon deposits. These deposits rapidly adhere to the surface of the iron concentrate, leading to strong agglomeration among powder particles. Once agglomerated areas form, gas channels are gradually blocked, resulting in uneven gas distribution within the furnace, creating "airflow dead zones" and "airflow dominant channels," further exacerbating carbon deposition and agglomeration. This vicious cycle creates localized high-temperature zones within the furnace, with drastic temperature gradient changes that not only affect the chemical reduction process of titanium but also cause titanium to form complexes with carbon and iron that are difficult to separate. Existing CO / CO2 ratio control systems typically rely on furnace outlet gas composition analysis, which suffers from severe time lag and spatial averaging effects, making it impossible for control signals to accurately reflect the local state within the furnace. By the time the system detects an anomaly and adjusts the gas injection parameters, the problem area has often progressed to an irreversible stage. Furthermore, conventional control strategies employ uniform parameter adjustments across the entire furnace, lacking the ability to specifically intervene in localized carbon deposition areas. This allows what could have been a controllable, small-scale problem to evolve into a systemic obstacle affecting the entire reduction process, ultimately leading to a series of technical and economic problems such as low titanium recovery rates, high energy consumption, and poor batch stability in the product.

[0004] In view of this, the present invention proposes an intelligent control system and method for titanium reduction process of iron concentrate using multi-sensor fusion to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a multi-sensor fusion-based intelligent control method for titanium reduction in iron concentrate, comprising:

[0006] Step 1: Obtain furnace temperature distribution data, furnace gas composition distribution data, furnace wall vibration signal, and furnace local pressure signal at each moment during the process of reducing titanium in iron concentrate;

[0007] Step 2: Based on the spatial distribution characteristics of the furnace temperature distribution data and the furnace gas composition distribution data, construct a furnace reaction thermodynamic constraint diagram;

[0008] Step 3: Based on the aforementioned in-furnace reaction thermodynamic constraint diagram, divide the in-furnace region into multiple reaction front zones;

[0009] Step 4: Within each reaction front partition, construct a carbon deposition triggering network based on vibration-pressure coupling, and obtain the carbon deposition triggering probability of each reaction front partition;

[0010] Step 5: Generate an adaptive gas injection strategy for each reaction front partition based on the carbon deposition trigger probability of each reaction front partition;

[0011] Step 6: Based on the adaptive gas injection strategy of all the reaction front zones, optimize the gas injection parameters in the furnace to achieve real-time suppression of local carbon deposition and effective separation of titanium during the titanium reduction process of iron concentrate.

[0012] Preferably, the construction of the in-furnace reaction thermodynamic constraint diagram includes:

[0013] Based on the furnace temperature distribution data, the boundary line between the high-temperature core region and the low-temperature edge region inside the furnace is identified, and a furnace temperature gradient boundary map is constructed.

[0014] Based on the gas composition distribution data inside the furnace, local extreme points of the CO / CO2 ratio inside the furnace are identified, and a gas reactivity map inside the furnace is constructed.

[0015] The furnace temperature gradient boundary map and the furnace gas reactivity map are spatially superimposed to generate an initial thermodynamic constraint map.

[0016] In the initial thermodynamic constraint diagram, spatial conflict regions between the boundary lines in the temperature gradient boundary diagram and the extreme points in the gas reactivity diagram are identified. Thermodynamic equilibrium corrections are performed on the spatial conflict regions to obtain the in-furnace reaction thermodynamic constraint diagram.

[0017] Preferably, dividing the furnace interior into multiple reaction front zones includes:

[0018] Multiple initial tracking points are randomly selected in the aforementioned in-furnace reaction thermodynamic constraint diagram;

[0019] For each initial tracking point, the propagation process of the virtual reaction front is simulated, wherein the propagation direction of the virtual reaction front is determined by the weighted vector of the temperature gradient and the CO / CO2 ratio gradient at the corresponding position point in the furnace reaction thermodynamic constraint diagram, and the propagation speed is determined by the thermodynamic entropy increase rate at the corresponding position point in the furnace reaction thermodynamic constraint diagram.

[0020] When the thermodynamic entropy increase rate is lower than a preset entropy increase rate threshold during the propagation of the virtual reaction front, the propagation stops, and the area covered by the virtual reaction front is marked as an initial reaction front partition.

[0021] Boundary optimization is performed on all the initial reaction front partitions. The boundary optimization is performed by identifying the continuity of the temperature gradient and CO / CO2 ratio gradient between adjacent initial reaction front partitions, and smoothing the boundaries with continuity below a preset continuity threshold to obtain the reaction front partitions.

[0022] Preferably, the step of constructing a carbon deposition triggering network based on vibration-pressure coupling within each reaction front partition, and obtaining the carbon deposition triggering probability of each reaction front partition, includes:

[0023] Within each reaction front zone, based on the time-domain waveform of the furnace wall vibration signal, local abrupt change points of the vibration waveform are identified, and a sequence of vibration abrupt change points is constructed.

[0024] Within each reaction front zone, local pulse points of the pressure waveform are identified based on the time-domain waveform of the local pressure signal inside the furnace, and a pressure pulse point sequence is constructed.

[0025] Based on the time synchronization between the vibration mutation point sequence and the pressure pulse point sequence, a vibration-pressure coupling network is constructed, wherein the nodes of the vibration-pressure coupling network are the vibration mutation points and the pressure pulse points, and the edges are the time delays between the vibration mutation points and the pressure pulse points;

[0026] The carbon deposition triggering probability of the reaction front partition is calculated based on the connectivity of the vibration-pressure coupling network, wherein the connectivity is quantified by the average path length and clustering coefficient of the vibration-pressure coupling network, and the carbon deposition triggering probability is negatively correlated with the average path length and positively correlated with the clustering coefficient.

[0027] Preferably, the adaptive gas injection strategy for generating each of the reaction front partitions includes:

[0028] Within each reaction front zone, local flow field bottleneck regions are identified according to the in-furnace reaction thermodynamic constraint diagram. These local flow field bottleneck regions are the areas in the in-furnace reaction thermodynamic constraint diagram where the spatial conflict between the temperature gradient and the CO / CO2 ratio gradient is most severe.

[0029] Within the local flow field bottleneck region, a turbulence-inducing path is designed, wherein the starting point of the turbulence-inducing path is the center of the local flow field bottleneck region, the ending point is the boundary of the reaction front zone, and the path direction is determined by the thermodynamic entropy increase rate gradient of the corresponding point in the furnace reaction thermodynamic constraint diagram.

[0030] An initial gas injection strategy is generated based on the turbulence-induced path, wherein the initial gas injection strategy includes an injection angle, an injection flow rate, and an injection frequency, the injection angle being determined by the direction of the turbulence-induced path, and the injection flow rate and the injection frequency being determined by the magnitude of the carbon deposition trigger probability;

[0031] Based on the predicted suppression effect of the initial gas injection strategy on the carbon deposition triggering probability of the reaction front zone, the initial gas injection strategy is dynamically adjusted to obtain the adaptive gas injection strategy, wherein the predicted suppression effect is quantified by simulating the influence of the initial gas injection strategy on the thermodynamic entropy increase rate of the in-furnace reaction thermodynamic constraint diagram.

[0032] Preferably, the optimization of the in-furnace gas injection parameters includes:

[0033] An initial global flow field distribution map is constructed based on the adaptive gas injection strategy of all the reaction front partitions, wherein the initial global flow field distribution map is generated by superimposing the local flow field vector corresponding to the adaptive gas injection strategy of each reaction front partition;

[0034] In the initial global flow field distribution map, global flow field conflict regions are identified, wherein the global flow field conflict regions are the regions in the initial global flow field distribution map where the direction conflict of the flow field vectors is most severe;

[0035] The flow field is reconstructed in the global flow field conflict region, wherein the flow field reconstruction is performed by iteratively adjusting the adaptive gas injection strategy of adjacent reaction front partitions in the global flow field conflict region to minimize the flow field vector direction conflict in the global flow field conflict region.

