WO2022037500A1 - 一种基于料层厚度预测的布料控制系统及方法 - Google Patents
一种基于料层厚度预测的布料控制系统及方法 Download PDFInfo
- Publication number
- WO2022037500A1 WO2022037500A1 PCT/CN2021/112557 CN2021112557W WO2022037500A1 WO 2022037500 A1 WO2022037500 A1 WO 2022037500A1 CN 2021112557 W CN2021112557 W CN 2021112557W WO 2022037500 A1 WO2022037500 A1 WO 2022037500A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- material layer
- speed
- layer thickness
- roller
- thickness
- Prior art date
Links
- 239000000463 material Substances 0.000 title claims abstract description 390
- 238000009826 distribution Methods 0.000 title claims abstract description 166
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000005245 sintering Methods 0.000 claims abstract description 214
- 238000005457 optimization Methods 0.000 claims abstract description 16
- 238000005096 rolling process Methods 0.000 claims abstract description 16
- 239000000203 mixture Substances 0.000 claims description 82
- 238000012549 training Methods 0.000 claims description 40
- 238000001514 detection method Methods 0.000 claims description 33
- 238000003062 neural network model Methods 0.000 claims description 32
- 238000012545 processing Methods 0.000 claims description 18
- 230000007246 mechanism Effects 0.000 claims description 17
- 238000013507 mapping Methods 0.000 claims description 15
- 239000013598 vector Substances 0.000 claims description 14
- 239000002994 raw material Substances 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 5
- 238000007619 statistical method Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 abstract description 23
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 abstract description 8
- 229910052742 iron Inorganic materials 0.000 abstract description 4
- 229910000831 Steel Inorganic materials 0.000 abstract description 3
- 239000010959 steel Substances 0.000 abstract description 3
- 238000003723 Smelting Methods 0.000 abstract description 2
- 239000004744 fabric Substances 0.000 description 14
- 238000002156 mixing Methods 0.000 description 12
- 238000005259 measurement Methods 0.000 description 9
- 238000004519 manufacturing process Methods 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000002955 isolation Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 239000000843 powder Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- 235000008733 Citrus aurantifolia Nutrition 0.000 description 1
- 235000011941 Tilia x europaea Nutrition 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000010924 continuous production Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000004571 lime Substances 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005453 pelletization Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27B—FURNACES, KILNS, OVENS, OR RETORTS IN GENERAL; OPEN SINTERING OR LIKE APPARATUS
- F27B21/00—Open or uncovered sintering apparatus; Other heat-treatment apparatus of like construction
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27D—DETAILS OR ACCESSORIES OF FURNACES, KILNS, OVENS, OR RETORTS, IN SO FAR AS THEY ARE OF KINDS OCCURRING IN MORE THAN ONE KIND OF FURNACE
- F27D19/00—Arrangements of controlling devices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27D—DETAILS OR ACCESSORIES OF FURNACES, KILNS, OVENS, OR RETORTS, IN SO FAR AS THEY ARE OF KINDS OCCURRING IN MORE THAN ONE KIND OF FURNACE
- F27D19/00—Arrangements of controlling devices
- F27D2019/0028—Regulation
- F27D2019/0075—Regulation of the charge quantity
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27M—INDEXING SCHEME RELATING TO ASPECTS OF THE CHARGES OR FURNACES, KILNS, OVENS OR RETORTS
- F27M2003/00—Type of treatment of the charge
- F27M2003/04—Sintering
Definitions
- the present application relates to the technical field of iron and steel smelting, and in particular, to a distribution control system and method based on material layer thickness prediction.
- Sintering refers to the process of converting powdery materials into dense bodies. In the iron and steel production process, it refers to mixing iron ore powder, coal powder and lime in a certain proportion, and sintering to obtain sintered ore with blast furnace production process requirements. , transporting raw materials for blast furnace ironmaking.
- the thickness of the material layer is appropriate is one of the important parameters affecting the output and quality of sintering production.
- it is usually used to adjust the opening of the auxiliary door to adjust the material in the area corresponding to the auxiliary door. The thickness of the layer is adjusted, or the rotation speed of the round roller feeder is adjusted to change the material layer in the width range of the trolley as a whole. If the opening of the auxiliary door is adjusted to a large extent, the flow of the corresponding cloth device to the sintering trolley will be large, and the thickness of the corresponding material layer will be thick, otherwise, the thickness of the material layer will be small.
- the material layer thickness of the sintering trolley is generally detected in real time to control the material layer.
- the specific process is, in the process of material distribution by the sintering trolley, at the downstream of the moving direction of the sintering trolley, the thickness of the material layer of the sintering trolley is monitored, and the feedback control of the material distribution is carried out according to the detected material layer thickness.
- this material layer thickness detection method detects the material layer thickness of the sintered trolley that has been clothed. It takes about 2 to 3 minutes from the auxiliary door adjustment point to the layer thickness detection point, and the feedback control has a large hysteresis. . If the thickness of the material layer is found to be inappropriate at this time, it cannot be changed, which is not conducive to realizing the timely and stable control of the material for the sintering system.
- the present application provides a distribution control system and method based on material layer thickness prediction to solve the problem of hysteresis in the distribution control of the sintering system by detecting the material layer thickness of the sintering trolley in the prior art.
- a first aspect of the present application provides a material layer thickness prediction-based material distribution control system
- the material distribution control system includes a round roller feeder, a roller type material distribution machine and a sintering trolley, and the round roller feeder is used to feed all materials
- the roller feeder supplies mixed material
- the roller feeder is used for material distribution to the sintering trolley
- the material distribution control system further includes a mixture detection mechanism, a feeding roller controller connected with the round roller feeder, and The distribution roller controller and auxiliary door controller connected with the roller distributor, and the sintering trolley controller connected with the sintering trolley; and the mixture detection mechanism, the feeding roller controller, the distribution roller controller, and the auxiliary door control central processing unit connected to the sintering trolley controller; of which:
- the central processing unit is configured to perform the following steps:
- the thickness of the material layer is predicted, and the predicted value of the thickness of the material layer is obtained;
- the feeding roller controller to adjust the speed of the feeding roller to the speed of the feeding roller to be adjusted; drive the distribution roller controller to adjust the speed of the distribution roller to the speed of the distribution roller to be adjusted; drive the auxiliary door controller Adjust the opening of the auxiliary door to the opening of the auxiliary door to be adjusted; drive the sintering trolley controller to adjust the speed of the sintering trolley to the speed of the sintering trolley to be adjusted;
- the rolling optimization model is used to calculate the corresponding feed roller speed, distribution roller speed, auxiliary door opening when the variance of the thickness deviation value of the material layer is the smallest under the condition that the bulk density of the mixture is unchanged. and sintering trolley speed.
- the variance of the material layer thickness deviation value is obtained in the following manner:
- ⁇ is the variance of the thickness deviation value of the material layer
- E(k) refers to the deviation value of the material layer thickness
- R(k) the preset target value of the material layer thickness
- Y(k) the predicted value of the material layer thickness
- the thickness of the material layer is predicted to obtain the predicted value of the thickness of the material layer, and the following steps are specifically performed:
- Data reduction is performed on the characteristic value of the thickness of the material layer to obtain the predicted value of the thickness of the material layer.
- the material distribution control system system further includes a material layer thickness detection device arranged above the sintering trolley, the material layer thickness detection device is connected to the central processing unit, and the material layer thickness detection device is used for pre-setting.
- the time interval of the sintering trolley is to detect the material layer thickness of the sintering trolley, and obtain the measured value of the material layer thickness; input the eigenvector of the predicted material layer thickness into the pre-established material layer thickness dynamic prediction model to generate the material layer thickness eigenvalues ,Also includes:
- the material layer thickness dynamic prediction model is updated online by using the learning sample, and the updated material layer thickness dynamic prediction model is obtained.
- hi is the obtained predicted value of the material layer thickness at time i
- ki is the generated characteristic value of the material layer thickness at time i
- H is the maximum allowable thickness of the material layer on the sintering trolley.
- the ratio of the speed of the sintering trolley to the maximum rotational speed of the sintering trolley is the ratio of the speed of the sintering trolley to the maximum rotational speed of the sintering trolley.
- the material layer thickness dynamic prediction model is generated based on neural network model training, and established according to the following steps:
- the bulk density, feed roller speed, distribution roller speed and sintering trolley speed of the N groups of independent mixtures are quantified to the same interval according to a certain shrinkage ratio, and combined with the opening of the auxiliary door, as the input of the N groups of training samples;
- the actual material layer thickness corresponding to the input of the above N groups of training samples on the sintering trolley is detected, and the actual characteristic value of the actual material layer thickness is calculated, and the actual characteristic value is used as the N groups of output training samples;
- the neural network model is trained by the time backpropagation method
- the material layer thickness dynamic prediction model can also be a material layer thickness prediction table, and is established according to the following steps:
- the preset time interval detect the actual material layer thickness corresponding to the above N groups of input training samples on the sintering trolley, and calculate the actual characteristic value of the actual material layer thickness;
- a second aspect of the present application provides a method for controlling distribution based on prediction of material layer thickness, the method for controlling distribution includes:
- the thickness of the material layer is predicted, and the predicted value of the thickness of the material layer is obtained;
- the rolling optimization model is used to calculate the corresponding feed roller speed, distribution roller speed, auxiliary door opening when the variance of the thickness deviation value of the material layer is the smallest under the condition that the bulk density of the mixture is unchanged. and sintering trolley speed.
- the step of predicting the thickness of the material layer according to the bulk density of the mixture, the rotating speed of the feeding roller, the rotating speed of the distribution roller, the opening of the auxiliary door and the speed of the sintering trolley, and obtaining the predicted value of the thickness of the material layer includes: :
- Data reduction is performed on the characteristic value of the thickness of the material layer to obtain the predicted value of the thickness of the material layer.
- the distribution control system includes a round roller feeder, a roller type distributor and a sintering cart.
- the mixed material is supplied to the roller distributor, which is used for material distribution to the sintering trolley;
- the distribution control system also includes a mixture detection mechanism, a feeding roller controller connected to the round roller feeder, and a roller distributor.
- the connected distribution roller controller and auxiliary door controller, and the sintering trolley controller connected with the sintering trolley; and the mixture detection mechanism, the feeding roller controller, the distribution roller controller, the auxiliary door controller and the sintering table
- the central processing unit to which the vehicle controller is connected.
- the sintering trolley is quantified to the same interval according to a certain shrinkage ratio, and combined with the opening of the auxiliary door, the eigenvector of the predicted material layer thickness is generated; then the eigenvector of the predicted material layer thickness is input into the pre-established material layer thickness dynamic prediction
- the characteristic value of the thickness of the material layer is generated; the data of the characteristic value of the material layer thickness is restored to obtain the predicted value of the material layer thickness, and then the material layer is calculated according to the predicted value of the material layer thickness and the target value of the material layer thickness.
- the layer thickness prediction realizes the distribution control of the sintering system.
- the material layer thickness prediction-based material distribution control system provided in the embodiment of the present application can predict the material layer thickness on the sintering trolley in advance through the collected mixture parameters and sintering system state parameters, so that the predicted material layer thickness and real-time Adjust the relevant key parameters that affect the cloth, and realize the timely and stable control of the cloth process of the sintering system.
