CN117932377A - Cluster learning-based cold-rolled aluminum plate strip thickness control method and related equipment - Google Patents
Cluster learning-based cold-rolled aluminum plate strip thickness control method and related equipment Download PDFInfo
- Publication number
- CN117932377A CN117932377A CN202410038008.0A CN202410038008A CN117932377A CN 117932377 A CN117932377 A CN 117932377A CN 202410038008 A CN202410038008 A CN 202410038008A CN 117932377 A CN117932377 A CN 117932377A
- Authority
- CN
- China
- Prior art keywords
- rolling
- learning
- cold
- current
- environment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- XAGFODPZIPBFFR-UHFFFAOYSA-N aluminium Chemical compound [Al] XAGFODPZIPBFFR-UHFFFAOYSA-N 0.000 title claims abstract description 124
- 229910052782 aluminium Inorganic materials 0.000 title claims abstract description 124
- 238000000034 method Methods 0.000 title claims abstract description 108
- 238000004519 manufacturing process Methods 0.000 claims abstract description 58
- 238000005096 rolling process Methods 0.000 claims description 169
- 238000005097 cold rolling Methods 0.000 claims description 40
- 230000008569 process Effects 0.000 claims description 36
- 230000009471 action Effects 0.000 claims description 23
- 230000002787 reinforcement Effects 0.000 claims description 19
- 238000001914 filtration Methods 0.000 claims description 17
- 238000012549 training Methods 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 10
- 239000003795 chemical substances by application Substances 0.000 claims description 9
- 238000011217 control strategy Methods 0.000 claims description 8
- 230000035945 sensitivity Effects 0.000 claims description 8
- 238000012546 transfer Methods 0.000 claims description 8
- 239000004411 aluminium Substances 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 230000004044 response Effects 0.000 claims description 3
- 230000006872 improvement Effects 0.000 abstract description 5
- 238000013461 design Methods 0.000 abstract description 4
- 238000005070 sampling Methods 0.000 abstract description 4
- 230000008859 change Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000004043 responsiveness Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Abstract
A method for controlling the thickness of a cold-rolled aluminum plate strip based on cluster learning. According to the method, only the historical data set is needed to carry out data driving modeling and strategy learning on the production environment, so that the problem of high labor cost of manual design automatic control mathematical formula is avoided, no priori knowledge of uncertain parameters is needed, the method is more convenient, the sampling efficiency is higher, the control precision and stability of the thickness deviation of the aluminum plate strip are improved, and the improvement of product quality and intelligent improvement are brought to the thickness control of the cold-rolled aluminum strip.
Description
Technical Field
The application relates to the technical field of aluminum cold ligature, in particular to a method, a device and equipment for controlling the thickness of a cold-rolled aluminum plate strip based on cluster learning and a computer readable storage medium.
Background
In recent years, with the continuous increase of the competition of the international and domestic markets and the high-speed development of the aluminum cold continuous rolling technology, the quality requirements of the market on cold-rolled aluminum plate strip products are increasing. One of the important indexes for representing the high quality level of the aluminum sheet strip product is the thickness tolerance of the aluminum sheet strip in the longitudinal direction. The reduction of the thickness difference of the product is not only beneficial to improving the yield of plate and strip production enterprises, but also is a precondition for realizing full-automatic mass production of downstream tank manufacturing enterprises.
In the current advanced cold rolling industry, the popular technology for controlling the online thickness difference is an automatic thickness control system, which is abbreviated as AGC (Automatic Gauge Control) and is derived from the modern control theory, and the modes are that the thickness of a rolled piece at the inlet and outlet of a rolling mill is continuously measured by a thickness gauge or various sensors, the comparison deviation from a target set value is obtained according to the actual measured value, and then the online thickness difference is controlled in an allowable range by changing the pressure, the rolling speed and the like of the rolling mill based on the deviation value and by means of a high-speed computer function and a field-level control loop.
The AGC system of the cold-rolled aluminum plate strip is a complex multivariable and strong coupling control system, and factors influencing the AGC system are increased along with the reduction of the thickness of a product and the improvement of the precision requirement, the interrelationship among the factors is more complex, and an accurate mathematical model between the control parameters and the offline thickness difference cannot be established. There are problems in that flexibility is lacking, it is difficult to adapt to the change of rolling conditions, and it is necessary to continuously adjust parameters depending on expert experience. Therefore, the modern control method is combined with the artificial intelligence modeling method, so that the method is a challenging work for improving the thickness control precision and the stability of the future cold-rolled sheet strip.
Disclosure of Invention
The application provides a method, a device, equipment and a computer readable storage medium for controlling the thickness of a cold-rolled aluminum plate strip based on cluster learning, which can solve the technical problems that the mathematical model of the traditional AGC control method in the prior art is complex in modeling, expert experience is needed, and the self-adaptive production working condition is difficult to change.
