A1-level inorganic fireproof core material production line intelligent adjusting method based on thickness change

Abstract

The invention discloses an A1-level inorganic fireproof core material production line intelligent adjusting method based on thickness variation, which comprises the following steps: respectively installing a thickness sensor and a pressure sensor at appointed positions of a double-roller mill, and controlling the double-roller mill through a PID (Proportion Integration Differentiation) controller; sensor data are collected in real time, sliding window filtering is adopted, the target middle thickness of each rolling mill is determined, and the change rate of pressure at adjacent time points of a hydraulic cylinder is determined; meanwhile, considering the thickness tracking precision and the pressure stability to construct a fitness function, optimizing by adopting a quantum genetic algorithm, and determining a global PID parameter and a pressure feed-forward compensation amount; determining a local PID parameter and a pressure feed-forward compensation amount by adopting an Elman neural network model of a dynamic structure adjustment strategy; and performing pressure closed-loop control based on the global and local PID parameters and the pressure feed-forward compensation amount, and controlling the operation of the double-roller mill production line. Compared with the prior art, the problems that the thicknesses of core materials are inconsistent, and the production efficiency is low are solved.

Description

Technical Field

[0001] The present invention belongs to the field of production line control, and in particular relates to an intelligent adjustment method for an A1-class inorganic fireproof core material production line based on thickness variation. Background Art

[0002] Aluminum-plastic composite panels are currently widely used in architectural decoration. Traditionally, their manufacturing process typically uses polyethylene (PE) as the core material. However, as an organic polymer, PE, while lightweight, flexible, and easy to process, is inherently flammable and lacks fire resistance, which, to a certain extent, limits its use in locations with strict fire protection requirements.

[0003] In recent years, major fires, particularly in high-rise buildings, have frequently occurred, resulting in significant casualties and property damage, garnering widespread public attention. These incidents have exposed the shortcomings of traditional building materials in terms of fire resistance, prompting relevant departments to comprehensively upgrade fire protection standards for building materials. Traditional aluminum-plastic composite panels, with a fire rating of only Class B, are flammable and highly susceptible to fires under extreme conditions such as high pressure, heat release, or electrical discharge, making them difficult to meet the stringent fire safety requirements of modern high-rise buildings.

[0004] Against this backdrop, panels made from A1-grade fireproofing materials have begun to gain market favor. A1-grade fireproofing materials are non-combustible and effectively contain the spread of fire in the event of a fire, buying time for evacuation and firefighting operations. Consequently, they have found widespread use in high-rise buildings, public spaces, and other areas with stringent fire protection requirements. However, despite the significant safety advantages of A1-grade fireproofing panels, their production still faces some technical challenges.

[0005] During the core board forming process of the existing A1-grade fireproof board core material production line, due to the significant differences in the physical properties of inorganic materials and organic materials, the differences in density and fiber distribution of different raw materials, and the fact that the thickness control system of the existing production line uses a traditional mechanical limit method and is therefore unable to respond in real time to changes in material properties and production thickness specifications, the thickness uniformity and surface smoothness of the core board are difficult to accurately control. This technological deficiency not only affects the aesthetics and performance of the product, but also leads to a high rate of defective products during the production process, increases the production costs of enterprises, and restricts the further promotion and application of A1-grade fireproof boards. In order to solve these problems, the industry urgently needs to introduce an intelligent control system to achieve precise regulation of the thickness and surface quality of the core board, providing more reliable protection for building fire safety. Summary of the Invention

[0006] Purpose of the invention: In response to the problems in the background technology, the present invention provides an intelligent adjustment method for an A1-class inorganic fireproof core material production line based on thickness changes, which optimizes the PID algorithm through quantum genetic algorithm and neural network to achieve thickness and pressure control of the core material.

[0007] Technical solution: The present invention discloses an intelligent adjustment method for an A1-grade inorganic fireproof core material production line based on thickness variation, comprising the following steps:

[0008] S1. Selecting a suitable twin-roll mill according to industrial needs, presetting mill operating parameters, target thickness values, and an appropriate pressure range based on the twin-roll mill, installing a thickness sensor and a pressure sensor at designated locations based on the structural characteristics of the twin-roll mill, and controlling the twin-roll mill using a PID controller;

[0009] S2. Real-time data collection from thickness sensors and pressure sensors is performed, the collected data is processed, and the collected thickness data is filtered using a sliding window. The target intermediate thickness of each rolling mill is determined based on the reduction ratio of each rolling mill, and the rate of change of the hydraulic cylinder pressure of the twin-roll mill at adjacent time points is determined.

[0010] S3. Build a thickness-velocity model, calculate the reference pressure and thickness error, and build a fitness function considering thickness tracking accuracy and pressure stability. Use a quantum genetic algorithm to optimize based on the fitness function to determine the global PID parameters and pressure feedforward compensation.

