Argon flow and micro-positive pressure linkage control method and system for electroslag furnace

Abstract

The invention provides an argon flow and micro-positive pressure linkage control method for an electroslag furnace. The method comprises the steps that the oxygen content and micro-positive pressure signals in the furnace are collected; the future furnace condition is predicted through the deep belief network model; optimizing initial parameters of the fuzzy PID controller in an off-line manner by using a particle swarm algorithm; on the basis of the real-time signal and the prediction result, adaptively adjusting controller parameters on line; and finally, an argon flow and dust removal valve opening control instruction is output in a linkage mode. According to the method, future furnace conditions are predicted through the deep belief network model, the control advancement is realized, and the problem of large lag of the system is solved; offline parameter optimization is carried out on the fuzzy PID controller through a particle swarm algorithm, and excellent initial performance adapting to different working conditions is provided for the fuzzy PID controller; the parameters of the controller are adaptively adjusted through online fuzzy reasoning, so that the system has robustness for coping with a dynamic process; finally, accurate and stable linkage control of the oxygen content in the furnace and the micro-positive pressure is achieved through linkage calculation of the argon flow and a dust removal valve opening instruction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of metallurgy, and in particular to an electric-shock furnace argon flow and micro-positive pressure linkage control method and system. BACKGROUND

[0002] In the electric-shock remelting smelting process, argon is filled into the furnace and a stable micro-positive pressure environment is maintained, which is a key process for preventing oxidation of the steel liquid, reducing element burnout and improving the purity of the ingot. At the same time, adjusting the exhaust capacity through a dust removal proportional valve is a core means for maintaining the above micro-positive pressure and removing harmful dust generated during the smelting process. At present, the traditional PID control or fixed flow mode is generally used in the industry to adjust the argon filling process, and the dust removal system also operates independently.

[0003] However, such methods have significant defects: Firstly, the control mode is rigid and lagging, and the control parameters of argon flow and dust removal valve are mostly dependent on manual experience setting or remain fixed, which cannot be adjusted in real time according to the dynamic changes of oxygen content and micro-positive pressure in the furnace, resulting in a mismatch between argon supply and actual demand, and the dust removal system cannot respond synchronously, causing fluctuations in the furnace atmosphere and insufficient control accuracy.

[0004] Secondly, the argon flow control and dust removal system pressure regulation are independent of each other, and no collaborative linkage mechanism is established between the two control loops, resulting in mutual interference during the adjustment process and high risk of system oscillation.

[0005] Finally, the control instructions of the existing system completely depend on the feedback signals at the current time, and have no prediction ability for the dynamic changes and large lag characteristics of the smelting process, and cannot implement advance compensation control, especially when the working conditions are switched, the system response is slow. SUMMARY

[0006] The present application proposes an electric-shock furnace argon flow and micro-positive pressure linkage control method and system, which solves the technical problems of insufficient control accuracy of oxygen content and micro-positive pressure in the furnace, response lag and system oscillation caused by independent control of argon and dust removal system, lack of prediction ability and parameter fixation of the existing electric-shock furnace control system.

[0007] The technical solution of the present application is as follows: The first aspect of the present application provides an electric-shock furnace argon flow and micro-positive pressure linkage control method, comprising the following steps: S1, collecting real-time oxygen content signals and micro-positive pressure signals in the electric-shock furnace; S2, predicting the oxygen content and micro-positive pressure in the furnace at future time based on the collected real-time signals through a deep belief network DBN model; S3, using a particle swarm PSO algorithm to optimize the initial parameters of the fuzzy PID controller offline; S4, initializing the fuzzy PID controller with the optimized initial parameters, and adaptively adjusting parameters of the fuzzy PID controller based on the real-time signal and the prediction result; S5, calculating and outputting the argon flow control instruction and the dedusting valve opening control instruction according to the adjusted fuzzy PID controller parameters, so as to cooperatively maintain the set oxygen content and micro-positive pressure in the furnace.

[0008] Specifically, the step S1 comprises: performing Kalman filtering processing on the collected oxygen content signal, and the filtering formula is: ; wherein, is the filtered oxygen content estimation value at the moment, k is the measured oxygen content at the moment, is the Kalman gain; k performing sliding average filtering processing on the collected pressure signal, and the filtering formula is: ; ; wherein, is the filtered pressure estimation value at the moment, k is the measured pressure at the moment, is the Kalman gain. k i .