- FIG. 1 is a schematic structural diagram of a distribution control system based on material layer thickness prediction provided by an embodiment of the application;
- Fig. 2 is the working flow chart of the cloth control system based on material layer thickness prediction provided by the embodiment of the application;
- FIG. 3 is a flow chart of predicting the thickness of the material layer by the cloth control system provided by the embodiment of the present application.
- Fig. 4 is the flow chart of the online update material layer thickness dynamic prediction model provided by the embodiment of this application.
- FIG. 5 is a flow chart of generating a dynamic prediction model of material layer thickness provided by an embodiment of the present application.
- FIG. 6 is a schematic structural diagram of an LSTM neural network model provided by an embodiment of the present application.
- FIG. 7 is another flowchart of generating a dynamic prediction model of layer thickness according to an embodiment of the present application.
- 1- batching room 2- mixer, 3- round roller feeder, 4- roller distributor, 5- sintering trolley, 6- ignition fan, 7- ignition fan, 8- single roller crusher, 9-ring cooler, 10-material layer thickness detection device, 101-mixing material detection mechanism, 102-feeding roller controller, 103-distributing roller controller, 104-auxiliary door controller, 105-sintering trolley controller , 106-CPU.
- the sintering system mainly includes a sintering trolley, a mixer, a main shaft fan, a ring cooler and other equipment.
- the overall process flow chart is shown in Figure 1: various raw materials are proportioned in the batching chamber 1 to form a mixture. After entering the mixer 2 for mixing and pelletizing, it is evenly distributed on the sintering trolley 5 through the round roller feeder 3 and the roller distributor 4 to form a mixed material layer, and the ignition fan 6 and the ignition fan 7 start mixing. The material is ignited to start the sintering process. After the sintering is completed, the obtained sintered ore is crushed by the single-roller crusher 8 and then enters the ring cooler 9 for cooling, and finally is screened and granulated before being sent to the blast furnace or the finished product silo.
- FIG. 1 it is a schematic structural diagram of a distribution control system based on material layer thickness prediction provided in an embodiment of the present application.
- the distribution control system includes a round roller feeder 3, a roller type distributor 4 and a sintering trolley 5.
- the round roller feeder 3 is used to supply the mixed material to the roller type distributor 4.
- the distribution machine 4 is used to distribute material to the sintering cart 5 .
- the material layer thickness prediction system also includes a mixture detection mechanism 101, the mixture detection mechanism 101 includes a sampling device and an offline detection device; the sampling device is used to obtain the pelletized mixture from the sintering system, and to The obtained mixture is input to the offline detection device, and the offline detection device is used to measure the density of each component of the test sample and the bulk density of the mixture, and the mixture detection mechanism 101 executes step S201, and the obtained mixture is The bulk density is sent to the central processing unit 106 .
- the mixture detection mechanism 101 includes a sampling device and an offline detection device; the sampling device is used to obtain the pelletized mixture from the sintering system, and to The obtained mixture is input to the offline detection device, and the offline detection device is used to measure the density of each component of the test sample and the bulk density of the mixture, and the mixture detection mechanism 101 executes step S201, and the obtained mixture is The bulk density is sent to the central processing unit 106 .
- Feeding roller controller 102 the feeding roller controller 102 is connected with the round roller feeder 3, used to control the feeding roller rotation speed of the round roller feeder 3, and can obtain the round roller feeding
- the rotation speed of the feeding roller of machine 3 for example, the rotation speed sensor is used to measure the rotation speed of the feeding roller.
- the rotation speed sensor is a sensor that converts the rotation speed of the rotating object into the power output. Install an encoder, or use a proximity switch combined with a high-speed counter for digital-to-analog conversion speed measurement.
- the feeding roller controller 102 is further configured to execute step S202, and send the obtained rotational speed of the feeding roller to the central processing unit 106.
- FIG. 2 it is a working flow chart of the cloth control system based on material layer thickness prediction provided by the embodiment of the present application.
- the distribution roller controller 103 the distribution roller controller 103 is connected with the roller distributor 4, is used to control the distribution roller rotation speed of the roller distributor 4, and can obtain the distribution roller rotation speed of the roller distributor 4, so
- the distribution roller controller 103 is further configured to execute step S203, and send the acquired rotation speed of the distribution roller to the central processing unit 106.
- the auxiliary door controller 104 is arranged above the round roller feeder 3, used to control the opening degree of the auxiliary door of the roller distributor 4, and can obtain the auxiliary door of the roller distributor 4 door opening degree, and is configured to execute step S204 , and send the acquired auxiliary door opening degree to the central processing unit 106 .
- auxiliary door controllers 104 there are at least four auxiliary doors under each silo, the number of auxiliary door controllers 104 used in the embodiment of the present application is the same as the number of auxiliary doors, and one auxiliary door controller 104 controls the opening of one auxiliary door.
- the sintering trolley controller 105 which is arranged on the sintering trolley 5, is used to control the running speed of the sintering trolley 5, and can obtain the running speed of the sintering trolley 5, and is configured to execute the steps 205. Send the acquired speed of the sintering trolley to the central processing unit 106.
- the central processing unit 106 is connected to the mixture detection mechanism 101 , the feeding roller controller 102 , the distributing roller controller 103 , the auxiliary door controller 104 and the sintering trolley controller 105 .
- steps S201 to S205 are not in order, and may be performed in any order or simultaneously.
- the central processing unit 106 is configured to perform the following steps S206 to S213.
- Step S206 receiving the bulk density of the mixture, the rotational speed of the feeding roller, the rotational speed of the distribution roller, the opening of the auxiliary door and the speed of the sintering trolley.
- Step S207 Predict the thickness of the material layer according to the bulk density of the mixture, the rotational speed of the feeding roller, the rotational speed of the distribution roller, the opening of the auxiliary door and the speed of the sintering trolley, and obtain the predicted value of the thickness of the material layer.
- Step S301 quantify the bulk density of the mixture, the rotational speed of the feeding roller, the rotational speed of the distribution roller, and the speed of the sintering trolley at the same time into the same interval according to a certain shrinkage ratio, and combine with the opening of the auxiliary door to generate a feature for predicting the thickness of the material layer vector.
- the bulk density, the rotation speed of the feeding roller, the rotation speed of the distribution roller and the speed of the sintering trolley at the same time are quantified into the same interval according to a certain shrinkage ratio, and the same interval after quantification is the interval (0,1).
- the quantitative model of bulk density is:
- the quantitative model of the feed roller speed is:
- the quantitative model of the rotation speed of the distribution roller is:
- the quantitative model of the speed of the sintering trolley is:
- Norm( ⁇ ) represents the quantized bulk density, ⁇ represents the bulk density, Represents the density of the component with the highest density among the components;
- Norm(n 1 ) represents the quantified feeding roller speed, n 1 represents the feeding roller speed, Represents the rated speed of the feeding roller;
- Norm(n 2 ) represents the quantized distribution roller speed, n 2 represents the distribution roller speed, Represents the rated speed of the distribution roller;
- Norm(s) represents the quantized sintering trolley speed, v represents the sintering trolley speed, Indicates the maximum speed of the sintering trolley.
- the feature vector for predicting the thickness of the material layer is to integrate the influencing factors of the thickness of the material layer according to a certain rule, for example:
- X(k) (x1(k), x2 (k), x3 (k), x4 (k), x5 (k))
- X(k) represents the feature vector of the predicted material layer thickness
- x 1 (k), x 2 (k), x 3 (k), x 4 (k), x 5 (k) represent the bulk density, Feed roller speed, distribution roller speed, auxiliary door opening and sintering trolley speed.
- Step S302 input the eigenvector of predicting the thickness of the material layer into the pre-established dynamic prediction model of the material layer thickness, and generate the characteristic value of the material layer thickness, and the dynamic prediction model of the material layer thickness includes the eigenvector of the predicted material layer thickness and the The mapping relationship between the eigenvalues of the layer thickness.
- the mapping relationship includes the eigenvector of the predicted material layer thickness and the corresponding eigenvalue of the material layer thickness, namely:
- y( k ) is the eigenvalue that affects the thickness of the material layer
- fk is the mapping relationship between the eigenvector of the predicted material layer thickness and the eigenvalue of the material layer thickness.
- Step S303 performing data restoration on the characteristic value of the material layer thickness to obtain a predicted value of the material layer thickness.
- the characteristic value of the material layer thickness output by the dynamic prediction model of material layer thickness certain data processing is required to obtain the final predicted value of the material layer thickness.
- the specific operation is to calculate the characteristic value of the material layer thickness and sintering.
- the product of the maximum allowable thickness of the material layer of the trolley is obtained to obtain the predicted value of the material layer thickness.
- hi is the obtained predicted value of the thickness of the material layer at time i
- ki is the generated characteristic value of the thickness of the material layer at time i
- H is the maximum allowable thickness of the material layer on the sintering trolley.
- Step S208 Calculate the deviation value of the thickness of the material layer according to the predicted value of the thickness of the material layer and the target value of the thickness of the material layer.
- Y(k) refers to the predicted value sequence of the layer thickness obtained at time k
- R(k) is the set target value sequence of the layer thickness at time k
- E(k) is the thickness deviation of the layer at time k sequence of values.
- Step S209 input the thickness deviation value of the material layer into the rolling optimization model, and obtain the speed of the feed roller to be adjusted, the speed of the distribution roller to be adjusted, the opening of the auxiliary door to be adjusted and the speed of the sintering trolley to be adjusted.
- the rolling optimization model is used to calculate the corresponding feed roller speed, distribution roller speed, auxiliary door opening when the variance of the thickness deviation value of the material layer is the smallest under the condition that the bulk density of the mixture is unchanged. and sintering trolley speed.
- the variance of the material layer thickness deviation value is obtained in the following manner:
- ⁇ is the variance of the material layer thickness deviation
- E(k) is the material layer thickness deviation value at time k
- R(k) is the preset target value sequence of material layer thickness at time k
- Y(k) is the material layer thickness at time k. Sequence of predicted values for layer thickness.
- the central processing unit 106 obtains the bulk density of the mixture, the rotational speed of the feeding roller, the rotational speed of the distribution roller, the opening of the auxiliary door and the speed of the sintering trolley, and predicts the Sequence of predicted values for layer thickness at time k:
- Y( k ) fk(x1(k), x2 (k), x3 (k), x4 (k), x5 (k))
- f k represents the mapping relationship between the eigenvectors of the predicted layer thickness at the kth moment and the eigenvalues of the layer thickness; at the kth moment, the sequence of predicted deviation values is expressed as follows:
- the variance of the thickness deviation value of the material layer is expressed as follows:
- Step S210 sending the speed of the sintering trolley to be adjusted.
- the sintering trolley controller 105 is driven to adjust the speed of the sintering trolley to the speed of the sintering trolley to be adjusted.
- Step S211 sending the opening degree of the auxiliary door to be adjusted.
- the auxiliary door controller 104 is driven to adjust the opening degree of the auxiliary door to the opening degree of the auxiliary door to be adjusted.
- Step S212 sending the rotation speed of the cloth roller to be adjusted.
- the distribution roller controller 103 is driven by sending the rotation speed of the distribution roller to be adjusted to the distribution roller controller 103 .
- Step S213 sending the rotation speed of the feeding roller to be adjusted.
- the feeding roller controller 102 is driven by sending the rotational speed of the feeding roller to be adjusted to the feeding roller controller 102 .
- steps S210 to S213 are not in order, and may be performed in any order or simultaneously.
