Intelligent blast furnace condition trend early warning system and method

Abstract

The invention discloses an intelligent blast furnace condition trend early warning system and method. The system comprises a data acquisition module, a dynamic detection module, a quality association module and a feedback control module. The data acquisition module generates original feature signals of time-space correlation; the preprocessing module receives the original characteristic signals and filters noise interference through a Gaussian process regression algorithm; the prediction module generates associated characteristics of gas flow distribution and charge level form, and then captures a long-period trend component in the standardized data flow through a time convolutional network; the feedback module receives the furnace condition evolution probability signal and generates virtual blast furnace state data based on a gas-solid two-phase flow and heat transfer coupled physical field simulation model; and the early warning module generates a grading early warning instruction according to the degree of the parameter calibration signal crossing the preset threshold. According to the intelligent blast furnace condition trend early warning system and method, the problem that early warning lags due to the fact that a traditional blast furnace condition early warning method depends on a single parameter, a static model and artificial experience can be solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent monitoring and fault early warning of blast furnace ironmaking process, and in particular to an intelligent blast furnace condition trend early warning system and method. BACKGROUND

[0002] As a core equipment of steel production, blast furnace involves a strong coupling process of gas-solid heat transfer, chemical reaction and material movement inside, and the stability of the furnace condition directly affects the production efficiency and safety. Traditional monitoring methods mainly rely on single parameter threshold alarm such as temperature and pressure, but the blast furnace has strong sealing and complex working conditions, and single parameter is difficult to fully reflect the state inside the furnace. In recent years, although machine learning models (such as LSTM) have been introduced for trend prediction, due to the problems of single data dimension and static model, there are still two bottlenecks: first, at the data level, only traditional process parameters are collected, and multi-physical field information such as vibration, acoustics and thermal imaging is ignored, resulting in incomplete feature expression; second, at the model level, traditional statistical models or single deep learning architectures are difficult to simultaneously depict the spatial correlation and temporal dynamics among blast furnace parameters, and lack online self-calibration capability.

[0003] For example, a certain steel plant uses BP neural network to predict furnace temperature, but the false alarm rate is higher than 15% for a long time because the spatial coupling relationship between gas flow and material surface morphology is not considered. In addition, existing digital twin technology is mainly used for equipment simulation and has not been deeply combined with real-time early warning, so the model parameter update lags behind the actual working condition changes. With the intelligent transformation of the steel industry, multi-source heterogeneous data fusion and dynamic model optimization have become the direction of technological breakthrough, but existing solutions have not formed a systematic solution in terms of data processing (such as noise filtering and feature extraction), model architecture (such as joint modeling of space-time features), and virtual-real interaction (such as closed-loop optimization driven by twin), resulting in that the early warning system is difficult to meet the needs of long-period stable operation of the blast furnace. SUMMARY

[0004] In view of the above shortcomings of the prior art, the present application aims to provide an intelligent blast furnace condition trend early warning system and method to solve the problem of early warning lag caused by relying on single parameters, static models and manual experience in traditional blast furnace condition early warning methods. The present application collects multi-dimensional data through vibration, ultrasonic and infrared sensors, and after noise filtering and standardization processing, it uses graph convolution network to extract spatial correlation features among parameters, combines time convolution network to capture long-period trends, and dynamically weights key events through attention mechanism; the digital twin module compares virtual simulation data with actual working conditions in real time, generates parameter calibration signals to optimize the prediction model; the early warning module outputs operation suggestions according to the deviation degree of calibration signals and threshold values, forms a "perception-prediction-calibration-decision" closed loop, and finally solves the problem of poor adaptability to complex working conditions of traditional methods.

[0005] The application provides a smart blast furnace condition trend early warning system, comprising: A data acquisition module acquires real-time parameters of the blast furnace and generates original feature signals associated with time and space; A preprocessing module receives the original feature signals and filters out noise interference through a Gaussian process regression algorithm to form standardized data streams after denoising; A prediction module receives the standardized data streams, extracts spatial topological relationships among blast furnace parameters through a graph convolution network, generates associated features of coal gas flow distribution and material surface morphology, captures long-period trend components in the standardized data streams through a time convolution network, dynamically weights the influence weight of key events in combination with an attention mechanism, and outputs blast furnace condition evolution probability signals after fusing the spatial topological relationships and the long-period trend components; A feedback module receives the blast furnace condition evolution probability signals, generates virtual blast furnace state data based on a physical field simulation model of gas-solid two-phase flow and heat transfer coupling, and generates parameter calibration signals by differentiating and comparing the virtual blast furnace state data with the actual collected data; An early warning module generates graded early warning instructions according to the degree of the parameter calibration signals crossing a preset threshold.