[0009] Specifically, in the step S2, the input features of the deep belief network DBN model include real-time signals, historical trends and smelting stage identifiers, and the output is the oxygen content prediction value and the pressure prediction value in the future specific sampling period; wherein the historical trend is the oxygen content deviation change rate and the pressure deviation change rate in the previous several sampling periods.

[0010] Further, the step S2 further comprises a prediction compensation step: When the deviation between the predicted future state and the set value exceeds the preset threshold, the basic argon flow is fed forwardly compensated, and the compensation formula is: ; wherein, is the argon flow after the prediction compensation, is the uncompensated basic argon flow, is the prediction gain, is the oxygen content prediction value at the moment, k + n in the future, is the oxygen content set value.

[0011] ​Specifically, step S3 includes: The Particle Swarm Optimization (PSO) algorithm iteratively optimizes the particle swarm based on historical smelting data, wherein the position code of each particle includes the initial parameters of the fuzzy PID controller and the center value of the fuzzy membership function. The particle velocity update formula for the PSO algorithm is as follows: ; in, Indicates the first i The particle in the first j Vie k +1 iteration speed; w Inertial weight; , For learning factors; , A random number in the interval [0,1]. For the first i The particle in the first j The optimal position of an individual in dimension. For the first i The particle in the first j Vie k Next iteration position For all particles in the first j The global optimal position of the dimension; Through iterative optimization using the PSO algorithm, an optimal initial parameter library corresponding to different steel grades or working conditions is obtained.

[0012] Specifically, in step S4, the method for adaptively adjusting the parameters of the fuzzy PID controller based on the real-time signal and the prediction result includes: Based on real-time oxygen content deviation and its rate of change The correction amount of PID parameters is generated in real time through fuzzy inference. And update the parameters according to the following formula: ; in, For fuzzy PID controller in k The parameters at time, These are adaptive coefficients.

[0013] Specifically, in step S5, the calculation formula for the argon flow control command is as follows: ; in, express k The control value of argon flow rate at any given time; This indicates the stage correction factor determined based on the smelting stage.

[0014] Specifically, in step S5, the calculation formula of the dust removal valve opening degree instruction is: ; wherein, represents k the control value of the dust removal valve opening degree at the moment; represents the initial value of the dust removal valve opening degree; is a flow change compensation coefficient, represents a pressure deviation compensation coefficient, represents k the micro-positive pressure deviation at the moment.

[0015] Preferably, the control method further comprises step S6: when the system performance index does not meet the requirements in a plurality of continuous smelting periods, triggering a self-evolution mechanism, re-executing the PSO algorithm for global parameter optimization, and incrementally training the DBN model using the newly added real-time data.

[0016] The second aspect of the present application provides an electric slag furnace argon flow and micro-positive pressure linkage control system, comprising: a data acquisition module for real-time acquisition of in-furnace oxygen content and micro-positive pressure data; a controller for calculating and outputting argon flow control instructions and dust removal valve opening degree control instructions based on the control method according to the real-time acquired data; an actuator comprising an argon flow control component and a dust removal proportional valve, for receiving and executing the argon flow control instructions and dust removal valve opening degree control instructions issued by the controller.

[0017] Compared with the prior art, the present application has the following advantages: (1) The present application realizes the advance of control by predicting the future furnace condition through a deep belief network (DBN) model, overcoming the problem of system large lag; provides excellent initial performance for the fuzzy PID controller through offline parameter optimization by a particle swarm optimization (PSO) algorithm, so that it can adapt to different working conditions; and makes the system robust to dynamic processes through online fuzzy reasoning for adaptive adjustment of the controller parameters; finally, through linkage calculation of argon flow and dust removal valve opening degree instructions, precise and stable linkage control of in-furnace oxygen content and micro-positive pressure is realized; (2) The present application introduces a self-evolution mechanism, which can automatically trigger the re-optimization of the PSO algorithm and the incremental training of the DBN model when the system performance declines, so that the control system has the ability to learn and optimize itself in the long-term running process, effectively overcomes the model mismatch and performance decline caused by equipment aging or raw material changes, and guarantees the long-term consistency and reliability of the control effect. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to make the technical solutions in the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments or prior art description will be briefly introduced. Obviously, the accompanying drawings in the following description only aim to some embodiments of the present application, and other drawings can be obtained by those of ordinary skill in the art without any creative effort.

