Method and system for predicting position of sintering end point
The artificial neural network model predicts the sintering end point position, which solves the problem of judging the sintering end point position deviation through the exhaust gas temperature in the prior art, and achieves more accurate prediction and adjustment, ensuring the smooth progress of the sintering process and the improvement of product quality.
Patent Information
- Application Number
- CN202510004259.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-02
- Publication Date
- 2025-03-28
AI Technical Summary
The prior art determines that there is a deviation in the sintering end point position through the bellows exhaust gas temperature, resulting in adjustment hysteresis, and it is impossible to accurately determine the sintering end point at different positions in the width direction of the trolley.
The artificial neural network model is used to predict the sintering end point position. By obtaining the sintering raw material parameters, mixture parameters and working parameters of the sintering machine, including the temperature of the bellows exhaust gas, the artificial neural network model is input to obtain the predicted sintering end point position.
It realizes a more accurate prediction of the sintering end point position, reduces adjustment hysteresis, and ensures that the sintering end point in the width direction of the trolley is consistent.
Smart Images
Figure CN119713892A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of metallurgical engineering technology, and in particular to a method and system for predicting a sintering endpoint position. Background Art
[0002] Sintering is an important part of the steel production process. Sintering is to mix various powdered iron-containing raw materials, fuel and flux, and then make a series of physical and chemical changes in the logistics on the sintering equipment to bond the ore powder particles into blocks to form sintered ore. The sintering process not only affects the output of sintered ore, but more importantly, the quality of sintered ore is directly related to the efficiency and product quality of subsequent blast furnace ironmaking.
[0003] Therefore, in the sintering process, the quality of sintered ore can be improved by controlling the sintering endpoint, which refers to the stage in the sintering process when the material is completely reacted and reaches the expected physical and chemical properties. The sintering endpoint is an important parameter reflecting the thermal state of sintering and is closely related to the thermal energy utilization rate of the sintering process. Correctly controlling the sintering endpoint position is of great significance to ensure the smooth progress of the entire sintering production, improve the output and quality of sintered ore, and improve the thermal energy utilization rate of the sintering process.
[0004] At present, the judgment of the sintering end point mainly depends on the exhaust gas temperature passing through the bellows. However, due to the slow response, large lag and long cycle of the sintering process, the sintering end point position judged by the exhaust gas temperature of the bellows can only indicate the state of the sintering process. When the judged sintering end point position deviates from the set position, it will cause adjustment hysteresis. In addition, due to the complexity of the sintering process, the vertical combustion speed of sintering is different, which will lead to differences in the position of the sintering end point in different areas in the width direction of the trolley. Judging the sintering end point by the exhaust gas temperature of a single-point bellows cannot accurately judge the sintering end point at different positions in the width direction of the trolley. Summary of the invention
[0005] The present application provides a sintering endpoint position prediction method and system to solve the problem of adjustment hysteresis when the sintering endpoint position determined by exhaust gas temperature deviates from the set position.
[0006] In a first aspect, the present application provides a method for predicting a sintering endpoint position, comprising:
[0007] Obtaining sintering raw material parameters, mixed material parameters and sintering machine operating parameters during the sintering process of the sintering machine; the sintering machine operating parameters include the wind box exhaust gas temperature;
[0008] Preprocessing the sintering raw material parameters, the mixed material parameters and the sintering machine operating parameters to obtain a sintering data group;
[0009] The sintering data set is input into an artificial neural network model to obtain a first sintering endpoint position output by the artificial neural network model; the artificial neural network model is an artificial neural network model trained according to training data, and the training data includes a number of randomly selected sintering data sets and a second sintering endpoint position calculated according to the wind box exhaust gas temperature; the number of sintering data sets are input values, and the second sintering endpoint position is an output reference value.
[0010] Optionally, the sintering raw material parameters are the material flow rate data of different types of raw materials during the sintering process; the mixture parameters are the moisture content, steam addition amount and particle size distribution data of different types of raw materials after mixing; the sintering machine working parameters are the round roller speed, nine-roller feeder speed, mixture layer thickness, ignition furnace temperature, sintering machine speed, main exhaust frequency, bellows exhaust gas temperature and bellows damper opening.
[0011] Optionally, preprocessing the sintering raw material parameters, the mixed material parameters and the sintering machine operating parameters to obtain the sintering data group includes:
[0012] Sequentially obtain the raw material types of different material discharge ports and the material flow velocity data corresponding to the raw material types;
[0013] Accumulating the material flow rate data corresponding to the same type of raw materials to obtain the raw material flow rate parameters;
[0014] Sequentially acquiring the operating parameters of the sintering machine in the area corresponding to the auxiliary door to obtain the area data corresponding to the auxiliary door;
[0015] The region data, the mixture parameters and the raw material flow rate parameters are stored in the same data group to obtain a sintering data group.