[0011] S4. Construct an Elman neural network model, adopt a dynamic structure adjustment strategy, and automatically increase or decrease the number of hidden layer nodes according to the thickness deviation. The Elman neural network model takes the thickness error, the thickness error change rate, the thickness error integral term, the temperature of the data point, the hydraulic cylinder pressure, and the change rate of the hydraulic cylinder pressure at adjacent time points as input, and ultimately outputs local PID parameters and pressure feedforward compensation;

[0012] S5. Perform pressure closed-loop control based on the global PID parameters and pressure feedforward compensation amount obtained in step S3 and the local PID parameters and pressure feedforward compensation amount obtained in step S4 to control the operation of the twin-roll mill production line.

[0013] Furthermore, the A1-grade inorganic fireproof core material production line includes a conveying system, an extrusion system and a baking system, wherein the conveying system includes conveying rollers and conveyor mesh belts, and the conveyor mesh belts are installed on the conveying rollers. The baking system includes multiple hot air circulation ovens, and the extrusion system includes a first double-roller mill, a second double-roller mill, a third double-roller mill and a thickness and pressure control device. The mills are arranged on a conveyor chain formed by conveyor rollers and conveyor mesh belts. The three double-roller mills are respectively arranged between the hot air circulation ovens. The thickness and pressure control device is provided with a thickness sensor and a pressure sensor, and the thickness sensor and the pressure sensor are connected to the thickness and pressure control device through a control circuit transmission wire.

[0014] Furthermore, the twin-roll mill includes an upper roll and a lower roll, and a rolling gap is formed between the upper roll and the lower roll for rolling the core material. The twin-roll mill also includes a driving device for driving the rotation of the upper roll and the lower roll. The twin-roll mill adjusts the gap of the twin-roll mill according to the control signal to change the rolling pressure.

[0015] Furthermore, the three twin-roll mills in S2 are arranged in sequence, and the target intermediate thickness of each mill is allocated by the reduction ratio:

[0016] h1=h0×(1-α1)

[0017] h2=h1×(1-α2)

[0018] h target =h2×(1-α3)

[0019] Where h0 is the initial thickness of the raw material (mm); h1 is the target thickness of the material after the first twin-roll mill (mm); h2 is the target thickness of the material after the second twin-roll mill (mm); target is the target thickness of the material (mm); α1, α2, α3 are the reduction rates of each rolling mill, ranging from 0.2 to 0.5.

[0020] Furthermore, in S2, the collected thickness data is filtered using a sliding window, specifically:

[0021]

[0022] Where, is the original thickness measurement value of the i-th rolling mill at the j-th time step (mm); is the filtered thickness value of the kth data point of the i-th rolling mill (mm).

[0023] Furthermore, the pressure change rate of the hydraulic cylinder at adjacent time points in S2 is specifically:

[0024] ΔP i (k) = Pi (k)-P i (k-1)

[0025] Where ΔP i (k) is the pressure change rate of the kth data point of the i-th rolling mill (MPa); P i (k) is the pressure of the kth data point of the i-th rolling mill (MPa); P i (k-1) is the pressure (MPa) of the k-1th data point of the i-th rolling mill.

[0026] Furthermore, in S3, a quantum genetic algorithm is used to perform optimization based on a fitness function to determine global PID parameters and pressure feedforward compensation, specifically:

[0027] 1) The chromosome encoding rule of quantum genetic algorithm is:

[0028]

[0029] Where q m is the state of the mth quantum bit, consisting of two probability amplitudes; cosθ m represents the probability amplitude of the qubit being in state |0>; sinθ m represents the probability amplitude of the qubit being in state |1>;

[0030] The quantum genetic algorithm takes on the role of global optimization in this process. Each control parameter, including the PID coefficient and pressure compensation, is encoded by an 8-qubit probability amplitude. A single chromosome contains 32 qubits. The quantum state is observed and converted into a classical value:

[0031]

[0032]

[0033] Where K p is the proportional coefficient of the PID controller; K i is the integral coefficient of the PID controller; K d is the differential coefficient of the PID controller; P ff is the pressure feedforward compensation amount (MPa); b m ∈{0,1}, is the quantum bit observation result; and The maximum value of proportional, integral, differential coefficients and pressure feedforward compensation, determined by historical data statistics;

[0034] 2) Calculate the thickness error, the expression is:

[0035]

[0036] Where, is the thickness error of the kth data point of the i-th rolling mill (mm);

[0037] 3) Calculate the reference pressure, which is obtained from the thickness-velocity model:

[0038]

[0039] Where, is the reference pressure of the i-th rolling mill (MPa); is the target thickness of the i-th rolling mill (mm); V is the conveyor belt speed (m / min); a, b, c are material property coefficients, calibrated through orthogonal rolling experiments;

[0040] 4) Design the fitness function, taking into account both thickness tracking accuracy and pressure stability, and achieve multi-objective optimization through weighted calculation. The formula is as follows:

[0041]

[0042] Where J is the fitness function of the quantum genetic algorithm; N is the total number of data points; and λ is the pressure fluctuation penalty weight, which is determined by sensitivity analysis.