In a first aspect, an embodiment of the present application provides a method for controlling a thickness of a cold-rolled aluminum sheet strip based on cluster learning, where the method for controlling a thickness of a cold-rolled aluminum sheet strip based on cluster learning includes:
acquiring historical rolling data of a cold rolling production line, and performing data processing on the historical rolling data to obtain an aluminum coil cold rolling process data set;
Determining an environment set based on the aluminum coil cold rolling process data set, and determining a current rolling production environment from the environment set by using a KNN clustering method;
performing data-driven rolling model modeling based on the current rolling production environment, and obtaining an integrated rolling model in the current rolling environment by using the integrated XGBoost;
Based on an integrated rolling model in the current rolling environment, a constraint filtering strategy learning method is used for learning an aluminum plate strip thickness control strategy.
In a second aspect, an embodiment of the present application provides a cold-rolled aluminum sheet strip thickness control device based on cluster learning, where the cold-rolled aluminum sheet strip thickness control device based on cluster learning includes:
the construction module is used for acquiring historical rolling data of the cold rolling production line, and carrying out data processing on the historical rolling data to obtain an aluminum coil cold rolling process data set;
The determining module is used for determining an environment set based on the aluminum coil cold rolling process data set, and determining the current rolling production environment from the environment set by using a KNN clustering method;
The generation module is used for performing data-driven rolling model modeling based on the current rolling production environment, and obtaining an integrated rolling model in the current rolling environment by using the integrated XGBoost;
And the learning module is used for learning the thickness control strategy of the aluminum plate strip by using a constraint filtering strategy learning method based on the integrated rolling model in the current rolling environment.
In a third aspect, an embodiment of the present application provides a cluster learning-based cold-rolled aluminum sheet strip thickness control apparatus, which includes a processor, a memory, and a cluster learning-based cold-rolled aluminum sheet strip thickness control program stored on the memory and executable by the processor, wherein the cluster learning-based cold-rolled aluminum sheet strip thickness control program, when executed by the processor, implements the steps of the cluster learning-based cold-rolled aluminum sheet strip thickness control method as described in the first aspect.
In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, where a cold-rolled aluminum sheet strip thickness control program based on cluster learning is stored on the computer readable storage medium, where the cold-rolled aluminum sheet strip thickness control program based on cluster learning implements the steps of the cold-rolled aluminum sheet strip thickness control method based on cluster learning according to the first aspect when the cold-rolled aluminum sheet strip thickness control program based on cluster learning is executed by a processor.
The technical scheme provided by the embodiment of the application has the beneficial effects that:
In the embodiment of the application, historical rolling data of a cold rolling production line is obtained, and the historical rolling data is subjected to data processing to obtain an aluminum coil cold rolling process data set; determining an environment set based on the aluminum coil cold rolling process data set, and determining a current rolling production environment from the environment set by using a KNN clustering method; performing data-driven rolling model modeling based on the current rolling production environment, and obtaining an integrated rolling model in the current rolling environment by using the integrated XGBoost; based on an integrated rolling model in the current rolling environment, a constraint filtering strategy learning method is used for learning an aluminum plate strip thickness control strategy. According to the embodiment of the application, only the historical data set is needed to carry out data driving modeling and strategy learning on the production environment, so that the problem of high labor cost of manually designing an automatic control mathematical formula is avoided, no priori knowledge of any uncertain parameter is needed, the method is more convenient, the sampling efficiency is higher, the control precision and stability of the thickness deviation of the aluminum plate strip are improved, and the improvement of product quality and intelligent improvement are brought to the thickness control of the cold-rolled aluminum strip.
Drawings
FIG. 1 is a schematic flow chart of an embodiment of a method for controlling the thickness of a cold-rolled aluminum sheet strip based on cluster learning;
FIG. 2 is a schematic view of a scenario of constructing an aluminum coil cold rolling process data set in an embodiment of a cluster learning-based cold rolled aluminum sheet strip thickness control method of the present application;
FIG. 3 is a diagram illustrating a state motion space of cold-cooling data in an embodiment of a method for controlling thickness of a cold-rolled aluminum sheet strip based on cluster learning according to the present application;
FIG. 4 is a schematic view of a scenario of an embodiment of a method for controlling the thickness of a cold rolled aluminum sheet strip based on cluster learning according to the present application;
FIG. 5 is a schematic diagram of functional modules of an embodiment of a cold rolled aluminum sheet strip thickness control device based on cluster learning according to the present application;
Fig. 6 is a schematic hardware structure diagram of a cluster learning-based cold-rolled aluminum sheet strip thickness control device according to an embodiment of the application.
Detailed Description
In order that those skilled in the art will better understand the present application, a technical solution in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the embodiments of the present application will be described in further detail with reference to the accompanying drawings.
In a first aspect, an embodiment of the application provides a method for controlling the thickness of a cold-rolled aluminum plate strip based on cluster learning.