[0016] Optionally, storing the region data, the mixture parameters and the raw material flow rate parameters in the same data group to obtain a sintering data group further comprises:
[0017] Comparing any one data in the sintering data set with a corresponding preset threshold range;
[0018] If any data in the sintering data set is not within the corresponding preset threshold range, the sintering data set is deleted.
[0019] Optionally, the sintering end point position calculated based on the wind box exhaust gas temperature includes:
[0020] Obtain the windbox exhaust gas temperature of windboxes at different positions;
[0021] Using the least square method to perform polynomial fitting on the wind box exhaust gas temperature and determining the order of the polynomial according to the complexity of the wind box exhaust gas temperature to obtain a polynomial function;
[0022] Adjusting the parameters of the polynomial function so as to minimize the sum of squares of vertical distances from the windbox exhaust gas temperature to the fitting curve formed by the polynomial function;
[0023] The wind box position corresponding to the maximum value of the wind box exhaust gas temperature is obtained; the wind box position is the second sintering end point position.
[0024] Optionally, the method further includes:
[0025] Acquiring a preset number of sintering data sets;
[0026] Randomly extracting a first preset proportion of the sintering data set as a training set;
[0027] Constructing an initial artificial neural network model; the initial artificial neural network model includes an input layer, a hidden layer and an output layer;
[0028] The sintering data set is input into the initial artificial neural network model, and back-propagation is performed according to the output result of the initial artificial neural network model to obtain the trained artificial neural network model; the back-propagation is to adjust the weight parameters, bias parameters and learning factors of the initial artificial network model according to the mapping relationship preset by the sintering data set; the mapping relationship is the mapping relationship between the sintering raw material parameters, the mixture parameters and the sintering machine operating parameters and the second sintering end point position.
[0029] Optionally, constructing an initial artificial neural network model includes:
[0030] The parameters of the initial artificial neural network model are input according to the artificial neural network model expression; the artificial neural network model expression is:
[0031]
[0032] Among them, WH ij is the weight of the connection between the i-th input and j-th hidden neuron, m is the number of input neurons, WO j is the connection weight between the jth hidden neuron and the jth output neuron, f h is the hidden neuron activation function, f o is the output neuron activation function, b j is the jth bias hidden neuron, b o is the output neuron bias, and HN is the number of hidden neurons in the output neuron.
[0033] Optionally, inputting the sintering data set into the initial artificial neural network model to perform back propagation according to the output result of the initial artificial neural network model to obtain the trained artificial neural network model comprises:
[0034] inputting the sintering data set into the input layer;
[0035] Inputting the weight of each data in the sintering data set into the hidden layer;
[0036] Obtaining the predicted end point position output by the output layer;
[0037] Calculating a mean square error based on the predicted endpoint position and the second sintering endpoint position;
[0038] If the mean square error is less than a preset threshold, outputting the trained artificial neural network model;
[0039] If the mean square error is greater than or equal to a preset threshold, the weight parameters, bias parameters and learning factors of the initial artificial neural network model are adjusted according to the mean square error.
[0040] Optionally, calculating a mean square error according to the predicted endpoint position and the second sintering endpoint position includes:
[0041] Substituting the predicted endpoint position and the second sintering endpoint position into a mean square error function;
[0042] The mean square error is calculated according to the mean square error function; the mean square error function expression is:
[0043]
[0044] Among them, Y a is the target output, Y o is the network output and N1 is the number of samples.
[0045] In a second aspect, the present application provides a sintering endpoint position prediction system, which is applied to the sintering endpoint prediction method described in the first aspect, comprising: a sensor module and a terminal; the sensor module is connected to the terminal; the sensor module comprises a thermocouple sensor, the thermocouple sensor is arranged in a wind box, and the number of the thermocouple sensors is associated with the number of auxiliary doors;
[0046] The thermocouple sensor is configured to: obtain the wind box exhaust gas temperature of the wind box;
[0047] The sensor module is configured to: obtain sintering raw material parameters, mixed material parameters and sintering machine working parameters during the sintering process of the sintering machine;
[0048] The terminal is configured as follows:
[0049] Preprocessing the sintering raw material parameters, the mixed material parameters and the sintering machine operating parameters to obtain the sintering data group;
[0050] The sintering data set is input into an artificial neural network model to obtain a first sintering endpoint position output by the artificial neural network model; the artificial neural network model is an artificial neural network model trained according to training data, and the training data includes a number of randomly selected groups of the sintering data sets and a second sintering endpoint position calculated according to the wind box exhaust gas temperature; the several sintering data sets are input values, and the second sintering endpoint position is an output reference value.