- the speed of the sintering trolley is adjusted to the speed of the sintering trolley to be adjusted
- the opening of the auxiliary door is adjusted to the opening of the auxiliary door to be adjusted
- the speed of the distribution roller is adjusted to the speed of the distribution roller to be adjusted
- the feeding The adjustment of the roller speed to the speed of the feed roller to be adjusted must be done at the same time.
- the central processing unit 106 is connected to the mixture detection mechanism 101 , the feeding roller controller 102 , the distributing roller controller 103 , the auxiliary door controller 104 and the sintering trolley controller 105 .
- the embodiment of the present application provides a distribution control system based on the prediction of material layer thickness. and the speed of the sintering trolley; then quantify the bulk density, feeding roller speed, distribution roller speed and sintering trolley speed at the same time to the same interval according to a certain shrinkage ratio, and combine the auxiliary door opening to generate the predicted layer thickness Then input the eigenvector of predicting the thickness of the material layer into the pre-established dynamic prediction model of the material layer thickness to generate the characteristic value of the material layer thickness; and restore the data of the characteristic value of the material layer thickness to obtain the material layer thickness.
- the predicted value of the thickness then according to the predicted value of the thickness of the material layer and the target value of the thickness of the material layer, the deviation value of the material layer thickness is calculated, and finally the deviation value of the material layer thickness is input into the rolling optimization model, and the speed of the feeding roller to be adjusted is obtained.
- the material layer thickness prediction-based material distribution control system provided in the embodiment of the present application can predict the material layer thickness on the sintering trolley in advance through the collected mixture parameters and the sintering system state parameters, so that the predicted material layer thickness can be realized. Precise cloth control of the sintering system.
- the parameters of the mixture refer to the bulk density of the mixture
- the state parameters of the sintering system refer to the rotation speed of the feeding roller, the rotation speed of the distribution roller, the opening of the auxiliary door and the speed of the sintering trolley.
- the thickness of the material layer of the sintering trolley there are many factors that affect the thickness of the material layer of the sintering trolley, such as the opening of the main door, the inclination of the feeding roller, the width of the sintering trolley and the number of auxiliary doors, etc., but these factors are generally fixed values, such as the width of the sintering trolley and the number of auxiliary doors ; or there are changes, but there will not be frequent changes, which are stable values for a relatively long time, such as the opening of the main door and the inclination of the feeding roller, so the cloth control system based on the prediction of the thickness of the material layer in the embodiment of the present application does not use these values.
- the influencing factors are taken into consideration, that is, these influencing factors have not changed in the actual application process and in the establishment of the dynamic prediction model of material layer thickness.
- the material layer thickness prediction system further includes a material layer thickness detection device 10.
- the layer thickness detection device The material layer thickness on the sintering trolley can be detected in real time.
- the material layer thickness detection device 10 is arranged on the upper position of the sintering trolley to detect the material layer thickness of the sintering trolley, for example, a material level gauge is used to detect the material layer thickness.
- a method for online updating a dynamic prediction model of layer thickness is provided.
- step S302 the feature vector of the predicted material layer thickness is input into the pre-established material layer thickness dynamic prediction model, and before the step of generating the characteristic value of the material layer thickness, the method further includes:
- Step S401 acquiring learning samples in time series adjacent to the feature vector of the predicted material layer thickness, wherein the learning samples include input samples and feature values of the material layer thickness corresponding to the input samples.
- the learning sample includes not only the mixture parameters of the sintering system and the state parameters of the sintering system, but also the corresponding measured value of the thickness of the material layer.
- the specific acquisition method of the learning sample is: acquiring and predicting the characteristics of the thickness of the material layer.
- the normalization process adopts the following models:
- hi is the obtained predicted value of the material layer thickness at time i
- ki is the generated characteristic value of the material layer thickness at time i
- H is the maximum allowable thickness of the material layer on the sintering trolley.
- the historical prediction samples collected here are used as learning samples. Since the sintering process of the sintering system is a long-term continuous process, the sintering system has been performing mixed operations before collecting the feature vector for predicting the thickness of the material layer.
- the historical prediction samples within the nearest preset time interval of the eigenvectors for predicting the thickness of the material layer can guarantee the working conditions of the sintering system corresponding to the obtained learning samples and the working conditions of the sintering system at the time when the eigenvectors for predicting the thickness of the material layers are collected. Consistent.
- Step S402 using the learning sample to update the material layer thickness dynamic prediction model online, and obtain the updated material layer thickness dynamic prediction model.
- the dynamic prediction model of the material layer thickness can be updated online, which further ensures that the characteristic value of the material layer thickness output by the material layer thickness dynamic prediction model is more accurate.
- the prediction deviation value between the eigenvalues of the material layer thickness corresponding to the measured value is determined.
- the dynamic prediction model has two update methods. The deviation value is small and relatively stable, and is within the allowable error range of the model quality index, then the predicted deviation value is directly added to the characteristic value of the material layer thickness of the dynamic prediction model, and the result is used as the updated material layer thickness. Eigenvalues.
- the mapping included in the dynamic prediction model of the material layer thickness is updated according to the learning sample and the characteristic value of the material layer thickness corresponding to the measured value. relation.
- the mean square error of the predicted deviation value can be generally used as the quality index, and then according to the statistical distribution law of the quality index, a statistical confidence limit is preset to determine whether an update needs to be triggered and the required update method.
- the characteristic value of the thickness of the material layer corresponding to the measured value although it is not suitable for the closed-loop control problem of the hybrid process due to the hysteresis problem, it can be used as a reference for the steady state, that is, when the sintering system reaches a steady state in a stable state, the thickness of the material layer will be maintained. At a certain level, under normal circumstances, the measured value fluctuates, but its distribution usually does not deviate from the confidence interval.
- the model update mechanism is triggered. If it is judged that the process characteristics belong to the gradual change according to the index analysis results, the model recursion method is selected, and the moving window recursion method is used to update the dynamic prediction model of the material layer thickness. The steps are as follows:
- a new measurement value [X m , Y m ] is obtained, it is added to the sample set, and the oldest sample is eliminated, the new sample set (learning sample) is:
- the dynamic prediction model of the material layer thickness is updated online by using the learning sample to obtain a new dynamic prediction model of the material layer thickness. If it is judged that the process characteristics belong to sudden changes according to the index analysis results, the real-time learning method is selected to reconstruct the dynamic prediction model of material layer thickness.
- the material layer thickness dynamic prediction model is established by using multiple sets of known mixture parameters, sintering system state parameters and corresponding material layer thickness measurement values.
- a method for training a neural network model is provided to generate the dynamic prediction model of the material layer thickness.
- the specific operation is to use multiple sets of known mixture parameters and sintering system state parameters as the input of the neural network model, and use the corresponding material layer thickness measurement value as the output of the neural network model to train the neural network model, that is, adjust the neural network model.
- the weight matrix and bias term corresponding to the middle layer of the network model are used to establish the mapping relationship between the mixture parameters and the corresponding mixer state parameters and the degree of mixing. Referring to FIG. 5 , a flow chart of generating a dynamic prediction model of material layer thickness provided by an embodiment of the present application, and the specific generating steps include:
- step S501 the bulk density, the rotational speed of the feeding roller, the rotational speed of the distribution roller, the opening of the auxiliary door and the speed of the sintering trolley of the N groups of independent mixtures are acquired.
- the N groups of independent mixture parameters and the corresponding state parameters of the sintering system can be the data of the same sintering system or the data of multiple sintering systems, and are divided into a group according to the corresponding relationship, that is, the same mixer and The data at the same time is regarded as a unified group of data.
- Step S502 quantify the N groups of independent bulk density, feed roller rotational speed, distribution roller rotational speed and sintering trolley speed into the same interval according to a certain shrinkage ratio, and combine with the auxiliary door opening to generate input of N groups of training samples.
- the bulk density in the parameters of the mixture, and the speed of the feed roller, the speed of the distribution roller and the speed of the sintering trolley in the state parameters of the sintering system need to be quantified to the interval (0,1), where the value of the auxiliary door opening is percentage data.
- Step S503 Obtain N groups of independent layer thickness measurement values sent by the material layer thickness detection device 10, and quantify the material layer thickness measurement values according to a certain shrinkage ratio to generate output of N groups of training samples.
- the measured value of the thickness of the material layer is quantified according to a certain shrinkage ratio, and the specific operation is to calculate the ratio of the measured value of the material layer thickness to the maximum allowable thickness of the material layer of the sintering trolley.
- Step S504 using the input of the training sample and the output of the training sample, the neural network model is trained by the time back-propagation method.
- the dynamic prediction training module uses the input of the training sample and the output of the training sample to train the neural network model by the time backpropagation method; wherein, the time backpropagation method is a learning algorithm suitable for multi-layer neuron networks, Excitation propagation and weight update are repeated in a loop to guide the response (output) of the multi-layer neuron network to the input until it reaches a predetermined target range.
- the time backpropagation method is a learning algorithm suitable for multi-layer neuron networks, Excitation propagation and weight update are repeated in a loop to guide the response (output) of the multi-layer neuron network to the input until it reaches a predetermined target range.
- Step S505 the weight parameters, bias parameters and learning factors of the neural network model are continuously updated through iterative training.
- Step S506 if the eigenvalues of the predicted material layer thickness of the neural network model and the measured eigenvalues of the material layer thickness reach the set tolerance range, or the neural network model reaches the set maximum number of iterations, the training ends, and Save the last updated weight parameters, bias parameters and learning factors to obtain a dynamic prediction model of layer thickness.
- a mapping relationship between the eigenvectors of the predicted layer thickness and the eigenvalues of the layer thickness is established in the neural network model, until the predicted layer thickness according to the mapping relationship is established.
- the eigenvalues of thickness meet production requirements.
- the specific judgment method is to judge whether the eigenvalues of the predicted material layer thickness and the detected eigenvalues of the material layer thickness of the neural network model reach the set tolerance range, and judge whether the neural network model reaches the set maximum number of iterations , if the eigenvalues of the predicted layer thickness of the neural network model and the detected eigenvalues of the layer thickness reach the set tolerance range, or the neural network model reaches the set maximum number of iterations, the training ends and saves the final
- the updated weight parameters, bias parameters and learning factors are used to obtain the dynamic prediction model of material layer thickness.
- the embodiment of the present application provides the neural network model training to generate the dynamic prediction model for the thickness of the material layer.
- RNN Recurrent Neural Network
- LSTM Long Short-Term Memory
- FIG. 6 it is a schematic diagram of the structure based on the LSTM neural network model.
- the structure of the middle layer mainly includes forget gate,
- the input gate and the output gate are composed of input gates and output gates, each gate has a corresponding weight matrix and bias term.
- the multi-layer neuron network is trained by training samples, and the weight parameters, bias parameters and learning factors are continuously updated to obtain the data.
- Layer thickness dynamic prediction model is trained by training samples, and the weight parameters, bias parameters and learning factors are continuously updated to obtain the data.
- the neural network model divides multiple groups of training samples into two parts, including 2/3 groups of training data and 1/3 groups of test data.
- the input of 3 sets of training samples and the output of training samples are used as the training data of the neural network model, and the weight parameters, bias parameters and learning factors are continuously updated; and the input of 1/3 sets of training samples and the output of training samples are used as neural network Tolerance test data for the network model.