[0036] Based on the global flow field distribution map after the flow field reconstruction, the gas injection parameters in the furnace are optimized. The optimization is achieved by minimizing the turbulence intensity of the global flow field distribution map and maximizing the uniformity of the thermodynamic entropy increase rate of the reaction thermodynamic constraint map in the furnace.

[0037] Preferably, the thermodynamic equilibrium correction is achieved in the following manner:

[0038] Within the spatial conflict region, a thermodynamic entropy increase optimization model is constructed, wherein the objective function of the thermodynamic entropy increase optimization model is the sum of thermodynamic entropy increases in the spatial conflict region, and the constraint condition is the physical feasibility of the temperature value and the CO / CO2 ratio within the spatial conflict region.

[0039] An iterative algorithm based on gradient descent is used to solve the optimal solution of the thermodynamic entropy increase optimization model, wherein the step size of the iterative algorithm is determined by the magnitude of the temperature gradient and the CO / CO2 ratio gradient in the spatial conflict region.

[0040] Based on the optimal solution, the temperature value and CO / CO2 ratio within the spatial conflict region are adjusted to obtain the corrected in-furnace reaction thermodynamic constraint diagram.

[0041] Preferably, the connectivity of the vibration-pressure coupling network is quantified in the following manner:

[0042] Calculate the shortest path length between all node pairs in the vibration-pressure coupled network to obtain the average path length;

[0043] For each node, calculate the edge connection density between the node and its neighboring nodes to obtain the local clustering coefficient of the node; take the average of the local clustering coefficients of all nodes to obtain the clustering coefficient.

[0044] The connectivity of the vibration-pressure coupled network is quantified by a weighted sum of the average path length and the clustering coefficients, wherein the weights of the weighted sum are determined by the signal-to-noise ratio of the furnace wall vibration signal and the local pressure signal inside the furnace.

[0045] Preferably, the uniformity of the flow field turbulence intensity and the thermodynamic entropy increase rate is quantified in the following manner:

[0046] Based on the flow field vector at each location point in the global flow field distribution map, the curl and divergence of the flow field vector are calculated to obtain the turbulence intensity at that location point; the average value of the turbulence intensity at all locations is then taken to obtain the overall flow field turbulence intensity.

[0047] Based on the thermodynamic entropy increase rate at each location point in the furnace reaction thermodynamic constraint diagram, the spatial distribution entropy of the thermodynamic entropy increase rate is calculated to obtain the uniformity of the thermodynamic entropy increase rate.

[0048] A multi-sensor fusion intelligent control system for the titanium reduction process of iron concentrate, used to implement the aforementioned intelligent control method for the titanium reduction process of iron concentrate, includes:

[0049] The data acquisition module is used to acquire data on furnace temperature distribution, furnace gas composition distribution, furnace wall vibration signal, and furnace local pressure signal at each moment during the process of reducing titanium in iron concentrate.

[0050] Thermodynamic constraint construction module is used to construct a furnace reaction thermodynamic constraint diagram based on the spatial distribution characteristics of the furnace temperature distribution data and the furnace gas composition distribution data.

[0051] The reaction partitioning module is used to divide the furnace area into multiple reaction front partitions according to the furnace reaction thermodynamic constraint diagram.

[0052] The carbon deposition analysis module is used to construct a carbon deposition triggering network based on vibration-pressure coupling within each of the reaction front partitions, and to obtain the carbon deposition triggering probability of each of the reaction front partitions.

[0053] The strategy generation module is used to generate an adaptive gas injection strategy for each reaction front partition based on the carbon deposition trigger probability of each reaction front partition.

[0054] The parameter optimization module is used to optimize the gas injection parameters in the furnace according to the adaptive gas injection strategy of all the reaction front zones, so as to achieve real-time suppression of local carbon deposition and effective separation of titanium during the titanium reduction process of iron concentrate.

[0055] The technical effects and advantages of the intelligent control system and method for titanium reduction process of iron concentrate based on multi-sensor fusion of this invention are as follows:

[0056] This invention enhances the prediction and suppression capabilities of localized carbon deposition during high-temperature reduction processes, resolving a series of cascading problems caused by lagging carbon deposition monitoring in traditional processes. By establishing a comprehensive, multi-dimensional real-time furnace state sensing system, potential risk areas can be accurately identified at the initial stage of carbon deposition formation, achieving a proactive control approach of "prevention before the event." It reduces iron concentrate agglomeration, maintains the unobstructed flow of gas channels within the reduction furnace, and ensures uniform contact of reactant gases with the material surface, thereby significantly improving the reduction and separation efficiency of titanium. Simultaneously, the thermodynamic constraint and vibration-pressure coupling analysis method enables the system to possess keen insight into microscopic changes within the furnace, eliminating the inherent response lag problem of traditional CO / CO2 proportional control systems and achieving precise adjustment of the reduction atmosphere. In practical production applications, this intelligent control method not only significantly improves the recovery rate of titanium in the product but also reduces energy consumption and equipment wear, minimizes unplanned downtime caused by carbon deposition, and improves production continuity and batch stability. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the intelligent control method for titanium reduction process of iron concentrate using multi-sensor fusion according to the present invention;

[0058] Figure 2 This is a schematic diagram of the intelligent control system for titanium reduction process of iron concentrate using multi-sensor fusion according to the present invention. Detailed Implementation

[0059] 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.

[0060] This application provides an intelligent control system and method for titanium reduction in iron concentrate using multi-sensor fusion. The system's execution entities include, but are not limited to, industrial control equipment, data processing servers, edge computing nodes, and intelligent sensor gateways, which can be considered general computing nodes in this application. The data processing platform includes, but is not limited to, at least one of the following: a process control system, a furnace temperature monitoring system, a gas analysis system, and a vibration monitoring system.

[0061] Please see Figure 1 This invention provides an intelligent control method for titanium reduction in iron concentrate using multi-sensor fusion, comprising the following steps:

[0062] Step 1: Obtain furnace temperature distribution data, furnace gas composition distribution data, furnace wall vibration signal, and furnace local pressure signal at each moment during the process of reducing titanium in iron concentrate;

[0063] Step 2: Based on the spatial distribution characteristics of the furnace temperature distribution data and the furnace gas composition distribution data, construct the furnace reaction thermodynamic constraint diagram;

[0064] Step 3: Based on the thermodynamic constraint diagram of the reaction inside the furnace, divide the furnace area into multiple reaction front zones;

[0065] Step 4: Within each reaction front partition, construct a carbon deposition triggering network based on vibration-pressure coupling to obtain the carbon deposition triggering probability of each reaction front partition;

[0066] Step 5: Generate an adaptive gas injection strategy for each reaction front partition based on the carbon deposition trigger probability of each reaction front partition;

[0067] Step 6: Based on the adaptive gas injection strategy of all reaction front zones, optimize the gas injection parameters in the furnace to achieve real-time suppression of local carbon deposition and effective separation of titanium during the titanium reduction process of iron concentrate.

[0068] This invention utilizes multi-sensor data fusion technology to ensure precise monitoring and control of the titanium reduction process in iron concentrate. It constructs an in-furnace reaction thermodynamic constraint map to provide a theoretical basis for dividing the reaction front into zones. Based on this map, the reaction front is divided into zones for refined management of the reaction area. A carbon deposition triggering network based on vibration-pressure coupling helps accurately predict carbon deposition risk areas. An adaptive gas injection strategy is generated according to the carbon deposition triggering probability to achieve precise intervention in the reaction process. Optimizing in-furnace gas injection parameters ensures effective suppression of local carbon deposition while improving titanium separation efficiency. The overall method chain constructs a complete closed-loop system from data acquisition and analysis to control execution, providing an intelligent solution for the titanium reduction process in iron concentrate.

