Simulation weight-fixed cutting data processing method, system, equipment and medium based on cognitive intelligent process model

Abstract

A method for processing simulated fixed-weight cutting data based on a cognitive intelligent process model includes: installing a hydraulic lifting platform scale at each stream at the conveying roller of the cooling bed of a multi-stream continuous casting machine; after each weighing, the platform scale weighing data is fed back to the fixed-weight cutting model; the fixed-weight cutting model uses a mathematical calculation formula to calculate the length of the steel billet; the steel billet length information is sent to a flame cutting machine for cutting to obtain the cut steel billet; the cut steel billet is weighed again by the platform scale to calculate the length required for the next steel billet cutting; ensure that the steel billet cutting is continuous and cyclical, and collect a series of parameters in the above process; and construct a neural network model to predict the simulated weight of the steel billet to replace the platform scale weighing data. The effect of simulated fixed-weight cutting is achieved, thereby ensuring the accuracy of cutting.

Description

Technical Field

[0001] The present invention belongs to the field of data modeling, and in particular relates to a method for processing simulated fixed-weight cutting data based on a cognitive intelligent process model. Background Art

[0002] At present, there are mainly the following fixed weight cutting methods:

[0003] The Chinese invention patent with application number CN201910336630.9 proposes a method and system for cutting with fixed weight for a continuous casting machine, including: first determining the standard weight of the billet, calculating and setting the standard length of the billet according to the standard weight, weighing the billet cut according to the set standard length of the billet, comparing the weighed weight with the standard weight, and adjusting the cutting length of the next billet according to the comparison result; the system includes a fixed-length cutting module, a weighing track control module, a fixed-weight control module, and a root number and weight statistics module, the fixed-length cutting module is used to cut the billet, the weighing roller control module is used to obtain the weighed weight of the billet, the fixed-weight control module is used to compare the weighed weight with the standard weight, calculate the most reasonable cutting position at present, and pass the cutting position information to the fixed-length cutting module for billet cutting. The present invention effectively improves the fluctuation of the length of the rolled steel, improves the yield rate of rolled steel, and reduces the labor intensity of employees. However, the technical characteristics of the fixed-weight cutting model mainly adopt the method of entering the fixed-weight control module with the weight value for fixed-weight cutting, and the practice of calculating the cognitive neural network model is not involved.

[0004] The Chinese patent with application number CN202011165738.5 provides a continuous casting billet fixed weight cutting control system and method, including: a continuous casting information acquisition device for acquiring continuous casting information related to the continuous casting machine; a temperature measuring device for acquiring the temperature distribution information of the current billet being cut; a weighing device for measuring the weight of the current billet; a control module for determining the target cutting length of the next billet according to the continuous casting information, the temperature distribution information of the current billet and the weight of the current billet; a cutting module for cutting the next billet according to the target cutting length. The present invention integrates the billet temperature distribution information when performing the billet fixed weight cutting, which is conducive to improving the accuracy of setting the next billet cutting length, thereby effectively improving the accuracy of the continuous casting billet fixed weight cutting. However, this invention patent uses a weighing machine as the target cutting method to perform the corresponding fixed weight cutting, but it is not applicable to the simulated fixed weight cutting method. Summary of the invention

[0005] In view of the above problems, in order to improve the yield of square billets, a fixed weight cutting method is adopted. This method uses a weighing machine to accurately weigh the square billet after each casting, and adjusts the next cutting length according to the result to ensure that the weight of the square billet cut each time is maintained within a stable range. However, under long-term operation, the weighing machine requires frequent maintenance, and its precision and the accuracy of the weighing instrument may decrease over time. These hardware factors often lead to inaccurate cutting results. To solve this problem, the present invention innovatively proposes a simulated fixed weight cutting data processing method based on a cognitive intelligent process model to achieve the effect of simulated fixed weight cutting, thereby ensuring the accuracy of cutting, specifically including:

[0006] S1: First, install a hydraulic lifting scale on each stream at the conveying roller table of the cooling bed of the 8-machine 8-stream continuous casting machine. To ensure the production rhythm, the lifting time of the hydraulic lifting scale is controlled within 20S.