In an embodiment, referring to fig. 1, fig. 1 is a schematic flow chart of a first embodiment of a method for controlling thickness of a cold rolled aluminum sheet strip based on cluster learning according to the present application. As shown in fig. 1, the method for controlling the thickness of the cold-rolled aluminum sheet strip based on cluster learning comprises the following steps:
step S10, historical rolling data of a cold rolling production line are obtained, and data processing is carried out on the historical rolling data to obtain an aluminum coil cold rolling process data set;
In this embodiment, historical rolling data of a factory aluminum strip production line is collected, and an original file format is converted into a usable format for machine learning and other approaches, and standardized.
Further, in one embodiment, an automatic thickness control AGC system applied in a cold rolling line is added with a Smith predictive controller.
In this embodiment, since the quality of the data set in the cold rolling process of the aluminum coil determines the effect of policy learning, the quality of the data set generated based on expert experience is more excellent. A production line based on a Smith predictive controller is selected, an aluminum coil cold rolling process data set containing expert experience knowledge is constructed, and the construction process is shown in fig. 2. Referring to fig. 2, first, rolling trajectory data is collected at intervals of 0.02s, and the state at each time is recorded. The offline rolling dataset is then generated by a Smith predictive controller based on expert experience to generate N markov process tuples containing states, actions, rewards, and next time state.
Because of a certain time lag in the measurement of the outlet thickness value and the time lag of the signal sampling processing and the execution system, the stability of the system control can be greatly affected by the larger system lag, so a controller monitoring module based on Smith estimation is introduced. Let the transfer function of the controller link be D(s), and the controlled object has a pure time lag term, let its transfer function be P(s) e -ts, and the time lag term be e -ts. The transfer function of the system added with the Smith estimation controller for monitoring AGC is as follows:
the influence of a pure hysteresis part on a control system is eliminated after the Smith pre-estimated compensation loop is arranged, e -ts is outside a closed-loop control loop, and the output of the controller is shifted by time t on a time coordinate by e -ts according to the displacement theorem of Law transformation, so that the transient process and other performance indexes of the control system are completely consistent with those of the object without time lag. Different from the control of the traditional AGC system, the monitoring AGC system of the Smith pre-estimated controller can lead the characteristic equation of the whole closed-loop system to contain no time lag term after pre-estimated compensation, thereby obtaining better theoretical control effect.
The rolling data of the production line, which consists of two rolling stands, contains expert experience knowledge by means of a Smith predictive controller, and the sensors continuously record and measure the thickness of the outlet aluminium strip and store it. For example, historical rolling mill production process data from 1 day to 30 days of 9 months of 22 are collected and stored in the. Dat format. Because the working condition and the production mode of the equipment are continuously changed, the data comprise different technological parameters, the ibaAnalyzer software is utilized for exporting the data, the original data are cleaned, the data are subjected to standard normalization, and null values are removed. And carrying out characteristic analysis on the sensor data according to the influence degree on the thickness deviation of the outlet of the aluminum coil, processing the sensor data into 16 action parameters, and analyzing the thickness deviation of the outlet of the two racks to serve as a state parameter. Sampling is carried out every 0.02s, the data set contains 1063 aluminum roll production data in total, and the data set comprises 8373540 (state, action, reward, state ') tuples for training of an offline reinforcement learning agent, wherein the state represents the system state at the current moment, the action represents the action taken at the current moment, the reward represents the reward obtained by the system after the action at the current moment is executed, and the state' represents the system state at the next moment after the action at the current moment is executed.
Step S20, determining an environment set based on an aluminum coil cold rolling process data set, and determining a current rolling production environment from the environment set by using a KNN clustering method;
in this embodiment, the working condition is constantly changed when the aluminum strip is rolled, and learning the data driving model only by using the offline data set may result in a decrease in model accuracy.
Further, in an embodiment, step S20 includes:
Determining an environment set B based on the aluminum coil cold rolling process data set; based on the current ambient external parameters W, the strip width D r,1 and the drum diameter D r,2, the current rolling production environment B is determined from the set of environments B using the KNN clustering method.
In this embodiment, the production environment during the rolling of the aluminum strip is different, and the production environment in the real world, such as the process parameters, may change over time. The direct application of the reinforcement learning model causes mismatching with the time-varying environment, and the reinforcement learning agent training effect is poor; meanwhile, the accuracy of the data-driven model constructed based on the whole offline data set is far lower than that of the model constructed based on the current environment. Therefore, a clustering method is adopted to sense the conditions of different operation parameters in the production environment, and the more similar the external preset parameters in the production process are, the more similar the environments are. Production environment similarity is measured by comparing the following factors: the current ambient external parameters W, the strip width D r,1, and the drum diameter D r,2 are entered. Using historical offline data sets under different operating strategies, historical environment B is defined as a set of environments B, namely:
Wherein [ b 1,bj,…,bN′ ] represents the set of all rolling working conditions, and N' production environments are provided.