[0069] In this embodiment of the invention, the detailed implementation steps of step 1 include:

[0070] A distributed temperature sensor array is installed inside the reactor to collect real-time temperature field data inside the furnace, obtain a high-precision temperature distribution matrix, and perform spatial interpolation and noise reduction on the temperature distribution matrix to obtain the temperature distribution data inside the furnace.

[0071] Multiple gas composition sampling probes are installed in the reactor to perform real-time gas composition analysis, obtain raw gas composition data, and then the raw gas composition data is calibrated and analyzed to obtain gas composition distribution data in the reactor.

[0072] High-sensitivity vibration sensors were installed at key locations on the reactor wall to collect the original waveform of the furnace wall vibration. The original waveform of the vibration was then subjected to spectrum analysis and feature extraction to obtain the furnace wall vibration signal.

[0073] High-temperature pressure sensors are installed in key reaction areas within the reactor to collect raw local pressure data. Dynamic response analysis and trend identification are then performed on the raw local pressure data to obtain the local pressure signal within the furnace.

[0074] In this embodiment, a distributed temperature sensor array layout scheme is first designed, considering the temperature gradient distribution characteristics and hot spot distribution patterns within the reactor. The type of temperature sensor (e.g., thermocouples, platinum resistance thermometers, fiber optic temperature sensors) and its technical parameters are determined, prioritizing sensors with high temperature resistance, fast response speed, and strong anti-interference capabilities. The spatial layout and installation positions of the sensors are determined to ensure a sufficiently high temperature monitoring density in key reaction areas while covering the entire reactor space. A distributed temperature sensor array is installed to collect real-time temperature field data within the reactor, including the temperature values ​​and timestamps of each measuring point. The sampling frequency is dynamically adjusted according to the reaction rate, typically 1-10Hz, forming a raw temperature data matrix. The raw temperature data undergoes preprocessing, including outlier detection and correction, signal denoising (e.g., median filtering, wavelet transform denoising), and data standardization. Spatial interpolation algorithms (e.g., Kriging interpolation, radial basis function interpolation) are used to interpolate the discrete temperature measuring point data, generating a continuous temperature field distribution and obtaining a high-precision temperature distribution matrix. This matrix represents the temperature value at any location within the reactor. Filtering algorithms (such as Kalman filtering and particle filtering) are applied to denoise the temperature distribution matrix, improving the signal-to-noise ratio and accuracy of the temperature data. This yields the furnace temperature distribution data, providing foundational data for subsequent thermodynamic constraint diagram construction. A multi-point gas composition sampling system is designed, determining the spatial distribution of sampling points, focusing on covering the reaction front, hotspot, and boundary regions. Suitable gas analysis techniques are selected, such as gas chromatography (GC), mass spectrometry (MS), infrared spectroscopy (IR), or laser absorption spectroscopy (LAS). Considering factors such as gas component characteristics, measurement accuracy requirements, and response time, multi-point gas composition sampling probes are installed within the reactor. The probe materials are selected from high-temperature resistant and corrosion-resistant special alloys or ceramic materials. The probe design considers gas flow dynamics to avoid interference with the reaction flow field during sampling. Real-time gas composition analysis is performed, monitoring the concentration distribution of key gas components such as CO, CO2, and O2. The sampling frequency is set according to the reaction kinetics characteristics, typically 0.At 5-5Hz, raw gas composition data is obtained. This raw data is then calibrated using standard gases periodically to establish calibration curves, eliminate instrument drift and systematic errors, and improve measurement accuracy using multivariate correction algorithms. Component analysis is performed on the calibrated data to calculate key reaction indicators such as the CO / CO2 ratio. A chemical equilibrium model is applied to assess the reaction process, obtaining gas composition distribution data within the furnace. This provides gas-phase reaction information for the thermodynamic constraint diagram, identifying key locations for furnace wall vibration monitoring, including furnace walls near areas of intense reaction, structural support points, and areas historically prone to carbon deposition. Suitable locations are then selected. Vibration sensor types, such as piezoelectric accelerometers and MEMS accelerometers, are selected. Considering factors such as measurement range, frequency response, sensitivity, and environmental resistance, high-sensitivity vibration sensors are installed at key locations on the reactor wall. Special high-temperature installation methods and heat insulation measures are used to protect the sensors. The sensor signals are transmitted to the data acquisition system via anti-interference cables to acquire the raw waveforms of the furnace wall vibration. The sampling frequency is typically 500-2000Hz to capture high-frequency vibration characteristics. Spectral analysis is performed on the raw vibration waveforms, and methods such as Fast Fourier Transform (FFT), wavelet transform, or Hilbert-Huang transform are applied to analyze the vibration signal. By analyzing the frequency domain characteristics of the signal, identifying characteristic frequency components, and extracting vibration characteristic parameters such as dominant frequency, frequency band energy distribution, and harmonic ratio from the spectrum, a vibration characteristic vector is constructed to obtain the furnace wall vibration signal. This signal serves as an important indicator for early warning of carbon deposition. Pressure monitoring points in key areas within the reactor are determined, with a focus on the reaction front, flow bottleneck areas, and areas with significant historical pressure fluctuations. Suitable high-temperature resistant pressure sensors, such as ceramic pressure sensors or silicon sapphire pressure sensors, are selected, considering measurement range, accuracy, response time, and high-temperature resistance. These sensors are installed in key reaction areas within the reactor and are water-cooled. Gas cooling technology protects the sensor, ensuring stable operation in high-temperature environments. The sensor signal, after isolation, amplification, and filtering, is transmitted to the data acquisition system to collect raw local pressure data. The sampling frequency is typically 50-200Hz, capturing mid-frequency pressure fluctuations. Dynamic response analysis is performed on the raw local pressure data, applying time series analysis methods (such as autoregressive moving average models and exponential smoothing) to identify pressure change patterns, calculate pressure pulsation frequency, amplitude, and phase characteristics, identify pressure fluctuation trends and patterns, detect abrupt changes and abnormal fluctuations, and obtain the local pressure signal within the furnace. This provides crucial input for the construction of the carbon deposition trigger network.

[0075] In this embodiment of the invention, the detailed implementation steps of step 2 include:

[0076] Based on the temperature distribution data inside the furnace, the boundary line between the high-temperature core region and the low-temperature edge region inside the furnace is identified, and a temperature gradient boundary map inside the furnace is constructed.

[0077] Based on the distribution data of gas composition in the furnace, local extreme points of the CO / CO2 ratio in the furnace are identified, and a gas reactivity map in the furnace is constructed.

[0078] The initial thermodynamic constraint map is generated by spatially overlaying the furnace temperature gradient boundary map with the furnace gas reactivity map.

[0079] In the initial thermodynamic constraint diagram, spatial conflict regions between the boundary lines in the temperature gradient boundary diagram and the extreme points in the gas reactivity diagram are identified. Thermodynamic equilibrium corrections are then performed on the spatial conflict regions to obtain the in-furnace reaction thermodynamic constraint diagram.