- the dynamic prediction model of material layer thickness can also be a mixing degree prediction table, and established according to the following steps:
- step S701 the bulk density, the rotational speed of the feeding roller, the rotational speed of the distribution roller, the opening degree of the auxiliary door and the speed of the sintering trolley of the N groups of independent mixtures are obtained.
- Step S702 acquiring N groups of independent material layer thickness measurement values sent by the material layer thickness detection device.
- Step S703 Statistical analysis is performed on the bulk density of the N groups of independent mixtures, the rotational speed of the feeding roller, the rotational speed of the distribution roller, the opening of the auxiliary door, the speed of the sintering trolley, and the measured values of the corresponding material layer thickness, and a layer thickness prediction is established. surface.
- layer thickness prediction table and the material layer thickness prediction table mentioned in this application have the same meaning, and all refer to the material layer thickness prediction table.
- the statistical analysis includes data preprocessing.
- the data preprocessing first quantifies the bulk density, the speed of the feeding roller, the speed of the distribution roller and the speed of the sintering trolley to the same interval according to a certain shrinkage ratio, and then quantifies the parameters of the mixture and the state of the sintering system.
- the same multiple sets of data are extracted, and the measured values of the corresponding material layer thicknesses in different groups are compared. If the corresponding measured values of the material layer thickness are also the same, only one set of data is retained, and the same data in other groups are removed.
- the average mixing degree corresponding to multiple sets of data is used as the corresponding material layer thickness, and only one set of data is retained.
- the range is set to 0.5%-2%.
- the corresponding measurement values of mixing degree are different, and the deviation exceeds the allowable range, and the source of multiple sets of data is marked and kept in the isolation area.
- the data source here refers to the data Collect the corresponding sintering system and collection time
- the isolation area refers to a separate area in the layer thickness prediction table, which is used to record abnormal data as reference data for equipment maintenance.
- the measured value of the material layer thickness here also needs to be preprocessed.
- the specific operation is to calculate the ratio of the measured value of the material layer thickness to the maximum allowable thickness of the material layer of the sintering trolley.
- sintering system state parameters and corresponding material layer thickness measurement values that have been preprocessed, they are sorted according to the set multiple index items, and each index item is sorted from small to large. Among them, at least 5 index items are included.
- the first-level index item is the bulk density of the mixture
- the second-level index item is the rotation speed of the feeding roller.
- the third-level index item is the speed of the cloth roller
- the fourth-level index item is the opening of the auxiliary door
- the fifth-level index item is the speed of the sintering trolley.
- the establishment of the mixing degree prediction table is completed according to the above rules.
- the mixing degree prediction table includes the mapping relationship between the feature vector of the predicted material layer thickness and the eigenvalue of the material layer thickness. For example, as shown in Table 1, it is the mixing degree prediction part of the table data.
- An embodiment of the present application provides a method for controlling distribution based on prediction of material layer thickness, and the method for controlling distribution includes:
- the thickness of the material layer is predicted, and the predicted value of the thickness of the material layer is obtained.
- the step of predicting the thickness of the material layer according to the bulk density of the mixture, the rotational speed of the feeding roller, the rotational speed of the distribution roller, the opening of the auxiliary door and the speed of the sintering trolley, and obtaining the predicted value of the thickness of the material layer includes:
- the material layer thickness dynamic prediction model includes the predicted material layer thickness eigenvector and material layer thickness The mapping relationship between the eigenvalues;
- Data reduction is performed on the characteristic value of the thickness of the material layer to obtain the predicted value of the thickness of the material layer.
- the deviation value of the layer thickness is calculated.
- the rolling optimization model is used to calculate the corresponding feed roller speed, distribution roller speed, auxiliary door opening when the variance of the thickness deviation value of the material layer is the smallest under the condition that the bulk density of the mixture is unchanged. and sintering trolley speed.
- the material distribution control system includes a round roller feeder 3, a roller type material distribution machine 4 and a sintering trolley 5,
- the round roller feeder 3 is used to supply the mixed material to the roller distributor 4, and the roller distributor 4 is used to distribute the material to the sintering trolley 5;
- the material distribution control system further includes a mixture detection mechanism 101 , the feeding roller controller 102 connected with the round roller feeder 3, the distribution roller controller 103 and the auxiliary door controller 104 connected with the roller distributor 4, and the sintering trolley controller connected with the sintering trolley 5 105 ; and the central processing unit 106 connected with the mixture detection mechanism 101 , the feeding roller controller 102 , the distribution roller controller 103 , the auxiliary door controller 104 and the sintering trolley controller 105 .
- the opening of the auxiliary door and the speed of the sintering trolley are quantified to the same interval according to a certain shrinkage ratio, and the eigenvector of the predicted layer thickness is generated; then the eigenvector of the predicted layer thickness is input into the pre-established dynamic prediction model of the layer thickness
- the characteristic value of the material layer thickness is generated; and the data of the characteristic value of the material layer thickness is restored to obtain the predicted value of the material layer thickness, and then the material layer thickness is calculated according to the predicted value of the material layer thickness and the target value of the material layer thickness.