[0051] It can be seen from the above technical scheme that the present application provides a sintering end point position prediction method and system, the method comprising: obtaining sintering raw material parameters, mixture parameters and sintering machine working parameters during the sintering process of the sintering machine; the sintering machine working parameters include the wind box exhaust gas temperature; performing preprocessing on the sintering raw material parameters, the mixture parameters and the sintering machine working parameters to obtain a sintering data group; inputting the sintering data group into an artificial neural network model to obtain a first sintering end point position output by the artificial neural network model; the artificial neural network model is an artificial neural network model trained according to training data, and the training data includes several randomly selected sintering data groups and a second sintering end point position calculated according to the wind box exhaust gas temperature; several of the sintering data groups are input values, and the second sintering end point position is an output reference value. By inputting the sintering raw material parameters, mixed material parameters and sintering machine working parameters into the artificial neural network model to obtain the first sintering end point position, the sintering machine speed, main exhaust fan frequency, etc. can be intervened in advance to ensure that the actual sintering end point position meets the set requirements, so as to solve the problem of adjustment delay when the sintering end point position judged by the exhaust gas temperature deviates from the set position. In the width direction of the trolley, the sintering end point is predicted in different regions according to the distribution of auxiliary doors, and the opening of the auxiliary doors in different regions is adjusted in advance to make the sintering end point in the width direction of the trolley tend to be consistent. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0053] Figure 1 A flow chart of a method for predicting the sintering endpoint position provided in an embodiment of the present application;
[0054] Figure 2A flow chart of a method for executing preprocessing of sintering raw material parameters, mixed material parameters and sintering machine operating parameters provided in an embodiment of the present application;
[0055] Figure 3 A flow chart of a method for training an initial artificial neural network model provided in an embodiment of the present application;
[0056] Figure 4 A schematic diagram of the artificial neural network structure provided in the embodiment of the present application;
[0057] Figure 5 A flow chart of the artificial neural network model for training provided in an embodiment of the present application. DETAILED DESCRIPTION
[0058] The following embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following embodiments do not represent all implementations consistent with the present application. They are only examples of systems and methods consistent with some aspects of the present application as detailed in the claims.
[0059] Sintering is an important part of the steel production process. Sintering is to mix various powdered iron-containing raw materials, fuel and flux, and then make a series of physical and chemical changes in the logistics on the sintering equipment to bond the ore powder particles into blocks to form sintered ore. The sintering process not only affects the output of sintered ore, but more importantly, the quality of sintered ore is directly related to the efficiency and product quality of subsequent blast furnace ironmaking.
[0060] Therefore, in the sintering process, the quality of sintered ore can be improved by controlling the sintering endpoint, where the sintering endpoint refers to the stage in the sintering process when the material is completely reacted and reaches the expected physical and chemical properties. The sintering endpoint is an important parameter reflecting the thermal state of sintering and is closely related to the thermal energy utilization rate of the sintering process. Correctly controlling the sintering endpoint position is of great significance to ensure the smooth progress of the entire sintering production, improve the output and quality of sintered ore, and improve the thermal energy utilization rate of the sintering process.
[0061] In some embodiments, the sintering end point position can be predicted by obtaining the bellows exhaust gas temperature. Specifically, the distribution of the bellows exhaust gas temperature can be checked by obtaining the bellows exhaust gas temperature data at different positions, and the maximum value of the bellows exhaust gas temperature can be obtained. The sintering end point position is near the maximum value of the bellows exhaust gas temperature. However, due to the characteristics of slow response, large lag, and long cycle of the sintering process, the sintering end point position determined by the exhaust gas temperature of the bellows can only indicate the state of the sintering process. When the determined sintering end point position deviates from the set position, adjustment delay will occur. In addition, due to the complexity of the sintering process, the vertical combustion speed of sintering is different, which will lead to differences in the position of the sintering end point in different areas in the width direction of the trolley. The exhaust gas temperature of a single-point bellows cannot accurately determine the sintering end point at different positions in the width direction of the trolley.