[0080] In this embodiment, firstly, numerical analysis methods are used to process the furnace temperature distribution data. Temperature gradient calculation and edge detection algorithms are applied to identify regions with significant temperature changes, locating the boundary between the high-temperature core region (the most active reaction region) and the low-temperature edge region (the relatively inactive reaction region). This boundary typically appears as a continuous curved surface with a large temperature gradient. These boundary lines are visualized using three-dimensional isosurface technology to form a furnace temperature gradient boundary map. This map visually displays the spatial structure and gradient distribution of the furnace temperature field. Next, the furnace gas composition distribution data is analyzed, with particular attention to the spatial distribution of the CO / CO2 ratio. This ratio is a key indicator for measuring the degree of reduction reaction and the risk of carbon deposition. Numerical optimization methods are used to identify local extreme points of the CO / CO2 ratio, including local maximum points (active reduction reaction regions) and local minimum points (reduction reaction lag regions). Interpolation algorithms are applied to generate a continuous CO / CO2 ratio distribution field, forming a furnace gas reactivity map. This map reflects the chemical reactivity state of different regions. The furnace temperature gradient boundary map and the furnace gas reactivity map are then imported into the same three-dimensional coordinate system. The system performs spatial overlay and registration, and uses spatial analysis techniques similar to Geographic Information System (GIS) to process the two layers, generating a composite layer containing both temperature gradient and gas reactivity information, forming an initial thermodynamic constraint map. This map comprehensively expresses the constraints of thermodynamic and kinetic factors on the reaction. In the initial thermodynamic constraint map, the spatial relationship between the temperature gradient boundary line and the extreme point of the CO / CO2 ratio is analyzed to identify regions where there are contradictions or inconsistencies, i.e., spatial conflict regions. These regions are usually characterized by the reaction direction predicted by the temperature gradient being inconsistent with the reaction direction predicted by the gas composition. Based on the basic principles of thermodynamics, thermodynamic equilibrium correction is performed on the spatial conflict regions, and a thermodynamic model based on the minimization of Gibbs free energy is established, considering the comprehensive influence of factors such as temperature, gas composition, and pressure. The equilibrium state of the system at each spatial point is calculated, and the inconsistent parts in the initial thermodynamic constraint map are corrected, finally forming the in-furnace reaction thermodynamic constraint map. This map accurately describes the thermodynamic constraint conditions in the process of reducing titanium in iron concentrate, providing a theoretical basis for subsequent reaction zone division and control strategy formulation.

[0081] In this embodiment of the invention, the detailed implementation steps of step 3 include:

[0082] Multiple initial tracking points are randomly selected in the thermodynamic constraint diagram of the reaction inside the furnace;

[0083] For each initial tracking point, the propagation process of the virtual reaction front is simulated. The propagation direction of the virtual reaction front is determined by the weighted vector of the temperature gradient and the CO / CO2 ratio gradient at the corresponding position in the furnace reaction thermodynamic constraint diagram, and the propagation speed is determined by the thermodynamic entropy increase rate at the corresponding position in the furnace reaction thermodynamic constraint diagram.

[0084] When the rate of increase in thermodynamic entropy falls below a preset threshold during the propagation of the virtual reaction front, the propagation stops and the area covered by the virtual reaction front is marked as an initial reaction front partition.

[0085] Boundary optimization is performed on all initial reaction front partitions. Boundary optimization is achieved by identifying the continuity of the temperature gradient and CO / CO2 ratio gradient between adjacent initial reaction front partitions, and smoothing the boundaries with continuity below a preset continuity threshold to obtain the reaction front partitions.

[0086] In this embodiment, multiple initial tracking points are first selected in the three-dimensional space of the furnace reaction thermodynamic constraint diagram using a stratified random sampling method. These points are distributed at different locations within the furnace, covering different temperature ranges and regions with different gas compositions, ensuring the representativeness and diversity of the tracking points. The number of initial points is adaptively determined based on the furnace size and complexity, typically ranging from 50 to 200. For each initial tracking point, the propagation process of the virtual reaction front is simulated using a coupled computational fluid dynamics (CFD) and chemical kinetics method, constructing a propagation direction vector field. This vector field consists of a temperature gradient vector and a CO / CO2 ratio gradient. The weighted combination of degree vectors is determined, with weighting coefficients based on reaction sensitivity analysis. For example, the temperature gradient weight is 0.6, and the CO / CO2 ratio gradient weight is 0.4. A propagation velocity field is constructed, with the propagation velocity determined by the local thermodynamic entropy increase rate. A higher entropy increase rate indicates a more active reaction and a faster propagation velocity. The entropy increase rate is calculated using local temperature, pressure, and gas composition. An iterative method is used to simulate the gradual propagation of the virtual reaction front. In each iteration step, the propagation direction and velocity at each point on the front are calculated, and the front position is updated. During propagation, the thermodynamic entropy increase rate at each front point is monitored in real time. When the entropy increase rate is lower than a certain value, the propagation velocity is recorded. When a preset threshold (e.g., 15% of the maximum entropy increase rate) is reached, the reactivity of that point is deemed insufficient to continue propagating, and the propagation process at that point is stopped. When all points in the entire virtual reaction front stop propagating or reach the furnace wall boundary, the entire area covered by the virtual reaction front from the initial point to the final front line is marked as an initial reaction front partition. The above process is repeated for all initial tracking points to obtain multiple initial reaction front partitions. These partitions may overlap or have irregular boundaries. The boundary characteristics between adjacent initial reaction front partitions are analyzed, and the continuity indices of the temperature gradient and CO / CO2 ratio gradient on the boundary are calculated, such as the angle of gradient direction and the rate of change of gradient magnitude. Boundary segments with continuity below the preset threshold are identified. These boundary segments usually represent areas with unreasonable partitioning. Boundary smoothing algorithms, such as B-spline curve fitting or morphological smoothing, are applied to boundary segments with low continuity to make the boundaries more consistent with physical meaning. Highly similar adjacent partitions are merged, and individual partitions with significant differences in physical properties are segmented to finally form optimized reaction front partitions. These partitions have clear physical meaning, smooth boundaries, and relatively uniform internal characteristics, providing a spatial basis for subsequent carbon deposition risk analysis and control strategy formulation.

[0087] In this embodiment of the invention, the detailed implementation steps of step 4 include:

[0088] Within each reaction front zone, based on the time-domain waveform of the furnace wall vibration signal, local abrupt change points of the vibration waveform are identified, and a sequence of vibration abrupt change points is constructed.

[0089] Within each reaction front zone, based on the time-domain waveform of the local pressure signal inside the furnace, local pulse points of the pressure waveform are identified, and a pressure pulse point sequence is constructed.

[0090] Based on the time synchronization between the vibration mutation point sequence and the pressure pulse point sequence, a vibration-pressure coupling network is constructed, where the nodes of the vibration-pressure coupling network are the vibration mutation points and the pressure pulse points, and the edges are the time delays between the vibration mutation points and the pressure pulse points.

[0091] Based on the connectivity of the vibration-pressure coupled network, the carbon deposition triggering probability of the reaction front partition is calculated. The connectivity is quantified by the average path length and clustering coefficient of the vibration-pressure coupled network. The carbon deposition triggering probability is negatively correlated with the average path length and positively correlated with the clustering coefficient.

[0092] In this embodiment, firstly, for each reaction front zone, the corresponding furnace wall vibration signal is processed. Time-frequency analysis methods such as wavelet analysis or Hilbert-Huang transform are applied to identify local abrupt changes in the vibration waveform, including amplitude abrupt changes, frequency abrupt changes, and phase abrupt changes. These abrupt changes are usually related to sudden movements of materials inside the furnace or changes in local reaction states. The timestamp, amplitude, and duration of each abrupt change point are recorded to form a vibration abrupt change point sequence. This sequence reflects the changing process of the furnace's dynamic state. Similarly, for each reaction front zone, the corresponding local furnace pressure signal is processed, and peak detection algorithms and outlier identification methods are used to identify... Local pulse points in the pressure waveform, including pressure peaks, pressure troughs, and abrupt pressure gradient changes, are typically associated with changes in gas flow patterns, chemical reaction rates, or localized gas flow impacts. The timestamp, amplitude, and duration of each pulse point are recorded to form a pressure pulse point sequence. This sequence reflects the changing fluid dynamics within the furnace. The temporal correlation between the vibration abrupt change sequence and the pressure pulse point sequence is analyzed. The time delay between any two points (one from the vibration sequence and one from the pressure sequence) is calculated. If the time delay is within a preset time window (e.g., 0-5 seconds), a potential causal relationship is considered to exist between these two points. To establish a correlation, a network structure is constructed where network nodes include all vibration abrupt change points and pressure pulse points. If the time delay between two nodes satisfies the correlation condition, an edge is established between these two nodes, with the edge weight determined by the magnitude of the time delay, forming a vibration-pressure coupled network. This network describes the spatiotemporal correlation pattern between vibration and pressure events. Using complex network analysis methods, key topological properties of the vibration-pressure coupled network are calculated, including the average path length (the average shortest path between any two nodes in the network) and the clustering coefficient (describing the degree of clustering of nodes in the network). The shorter the average path length, the stronger the correlation between vibration and pressure events. The faster the propagation speed, the tighter the coupling; the higher the clustering coefficient, the more local clusters are formed in the network, and the stronger the system's synergy. Based on network theory and experimental verification, a calculation model for the carbon deposition triggering probability is established. This model expresses the quantitative relationship between network topology characteristics and carbon deposition risk. For example, the carbon deposition triggering probability is negatively correlated with the average path length (the shorter the path, the higher the risk) and positively correlated with the clustering coefficient (the stronger the clustering, the higher the risk). The carbon deposition triggering probability of each reaction front partition is calculated using this model. The probability value is between 0 and 1, which represents the degree of risk of carbon deposition in that region. This probability value provides a quantitative basis for the subsequent control strategy formulation.