- Layer thickness prediction enables cloth control of the sintering system.
- the material layer thickness prediction-based material distribution control system provided by the embodiment of the present application can predict the material layer thickness on the sintering trolley in advance through the collected mixture parameters and the sintering system state parameters, so that the predicted material layer thickness can be realized. Timely and stable feedback cloth control for sintering system.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Mechanical Engineering (AREA)
- Manufacture And Refinement Of Metals (AREA)
Abstract
本申请涉及钢铁冶炼技术领域，提供了一种基于料层厚度预测的布料控制系统及方法，在实际应用过程中，首先获取混合料的堆密度、给料辊转速、布料辊转速、辅门开度以及烧结台车速度；并利用预先建立的料层厚度动态预测模型，生成料层厚度的特征值，通过对料层厚度特征值进行数据还原，获得料层厚度的预测值；然后根据料层厚度预测值和料层厚度目标值，计算料层厚度的偏差值；最后将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度，从而实现基于料层厚度预测的烧结系统布料控制，以实现对烧结系统布料过程的精准控制。
Description
本申请涉及钢铁冶炼技术领域，尤其涉及一种基于料层厚度预测的布料控制系统及方法。
烧结是指将粉状物料转变为致密体的工艺，在钢铁生产工艺中，则是指将铁矿粉、煤粉和石灰按一定配比混匀，通过烧结得到具有高炉生产工艺要求的烧结矿，为高炉炼铁输送原料。针对烧结生产工艺流程中的布料过程，料层厚度是否合适是影响烧结生产产量与质量的重要参数之一，目前，布料过程中通常是利用调节辅门开度大小来调节与辅门对应区域物料的层厚，或利用调节圆辊给料机转速，对台车宽度范围料层进行整体改变。若辅门开度调节大，则对应的布料装置向烧结台车布料的流量就大，相应的料层的厚度就厚，反之，料层厚度小。
为了获得厚度合适的料层，现有技术中，一般通过实时检测烧结台车的料层厚度，进行布料的调控。具体过程为，在烧结台车布料过程中，在烧结台车运动方向的下游，监测烧结台车的料层厚度，并根据检测的料层厚度，进行布料的反馈控制。
在实际运行中，这种料层厚度检测方法，检测的是布料已经完成的烧结台车料层厚度，从辅门调节点到层厚检测点约2到3分钟，采用反馈控制滞后性较大。如果此时发现料层厚度不合适，已经无法改变，不利于实现对烧结系统及时稳定控制布料。
发明内容
本申请提供了一种基于料层厚度预测的布料控制系统及方法，以解决现有技术中，通过检测烧结台车的料层厚度，对烧结系统的布料控制存在滞后性的问题。
本申请第一方面提供一种基于料层厚度预测的布料控制系统，所述布料控制系统包括圆辊给料机、辊式布料机和烧结台车，所述圆辊给料机用于向所述辊式布料机供给混合料，所述辊式布料机用于向烧结台车布料；所述布料控制系统还包括混合料检测机构、与圆辊给料机连接的给料辊控制器、与辊式布料机连接的布料辊控制器和辅门控制器，以及与烧结台车连接的烧结台车控制器；以及与混合料检测机构、给料辊控制器、布料辊控制器、辅门控制器和烧结台车控制器连接的中央处理器；其中：
所述中央处理器被配置为执行以下步骤：
接收所述混合料检测机构发送的混合料堆密度，接收给料辊控制器发送的给料辊转速，接收布料辊控制器发送的布料辊转速，接收辅门控制器发送的辅门开度，接收烧结台车控制器发送的烧结台车速度；
根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值；
根据料层厚度的预测值和料层厚度的目标值，计算料层厚度偏差值；
将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转 速、待调整的辅门开度和待调整的烧结台车速度；
驱动所述给料辊控制器将给料辊转速调整为待调整的给料辊转速；驱动所述布料辊控制器将布料辊转速调整为待调整的布料辊转速；驱动所述辅门控制器将辅门开度调整为待调整的辅门开度；驱动所述烧结台车控制器将烧结台车速度调整为待调整的烧结台车速度；
其中，所述滚动优化模型用于在所述混合料的堆密度不变的条件下，计算出料层厚度偏差值的方差最小时，对应的给料辊转速、布料辊转速、辅门开度和烧结台车速度。
可选的，所述料层厚度偏差值的方差通过以下方式得到：
其中，σ是料层厚度偏差值的方差，E(k)是指料层厚度偏差值，R(k)预设的料层厚度的目标值，Y(k)料层厚度的预测值。
可选的，根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值，具体执行以下步骤：
将同一时刻的混合料的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，生成预测料层厚度的特征向量；
将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值，所述料层厚度动态预测模型中包含预测料层厚度的特征向量与料层厚度的特征值之间的映射关系；
对料层厚度的特征值进行数据还原，获得料层厚度的预测值。
可选的，所述布料控制系统系统还包括设置在烧结台车上方的料层厚度检测装置，所述料层厚度检测装置连接中央处理器，所述料层厚度检测装置用于按照预先设定的时间间隔，检测烧结台车的料层厚度，并获得料层厚度的测量值；将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值，还包括：
获取与所述预测料层厚度的特征向量邻近时序上的数据并作为样本，其中，所述样本包括输入样本，以及所述输入样本对应的料层厚度的特征值；
利用所述学习样本在线更新所述料层厚度动态预测模型，获取更新后的料层厚度动态预测模型。
可选的，采用以下方法对料层厚度的特征值进行数据还原：
h
i＝k
i×H
其中，h
i为获得的i时刻料层厚度的预测值，k
i为生成的i时刻料层厚度的特征值，H为烧结台车上的料层最大允许厚度。
可选的，将同一时刻的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，具体执行以下步骤：
计算堆密度与各原料中密度最大的原料的密度的比值；
计算给料辊转速与给料辊的最大转速的比值；
计算布料辊转速与布料辊的最大转速的比值；
烧结台车速度与烧结台车的最大转速的比值。
可选的，所述料层厚度动态预测模型是基于神经网络模型训练生成，并按照以下步骤建立：
按照预先设定的时间间隔，获取N组独立的混合料的堆密度、给料辊转速、布料辊转 速、辅门开度和烧结台车速度；
将N组独立的混合料的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，作为N组训练样本的输入；
按照预先设定的时间间隔，检测烧结台车上与上述N组训练样本输入对应的实际料层厚度，并计算实际料层厚度的实际特征值，将实际特征值作为N组输出训练样本；
利用输入训练样本以及输出训练样本，采用时间反向传播法训练神经网络模型；
通过迭代训练不断更新神经网络模型的权重参数、偏置参数以及学习因子；
若神经网络模型的预测值与测量值达到设定的允差范围，或神经网络模型达到设定的最大迭代次数，则训练结束，并保存最后更新的权重参数、偏置参数以及学习因子，获得料层厚度动态预测模型。
可选的，所述料层厚度动态预测模型还可以是料层厚度预测表，并按照以下步骤建立：
按照预先设定的时间间隔，获取N组独立的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度；
按照预先设定的时间间隔，检测烧结台车上与上述N组输入训练样本对应的实际料层厚度，并计算实际料层厚度的实际特征值；
对N组独立的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度以及对应的实际特征值，进行统计分析，建立料层厚度预测表。
本申请第二方面提供一种基于料层厚度预测的布料控制方法，所述布料控制方法包括：
获取混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度；
根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值；
根据料层厚度的预测值和料层厚度的目标值，计算料层厚度偏差值；
将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度；
将给料辊转速调整为待调整的给料辊转速；将布料辊转速调整为待调整的布料辊转速；将辅门开度调整为待调整的辅门开度；将烧结台车速度调整为待调整的烧结台车速度；
其中，所述滚动优化模型用于在所述混合料的堆密度不变的条件下，计算出料层厚度偏差值的方差最小时，对应的给料辊转速、布料辊转速、辅门开度和烧结台车速度。
可选的，所述根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值的步骤，包括：
将同一时刻的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，生成预测料层厚度的特征向量；
将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值，所述料层厚度动态预测模型中包含预测料层厚度的特征向量与料层厚度的特征值之间的映射关系；
对料层厚度的特征值进行数据还原，获得料层厚度的预测值。
由上述技术方案可知，本申请提供的一种基于料层厚度预测的布料控制系统及方法，布料控制系统包括圆辊给料机、辊式布料机和烧结台车，圆辊给料机用于向辊式布料机供给混合料，辊式布料机用于向烧结台车布料；布料控制系统还包括混合料检测机构、与圆 辊给料机连接的给料辊控制器、与辊式布料机连接的布料辊控制器和辅门控制器，以及与烧结台车连接的烧结台车控制器；以及与混合料检测机构、给料辊控制器、布料辊控制器、辅门控制器和烧结台车控制器连接的中央处理器。
在实际应用过程中，首先获取混合料的堆密度、给料辊转速、布料辊转速、辅门开度以及烧结台车速度；然后将同一时刻的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，生成预测料层厚度的特征向量；再然后将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值；并对料层厚度的特征值进行数据还原，获得料层厚度的预测值，之后根据料层厚度的预测值和料层厚度的目标值，计算料层厚度偏差值，最后将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度，从而基于料层厚度预测实现烧结系统的布料控制。本申请实施例提供的基于料层厚度预测的布料控制系统，可以通过采集的混合料参数以及烧结系统状态参数，提前预测出烧结台车上的料层厚度，以便通过预测的料层厚度及实时调节相关影响布料的关键参数，实现对烧结系统的布料过程及时、稳定的控制。
为了更清楚地说明本申请的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，显而易见地，对于本领域普通技术人员而言，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1为本申请实施例提供的一种基于料层厚度预测的布料控制系统的结构示意图；
图2为本申请实施例提供的基于料层厚度预测的布料控制系统的工作流程图；
图3为本申请实施例提供的布料控制系统预测料层厚度的流程图；
图4为本申请实施例提供的在线更新料层厚度动态预测模型的流程图；
图5为本申请实施例提供的一种生成料层厚度动态预测模型的流程图；
图6为本申请实施例提供的LSTM神经网络模型的结构示意图；
图7为本申请实施例提供的另一种生成料层厚度动态预测模型的流程图。
图示说明：
其中，1-配料室，2-混合机，3-圆辊给料机，4-辊式布料机，5-烧结台车，6-点火风机，7-引火风机，8-单辊破碎机，9-环冷机，10-料层厚度检测装置，101-混合料检测机构，102-给料辊控制器，103-布料辊控制器，104-辅门控制器，105-烧结台车控制器，106-中央处理器。
下面将详细地对实施例进行说明，其示例表示在附图中。下面的描述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下实施例中描述的实施方式并不代表与本申请相一致的所有实施方式。仅是与权利要求书中所详述的、本申请的一些方面相一致的系统和方法的示例。
烧结系统主要包括烧结台车、混合机、主轴风机、环冷机等多个设备，其总的工艺流程图参见图1所示：各种原料经配料室1配比，形成混合料，混合料进入混合机2混匀和造球后，在通过圆辊给料机3和辊式布料机4将其均匀的散布在烧结台车5上形成混合料 层，点火风机6和引火风机7启动混合料点火开始烧结过程。烧结完成后，得到的烧结矿经单辊破碎机8破碎后进入环冷机9冷却，最后经筛分整粒后送至高炉或成品矿仓。
为了能够及时根据当前布料状态，获得烧结台车5料层厚度，从而根据烧结台车5料层厚度，及时实施对烧结台车布料的控制。参见图1，为本申请实施例提供的一种基于料层厚度预测的布料控制系统的结构示意图。所述布料控制系统包括圆辊给料机3、辊式布料机4和烧结台车5，所述圆辊给料机3用于向所述辊式布料机4供给混合料，所述辊式布料机4用于向烧结台车5布料。
所述料层厚度预测系统还包括混合料检测机构101，所述混合料检测机构101包括取样装置和离线检测装置；所述取样装置用于从烧结系统中获取经过造球的混合料，并将获取的混合料输入所述离线检测装置，所述离线检测装置用于测量检测样本的各组分密度以及混合料的堆密度，以及所述混合料检测机构101执行步骤S201，将获取的混合料堆密度发送给中央处理器106。
需要说明的是，本申请提及的混合料堆密度、混合料的堆密度，含义相同，指代的均为混合料的堆密度。
给料辊控制器102，所述给料辊控制器102与圆辊给料机3连接，用于控制所述圆辊给料机3的给料辊转速，且能够获取所述圆辊给料机3的给料辊转速，例如采用转速传感器测量给料辊转速，转速传感器是将旋转物体的转速转换为电量输出的传感器，例如采用磁敏式转速传感器或者激光式转速传感器，或者在旋转端安装编码器，或者采用接近开关结合高速记数器进行数模转换测速。所述给料辊控制器102还被配置为执行步骤S202，将获取的给料辊转速发送给中央处理器106。
参见图2，为本申请实施例提供的基于料层厚度预测的布料控制系统的工作流程图。