[0006] In an embodiment of the application, the data acquisition module includes a vibration sensor array, which is arranged at a non-uniform density according to the stress gradient distribution of the blast furnace body, wherein the sensor spacing in the furnace throat area is smaller than that in the furnace waist area, and each sensor is configured with a dual-channel acquisition mode to capture vertical and horizontal vibration components respectively. After the material surface morphology distribution obtained by the vibration waveform and the ultrasonic scanning device is aligned by time stamp, multi-source data fusion is performed using a frequency domain convolution kernel to generate enhanced original feature signals containing material column collapse characteristic frequencies. The scanning period of the infrared thermal imaging device is dynamically adjusted according to the blast furnace smelting stage, and a continuous scanning mode is used in the tap hole area to generate a thermal radiation distribution time spectrum.

[0007] In an embodiment of the application, when the preprocessing module executes the Gaussian process regression algorithm, the noise filtering threshold is dynamically adjusted according to the working state of the blast furnace. When the fluctuation of the furnace top pressure exceeds the set range, an adaptive covariance function is automatically enabled. The kernel parameters of the covariance function are updated online based on the historical noise distribution characteristics, and after denoising, the standardized data streams are subjected to multi-scale decomposition to extract low-frequency components containing the material surface descent rate and high-frequency components reflecting the change of the coal gas flow rate. The standardized data streams with working condition adaptive characteristics are generated by time-frequency domain reorganization.

[0008] In an embodiment of the present application, the graph convolution network in the prediction module constructs a spatial topology graph based on the three-dimensional structure of the blast furnace, the nodes contain temperature, pressure, material level and gas composition parameters, the edge weights are dynamically calculated through the correlation strength of gas flow rate and material surface porosity, the time convolution network adopts an inflation causal convolution structure, the receptive field covers a time span of at least three smelting periods, when the attention mechanism detects that the weight distribution of key events exceeds a preset proportion, the edge weight reconstruction mechanism of the spatial topology graph is triggered, and the reconstructed topology relationship is spliced with the trend component to input the gate fusion unit.

[0009] In an embodiment of the present application, the physical field simulation model in the feedback module includes a coupled solver of gas-solid two-phase flow dynamics equation and radiation heat transfer equation, and after receiving the furnace condition evolution probability signal, it starts multi-thread parallel calculation, wherein the gas phase motion adopts discrete element method to simulate the coal gas flow field distribution, the solid phase material motion traces the ore falling trajectory through the improved DEM-CFD coupling algorithm, the difference comparison process includes establishing a multi-dimensional feature space mapping relationship between actual data and simulation data, generating a parameter calibration signal by calculating Mahalanobis distance, and encoding the calibration signal into a gradient update vector to inject into the network weight adjustment process of the prediction module.

[0010] In an embodiment of the present application, the early warning module includes a multi-level threshold dynamic generation mechanism, when the parameter calibration signal first touches the primary threshold, it starts the early warning prediction mode, estimates the risk evolution path by searching for similar signal patterns in the historical working condition database, when the signal continuously crosses the secondary threshold, it activates the emergency response protocol, which includes automatically retrieving equipment maintenance records and spare parts inventory information to generate a repair scheme, and superimposes and displays the deviation thermodynamic map of the current furnace condition parameters and the standard working condition through a three-dimensional visualization interface.

[0011] In an embodiment of the present application, it also includes a self-learning optimization module, which receives the false alarm records and operation feedback signals generated by the early warning module, constructs a loss function by comparing the deviation degree of the actual furnace condition evolution path and the prediction result, optimizes the hyperparameters of the graph convolution network and the time convolution network in the prediction module using a meta-learning framework, and synchronizes the optimized network structure parameters to the simulation calculation engine of the digital twin model, forming a closed-loop knowledge evolution link.

[0012] In an embodiment of the present application, a data quality evaluation unit is provided between the data acquisition module and the preprocessing module, which monitors the integrity and consistency of the original feature signals in real time, when it detects that the sensor fails or the data transmission is interrupted, it starts a missing data compensation mechanism based on a generative adversarial network, which generates a substitute signal using the joint distribution characteristics of multi-modal data under normal working conditions, and automatically performs residual correction of the compensation signal and the real signal after data recovery.