[0093] In this embodiment of the invention, the detailed implementation steps of step 5 include:

[0094] Within each reaction front zone, based on the in-furnace reaction thermodynamic constraint diagram, local flow field bottleneck regions are identified within the reaction front zone. These local flow field bottleneck regions are the areas in the in-furnace reaction thermodynamic constraint diagram where the spatial conflict between the temperature gradient and the CO / CO2 ratio gradient is most severe.

[0095] Within the local flow field bottleneck region, a turbulence-induced path is designed. The starting point of the turbulence-induced path is the center of the local flow field bottleneck region, and the ending point is the boundary of the reaction front zone. The path direction is determined by the thermodynamic entropy increase rate gradient of the corresponding point in the furnace reaction thermodynamic constraint diagram.

[0096] Based on the turbulence-induced path, an initial gas injection strategy is generated, which includes injection angle, injection flow rate and injection frequency. The injection angle is determined by the direction of the turbulence-induced path, and the injection flow rate and injection frequency are determined by the magnitude of the carbon deposition trigger probability.

[0097] Based on the predicted suppression effect of the initial gas injection strategy on the carbon deposition triggering probability of the reaction front zone, the initial gas injection strategy is dynamically adjusted to obtain an adaptive gas injection strategy. The predicted suppression effect is quantified by simulating the influence of the initial gas injection strategy on the thermodynamic entropy increase rate of the reaction thermodynamic constraint diagram in the furnace.

[0098] In this embodiment, firstly, for each reaction front zone, the in-furnace reaction thermodynamic constraint diagram of that region is analyzed, with particular attention paid to the spatial distribution relationship of the temperature gradient and the CO / CO2 ratio gradient. The spatial conflict index of the two gradient fields is calculated, such as the negative value of the cosine of the angle between the gradient directions. The higher the conflict index, the more inconsistent the directions of the two gradient fields are, and the greater the resistance to fluid flow. Regions with conflict indices exceeding a threshold (e.g., 0.7) are identified. These regions are usually bottleneck areas where gas flow is obstructed and the exchange of reactants and products is hindered. These are marked as local flow field bottleneck regions. Then, within each identified local flow field bottleneck region, gas turbulence inducement is designed. The guiding path starts at the geometric center of the bottleneck region, i.e., the point with the highest conflict index, and ends at the boundary of the reaction front zone, enabling effective gas discharge from the bottleneck region. The path direction is designed based on the thermodynamic entropy increase rate gradient field, arranging the path along the negative direction of the entropy increase rate gradient. This allows the gas flow to move from the high entropy increase region (active reaction zone) to the low entropy increase region (slow reaction zone), promoting the discharge of reaction products and the introduction of fresh reactants. The path shape uses a three-dimensional spline curve to ensure a smooth and continuous path, avoiding flow losses caused by sharp turns. Based on the designed turbulence-induced path, key gas injection parameters are determined, including injection... The injection angle, injection flow rate, and injection frequency are all considered. The injection angle is aligned with the tangent of the turbulence-induced path at the injection point to ensure the initial airflow direction follows the designed path. The injection flow rate is positively correlated with the carbon deposition triggering probability of the reaction front zone; the higher the risk, the stronger the required gas disturbance. A piecewise linear or exponential function is used to establish the mapping relationship between flow rate and probability. The injection frequency is also related to the carbon deposition triggering probability, but upper and lower limits are set to consider the system response characteristics. The combination of these parameters forms the initial gas injection strategy, and a predictive model for the gas injection effect is constructed. This model is based on computational fluid dynamics principles to simulate the effect of gas injection on the flow field and thermodynamic state within the furnace. The effects of the state, especially the change in the thermodynamic entropy increase rate distribution, are considered. The average entropy increase rate change in the bottleneck region before and after injection is calculated as a quantitative indicator of the predicted suppression effect. If the predicted suppression effect is not ideal (e.g., the entropy increase rate is reduced by less than 20%), the parameters of the initial injection strategy are adjusted. Optimization methods such as gradient descent or genetic algorithms are usually used to find the optimal parameter combination. The prediction and adjustment process is repeated until a satisfactory suppression effect is achieved or the maximum number of iterations is reached, ultimately forming an adaptive gas injection strategy. This strategy can adaptively adjust the gas injection parameters according to the characteristics of the bottleneck region and the degree of carbon deposition risk, thereby achieving effective suppression of carbon deposition.

[0099] In this embodiment of the invention, the detailed implementation steps of step 6 include:

[0100] An initial global flow field distribution map is constructed based on the adaptive gas injection strategies of all reaction front partitions. The initial global flow field distribution map is generated by superimposing the local flow field vectors corresponding to the adaptive gas injection strategies of each reaction front partition.

[0101] In the initial global flow field distribution map, identify the global flow field conflict region, which is the region in the initial global flow field distribution map where the direction conflict of the flow field vector is the most serious;

[0102] The flow field is reconstructed in the global flow field conflict region. The flow field reconstruction is carried out by iteratively adjusting the adaptive gas injection strategy of adjacent reaction front partitions in the global flow field conflict region to minimize the flow field vector direction conflict in the global flow field conflict region.

[0103] Based on the global flow field distribution map after flow field reconstruction, the gas injection parameters in the furnace are optimized. The optimization is achieved by minimizing the turbulence intensity of the global flow field distribution map and maximizing the uniformity of the thermodynamic entropy increase rate of the reaction thermodynamic constraint map in the furnace.

[0104] In this embodiment, adaptive gas injection strategies for all reaction front zones are first collected. A computational fluid dynamics model is used to simulate the local flow field generated by each injection strategy in its corresponding zone, calculating flow field parameters such as velocity vector field, pressure distribution, and turbulence characteristics. The local flow field vectors of all reaction front zones are interpolated and superimposed across the entire furnace space. Considering the superposition effect of the flow field and boundary condition constraints, an initial global flow field distribution map covering the entire furnace space is generated. This map describes the overall flow pattern of the gas in the furnace when all injection strategies are applied simultaneously. In the initial global flow field distribution map, the directional consistency index of the flow field vectors is calculated, such as the flow field vector of adjacent grid points. The cosine of the angle between the flow paths is used to identify regions where the directional consistency index is below a threshold (e.g., 0.3). These regions indicate significant conflicts in the flow field directions generated by the injection strategies of different reaction front zones, and are marked as global flow field conflict regions. Flow field reconstruction is then performed on these identified global flow field conflict regions. First, the adjacent reaction front zones and their injection strategies involved in the conflict regions are analyzed to establish a conflict resolution priority. Zones with lower carbon deposition triggering probabilities are adjusted first to ensure control of high-risk areas. For each zone requiring adjustment, the parameters of its injection strategy, such as injection angle, flow rate, and frequency, are modified using iterative optimization methods, such as the Alternating Direction Multiplier (AD) method. The algorithm uses a multi-modal (MM) or collaborative optimization algorithm to adjust the strategy of one partition in each iteration, then recalculates the global flow field, evaluates the changes in the degree of conflict, and continues the iteration process until the consistency of the flow field direction in the conflict area reaches a satisfactory level or the maximum number of iterations is reached. Finally, a global flow field distribution map after flow field reconstruction is obtained. Based on the results of flow field reconstruction, the specific parameters of each gas injector are further optimized, including hardware parameters such as nozzle type, orifice size, installation position, and injection pressure, as well as control parameters such as injection timing, duration, and alternation mode. The optimization objective is twofold: firstly, to minimize the turbulence intensity of the global flow field to avoid excessive turbulence disturbances from disrupting the stability of the response. The turbulence intensity is quantified by the Reynolds stress tensor or turbulent kinetic energy distribution. On the other hand, maximizing the spatial uniformity of the thermodynamic entropy increase rate promotes uniform reaction. The uniformity of the entropy increase rate is quantified by the spatial standard deviation of the entropy increase rate or the inverse of the coefficient of variation. Using multi-objective optimization methods, such as Pareto front search or weighted objective function method, these two potentially conflicting objectives are balanced to obtain the optimal combination of gas injection parameters. Finally, a complete in-furnace gas injection parameter optimization scheme is formed. This scheme considers local control requirements and global synergistic effects, and can achieve real-time suppression of local carbon deposition and effective separation of titanium during the titanium reduction process of iron concentrate, thereby improving production efficiency and product quality.

