A method and system for end zoning of hot strip head monitoring

Abstract

The present application belongs to the technical field of hot continuous rolling, and particularly relates to a hot rolling coiler strip head monitoring end partition method and system, which comprises monitoring points arranged on the path through which the strip head passes, sensors on the monitoring points collecting information on the strip head position and obtaining an input data set, constructing a REF neural network based on the input data set and training the neural network, and continuously predicting before ideal results are obtained. The method proposed by the present application can realize online, real-time, automatic and non-contact measurement of the strip head, and realizes information acquisition and information processing functions of the sensors. Through the RBF neural network, the present application realizes qualitative tracking of different kinds of strip heads. When used online, the present application is fast, stable in performance, accurate in tracking effect, can adapt to harsh working conditions, effectively reduces labor intensity of workers, reduces tracking failure rate, improves automatic step control effect, and meets the needs of the strip process.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of hot continuous rolling, and particularly relates to a hot rolling coiler strip head monitoring end partition method and system. BACKGROUND

[0002] Strip head tracking is the basis of automatic jog control (AJC) of the coiler. In engineering practice, strip head tracking plays the role of a "traffic light". Higher strip head tracking accuracy can indicate the optimal take-off timing for the coiling roller.

[0003] The traditional strip head tracking method uses a rear sensor close to the coiling roller to correct a front sensor away from the coiling roller. This method relies too much on the accuracy of the rear sensor and ignores useful information extracted by the front sensor. Visual sensors are easily disturbed by external disturbances and sudden conditions, such as strip head warping and deviation. Temperature sensors are easily disturbed by environmental temperature. The detection method using a single sensor is not reliable.

[0004] In automatic jog control, if the take-off is too early, the pressure control loop will start before the strip head passes, causing head impact. If the take-off is too late, the position loop will take off late, also causing head impact, and even causing loose coiling and steel piling. The ideal take-off point should be to avoid the strip head lagging as soon as possible, to press the strip tightly, and to complete the switching of the position control loop to the pressure control loop. This puts high requirements on the accuracy and performance of strip head tracking. SUMMARY

[0005] The application provides a hot rolling coiler strip head monitoring end partition method and system to solve the problem of lacking an accurate strip head tracking method in the prior art.

[0006] To achieve the above purpose, the technical scheme of the application is as follows:

[0007] A hot rolling coiler strip head monitoring end partition method, comprising the following steps:

[0008] Step 1: Setting monitoring points on the path of the strip head: seven monitoring points are set to divide the strip head path into six monitoring sections, which are the first monitoring section, the second monitoring section, the third monitoring section, the fourth monitoring section, the fifth monitoring section, and the sixth monitoring section. Information acquisition devices are set at the seven monitoring points.

[0009] Step 2: The monitoring points acquire information about the position of the strip head: a data set is established according to the information obtained in step 1 as the input of the neural network.

[0010] Step 3: Construct the RBF neural network training model and train the neural network training model based on the dataset from Step 2 until the error between the prediction result and the actual result is within the preset range.

[0011] Preferably, in step 1, a thermal detection sensor, a pressure sensor, an industrial camera, and a laser detection sensor are sequentially installed at the first monitoring point, the second monitoring point, the third monitoring point, and the fourth monitoring point; and a pressure sensor is installed at the fifth monitoring point, the sixth monitoring point, and the seventh monitoring point.

[0012] Preferably, the second monitoring point is the pinch roll feedback point on the strip head path, and the fifth, sixth and seventh monitoring points are the first auxiliary roll feedback point, the second auxiliary roll feedback point and the third auxiliary roll feedback point, respectively.

[0013] Preferably, step 2 includes obtaining the average speed of the six monitoring segments based on the time it takes for the strip head to pass through the monitoring point:

[0014] Setting t0, t1, t2, t3, t4, t5, and t6 as the time it takes for the strip head to reach each monitoring point, we obtain:

[0015]

[0016] The v n Let n be the average speed of the nth monitoring segment, where n = 1, 2, 3, 4, 5, 6; Representing the The distance of the head section of the strip steel.

[0017] Preferably, the RBF neural network training model includes an input layer, a hidden layer, and an output layer.