[0062] To solve the problem of adjustment hysteresis when the sintering end point position determined by the exhaust gas temperature deviates from the set position, refer to Figure 1 , Figure 1 This is a flow chart of a method for predicting the sintering endpoint position provided in an embodiment of the present application. Some embodiments of the present application provide a method for predicting the sintering endpoint position, including:
[0063] S100: Acquire sintering raw material parameters, mixed material parameters and sintering machine working parameters during the sintering process of the sintering machine.
[0064] Among them, the sintering machine working parameters include the wind box exhaust gas temperature. In some embodiments, the sintering raw material parameters are the material flow rate data of different types of raw materials during the sintering process; the mixed material parameters are the moisture content, steam addition amount and particle size distribution data of different types of raw materials after mixing; the sintering machine working parameters are the round roller speed, the nine-roller distributor speed, the mixed material layer thickness, the ignition furnace temperature, the sintering machine speed, the main extraction frequency, the wind box exhaust gas temperature and the wind box damper opening.
[0065] Specifically, the sintering raw material parameters include the flow rate data of return ore, flux, fuel, dolomite, quicklime, etc. The mixed material parameters include the moisture content after the second mixing (granulator), the amount of steam added, and the mixed material particle size distribution data. The sintering machine working parameters include the round roller speed, the nine-roller distributor speed, the mixed material layer thickness, the ignition furnace temperature, the sintering machine speed, the main extraction frequency, the wind box exhaust gas temperature, and the wind box damper opening. The above data can be obtained by setting corresponding sensors at different structures of the sintering machine for detection. For example, according to the operation process of the sintering machine, the material flow rate of the first material drop point such as the 1# return ore drop port is recorded. After a period of time, the total belt of the mixing reaches the second material drop point, and the material flow rate of the second material drop point such as the 2# return ore drop port is recorded. By analogy, the material flow rate of all the material drop points in the mixing process is recorded. After the mixing is completed, the moisture content, steam addition amount, and mixed material particle size distribution data after the second mixing are continued to be recorded. When the material after the second mixing reaches the trolley, the relevant round roller speed, nine-roller distributor speed, mixed material layer thickness, ignition furnace temperature, sintering machine speed, main pumping frequency, and bellows damper opening are recorded. At the thermocouple sensor of the 8# bellows, the current bellows exhaust gas temperature of the 8# bellows is recorded until the bellows exhaust gas temperature of the 16# bellows is passed. Among them, the 8# bellows is the middle bellows and the 16# bellows is the last bellows, completing the first set of logistics tracking data. At the same time, after the preset interval time, the second set of logistics tracking data and the third set of logistics tracking data are recorded...
[0066] S200: Preprocessing the sintering raw material parameters, the mixed material parameters and the sintering machine operating parameters to obtain a sintering data set.
[0067] After obtaining several groups of logistics tracking data, it is necessary to pre-process the sintering raw material parameters, mixture parameters and sintering machine operating parameters to meet the input or training requirements of the artificial neural network model. Different types of data need to be processed with different requirements.
[0068] Among them, Figure 2 As shown, preprocessing is performed on the sintering raw material parameters, the mixed material parameters and the sintering machine operating parameters to obtain a sintering data group including:
[0069] S210: sequentially acquiring raw material types of different material discharge silo openings and material flow velocity data corresponding to the raw material types.
[0070] S220: Accumulate the material flow rate data corresponding to the same raw material type to obtain the raw material flow rate parameter.
[0071] Among them, the preprocessing of sintering raw material parameters requires the accumulation of the material flow rate data of the same raw material type according to the raw material type. For example, if there are two silos for return ore, the material flow rate data of 1# return ore and 2# return ore need to be accumulated for the same set of tracked data. Similarly, if there are multiple silos for flux, fuel, dust, dolomite, and quicklime, the material flow rate data of the same raw material type should be accumulated according to the raw material type to obtain the raw material flow rate parameters.
[0072] S230: sequentially acquiring the operating parameters of the sintering machine in the area corresponding to the auxiliary door to obtain the area data corresponding to the auxiliary door.
[0073] Among them, the preprocessing of the sintering machine working parameters requires dividing the trolley width direction into several areas through the auxiliary door position. The number of divided areas is the same as the number of auxiliary doors. For example, the trolley is equipped with six auxiliary doors, which can be divided into six areas in the trolley width direction to obtain the mixture layer thickness in each area and the bellows damper opening corresponding to the auxiliary door, and then record the same data of different auxiliary doors, such as round roller speed, nine-roller feeder speed, ignition furnace temperature, sintering machine speed and main pumping frequency.