[0043] Furthermore, the Elman neural network model in S4 is specifically as follows:

[0044] 1) The Elman neural network model adopts a dynamic structural adjustment strategy. The number of hidden layer nodes automatically increases or decreases according to the thickness deviation. When the thickness error is detected to continuously exceed the threshold, the network automatically adds hidden nodes to enhance the nonlinear fitting ability. When the node activity is insufficient, the streamlining mechanism is triggered to ensure computational efficiency. The network output layer not only generates PID control parameters, but also outputs additional pressure feedforward compensation.

[0045] 2) Construction of Elman neural network dynamic control input layer:

[0046]

[0047]

[0048]

[0049] Where x i (k) is the input vector of the kth data point of the i-th rolling mill; is the thickness error change rate of the kth data point of the i-th rolling mill (mm); is the thickness error integral term of the kth data point of the i-th rolling mill (s·mm); T i (k) is the temperature of the kth data point of the i-th rolling mill (°C); T sis the control period (s);

[0050] Conditions for adding hidden layer nodes:

[0051]

[0052] Node deletion conditions:

[0053]

[0054] Where, is the average activation of the kth data point of the uth node; d u (j) is the activation value of the u-th node at the j-th time step;

[0055] Output calculation:

[0056]

[0057] Where, is the proportional coefficient of the i-th rolling mill; is the integral coefficient of the i-th rolling mill; is the differential coefficient of the i-th rolling mill; is the pressure feedforward compensation amount of the i-th rolling mill (MPa).

[0058] Furthermore, the pressure closed-loop control in S5 is as follows:

[0059]

[0060] Where K p is the proportional coefficient of the PID controller; K i is the integral coefficient of the PID controller; K d is the differential coefficient of the PID controller; P ff is the pressure feedforward compensation amount (MPa), is the proportional coefficient of the i-th rolling mill; is the integral coefficient of the i-th rolling mill; is the differential coefficient of the i-th rolling mill; is the pressure feedforward compensation of the i-th rolling mill (MPa), is the thickness error of the kth data point of the i-th rolling mill (mm), is the thickness error change rate of the kth data point of the i-th rolling mill (mm), is the pressure setting value (MPa) of the kth data point of the i-th rolling mill.

[0061] Furthermore, when abnormal pressure fluctuations or decreased population diversity are detected, the quantum genetic algorithm catastrophe operator automatically resets part of the chromosomes and injects the historical optimal solution to prevent the algorithm from falling into the local optimum. The triggering condition of the catastrophe operator is that the population diversity is less than 0.1 or the pressure fluctuation rate is less than 0.1. in:

[0062] The formula for calculating population diversity is:

[0063]

[0064] Where L is the total number of chromosomes in the population; q l,m is the probability amplitude of the mth quantum bit of the lth chromosome; is the average probability amplitude of all chromosomes in the population at the mth qubit;

[0065] The pressure fluctuation rate is:

[0066]

[0067] Where, is the mean pressure of the i-th rolling mill (MPa).

[0068] Beneficial effects:

[0069] The present invention provides an intelligent adjustment method for an A1-class inorganic fireproof core material production line based on thickness change. The first double-roll mill and the second double-roll mill can adjust the surface smoothness of the core material and reduce the thickness of the core material. The third double-roll mill can further adjust the thickness of the core material. The thickness and pressure control device uses a thickness sensor, a pressure sensor and a quantum genetic algorithm next to the double-roll mill, and a neural network model to optimize the PID algorithm to achieve thickness control of the core material. The global PID parameters and the global pressure feedforward compensation amount are obtained by optimizing the fitness function using the quantum genetic algorithm. The local PID parameters and the pressure feedforward compensation amount are obtained by dynamically adjusting the hidden layer through the neural network. The global and local aspects are taken into consideration simultaneously, and the thickness is gradually reduced to finally reach the target thickness. In addition, the pressure control of the core material during rolling is achieved, and the pressure on the core material is controlled within a reasonable range, which greatly reduces the defective rate and reduces the input cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 Schematic diagram of the structure of the A1 grade inorganic fireproof core material production line of the present invention;

[0071] Figure 2 This is a flow chart of the intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness changes of the present invention.