After defining the environment, a clustering algorithm k-nearest neighbor (KNN) is used to identify similar environments, and the current rolling production environment is determined by identifying external parameters. Namely:
b＝KNN(B,W,Dr,1,Dr,2)
Step S30, performing data-driven rolling model modeling based on the current rolling production environment, and obtaining an integrated rolling model in the current rolling environment by using the integrated XGBoost;
In this embodiment, the interaction with the real production environment directly results in inefficiency and unsafe. A data-driven rolling model is thus constructed to represent the state transition process in reinforcement learning based on the current rolling environment. The method of integrated machine learning is used, a supervised learning mode is used, the state of the next time step is used as a label, and the average mean square error is used for learning state change so as to better fit the current rolling process.
Further, in an embodiment, the cold rolling line includes two rolling stands, and step S30 includes:
developing a data-driven rolling model based on the current rolling production environment, structuring
With action a t at time t and outlet thickness of two mill standsFor input, prediction/>, of the thickness deviation of the aluminium coil produced by the two rolling stands in the following time step t+1As a tag, two supervised learning problems:
wherein, S t is a state parameter at the time t and is used as a state transfer function, and the thickness deviation of aluminum coils at the outlets of two rolling mill frames at the next time is predicted through the state and the action at the current time;
Learning a rolling model by minimizing a mean square error between an actual value and a predicted value;
and training the independent rolling models according to different states to obtain an integrated rolling model in the current rolling environment.
In this embodiment, after defining the current rolling production environment, it is also necessary to construct a state transfer function, and since more than 10 process parameter controls are involved, it is necessary to obtain a state change after each time step is performed. Rolling lines of different models are based on different mathematical physical models defined by expert experience, so it is difficult to accurately model the rolling process using mathematical equations. For safety reasons, a data-driven rolling model is developed using current environmental data obtained from KNN, structured asSince the thickness deviation is in the micrometer level, the thickness deviation/>, of the two frames at the current moment tThe variation between adjacent time steps is relatively small, requiring a large amount of training data to capture the small variation. Designing an integrated XGBoost rolling process model, and setting the real-time thickness of two normalized frames at the time t as the current state/> And action a t (containing 16 parameters) as inputs, as shown in fig. 3. Referring to fig. 3, 16 motion parameters are depicted with their minimum, maximum and units shown. The thickness of the two frames at the current moment is selected as a state, and 2 state parameters are described and the minimum, maximum and unit of the state parameters are displayed. Prediction/>, of the thickness deviation of the aluminium coil produced by two frames in the next time step t+1As a tag, two supervised learning problems:
Wherein the method comprises the steps of As a state transfer function, S t is a state parameter at the current moment t, a t is an action parameter at the current moment t, and thickness deviation of aluminum coils produced by two frames at the next moment is predicted through the state and the action parameters at the current moment.
And learning an integrated rolling model in the current rolling environment by minimizing the mean square error between the actual value and the predicted value by using the current rolling environment b obtained after the offline data set clustering. Training individual data-driven rolling models for different statesAn integrated data-driven rolling model is formed to simulate the current aluminum strip cold rolling production environment.
And S40, learning an aluminum plate strip thickness control strategy by using a constraint filtering strategy learning method based on an integrated rolling model in the current rolling environment.
In this embodiment, an offline reinforcement learning algorithm based on constraint filtering is used to train the agent, screen out low-quality data that makes the model sensitive, and design multiple groups of learning networks so as to obtain an optimal production behavior strategy. The average rewards in the training process are used to evaluate the performance of the algorithm framework and different metrics are used to measure the effectiveness of the algorithm. If the algorithm effect is poor, the algorithm parameters are readjusted, and finally the thickness deviation of the aluminum strip is controlled in a reasonable range.
Further, in an embodiment, step S40 includes:
Based on an integrated rolling model in a current rolling environment, adopting an Actor-Critic reinforcement learning algorithm to perform reinforcement learning agent strategy learning, wherein two groups of Critic networks are adopted, and when a target value is calculated, a smaller value in two network predictions is taken, so that the problem of value overestimation in reinforcement learning is solved:
Wherein r is the current prize value, gamma is the prize attenuation, pi φ′ (s ') refers to the next state s' in the target network, and the smaller next state Q value in the target network and the current network is calculated;
meanwhile, the sensitivity of the rolling model is integrated through a constraint filtering strategy learning method;
Evaluating the response capability of the integrated rolling model to input data to explore generalization thereof, selecting and retaining data of which the integrated rolling model shows high confidence to output prediction of the integrated rolling model from training data1 to training data M, and filtering out samples which are difficult to predict accurately:
Wherein V represents the variance of the output, f is the integrated rolling model, (s, a) is the current state action pair, and co i is the added disturbance, if the sensitivity degree is larger, the integrated rolling model is sensitive to the input disturbance at (s, a), and the prediction uncertainty of the integrated rolling model to the integrated rolling model is larger.
In this embodiment, after the data-driven aluminum strip cold rolling model is established, an Actor-Critic reinforcement learning algorithm is adopted to perform reinforcement learning agent policy learning, so that decisions can be made in continuous actions and state space. Two sets of Critic networks were used. When the target value is calculated, the smaller value of the two network predictions is taken, so that the problem of value overestimation in reinforcement learning is solved:
Where r is the current prize value, γ is the prize decay, pi φ′ (s ') refers to the next state s' in the target network, and the smaller next state Q value in the target network and the current network is calculated.