[0019] Figure 1 A flowchart of a method for controlling the argon flow and the micro-positive pressure of an electroslag furnace according to the present application. DETAILED DESCRIPTION

[0020] The technical solutions of the present application will be described clearly and completely in combination with the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without any creative effort are within the protection scope of the present application.

[0021] Reference Figure 1 The first aspect of the present application provides a method for controlling the argon flow and the micro-positive pressure of an electroslag furnace, comprising the following steps: S1, collecting the real-time oxygen content signal and the micro-positive pressure signal in the electroslag furnace by an oxygen content sensor and a micro-positive pressure sensor; In order to ensure the control precision, the original collected signals need to be filtered: The collected oxygen content signal is subjected to Kalman filtering to suppress random noise and improve the signal quality, and the filtering formula is: ; Wherein, is the filtered oxygen content estimation value at the moment, k is the measured oxygen content at the moment, is the Kalman gain, and the value range is usually 0.1 to 0.3; k The collected pressure signal is subjected to sliding average filtering to smooth the short-term fluctuations, and the filtering formula is: ; ; Wherein, is the filtered pressure estimation value at the moment, k is the measured pressure at the moment, k i .

[0022] ​​S2, based on the collected real-time signals, uses a deep belief network (DBN) model to predict the oxygen content and micro-positive pressure in the furnace at future moments, and performs feedforward compensation based on the prediction results; The DBN model is a three-layer network structure, consisting of one input layer (6 neurons) and two layers of Restricted Boltzmann Machines (RBMs), each with 12 neurons. Its input features include: Real-time parameter: Oxygen content at the current moment Slight positive pressure Argon flow rate Dust collector valve opening ; Historical trend: Rate of change in oxygen content deviation over the first 3 sampling periods (corresponding to 0.3 seconds) and pressure deviation change rate ; Smelting stage identifiers: Different stages are represented by numbers, such as arc initiation period = 1, arc stabilization period = 2, crystallization period = 3.

[0023] The model was trained using at least 5000 sets of historical data covering different steel grades and furnace conditions. It employed the Adam optimizer with mean squared error (MSE) as the loss function. The model output is a predicted oxygen content for the next two sampling periods (0.2 seconds). and pressure forecast values .

[0024] Furthermore, a predictive compensation step is introduced: When the deviation between the predicted future state and the set value satisfies or hour, Feedforward compensation is applied to the base argon flow rate, and the compensation formula is as follows: ; in, The predicted and compensated argon flow rate, The uncompensated base argon flow rate, To predict gain, For the future k + n Predicted oxygen content at time [time]. Set the oxygen content value.

[0025] S3, using the particle swarm optimization (PSO) algorithm to optimize the initial parameters of the fuzzy PID controller offline; The PSO algorithm is used to solve the problem of blind initial setting of controller parameters. The PSO algorithm iteratively optimizes the particle swarm based on historical smelting data, wherein the position code of each particle contains the initial parameters of the fuzzy PID controller and the center value of the fuzzy membership function; its specific implementation process is as follows: Particle encoding: Each particle is a 19-dimensional vector containing the initial parameters of 3 fuzzy PIDs ( ) and the center values ​​of 14 fuzzy membership functions (corresponding to the oxygen content deviation of the two input variables). and slight positive pressure deviation Each variable has 7 fuzzy subsets {NB, NM, NS, ZO, PS, PM, PB}.

[0026] Parameter range: , , The membership function center value is allowed to fluctuate by ±20% from the base value.

[0027] Velocity update: A velocity update formula with adaptive inertia weights is used. ; in, Indicates the first i The particle in the first j Vie k +1 iteration speed; w For inertial weights, , Maximum number of iterations; learning factor ; , A random number in the interval [0,1]. For the first i The particle in the first j The optimal position of an individual in dimension. For the first i The particle in the first j Vie k Next iteration position For all particles in the first j The global optimal position of the dimension; Optimization objective: Minimize the expression of the objective function as follows: ; in, T The end time, for t Oxygen content deviation at any given time for t The slight positive pressure deviation at any given time, To control the penalty coefficient, The maximum absolute value of the output quantity (argon flow rate or valve opening change command) of the controller over the entire time period [0, T ] is represented.

[0028] The optimization process is based on historical data of 100 heats classified by steel grade, and finally generates an optimal initial parameter library corresponding to different steel grades (such as alloy structural steel, stainless steel).

[0029] S4, initialize the fuzzy PID controller with the optimized initial parameters, and based on the real-time signal and the prediction result, adaptively adjust the parameters of the fuzzy PID controller; In the real-time control phase (sampling period 0.1s), the system calls the initial parameters matching the current steel grade from the parameter library Initialize the fuzzy PID controller.