[0072] In the figure: 1- conveying system, 101- conveyor belt, 102- conveyor roller, 2- extrusion system, 201- double-roll mill, 201-1- first double-roll mill, 201-2- second double-roll mill, 201-3- third double-roll mill, 202- thickness sensor, 202-1- first thickness sensor, 202-2- second thickness sensor, 202-3- third thickness sensor, 202-4- fourth thickness sensor, 202-5- fifth thickness sensor, 202-6- sixth thickness sensor, 203- pressure sensor, 203-1- first pressure sensor, 203-2 -Second pressure sensor, 203-3-Third pressure sensor, 204-Thickness and pressure control device, 204-1 Sensing display module, 204-2 Display screen, 204-3 Functional area, 205-Control circuit transmission wire, 3-Baking system, 301-Hot air circulation oven, 301-1-First hot air circulation oven, 301-5-Fifth hot air circulation oven, 301-6-Sixth hot air circulation oven, 301-7-Seventh hot air circulation oven, 301-8-Eighth hot air circulation oven, 301-10-Tenth hot air circulation oven, 301-11-Eleventh hot air circulation oven. DETAILED DESCRIPTION

[0073] The specific embodiments of the present invention are described below with reference to the accompanying drawings.

[0074] like Figure 1 As shown, the present invention provides an intelligent adjustment method for an A1-level inorganic fireproof core material production line based on thickness changes. The A1-level inorganic fireproof core material production line of this embodiment is composed of a conveying system 1, an extrusion system 2 and a baking system 3. The conveying system 1 includes a conveying mesh belt 101 and a conveying roller 102. The conveying mesh belt 101 is installed on the conveying roller 102, and the transmission speed of the conveying roller 102 is 1.5 to 3.5 m / min.

[0075] The baking system 3 includes a hot air circulation oven 301, and the baking system 3 has eleven hot air circulation ovens 301. The extrusion system 2 includes a first double-roller rolling mill 201-1, a second double-roller rolling mill 201-2 and a third double-roller rolling mill 201-3. The rolling mills are arranged on a conveyor chain formed by a conveyor mesh belt 101 and a conveyor roller 102. The first double-roller rolling mill 201-1, the second double-roller rolling mill 201-2 and the third double-roller rolling mill 201-3 are respectively arranged between the hot air circulation ovens 301. The extrusion system 2 is provided with a thickness sensor 202 and a pressure sensor 203. The thickness sensor 202 and the pressure sensor 203 are connected to the thickness and pressure control device 204 through a control circuit transmission wire 205.

[0076] A sensing display module 204-1 is provided in the thickness and pressure control device 204, a display screen 204-2 is provided on the thickness and pressure control device 204, a functional area 204-3 is provided under the display screen 204-2, the thickness sensor 202 and the pressure sensor 203 are connected to the sensing display module 204-1 through a control circuit transmission wire 205, and the sensing display module 204-1 is connected to the display screen 204-2 through a wire.

[0077] The hot air circulation oven 301 includes a first hot air circulation oven 301-1, a fifth hot air circulation oven 301-5, a sixth hot air circulation oven 301-6, a seventh hot air circulation oven 301-7, an eighth hot air circulation oven 301-8, a tenth hot air circulation oven 301-10, and an eleventh hot air circulation oven 301-11. Other hot air circulation ovens are not further introduced. A first double-roll rolling mill 201-1 is provided between the fifth hot air circulation oven 301-5 and the sixth hot air circulation oven 301-6, a second double-roll rolling mill 201-2 is provided between the seventh hot air circulation oven 301-7 and the eighth hot air circulation oven 301-8, and a third double-roll rolling mill 201-3 is provided between the tenth hot air circulation oven 301-10 and the eleventh hot air circulation oven 301-11.

[0078] The temperature control range of the first hot air circulation oven 301-1 is 280℃~340℃, the temperature control range of the fifth hot air circulation oven 301-5 is 120℃~160℃, the temperature control range of the sixth hot air circulation oven 301-6 is 145℃~185℃, the temperature control range of the seventh hot air circulation oven 301-7 is 145℃~185℃, the temperature control range of the eighth hot air circulation oven 301-8 is 155℃~190℃, the temperature control range of the tenth hot air circulation oven 301-10 is 155℃~190℃, and the temperature control range of the eleventh hot air circulation oven 301-11 is 210℃~255℃.

[0079] The twin-roll mill comprises an upper roll and a lower roll, which form a rolling gap, and the upper roll and the lower roll are driven to rotate by a driving device.

[0080] The drive device can be a servo motor and a hydraulic system to drive the upper and lower rolls to rotate. The servo motor can adopt a high-performance Yaskawa Sigma-7 series servo motor, and the hydraulic system adopts a high-response hydraulic servo system to ensure precise control and fast response of the rolls.