Meanwhile, the data driving model is imperfect and prediction is inaccurate due to direct learning from a historical data set, and the model is difficult to fully fit data due to uneven data distribution and complex coupling in a system. A constraint filtering method is proposed to adjust the sensitivity of a data driven model.
The responsiveness of the model to input data is evaluated to explore its generalization. From training data 1 to M, selecting data for which the retention model exhibits high confidence in the output prediction, while filtering out samples that are difficult to predict accurately:
Where V denotes the variance of the output, f is the integrated rolling model, (s, a) is the current state action pair and co i is the disturbance applied. If the sensitivity is high, it is indicated that the integrated rolling model is sensitive to input disturbance at (s, a), and the prediction uncertainty of the prediction integrated rolling model is high, in other words, the prediction error rate is high. The imperfect data are screened out according to the preset proportion value by the method, so that the training effect of the reinforcement learning agent is improved.
Further, referring to fig. 4, fig. 4 is a schematic view of a scenario of an embodiment of a method for controlling thickness of a cold rolled aluminum sheet strip based on cluster learning according to the present application. As shown in fig. 4, the method for controlling the thickness of the cold-rolled aluminum sheet strip based on cluster learning is divided into three stages:
1. And a data acquisition stage. In the process of rolling the aluminum strip, the embodiment selects a rolling system with two racks to collect data, collects historical rolling data of a production line, and the data comprises contents such as an action state, wherein the action has 16 parameters, and the thickness of the two current racks is selected as the state.
2. A data-driven modeling phase. According to the collected historical rolling data on the aluminum strip cold rolling production line, the historical rolling data set D4ASCR is obtained through data cleaning and preprocessing, then the historical rolling data set D4ASCR is clustered into a current production environment b through a clustering KNN algorithm, and meanwhile, a data-driven integrated XGBoost model is trained and used for representing a rolling model in the current environment.
3. Offline training and deployment phase. The rolling model under the current environment of general generation combines with the constraint filtering method to train the offline reinforcement learning agent and continuously interacts with the dynamic model. Eventually it is deployed to the production system controller, forming a controlled loop.
In a second aspect, the embodiment of the application also provides a cold-rolled aluminum plate strip thickness control device based on cluster learning.
In one embodiment, referring to fig. 5, fig. 5 is a schematic functional block diagram of an embodiment of a cold rolled aluminum sheet strip thickness control device based on cluster learning according to the present application. As shown in fig. 5, the cold rolled aluminum sheet strip thickness control device based on cluster learning includes:
The construction module 10 is used for acquiring historical rolling data of the cold rolling production line, and performing data processing on the historical rolling data to obtain an aluminum coil cold rolling process data set;
A determining module 20, configured to determine an environment set based on the aluminum coil cold rolling process data set, and determine a current rolling production environment from the environment set using a KNN clustering method;
A generating module 30, configured to perform data-driven rolling model modeling based on a current rolling production environment, and obtain an integrated rolling model under the current rolling environment using the integrated XGBoost;
The learning module 40 is configured to learn an aluminum sheet strip thickness control strategy using a constraint filtering strategy learning method based on an integrated rolling model in a current rolling environment.
Further, in one embodiment, the automatic thickness control AGC system applied in the cold rolling line is added with a Smith predictive controller.
Further, in an embodiment, the determining module is configured to:
determining an environment set B based on the aluminum coil cold rolling process data set;
Based on the current ambient external parameters W, the strip width D r,1 and the drum diameter D r,2, the current rolling production environment B is determined from the set of environments B using the KNN clustering method.
Further, in an embodiment, the cold rolling line comprises two rolling stands, a generating module 30 for:
developing a data-driven rolling model based on the current rolling production environment, structuring
With action a t at time t and outlet thickness of two mill standsFor input, prediction/>, of the thickness deviation of the aluminium coil produced by the two rolling stands in the following time step t+1As a tag, two supervised learning problems:
wherein, S t is a state parameter at the time t and is used as a state transfer function, and the thickness deviation of aluminum coils at the outlets of two rolling mill frames at the next time is predicted through the state and the action at the current time;
Learning a rolling model by minimizing a mean square error between an actual value and a predicted value;
and training the independent rolling models according to different states to obtain an integrated rolling model in the current rolling environment.