[0105] In this embodiment of the invention, the thermodynamic equilibrium correction is achieved in the following manner:

[0106] Within the spatial conflict region, a thermodynamic entropy increase optimization model is constructed. The objective function of the thermodynamic entropy increase optimization model is the sum of thermodynamic entropy increases in the spatial conflict region, and the constraint condition is the physical feasibility of the temperature value and the CO / CO2 ratio within the spatial conflict region.

[0107] An iterative algorithm based on gradient descent is used to solve the optimal solution of the thermodynamic entropy increase optimization model. The step size of the iterative algorithm is determined by the magnitude of the temperature gradient and the CO / CO2 ratio gradient in the spatial conflict region.

[0108] Based on the optimal solution, the temperature and CO / CO2 ratio within the spatial conflict region are adjusted to obtain the corrected in-furnace reaction thermodynamic constraint diagram.

[0109] In this embodiment, the regions where the identified temperature gradient and CO / CO2 ratio gradient spatially conflict are first finely divided, and a discretized grid model is constructed. Each grid point contains state variables such as temperature and CO / CO2 ratio. Based on the principle of the second law of thermodynamics, a thermodynamic entropy increase optimization model is constructed. The objective function of this model is defined as the sum of the thermodynamic entropy increase rates of all grid points in the spatial conflict region. The entropy increase rate is calculated using the Gibbs free energy change rate, taking into account factors such as temperature, pressure, and gas composition. Constraints are set to ensure that the solution results are physically feasible, including temperature constraints (e.g., the temperature value is not lower than the minimum reaction temperature and not higher than the upper limit of the equipment tolerance), CO / CO2 ratio constraints (e.g., the ratio is within the range allowed by thermodynamic equilibrium), continuity constraints of state variables (to prevent physically unreasonable jumps), energy conservation and mass conservation constraints, etc. The gradient descent method is used to solve the optimization model. First, the initial state values ​​are randomly initialized or the original state values ​​are used as the starting point. The objective function value and the constraint conditions are satisfied under the current state. The gradient of the objective function with respect to each state variable is calculated to determine the search direction. The step size of the gradient descent is adaptive. The adjustment process is proportional to the magnitudes of the local temperature gradient and the CO / CO2 ratio gradient. Smaller step sizes are used in regions with larger gradients to improve accuracy, while larger step sizes are used in regions with smaller gradients to accelerate convergence. State variables are updated along the search direction, and the updated state is checked to ensure it meets the constraints. If not, projection correction is performed to bring the state back into the feasible region. This process is repeated until convergence conditions are met (e.g., the gradient norm is less than a threshold or the number of iterations reaches the upper limit), yielding the optimal solution of the optimization model. Based on the optimal solution, the temperature and CO / CO2 ratio values ​​in spatial conflict regions are adjusted to ensure the adjusted state satisfies the thermodynamic equilibrium principle. Interpolation methods are used during the adjustment process to ensure the smoothness of the spatial distribution and avoid artificially introducing new discontinuities. The adjusted temperature field and gas composition field are remapped onto the original in-furnace reaction thermodynamic constraint map, updating the data in conflict regions while keeping the original data in non-conflict regions unchanged. Finally, a corrected in-furnace reaction thermodynamic constraint map is obtained. This map accurately describes the constraints of the in-furnace reaction while satisfying thermodynamic principles, providing a reliable theoretical basis for subsequent reaction front partitioning.

[0110] In this embodiment of the invention, the connectivity of the vibration-pressure coupling network is quantified in the following way:

[0111] Calculate the shortest path length between all node pairs in the vibration-pressure coupled network, and obtain the average path length;

[0112] For each node, calculate the edge connection density between the node and its neighboring nodes to obtain the local clustering coefficient of the node; take the average of the local clustering coefficients of all nodes to obtain the clustering coefficient.

[0113] The connectivity of the vibration-pressure coupled network is quantified by the weighted sum of the average path length and the clustering coefficient, where the weights of the weighted sum are determined by the signal-to-noise ratio of the furnace wall vibration signal and the local pressure signal inside the furnace.

[0114] In this embodiment, the constructed vibration-pressure coupled network is first represented as an adjacency matrix or adjacency list, where matrix elements or list entries indicate whether there is a connection between nodes and the strength of that connection. For each pair of nodes (i,j) in the network, the shortest path length from node i to node j is calculated using breadth-first search or Dijkstra's algorithm, i.e., the minimum number of edges traversed between the two nodes. If there is no path between the two nodes, the distance is defined as infinity or the maximum diameter of the network plus 1. The average path length of the entire network is calculated for all reachable node pairs. This index reflects the efficiency of information propagation in the network. In carbon deposition triggered networks, a shorter average path length means that vibration and pressure events can quickly interact with each other, and the triggering process is more agile. For each node i in the network, its set of all directly connected neighbor nodes N(i) is identified, and the actual number of connections E(i) between nodes in N(i) is calculated. The local clustering coefficient of node i is calculated, defined as the ratio of E(i) to the maximum number of connections that nodes in N(i) can form. For a node with k neighbors, its local clustering coefficient is... This coefficient describes the density of a node's neighborhood. The arithmetic mean of the local clustering coefficients of all nodes in the network is used to obtain the clustering coefficient of the entire network. This index reflects the trend of node clustering in the network. In carbon deposition triggered networks, a higher clustering coefficient indicates that vibration and pressure events tend to form closely related groups, enhancing the triggering effect. The quality of furnace wall vibration signals and local pressure signals within the furnace is analyzed, and the signal-to-noise ratio (SNR) of each signal is calculated. Power spectral density analysis or wavelet transform methods can be used to estimate the effective and noise components in the signal. The weight of the average path length and clustering coefficient in connectivity calculation is determined based on signal quality. Signals with higher SNR receive greater weight, ensuring that high-quality signals play a greater role in the evaluation. A connectivity quantification formula is constructed. Where LT represents connectivity, w1 and w2 are weight coefficients, and w1+w2=1; PU represents average path length, and GU represents clustering coefficient;

[0115] The formula for calculating the weighting coefficient is:

[0116] , Wherein, SNR_Y is the signal-to-noise ratio of the pressure signal and SNR_Z is the signal-to-noise ratio of the vibration signal; in this way, the connectivity index comprehensively considers the network propagation efficiency (the inverse of the average path length) and the clustering effect (clustering coefficient), and performs reasonable weighting according to the signal quality, providing a reliable quantitative basis for the calculation of carbon deposition triggering probability of network topology features.

[0117] In this embodiment of the invention, the uniformity of turbulence intensity and thermodynamic entropy increase rate in the flow field is quantified in the following way:

[0118] Based on the flow field vector at each location point in the global flow field distribution diagram, calculate the curl and divergence of the flow field vector to obtain the turbulence intensity at that location point; take the average value of the turbulence intensity at all locations to obtain the overall turbulence intensity of the flow field.