布料辊控制器103，所述布料辊控制器103与辊式布料机4连接，用于控制辊式布料机4的布料辊转速，且能够获取所述辊式布料机4的布料辊转速，所述布料辊控制器103还被配置为执行步骤S203，将获取的布料辊转速发送给中央处理器106。
辅门控制器104，所述辅门控制器104设置在所述圆辊给料机3的上方，用于控制辊式布料机4的辅门开度，且能够获取辊式布料机4的辅门开度，并被配置为执行步骤S204，将获取的辅门开度发送给所述中央处理器106。
需要说明的是，每个料仓下有至少四个辅门，本申请实施例中采用的辅门控制器104的数量与辅门数量一致，一个辅门控制器104控制一个辅门开度。
烧结台车控制器105，烧结台车控制器105设置在所述烧结台车5上，用于控制烧结台车5的运行速度，且能够获取烧结台车5的运行速度，并被配置执行步骤205，将获取的烧结台车速度发送给所述中央处理器106。
中央处理器106，所述中央处理器连接混合料检测机构101、给料辊控制器102、布料辊控制器103、辅门控制器104和烧结台车控制器105。
需要说明的是，步骤S201至步骤S205并没有先后之分，可以按照任意顺序执行，或者同时执行。
所述中央处理器106被配置执行以下步骤S206至S213。
步骤S206，接收混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度。
步骤S207，根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值。
其中，参见图3，为本申请实施例提供的布料控制系统预测料层厚度的流程图，预测过程包括以下步骤：
步骤S301，将同一时刻的混合料的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，生成预测料层厚度的特征向量。
由于同一时刻的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度的数据量及数据类型不同，无法直接对这些数据量进行运算。在本申请实施例中，将同一时刻的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，量化后的同一区间为区间(0,1)。
即，计算堆密度与各组份中密度最大的原料的密度的比值；计算给料辊转速与给料辊额定转速的比值；计算布料辊转速与布料辊额定转速的比值；计算烧结台车速度与烧结台车最大速度的比值。
其中，堆密度的量化模型为：
给料辊转速的量化模型为：
布料辊转速的量化模型为：
烧结台车速度的量化模型为：
其中，Norm(ρ)表示量化后堆密度，ρ表示堆密度，
表示各组份中密度最大的组份的密度；Norm(n
1)表示量化后的给料辊转速，n
1表示给料辊转速，
表示给料辊的额定转速；Norm(n
2)表示量化后的布料辊转速，n
2表示布料辊转速，
表示布料辊的额定转速；Norm(s)表示量化后的烧结台车速度，v表示烧结台车速度，
表示烧结台车最大速度。
所述预测料层厚度的特征向量是将料层厚度的影响因素按照一定规律整合，例如：
X(k)＝(x
1(k),x
2(k),x
3(k),x
4(k),x
5(k))
其中,X(k)表示预测料层厚度的特征向量,x
1(k)，x
2(k)，x
3(k)，x
4(k)，x
5(k),分别表示堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度。
步骤S302，将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值，所述料层厚度动态预测模型中包含预测料层厚度的特征向量与料层厚度的特征值之间的映射关系。
所述映射关系中包含了预测料层厚度的特征向量和对应的料层厚度的特征值，即：
y(k)＝f
k(X(k))
其中，y(k)为影响料层厚度的特征值，f
k为预测料层厚度的特征向量与料层厚度的特征值的映射关系。
步骤S303，对料层厚度的特征值进行数据还原，获得料层厚度的预测值。
对于所述料层厚度动态预测模型输出的料层厚度的特征值，需要进行一定的数据处理，才能够得到最终的料层厚度的预测值，具体操作是，计算料层厚度的特征值与烧结台车的料层最大允许厚度的乘积，获得料层厚度的预测值。
具体采用以下方法对料层厚度的特征值进行数据还原：
h
i＝k
i×H
其中，h
i为获得的i时刻料层厚度的预测值，k
i为生成的i时刻料层厚度的特征值，H 为烧结台车上的料层最大允许厚度。
步骤S208，根据料层厚度的预测值和料层厚度的目标值，计算料层厚度偏差值。
E(k)＝R(k)-Y(k)
其中，Y(k)是指k时刻获得的料层厚度的预测值序列，R(k)为设定的k时刻料层厚度的目标值序列，E(k)为k时刻的料层厚度偏差值序列。
步骤S209，将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度。
其中，所述滚动优化模型用于在所述混合料的堆密度不变的条件下，计算出料层厚度偏差值的方差最小时，对应的给料辊转速、布料辊转速、辅门开度和烧结台车速度。
所述料层厚度偏差值的方差通过以下方式得到：
其中，σ是料层厚度偏差值的方差，E(k)是k时刻料层厚度偏差值，R(k)为k时刻预设的料层厚度的目标值序列，Y(k)k时刻料层厚度的预测值序列。
举个例子，以第k时刻为例，中央处理器106在第k时刻，获取混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，并预测出第k时刻的料层厚度的预测值序列：
Y(k)＝f
k(x
1(k),x
2(k),x
3(k),x
4(k),x
5(k))
f
k表示第k时刻的预测料层厚度的特征向量与料层厚度的特征值之间的映射关系；在第k时刻，预测偏差值序列表示如下：
E(k)＝R(k)-Y(k)
在第k时刻，料层厚度偏差值的方差表示如下：
将Y(k)＝f
k(x
1(k),x
2(k),x
3(k),x
4(k),x
5(k))代入上述方差公式中，在所述混合料的堆密度不变的条件下，可以通过滚动优化模型，计算出σ最小时，对应的待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度。
以上是对第k时刻内，给料辊转速、布料辊转速、辅门开度和烧结台车速度进行控制的过程，对于第k+1时刻、……、第k+j时刻的控制过程也是类似的，此处不再赘述。
在所述中央处理器106得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度后，将执行以下步骤S210至步骤S213。
步骤S210，发送待调整的烧结台车速度。通过发送待调整的烧结台车速度给所述烧结台车控制器105，以驱动所述烧结台车控制器105将烧结台车的速度调整为待调整的烧结台车速度。
步骤S211，发送待调整的辅门开度。通过发送待调整的辅门开度给辅门控制器104，以驱动所述辅门控制器104将辅门开度调整为待调整的辅门开度。
步骤S212，发送待调整的布料辊转速。通过发送待调整的布料辊转速给布料辊控制器103，以驱动所述布料辊控制器103。
步骤S213，发送待调整的给料辊转速。通过发送待调整的给料辊转速至给料辊控制器102，以驱动所述给料辊控制器102。
需要说明的是，步骤S210至步骤S213并没有先后之分，可以按照任意顺序执行，或 者同时执行。但是，将烧结台车的速度调整为待调整的烧结台车速度，将辅门开度调整为待调整的辅门开度，将布料辊转速调整为待调整的布料辊转速，以及将给料辊转速调整为待调整的给料辊转速，必须在同一时刻进行。
中央处理器106，所述中央处理器连接混合料检测机构101、给料辊控制器102、布料辊控制器103、辅门控制器104和烧结台车控制器105。
由以上技术方案可知，本申请实施例提供了一种基于料层厚度预测的布料控制系统，在实际应用过程中，首先获取混合料的堆密度、给料辊转速、布料辊转速、辅门开度以及烧结台车速度；然后将同一时刻的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，生成预测料层厚度的特征向量；再然后将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值；并对料层厚度的特征值进行数据还原，获得料层厚度的预测值，之后根据料层厚度的预测值和料层厚度的目标值，计算料层厚度偏差值，最后将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度，从而基于料层厚度预测实现烧结系统的布料控制。本申请实施例提供的基于料层厚度预测的布料控制系统，可以通过采集的混合料参数以及烧结系统状态参数，提前预测出烧结台车上的料层厚度，以便通过预测的料层厚度，实现对烧结系统的精准布料控制。
需要说明的是，混合料参数是指混合料的堆密度，烧结系统状态参数是指给料辊转速、布料辊转速、辅门开度和烧结台车速度。影响烧结台车的料层厚度的因素很多，例如主门开度、给料辊倾角、烧结台车宽度和辅门数量等，但是这些因素一般属于定值，例如烧结台车宽度和辅门数量；或者有变化，但是不会出现频繁变化，属于比较长时间的稳定值，例如主门开度和给料辊倾角，所以本申请实施例的基于料层厚度预测的布料控制系统，没有将这些影响因素纳入考虑的范围，即这些影响因素在实际应用过程中和在建立料层厚度动态预测模型中，没有发生变化。
由于料层厚度动态预测模型是依据部分烧结系统的实际生产数据预先建立，而在应用阶段，料层厚度动态预测模型会使用在所有烧结系统中，而不同烧结系统的实际工况存在一定的差异性，以及同一烧结系统在长期使用过程中，随着时间推移，工况也会发生一定的变化，在此情况下，若料层厚度动态预测模型没有适应性的调整，可能会出现料层厚度的预测值与实际结果偏差较大的情况，为了避免这一技术问题，在本申请的部分实施例中，所述料层厚度预测系统还包括料层厚度检测装置10，所述层厚检测装置可以实时检测烧结台车上的料层厚度，所述料层厚度检测装置10设置在烧结台车上位，以检测烧结台车的料层厚度，例如采用料位计检测料层厚度。
如图4所示，在本申请的部分实施例中提供了一种在线更新料层厚度动态预测模型的方法。
在步骤S302将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值的步骤之前，还包括：
步骤S401，获取与预测料层厚度的特征向量邻近时序上的学习样本，其中，学习样本包括输入样本，以及输入样本对应的料层厚度的特征值。
需要说明的是，本申请中提及的样本、学习样本，含义相同，指代的均为学习样本。
其中，所述学习样本不仅包含了烧结系统的混合料参数及烧结系统状态参数，还包含了对应的料层厚度测量值，所述学习样本的具体获取方法是：获取与预测料层厚度的特征向量邻近时序上的学习样本，其中，学习样本包括输入样本，以及输入样本对应的料层厚度的特征值；需要说明的是，料层厚度的测量值需要经过量化处理后，才能用作料层厚度动态预测模型的更新或者训练，归一化化处理采用以下模型：
其中，h
i为获得的i时刻料层厚度的预测值，k
i为生成的i时刻料层厚度的特征值，H为烧结台车上的料层最大允许厚度。
这里采集的历史预测样本是用作学习样本使用，由于烧结系统的烧结过程是一个长时间连续的过程，在进行预测料层厚度的特征向量采集之前，烧结系统一直在进行混合作业，采集与所述预测料层厚度的特征向量最近的预设时间间隔内的历史预测样本，可以保证得到的学习样本所对应的烧结系统工况，与预测料层厚度的特征向量采集时间点的烧结系统工况一致。
步骤S402，利用所述学习样本在线更新所述料层厚度动态预测模型，获取更新后的料层厚度动态预测模型。
通过步骤S401中获取的学习样本，可以对料层厚度动态预测模型进行在线更新，进一步保证所述料层厚度动态预测模型输出的料层厚度的特征值更精准。在具体更新过程中，根据学习样本预测的料层厚度的特征值，确定与测量值对应的料层厚度的特征值之间的预测偏差值，所述动态预测模型具备两种更新方式，若预测偏差值较小且相对稳定，处于模型质量指标允许误差范围内，则将预测偏差值直接加到所述动态预测模型的料层厚度的特征值上，并将结果作为更新后的料层厚度的特征值。若预测偏差值较大，且根据模型质量指标判断为映射关系发生变化，则根据所述学习样本以及测量值对应的料层厚度的特征值，更新所述料层厚度动态预测模型中包含的映射关系。
需要说明的是，通常可采用预测偏差值的均方差作为质量指标，然后根据质量指标的统计分布规律，预设统计置信限，判断是否需要触发更新以及需要的更新方法。测量值对应的料层厚度的特征值，虽然因为滞后问题不适合用于混合工艺闭环控制问题，但可用于稳定状态的参考，即烧结系统在稳定状态下达到稳态时，料层厚度会维持在一定的水平，在正常的情况下，测量值虽有波动，但其分布通常不会偏离置信区间，当超出这个置信区间，触发模型更新机制。如果根据指标分析结果判断过程特征属于渐变，则选择模型递推法，利用移动窗递推法来更新料层厚度动态预测模型，步骤如下：
设原动态预测模型的样本集为S＝{[X
1,Y
1],...,[X
t,Y
t]}，t为总的样本数。当获取新测量值[X
m,Y
m]，则将其加入样本集，并淘汰最陈旧的样本，则新的样本集(学习样本)为：
S＝{[X
2,Y
2],...,[X
t,Y
t],[X
m,Y
m]}
然后利用学习样本在线更新所述料层厚度动态预测模型，获得新的料层厚度动态预测模型。如果根据指标分析结果判断过程特征属于突变，则选择即时学习法，重构料层厚度动态预测模型。
所述料层厚度动态预测模型，是利用多组已知混合料参数、烧结系统状态参数以及对应的料层厚度测量值建立。在本申请实施例中提供一种利用神经网络模型训练的方式，生成所述料层厚度动态预测模型。具体操作是将多组已知混合料参数、烧结系统状态参数作 为神经网络模型的输入，而将对应的料层厚度测量值，作为神经网络模型的输出，对神经网络模型进行训练，即调整神经网络模型中间层对应的权重矩阵和偏置项，从而建立混合料参数以及对应的混合机状态参数，与混匀度之间的映射关系。参见图5，为本申请实施例提供的一种生成料层厚度动态预测模型的流程图，具体生成步骤包括：
步骤S501，获取N组独立的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度。
其中，N组独立的混合料参数以及对应的烧结系统状态参数，可以是同一烧结系统的数据，也可以是多个烧结系统的数据，并按照对应关系分为一组，即，同一混合机且同一时刻的数据作为统一组数据。
步骤S502，将N组独立的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，生成N组训练样本的输入。其中，混合料参数中的堆密度，以及烧结系统状态参数中的给料辊转速、布料辊转速和烧结台车速度，需要量化到区间(0,1)，其中，辅门开度的数值为百分比数据。
步骤S503，获取料层厚度检测装置10发送的N组独立的料层厚度的测量值，并将料层厚度的测量值按照一定收缩比例量化，生成N组训练样本的输出。
其中，将料层厚度的测量值按照一定收缩比例量化，具体操作是计算料层厚度的测量值与烧结台车的料层最大允许厚度的比值。
步骤S504，利用训练样本的输入以及训练样本的输出，采用时间反向传播法训练神经网络模型。
所述动态预测训练模块利用训练样本的输入以及训练样本的输出，采用时间反向传播法训练神经网络模型；其中，时间反向传播法是一种适用于多层神经元网络的学习算法，通过激励传播和权重更新反复循环迭代，指导多层神经元网络对输入的响应(输出)达到预定的目标范围为止。
步骤S505，通过迭代训练不断更新神经网络模型的权重参数、偏置参数以及学习因子。
步骤S506，若神经网络模型的预测料层厚度的特征值与测量的料层厚度的特征值，达到设定的容差范围，或神经网络模型达到设定的最大迭代次数，则训练结束，并保存最后更新的权重参数、偏置参数以及学习因子，获得料层厚度动态预测模型。
通过不断更新神经网络模型的权重参数、偏置参数以及学习因子，在神经网络模型建立预测料层厚度的特征向量与料层厚度的特征值的映射关系，直到根据所述映射关系预测的料层厚度的特征值满足生产需求。具体的判断方法是，判断神经网络模型的预测的料层厚度的特征值与检测的料层厚度的特征值是否达到设定的容差范围，以及判断神经网络模型是否达到设定的最大迭代次数，若神经网络模型的预测的料层厚度的特征值与检测的料层厚度的特征值达到设定的容差范围，或神经网络模型达到设定的最大迭代次数，则训练结束，并保存最后更新的权重参数、偏置参数以及学习因子，获得料层厚度动态预测模型。
本申请实施例提供的利用神经网络模型训练生成所述料层厚度动态预测模型。具体可以采用RNN(Recurrent Neural Network)神经网络或者LSTM(Long Short-Term Memory)神经网络，如图6所示，为基于LSTM神经网络模型的结构示意图，其中，中间层的结构主要有遗忘门、输入门和输出门组成，每个门都有对应的权重矩阵和偏置项，通过训练样本对多层神经元网络进行训练，不断更新权重参数、偏置参数以及学习因子，从而获得所 述料层厚度动态预测模型。
在本申请实施例的神经网络模型具体训练过程中，所述神经网络模型将多组训练样本分为两个部分，包括2/3组的训练数据和1/3组的测试数据，将2/3组训练样本的输入以及训练样本的输出，作为神经网络模型的训练数据，不断更新权重参数、偏置参数以及学习因子；并利用1/3组训练样本的输入以及训练样本的输出，作为神经网络模型的容差测试数据。
参见图7，为本申请实施例提供的另一种生成料层厚度动态预测模型的流程图，所述料层厚度动态预测模型还可以是混匀度预测表，并按照以下步骤建立：
步骤S701，获取N组独立的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度。
步骤S702，获取料层厚度检测装置发送的N组独立的料层厚度的测量值。
步骤S703，对N组独立的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度以及对应的料层厚度的测量值，进行统计分析，建立层厚预测表。
需要说明的是，本申请提及的层厚预测表、料层厚度预测表，含义相同，指代的均为料层厚度预测表。
其中，统计分析包括数据预处理，数据预处理先将堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，然后将混合料参数、烧结系统状态参数相同的多组数据提取出，并比较不同组对应的料层厚度的测量值，若对应的料层厚度的测量值也相同，则只保留一组数据，去除其他组相同的数据。若对应的混匀度不同，但是偏差处于允许范围内，则将多组数据对应的混匀度平均值作为对应的料层厚度，且只保留一组数据，一般根据生产精度要求，偏差的允许范围设置为0.5％-2％。对于混合料参数、烧结系统状态参数相同的多组数据，对应的混匀度测量值不同，且偏差超过允许范围，且标记多组数据的来源，保留在隔离区，这里的数据来源是指数据采集对应的烧结系统及采集时间，隔离区是指在层厚预测表单独的一块区域，用于记录异常数据，以作为设备检修的参考数据。
需要说明的是，这里的料层厚度的测量值同样需要进行预处理，具体操作是，计算料层厚度的测量值与烧结台车的料层最大允许厚度的比值。
对于经过数据预处理的多组独立的混合料参数、烧结系统状态参数以及对应的料层厚度测量值，按照设定的多个索引项进行排序，每一个索引项的排序为从小到大排列。其中，至少包括5个索引项，例如本申请实施例中，设置的5个索引项，以及索引顺序，第一级索引项为混合料的堆密度，第二级索引项为给料辊转速，第三级索引项为布料辊转速，第四级索引项为辅门开度，第五级索引项为烧结台车速度。按照以上规则完成混匀度预测表的建立，所述混匀度预测表中包含预测料层厚度的特征向量与料层厚度的特征值的映射关系，例如表1所示，为混匀度预测表的部分数据。
表1层厚预测表部分示例数据
下述为本申请方法实施例，用于实施本申请方法实施例。对于本申请方法实施例中未披露的细节，请参照本申请系统实施例。
本申请实施例提供一种基于料层厚度预测的布料控制方法，所述布料控制方法包括：
获取混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度。
根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值。
其中，所述根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值的步骤，包括：
将同一时刻的混合料的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，生成预测料层厚度的特征向量；
将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值，所述料层厚度动态预测模型中包含预测料层厚度的特征向量与料层厚度的特征值之间的映射关系；
对料层厚度的特征值进行数据还原，获得料层厚度的预测值。
根据料层厚度的预测值和料层厚度的目标值，计算料层厚度偏差值。
将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度。
将给料辊转速调整为待调整的给料辊转速；将布料辊转速调整为待调整的布料辊转速；将辅门开度调整为待调整的辅门开度；将烧结台车速度调整为待调整的烧结台车速度。
其中，所述滚动优化模型用于在所述混合料的堆密度不变的条件下，计算出料层厚度偏差值的方差最小时，对应的给料辊转速、布料辊转速、辅门开度和烧结台车速度。
由上述技术方案可知，本申请实施例提供的一种基于料层厚度预测的布料控制系统及方法，所述布料控制系统包括圆辊给料机3、辊式布料机4和烧结台车5，所述圆辊给料机3用于向所述辊式布料机4供给混合料，所述辊式布料机4用于向烧结台车5布料；所述布料控制系统还包括混合料检测机构101、与圆辊给料机3连接的给料辊控制器102、与辊式布料机4连接的布料辊控制器103和辅门控制器104，以及与烧结台车5连接的烧结台车控制器105；以及与混合料检测机构101、给料辊控制器102、布料辊控制器103、辅门控制器104和烧结台车控制器105连接的中央处理器106。
在实际应用过程中，首先获取混合料的堆密度、给料辊转速、布料辊转速、辅门开度以及烧结台车速度；然后将同一时刻的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，按照一定收缩比例量化到同一区间，并生成预测料层厚度的特征向量；再然后将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值；并对料层厚度的特征值进行数据还原，获得料层厚度的预测值，之后根据料层 厚度的预测值和料层厚度的目标值，计算料层厚度偏差值，最后将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度，从而基于料层厚度预测实现烧结系统的布料控制。本申请实施例提供的基于料层厚度预测的布料控制系统，可以通过采集的混合料参数以及烧结系统状态参数，提前预测出烧结台车上的料层厚度，以便通过预测的料层厚度，实现对烧结系统及时、稳定的反馈布料控制。
本申请提供的实施例之间的相似部分相互参见即可，以上提供的具体实施方式只是本申请总的构思下的几个示例，并不构成本申请保护范围的限定。对于本领域的技术人员而言，在不付出创造性劳动的前提下依据本申请方案所扩展出的任何其他实施方式都属于本申请的保护范围。
Claims (10)
- 一种基于料层厚度预测的布料控制系统，所述布料控制系统包括圆辊给料机、辊式布料机和烧结台车，所述圆辊给料机用于向所述辊式布料机供给混合料，所述辊式布料机用于向烧结台车布料；其特征在于，所述布料控制系统还包括混合料检测机构、与圆辊给料机连接的给料辊控制器、与辊式布料机连接的布料辊控制器和辅门控制器，以及与烧结台车连接的烧结台车控制器；以及与混合料检测机构、给料辊控制器、布料辊控制器、辅门控制器和烧结台车控制器连接的中央处理器；其中：所述中央处理器被配置为执行以下步骤：接收所述混合料检测机构发送的混合料堆密度，接收给料辊控制器发送的给料辊转速，接收布料辊控制器发送的布料辊转速，接收辅门控制器发送的辅门开度，接收烧结台车控制器发送的烧结台车速度；根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值；根据料层厚度的预测值和料层厚度的目标值，计算料层厚度偏差值；将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度；驱动所述给料辊控制器将给料辊转速调整为待调整的给料辊转速；驱动所述布料辊控制器将布料辊转速调整为待调整的布料辊转速；驱动所述辅门控制器将辅门开度调整为待调整的辅门开度；驱动所述烧结台车控制器将烧结台车速度调整为待调整的烧结台车速度；其中，所述滚动优化模型用于在所述混合料的堆密度不变的条件下，计算出料层厚度偏差值的方差最小时，对应的给料辊转速、布料辊转速、辅门开度和烧结台车速度。
- 根据权利要求1所述的基于料层厚度预测的布料控制系统，其特征在于，所述料层厚度偏差值的方差通过以下方式得到：
其中，σ是料层厚度偏差值的方差，E(k)是指料层厚度偏差值，R(k)预设的料层厚度的目标值，Y(k)料层厚度的预测值。 - 根据权利要求1所述的基于料层厚度预测的布料控制系统，其特征在于，根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值，具体执行以下步骤：将同一时刻的混合料的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，生成预测料层厚度的特征向量；将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值，所述料层厚度动态预测模型中包含预测料层厚度的特征向量与料层厚度的特征值之间的映射关系；对料层厚度的特征值进行数据还原，获得料层厚度的预测值。
- 根据权利要求3所述的基于料层厚度预测的布料控制系统，其特征在于，所述布料控制系统还包括设置在烧结台车上方的料层厚度检测装置，所述料层厚度检测装置连接中央处理器，所述料层厚度检测装置用于按照预先设定的时间间隔，检测烧结台车的料层厚度，并获得料层厚度的测量值；将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值，还包括：获取与所述预测料层厚度的特征向量邻近时序上的数据并作为样本，其中，所述样本包括输入样本，以及所述输入样本对应的料层厚度的特征值；利用所述学习样本在线更新所述料层厚度动态预测模型，获取更新后的料层厚度动态预测模型。
- 根据权利要求3所述的基于料层厚度预测的布料控制系统，其特征在于，采用以下方法对料层厚度的特征值进行数据还原：h i＝k i×H其中，h i为获得的i时刻料层厚度的预测值，k i为生成的i时刻料层厚度的特征值，H为烧结台车上的料层最大允许厚度。
- 根据权利要求3所述的基于料层厚度预测的布料控制系统，其特征在于，将同一时刻的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，具体执行以下步骤：计算堆密度与各原料中密度最大的原料的密度的比值；计算给料辊转速与给料辊的最大转速的比值；计算布料辊转速与布料辊的最大转速的比值；计算烧结台车速度与烧结台车的最大转速的比值。
- 根据权利要求3所述的基于料层厚度预测的布料控制系统，其特征在于，所述料层厚度动态预测模型是基于神经网络模型训练生成，并按照以下步骤建立：按照预先设定的时间间隔，获取N组独立的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度；将N组独立的混合料的堆密度、给料辊转速、布料辊转速和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，作为N组训练样本的输入；按照预先设定的时间间隔，检测烧结台车上与上述N组训练样本输入对应的实际料层厚度，并计算实际料层厚度的实际特征值，将实际特征值作为N组输出训练样本；利用输入训练样本以及输出训练样本，采用时间反向传播法训练神经网络模型；通过迭代训练不断更新神经网络模型的权重参数、偏置参数以及学习因子；若神经网络模型的预测值与测量值达到设定的允差范围，或神经网络模型达到设定的最大迭代次数，则训练结束，并保存最后更新的权重参数、偏置参数以及学习因子，获得料层厚度动态预测模型。
- 根据权利要求3所述的基于料层厚度预测的布料控制系统，其特征在于，所述料层厚度动态预测模型还可以是料层厚度预测表，并按照以下步骤建立：按照预先设定的时间间隔，获取N组独立的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度；按照预先设定的时间间隔，检测烧结台车上与上述N组输入训练样本对应的实际料层厚度，并计算实际料层厚度的实际特征值；对N组独立的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度以及对应的实际特征值，进行统计分析，建立料层厚度预测表。
- 一种基于料层厚度预测的布料控制方法，其特征在于，所述布料控制方法包括：获取混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度；根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值；根据料层厚度的预测值和料层厚度的目标值，计算料层厚度偏差值；将料层厚度偏差值输入滚动优化模型，得到待调整的给料辊转速、待调整的布料辊转速、待调整的辅门开度和待调整的烧结台车速度；将给料辊转速调整为待调整的给料辊转速；将布料辊转速调整为待调整的布料辊转速；将辅门开度调整为待调整的辅门开度；将烧结台车速度调整为待调整的烧结台车速度；其中，所述滚动优化模型用于在所述混合料的堆密度不变的条件下，计算出料层厚度偏差值的方差最小时，对应的给料辊转速、布料辊转速、辅门开度和烧结台车速度。
- 根据权利要求9所述的基于料层厚度预测的布料控制方法，其特征在于，所述根据混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，对料层厚度进行预测，获得料层厚度的预测值的步骤，包括：将同一时刻的混合料的堆密度、给料辊转速、布料辊转速、辅门开度和烧结台车速度，按照一定收缩比例量化到同一区间，并结合辅门开度，生成预测料层厚度的特征向量；将预测料层厚度的特征向量输入到预先建立的料层厚度动态预测模型中，生成料层厚度的特征值，所述料层厚度动态预测模型中包含预测料层厚度的特征向量与料层厚度的特征值之间的映射关系；对料层厚度的特征值进行数据还原，获得料层厚度的预测值。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| BR112022023796A BR112022023796A2 (pt) | 2020-08-20 | 2021-08-13 | Sistema de controle de distribuição de material e método com base na previsão de espessura de camada de material |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010844822.3A CN113295000B (zh) | 2020-08-20 | 2020-08-20 | 一种基于料层厚度预测的布料控制系统及方法 |
| CN202010844822.3 | 2020-08-20 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2022037500A1 true WO2022037500A1 (zh) | 2022-02-24 |
Family
ID=77318321
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2021/112557 WO2022037500A1 (zh) | 2020-08-20 | 2021-08-13 | 一种基于料层厚度预测的布料控制系统及方法 |
Country Status (3)
| Country | Link |
|---|---|
| CN (1) | CN113295000B (zh) |
| BR (1) | BR112022023796A2 (zh) |
| WO (1) | WO2022037500A1 (zh) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114807596A (zh) * | 2022-05-07 | 2022-07-29 | 北京首钢自动化信息技术有限公司 | 一种矿堆的配料控制方法和装置 |
| CN115061427A (zh) * | 2022-06-28 | 2022-09-16 | 浙江同发塑机有限公司 | 吹塑机的料层均匀性控制系统及其控制方法 |
| CN115478159A (zh) * | 2022-09-01 | 2022-12-16 | 马鞍山钢铁股份有限公司 | 一种适用于超厚料层烧结的梯形布料装置 |
| CN117213260A (zh) * | 2023-10-13 | 2023-12-12 | 湖南科技大学 | 一种节能降耗的环冷机分布式智能协调控制方法 |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114147851B (zh) * | 2021-12-15 | 2023-03-14 | 筑友智造建设科技集团有限公司 | 一种混凝土布料控制方法及系统 |
| CN114290507A (zh) * | 2021-12-28 | 2022-04-08 | 筑友智造科技投资有限公司 | 一种多模具并行布料控制方法、系统、设备和存储介质 |
| CN114739182A (zh) * | 2022-03-17 | 2022-07-12 | 北京首钢自动化信息技术有限公司 | 烧结台车布料闸门的卡料判定方法、装置、设备及介质 |
| CN117420807B (zh) * | 2023-12-14 | 2024-03-12 | 深圳市德镒盟电子有限公司 | 一种智能控制去粘层厚度的方法、系统及生产设备 |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| IT1139918B (it) * | 1980-12-08 | 1986-09-24 | Bostroem Olle | Metodo di sinterizzazione per aspirazione e apparecchiatura relativa |
| KR100530081B1 (ko) * | 2002-12-12 | 2005-11-22 | 주식회사 포스코 | 고로용 소결광 장입 제어방법 |
| CN1776338A (zh) * | 2005-11-24 | 2006-05-24 | 广东韶钢松山股份有限公司 | 一种烧结自动布料方法 |
| CN101560599A (zh) * | 2009-04-17 | 2009-10-21 | 中冶长天国际工程有限责任公司 | 一种混合料层厚的控制方法及控制系统 |
| CN102072657A (zh) * | 2010-12-30 | 2011-05-25 | 中南大学 | 一种基于多目标遗传算法的烧结布料过程优化控制方法 |
| CN102072658A (zh) * | 2010-12-30 | 2011-05-25 | 中南大学 | 一种稳定料层厚度的烧结偏析布料控制方法 |
| CN204665915U (zh) * | 2015-03-23 | 2015-09-23 | 宝钢不锈钢有限公司 | 一种烧结机料层厚度控制系统 |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3950244B2 (ja) * | 1998-11-13 | 2007-07-25 | 新日本製鐵株式会社 | 焼結原料の装入制御方法 |
| CN101339115B (zh) * | 2008-08-20 | 2010-11-03 | 中冶长天国际工程有限责任公司 | 一种混合料密度的检测方法及检测系统 |
| JP5400555B2 (ja) * | 2009-03-31 | 2014-01-29 | 株式会社神戸製鋼所 | 高炉の操業条件導出方法、及びこの方法を用いた高炉の操業条件導出装置 |
| TWI513948B (zh) * | 2013-03-29 | 2015-12-21 | China Steel Corp | 燒結機佈料控制系統與方法 |
| JP2014201827A (ja) * | 2013-04-10 | 2014-10-27 | Ｊｆｅスチール株式会社 | 焼結鉱の冷却制御方法 |
| CN104180659B (zh) * | 2013-05-22 | 2016-03-30 | 宝山钢铁股份有限公司 | 烧结机头部组合偏析布料方法 |
| KR101719516B1 (ko) * | 2015-11-05 | 2017-03-24 | 주식회사 포스코 | 소결광 제조 방법 |
-
2020
- 2020-08-20 CN CN202010844822.3A patent/CN113295000B/zh active Active
-
2021
- 2021-08-13 WO PCT/CN2021/112557 patent/WO2022037500A1/zh active Application Filing
- 2021-08-13 BR BR112022023796A patent/BR112022023796A2/pt unknown
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| IT1139918B (it) * | 1980-12-08 | 1986-09-24 | Bostroem Olle | Metodo di sinterizzazione per aspirazione e apparecchiatura relativa |
| KR100530081B1 (ko) * | 2002-12-12 | 2005-11-22 | 주식회사 포스코 | 고로용 소결광 장입 제어방법 |
| CN1776338A (zh) * | 2005-11-24 | 2006-05-24 | 广东韶钢松山股份有限公司 | 一种烧结自动布料方法 |
| CN101560599A (zh) * | 2009-04-17 | 2009-10-21 | 中冶长天国际工程有限责任公司 | 一种混合料层厚的控制方法及控制系统 |
| CN102072657A (zh) * | 2010-12-30 | 2011-05-25 | 中南大学 | 一种基于多目标遗传算法的烧结布料过程优化控制方法 |
| CN102072658A (zh) * | 2010-12-30 | 2011-05-25 | 中南大学 | 一种稳定料层厚度的烧结偏析布料控制方法 |
| CN204665915U (zh) * | 2015-03-23 | 2015-09-23 | 宝钢不锈钢有限公司 | 一种烧结机料层厚度控制系统 |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114807596A (zh) * | 2022-05-07 | 2022-07-29 | 北京首钢自动化信息技术有限公司 | 一种矿堆的配料控制方法和装置 |
| CN114807596B (zh) * | 2022-05-07 | 2023-11-07 | 北京首钢自动化信息技术有限公司 | 一种矿堆的配料控制方法和装置 |
| CN115061427A (zh) * | 2022-06-28 | 2022-09-16 | 浙江同发塑机有限公司 | 吹塑机的料层均匀性控制系统及其控制方法 |
| CN115061427B (zh) * | 2022-06-28 | 2023-04-14 | 浙江同发塑机有限公司 | 吹塑机的料层均匀性控制系统及其控制方法 |
| CN115478159A (zh) * | 2022-09-01 | 2022-12-16 | 马鞍山钢铁股份有限公司 | 一种适用于超厚料层烧结的梯形布料装置 |
| CN115478159B (zh) * | 2022-09-01 | 2023-11-21 | 马鞍山钢铁股份有限公司 | 一种适用于超厚料层烧结的梯形布料装置 |
| CN117213260A (zh) * | 2023-10-13 | 2023-12-12 | 湖南科技大学 | 一种节能降耗的环冷机分布式智能协调控制方法 |
| CN117213260B (zh) * | 2023-10-13 | 2024-05-24 | 湖南科技大学 | 一种节能降耗的环冷机分布式智能协调控制方法 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113295000B (zh) | 2022-04-12 |
| BR112022023796A2 (pt) | 2023-03-28 |
| CN113295000A (zh) | 2021-08-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022037500A1 (zh) | 一种基于料层厚度预测的布料控制系统及方法 | |
| EP2329327B1 (en) | Method and system for controlling an industrial process | |
| Zhou et al. | Intelligence-based supervisory control for optimal operation of a DCS-controlled grinding system | |
| WO2022037499A1 (zh) | 一种用于造球机的控制系统及控制方法 | |
| CN111079906A (zh) | 基于长短时记忆网络的水泥成品比表面积预测方法及系统 | |
| CN110782067B (zh) | 一种基于模糊信息粒化的烧结终点波动范围预测方法 | |
| CN110090728B (zh) | 用于控制水泥生料立磨中喂料量的方法、装置及设备 | |
| CN113299352B (zh) | 一种烧结台车的料层厚度动态预测系统及方法 | |
| CN104913639A (zh) | 基于数据融合的烧结终点控制系统及控制方法 | |
| CN113289541B (zh) | 一种基于混匀度预测的混合机控制系统及方法 | |
| CN113299353B (zh) | 一种混合机的混匀度预测方法及系统 | |
| CN114326630A (zh) | 沥青混合料智能称重配料控制方法 | |
| Zhang et al. | Application of neural network predictive control in cement combined grinding | |
| CN113289542B (zh) | 一种基于混合能效比最优的混合机控制系统及方法 | |
| RU2812444C1 (ru) | Система и способ управления распределением материала на основе прогнозирования толщины слоя материала | |
| CN114169640A (zh) | 一种烧结流程冷却圆筒返粉料水分预测方法及预测系统 | |
| CN109142167B (zh) | 一种磨矿粒度的在线鲁棒正则软测量模型的建立方法 | |
| CN113297542A (zh) | 一种造球机的生球占比预测方法及装置 | |
| Sun et al. | Development of an Online Updating Stochastic Configuration Network for the Soft-sensing of the Semi-autogenous Ball Mill Crusher System | |
| Liu et al. | Online Modeling for Combined Cement Grinding System via Extreme Learning Machine | |
| Moshgbar et al. | An intelligent distributed system for real-time control of cone crushers | |
| Shen et al. | Modeling and analysis of material layer thickness in slag powder process | |
| CN117111558A (zh) | 一种融合工艺机理与数据模型的联合粉磨系统实时控制方法 | |
| CN116956140B (zh) | 一种智能分拣过程中参数动态预测方法及系统 | |
| Araromi et al. | Neuro-fuzzy modelling of blending process in cement plant |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21857595 Country of ref document: EP Kind code of ref document: A1 |
|
| REG | Reference to national code |
Ref country code: BR Ref legal event code: B01A Ref document number: 112022023796 Country of ref document: BR |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 2022131358 Country of ref document: RU |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| ENP | Entry into the national phase |
Ref document number: 112022023796 Country of ref document: BR Kind code of ref document: A2 Effective date: 20221122 |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 21857595 Country of ref document: EP Kind code of ref document: A1 |