[0074] In addition, in an embodiment of the present application, each bellows is provided with a number of thermocouple sensors, and the number of thermocouple sensors provided on each bellows is the same as the number of auxiliary doors. For example, the sintering machine is provided with 16 groups of bellows along the length direction of the trolley, and each group of bellows is divided into two sides AB. Six auxiliary doors are provided in the width direction of the trolley, and the installation position of each group of bellows thermocouple sensors is consistent with the midpoint position of the auxiliary door, that is, six thermocouple sensors are provided on each group of bellows. Therefore, the corresponding bellows exhaust gas temperature in the auxiliary door direction can be obtained respectively through the thermocouple sensors.
[0075] Finally, the width of the trolley is divided into several areas according to the position of the auxiliary door, and divided into several corresponding data groups. The thickness of the mixed material layer and the temperature of the wind box exhaust gas in each data group have corresponding auxiliary doors. Among them, each data group can be called a sintering data group.
[0076] In addition, the sintering end position in the current auxiliary door direction needs to be calculated based on the wind box exhaust gas temperature of different wind boxes in the same group. The sintering end position calculated based on the wind box exhaust gas temperature includes:
[0077] S231: Obtain the wind box exhaust gas temperature of wind boxes at different positions.
[0078] S232: Perform polynomial fitting on the wind box exhaust gas temperature using the least square method and determine the order of the polynomial according to the complexity of the wind box exhaust gas temperature to obtain a polynomial function.
[0079] S233: Adjust the parameters of the polynomial function so that the sum of the squares of the vertical distances from the wind box exhaust gas temperature to the fitting curve formed by the polynomial function is minimized.
[0080] S234: Obtaining the wind box position corresponding to the maximum value of the wind box exhaust gas temperature; the wind box position is the second sintering end point position.
[0081] Specifically, the bellows exhaust gas temperatures of bellows #8 to #16 are fitted using the least squares method to find a polynomial function, and then the order of the polynomial function is determined according to the complexity of the bellows temperature. For example, if the bellows exhaust gas temperature satisfies the variation law of the quadratic function, the order of the polynomial function is second order. After the order of the polynomial function is determined, the square sum of the vertical distances from all bellows exhaust gas temperatures to the fitting curve formed by the polynomial function can be calculated, and the parameters in the polynomial function can be adjusted to minimize the square sum. The highest point of the bellows exhaust gas temperature can be found by curve fitting, and then the bellows position corresponding to the highest point of the bellows exhaust gas temperature can be obtained, that is, the second sintering end point position.
[0082] S240: storing the region data, the mixture parameters and the raw material flow velocity parameters in the same data group to obtain a sintering data group.
[0083] When storing the regional data, mixture parameters and raw material flow velocity parameters in the same data group, it is also necessary to remove the abnormal data in the data group. Therefore, any data in the sintering data group is compared with the corresponding preset threshold range; if any data in the sintering data group is not within the corresponding preset threshold range, the sintering data group is deleted. For example, if the round roller speed in the sintering data group is not within the set round roller speed preset threshold range, the entire sintering data group needs to be deleted. In addition, an alarm prompt can also be made so that the operation and maintenance personnel can handle the corresponding abnormalities.
[0084] S300: Inputting the sintering data set into the artificial neural network model to obtain the first sintering end point position output by the artificial neural network model.
[0085] The artificial neural network model is an artificial neural network model trained according to training data, and the training data includes a number of randomly selected sintering data groups and the second sintering end point position calculated according to the wind box exhaust gas temperature; the number of sintering data groups are input values, and the second sintering end point position is an output reference value. Therefore, before inputting the sintering data groups into the artificial neural network model, it is necessary to establish and train the artificial neural network model, such as Figure 3 and Figure 4 As shown, Figure 3 Flowchart of the method for training the initial artificial neural network model, Figure 4 Schematic diagram of the artificial neural network model structure. The steps of training the artificial neural network model include:
[0086] S310: Acquire a preset number of sintering data sets.
[0087] S320: Randomly select a first preset proportion of the sintering data set as a training set.
[0088] Specifically, a preset number of sintering data groups are randomly selected from the recorded sintering data groups for model training and model testing, and the preset number of sintering data groups is 50%-80% of the recorded sintering data groups. For example, 65% of the sintering data groups are randomly selected from the sintering data groups as input data for training the model, and the first preset ratio is 65%. When 70% of the sintering data groups are randomly selected from the sintering data as input data for training the model, the first preset ratio is 70%.
[0089] S330: Construct an initial artificial neural network model.