[0030] The adaptive adjustment of parameters is realized through fuzzy reasoning: Calculate the current oxygen content deviation And its rate of change ; Input and into the fuzzy reasoning system, and according to the preset fuzzy rule base, generate the correction amount of PID parameters in real time ; Update the PID parameters as follows: ; Where, is the parameter of the fuzzy PID controller at k time, is the adaptive coefficient, which is used to limit the adjustment amplitude of a single step to avoid system oscillation.

[0031] S5, according to the adjusted fuzzy PID controller parameters, calculate and output the argon flow rate control command and the dedusting valve opening control command to cooperatively maintain the set oxygen content and micro-positive pressure in the furnace.

[0032] The argon flow rate control command is derived from the predicted and compensated flow rate combined with the smelting stage factor, and the calculation formula is: ; Where, represents k the regulation value of argon flow rate at time; represents the stage correction factor determined according to the smelting stage, which takes a value between 0.8 and 1.2 according to different smelting stages such as arc starting, arc stabilizing, and crystallization.

[0033] The dust removal valve opening degree control instruction is calculated through a linkage algorithm, to ensure coordinated action with the argon gas flow, to quickly and stably control the furnace pressure, and the calculation formula is: ; wherein, represents the control value of the dust removal valve opening degree at the moment t; k represents the initial value of the dust removal valve opening degree; is a flow change compensation coefficient, when the argon gas flow increases, the dust removal valve is appropriately closed in advance to inhibit the pressure rise; represents a pressure deviation compensation coefficient, for quickly eliminating static deviation; represents the micro-positive pressure deviation at the moment t. k

[0034] Constraint processing: the final instruction needs to satisfy , to ensure the safe operation of the actuator.

[0035] As an optional technical solution, the control method further comprises a step S6 of triggering a self-evolution mechanism to realize long-term self-optimization of the control system. To ensure the long-term consistency of the control effect, the application also designs a self-evolution mechanism, when the system detects that the steady-state error of oxygen content of three consecutive furnace times is greater than 0.5ppm, the mechanism is automatically triggered: re-execute PSO global optimization (30 iterations), and update the initial parameter library by using the new data; At the same time, the DBN prediction model is incrementally trained by using the newly added 50 groups of real-time data, so that the model adapts to the latest working condition changes.

[0036] The second aspect of the application provides an argon gas flow and micro-positive pressure linkage control system of an electroslag furnace, comprising: A data acquisition module comprising an oxygen content sensor, a micro-positive pressure sensor and a signal conditioning circuit, responsible for real-time acquisition of in-furnace data, for real-time acquisition of in-furnace oxygen content and micro-positive pressure data; A controller, which is the core of the system, can be an industrial PLC or an embedded industrial computer, and has an executable program stored therein, and can calculate and output argon gas flow control instructions and dust removal valve opening degree control instructions based on the control method according to the real-time acquired data; An actuator, comprising an argon gas flow control component (such as a mass flow meter and an electric regulating valve) and a dust proportional valve, for receiving and executing the instructions issued by the controller.

[0037] ​​In order to ensure the control effect, in the furnace structure, the dynamic seal adopts the metal bellows made of Inconel 625 material, which can resist 1000 DEG C temperature and compensate displacement of ±15mm; the static seal adopts the silicone rubber sealing rubber strip, which has hardness of Shore A70 and compression permanent deformation of <10% (200 DEG C x 72h), thereby providing a reliable hardware foundation for maintaining the micro-positive pressure.

[0038] The above merely describes the preferred embodiments of the present application and is not used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for controlling the argon flow rate and micro-positive pressure in an electroslag furnace, characterized in that, Includes the following steps: S1, collects real-time oxygen content signals and micro-positive pressure signals inside the electroslag furnace; S2, based on the collected real-time signals, uses a deep belief network (DBN) model to predict the oxygen content and micro-positive pressure in the furnace at future moments; S3, using the particle swarm optimization (PSO) algorithm to optimize the initial parameters of the fuzzy PID controller offline; S4. Initialize the fuzzy PID controller using the optimized initial parameters, and adaptively adjust the parameters of the fuzzy PID controller based on the real-time signal and prediction results. S5 calculates and outputs argon flow control commands and dust removal valve opening control commands based on the adjusted fuzzy PID controller parameters, in order to work together to maintain the set oxygen content and slight positive pressure inside the furnace.