Further, in an embodiment, the learning module 40 is configured to:
Based on an integrated rolling model in a current rolling environment, adopting an Actor-Critic reinforcement learning algorithm to perform reinforcement learning agent strategy learning, wherein two groups of Critic networks are adopted, and when a target value is calculated, a smaller value in two network predictions is taken, so that the problem of value overestimation in reinforcement learning is solved:
Wherein r is the current prize value, gamma is the prize attenuation, pi φ′ (s ') refers to the next state s' in the target network, and the smaller next state Q value in the target network and the current network is calculated;
meanwhile, the sensitivity of the rolling model is integrated through a constraint filtering strategy learning method;
Evaluating the response capability of the integrated rolling model to input data to explore generalization thereof, selecting and retaining data of which the integrated rolling model shows high confidence to output prediction of the integrated rolling model from training data1 to training data M, and filtering out samples which are difficult to predict accurately:
Wherein V represents the variance of the output, f is the integrated rolling model, (s, a) is the current state action pair, and co i is the added disturbance, if the sensitivity degree is larger, the integrated rolling model is sensitive to the input disturbance at (s, a), and the prediction uncertainty of the integrated rolling model to the integrated rolling model is larger.
The function implementation of each module in the cold-rolled aluminum sheet strip thickness control device based on cluster learning corresponds to each step in the cold-rolled aluminum sheet strip thickness control method embodiment based on cluster learning, and the function and implementation process of the function implementation are not described in detail herein.
In a third aspect, an embodiment of the present application provides a cold-rolled aluminum sheet strip thickness control device based on cluster learning, where the cold-rolled aluminum sheet strip thickness control device based on cluster learning may be a device with a data processing function, such as a personal computer (personal computer, PC), a notebook computer, a server, or the like.
Referring to fig. 6, fig. 6 is a schematic hardware structure diagram of a cold rolled aluminum sheet strip thickness control apparatus based on cluster learning according to an embodiment of the present application. In the embodiment of the application, the cold-rolled aluminum sheet strip thickness control equipment based on cluster learning can comprise a processor, a memory, a communication interface and a communication bus.
The communication bus may be of any type for implementing the processor, memory, and communication interface interconnections.
The communication interfaces include input/output (I/O) interfaces, physical interfaces, logical interfaces, and the like for realizing interconnection of devices inside the cluster learning-based cold-rolled aluminum sheet strip thickness control apparatus, and interfaces for realizing interconnection of the cluster learning-based cold-rolled aluminum sheet strip thickness control apparatus with other apparatuses (e.g., other computing apparatuses or user apparatuses). The physical interface may be an ethernet interface, a fiber optic interface, an ATM interface, etc.; the user device may be a Display, a Keyboard (Keyboard), or the like.
The memory may be various types of storage media such as random access memory (randomaccess memory, RAM), read-only memory (ROM), nonvolatile RAM (non-volatileRAM, NVRAM), flash memory, optical memory, hard disk, programmable ROM (PROM), erasable PROM (erasable PROM, EPROM), electrically erasable PROM (ELECTRICALLY ERASABLE PROM, EEPROM), and the like.
The processor can be a general processor, and the general processor can call a cold-rolled aluminum plate strip thickness control program based on cluster learning stored in the memory and execute the cold-rolled aluminum plate strip thickness control method based on cluster learning provided by the embodiment of the application. For example, the general purpose processor may be a central processing unit (central processing unit, CPU). The method executed when the cold-rolled aluminum sheet strip thickness control program based on cluster learning is called can refer to various embodiments of the cold-rolled aluminum sheet strip thickness control method based on cluster learning, and are not described herein.
Those skilled in the art will appreciate that the hardware configuration shown in fig. 6 is not limiting of the application and may include more or fewer components than shown, or may combine certain components, or a different arrangement of components.
In a fourth aspect, embodiments of the present application also provide a computer-readable storage medium.
The computer readable storage medium of the application stores a cold-rolled aluminum plate strip thickness control program based on cluster learning, wherein when the cold-rolled aluminum plate strip thickness control program based on cluster learning is executed by a processor, the steps of the cold-rolled aluminum plate strip thickness control method based on cluster learning are realized.
The method implemented when the cluster learning-based cold-rolled aluminum sheet strip thickness control program is executed can refer to various embodiments of the cluster learning-based cold-rolled aluminum sheet strip thickness control method of the present application, and will not be described herein.
It should be noted that, the foregoing reference numerals of the embodiments of the present application are merely for describing the embodiments, and do not represent the advantages and disadvantages of the embodiments.
The terms "comprising" and "having" and any variations thereof in the description and claims of the application and in the foregoing drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those listed steps or elements but may include other steps or elements not listed or inherent to such process, method, article, or apparatus. The terms "first," "second," and "third," etc. are used for distinguishing between different objects and not necessarily for describing a sequential or chronological order, and are not limited to the fact that "first," "second," and "third" are not identical.
In describing embodiments of the present application, "exemplary," "such as," or "for example," etc., are used to indicate by way of example, illustration, or description. Any embodiment or design described herein as "exemplary," "such as" or "for example" is not necessarily to be construed as preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary," "such as" or "for example," etc., is intended to present related concepts in a concrete fashion.
In the description of the embodiments of the present application, unless otherwise indicated, "/" means or, for example, a/B may represent a or B; the text "and/or" is merely an association relation describing the associated object, and indicates that three relations may exist, for example, a and/or B may indicate: the three cases where a exists alone, a and B exist together, and B exists alone, and furthermore, in the description of the embodiments of the present application, "plural" means two or more than two.