[0090] The initial artificial neural network model includes an input layer, a hidden layer, and an output layer. The structure of the artificial neural network model is as follows: Figure 4 As shown, when constructing the initial artificial neural network model, parameters can be set according to the artificial neural network model expression. Specifically, the artificial neural network model expression is:
[0091]
[0092] Among them, WH ij is the weight of the connection between the i-th input and j-th hidden neuron, m is the number of input neurons, WO j is the connection weight between the jth hidden neuron and the jth output neuron, f h is the hidden neuron activation function, f o is the output neuron activation function, b j is the jth bias hidden neuron, b o is the output neuron bias, and HN is the number of hidden neurons in the output neuron. After the parameters of the initial artificial neural network model are set, the initial neural network model can be trained.
[0093] S340: Inputting the sintering data set into the initial artificial neural network model to perform back propagation according to the output result of the initial artificial neural network model to obtain a trained artificial neural network model.
[0094] Among them, back propagation is to adjust the weight parameters, bias parameters and learning factors of the initial artificial network model according to the mapping relationship pre-set in the sintering data group; the mapping relationship is the mapping relationship between the sintering raw material parameters, the mixture parameters and the sintering machine working parameters and the second sintering end point position.
[0095] Therefore, in the process of model training, it is also necessary to calculate the second sintering end point position through the bellows exhaust gas temperature as the target output, and compare the predicted end point position obtained after inputting the sintering data group into the initial artificial network model with the second sintering end point position. According to the comparison result between the preset end point position and the second sintering end point position, the weight parameters, bias parameters and learning factors of the initial artificial neural network model are updated.
[0096] Among them, Figure 5 As shown, the sintering data set is input into the initial artificial neural network model, and the output result of the initial artificial neural network model is back-propagated to obtain the trained artificial neural network model, which specifically includes:
[0097] The sintering data set is fed into the input layer.
[0098] The weight of each data in the sintering data set is input into the hidden layer.
[0099] Get the predicted end point position output by the output layer.
[0100] The mean square error is calculated based on the predicted endpoint position and the second sintering endpoint position.
[0101] Wherein, calculating the mean square error according to the predicted end point position and the second sintering end point position comprises:
[0102] The predicted endpoint position and the second sintering endpoint position are substituted into the mean square error function.
[0103] The mean square error is calculated according to the mean square error function; the expression of the mean square error function is:
[0104]
[0105] Among them, Y a is the target output, Y o is the network output and N1 is the number of samples.
[0106] If the mean square error is less than the preset threshold, the trained artificial neural network model is output.
[0107] If the mean square error is greater than or equal to the preset threshold, the weight parameters, bias parameters and learning factors of the initial artificial neural network model are adjusted according to the mean square error.
[0108] Specifically, the input layer inputs the sintering data set, which includes sintering raw material parameters, mixture parameters and sintering machine operating parameters, and the hidden layer inputs the weights between the variables. Then the predicted end point position output by the initial artificial neural network model is obtained, and then the predicted end point position and the second sintering end point position are substituted into the mean square error function to calculate the mean square error, and it is determined whether the mean square error is less than the preset threshold, where the preset threshold is 0.001. If the mean square error is greater than 0.001, the parameter error terms in the input layer and the hidden layer are calculated by the automatic correction method, the bias terms of the hidden neuron activation function and the output neuron activation function are calculated, and the weight parameters and learning factors are readjusted. The model network parameters to be adjusted are calculated according to the parameter error terms and the bias terms, wherein the network parameters to be adjusted are used to update the network parameters of the network model. When the mean square error is less than 0.001, it means that the initial artificial neural network training is completed. During the generation process, the sintering end point position can be predicted by obtaining the sintering raw material parameters, mixture parameters and sintering machine working parameters, and then intervention can be made in advance by adjusting the sintering machine speed, main exhaust fan frequency, etc. to ensure that the actual position of the sintering end point meets the set requirements, so as to solve the problem of adjustment delay when the sintering end point position judged by the exhaust gas temperature deviates from the set position.
[0109] In addition, the present application divides the width direction of the trolley into multiple areas according to the auxiliary door, predicts the sintering end point position of each area, and adjusts the opening of the auxiliary door in advance to achieve consistency of the sintering end points in the width direction of the trolley.
[0110] In some embodiments, the embodiments of the present application also provide a sintering end point position prediction system, which is applied to the sintering end point prediction method provided in the above embodiments, including: a sensor module and a terminal; the sensor module is connected to the terminal; the sensor module includes a thermocouple sensor, the thermocouple sensor is arranged in the bellows, and the number of thermocouple sensors is associated with the number of auxiliary doors.