2. The method for linkage control of argon flow rate and micro-positive pressure in an electroslag furnace as described in claim 1, characterized in that, Step S1 includes: The collected oxygen content signal is processed by Kalman filtering. The filtering formula is as follows: ； in, for k The oxygen content estimate after time-lapse filtering. for k Measured oxygen content at any given time Kalman gain; The collected pressure signal is processed by moving average filtering. The filtering formula is as follows: ； in, for k The pressure estimate after filtering at any given time. for k i Real-time pressure measurement.

3. The method for linkage control of argon flow rate and micro-positive pressure in an electroslag furnace as described in claim 1, characterized in that, In step S2, the input features of the Deep Belief Network (DBN) model include real-time signals, historical trends, and smelting stage identifiers, and the output is the predicted oxygen content and pressure for a specific future sampling period; wherein, the historical trend is the rate of change of oxygen content deviation and the rate of change of pressure deviation over the previous several sampling periods.

4. The method for linkage control of argon flow rate and micro-positive pressure in an electroslag furnace as described in claim 3, characterized in that, Step S2 also includes a prediction compensation step: When the deviation between the predicted future state and the set value exceeds a preset threshold, feedforward compensation is applied to the basic argon flow rate. The compensation formula is as follows: ； in, The predicted and compensated argon flow rate, The uncompensated base argon flow rate, To predict gain, For the future k + n Predicted oxygen content at time [time]. Set the oxygen content value.

5. The method for linkage control of argon flow rate and micro-positive pressure in an electroslag furnace as described in claim 4, characterized in that, Step S3 specifically includes: The Particle Swarm Optimization (PSO) algorithm iteratively optimizes the particle swarm based on historical smelting data, wherein the position code of each particle includes the initial parameters of the fuzzy PID controller and the center value of the fuzzy membership function. The particle velocity update formula for the PSO algorithm is as follows: ； in, Indicates the first i The particle in the first j Vie k +1 iteration speed; w Inertial weight; , For learning factors; , A random number in the interval [0,1]. For the first i The particle in the first j The optimal position of an individual in dimension. For the first i The particle in the first j Vie k Next iteration position For all particles in the first j The global optimal position of the dimension; Through iterative optimization using the PSO algorithm, an optimal initial parameter library corresponding to different steel grades or working conditions is obtained.

6. The method for linkage control of argon flow rate and micro-positive pressure in an electroslag furnace as described in claim 5, characterized in that, In step S4, the method for adaptively adjusting the parameters of the fuzzy PID controller based on the real-time signal and the prediction result includes: Based on real-time oxygen content deviation and its rate of change The correction amount of PID parameters is generated in real time through fuzzy inference. And update the parameters according to the following formula: ； in, For fuzzy PID controller in k The parameters at time, These are adaptive coefficients.

7. The method for linkage control of argon flow rate and micro-positive pressure in an electroslag furnace as described in claim 6, characterized in that, In step S5, the calculation formula for the argon flow control command is as follows: ； in, express k The control value of argon flow rate at any given time; This indicates the stage correction factor determined based on the smelting stage.

8. The method for linkage control of argon flow rate and micro-positive pressure in an electroslag furnace as described in claim 7, characterized in that, In step S5, the calculation formula for the dust collector valve opening command is: ； in, express k The control value for the opening degree of the dust collector valve at all times; This indicates the initial value of the dust collector valve opening. This is the flow rate change compensation coefficient. This represents the pressure deviation compensation coefficient. express k The slight positive pressure deviation at any given moment.

9. The method for linkage control of argon flow rate and micro-positive pressure in an electroslag furnace as described in claim 1, characterized in that, It also includes step S6, which triggers the self-evolution mechanism when the system performance indicators fail to meet the requirements in multiple consecutive smelting cycles, re-executes the PSO algorithm to optimize global parameters, and uses newly added real-time data to incrementally train the DBN model.

10. A linkage control system for argon flow rate and micro-positive pressure in an electroslag furnace, characterized in that, include: The data acquisition module is used to collect real-time data on oxygen content and micro-positive pressure inside the furnace; The controller calculates and outputs argon flow control commands and dust removal valve opening control commands based on real-time collected data and the control method described in any one of claims 1-9. The actuator, including the argon flow control component and the dust removal proportional valve, is used to receive and execute the argon flow control command and the dust removal valve opening control command issued by the controller.

Images