[0081] The thickness sensor uses a KEYENCE LK-G5000 series laser thickness gauge with an accuracy of ±0.1μm and a response time of 50μs. The thickness sensor is fixed at the entrance and exit of the twin-roll mill with brackets to ensure accurate thickness data as the core material enters and exits the mill. The sensor maintains an appropriate distance from the core material surface to prevent vibration and other factors from affecting measurement accuracy. The thickness sensor acquires thickness data at a high frequency for real-time monitoring. A NIPXI-4472 high-precision data acquisition card with a sampling frequency of 1000Hz is used to ensure high-frequency data acquisition. The data acquisition card is connected to the thickness and pressure control device 204 via a PXI bus, ensuring fast and stable data transmission. The thickness and pressure control device 204 receives the thickness data collected by the data acquisition card and calculates the thickness deviation. Through real-time data processing and control signal output, dynamic adjustment of the rolling pressure is achieved. In this embodiment, the thickness and pressure control device 204 adopts Siemens S7-1500 series PLC or industrial PC, and connects the data acquisition card and the pressure control device through industrial Ethernet or field bus (such as EtherCAT, Profibus, CAN bus).

[0082] In this embodiment, if Figure 2 As shown, the present invention proposes an intelligent adjustment method for an A1-level inorganic fireproof core material production line based on thickness changes. The thickness and pressure control device 204 is based on a neural network PID algorithm optimized by a quantum genetic algorithm. The algorithm takes the neural network optimized by quantum genetic algorithm as the core, deeply integrates thickness and pressure parameters, and forms a multivariable collaborative control architecture. The system collects core material thickness data in real time through a high-precision laser thickness gauge, and integrates a pressure transmitter to monitor the pressure value of the hydraulic system. Parameters such as thickness error, pressure change rate, and temperature together constitute the input vector of the neural network. The input layer is expanded into a multidimensional data space containing 8 nodes to achieve comprehensive perception of the production status, which specifically includes the following steps:

[0083] S1. Select a suitable twin-roll mill according to industrial needs, pre-set mill operating parameters, target thickness values, and suitable pressure ranges based on the twin-roll mill, install thickness sensors and pressure sensors at designated locations based on the structural characteristics of the twin-roll mill, and control the twin-roll mill through a PID controller.

[0084] S2. Real-time data collection from thickness sensors and pressure sensors is performed, and the collected data is processed. The collected thickness data is filtered using a sliding window. The target intermediate thickness of each rolling mill is determined based on the reduction rate of each rolling mill, and the rate of change of the hydraulic cylinder pressure of the twin-roll mill at adjacent time points is determined.

[0085] Three twin-roll mills are arranged in sequence, and the target intermediate thickness of each mill is allocated by the reduction ratio:

[0086] h1=h0×(1-α1)

[0087] h2=h1×(1-α2)

[0088] h target =h2×(1-α3)

[0089] Where h0 is the initial thickness of the raw material (mm); h1 is the target thickness of the material after the first twin-roll mill (mm); h2 is the target thickness of the material after the second twin-roll mill (mm); target is the target thickness of the material (mm); α1, α2, α3 are the reduction ratios of each rolling mill (dimensionless, ranging from 0.2 to 0.5), which need to be calibrated through material compression tests.

[0090] S3. Each rolling mill is equipped with a laser thickness gauge and uses a sliding window filter:

[0091]

[0092] Where, is the original thickness measurement value of the i-th rolling mill at the j-th time step (mm); is the filtered thickness value of the kth data point of the i-th rolling mill (mm).

[0093] Collect the hydraulic cylinder pressure and calculate the rate of change:

[0094] ΔP i (k) = P i (k)-P i (k-1)

[0095] Where ΔP i (k) is the pressure change rate of the kth data point of the i-th rolling mill (MPa); P i (k) is the pressure of the kth data point of the i-th rolling mill (MPa); P i (k-1) is the pressure (MPa) of the k-1th data point of the i-th rolling mill.

[0096] S4. Construct a thickness-velocity model, calculate the reference pressure and thickness error, and construct a fitness function considering thickness tracking accuracy and pressure stability. Use quantum genetic algorithm to optimize based on the fitness function to determine the global PID parameters and pressure feedforward compensation.

[0097] The quantum genetic algorithm is used for chromosome encoding, and the rules are:

[0098]

[0099] Where q mis the state of the mth quantum bit, consisting of two probability amplitudes; cosθ m represents the probability amplitude of the qubit being in state |0>; sinθ m represents the probability amplitude of the qubit being in state |1>.

[0100] The quantum genetic algorithm performs global optimization in this process. Each control parameter (including PID coefficients and pressure compensation) is encoded by an 8-qubit probability amplitude, and a single chromosome contains 32 qubits. After quantum state observation, it is converted into a classical value:

[0101]

[0102] Where K p is the proportional coefficient of the PID controller; K i is the integral coefficient of the PID controller; K d is the differential coefficient of the PID controller; P ff is the pressure feedforward compensation amount (MPa); b m ∈{0,1}, is the quantum bit observation result; and It is the maximum value of proportional, integral, differential coefficient and pressure feedforward compensation (determined by historical data statistics).