[0111] The thermocouple sensor is configured to: obtain a windbox exhaust temperature of the windbox;
[0112] The sensor module is configured to: obtain sintering raw material parameters, mixed material parameters and sintering machine working parameters during the sintering process of the sintering machine;
[0113] The terminal is configured as:
[0114] Preprocessing the sintering raw material parameters, the mixed material parameters and the sintering machine operating parameters to obtain a sintering data group;
[0115] The sintering data set is input into the artificial neural network model to obtain the first sintering endpoint position output by the artificial neural network model; the artificial neural network model is an artificial neural network model trained according to training data, and the training data includes several groups of sintering data sets randomly selected and the second sintering endpoint position calculated according to the wind box exhaust gas temperature; the several sintering data sets are input values, and the second sintering endpoint position is the output reference value.
[0116] The effects of the above-mentioned system when applying the above-mentioned method can be found in the description of the above-mentioned method embodiment, which will not be repeated here.
[0117] It can be seen from the above technical scheme that the present application provides a sintering end point position prediction method and system, and the sintering end point position prediction method includes: obtaining sintering raw material parameters, mixed material parameters and sintering machine working parameters during the sintering process of the sintering machine; the sintering machine working parameters include the wind box exhaust gas temperature; pre-processing the sintering raw material parameters, mixed material parameters and sintering machine working parameters to obtain a sintering data group; inputting the sintering data group into an artificial neural network model to obtain a first sintering end point position output by the artificial neural network model; the artificial neural network model is an artificial neural network model obtained by training according to training data, and the training data includes a number of randomly selected sintering data groups and a second sintering end point position calculated according to the wind box exhaust gas temperature; the number of sintering data groups are input values, and the second sintering end point position is an output reference value. By inputting the sintering raw material parameters, mixed material parameters and sintering machine working parameters into the artificial neural network model to obtain the first sintering end point position, the sintering machine speed, the main exhaust fan frequency, etc. can be intervened in advance to ensure that the actual position of the sintering end point meets the set requirements, so as to solve the problem of adjustment hysteresis when the sintering end point position judged by the exhaust gas temperature deviates from the set position. The sintering endpoint is predicted for each area in the width direction of the trolley according to the distribution of auxiliary doors. Early intervention is carried out by adjusting the opening of auxiliary doors in different areas to achieve a consistent sintering endpoint in the width direction of the trolley.
[0118] Similar parts between the embodiments provided in this application can be referenced to each other. The specific implementation methods provided above are only a few examples under the general concept of this application and do not constitute a limitation on the protection scope of this application. For those skilled in the art, any other implementation methods expanded based on the scheme of this application without creative work belong to the protection scope of this application.
Claims
1. A method for predicting the sintering endpoint position, characterized in that: include: Obtaining sintering raw material parameters, mixed material parameters and sintering machine operating parameters during the sintering process of the sintering machine; the sintering machine operating parameters include the wind box exhaust gas temperature; Preprocessing the sintering raw material parameters, the mixed material parameters and the sintering machine operating parameters to obtain a sintering data group; The sintering data set is input into an artificial neural network model to obtain a first sintering endpoint position output by the artificial neural network model; the artificial neural network model is an artificial neural network model trained according to training data, and the training data includes a number of randomly selected sintering data sets and a second sintering endpoint position calculated according to the wind box exhaust gas temperature; the number of sintering data sets are input values, and the second sintering endpoint position is an output reference value.
2. The method for predicting the sintering endpoint position according to claim 1, characterized in that: The sintering raw material parameters are the material flow rate data of different types of raw materials during the sintering process; the mixed material parameters are the moisture content, steam addition amount and particle size distribution data of different types of raw materials after mixing; the sintering machine working parameters are the round roller speed, nine-roller distributor speed, mixed material layer thickness, ignition furnace temperature, sintering machine speed, main extraction frequency, bellows exhaust gas temperature and bellows damper opening.
3. The method for predicting the sintering end point position according to claim 2, characterized in that: Preprocessing the sintering raw material parameters, the mixed material parameters and the sintering machine operating parameters to obtain the sintering data group includes: Sequentially obtain the raw material types of different material discharge ports and the material flow velocity data corresponding to the raw material types; Accumulating the material flow rate data corresponding to the same type of raw materials to obtain the raw material flow rate parameters; Sequentially acquiring the operating parameters of the sintering machine in the area corresponding to the auxiliary door to obtain the area data corresponding to the auxiliary door; The region data, the mixture parameters and the raw material flow rate parameters are stored in the same data group to obtain a sintering data group.