In some of the processes described in the embodiments of the present application, a plurality of operations or steps occurring in a particular order are included, but it should be understood that the operations or steps may be performed out of the order in which they occur in the embodiments of the present application or in parallel, the sequence numbers of the operations merely serve to distinguish between the various operations, and the sequence numbers themselves do not represent any order of execution. In addition, the processes may include more or fewer operations, and the operations or steps may be performed in sequence or in parallel, and the operations or steps may be combined.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) as described above, comprising several instructions for causing a terminal device to perform the method according to the embodiments of the present application.
The foregoing description is only of the preferred embodiments of the present application, and is not intended to limit the scope of the application, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.
Claims (10)
1. The cold-rolled aluminum plate strip thickness control method based on cluster learning is characterized by comprising the following steps of:
acquiring historical rolling data of a cold rolling production line, and performing data processing on the historical rolling data to obtain an aluminum coil cold rolling process data set;
Determining an environment set based on the aluminum coil cold rolling process data set, and determining a current rolling production environment from the environment set by using a KNN clustering method;
performing data-driven rolling model modeling based on the current rolling production environment, and obtaining an integrated rolling model in the current rolling environment by using the integrated XGBoost;
Based on an integrated rolling model in the current rolling environment, a constraint filtering strategy learning method is used for learning an aluminum plate strip thickness control strategy.
2. The method for controlling thickness of a cold rolled aluminum sheet strip based on cluster learning as claimed in claim 1, wherein an automatic thickness control AGC system applied to the cold rolling line is added with a Smith predictive controller.
3. The method for controlling thickness of a cold rolled aluminum sheet strip based on cluster learning as claimed in claim 1, wherein the step of determining an environment set based on the aluminum coil cold rolling process data set, and determining a current rolling production environment from the environment set using KNN cluster method comprises:
determining an environment set B based on the aluminum coil cold rolling process data set;
Based on the current ambient external parameters W, the strip width D r,1 and the drum diameter D r,2, the current rolling production environment B is determined from the set of environments B using the KNN clustering method.
4. The method for controlling thickness of a cold rolled aluminum sheet strip based on cluster learning as claimed in claim 1, wherein the cold rolling line comprises two rolling stands, wherein the step of performing data driven rolling model modeling based on a current rolling production environment and obtaining an integrated rolling model in the current rolling environment using the integrated XGBoost comprises:
developing a data-driven rolling model based on the current rolling production environment, structuring
With action a t at time t and outlet thickness of two mill standsFor input, prediction/>, of the thickness deviation of the aluminium coil produced by the two rolling stands in the following time step t+1As a tag, two supervised learning problems:
wherein, S t is a state parameter at the time t and is used as a state transfer function, and the thickness deviation of aluminum coils at the outlets of two rolling mill frames at the next time is predicted through the state and the action at the current time;
Learning a rolling model by minimizing a mean square error between an actual value and a predicted value;
and training the independent rolling models according to different states to obtain an integrated rolling model in the current rolling environment.
5. The method for controlling thickness of a cold rolled aluminum sheet strip based on cluster learning as claimed in claim 4, wherein the step of learning the thickness control strategy of the aluminum sheet strip using the constraint filter strategy learning method based on the integrated rolling model in the current rolling environment comprises:
Based on an integrated rolling model in a current rolling environment, adopting an Actor-Critic reinforcement learning algorithm to perform reinforcement learning agent strategy learning, wherein two groups of Critic networks are adopted, and when a target value is calculated, a smaller value in two network predictions is taken, so that the problem of value overestimation in reinforcement learning is solved:
Wherein r is the current prize value, gamma is the prize attenuation, pi φ′ (s ') refers to the next state s' in the target network, and the smaller next state Q value in the target network and the current network is calculated;
meanwhile, the sensitivity of the rolling model is integrated through a constraint filtering strategy learning method;
Evaluating the response capability of the integrated rolling model to input data to explore generalization thereof, selecting and retaining data of which the integrated rolling model shows high confidence to output prediction of the integrated rolling model from training data1 to training data M, and filtering out samples which are difficult to predict accurately:
Wherein V represents the variance of the output, f is the integrated rolling model, (s, a) is the current state action pair, and co i is the added disturbance, if the sensitivity degree is larger, the integrated rolling model is sensitive to the input disturbance at (s, a), and the prediction uncertainty of the integrated rolling model to the integrated rolling model is larger.
6. The utility model provides a cold rolled aluminum sheet strip thickness control device based on cluster study which characterized in that, cold rolled aluminum sheet strip thickness control device based on cluster study includes:
the construction module is used for acquiring historical rolling data of the cold rolling production line, and carrying out data processing on the historical rolling data to obtain an aluminum coil cold rolling process data set;
The determining module is used for determining an environment set based on the aluminum coil cold rolling process data set, and determining the current rolling production environment from the environment set by using a KNN clustering method;
The generation module is used for performing data-driven rolling model modeling based on the current rolling production environment, and obtaining an integrated rolling model in the current rolling environment by using the integrated XGBoost;
And the learning module is used for learning the thickness control strategy of the aluminum plate strip by using a constraint filtering strategy learning method based on the integrated rolling model in the current rolling environment.