[0119] Based on the thermodynamic entropy increase rate at each location point in the furnace reaction thermodynamic constraint diagram, the spatial distribution entropy of the thermodynamic entropy increase rate is calculated to obtain the uniformity of the thermodynamic entropy increase rate.

[0120] In this embodiment, firstly, based on the global flow field distribution map after flow field reconstruction, the flow field vector V(x,y,z) at each location point on the discrete grid in the furnace space is obtained, including velocity components in three directions (x,y,z). The curl of the flow field vector is calculated using the numerical differentiation method, expressed as... ×V(x,y,z) reflects the local rotational characteristics of the fluid. Curl is a vector field; mapping this vector field to a vector shows that the magnitude of the curl is proportional to the intensity of the fluid rotation. The divergence of the flow field vector is calculated and expressed as... V(x,y,z) reflects the local expansion or contraction characteristics of the fluid. Divergence is a scalar field that maps a vector field to a scalar. The absolute value of the divergence is proportional to the rate of change of fluid volume. Combining information from curl and divergence, a turbulence intensity index is defined for the location point (x,y,z): Where α and β are weighting coefficients, determined based on the relative contributions of rotation and expansion to turbulence formation. Typically, α > β because the rotation effect plays a dominant role in turbulence formation. The overall turbulence intensity of the flow field is obtained by averaging the turbulence intensity indices of all grid points within the furnace. Where N is the total number of grid points, this index reflects the degree of turbulence in the entire flow field. Moderate turbulence is beneficial for the mixing of matter and heat, but excessive turbulence will interfere with the stability of the reaction. The thermodynamic entropy increase rate s(x,y,z) of each location point is extracted from the thermodynamic constraint diagram of the reaction in the furnace. This value reflects the local reaction activity and energy conversion efficiency. The concept of information entropy is used to quantify the spatial distribution uniformity of the thermodynamic entropy increase rate. First, the entropy increase rate range is divided into m equally spaced intervals. The number of grid points in each interval is calculated to obtain the frequency distribution p_i, where p_i represents the proportion of grid points whose entropy increase rate falls within the interval [s_i,s_i+1). The Shannon entropy formula is applied to calculate the distribution entropy. When the entropy increase rate of the distribution entropy is completely uniformly distributed, all p_i are equal, and the distribution entropy reaches its maximum value log m, where m is the total number of equally spaced intervals. When the entropy increase rate is completely concentrated in one interval, only one p_i is 1, and the rest are 0, and the distribution entropy reaches its minimum value of 0. The calculated distribution entropy H is standardized as a thermodynamic entropy increase rate uniformity index: The value range of this index is [0,1]. The closer the value is to 1, the more uniform the entropy increase rate distribution; the closer the value is to 0, the more uneven the distribution. An ideal in-furnace reaction should have a moderately uniform entropy increase rate distribution, avoiding both reaction "hot spots" leading to local overheating and reaction "cold spots" leading to incomplete reaction. In this way, turbulence intensity TI and entropy increase rate uniformity HU serve as two key indicators to guide the optimization of gas injection parameters. In the actual optimization process, upper limits for TI and lower limits for HU are usually set, or a weighted objective function is constructed. Minimize the flow field, where γ1 and γ2 are weighting coefficients that are adjusted according to process requirements to balance the requirements of flow field stability and reaction uniformity.

[0121] The above describes the intelligent control method for titanium reduction in iron concentrate using multi-sensor fusion in the embodiments of this application. The following describes the intelligent control system for titanium reduction in iron concentrate using multi-sensor fusion in the embodiments of this application. Please refer to [link to relevant documentation]. Figure 2 One embodiment of the intelligent control system for titanium reduction process of iron concentrate using multi-sensor fusion in this application includes:

[0122] The data acquisition module is used to acquire data on furnace temperature distribution, furnace gas composition distribution, furnace wall vibration signal, and furnace local pressure signal at each moment during the process of reducing titanium in iron concentrate.

[0123] The thermodynamic constraint construction module is used to construct a furnace reaction thermodynamic constraint diagram based on the spatial distribution characteristics of furnace temperature distribution data and furnace gas composition distribution data.

[0124] The reaction partitioning module is used to divide the furnace area into multiple reaction front zones based on the furnace reaction thermodynamic constraint diagram;

[0125] The carbon deposition analysis module is used to construct a carbon deposition triggering network based on vibration-pressure coupling within each reaction front partition, and to obtain the carbon deposition triggering probability of each reaction front partition.

[0126] The strategy generation module is used to generate an adaptive gas injection strategy for each reaction front partition based on the carbon deposition trigger probability of each reaction front partition.

[0127] The parameter optimization module is used to optimize the gas injection parameters in the furnace according to the adaptive gas injection strategy of all reaction front zones, so as to achieve real-time suppression of local carbon deposition and effective separation of titanium during the titanium reduction process of iron concentrate.

[0128] The modules are connected via wired and / or wireless means to enable data transmission between them.

[0129] This invention acquires comprehensive information about the furnace state through multi-sensor data fusion, constructs a reaction thermodynamic constraint diagram, and divides the reaction front into zones, achieving accurate modeling of complex reaction processes within the furnace. It accurately predicts carbon deposition risks through a vibration-pressure coupled carbon deposition triggering network. Based on the prediction results, it generates an adaptive gas injection strategy and performs global optimization, achieving precise and intelligent control of the titanium reduction process in iron concentrate. This effectively suppresses local carbon deposition problems, improves titanium separation efficiency, and enhances product quality.

[0130] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0131] It should be noted that all formulas in this manual are calculated by removing dimensions and taking their numerical values. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0132] Although embodiments of the invention have been shown and described, those skilled in the art will understand 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 claims and their equivalents.

Claims

1. A method for intelligent control of a multi-sensor fusion-based titanium reduction process for iron fines, characterized by, The method comprises the following steps: Step 1: obtaining the temperature distribution data, gas composition distribution data, furnace wall vibration signal and local pressure signal of the reaction furnace at each moment during the titanium reduction process of the iron concentrate; Step 2: constructing a furnace reaction thermodynamic constraint graph according to the spatial distribution characteristics of the temperature distribution data and the gas composition distribution data in the furnace; Step 3: dividing the furnace area into multiple reaction front subareas according to the furnace reaction thermodynamic constraint graph; Step 4: constructing a carbon deposition trigger network based on vibration-pressure coupling in each reaction front subarea to obtain the carbon deposition trigger probability of each reaction front subarea; Step 5: generating an adaptive gas injection strategy for each reaction front subarea according to the carbon deposition trigger probability of each reaction front subarea; Step 6: optimizing the gas injection parameters in the furnace according to the adaptive gas injection strategies of all reaction front subareas to realize real-time inhibition of local carbon deposition and effective separation of titanium during the titanium reduction process of the iron concentrate.

2. The process intelligent control method of multi-sensor fusion of iron ore fines and titanium reduction as claimed in claim 1, wherein, The construction of the furnace reaction thermodynamic constraint graph comprises: According to the temperature distribution data in the furnace, the boundary line between the high-temperature core area and the low-temperature edge area in the furnace is identified, and a temperature gradient boundary graph in the furnace is constructed; According to the gas composition distribution data in the furnace, the local extreme points of the CO / CO2 ratio in the furnace are identified, and a gas reaction activity graph in the furnace is constructed; The temperature gradient boundary graph and the gas reaction activity graph in the furnace are spatially superimposed to generate an initial thermodynamic constraint graph; In the initial thermodynamic constraint graph, the spatial conflict area between the boundary line in the temperature gradient boundary graph and the extreme point in the gas reaction activity graph is identified, and the spatial conflict area is corrected by thermodynamic equilibrium to obtain the furnace reaction thermodynamic constraint graph.

