Sintering endpoint prediction and regulation method based on characteristics of mixture and post-ignition state

Abstract

The application discloses a sintering endpoint advanced prediction and intelligent regulation method based on mixture characteristics and ignition initial state. The method belongs to the technical field of steel metallurgy automatic control. The technical scheme firstly collects physical and chemical characteristic data of the mixture and initial state parameters of the mixture surface ignition, and constructs a multi-dimensional feature data set; then, a sintering endpoint advanced prediction model is trained and established by using historical production data, so that the sintering endpoint position is predicted in advance; when the prediction result deviates from the set target value, the optimal control parameter adjustment amount is calculated by an intelligent algorithm, and instructions are automatically issued to adjust the sintering machine speed and the main exhaust speed. By introducing the mixture characteristics and the mixture surface ignition state as feedforward variables, the sintering endpoint advanced prediction and closed-loop control are realized, the problem of large hysteresis and poor stability of the traditional control method is effectively solved, and the yield and energy utilization rate of the sintered ore are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of sintering process control technology in iron and steel metallurgy, specifically to a method for predicting and controlling the sintering endpoint based on the characteristics of the mixture and the state after ignition, and a method for intelligently controlling the speed of the main exhaust fan and the running speed of the sintering machine accordingly. Background Technology

[0002] Sintering is a crucial step in steelmaking, with the core objective of sintering a mixture of iron ore powder, fuel, and flux into a stable sinter. The final sintering temperature (BTP) is a key control indicator in the sintering process, directly determining the yield, quality, and energy consumption of the sinter. If the BTP is too high, the effective area utilization of the sintering machine is low, resulting in reduced output and shortened lifespan of the sintering machine trolley and grate bars. Conversely, if the BTP is too low, the mixture is not fully burned, leading to a higher return rate, shortened lifespan of the annular cooler trolley, and potentially causing fires due to the "red ore" being pushed into the furnace.

[0003] Existing methods for predicting the sintering endpoint temperature suffer from the following drawbacks: First, the prediction dimension is singular, relying heavily on single process parameters such as exhaust gas temperature, failing to fully consider the fundamental influence of the mixture's basic characteristics (such as chemical composition and particle size distribution) on the sintering reaction. This results in poor predictive adaptability, with a significant decrease in prediction accuracy when the batching scheme is adjusted. Second, the prediction exhibits significant lag. Existing technologies mostly estimate the endpoint by observing the trend of the sintering inflection point (BRP) during the sintering process, lacking real-time capture of the initial state after ignition, making it difficult to achieve "advanced prediction." Third, the control strategies are isolated, often adjusting the sintering machine speed or the main exhaust fan volume individually, without forming a coordinated control mechanism. This fails to address the strongly coupled characteristics of the sintering process, resulting in insufficient control stability. Therefore, there is an urgent need to develop a multi-dimensional, advanced, and coordinated sintering endpoint control method to address the pain points of existing technologies.

[0004] The purpose of this invention is to overcome the defects of the prior art and provide a method for predicting and intelligently controlling the sintering endpoint temperature based on the basic characteristics of the mixture and the initial state after ignition. This method achieves accurate prediction of the sintering endpoint temperature and ensures that the sintering endpoint is stable within the optimal range through coordinated intelligent control of the main exhaust fan speed and the sintering machine operating speed, thereby improving the quality and yield of sintered ore and reducing energy consumption. Summary of the Invention

[0005] The purpose of this invention is to address the above-mentioned problems by providing a method for predicting and controlling the sintering endpoint based on the characteristics of the mixture and its state after ignition.

[0006] The objective of this invention is achieved as follows: a method for predicting and controlling the sintering endpoint based on the characteristics of the mixture and its post-ignition state, comprising the following steps: Step 1: Constructing a database of basic characteristic parameters of the mixture: collecting basic characteristic parameters of the mixture to be sintered on the sintering machine trolley and constructing a parameter database that is updated in real time; Step 2: Real-time monitoring of initial state parameters after ignition: setting up a multi-dimensional monitoring module at the downstream of the ignition furnace outlet of the sintering machine, i.e., the initial area where the mixture has not entered the stable combustion stage after ignition, and collecting the following initial state parameters in real time: Temperature parameters: the surface temperature distribution of the mixture after ignition is detected by an infrared thermal imager; Physical state parameters: the initial thickness of the sintering material layer corresponding to each auxiliary door on the surface of the mixture is detected by an infrared rangefinder; real-time preprocessing of the monitoring data, removing outliers, and calculating characteristic indicators to obtain the initial state feature vector Y=[y1,y2,...,y m ], where m is the dimension of the initial state parameters; Step 3: Training and prediction of the sintering endpoint temperature prediction model: 3.1 Constructing the prediction model: Using the basic characteristic feature vector X of the mixture and the initial state feature vector Y after ignition as input, and the sintering endpoint position, i.e., the sintering length, as output, construct the sintering endpoint temperature prediction model; The model structure includes: a feature fusion layer to perform spatiotemporal feature fusion of X and Y, a long-term feature extraction layer, a short-term dynamic capture layer, and a regression prediction layer to output the predicted value of the sintering endpoint temperature P_BTP; 3.2 Model training: Collect historical production data under different working conditions covering different The batching scheme and ignition parameter combination are used to construct a training dataset and iteratively train the prediction model to make the model prediction error ≤ ±0.5m position; 3.3 Advance prediction: The standardized feature vector X from step one and the initial state feature vector Y from step two are input into the trained prediction model in real time, and the predicted value of the sintering endpoint temperature P_BTP is output. The prediction advance is 75%-80% of the total sintering process time; Step four: Intelligent coordinated control of main exhaust fan speed and sintering machine speed: 4.1 Setting the optimal range of sintering endpoint: According to the preset optimal sintering endpoint position range [P] of the sintering machine equipment. min ,P max 4.2 Deviation Calculation: Calculate the predicted sintering endpoint value P_BTP and the midpoint of the optimal interval P. mid The deviation ΔP = P_BTP - P mid4.3 Regulation Strategy Formulation: Based on the deviation ΔP, a fuzzy PID control algorithm is used to formulate a coordinated regulation strategy for the main exhaust fan speed n and the sintering machine operating speed v: When ΔP > 0.5 and the predicted endpoint is later: increase the main exhaust fan speed Δn = k1 × ΔP, where k1 is the speed regulation coefficient, with a value of 1.0-1.2 r / (min·m), and simultaneously decrease the sintering machine speed Δv = k2 × ΔP, where k2 is the speed regulation coefficient, with a value of 0.02-0.03 m / (min·m) to accelerate the heat transfer and combustion process of the mixture; When ΔP < 0.5 and the predicted endpoint is earlier: decrease the main exhaust fan speed Δn = k3 × |ΔP|, where k3 is the speed regulation coefficient, and k2 is the speed regulation coefficient, and ... The rotational speed control coefficient is 0.8-1.0 r / (min·m), while simultaneously increasing the sintering machine speed Δv=k4×|ΔP|, where k4 is the speed control coefficient, with a value of 0.01-0.02 m / (min·m), to slow down the heat transfer and combustion process of the mixture; when |ΔP|≤0.5, the predicted endpoint is within the optimal range: maintain the current main exhaust fan speed and sintering machine speed, and achieve stable control by fine-tuning the PID parameters; 4.4 Dynamic feedback correction: collect the exhaust gas temperature and negative pressure data of each air box during the sintering process in real time, calculate the actual deviation of the sintering endpoint, and dynamically correct the control parameters to ensure that the sintering endpoint is finally stable within [P]. min ,P max Within the range.

[0007] The basic characteristic parameters mentioned in step one include: physical characteristic parameters: particle size distribution of the mixture: percentage of particles with diameters of 0-1mm, 1-3mm, 3-5mm, and >5mm; moisture content; and granulation characteristic parameters: average particle size of the mixture after granulation. The above parameters are normalized to obtain a standardized feature vector X=[x1,x2,...,x...]. n ], where n is the dimension of the basic characteristic parameter, x n ∈[0,1].

[0008] The beneficial effects of this invention are: accurate and forward-looking prediction: by integrating the basic characteristics of the mixture (essential influencing factors) and the initial state of ignition (immediate influencing factors), a two-dimensional input prediction model is constructed, which not only improves the predictive adaptability under different working conditions, but also the prediction advance reaches 75%-80% of the total sintering time, leaving sufficient time for regulation.

[0009] Coordination and intelligence of control: A coordinated control strategy of main exhaust fan speed and sintering machine speed is adopted. In view of the strong coupling characteristics of the sintering process, the parameters are dynamically optimized through fuzzy PID algorithm. Compared with single parameter control, the control stability is improved by more than 40%.

[0010] Significant economic and technical benefits: It can increase the sintering endpoint hit rate to over 95%, reduce the sinter return rate by 2%-3%, reduce energy consumption per ton of ore by 1%-3%, and extend the service life of sintering machines, thus having significant industrial application value. Attached Figure Description

[0011] The present invention will now be further described with reference to the accompanying drawings.

[0012] Figure 1 This is a flowchart of the present invention.

[0013] 1. Belt conveyor for sintering, 2. Thermometer, 3. Particle size analyzer, 4. Server, 5. Shuttle-type fabric conveyor, 6. Mixing silo, 7. Circular roller feeder, 8. Ignition furnace, 9. Sintering machine.

[0014] The basic characteristic parameters of the mixture to be sintered on the sintering machine trolley and the initial state parameters of the ignition of the material surface in the sintering machine are collected on the sintering-1 belt conveyor and aggregated on the same server to establish a model, make predictions, and regulate, so as to realize closed-loop control. Detailed Implementation

[0015] Currently, sintering control prediction has a single dimension, relying heavily on single process parameters such as exhaust gas temperature, resulting in poor predictive adaptability. Secondly, the prediction lag is obvious, and the endpoint is often estimated by the trend of the sintering inflection point (BRP) during the sintering process, making it difficult to achieve "advanced prediction". Thirdly, the control strategy is isolated, often adjusting the sintering machine speed or the main exhaust fan air volume separately, resulting in insufficient control stability.

[0016] The method of predicting and intelligently controlling the sintering endpoint temperature in advance based on the basic characteristics of the mixture and the initial state after ignition enables accurate prediction of the sintering endpoint temperature. Through the coordinated intelligent control of the main exhaust fan speed and the sintering machine running speed, the sintering endpoint is ensured to remain stable within the optimal range.

[0017] This invention can increase the sintering endpoint hit rate to over 95%, reduce the sinter return rate by 2%-3%, reduce energy consumption per ton of ore by 1%-3%, and extend the service life of the sintering machine.

[0018] A multi-dimensional data fusion method for mixture characteristics and initial state after ignition is proposed, which involves acquiring the particle size distribution, moisture content, and initial state of the mixture as initial input parameters. The aforementioned multi-source heterogeneous data are preprocessed to construct feature vectors for prediction.

[0019] An advanced prediction model for the sintering endpoint (BTP) is constructed, based on a mathematical model using machine learning or deep learning, to describe the influence of mixture characteristics and initial ignition state on the location of the sintering endpoint. This model can predict the location and arrival time of the sintering endpoint in advance based on the input parameters at the current moment, rather than relying solely on real-time monitoring data.

[0020] A dynamic intelligent control strategy is employed to set a target control range for the sintering endpoint. When the predicted sintering endpoint deviates from the target range, the system automatically calculates and outputs control commands, including at least one adjustment to the sintering machine speed or the main pumping speed.

[0021] The feedback closed-loop control mechanism includes a method that feeds back the sintering endpoint position measured in actual production to the prediction model for online correction, thereby achieving adaptive updating of model parameters.

[0022] A method for predicting and intelligently controlling the sintering endpoint temperature based on the basic characteristics of the mixture and its initial state after ignition includes the following steps: Step 1: Constructing a database of basic characteristic parameters of the mixture: Collecting the basic characteristic parameters of the mixture to be sintered on the sintering machine trolley and constructing a parameter database that is updated in real time. The basic characteristic parameters include: physical characteristic parameters: particle size distribution of the mixture (percentage of particles with diameters of 0-1mm, 1-3mm, 3-5mm, and >5mm), moisture content; granulation characteristic parameters: average particle size of the mixture after granulation. The above parameters are normalized to obtain a standardized feature vector X=[x1,x2,...,x...]. n ], where n is the dimension of the basic characteristic parameter, x n ∈[0,1].

[0023] Step 2: Real-time monitoring of initial state parameters after ignition: At point 1 downstream of the sintering machine ignition furnace outlet (the initial area where the mixture has not yet entered the stable combustion stage after ignition), a multi-dimensional monitoring module is set up (the material surface real-time temperature monitoring module, i.e., the material surface temperature distribution and the material surface thickness monitoring module, i.e., six sets of two databases of longitudinal thickness distribution of the material surface; the layer thickness is an existing module, and an additional material surface real-time temperature monitoring module is added). The following initial state parameters are collected in real time: Temperature parameters: the surface temperature distribution of the mixture after ignition (detected by an infrared thermal imager); Physical state parameters: the initial thickness of the sintering material layer corresponding to each auxiliary door on the surface of the mixture (detected by an infrared rangefinder). The monitoring data is preprocessed in real time, and outliers are removed (outliers are values ​​outside the normal range, such as materials within about 10 minutes before or after machine start-up or shutdown, and materials whose fluctuations are caused by abnormal fluctuations in material quality or material interruption, which cannot meet the production process requirements and are discharged as waste). After that, the characteristic index is calculated to obtain the initial state characteristic vector Y=[y1,y2,...,ym], where m is the dimension of the initial state parameters.

[0024] Step 3: Training and Prediction of Sintering Endpoint Temperature Prediction Model: 3.1 Constructing the Prediction Model: Using the basic characteristic feature vector X of the mixture and the initial state feature vector Y after ignition as inputs, and the sintering endpoint position (sintering length) as output, a sintering endpoint temperature prediction model is constructed. The model structure includes: a feature fusion layer (for spatiotemporal feature fusion of X and Y), a long-term feature extraction layer, a short-term dynamic capture layer, and a regression prediction layer (outputting the predicted sintering endpoint temperature value P_BTP); 3.2 Model Training: Historical production data under different working conditions (covering different batching schemes and ignition parameter combinations) are collected to construct a training dataset. The prediction model is iteratively trained to ensure that the model prediction error is ≤ ±0.5m; 3.3 Advance Prediction: The standardized feature vector X from Step 1 and the initial state feature vector Y from Step 2 are input into the trained prediction model in real time to output the predicted sintering endpoint temperature advance value P_BTP. The prediction advance is 75%-80% of the total sintering process time.

[0025] Step 4: Intelligent Coordinated Control of Main Exhaust Fan Speed ​​and Sintering Machine Speed: 4.1 Setting the Optimal Range for Sintering End Point: Based on the sintering machine equipment parameters (this invention is based on a sintering machine area of ​​450m²), 2 That is, the effective exhaust length is 90 meters and the width is 5 meters), and the preset optimal sintering endpoint position range is [P]. min ,P max (For the sintering position length, such as 84-87m); 4.2 Deviation Calculation: Calculate the predicted value P_BTP of the sintering endpoint and the midpoint P of the optimal interval. mid The deviation ΔP = P_BTP - P mid4.3 Control Strategy Formulation: Based on the deviation ΔP, a fuzzy PID control algorithm (a reasonable algorithm formed by setting a basic control algorithm after establishing a database based on different raw material structures through data collection, followed by system self-learning and continuous correction) is adopted to formulate a coordinated control strategy for the main exhaust fan speed (n) and sintering machine operating speed (v): When ΔP > 0.5 (predicted endpoint is too far back): increase the main exhaust fan speed Δn = k1 × ΔP (k1 is the speed control coefficient, with a value of 1.0-1.2 r / (min·m)), and at the same time decrease the sintering machine speed Δv = k2 × ΔP (k2 is the speed control coefficient, with a value of 0.02-0.03 m / (min·m)) to accelerate the heat transfer and combustion process of the mixture; when ΔP < 0.5 (Predicted endpoint slightly ahead): Reduce the main exhaust fan speed Δn = k3 × |ΔP| (k3 is the speed control coefficient, with a value of 0.8-1.0 r / (min·m)), and simultaneously increase the sintering machine speed Δv = k4 × |ΔP| (k4 is the speed control coefficient, with a value of 0.01-0.02 m / (min·m)), slowing down the heat transfer and combustion process of the mixture; when |ΔP| ≤ 0.5 (predicted endpoint within the optimal range): maintain the current main exhaust fan speed and sintering machine speed, and achieve stable control by fine-tuning the PID parameters; 4.4 Dynamic feedback correction: collect the exhaust gas temperature and negative pressure data of each air box during the sintering process in real time, calculate the actual deviation of the sintering endpoint, and dynamically correct the control parameters to ensure that the sintering endpoint is finally stable within [P]. min ,P max Within the range. Example

[0026] The present invention will be further described in detail below with reference to specific embodiments: a steel plant 450m 2 The sintering machine uses the method of this invention, and the specific steps are as follows: Basic characteristic parameters of the mixture are collected: the TFe content of the mixture to be sintered is 57.3%, CaO content is 11.55%, SiO2 content is 5.3%, MgO content is 1.8%, and coke ratio is 4.5%; the particle size distribution is 0-1mm 25%, 1-3mm 45%, 3-5mm 20%, and >5mm 10%; the bulk density is 1.75 g / cm³. 3 The moisture content was 7.9%; the average particle size after granulation was 3.21 mm, and the eigenvector X was obtained after normalization.

[0027] Initial state monitoring after ignition: The average surface temperature of the mixture was detected as 80℃ by an infrared thermal imager; the initial thickness of the sintered material layer was detected as 820mm by an infrared rangefinder, and the feature vector Y was obtained after preprocessing.

[0028] Sintering endpoint prediction: Input X and Y into the training of the prediction model, and output the sintering endpoint prediction value P_BTP=84m (optimal interval [84,87]), with a deviation ΔP=1.5.

[0029] Collaborative intelligent control: Based on ΔP=1.5, k1=1.1r / (min・m), Δn=1.5×1.1=1.7r / min; k2=0.02m / (min・m), Δv=1.5×0.02=0.03m / min. After control, the main exhaust fan speed decreased from 825r / min to 823.4r / min, and the sintering machine speed increased from 1.8m / min to 1.83m / min.

[0030] Feedback correction: Real-time monitoring of the surface temperature of the mixture after ignition, calculation of the actual sintering endpoint deviation ≤0.5m, and the final sintering endpoint stabilized at 85m, meeting the requirements of the optimal range.

[0031] After applying the method of this embodiment, the sintering endpoint hit rate of the sintering machine increased from 82% to 96%, the sinter return rate decreased from 21% to 19%, and the comprehensive energy consumption per ton of ore decreased by 1.6%.

[0032] The above description is only a specific embodiment of the present invention, but the structural features protected by the present invention are not limited thereto. Any changes or modifications made by those skilled in the art within the scope of the present invention are covered by the patent scope of the present invention.

Claims

1. A method for predicting and controlling the sintering endpoint based on the characteristics of the mixture and its post-ignition state, characterized in that: Includes the following steps: Step 1: Construct a database of basic characteristic parameters of the mixture: Collect the basic characteristic parameters of the mixture to be sintered on the sintering machine trolley and construct a parameter database that is updated in real time; Step Two: Real-time Monitoring of Initial State Parameters After Ignition: At the downstream end of the sintering machine ignition furnace outlet (i.e., the initial area before the mixture enters the stable combustion stage after ignition), a multi-dimensional monitoring module is set up to collect the following initial state parameters in real time: Temperature parameters: The surface temperature distribution of the mixture after ignition is detected by an infrared thermal imager; Physical state parameters: The initial thickness of the sintering material layer corresponding to each auxiliary door on the surface of the mixture is detected by an infrared rangefinder. The monitoring data is preprocessed in real time, outliers are removed, and characteristic indicators are calculated to obtain the initial state feature vector Y=[y1,y2,...,y...]. m ], where m is the dimension of the initial state parameters; Step 3: Training and Prediction of Sintering Endpoint Temperature Prediction Model: 3.1 Constructing the Prediction Model: Using the basic characteristic feature vector X of the mixture and the initial state feature vector Y after ignition as input, and the sintering endpoint position (i.e., sintering length) as output, a sintering endpoint temperature prediction model is constructed. The model structure includes: a feature fusion layer for spatiotemporal feature fusion of X and Y, a long-term feature extraction layer, a short-term dynamic capture layer, and a regression prediction layer to output the predicted sintering endpoint temperature value P_BTP; 3.2 Model Training: Historical production data under different working conditions, covering different batching schemes and ignition parameter combinations, are collected to construct a training dataset. The prediction model is iteratively trained to ensure that the model prediction error is ≤ ±0.5m; 3.3 Advance Prediction: The standardized feature vector X from Step 1 and the initial state feature vector Y from Step 2 are input into the trained prediction model in real time to output the predicted sintering endpoint temperature advance value P_BTP. The prediction advance is 75%-80% of the total sintering process time. Step 4: Intelligent Coordinated Control of Main Exhaust Fan Speed ​​and Sintering Machine Speed: 4.1 Setting the Optimal Range for Sintering End Point: Based on the preset optimal range for the sintering end point position of the sintering machine [P] min ,P max 4.2 Deviation Calculation: Calculate the predicted sintering endpoint value P_BTP and the midpoint of the optimal interval P. mid The deviation ΔP = P_BTP - P mid 4.3 Regulation Strategy Formulation: Based on the deviation ΔP, a fuzzy PID control algorithm is used to formulate a coordinated regulation strategy for the main exhaust fan speed n and the sintering machine operating speed v: When ΔP > 0.5 and the predicted endpoint is later: increase the main exhaust fan speed Δn = k1 × ΔP, where k1 is the speed regulation coefficient, with a value of 1.0-1.2 r / (min·m), and simultaneously decrease the sintering machine speed Δv = k2 × ΔP, where k2 is the speed regulation coefficient, with a value of 0.02-0.03 m / (min·m) to accelerate the heat transfer and combustion process of the mixture; When ΔP < 0.5 and the predicted endpoint is earlier: decrease the main exhaust fan speed Δn = k3 × |ΔP|, where k3 is the speed regulation coefficient, and k2 is the speed regulation coefficient, and ... The rotational speed control coefficient is 0.8-1.0 r / (min·m), while simultaneously increasing the sintering machine speed Δv=k4×|ΔP|, where k4 is the speed control coefficient, with a value of 0.01-0.02 m / (min·m), to slow down the heat transfer and combustion process of the mixture; when |ΔP|≤0.5, the predicted endpoint is within the optimal range: maintain the current main exhaust fan speed and sintering machine speed, and achieve stable control by fine-tuning the PID parameters; 4.4 Dynamic feedback correction: collect the exhaust gas temperature and negative pressure data of each air box during the sintering process in real time, calculate the actual deviation of the sintering endpoint, and dynamically correct the control parameters to ensure that the sintering endpoint is finally stable within [P]. min ,P max Within the range.

2. The method for predicting and controlling the sintering endpoint based on the characteristics of the mixture and the state after ignition, as described in claim 1, is characterized in that: The basic characteristic parameters mentioned in step one include: physical characteristic parameters: particle size distribution of the mixture: percentage of particles with diameters of 0-1mm, 1-3mm, 3-5mm, and >5mm; moisture content; and granulation characteristic parameters: average particle size of the mixture after granulation. The above parameters are normalized to obtain a standardized feature vector X=[x1,x2,...,x...]. n ], where n is the dimension of the basic characteristic parameter, x n ∈[0,1].

Images