[0007] S2: After each hydraulic platform scale is lifted and weighed, the weighing data is fed back to the fixed weight cutting model. The fixed weight cutting model uses the following mathematical calculation method to calculate the length:

[0008] ρ*L*S=H

[0009] Wherein, H is the weight of the hydraulic scale; ρ is the density of the billet; L is the length of the billet; S is the cross-sectional area of ​​the billet;

[0010] H and S are known, ρ is the theoretical density, and the size of L is obtained by calculation. The length of the steel billet is calculated iteratively each time through H;

[0011] The size of ρ is related to the billet drawing speed, billet nozzle size, billet cooling water volume, billet cooling temperature, etc. The calculation method is detailed in the subsequent steps.

[0012] S3: After the length of the billet is calculated, the length information is sent to the flame cutting machine for cutting. After each cutting, it is re-weighed by the billet scale and the next billet cutting length is calculated.

[0013] This cycle ensures that the weight of the billet after cutting is stable within a very small range, achieving fixed weight cutting.

[0014] S4: After the fixed weight cutting is achieved and the program runs continuously for 3 to 4 months, the information such as the billet weight, billet number, billet drawing speed, billet nozzle size, billet cooling water volume, billet cooling temperature and secondary cooling water volume measured each time is collected during the process to obtain process data on billet production within 3 to 4 consecutive months.

[0015] S5: By studying the dynamic relationship between the above variables and ρ, artificial intelligence data modeling is performed between the above process parameters and actual weighing to accurately predict the weight of the steel billet after each production to replace the data of the weighing machine.

[0016] The specific steps include:

[0017] (1) Analyze the steel billets in stream 1, align the data of each steel billet in stream 1 in time, granulate the billet drawing speed in time, granulate the drawing speed into 100 parts according to the time required for the billet to be produced in stream 1 (about 10 minutes), obtain the list value of the drawing speed [X,X,X,X…,X], and then use the list value of the drawing speed as the input of the neural network.

[0018] (2) The same method as above is adopted for the process parameters such as billet nozzle size, billet cooling water volume, billet cooling temperature and secondary cooling water volume. The production time of each billet is granulated and then the list value is obtained for neural network input. Time granulation means: sampling continuous time and then discretizing it.

[0019] (3) The neural network adopts a multi-dimensional feedforward neural network, which is different from the ordinary feedforward neural network in that the neural network adopts the form of multi-dimensional numerical input. The matrix operation input expression of f(WX+B)+δY is used inside the neural network to perform neural network reasoning, where X represents the input of a single node of the neural network, W represents the weight, B represents the offset, which is used to fit the neural network, and δY is used as a compensation item to correct the neural network model. The inputs are the list of drawing speeds of each stream and each billet (100), the list of billet nozzle sizes (100), the list of billet cooling water volumes (100), and the list of billet cooling temperatures (100). Four sets of numerical values ​​are taken as the input of a single node of the neural network for calculation.

[0020] Each input value adjusts the proportion of input data by adjusting the weight w. The setting of weight w is calculated accordingly through the cognitive computing engine. The calculation mode mainly uses the steel grade process weight calculation library to perform corresponding calculations. The relationship between the steel grade list and the nozzle size, cooling water volume, and cooling temperature is used as the basis for calculation. The cognitive computing engine calculates the list for corresponding analysis and calculation to obtain the weight value of the process parameters under the precise steel grade, and then enters the corresponding neural network model calculation.

[0021] (4) The output of the neuron uses the weight value obtained by comparative weighing as the output of the neural network.

[0022] (5) The weighing data AY11, AY21, and AY31 obtained by weighing are fed back to the neurons, and δY is used as a compensation item to correct the model.

[0023] (6) The model is trained using the data of each billet in each stream, and a total of 12,000 billets in about three months of production are used to train the model. The neural network parameter values ​​under the fitting effect are obtained, the values ​​of W and B are obtained, and the simulated neural network data model for this stream is obtained.

[0024] (7) During the training process, the weighing data of the body is compared with the simulated weighing data after the simulated fixed weight cutting data, and the model is dynamically refined and adjusted step by step. At the same time, continuous training is performed to stabilize the W and B values, and the model is stabilized by finally determining the size of the W and B values.