4. The method for predicting the sintering end point position according to claim 3, characterized in that: Storing the region data, the mixture parameters and the raw material flow rate parameters in the same data group to obtain a sintering data group further includes: Comparing any one data in the sintering data set with a corresponding preset threshold range; If any data in the sintering data set is not within the corresponding preset threshold range, the sintering data set is deleted.
5. The method for predicting the sintering endpoint position according to claim 1, characterized in that: The sintering end point position calculated based on the wind box exhaust gas temperature includes: Obtain the windbox exhaust gas temperature of windboxes at different positions; Using the least square method to perform polynomial fitting on the wind box exhaust gas temperature and determining the order of the polynomial according to the complexity of the wind box exhaust gas temperature to obtain a polynomial function; Adjusting the parameters of the polynomial function so as to minimize the sum of squares of vertical distances from the windbox exhaust gas temperature to the fitting curve formed by the polynomial function; The wind box position corresponding to the maximum value of the wind box exhaust gas temperature is obtained; the wind box position is the second sintering end point position.
6. The method for predicting the sintering end point position according to claim 1, characterized in that: The method further comprises: Acquiring a preset number of sintering data sets; Randomly extracting a first preset proportion of the sintering data set as a training set; Constructing an initial artificial neural network model; the initial artificial neural network model includes an input layer, a hidden layer and an output layer; The sintering data set is input into the initial artificial neural network model, and back-propagation is performed according to the output result of the initial artificial neural network model to obtain the trained artificial neural network model; the back-propagation is to adjust the weight parameters, bias parameters and learning factors of the initial artificial network model according to the mapping relationship preset by the sintering data set; the mapping relationship is the mapping relationship between the sintering raw material parameters, the mixture parameters and the sintering machine operating parameters and the second sintering end point position.
7. The method for predicting the sintering end point position according to claim 6, characterized in that: Building the initial artificial neural network model includes: The parameters of the initial artificial neural network model are input according to the artificial neural network model expression; the artificial neural network model expression is: Among them, WH ij is the weight of the connection between the i-th input and j-th hidden neuron, m is the number of input neurons, WO j is the connection weight between the i-th hidden neuron and the j-th output neuron, f h is the hidden neuron activation function, f o is the output neuron activation function, b j is the jth bias hidden neuron, b o is the output neuron bias, and HN is the number of hidden neurons in the output neuron.
8. The method for predicting the sintering end point position according to claim 6, characterized in that: Inputting the sintering data set into the initial artificial neural network model to perform back propagation according to the output result of the initial artificial neural network model, and obtaining the trained artificial neural network model comprises: inputting the sintering data set into the input layer; Inputting the weight of each data in the sintering data set into the hidden layer; Obtaining the predicted end point position output by the output layer; Calculating a mean square error based on the predicted endpoint position and the second sintering endpoint position; If the mean square error is less than a preset threshold, outputting the trained artificial neural network model; If the mean square error is greater than or equal to a preset threshold, the weight parameters, bias parameters and learning factors of the initial artificial neural network model are adjusted according to the mean square error.
9. The method for predicting the sintering end point position according to claim 8, characterized in that: Calculating the mean square error according to the predicted endpoint position and the second sintering endpoint position comprises: Substituting the predicted endpoint position and the second sintering endpoint position into a mean square error function; The mean square error is calculated according to the mean square error function; the mean square error function expression is: Among them, Y a is the target output, Y o is the network output and N1 is the number of samples.
10. A sintering end point position prediction system, characterized in that: The sintering endpoint prediction method applied to any one of claims 1 to 9 comprises: a sensor module and a terminal; the sensor module is connected to the terminal; the sensor module comprises a thermocouple sensor, the thermocouple sensor is arranged in a wind box, and the number of the thermocouple sensors is associated with the number of auxiliary doors; The thermocouple sensor is configured to: obtain the wind box exhaust gas temperature of the wind box; The sensor module is configured to: obtain sintering raw material parameters, mixed material parameters and sintering machine working parameters during the sintering process of the sintering machine; The terminal is configured as follows: Preprocessing the sintering raw material parameters, the mixed material parameters and the sintering machine operating parameters to obtain the sintering data group; The sintering data set is input into an artificial neural network model to obtain a first sintering endpoint position output by the artificial neural network model; the artificial neural network model is an artificial neural network model trained according to training data, and the training data includes a number of randomly selected groups of the sintering data sets and a second sintering endpoint position calculated according to the wind box exhaust gas temperature; the several sintering data sets are input values, and the second sintering endpoint position is an output reference value.