[0018] Preferably, during the construction of the hidden layer, the structure of the hidden layer is determined by Gaussian radial basis functions and clustering algorithms:

[0019]

[0020] Where G represents the Gaussian radial basis function, Let X be the variance, and let X be the center of the hidden layer nodes in the X network. p Let p be the p-th input sample.

[0021] Preferably, the elbow method achieves a balance between the number of cluster centers and clustering performance in the clustering algorithm, specifically as follows:

[0022]

[0023] in It is the first There are clusters, i = 1, 2, 3…h. is the sample in is the centroid of the sample, SSE is the error of all samples;

[0024] When the elbow point of the SSE function is reached, the number of clustering centers and the clustering performance reach a balance, and the elbow point is a point at which the rate of decrease of the SSE begins to slow down.

[0025] Preferably, in the RBF neural network training model, the weights from the hidden layer to the output layer are adjusted by a back propagation algorithm, and specifically:

[0026]

[0027] wherein is an output value, is an actual value, represents the weight of the i-th hidden layer neuron to the j-th output layer neuron, is the value of the last iteration, is the value of the last iteration, is the value of the last iteration, is the value of the last iteration, and respectively represent a current learning rate and a historical learning rate, and take values of 0.1-0.3.

[0028] A hot rolling coiler strip head monitoring end partition system based on the hot rolling coiler strip head monitoring end partition method, comprising an information acquisition module, a data processing module and a neural network construction module;

[0029] The information acquisition module acquires position information of the strip head;

[0030] The data processing module stores and processes information collected by the information acquisition module;

[0031] The neural network construction module constructs and trains a neural network, and the trained neural network is used to predict the position of the strip head.

[0032] Preferably, the data processing module receives a prediction result from the neural network construction module, judges the prediction result, and if the error between the prediction result and the actual result is within a preset range, the result is retained and output, otherwise a signal is sent to the neural network construction module, and the neural network construction module re-trains the model and the prediction result.

[0033] The present application has the following advantages:

[0034] The present application proposes to set several monitoring points, and set information collection devices at the monitoring points, construct and train the RBF neural network through the collected information, which overcomes the problems of uncertain sensor accuracy, large difference and low utilization rate in the prior art, can be performed offline according to production data, can be trained without affecting normal production, and the trained parameters are imported into the online system for use, which overcomes the winding machine quality problem of AJC failure caused by inaccurate tracking of the strip head. BRIEF DESCRIPTION OF DRAWINGS

[0035] The drawings accompanying the specification of this application form a part thereof, serve to provide further understanding of the application, and together with the description of the exemplary embodiments of the application, serve to explain the application, and do not constitute an improper limitation on the application. In the drawings:

[0036] Figure 1 It is a hot rolling coiler strip head monitoring end zoning method flow chart. DETAILED DESCRIPTION

[0037] The application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments. It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0038] The following detailed description is exemplary description, which is intended to provide further detailed description of the application. Unless otherwise specified, all technical terms used in the application have the same meaning as understood by those skilled in the art. The terms used in the application are only for the purpose of describing the specific embodiments, and are not intended to limit the exemplary embodiments according to the application.

[0039] Please refer to Figure 1 The present application is a hot rolling coiler strip head monitoring end zoning method, which specifically comprises the following steps:

[0040] Step 1, setting monitoring points: seven monitoring points are set in sequence on the path of the strip head, the first monitoring point, the second monitoring point, the third monitoring point, the fourth monitoring point, the fifth monitoring point, the sixth monitoring point and the seventh monitoring point;

[0041] A thermal detection sensor, a pressure sensor, an industrial camera and a laser detection sensor are arranged in sequence at the first monitoring point, the second monitoring point, the third monitoring point and the fourth monitoring point; pressure sensors are arranged at the fifth monitoring point, the sixth monitoring point and the seventh monitoring point.

[0042] The second monitoring point is arranged at the pinch roll feedback point on the strip head path, and the fifth monitoring point, the sixth monitoring point and the seventh monitoring point are arranged at the first coiler feedback point, the second coiler feedback point and the third coiler feedback point respectively.