[0025] (8) The weighing data of the scale is separated from the feedback link, and then connected to the actual production data, and the model output value of the simulated neural network data is inferred to obtain the weight data of the simulated steel billet, and the corresponding cutting optimization information feedback is performed, thereby achieving the effect of simulating fixed weight cutting optimization.

[0026] Based on the above-mentioned simulated weight-fixed cutting data processing method based on cognitive intelligent process model, in order to better realize the present invention, a simulated weight-fixed cutting data processing system based on cognitive intelligent process model is further proposed, which is used to realize the above-mentioned simulated weight-fixed cutting data processing method based on cognitive intelligent process model.

[0027] Based on the above-mentioned simulated weight-fixed cutting data processing method based on the cognitive intelligent process model, in order to better realize the present invention, an electronic device is further proposed, including a memory and a processor; a computer program is stored on the memory; when the computer program is executed on the processor, the above-mentioned simulated weight-fixed cutting data processing method based on the cognitive intelligent process model is implemented.

[0028] Based on the above-mentioned simulated weight-fixed cutting data processing method based on the cognitive intelligent process model, in order to better realize the present invention, a computer-readable storage medium is further proposed, on which computer instructions are stored; when the computer instructions are executed on the above-mentioned electronic device, the above-mentioned simulated weight-fixed cutting data processing method based on the cognitive intelligent process model is implemented.

[0029] Beneficial effects:

[0030] At the cooling bed conveyor roller of the casting machine, a hydraulic lifting scale is installed on each flow to accurately measure the weight of the billet. However, the weighing of the feedback scale is cancelled, and a simulation model is used instead for fixed weight cutting. After each weighing by the hydraulic scale, the relevant data will be quickly fed back to the fixed weight cutting model, which uses a unique mathematical calculation method to accurately calculate the required cutting length. Furthermore, by studying the dynamic relationship between process parameters and actual weighing, these relationships are used to establish an accurate artificial intelligence data model. This model can accurately predict the weight of the billet after each production, realizing a more efficient and intelligent production process. The present invention not only eliminates the influence of the weighing machine on the production process, but also uses virtual weighing data as signal feedback to achieve the effect of simulated weighing and cutting, significantly improving production efficiency and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a multi-dimensional feedforward neural network in a simulated fixed-weight cutting data processing method based on a cognitive intelligent process model. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0034] At present, there are two main cutting modes in the continuous casting industry production line, namely fixed length and fixed weight. The fixed length method often affects the range of the scale, resulting in a low product yield at the rolling mill. Therefore, in order to improve the yield of square billets, a fixed weight cutting method is adopted, that is, after each casting, a weighing machine is used to weigh the square billet to correct the next cutting length, so that the weight of the square billet cut each time is within a stable range. However, the weighing machine requires long-term maintenance, and the accuracy of the weighing machine and the weighing instrument will decrease over time. Often, hardware will cause inaccurate factors in each cutting. Therefore, the present invention adopts a simulated fixed weight cutting method to perform data modeling, which can eliminate the weighing influence of the weighing machine, achieve the effect of virtual weighing data as signal feedback, and realize the effect of simulated fixed weight and cutting.

[0035] Here’s how:

[0036] S1: First, install a hydraulic lifting scale on each stream at the conveying roller table of the cooling bed of the 8-machine 8-stream continuous casting machine. To ensure the production rhythm, the lifting time of the hydraulic lifting scale is controlled within 20S.

[0037] S2: After each hydraulic platform scale is lifted and weighed, the weighing data is fed back to the fixed weight cutting model. The fixed weight cutting model uses the following mathematical calculation method to calculate the length:

[0038] ρ*L*S=H

[0039] Wherein, H is the weight of the hydraulic scale; ρ is the density of the billet; L is the length of the billet; S is the cross-sectional area of ​​the billet;

[0040] H and S are known, ρ is the theoretical density, and the size of L is obtained by calculation. The length of the steel billet is calculated iteratively each time through H;

[0041] The size of ρ is related to the billet drawing speed, billet nozzle size, billet cooling water volume, billet cooling temperature, etc. The calculation method is detailed in the subsequent steps.