3. The process intelligent control method of multi-sensor fusion of iron ore fines and reduction of titanium content as claimed in claim 1, wherein, The division of the furnace area into multiple reaction front subareas comprises: In the furnace reaction thermodynamic constraint graph, a plurality of initial tracking points are randomly selected; For each initial tracking point, the propagation process of a virtual reaction front is simulated, wherein the propagation direction of the virtual reaction front is determined by the weighted vector of the temperature gradient and the CO / CO2 ratio gradient of the corresponding position point in the furnace reaction thermodynamic constraint graph, and the propagation speed of the virtual reaction front is determined by the thermodynamic entropy increase rate of the corresponding position point in the furnace reaction thermodynamic constraint graph; When the thermodynamic entropy increase rate is lower than a preset entropy increase rate threshold during the propagation process of the virtual reaction front, the propagation is stopped, and the area covered by the virtual reaction front is marked as an initial reaction front subarea; The boundary optimization is performed on all the initial reaction front subareas, wherein the boundary optimization is performed by identifying the continuity of the temperature gradient and the CO / CO2 ratio gradient between adjacent initial reaction front subareas, and the boundary with a continuity lower than a preset continuity threshold is smoothed to obtain the reaction front subarea.

4. The process intelligent control method of multi-sensor fusion of iron ore fines and reduction of titanium content as claimed in claim 1, wherein, The construction of the carbon deposition trigger network based on vibration-pressure coupling in each reaction front subarea to obtain the carbon deposition trigger probability of each reaction front subarea comprises: In each of the reaction front sub-zones, according to the time-domain waveform of the furnace wall vibration signal, a local mutation point of the vibration waveform is identified, and a vibration mutation point sequence is constructed; In each of the reaction front sub-zones, according to the time-domain waveform of the local pressure signal in the furnace, a local pulse point of the pressure waveform is identified, and a pressure pulse point sequence is constructed; According to the time synchronization of the vibration mutation point sequence and the pressure pulse point sequence, a vibration-pressure coupling network is constructed, wherein the nodes of the vibration-pressure coupling network are the vibration mutation points and the pressure pulse points, and the edges are the time delays between the vibration mutation points and the pressure pulse points; According to the connectivity of the vibration-pressure coupling network, the carbon deposition trigger probability of the reaction front sub-zone is calculated, wherein the connectivity is quantified by the average path length and the clustering coefficient of the vibration-pressure coupling network.

5. The process intelligent control method of multi-sensor fusion of iron ore fines and reduction of titanium content as claimed in claim 1, wherein, The generation of the adaptive gas injection strategy for each of the reaction front sub-zones includes: In each of the reaction front sub-zones, according to the reaction thermodynamic constraint graph in the furnace, a local flow field bottleneck region in the reaction front sub-zone is identified; In the local flow field bottleneck region, a disturbance-induced path is designed, wherein the starting point of the disturbance-induced path is the center of the local flow field bottleneck region, the ending point is the boundary of the reaction front sub-zone, and the path direction is determined by the thermodynamic entropy increase rate gradient of the corresponding position point in the reaction thermodynamic constraint graph in the furnace; According to the disturbance-induced path, an initial gas injection strategy is generated, wherein the initial gas injection strategy includes injection angle, injection flow rate and injection frequency; According to the prediction suppression effect of the initial gas injection strategy on the carbon deposition trigger probability of the reaction front sub-zone, the initial gas injection strategy is dynamically adjusted to obtain the adaptive gas injection strategy, wherein the prediction suppression effect is quantified by simulating the influence of the initial gas injection strategy on the thermodynamic entropy increase rate of the reaction thermodynamic constraint graph in the furnace.

6. The process intelligent control method of multi-sensor fusion of iron ore fines and reduction of titanium dioxide as claimed in claim 1, wherein, The optimization of the furnace gas injection parameters includes: According to the adaptive gas injection strategy of all the reaction front sub-zones, an initial global flow field distribution graph is constructed; In the initial global flow field distribution graph, a global flow field conflict region is identified; The global flow field conflict region is subjected to flow field reconstruction, wherein the flow field reconstruction minimizes the direction conflict of the flow field vector in the global flow field conflict region by iteratively adjusting the adaptive gas injection strategy of the adjacent reaction front sub-zones in the global flow field conflict region; According to the global flow field distribution graph after the flow field reconstruction, the furnace gas injection parameters are optimized, wherein the optimization is realized by minimizing the flow field turbulence intensity of the global flow field distribution graph and maximizing the uniformity of the thermodynamic entropy increase rate of the reaction thermodynamic constraint graph in the furnace.

7. The intelligent control method of process of iron ore fines de-titanium of multi-sensor fusion as claimed in claim 2, wherein, The thermodynamic equilibrium correction is realized by the following way: In the spatial conflict region, a thermodynamic entropy increase optimization model is constructed, wherein the objective function of the thermodynamic entropy increase optimization model is the total thermodynamic entropy increase of the spatial conflict region, and the constraint condition is the physical feasibility of the temperature value and the CO / CO2 ratio value in the spatial conflict region; An iterative algorithm based on gradient descent is used to solve the optimal solution of the thermodynamic entropy increase optimization model, wherein the step length of the iterative algorithm is determined by the modulus of the temperature gradient and the CO / CO2 ratio gradient in the spatial conflict area; According to the optimal solution, the temperature value and the CO / CO2 ratio value in the spatial conflict area are adjusted to obtain a corrected furnace reaction thermodynamic constraint graph.

8. The process intelligent control method of multi-sensor fusion of iron ore fines and reduction of titanium content as claimed in claim 4 wherein, The connectivity of the vibration-pressure coupling network is quantified by: Calculate the shortest path length between all node pairs in the vibration-pressure coupling network to obtain the average path length; For each node, calculate the edge connection density between the node and its neighbor nodes to obtain the local clustering coefficient of the node; and average the local clustering coefficients of all nodes to obtain the clustering coefficient; The connectivity of the vibration-pressure coupling network is quantified by the weighted sum of the average path length and the clustering coefficient.

9. The intelligent control method of process of iron ore fines de-titanium of multi-sensor fusion as claimed in claim 6, wherein, The flow field turbulence intensity and the thermodynamic entropy increase rate uniformity are quantified by: According to the flow field vector of each position point in the global flow field distribution map, the curl and divergence of the flow field vector are calculated to obtain the turbulence intensity of the position point; and the turbulence intensities of all position points are averaged to obtain the flow field turbulence intensity; According to the thermodynamic entropy increase rate of each position point in the furnace reaction thermodynamic constraint graph, the spatial distribution entropy of the thermodynamic entropy increase rate is calculated to obtain the thermodynamic entropy increase rate uniformity.

10. The multi-sensor fusion intelligent control system for iron concentrate titanium reduction process, which is used to realize the multi-sensor fusion intelligent control method for iron concentrate titanium reduction process according to any one of claims 1 to 9, characterized in that, It comprises: A data acquisition module for acquiring furnace temperature distribution data, furnace gas composition distribution data, furnace wall vibration signals, and local pressure signals in the reaction furnace at each moment during the titanium reduction process of iron concentrate; A thermodynamic constraint construction module for constructing a furnace reaction thermodynamic constraint graph based on the spatial distribution characteristics of the furnace temperature distribution data and the furnace gas composition distribution data; A reaction partitioning module for dividing the furnace area into multiple reaction front partitions according to the furnace reaction thermodynamic constraint graph; A carbon deposition analysis module for constructing a vibration-pressure coupling-based carbon deposition trigger network in each reaction front partition to obtain the carbon deposition trigger probability of each reaction front partition; A strategy generation module for generating an adaptive gas injection strategy for each reaction front partition based on the carbon deposition trigger probability of each reaction front partition; A parameter optimization module for optimizing the furnace gas injection parameters based on the adaptive gas injection strategies of all reaction front partitions to achieve real-time suppression of local carbon deposition and effective separation of titanium during the titanium reduction process of iron concentrate.

Images