[0043] The pinch roll feedback point receives a pinch roll pressure feedback fluctuation signal, and the first coiler feedback point, the second coiler feedback point and the third coiler feedback point are respectively used for receiving coiler pressure feedback fluctuation signals.

[0044] The above monitoring points divide the strip head path into six monitoring sections, which are the first monitoring section, the second monitoring section, the third monitoring section, the fourth monitoring section, the fifth monitoring section and the sixth monitoring section.

[0045] Step 2, information collection of the strip head position: the time point when the strip head reaches the hot metal detector is t1, the time point when the strip head reaches the pinch roll feedback point to receive the pinch roll pressure feedback fluctuation signal is t2, the time point when the strip head reaches the industrial camera is t3, the time point when the strip head reaches the laser detector is t4, the time point when the strip head reaches the first coiler feedback point to receive the first coiler pressure feedback fluctuation signal is t5, the time point when the strip head reaches the second coiler feedback point to receive the second coiler pressure feedback fluctuation signal is t6, and the time point when the strip head reaches the third coiler feedback point to receive the third coiler pressure feedback fluctuation signal is t7. .

[0046] According to the above time points, the average speeds of the six monitoring sections can be obtained, which are used as the input of the neural network and the strip head position for neural network training error calculation:

[0047]

[0048] The v n is the average speed of the nth monitoring section, n=1, 2, 3, 4, 5, 6; and the represents the distance of the strip head path of the nth monitoring section.

[0049] The strip head position is calculated in real time according to the average speed, and as the strip head travels, more and more road sections are passed through, which affect the subsequent tracking strategy.

[0050] In this embodiment, according to the practical needs, the final time point is t8, the final time point t8 , the control is in the normal strip head has not reached the coiler to start step.

[0051] The time period is the data acquisition stage, and the training model with time difference as the input is constructed, and the network input is:

[0052] ​​​​​​​​

[0053] x is a dataset, x n This represents the time difference between adjacent sampling points detecting the position of the strip head, i.e., t. n With t n-1 The difference between them, n=1, 2, 3, 4, 5, 6.

[0054] Step 3, Constructing the Neural Network Training Model: The neural network is an RBF neural network training model, including an input layer, hidden layers, and an output layer; the mapping from input to output is non-linear, possessing the ability to approximate continuous functions of arbitrary precision. An RBF neural network model is used to establish a function mapping model between the input time difference and the output speed compensation coefficient.

[0055] After verifying the effect of multiple automatic stepping work cycles of the strip head, it is determined whether the neural network training model can achieve accurate prediction. If it cannot, training continues until the error of the calculation output of the verification group is within the ideal range, thus achieving accurate prediction of the strip head position.

[0056] In one specific implementation, the structure of the hidden layer is determined by Gaussian radial basis function and clustering algorithm, and the balance between the number of cluster centers and clustering performance in the clustering algorithm is achieved by elbow method;

[0057] The Gaussian radial basis function is:

[0058]

[0059] in, Let X be the variance, and let X be the center of the hidden layer nodes in the X network. p Let p be the p-th input sample.

[0060] The elbow method is specifically as follows:

[0061]

[0062] in It is the first There are clusters, i = 1, 2, 3…h. yes The samples in yes The centroid of the sample is SSE, which is the error of all samples.

[0063] When the SSE function reaches the "elbow point," the number of cluster centers and clustering performance reach a balance. The elbow point is the point where the rate of decline of SSE begins to slow down.

[0064] variance for:

[0065]

[0066] wherein is the maximum distance between the selected centers.

[0067] In a specific embodiment, the weights of the hidden layer to the output layer are set by the back propagation algorithm, specifically:

[0068]

[0069] wherein is the output value, is the actual value, represents the weight of the i-th hidden layer neuron to the j-th output layer neuron, is the value of the last iteration, and respectively represent the current learning rate and the historical learning rate, usually taking values of 0.1~0.3. In a specific embodiment, a strip head position prediction model is constructed according to the trained neural network, the speed weight coefficient and the restriction term are determined, and the prediction result that makes the error calculation value minimum is selected as the strip head tracking signal of the controller,

[0070] In a specific embodiment, in order to establish the relationship between the speed coefficient output by the neural network and the strip head tracking, after the speed coefficient is calculated, an integral layer is added to calculate the tracking position of the strip head, and the manual setting error is used as the criterion to calculate the error and input it into the back propagation algorithm to compensate the weights between the hidden layer and the output layer.

[0071] In a specific embodiment, the learning time of the neural network is determined by the number of iterations of the neural network (the number of data groups collected), the order of combination of the iteration data groups, the number of times of back propagation of the back propagation algorithm in each iteration, the influence coefficient of the current time back propagation on the calculation of the weights, and the influence strategy of the historical time weights on the calculation of the weights of the current time back propagation.

[0072] In a specific embodiment, after the controller writes the neural network parameters, online verification is performed in actual production, the tracking error of the strip head is continuously recorded, and the latest production data is imported into the offline system for continuous training until the calculation error of the model verification written into the online system is within the ideal range, and the accurate prediction of the strip head position is realized.

[0073] In a specific embodiment, after the controller writes the neural network parameters, online verification is performed in actual production, the tracking error of the strip head is continuously recorded, and the latest production data is imported into the offline system for continuous training until the calculation error of the model verification written into the online system is within the ideal range, and the accurate prediction of the strip head position is realized.

[0074] The application also provides a hot rolling coiler strip head monitoring end partition system, which comprises an information acquisition module, a data processing module and a neural network construction module.

[0075] ​​The information acquisition module comprises a thermal detection sensor, four pressure sensors, an industrial camera and a laser detection sensor, wherein the thermal detection sensor, the four pressure sensors, the industrial camera and the laser detection sensor are arranged on the path of the strip head.

[0076] The data processing module receives the position information of the strip head acquired by the information acquisition module and transmits the processed information to the neural network construction module as the input of the neural network.

[0077] The neural network construction module constructs a neural network and trains the input content from the data processing module through the neural network, and after the training is completed, the position of the strip head is predicted.

[0078] In a specific embodiment, the data processing module obtains the edge information of the strip head from the original image information acquired by the industrial camera, and judges the prediction result of the neural network construction module based on the data processing module, and if the calculation error of the model verification is within the ideal range, the result is accepted, and if not, the model is retrained.

[0079] The RBF neural network-based strip head tracking method of the hot rolling coiler provided by the present application overcomes the technical problems of uncertain sensor accuracy, large difference and low utilization rate in the prior art, and the training process of the RBF neural network can be performed offline according to production data, which can be trained without affecting normal production, and the training parameters are imported into the online system for use, overcoming the quality problem of the coiler caused by AJC failure due to inaccurate tracking of the strip head. Greatly improve product quality, ensure real-time and computing power, avoid disturbance of complex working conditions to the control system, maintain system stability, improve production efficiency, and the algorithm complexity is relatively low.

[0080] In summary, the strip head tracking and optimization cutting method provided by the present application can realize online, real-time, automatic and non-contact measurement of the strip head, realize the information acquisition and information processing functions of the sensor. Through the fitting ability of the RBF neural network to any function, the tracking of different types of strip head is realized. The system is fast, stable and accurate in tracking effect when used online, and can adapt to harsh working conditions, improve production efficiency and product quality, effectively reduce labor intensity, reduce tracking failure rate, improve automatic step control effect, and meet the needs of strip process.

[0081] It is apparent that the application can be implemented in a variety of other forms, without departing from the spirit or essential characteristics thereof. The disclosed embodiments are to be considered in all respects only as illustrative and not restrictive. The scope of the application is indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are intended to be embraced therein.

[0082] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code thereon.

[0083] The present application is described in reference to the flowchart and / or block diagrams of the method, apparatus (system) and computer program product according to embodiments of the application. It is understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing system or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for performing the function specified by the flowchart and / or block diagram block or blocks.

[0084] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for performing the function specified by the flowchart and / or block diagram block or blocks.

[0085] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for performing the function specified by the flowchart and / or block diagram block or blocks.