[0042] S3: After the length of the billet is calculated, the length information is sent to the flame cutting machine for cutting. After each cutting, it is re-weighed by the billet scale and the next billet cutting length is calculated.

[0043] This cycle ensures that the weight of the billet after cutting is stable within a very small range, achieving fixed weight cutting.

[0044] S4: After the fixed weight cutting is achieved and the program runs continuously for 3 to 4 months, the information such as the billet weight, billet number, billet drawing speed, billet nozzle size, billet cooling water volume, billet cooling temperature and secondary cooling water volume measured each time is collected during the process to obtain process data on billet production within 3 to 4 consecutive months.

[0045] S5: By studying the dynamic relationship between the above variables and ρ, artificial intelligence data modeling is performed between the above process parameters and actual weighing to accurately predict the weight of the steel billet after each production to replace the data of the weighing machine.

[0046] The specific steps include:

[0047] (1) Analyze the steel billets in stream 1, align the data of each steel billet in stream 1 in time, granulate the billet drawing speed in time, granulate the drawing speed into 100 parts according to the time required for the billet to be produced in stream 1 (about 10 minutes), obtain the list value of the drawing speed [X,X,X,X…,X], and then use the list value of the drawing speed as the input of the neural network.

[0048] (2) The same method as above is adopted for the process parameters such as billet nozzle size, billet cooling water volume, billet cooling temperature and secondary cooling water volume. The production time of each billet is granulated and then the list value is obtained for neural network input. Time granulation means: sampling continuous time and then discretizing it.

[0049] (3) Figure 1 As shown, the neural network adopts a multi-dimensional feedforward neural network, which is different from the ordinary feedforward neural network in that the neural network adopts the form of multi-dimensional numerical input. The matrix operation input expression of f(WX+B)+δY is used inside the neural network to perform neural network reasoning, where X represents the input of a single node of the neural network, W represents the weight, B represents the offset, which is used to fit the neural network, and δY is used as a compensation item to correct the neural network model. The inputs are the list of drawing speeds of each stream and each billet (100), the list of billet nozzle sizes (100), the list of billet cooling water volume (100), and the list of billet cooling temperature (100). Four sets of values ​​are taken as the input of a single node of the neural network for calculation.

[0050] Each input value adjusts the proportion of input data by adjusting the weight w. The setting of weight w is calculated accordingly through the cognitive computing engine. The calculation mode mainly uses the steel grade process weight calculation library to perform corresponding calculations. The relationship between the steel grade list and the nozzle size, cooling water volume, and cooling temperature is used as the basis for calculation. The cognitive computing engine calculates the list for corresponding analysis and calculation to obtain the weight value of the process parameters under the precise steel grade, and then enters the corresponding neural network model calculation.

[0051] (4) The output of the neuron uses the weight value obtained by comparative weighing as the output of the neural network.

[0052] (5) The weighing data AY11, AY21, and AY31 obtained by weighing are fed back to the neurons, and δY is used as a compensation item to correct the model.

[0053] (6) The model is trained using the data of each billet in each stream, and a total of 12,000 billets in about three months of production are used to train the model. The neural network parameter values ​​under the fitting effect are obtained, the values ​​of W and B are obtained, and the simulated neural network data model for this stream is obtained.

[0054] (7) During the training process, the weighing data of the body is compared with the simulated weighing data after the simulated fixed weight cutting data, and the model is dynamically refined and adjusted step by step. At the same time, continuous training is performed to stabilize the W and B values, and the model is stabilized by finally determining the size of the W and B values.

[0055] (8) The weighing data of the scale is separated from the feedback link, and then connected to the actual production data, and the model output value of the simulated neural network data is inferred to obtain the weight data of the simulated steel billet, and the corresponding cutting optimization information feedback is performed, thereby achieving the effect of simulating fixed weight cutting optimization.

[0056] This embodiment also proposes a simulated weight-fixed cutting data processing system based on a cognitive intelligent process model, which is used to implement the above-mentioned simulated weight-fixed cutting data processing method based on a cognitive intelligent process model.