[0103] S5. Calculate the thickness error, which is:

[0104]

[0105] Where, is the thickness error (mm) of the kth data point of the i-th rolling mill.

[0106] Calculate the reference pressure, which is obtained from the thickness-velocity model:

[0107]

[0108] Where, is the reference pressure of the i-th rolling mill (MPa); is the target thickness of the i-th rolling mill (mm); V is the conveyor belt speed (m / min); a, b, c are material property coefficients, which are calibrated through orthogonal rolling experiments.

[0109] The fitness function is designed. The algorithm fitness function considers both thickness tracking accuracy and pressure stability, and achieves multi-objective optimization through weighted calculation. The formula is as follows:

[0110]

[0111] Where J is the fitness function of the quantum genetic algorithm; N is the total number of data points; λ is the pressure fluctuation penalty weight (dimensionless, default value 0.3, determined by sensitivity analysis).

[0112] S6. Construct an Elman neural network model and adopt a dynamic structure adjustment strategy. The number of hidden layer nodes automatically increases or decreases according to the thickness deviation. The Elman neural network model takes the thickness error, the thickness error change rate, the thickness error integral term, the temperature of the data point, the hydraulic cylinder pressure, and the change rate of the hydraulic cylinder pressure at adjacent time points as input, and finally outputs the local PID parameters and the pressure feedforward compensation.

[0113] The neural network employs a dynamic structural adjustment strategy, with the number of hidden layer nodes automatically increasing or decreasing based on thickness deviation. When thickness errors consistently exceed a threshold, the network automatically adds hidden nodes to enhance nonlinear fitting capabilities. When node activity is insufficient, a streamlined mechanism is triggered to ensure computational efficiency. The network's output layer not only generates PID control parameters (proportional, integral, and differential coefficients) but also outputs additional pressure feedforward compensation, creating a dual-channel "PID + feedforward" control signal.

[0114] Elman neural network dynamic control input layer construction:

[0115]

[0116] Where x i (k) is the input vector of the kth data point of the i-th rolling mill; is the thickness error change rate of the kth data point of the i-th rolling mill (mm); is the thickness error integral term of the kth data point of the i-th rolling mill (s·mm); T i (k) is the temperature of the kth data point of the i-th rolling mill (°C); T s is the control period (s).

[0117] Conditions for adding hidden layer nodes:

[0118]

[0119] Node deletion conditions:

[0120]

[0121] Where, is the average activation of the kth data point of the uth node; d u (j) is the activation value of the u-th node at the j-th time step.

[0122] Output calculation:

[0123]

[0124] Where, is the proportional coefficient of the i-th rolling mill; is the integral coefficient of the i-th rolling mill; is the differential coefficient of the i-th rolling mill; is the pressure feedforward compensation amount of the i-th rolling mill (MPa).

[0125] S7. Perform pressure closed-loop control based on the global PID parameters and pressure feedforward compensation amount obtained in step S3 and the local PID parameters and pressure feedforward compensation amount obtained in step S4 to control the operation of the twin-roll mill production line.

[0126] The pressure closed-loop control formula is as follows:

[0127]

[0128] Where, is the pressure setting value (MPa) of the kth data point of the i-th rolling mill.

[0129] S8. When abnormal pressure fluctuations or a decrease in population diversity are detected, the catastrophe operator automatically resets some chromosomes and injects historical optimal solutions to prevent the algorithm from falling into local optimality. The triggering condition for the catastrophe operator is that the population diversity is less than 0.1 or the pressure fluctuation rate is

[0130] The formula for calculating population diversity is:

[0131]

[0132] Where L is the total number of chromosomes in the population; q l,m is the probability amplitude of the mth quantum bit of the lth chromosome; is the average probability amplitude of all chromosomes in the population at the mth qubit.

[0133] The pressure fluctuation rate is:

[0134]

[0135] Where, is the mean pressure of the i-th rolling mill (MPa).

[0136] At the hardware level, this invention uses time-sensitive networking (TSN) to synchronize sensor data, achieving a sampling time deviation of less than 1 millisecond between the laser thickness gauge and the pressure transmitter. The iterative optimization of the quantum genetic algorithm is accelerated by an FPGA, keeping a single iteration time under 20 milliseconds, meeting real-time control requirements. The actuator receives the integrated control instructions and synchronously adjusts the roller pressure, conveyor belt speed, and temperature, forming a closed-loop control system.

[0137] This implementation utilizes a display screen to display thickness, pressure, error, and alarm information in real time, allowing operators to monitor and adjust the system. Designed using Siemens WinCC, the display screen provides real-time monitoring of rolling pressure, thickness data, and their errors. Operators use the screen to adjust parameters and address fault alarms, ensuring stable and efficient production.

[0138] In another embodiment, the present invention provides a computer-readable storage medium storing a computer program, which enables a computer to execute the above-mentioned intelligent adjustment method for an A1-level inorganic fireproof core material production line based on thickness changes.

[0139] In another embodiment, the present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, it implements the above-mentioned intelligent adjustment method for the A1-level inorganic fire-proof core material production line based on thickness changes.

[0140] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. An intelligent adjustment method for an A1-level inorganic fireproof core material production line based on thickness variation, characterized in that: The following steps are involved: S1. Selecting a suitable twin-roll mill according to industrial needs, presetting mill operating parameters, target thickness values, and an appropriate pressure range based on the twin-roll mill, installing a thickness sensor and a pressure sensor at designated locations based on the structural characteristics of the twin-roll mill, and controlling the twin-roll mill using a PID controller; S2. Real-time data collection from thickness sensors and pressure sensors is performed, the collected data is processed, and the collected thickness data is filtered using a sliding window. The target intermediate thickness of each rolling mill is determined based on the reduction ratio of each rolling mill, and the rate of change of the hydraulic cylinder pressure of the twin-roll mill at adjacent time points is determined. S3. Build a thickness-velocity model, calculate the reference pressure and thickness error, and build a fitness function considering thickness tracking accuracy and pressure stability. Use a quantum genetic algorithm to optimize based on the fitness function to determine the global PID parameters and pressure feedforward compensation. S4. Construct an Elman neural network model, adopt a dynamic structure adjustment strategy, and automatically increase or decrease the number of hidden layer nodes according to the thickness deviation. The Elman neural network model takes the thickness error, the thickness error change rate, the thickness error integral term, the temperature of the data point, the hydraulic cylinder pressure, and the change rate of the hydraulic cylinder pressure at adjacent time points as input, and ultimately outputs local PID parameters and pressure feedforward compensation; S5. Perform pressure closed-loop control based on the global PID parameters and pressure feedforward compensation amount obtained in step S3 and the local PID parameters and pressure feedforward compensation amount obtained in step S4 to control the operation of the twin-roll mill production line.

2. The intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness change according to claim 1 is characterized in that: The A1-grade inorganic fireproof core material production line includes a conveying system, an extrusion system and a baking system. The conveying system includes conveying rollers and conveyor mesh belts, and the conveyor mesh belts are installed on the conveying rollers. The baking system includes multiple hot air circulation ovens. The extrusion system includes a first double-roller rolling mill, a second double-roller rolling mill, a third double-roller rolling mill and a thickness and pressure control device. The rolling mills are arranged on a conveyor chain formed by conveying rollers and conveyor mesh belts. The three double-roller rolling mills are respectively arranged between the hot air circulation ovens. The thickness and pressure control device is provided with a thickness sensor and a pressure sensor. The thickness sensor and the pressure sensor are connected to the thickness and pressure control device through a control circuit transmission wire.

3. The intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness variation according to claim 2 is characterized in that: The twin-roll mill includes an upper roll and a lower roll, and a rolling gap is formed between the upper roll and the lower roll for rolling the core material. The twin-roll mill also includes a driving device for driving the rotation of the upper roll and the lower roll. The twin-roll mill adjusts the gap of the twin-roll mill according to the control signal to change the rolling pressure.

4. The intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness variation according to claim 2 is characterized in that: The three twin-roll mills in S2 are arranged in sequence, and the target intermediate thickness of each mill is allocated by the reduction ratio: h1=h0×(1-α1) h2=h1×(1-α2) h target =h2×(1-α3) Where h0 is the initial thickness of the raw material (mm); h1 is the target thickness of the material after the first twin-roll mill (mm); h2 is the target thickness of the material after the second twin-roll mill (mm); target is the target thickness of the material (mm); α1, α2, α3 are the reduction rates of each rolling mill, ranging from 0.2 to 0.

5.

5. The intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness variation according to claim 1 is characterized in that: In S2, the collected thickness data is filtered using a sliding window, specifically: Where, is the original thickness measurement value of the i-th rolling mill at the j-th time step (mm); is the filtered thickness value of the kth data point of the i-th rolling mill (mm).

6. The intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness variation according to claim 1 is characterized in that: The pressure change rate of the hydraulic cylinder at adjacent time points in S2 is specifically: ΔP i (k)=P i (k)-P i (k-1) Where, ΔP i (k) is the pressure change rate of the kth data point of the i-th rolling mill (MPa); P i (k) is the pressure of the kth data point of the i-th rolling mill (MPa); P i (k-1) is the pressure (MPa) of the k-1th data point of the i-th rolling mill.

7. The intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness variation according to claim 1 is characterized in that: In S3, a quantum genetic algorithm is used to optimize based on the fitness function to determine the global PID parameters and the pressure feedforward compensation, specifically: 1) The chromosome encoding rule of quantum genetic algorithm is: Where q m is the state of the mth quantum bit, consisting of two probability amplitudes; cosθ m represents the probability amplitude of the qubit being in state |0>; sinθ m represents the probability amplitude of the qubit being in state |1>; The quantum genetic algorithm takes on the role of global optimization in this process. Each control parameter, including the PID coefficient and pressure compensation, is encoded by an 8-qubit probability amplitude. A single chromosome contains 32 qubits. The quantum state is observed and converted into a classical value: Where K p is the proportional coefficient of the PID controller; K i is the integral coefficient of the PID controller; K d is the differential coefficient of the PID controller; P ff is the pressure feedforward compensation amount (MPa); b m ∈{0,1}, is the quantum bit observation result; and The maximum value of proportional, integral, differential coefficients and pressure feedforward compensation, determined by historical data statistics; 2) Calculate the thickness error, the expression is: Where, is the thickness error of the kth data point of the i-th rolling mill (mm); 3) Calculate the reference pressure, which is obtained from the thickness-velocity model: Where, is the reference pressure of the i-th rolling mill (MPa); is the target thickness of the i-th rolling mill (mm); V is the conveyor belt speed (m / min); a, b, c are material property coefficients, calibrated through orthogonal rolling experiments; 4) Design the fitness function, taking into account both thickness tracking accuracy and pressure stability, and achieve multi-objective optimization through weighted calculation. The formula is as follows: Where J is the fitness function of the quantum genetic algorithm; N is the total number of data points; and λ is the pressure fluctuation penalty weight, which is determined by sensitivity analysis.

8. The intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness variation according to claim 1, characterized in that: The Elman neural network model in S4 is as follows: 1) The Elman neural network model adopts a dynamic structural adjustment strategy. The number of hidden layer nodes automatically increases or decreases according to the thickness deviation. When the thickness error is detected to continuously exceed the threshold, the network automatically adds hidden nodes to enhance the nonlinear fitting ability. When the node activity is insufficient, the streamlining mechanism is triggered to ensure computational efficiency. The network output layer not only generates PID control parameters, but also outputs additional pressure feedforward compensation. 2) Construction of Elman neural network dynamic control input layer: Where x i (k) is the input vector of the kth data point of the i-th rolling mill; is the thickness error change rate of the kth data point of the i-th rolling mill (mm); is the thickness error integral term of the kth data point of the i-th rolling mill (s·mm); T i (k) is the temperature of the kth data point of the i-th rolling mill (°C); T s is the control period (s); Conditions for adding hidden layer nodes: Node deletion conditions: Where, is the average activation of the kth data point of the uth node; d u (j) is the activation value of the u-th node at the j-th time step; Output calculation: Where, is the proportional coefficient of the i-th rolling mill; is the integral coefficient of the i-th rolling mill; is the differential coefficient of the i-th rolling mill; is the pressure feedforward compensation amount of the i-th rolling mill (MPa).

9. The intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness variation according to claim 1, characterized in that: The pressure closed-loop control in S5 is as follows: Where K p is the proportional coefficient of the PID controller; K i is the integral coefficient of the PID controller; K d is the differential coefficient of the PID controller; P ff is the pressure feedforward compensation amount (MPa), is the proportional coefficient of the i-th rolling mill; is the integral coefficient of the i-th rolling mill; is the differential coefficient of the i-th rolling mill; is the pressure feedforward compensation of the i-th rolling mill (MPa), is the thickness error of the kth data point of the i-th rolling mill (mm), is the thickness error change rate of the kth data point of the i-th rolling mill (mm), is the pressure setting value (MPa) of the kth data point of the i-th rolling mill.

10. The intelligent adjustment method for the A1-level inorganic fireproof core material production line based on thickness variation according to claim 9, characterized in that: When abnormal pressure fluctuations or decreased population diversity are detected, the quantum genetic algorithm catastrophe operator automatically resets some chromosomes and injects historical optimal solutions to prevent the algorithm from falling into local optimality. The triggering condition for the catastrophe operator is that the population diversity is <0.1 or the pressure fluctuation rate is in: The formula for calculating population diversity is: Where L is the total number of chromosomes in the population; q l,m is the probability amplitude of the mth quantum bit of the lth chromosome; is the average probability amplitude of all chromosomes in the population at the mth qubit; The pressure fluctuation rate is: Where, is the mean pressure of the i-th rolling mill (MPa).

Images