[0086] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application but not to limit it. Although the present application has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that the specific embodiments of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A method of end zone monitoring for hot strip head in a hot strip mill, characterized by, It comprises the following steps: Step 1: monitoring points are set on the path of the strip head: seven monitoring points are set, which divide the strip head path into six monitoring sections, namely the first monitoring section, the second monitoring section, the third monitoring section, the fourth monitoring section, the fifth monitoring section and the sixth monitoring section; information acquisition devices are set at the seven monitoring points; Step 2: the monitoring points acquire information of the strip head position: a data set is established according to the information obtained in step 1, which is used as the input of the neural network; Step 3: an RBF neural network training model is constructed, and the neural network training model is trained according to the data set from step 2 until the error between the predicted result and the actual result is within the preset range.

2. A method of end zone monitoring of a hot strip head in a hot strip mill as defined in claim 1, characterized in that In step 1, a thermal detection sensor, a pressure sensor, an industrial camera and a laser detection sensor are sequentially arranged at the first monitoring point, the second monitoring point, the third monitoring point and the fourth monitoring point; pressure sensors are arranged at the fifth monitoring point, the sixth monitoring point and the seventh monitoring point.

3. A method of end zone monitoring of a hot strip head in a hot strip mill as defined in claim 2, characterized in that The second monitoring point is a pinch roll feedback point on the strip head path, and the fifth monitoring point, the sixth monitoring point and the seventh monitoring point are respectively a first coiler feedback point, a second coiler feedback point and a third coiler feedback point.

4. A method of end zone monitoring of a hot strip head in a hot strip mill as defined in claim 1, characterized in that, In step 2, the average speed of the six monitoring sections is obtained according to the time when the strip head passes through the monitoring points: t0, t1, t2, t3, t4, t5 and t6 are respectively set as the time when the strip head reaches each monitoring point, and the following is obtained: The v n The average speed of the nth monitoring section, n = 1, 2, 3, 4, 5, 6; the The average speed of the nth monitoring section, n = 1, 2, 3, 4, 5, 6; the The distance of the strip head path of the nth monitoring section, n = 1, 2, 3, 4, 5, 6.

5. A method of end zone monitoring of a hot strip head in a hot strip mill as defined in claim 1, characterized in that The RBF neural network training model comprises an input layer, a hidden layer and an output layer.

6. A method of end zone monitoring of a hot strip head in a hot strip mill as defined in claim 5, characterized in that In the process of constructing the hidden layer, the structure of the hidden layer is determined by the Gaussian radial basis function and the clustering algorithm: where G denotes a Gaussian radial basis function, is the variance, X the center of the hidden layer node of the X network, p is the p-th input sample.

7. A method of end zone monitoring of a hot strip head in a hot strip mill as defined in claim 6, characterized in that The elbow method is used to balance the number of cluster centers and the clustering performance in the clustering algorithm, specifically as follows: in It is the first There are clusters, i = 1, 2, 3…h. yes The samples in yes The centroid of the sample is SSE, which is the error of all samples. When the SSE function reaches the elbow point, the number of cluster centers and the clustering performance reach balance, and the elbow point is the point at which the descending rate of SSE begins to slow down.

8. A method of end zone monitoring of a hot strip head in a hot strip mill as defined in claim 7, characterized in that In the RBF neural network training model, the weights from the hidden layer to the output layer are adjusted by the back propagation algorithm, specifically as follows: wherein is the output value, is the actual value, denotes the weight of the jth hidden layer neuron to the ith output layer neuron, is the value at the previous iteration, is the value at the previous iteration, is the value at the previous iteration, and denote the current learning rate and the historical learning rate, respectively, and take values between 0.1 and 0.3.​​ 9. A hot strip mill coiler strip head monitoring end zoning system based on the hot strip mill coiler strip head monitoring end zoning method of any of claims 1-8, characterized by, It comprises an information acquisition module, a data processing module and a neural network construction module; The information acquisition module acquires the position information of the strip head; The data processing module stores and processes the information collected by the information acquisition module; The neural network construction module constructs and trains the neural network, and the trained neural network is used to predict the position of the strip head.

10. A hot strip mill reel strip head monitoring end zoning system as claimed in claim 9 wherein, The data processing module receives the prediction result from the neural network construction module and judges the prediction result, if the error between the prediction result and the actual result is within the preset range, the result is retained and output, otherwise a signal is sent to the neural network construction module, and the neural network construction module re-trains the model and the prediction result.

Images