[0057] This embodiment also proposes an electronic device, including a memory and a processor; a computer program is stored in the memory; when the computer program is executed on the processor, the above-mentioned simulated fixed-weight cutting data processing method based on the cognitive intelligent process model is implemented.

[0058] This embodiment also proposes a computer-readable storage medium, on which computer instructions are stored; when the computer instructions are executed on the above-mentioned electronic device, the above-mentioned simulated fixed-weight cutting data processing method based on the cognitive intelligent process model is implemented.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person having ordinary knowledge in the art can still modify the technical solutions recorded in the aforementioned embodiments, or replace part or all of the technical features therein with equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A method for processing simulated fixed-weight cutting data based on a cognitive intelligent process model, characterized in that: include: A hydraulic lifting scale is installed on each strand at the conveying roller table of the cooling bed of the multi-strand continuous casting machine; After each weighing, the platform scale feeds back the weighing data to the fixed weight cutting model; The fixed weight cutting model uses a mathematical formula to calculate the length of the steel billet; The mathematical calculation formula is: ρ×L×S=H, where H is the weight of the hydraulic scale, ρ is the density of the steel billet, L is the length of the steel billet, and S is the cross-sectional area of ​​the steel billet; The size of L is calculated, and the length of the billet is calculated iteratively each time through H; Send the length information of the steel billet to the flame cutting machine, cut the steel billet, and obtain the remaining steel billet after cutting; Weighing the remaining steel billet after cutting on the platform scale to calculate the length required for the next steel billet cutting; Ensure that billet cutting is continuous and cyclical, and collect process parameters; The parameters include: billet weight, billet number, billet drawing speed, billet nozzle size, billet cooling water volume, billet cooling temperature and secondary cooling water volume, billet production process data information; Construct a neural network model to predict the simulated weight of the billet to replace the weighing data of the scale. The specific steps include: Analyze the parameters of each stream, align the data of the parameters of each stream in time, granulate the billet drawing speed, billet nozzle size, billet cooling water volume, billet cooling temperature in time, and obtain the list values ​​of each parameter; Use the values ​​of each parameter list as the input of a single node of the neural network; The neural network model adopts a multi-dimensional feedforward neural network; The neural network model uses a matrix operation input expression of f(WX+B)+δY to perform neural network reasoning, where X represents the input of a single node of the neural network, W represents the weight, B represents the offset, and is used to fit the neural network. δY is used as a compensation term to correct the neural network model. The output of the neural network model is the weighing value of the platform scale weighing data; By training the neural network model on the data of each billet of each stream, the neural network parameter values, W and B values ​​under the fitting effect are obtained; During the training of the neural network model, the weighing data of the platform scale is compared with the simulated weight of the steel billet, the neural network model is dynamically adjusted, the values ​​of W and B are determined, and the neural network model is stabilized; The weighing data of the platform scale is separated from the feedback link, connected to the actual production data, and the simulated weight of the steel billet is output to optimize the information feedback.

2. The method for processing simulated fixed-weight cutting data based on a cognitive intelligent process model according to claim 1 is characterized in that: The lifting time of the hydraulic lifting scale is controlled within 20S.

3. The method for processing simulated fixed-weight cutting data based on a cognitive intelligent process model according to claim 1 is characterized in that: Artificial intelligence data modeling between the parameters and actual weighing is used to accurately predict the weight of the steel billet after each production to replace the data of the scale.

4. A simulated fixed-weight cutting data processing system based on a cognitive intelligent process model, characterized in that: The system is used to implement the simulated fixed-weight cutting data processing method based on the cognitive intelligent process model as described in any one of claims 1-3.

5. An electronic device, characterized in that: It comprises a memory and a processor; a computer program is stored in the memory; when the computer program is executed on the processor, the simulated fixed-weight cutting data processing method based on the cognitive intelligent process model as described in any one of claims 1 to 3 is implemented.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions; when the computer instructions are executed on the electronic device as described in claim 5, the simulated fixed-weight cutting data processing method based on the cognitive intelligent process model as described in any one of claims 1-3 is implemented.

Images