[0013] In one embodiment of the present application, the prediction module configures a model credibility evaluation mechanism, when the standard deviation of the furnace condition evolution probability signal exceeds the set range, automatically triggers a multi-model voting mechanism, the voting mechanism simultaneously runs three parallel prediction models containing a spatio-temporal convolution network, a deep belief network and a gated recurrent unit, selects the optimal prediction value by comparing the KL divergence of the output results of each model, and stores the divergent samples into a feature library for subsequent model training.

[0014] The present application also provides an intelligent blast furnace condition trend early warning method, comprising: S1: collecting real-time parameters of the blast furnace and generating original feature signals associated with space and time; S2: receiving the original feature signals and filtering out noise interference through a Gaussian process regression algorithm to form a standardized data stream after denoising; S3: receiving the standardized data stream, extracting the spatial topological relationship between the blast furnace parameters through a graph convolution network, generating the associated features of the gas flow distribution and the material surface morphology, capturing the long-period trend components in the standardized data stream through a time convolution network, and combining the attention mechanism to dynamically weight the influence weight of the key events, fusing the spatial topological relationship and the long-period trend components to output the furnace condition evolution probability signal; S4: receiving the furnace condition evolution probability signal and generating virtual blast furnace state data based on the physical field simulation model of gas-solid two-phase flow and heat transfer coupling, and differentiating and comparing the virtual blast furnace state data with the actual collected data to generate a parameter calibration signal; S5: generating a hierarchical early warning instruction according to the degree of the parameter calibration signal crossing the preset threshold.

[0015] The intelligent blast furnace condition trend early warning system and method provided by the present application collects multi-dimensional data through vibration, ultrasonic and infrared sensors, filters and standardizes the data, extracts the spatial correlation features between the parameters using a graph convolution network, captures the long-period trend by combining a time convolution network, and dynamically weights the key events through an attention mechanism; the digital twin module compares the virtual simulation data with the actual working condition in real time to generate a parameter calibration signal to optimize the prediction model; the early warning module outputs operation suggestions according to the deviation degree of the calibration signal from the threshold, forms a "perception-prediction-calibration-decision" closed loop, and finally solves the problem of poor adaptability to complex working conditions of traditional methods. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0017] Figure 1 A system architecture diagram of a smart blast furnace condition trend early warning system; Figure 2 A method flow diagram of a smart blast furnace condition trend early warning method. DETAILED DESCRIPTION

[0018] The advantages and effects of the present application can be easily understood by those skilled in the art from the content disclosed in the specification. The present application can also be implemented or applied by other different specific embodiments, and the details in the specification can be modified or changed based on different views and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0019] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and only the components related to the present application are shown in the diagrams, not the number, shape and size of the components when actually implemented. The actual implementation of each component may be a random change, and the component layout pattern may be more complex.

[0020] In the following description, a large number of details are discussed to provide a more thorough explanation of the embodiments of the present application, however, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details, and in other embodiments, the known structures and devices are shown in the form of block diagrams rather than in the form of details to avoid making the embodiments of the present application difficult to understand.

[0021] Please refer to Figures 1-2, shown as a kind of intelligent blast furnace condition trend early warning system and method of the application. The intelligent blast furnace condition trend early warning system of the application includes data acquisition module, preprocessing module, prediction module, feedback module and early warning module. The data acquisition module collects the real-time parameters of the blast furnace and generates original characteristic signals associated with time and space;The preprocessing module receives the original characteristic signals and filters out noise interference through Gaussian process regression algorithm to form standardized data stream after denoising;The prediction module receives the standardized data stream, extracts the spatial topological relationship between blast furnace parameters through graph convolution network, generates the correlation characteristics of gas flow distribution and material surface morphology, captures the long-period trend component in the standardized data stream through time convolution network, and combines the attention mechanism to dynamically weight the influence weight of key events, and outputs the blast furnace condition evolution probability signal after the fusion of spatial topological relationship and long-period trend component;The feedback module receives the blast furnace condition evolution probability signal and generates virtual blast furnace state data based on the physical field simulation model of gas-solid two-phase flow and heat transfer coupling, and generates parameter calibration signal by differentiating and comparing the virtual blast furnace state data with the actual collected data;The early warning module generates graded early warning instructions according to the degree of parameter calibration signal crossing the preset threshold.

[0022] As Figure 1As shown, the system consists of a data acquisition module, a preprocessing module, a prediction module, a feedback module, and an early warning module to form a complete technical chain. The data acquisition module, as the perception layer of the system, captures multi-dimensional physical signals in real time through the deployment of vibration sensor arrays, ultrasonic scanning devices, and infrared thermal imaging equipment at different positions of the furnace. The vibration sensor array adopts a non-uniform distribution strategy to monitor the stress changes in the furnace throat area. Its sampling frequency is dynamically adjusted according to the smelting stage to ensure the capture of characteristic vibration waveforms before the collapse of the charge column. The ultrasonic scanning device works in pulse reflection mode, transmits high-frequency sound waves, and receives echo signals to construct a three-dimensional distribution map of the charge surface shape and record the time sequence changes of the charge surface descent rate. The infrared thermal imaging device uses a rotating scanning mechanism to periodically measure the thermal radiation distribution of the furnace surface, especially in the tap hole area using a continuous scanning mode to generate temperature field data with spatial continuity. These heterogeneous sensor-collected raw signals are aligned through a timestamp synchronization mechanism to form spatio-temporally correlated raw feature signals, providing a multi-modal data basis for subsequent analysis. The preprocessing module undertakes the core functions of data cleaning and feature enhancement. After receiving the raw feature signals from the data acquisition module, it first performs signal integrity checking to detect sensor failure or data transmission abnormalities. For normal data streams, a Gaussian process regression algorithm is used for noise filtering. The covariance function parameters of this algorithm are dynamically optimized according to the current working state of the blast furnace. For example, when the furnace top pressure fluctuates dramatically, it automatically switches to the Matern kernel function with stronger robustness and adjusts the length scale parameter based on the historical noise statistical characteristics within the sliding window. After denoising, the data stream is processed through multi-scale decomposition, using wavelet transform to decompose the signal into low-frequency components reflecting the macro motion trend of the charge surface and high-frequency components representing the transient fluctuations of the gas flow rate. After normalization of each sub-band signal, the standardized data stream is generated through time-frequency domain reconstruction techniques. This data stream not only retains the physical meaning of the original signal but also enhances the expression dimension of the working condition characteristics.

[0023] Further, the prediction module serves as the intelligent analysis hub of the system, and its core is to build a hybrid space-time prediction model to achieve accurate prediction of the evolution trend of the furnace condition. After receiving the standardized data stream output by the preprocessing module, the module simultaneously starts two calculation paths of spatial feature extraction and time trend analysis. In the spatial dimension, the graph convolution network constructs a parameter correlation topological graph based on the three-dimensional structure of the blast furnace, taking temperature, pressure, material surface height, and gas composition as graph nodes, and the edge weight between nodes is calculated in real time according to the dynamic correlation of gas flow rate and material surface porosity. Through multi-layer graph convolution operation, the network learns the deep correlation features of gas flow distribution and material surface morphology. In the time dimension, the time convolution network adopts an inflation causal convolution structure, with an exponentially increasing dilation factor, forming a receptive field covering multiple smelting cycles, effectively capturing the slow-changing rules in long-term trends. At the same time, the attention mechanism dynamically allocates weights to key events (such as charging period and tapping operation) in the input sequence, enhancing their influence on the prediction results. The spatial topological features and time trend components are tensor-spliced and nonlinearly transformed through a gating fusion unit, and finally output a probability signal representing the future evolution direction of the furnace condition. This signal not only contains the possibility evaluation of abnormality, but also indicates the time window of risk evolution. The specific implementation of the data acquisition module is technically deepened and innovatively designed. The layout scheme of the vibration sensor array adopts a non-uniform density configuration guided by stress gradient, with 6-8 sensor nodes per square meter in the throat area, and 3-4 in the waist area. This differentiated layout effectively matches the stress distribution characteristics generated by the movement of materials in the furnace. Each vibration sensor node integrates a dual-channel acquisition circuit, which configures vertical and horizontal piezoelectric sensing elements, respectively, and synchronously acquires three-dimensional vibration components through orthogonally arranged accelerometers. The raw vibration waveform collected is subjected to anti-aliasing filtering, and is spatio-temporally aligned with the material surface morphology data obtained by the ultrasonic scanning device, with microsecond-level timestamp synchronization achieved through a GPS synchronous clock module. In the data fusion stage, the frequency domain convolution kernel technology is used to perform convolution operation on the frequency spectrum features of the vibration signal and the ultrasonic material surface data in the frequency domain, extracting specific frequency band energy features representing the risk of material column collapse. The infrared thermal imaging device uses dual-band imaging technology to simultaneously capture mid-infrared and long-infrared spectral data, and eliminates the interference of the oxide layer on the surface of the furnace body on the thermal radiation measurement through a multispectral fusion algorithm, generating a high-precision temperature distribution map. In the tapping hole area, the system starts a high-speed scanning mode, increasing the sampling frequency to 5 times that of the regular mode, and records the transient temperature field changes caused by the flow of molten iron, which provides key basis for judging the erosion state of the furnace.

[0024] Specifically, the dynamic optimization mechanism of the preprocessing module and the feature enhancement method. The module is built-in working condition perception unit, real-time monitoring of blast furnace smelting stage and working state, when the detection of furnace top pressure fluctuation exceeds the set threshold, automatically trigger noise filtering parameter adjustment mechanism. The covariance function of Gaussian process regression algorithm is switched to composite kernel structure in this state, which is composed of radial basis function kernel and periodic kernel linear combination, which can capture the transient characteristics of sudden noise and filter out periodic interference. The hyperparameters of the kernel function are continuously optimized through the online learning mechanism, using the noise samples in the sliding window to calculate the marginal likelihood function, and using the gradient ascent method to iteratively update the length scale and signal variance parameters. The denoised data stream enters the multi-scale analysis stage, and the signal is adaptively decomposed into multiple intrinsic mode functions through empirical mode decomposition. Select the low-frequency IMF component reflecting the macro trend of the material surface movement and the high-frequency IMF component representing the gas flow velocity fluctuation for processing. The low-frequency component is smoothed to eliminate random fluctuations, while the high-frequency component is subjected to envelope analysis to extract its modulation features. In the recombination process, time-frequency matrix fusion technology is introduced to weight and superimpose the time-frequency spectrum of each sub-band signal to generate an enhanced standardized data stream with both time and frequency resolution. This data stream particularly emphasizes the following features: the first derivative feature of the material surface descending rate, the second moment statistical quantity of the gas flow velocity change, and the spatial gradient feature of the temperature field distribution, providing a high information density input feature set for the subsequent prediction model.

[0025] In an embodiment of the present application, the physical field simulation model and the parameter calibration mechanism work together. The feedback module converts the furnace condition evolution probability signal output by the prediction module into real-time state data of the virtual blast furnace by constructing a high-precision digital twin model. The physical field simulation model uses a multi-physical field coupling solution technology to integrate the gas-solid two-phase flow dynamics equation and the radiation heat transfer equation. The gas phase motion simulation is based on the discrete element method, which decomposes the coal gas flow into a discrete particle group, and constructs a three-dimensional flow field distribution by calculating the collision and momentum transfer between particles. The solid phase material motion uses an improved DEM-CFD coupling algorithm, which introduces a visco-plastic contact model into the traditional discrete element framework to accurately simulate the deformation characteristics caused by softening at high temperature during the ore falling process. When the simulation calculation is started, the module automatically allocates computing resources, opens independent threads for parallel solving of gas phase and solid phase motion respectively, and realizes real-time data interaction between the two phases through shared memory technology. After the virtual blast furnace state data is generated, the system establishes a multi-dimensional feature space mapping relationship between the actual collected data and the simulation data, selects the temperature gradient distribution, coal gas flow velocity peak value and material surface settlement rate as the key comparison dimensions, and evaluates the deviation degree by calculating the Mahalanobis distance of the two groups of data in the feature space. After the distance value is normalized, it is compared with the preset dynamic tolerance threshold to generate a parameter calibration signal. In the calibration signal coding process, the system uses the gradient projection method to decompose the multi-dimensional deviation vector into the weight update direction of the graph convolution network and the time convolution network in the prediction module, and injects the network parameter adjustment process through the momentum optimization algorithm with a time decay factor to ensure that the model can quickly respond to the current deviation during the calibration process, and avoid overfitting caused by short-term fluctuations.

[0026] As Figure 1As shown, the hierarchical decision mechanism of the early warning module is deeply technically expanded with the visual auxiliary function. The early warning module is built-in with a multi-level threshold dynamic generation engine, which automatically adjusts the threshold curve according to the blast furnace smelting cycle, raw material ratio change and equipment running state. When the parameter calibration signal first touches the primary threshold, the system starts the early warning prediction mode, retrieves the historical records of similar working conditions in the past year through the graph database, extracts multi-dimensional features including abnormal types, evolution speed and disposal effect, aligns the time series after using the dynamic time warping algorithm, constructs a risk path prediction model based on random forest, and outputs the probability distribution of furnace condition deterioration in the next three hours. If the signal continuously crosses the secondary threshold, the module activates the emergency response protocol, first calls the maintenance log and sensor health state data in the device management system, combines the spare parts inventory information to generate a repair scheme list containing priority, and at the same time starts the three-dimensional visualization engine to map and compare the current furnace condition parameters with the standard working condition benchmark value in space, superimposes the temperature deviation thermal map, stress abnormal area and coal gas flow velocity exceeding position on the surface of the digital twin model, and distinguishes the risk level through color gradient and flicker frequency. The visualization interface also integrates operation suggestion deduction function, when the operator selects a certain disposal measure, the system simulates the furnace condition evolution trend in the next hour after the implementation of the measure in real time, displays the prediction of the change of key parameters in the form of dynamic curve, and assists the decision maker to evaluate the expected effect of different schemes. In addition, the module is built-in with a feedback learning mechanism, which compares and analyzes the results of each early warning response and the actual furnace condition change, automatically corrects the weight coefficients in the threshold generation algorithm, and forms a continuously optimized decision support system.

[0027] Further, a self-learning optimization module is constructed to form a closed-loop learning link from early warning feedback to model evolution. The module continuously receives false alarm cases recorded by the early warning module, effective alarms confirmed by the operating personnel, and subsequent disposal effect data, describes the environmental context of each early warning event through the construction of a multi-dimensional feature vector, including the fluctuation range of raw material components, the percentage value of device running time, and environmental temperature and humidity conditions, and other auxiliary parameters. To address the prediction bias problem, the module designs a composite loss function to decompose the prediction error of the furnace condition evolution probability signal into three components: spatial feature error, time trend error, and correlation weight error. The cosine similarity loss between the coal gas flow-material surface correlation features output by the graph convolution network and the actual observation value, the mean square error loss between the prediction trend of the time convolution network and the true value, and the KL divergence loss between the attention mechanism weight distribution and the expert-labeled key events are calculated. Based on this loss function, a meta-learning framework is used to optimize the prediction model in two layers: in the inner loop, the network weights are updated using the current batch of data; in the outer loop, the network hyperparameters are optimized through cross-task knowledge transfer, including the neighborhood aggregation radius of the graph convolution layer, the inflation coefficient sequence of the time convolution network, and the temperature coefficient of the attention mechanism. The optimized network parameters are synchronized to the simulation calculation engine of the digital twin model. Specifically, the spatial correlation rules learned by the graph convolution network are converted into the coupling coefficients of the coal gas flow and the material surface movement in the physical field simulation model through a parameter mapping interface, and the long-period trend features captured by the time convolution network are embedded into the initial condition generation algorithm, realizing the knowledge co-evolution of the prediction model and the simulation model. This process forms a positive feedback loop, enabling the system to continuously improve its adaptability to complex working conditions during continuous operation.

[0028] As shown in Figure 2 , it is an intelligent blast furnace condition trend early warning method. S1: Collecting real-time parameters of the blast furnace and generating original feature signals with space-time correlation; S2: Receiving the original feature signals and filtering out noise interference through Gaussian process regression algorithm to form standardized data stream after denoising; S3: Receiving the standardized data stream, extracting the spatial topological relationship between blast furnace parameters through a graph convolution network to generate correlation features of coal gas flow distribution and material surface morphology, capturing long-period trend components in the standardized data stream through a time convolution network, and dynamically weighting the influence weight of key events through an attention mechanism, and outputting the furnace condition evolution probability signal after fusing the spatial topological relationship and the long-period trend component; S4: Receiving the furnace condition evolution probability signal and generating virtual blast furnace state data based on the physical field simulation model of gas-solid two-phase flow and heat transfer coupling, and comparing the virtual blast furnace state data with the actual collected data to generate a parameter calibration signal; S5: Generating a hierarchical early warning instruction according to the degree of crossing the preset threshold of the parameter calibration signal.

[0029] Further, a data quality assurance system is added between the data acquisition module and the preprocessing module. The unit monitors the integrity of the original feature signal in real time, including sensor sampling rate stability, data packet time sequence continuity and cross-modal data alignment accuracy. When a specific sensor node failure or data transmission delay is detected to exceed the tolerance threshold, a data compensation mechanism based on a deep generative adversarial network is immediately started. The generator network uses a U-Net architecture, inputs the joint distribution features of multi-sensor data under normal working conditions, learns the implicit correlation between vibration waveforms, material surface morphology and thermal radiation distribution through spatial attention gate mechanism, and outputs synthesized signals that meet the context features of the current working condition. The discriminator network builds a multi-scale feature pyramid to identify real signals and generated signals from three dimensions of time domain waveform, frequency domain energy distribution and spatio-temporal correlation. During the adversarial training process, a curriculum learning strategy is introduced to gradually increase the proportion of missing data to improve the robustness of the generator. When the data transmission is restored, the system automatically performs residual correction of the compensation signal and the real signal: first, calculate the mean square error of the two in the sliding time window, if the error exceeds the allowed range, trigger the Kalman filter algorithm to dynamically correct the subsequent generated compensation signal; At the same time, the compensation data of the abnormal period is marked as a low-confidence sample for subsequent model retraining and weighted processing. In addition, the unit has a built-in sensor health assessment model that can predict sensor performance degradation risks in advance by analyzing the signal-to-noise ratio, zero drift and frequency spectrum characteristics of each node signal, and generate maintenance warning prompts to the device management system.

[0030] The intelligent blast furnace condition trend early warning system and method of the present application collects multi-dimensional data through vibration, ultrasonic and infrared sensors, filters and standardizes the data, extracts spatial correlation features between parameters using a graph convolution network, captures long-period trends using a time convolution network, and dynamically weights key events using an attention mechanism. The digital twin module compares virtual simulation data with actual working conditions in real time to generate parameter calibration signals to optimize the prediction model. The early warning module outputs operation suggestions according to the deviation of the calibration signal from the threshold value, forms a "perception-prediction-calibration-decision" closed loop, and finally solves the problem of poor adaptability of traditional methods to complex working conditions.

[0031] Therefore, the intelligent blast furnace condition trend early warning system and method of the present application solves the problem of early warning lag caused by relying on single parameters, static models and manual experience in traditional blast furnace condition early warning methods.

[0032] The above embodiments are only illustrative of the principles of the present application and its efficacy, and are not intended to limit the present application. Any modification or change made by any person skilled in the art without departing from the spirit and scope of the present application shall be covered by the claims of the present application.

Claims

1. An intelligent blast furnace condition trend early warning system, characterized in that, include: The data acquisition module collects real-time parameters of the blast furnace and generates spatiotemporally correlated raw feature signals. The preprocessing module receives the original feature signal and filters out noise interference through a Gaussian process regression algorithm to form a denoised standardized data stream. The prediction module receives the standardized data stream, extracts the spatial topological relationship between blast furnace parameters through a graph convolutional network, generates the correlation features between gas flow distribution and material surface morphology, captures the long-period trend component in the standardized data stream through a temporal convolutional network, and dynamically weights the influence weight of key events by combining an attention mechanism. After fusing the spatial topological relationship with the long-period trend component, the module outputs a probability signal of furnace condition evolution. The feedback module receives the furnace condition evolution probability signal and generates virtual blast furnace state data based on the physical field simulation model of gas-solid two-phase flow and heat transfer coupling. The feedback module then performs differential comparison between the virtual blast furnace state data and the actual collected data to generate a parameter calibration signal. The early warning module generates graded early warning instructions based on the degree to which the parameter calibration signal crosses a preset threshold.

2. The intelligent blast furnace condition trend early warning system according to claim 1, characterized in that, The data acquisition module includes a vibration sensor array, which is arranged with a non-uniform density according to the stress gradient distribution of the blast furnace body. The sensor spacing in the throat area is smaller than that in the waist area, and each sensor is configured with a dual-channel acquisition mode to capture the vibration components in the vertical and horizontal directions respectively. The vibration waveform and the material surface morphology distribution obtained by the ultrasonic scanning device are aligned with the timestamp, and multi-source data are fused using a frequency domain convolution kernel to generate an enhanced original feature signal containing the characteristic frequency of the material column collapse. The scanning cycle of the infrared thermal imaging device is dynamically adjusted according to the blast furnace smelting stage, and a continuous scanning mode is used in the tapping area to generate a time series map of thermal radiation distribution.

3. The intelligent blast furnace condition trend early warning system according to claim 1, characterized in that, When executing the Gaussian process regression algorithm, the preprocessing module dynamically adjusts the noise filtering threshold according to the blast furnace operating status. When the top pressure fluctuation exceeds the set range, the adaptive covariance function is automatically activated. The kernel parameters of the covariance function are updated online based on historical noise distribution characteristics. After completing the denoising process, the standardized data stream is decomposed into multiple scales to extract the low-frequency component containing the material surface descent rate and the high-frequency component reflecting the gas flow rate change. The standardized data stream with operating condition adaptive characteristics is generated through time-frequency domain reconstruction.

4. The intelligent blast furnace condition trend early warning system according to claim 1, characterized in that, The graph convolutional network in the prediction module constructs a spatial topology graph based on the three-dimensional structure of the blast furnace. The nodes include parameters such as temperature, pressure, material level, and gas composition. The edge weights are dynamically calculated based on the correlation strength between the gas flow rate and the porosity of the material holes. The temporal convolutional network adopts a dilated causal convolutional structure, and its receptive field covers a time span of at least three smelting cycles. When the attention mechanism is detected to assign more weights to key events than a preset ratio, the edge weight reconstruction mechanism of the spatial topology graph is triggered. The reconstructed topological relationships and trend components are then tensor-concatenated and input into the gating fusion unit.

5. The intelligent blast furnace condition trend early warning system according to claim 1, characterized in that, The physical field simulation model in the feedback module includes a coupled solver for the gas-solid two-phase flow dynamics equation and the radiation heat transfer equation. After receiving the furnace condition evolution probability signal, it starts multi-threaded parallel calculation. The gas phase motion uses the discrete element method to simulate the gas flow field distribution, and the solid phase material motion tracks the ore falling trajectory through an improved DEM-CFD coupling algorithm. The differential comparison process includes establishing a multi-dimensional feature space mapping relationship between actual data and simulation data, generating parameter calibration signals by calculating Mahalanobis distance, and encoding the calibration signals as gradient update vectors into the network weight adjustment process of the prediction module.

6. The intelligent blast furnace condition trend early warning system according to claim 1, characterized in that, The early warning module includes a multi-level threshold dynamic generation mechanism. When the parameter calibration signal first touches the primary threshold, the early warning prediction mode is activated. The risk evolution path is estimated by searching the historical operating condition database for similar signal patterns. When the signal continuously crosses the secondary threshold, the emergency response protocol is activated. The protocol includes automatically retrieving equipment maintenance records and spare parts inventory information to generate a maintenance plan, and displaying the deviation thermal map between the current furnace condition parameters and the standard operating condition through a three-dimensional visualization interface.

7. The intelligent blast furnace condition trend early warning system according to claim 1, characterized in that, It also includes a self-learning optimization module, which receives false alarm records and operation feedback signals generated by the early warning module, constructs a loss function by comparing the deviation between the actual furnace condition evolution path and the prediction result, uses a meta-learning framework to jointly optimize the hyperparameters of the graph convolutional network and the temporal convolutional network in the prediction module, and synchronizes the optimized network structure parameters to the simulation calculation engine of the digital twin model to form a closed-loop knowledge evolution link.

8. The intelligent blast furnace condition trend early warning system according to claim 1, characterized in that, A data quality assessment unit is provided between the data acquisition module and the preprocessing module. The assessment unit monitors the integrity and consistency of the original feature signals in real time. When sensor failure or data transmission interruption is detected, a missing data compensation mechanism based on generative adversarial networks is activated. The compensation mechanism uses the joint distribution characteristics of multimodal data under normal operating conditions to generate a substitute signal, and automatically performs residual correction between the compensation signal and the real signal after data recovery.

9. The intelligent blast furnace condition trend early warning system according to claim 1, characterized in that, The prediction module is configured with a model credibility evaluation mechanism. When the standard deviation of the furnace condition evolution probability signal exceeds the set range, a multi-model voting mechanism is automatically triggered. The voting mechanism simultaneously runs three parallel prediction models, including a spatiotemporal convolutional network, a deep belief network, and a gated recursive unit. The optimal prediction value is selected by comparing the KL divergence of the output results of each model, and the divergence samples are stored in the feature library for subsequent model training.

10. A method for early warning of intelligent blast furnace condition trends using any one of claims 1-9, characterized in that, include: S1: Collect real-time parameters of the blast furnace and generate spatiotemporally correlated original feature signals; S2: Receive the original feature signal and filter out noise interference through Gaussian process regression algorithm to form a denoised standardized data stream; S3: Receive the standardized data stream, extract the spatial topological relationship between blast furnace parameters through graph convolutional network, generate the correlation features between gas flow distribution and material surface morphology, capture the long-period trend component in the standardized data stream through temporal convolutional network, and dynamically weight the influence weight of key events by combining the attention mechanism, and output the furnace condition evolution probability signal after fusing the spatial topological relationship with the long-period trend component. S4: Receive the furnace condition evolution probability signal and generate virtual blast furnace state data based on the physical field simulation model of gas-solid two-phase flow and heat transfer coupling, and perform differential comparison between the virtual blast furnace state data and the actual collected data to generate parameter calibration signal. S5: Generate a graded early warning command based on the degree to which the calibration signal crosses a preset threshold according to the parameters.

Images