7. The cluster-learning-based cold-rolled aluminum sheet strip thickness control device as claimed in claim 6, wherein an automatic thickness control AGC system applied to the cold-rolling production line is added with a Smith predictive controller.
8. The cluster-learning-based cold-rolled aluminum sheet strip thickness control device as claimed in claim 6, wherein the determining module is configured to:
determining an environment set B based on the aluminum coil cold rolling process data set;
Based on the current ambient external parameters W, the strip width D r,1 and the drum diameter D r,2, the current rolling production environment B is determined from the set of environments B using the KNN clustering method.
9. A cluster-learning-based cold-rolled aluminum sheet strip thickness control apparatus, characterized in that the cluster-learning-based cold-rolled aluminum sheet strip thickness control apparatus comprises a processor, a memory, and a cluster-learning-based cold-rolled aluminum sheet strip thickness control program stored on the memory and executable by the processor, wherein the cluster-learning-based cold-rolled aluminum sheet strip thickness control program, when executed by the processor, implements the steps of the cluster-learning-based cold-rolled aluminum sheet strip thickness control method as set forth in any one of claims 1 to 5.
10. A computer-readable storage medium, wherein a cluster-learning-based cold-rolled aluminum sheet strip thickness control program is stored on the computer-readable storage medium, wherein the cluster-learning-based cold-rolled aluminum sheet strip thickness control program, when executed by a processor, implements the steps of the cluster-learning-based cold-rolled aluminum sheet strip thickness control method as set forth in any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410038008.0A CN117932377A (en) | 2024-01-10 | 2024-01-10 | Cluster learning-based cold-rolled aluminum plate strip thickness control method and related equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410038008.0A CN117932377A (en) | 2024-01-10 | 2024-01-10 | Cluster learning-based cold-rolled aluminum plate strip thickness control method and related equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117932377A true CN117932377A (en) | 2024-04-26 |
Family
ID=90769459
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410038008.0A Pending CN117932377A (en) | 2024-01-10 | 2024-01-10 | Cluster learning-based cold-rolled aluminum plate strip thickness control method and related equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117932377A (en) |
-
2024
- 2024-01-10 CN CN202410038008.0A patent/CN117932377A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Solak et al. | Derivative observations in Gaussian process models of dynamic systems | |
| CN108764517B (en) | Method, equipment and storage medium for predicting change trend of silicon content in molten iron of blast furnace | |
| US5581459A (en) | Plant operation support system | |
| CN109992921B (en) | On-line soft measurement method and system for thermal efficiency of boiler of coal-fired power plant | |
| CN102257448B (en) | Method and device for filtering signal using switching models | |
| CN112580257B (en) | Method, device and equipment for constructing prediction model of complex industrial system based on deep learning and storage medium | |
| CN109856973A (en) | A kind of Technique for Controlling Greenhouse Environment and system based on fuzzy neural network | |
| CN110032799A (en) | A kind of the angle similarity divided stages and monitoring method of microbiological pharmacy process | |
| CN110097929A (en) | A kind of blast furnace molten iron silicon content on-line prediction method | |
| CN111222798B (en) | Complex industrial process key index soft measurement method | |
| Liu et al. | Long–short-term memory encoder–decoder with regularized hidden dynamics for fault detection in industrial processes | |
| CN109115446B (en) | Wind speed precision control method and system based on transonic wind tunnel | |
| Bouhouche et al. | Combined use of principal component analysis and self organisation map for condition monitoring in pickling process | |
| CN114117919A (en) | Instant learning soft measurement modeling method based on sample collaborative representation | |
| CN106599391B (en) | Association vector machine soft measurement modeling method based on dynamic weighting of triangle angle values | |
| CN116020879B (en) | Technological parameter-oriented strip steel hot continuous rolling space-time multi-scale process monitoring method and device | |
| CN117932377A (en) | Cluster learning-based cold-rolled aluminum plate strip thickness control method and related equipment | |
| CN115169721B (en) | Method and system for predicting single ton energy consumption in rectification process based on migration identification | |
| CN113988311B (en) | Quality variable prediction method, device, terminal and storage medium | |
| CN115631804A (en) | Method for predicting outlet concentration of sodium aluminate solution in evaporation process based on data coordination | |
| CN114139446A (en) | On-line detection soft measurement method for components in special rectification process | |
| CN114004138A (en) | Building monitoring method and system based on big data artificial intelligence and storage medium | |
| Agarwal | Application of Deep Learning in Chemical Processes: Explainability, Monitoring and Observability | |
| US20210232132A1 (en) | Online fault localization in industrial processes without utilizing a dynamic system model | |
| CN111291020A (en) | Dynamic process soft measurement modeling method based on local weighted linear dynamic system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination |