Intelligent monitoring and optimizing method for industrial solid waste high-strength mineral fiber preparation process

Abstract

The invention belongs to the technical field of industrial fiber preparation, and discloses an intelligent monitoring and optimizing method for an industrial solid waste high-strength mineral fiber preparation process. The method comprises the steps of obtaining smelting furnace temperature data in real time; obtaining furnace temperature influence data; acquiring smelting furnace control data in real time; inputting the smelting furnace temperature data, the furnace temperature influence data and the smelting furnace control data into a pre-trained furnace temperature prediction model, and outputting a furnace temperature prediction value after unit time; based on the smelting furnace temperature data, the smelting furnace control data and the furnace temperature predicted value, strategy optimization is carried out on a furnace temperature control strategy according to a target condition through a self-adaptive feedback algorithm, and an optimal control strategy is output; the target conditions comprise the minimum target temperature difference and the minimum heat loss; generating control information based on the furnace temperature prediction value and a preset warning value; the working efficiency of the smelting furnace is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial fiber preparation, and more specifically, to an intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste. Background Art

[0002] In the process of preparing high-strength mineral fibers from industrial solid waste, the molten solid waste raw material is a crucial link, which determines whether the solid waste raw material can form a uniform high-temperature melt and provides a basis for subsequent fiberization steps and ultimately excellent finished products.

[0003] The existing Chinese patent application with the publication number CN110673563A discloses a high-temperature molten metal container monitoring and early warning system and implementation method, including a control center, a container number matching system, a container wall infrared detection imaging system, a container wall infrared image real-time analysis system, a container wall full-life temperature monitoring database, a container wall temperature early warning and alarm system, and a container wall refractory full-life tracking monitoring analysis and query system. The control center is provided with a subsystem for driving management; this invention uses infrared detection imaging technology to perform infrared detection imaging and analysis processing on the high-temperature molten metal container during hoisting or rotation through an infrared camera, obtains effective data information corresponding to the container number, and gives early warnings and alarms for abnormal wall temperatures according to the warning and alarm values set by the program, forming historical trends and data of the corresponding container for easy query and analysis, and eliminating the risk of leakage of high-temperature molten metal containers.

[0004] The existing technology still has the following problems:

[0005] Most of the existing technologies are for the safety monitoring of high-temperature molten metal containers to judge whether there is a risk of leakage. Only the safety is monitored, and a closed-loop feedback optimization system based on real-time temperature data and predicted temperature data for the control parameters in the high-temperature molten metal container is not formed, resulting in a lack of high automation in the operation of the high-temperature molten metal container, so the working efficiency of the high-temperature molten metal container needs to be further improved.

[0006] In view of this, the present invention proposes an intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste to solve the above problems. Summary of the Invention

[0007] In order to overcome the above defects of the existing technology and achieve the above object, the present invention provides the following technical solutions: An intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste, including:

[0008] Obtain furnace temperature data in real time;

[0009] Obtain data affecting furnace temperature;

[0010] Obtain the furnace control data in real time;

[0011] Based on the furnace temperature data, furnace temperature influence data, and furnace control data, input them into a pre-trained furnace temperature prediction model to output the predicted furnace temperature value after a unit time;

[0012] Based on the furnace temperature data, furnace control data, and predicted furnace temperature value, through an adaptive feedback algorithm, optimize the control strategy of the furnace temperature according to the target conditions to output the optimal control strategy; the target conditions include the minimum target temperature difference and the minimum heat loss;

[0013] Generate control information based on the predicted furnace temperature value and the preset warning value.

[0014] Furthermore, the furnace control data includes the heating power and the stirring rate;

[0015] The specific method for optimizing the control strategy of the furnace temperature according to the target conditions through the adaptive feedback algorithm includes:

[0016] Step A1: Randomly generate kinds of control strategies; the control strategies include the heating power update value and the stirring rate update value;

[0017] Step A2: Initialize the particle swarm, and each particle represents a control strategy;

[0018] Step A3: Calculate the target temperature difference and heat loss of each control strategy;

[0019] Step A4: Calculate the comprehensive fitness of kinds of control strategies according to the target temperature difference and heat loss ;

[0020] Step A5: Update the velocity and position of the particles, that is, update the heating power update value and the stirring rate update value;

[0021] Step A6: If the target temperature difference of the th particle after updating the position is less than the target temperature difference corresponding to the historical best position, then let ; if the target temperature difference of the th particle after updating the position is less than the target temperature difference corresponding to the global best position, then let ; is the position of the th particle at the th iteration; is the historical best position found by the th particle itself in the corresponding control strategy, and the best position in the corresponding control strategy is the one with the minimum target temperature difference; The global (i.e., all control strategies) best position found among all particles;

[0022] Step A7. Repeat steps A3 - A6 until the preset number of repetitions is reached, then select the control strategy with the lowest comprehensive fitness and output the optimized control strategy as the optimal control strategy.

[0023] Furthermore, the training method of the furnace temperature prediction model includes:

[0024] Obtain a set of historical furnace temperature datasets, where the furnace temperature datasets include furnace temperature data, furnace temperature influence data, furnace control data, and the corresponding temperature calibration values after a unit time; use the furnace temperature datasets as the sample set, divide the sample set into a training set and a test set; construct an RNN classifier, use the furnace temperature data, furnace temperature influence data, and furnace control data in the training set as the input of the furnace temperature prediction model, use the temperature calibration values after a unit time in the training set as the target output, perform iterative training on the classifier to obtain a preliminary prediction model; use the test set to test the preliminary prediction model, if the accuracy of the preliminary prediction model meets the preset accuracy, then use the preliminary prediction model as the furnace temperature prediction model; the preliminary prediction model is an RNN neural network model; is a positive integer greater than 1.

[0025] Furthermore, subtract the preset furnace temperature target value from the furnace temperature prediction value to calculate the target temperature difference.

[0026] Furthermore, the heat loss is calculated from heat conduction loss, radiation loss, and convection loss.

[0027] Furthermore, the comprehensive fitness is calculated by the dynamic Chebyshev formula.

[0028] Furthermore, the velocity of the updated particle is calculated from the inertial part, individual cognition, and swarm cognition; the position of the updated particle is obtained by summing the current position of the particle and the velocity of the particle in the current iteration.

[0029] Furthermore, the control information specifically includes:

[0030] If the furnace temperature prediction value is less than the preset warning value, execute the optimal control strategy;

[0031] If the furnace temperature prediction value is greater than or equal to the preset warning value, generate a furnace maintenance instruction.

[0032] Furthermore, it also includes:

[0033] The furnace temperature prediction model is trained based on meta - learning, and the training method includes:

[0034] Step B1: Classify the historical furnace temperature dataset according to different working conditions to obtain subtask sets corresponding to different working conditions. Each subtask set includes a corresponding sub-training set and a sub-test set. The working conditions are preset furnace temperature influence data intervals.

[0035] Step B2: Select an RNN neural network model and obtain the initial model parameters ;

[0036] Step B3: Perform inner-loop updates on the initial model parameters. Randomly select subtask sets and perform times of gradient descent on the model, update and record the current model parameters ; is a positive integer less than or equal to ; is any positive integer;

[0037] Step B4: Perform outer-loop updates on the initial model parameters. Calculate the loss value for the validation set corresponding to the subtask set through the loss function, backpropagate the loss value to the initial model parameters, and perform meta-updates through the outer-loop update formula to obtain the updated model parameters ;

[0038] Step B5: Repeat Step B3 and Step B4 until the average prediction error reaches the preset value, and use the model parameters updated last time as the corresponding parameters of the finally deployed furnace temperature prediction model.

[0039] Furthermore, the furnace temperature influence data includes furnace physical data, raw material characteristic data, and environmental data. The furnace physical data is the inner surface area of the furnace. The raw material characteristic data includes the type of raw materials, the corresponding raw material ratio, and the total amount of raw materials. The environmental data includes environmental temperature and environmental humidity.

[0040] The technical effects and advantages of the intelligent monitoring and optimization method for the industrial solid waste high-strength mineral fiber preparation process of the present invention:

[0041] By obtaining the furnace temperature data, furnace temperature influence data, and furnace control data, the present invention first outputs the predicted furnace temperature value after a unit time through a pre-trained furnace temperature prediction model, then optimizes the control strategy of the furnace temperature according to the minimum target temperature difference and the minimum heat loss through an adaptive feedback algorithm, and finally makes a decision on whether to execute the optimized strategy or perform furnace maintenance based on the predicted furnace temperature value and the preset warning value, forming a closed-loop feedback optimization based on real-time temperature data and predicted temperature data, which not only improves the automation level of the entire preparation process but also effectively improves the working efficiency of the furnace.

[0042] In addition, the present invention optimizes the control strategy through an adaptive feedback algorithm based on the minimum target temperature difference and the minimum heat loss, not only effectively maintaining the stability and uniformity of the melt temperature, providing an ideal melting basis for the subsequent fiber chemical process, but also effectively reducing energy consumption and reducing energy waste.

[0043] Finally, control information is generated by comparing the predicted furnace temperature with the preset warning value: when the furnace is within the safe range, executing the optimized strategy can maximize the process efficiency; while when there are potential safety hazards in the furnace, maintenance and warning instructions are triggered in a timely manner, taking into account both process efficiency and equipment safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the intelligent monitoring and optimization system for the preparation process of industrial solid waste high-strength mineral fibers in Embodiment 1 of the present invention;

[0045] Figure 2 It is a flowchart of the intelligent monitoring and optimization method for the preparation process of industrial solid waste high-strength mineral fibers in Embodiment 2 of the present invention;

[0046] Figure 3 It is a flowchart of the adaptive feedback algorithm in Embodiment 1 of the present invention;

[0047] Figure 4 It is a schematic diagram of the closed-loop feedback system in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0049] Embodiment 1

[0050] Please refer to Figure 1 As shown, the intelligent monitoring and optimization system for the preparation process of industrial solid waste high-strength mineral fibers in this embodiment includes: a first acquisition module, a second acquisition module, a third acquisition module, a temperature prediction module, a feedback control module, and an execution module, and each module is connected by wire and / or wirelessly.

[0051] The first acquisition module acquires furnace temperature data in real time.

[0052] The temperature of the melt directly affects the melting process of industrial solid waste raw materials. Too high or too low a temperature will affect the product quality and may even cause safety accidents. Too high a temperature may damage the furnace equipment, and too low a temperature may affect the quality of subsequent products. The real-time furnace temperature data is the main input feature of the subsequent furnace temperature prediction model, providing accurate initial conditions for temperature prediction. At the same time, by obtaining the real-time furnace temperature data and providing closed-loop feedback to the algorithm through real-time update, the furnace can maintain the best temperature control and energy consumption level in a changing production environment.

[0053] The furnace temperature data is obtained by an optical fiber temperature sensor.

[0054] A second acquisition module acquires furnace temperature influence data. The furnace temperature influence data includes furnace physical data, raw material characteristic data, and environmental data. The furnace physical data is the inner surface area of the furnace. The raw material characteristic data includes the types of raw materials, the corresponding raw material ratios, and the total amount of raw materials. The environmental data includes the environmental temperature and the environmental humidity.

[0055] The inner surface area of the furnace is an important parameter affecting the magnitude of radiation loss, and the radiation loss is proportional to the inner surface area. The larger the inner surface area, the more heat loss of the furnace. In the furnace temperature prediction model, the inner surface area is input as a static feature, effectively enhancing the adaptability and robustness of the furnace temperature prediction model to the thermal characteristics of the furnace.

[0056] In the preparation of high-strength mineral fibers from industrial solid waste, various solid wastes (such as fly ash, slag, construction waste, etc.) are mixed and used in a certain proportion. For raw materials with different batches and different ratios, due to the differences in the thermophysical and chemical properties of various solid waste raw materials, the temperature rise characteristics of the furnace will fluctuate.

[0057] The types of raw materials and the corresponding raw material ratios affect the specific heat capacity, thermal conductivity, and melting point of the melt. When the specific heat capacity of the melt in the furnace is relatively high, when the furnace rises from room temperature to the target temperature, for example, from 20 degrees Celsius to 1500 degrees Celsius, the required heating time is longer and the energy consumed by the furnace is also more. Raw materials with high thermal conductivity can conduct heat quickly, making the heat distribution in the furnace more uniform and helping to accelerate the overall heating speed. Different melts have different melting temperatures. If there are components with high melting points, their melting requires a higher temperature, resulting in more energy consumption and more time spent by the furnace during heating, affecting the setting of subsequent temperature target values. The larger the total amount of raw materials, the lower the temperature that the furnace can reach under the condition of the same heating power. By obtaining the raw material characteristic data during each heating, it helps to improve the prediction ability of the subsequent furnace temperature prediction model for the actual working conditions of the furnace.

[0058] The ambient temperature and the furnace temperature jointly affect the heat conduction loss and radiation loss of the furnace during the melting of the melt; the ambient humidity affects the convection loss of the furnace during the melting of the melt; the weighted sum of the heat conduction loss, radiation loss and convection loss reflects the heat loss of the furnace during the melting of the melt; by obtaining the ambient data, not only the prediction accuracy of the subsequent furnace temperature prediction model is improved, but also the accuracy of the heat loss calculation in the subsequent target conditions is improved.

[0059] The inner surface area is obtained by calculating the CAD model of the furnace; the raw material characteristic data is updated every time batching is carried out and is obtained through the batching system of the factory; the ambient temperature is obtained in real time through an industrial temperature sensor; the ambient humidity is obtained in real time through a high-temperature resistant humidity sensor.

[0060] A third acquisition module acquires furnace control data in real time; the furnace control data includes heating power and stirring rate.

[0061] The heating power directly determines the heating rate of the furnace and the temperature level maintained by the melt; the greater the heating power, the faster the furnace temperature rises, that is, the higher the temperature after a unit time; by adjusting and maintaining the heating power per unit time, the use of energy can be effectively optimized; by maintaining a stable heating power, the production process can operate in a relatively balanced state, reducing the time of fault shutdown or repeated debugging and improving the overall production efficiency.

[0062] In high-temperature furnace processes, such as the preparation of high-strength mineral fibers from industrial solid waste, the stirring rate will have a certain impact on the rate of temperature rise (or fall) in the furnace; when the stirring rate is high, the temperature gradient in the melt decreases and the overall temperature rises more rapidly; at the same time, under the same heating power, high-speed stirring can quickly disperse the heat inside the melt and accelerate the time for the overall temperature to reach the standard; conversely, if the stirring is insufficient, the heating process is slower.

[0063] The heating power is obtained through a power quality analyzer; the stirring rate is obtained through a motor speed sensor.

[0064] A temperature prediction module inputs the furnace temperature data, furnace temperature influence data and furnace control data into a pre-trained furnace temperature prediction model and outputs the furnace temperature prediction value after a unit time.

[0065] The training method of the furnace temperature prediction model includes:

[0066] Obtain A set of historical furnace temperature datasets, where the furnace temperature datasets include furnace temperature data, furnace temperature influence data, furnace control data, and the corresponding temperature calibration values after a unit time; using the furnace temperature datasets as a sample set, dividing the sample set into a training set and a test set; constructing an RNN classifier, using the furnace temperature data, furnace temperature influence data, and furnace control data in the training set as the input of the furnace temperature prediction model, using the temperature calibration values after a unit time in the training set as the target output, iteratively training the classifier to obtain a preliminary prediction model; using the test set to test the preliminary prediction model, if the accuracy of the preliminary prediction model meets the preset accuracy, then using the preliminary prediction model as the furnace temperature prediction model; the preliminary prediction model is an RNN neural network model; is a positive integer greater than 1.

[0067] It should be noted that obtaining the historical furnace temperature datasets requires integrating the furnace temperature data, furnace temperature influence data, and furnace control data, aligning them at the same timestamp, and combining the measured temperature values after a certain time as calibration labels to form a complete "input-output" sample.

[0068] A feedback control module, based on the furnace temperature data, furnace control data, and furnace temperature prediction value, through an adaptive feedback algorithm, optimizes the control strategy of the furnace temperature according to the target conditions, and outputs the optimized optimal control strategy; the target conditions include the minimum target temperature difference and the minimum heat loss.

[0069] The target temperature difference is the difference between the furnace temperature prediction value and the furnace temperature target value; the furnace temperature target value is a preset value by the staff according to the raw material characteristic data; the minimum heat loss is the weighted sum of the heat conduction loss, radiation loss, and convection loss during heating.

[0070] Please refer to Figure 3 As shown, the specific method for optimizing the control strategy of the furnace temperature according to the target conditions through the adaptive feedback algorithm includes:

[0071] Step A1, randomly generate kinds of control strategies; the control strategies include the heating power update value and the stirring rate update value.

[0072] Step A2, initialize the particle swarm, and each particle represents a control strategy.

[0073] Step A3, calculate the target temperature difference and heat loss of each control strategy.

[0074] The specific method for obtaining the target temperature difference specifically includes:

[0075] ;

[0076] In the formula, is the target temperature difference; is the predicted furnace temperature value; is the target furnace temperature value.

[0077] The method for obtaining the heat loss specifically includes:

[0078] ;

[0079] In the formula, is the heat loss; is the thermal conductivity of the furnace material, which is a preset value; is the inner surface area of the furnace; is the ambient temperature at the current moment; is the unit time length; is the average wall thickness of the furnace, which is a preset value; is the Boltzmann constant, which is a preset value; is the surface emissivity of the furnace, obtained from the material reference manual of the furnace; is the effective radiation area of the furnace, which is a preset value; is the convective heat transfer coefficient, preset by the staff according to experience or industrial manuals; is the convective heat transfer area of the furnace, preset by the staff according to the size of the actual outer wall area; , and are the corresponding preset weights of the heat conduction loss, radiation loss, and convective loss respectively.

[0080] It should be noted that the error caused by using the ambient temperature at the current moment to represent the average temperature within the unit time can be corrected through the closed-loop control of the entire system; in the calculation of the heat loss, by using the ambient temperature at the current moment, the algorithm is greatly simplified in the case of short cycles and relatively stable environmental fluctuations.

[0081] Step A4, calculate the comprehensive fitness of control strategies according to the target temperature difference and the heat loss ; The comprehensive fitness is obtained by calculating through the dynamic Chebyshev formula; The dynamic Chebyshev formula specifically includes:

[0082] ;

[0083] In the formula, and are preset weight coefficients; is the maximum value of the target temperature difference in all current particle swarms; is the minimum value of the target temperature difference in all current particle swarms; is the maximum value of the heat loss in all current particle swarms; is the minimum value of heat loss in all current particle swarms, is the function to find the maximum value.

[0084] Step A5: Update the velocity and position of the particles, that is, update the updated value of the heating power and the updated value of the stirring rate; The method for updating the velocity of the particles includes:

[0085] ;

[0086] In the formula, is the velocity of the th particle at the th iteration; is the velocity of the th particle at the th iteration; is the position of the th particle at the th iteration; is the historical best position found by the th particle itself in the corresponding control strategy. The best position in the corresponding control strategy is the minimum target temperature difference; is the global (i.e., all control strategies) best position found among all particles; is the inertia weight, which controls the continuity of the particle velocity; is the individual learning factor, The larger the value of is the swarm learning factor, The larger the value of and are both random numbers in [0, 1], which increases the randomness of the search.

[0087] It should be noted that the smaller the inertia weight, the smaller the amplitude of each movement, which is convenient for fine-tuning; for example, when the furnace operation enters the stable stage, it is necessary to fine-tune near the optimal heating power and stirring rate to reduce temperature fluctuations and save energy consumption.

[0088] Exemplarily, the parameters of the th particle are as follows: ; ; ; ; ; ; ; ; .

[0089] Inertia term calculation:

[0090] 。

[0091] Individual cognitive item calculation:

[0092] 。

[0093] Group cognitive item calculation:

[0094] 。

[0095] Update speed calculation:

[0096] 。

[0097] The method for updating the position of the particle includes:

[0098] 。

[0099] Exemplarily, the updated position of the th particle is:

[0100] 。

[0101] Step A6. If the target temperature difference of the th particle after updating the position is less than the target temperature difference corresponding to the historical best position, then let ; if the target temperature difference of the th particle after updating the position is less than the target temperature difference corresponding to the global best position, then let 。

[0102] Step A7. Repeat steps A3 - A6 until the preset number of repetitions is reached, then select the control strategy with the lowest comprehensive fitness, and output the optimized control strategy as the optimal control strategy to meet the set target conditions.

[0103] An execution module generates control information based on the predicted furnace temperature value and the preset warning value.

[0104] The control information specifically includes:

[0105] If the predicted furnace temperature value is less than the preset warning value, execute the optimal control strategy;

[0106] If the predicted furnace temperature value is greater than or equal to the preset warning value, generate a furnace maintenance instruction to ensure the safety of the process.

[0107] It should be noted that the preset warning value of the furnace temperature is greater than the preset furnace temperature target value.

[0108] Please refer to Figure 4As shown in the figure, the present invention forms a complete closed-loop feedback system by building the above modules:

[0109] First, obtain the temperature data of the furnace in real time; then, generate the predicted furnace temperature value through the furnace temperature prediction model; subsequently, make a decision based on the predicted furnace temperature value: if the predicted furnace temperature value is lower than the preset warning value, execute the optimal control strategy through the execution module; if the predicted furnace temperature value is greater than or equal to the preset warning value, generate the corresponding furnace maintenance instruction; finally, dynamically adjust the future furnace temperature data according to the optimal control strategy to achieve precise optimization of the closed-loop control and the feedback mechanism.

[0110] Embodiment 2

[0111] Please refer to Figure 2 As shown in the figure, this embodiment provides an intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste, including:

[0112] Obtain the furnace temperature data in real time;

[0113] Obtain the data affecting the furnace temperature;

[0114] Obtain the furnace control data in real time;

[0115] Based on the furnace temperature data, the data affecting the furnace temperature, and the furnace control data, input them into the pre-trained furnace temperature prediction model, and output the predicted furnace temperature value after a unit time;

[0116] Based on the furnace temperature data, the furnace control data, and the predicted furnace temperature value, through the adaptive feedback algorithm, optimize the control strategy of the furnace temperature according to the target conditions, and output the optimal control strategy; the target conditions include the minimum target temperature difference and the minimum heat loss;

[0117] Generate control information based on the predicted furnace temperature value and the preset warning value.

[0118] Embodiment 3

[0119] When there is a new working condition requirement, the prediction accuracy of the traditional furnace temperature prediction model will have a large error; however, in actual production, in order to improve production efficiency, the demand for quickly switching working conditions is increasing. Therefore, the present invention also provides a furnace temperature prediction model based on meta-learning to further improve the generalization ability and rapid adaptation ability of the model.

[0120] The core of the meta - learning lies in enabling the furnace temperature prediction model to conduct meta - training on multiple training tasks with diverse distributions, so that the furnace temperature prediction model can converge quickly when facing new tasks, efficiently and improve the accuracy of the model; traditional furnace temperature prediction models rely more on a "single large - scale dataset". Once the working conditions change (the distribution changes significantly), they often need to start from scratch or conduct large - scale fine - tuning; while meta - learning aims to extract general "initial parameters" or "learning strategies" from multi - task training, and only requires very few iterations to achieve considerable accuracy in new working conditions.

[0121] The training method of the furnace temperature prediction model based on meta - learning includes:

[0122] Step B1: Classify the historical furnace temperature dataset according to different working conditions to obtain sub - task sets corresponding to different working conditions. Each sub - task set includes a corresponding sub - training set and a sub - test set; the working condition is a preset data interval affecting the furnace temperature.

[0123] Step B2: Select an RNN neural network model and obtain the initial model parameters .

[0124] Step B3: Conduct inner - loop update on the initial model parameters. Randomly select sub - task sets, conduct times of gradient descent on the model, update and record the current model parameters ; is a positive integer less than or equal to ; is an arbitrary positive integer.

[0125] Step B4: Conduct outer - loop update on the initial model parameters. Calculate the loss value for the validation set corresponding to the sub - task set through the loss function, back - propagate the loss value to the initial model parameters, and conduct meta - update through the outer - loop update formula to obtain the updated model parameters , so that the model parameters can quickly adapt to multiple tasks with a small amount of data and update steps, thus enabling the furnace temperature prediction model to perform well when facing new tasks; the outer - loop update formula is:

[0126] ;

[0127] where, is the learning rate of the outer - loop update; is the descending gradient of the initial model parameters ; is the th current model parameter corresponding to the sub - task set; is the loss function.

[0128] Step B5: Repeat Step B3 and Step B4 until the average prediction error reaches a preset value, and use the model parameters updated last time as the corresponding parameters of the finally deployed furnace temperature prediction model.

[0129] By building a furnace temperature prediction model based on meta-learning, the prediction model can be quickly adjusted according to new working conditions, improving the flexibility and automation level of the production process. At the same time, when quickly switching working conditions in actual production, a high prediction accuracy can still be maintained, significantly improving production efficiency.

[0130] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. An intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste, characterized in that, Including: Obtaining the furnace temperature data in real time; Obtaining the data affecting the furnace temperature; Obtaining the furnace control data in real time; Based on the furnace temperature data, the data affecting the furnace temperature, and the furnace control data, inputting them into a pre-trained furnace temperature prediction model, and outputting the predicted furnace temperature value after a unit time; Based on the furnace temperature data, the furnace control data, and the predicted furnace temperature value, through an adaptive feedback algorithm, optimizing the control strategy of the furnace temperature according to the target conditions, and outputting an optimal control strategy; the target conditions include the minimum target temperature difference and the minimum heat loss; Generating control information based on the predicted furnace temperature value and a preset warning value.

2. The intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste according to claim 1, characterized in that The furnace control data includes the heating power and the stirring rate; The specific method for optimizing the control strategy of the furnace temperature according to the target conditions through the adaptive feedback algorithm includes: Step A1: Randomly generate control strategies; the control strategies include a heating power update value and a stirring rate update value; Step A2: Initializing the particle swarm, where each particle represents a control strategy; Step A3: Calculating the target temperature difference and the heat loss of each control strategy; Step A4. Calculate the comprehensive fitness of the control strategy according to the target temperature difference and heat loss of a certain control strategy ; Step A5: Updating the velocity and position of the particle, that is, updating the updated value of the heating power and the updated value of the stirring rate; Step A6. If the target temperature difference of the th particle after updating the position is less than the target temperature difference corresponding to the historical best position, then let ; if the target temperature difference of the th particle after updating the position is less than the target temperature difference corresponding to the global best position, then let ; is the position of the th particle at the th iteration; is the historical best position found by the th particle itself in the corresponding control strategy, and the best position in the corresponding control strategy is the one with the minimum target temperature difference; is the global best position found among all particles; Step A7: Repeating steps A3 - A6 until the preset number of repetitions is reached, then selecting the control strategy with the lowest comprehensive fitness and outputting the optimized control strategy as the optimal control strategy.

3. The intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste according to claim 1, characterized in that The training method of the furnace temperature prediction model includes: Obtain A set of historical furnace temperature data sets, where the furnace temperature data sets include furnace temperature data, furnace temperature influence data, furnace control data, and the corresponding temperature calibration values after a unit time; taking the furnace temperature data sets as the sample set, dividing the sample set into a training set and a test set; constructing an RNN classifier, using the furnace temperature data, furnace temperature influence data, and furnace control data in the training set as the input of the furnace temperature prediction model, using the temperature calibration value after a unit time in the training set as the target output, iteratively training the classifier to obtain a preliminary prediction model; using the test set to test the preliminary prediction model, if the accuracy of the preliminary prediction model meets the preset accuracy, then taking the preliminary prediction model as the furnace temperature prediction model; the preliminary prediction model is an RNN neural network model; Is a positive integer greater than 1.

4. The intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste according to claim 2, characterized in that, Subtracting the preset furnace temperature target value from the predicted furnace temperature value to calculate the obtained target temperature difference.

5. The intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste according to claim 2, characterized in that, The heat loss is obtained by calculating the heat conduction loss, the radiation loss, and the convection loss.

6. The intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste according to claim 2, characterized in that, The comprehensive fitness is obtained by calculating through the dynamic Chebyshev formula.

7. The intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste according to claim 2, characterized in that The velocity of updating the particle is obtained by calculating the inertia part, the individual cognition, and the group cognition; the position of updating the particle is obtained by summing the current position of the particle and the velocity of the current iteration of the particle.

8. The intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste according to claim 1, characterized in that, The control information specifically includes: If the predicted furnace temperature value is less than the preset warning value, execute the optimal control strategy; If the predicted furnace temperature value is greater than or equal to the preset warning value, generate a furnace maintenance instruction.

9. The intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste according to claim 1, characterized in that, The data affecting the furnace temperature includes the furnace physical data, the raw material characteristic data, and the environmental data; the furnace physical data is the inner surface area of the furnace; the raw material characteristic data includes the raw material type, the corresponding raw material ratio, and the total amount of raw materials; the environmental data includes the environmental temperature and the environmental humidity.

10. The intelligent monitoring and optimization method for the preparation process of high-strength mineral fibers from industrial solid waste according to claim 1, characterized in that, Also including: The furnace temperature prediction model is trained based on meta - learning, and the training method includes: Step B1: Classifying the historical furnace temperature dataset according to different working conditions to obtain sub - task sets corresponding to different working conditions, and each sub - task set includes a corresponding sub - training set and a sub - test set; the working condition is a preset furnace temperature influence data interval; Step B2: Select an RNN neural network model and obtain initial model parameters ; Step B3: Perform inner-loop update on the initial model parameters, randomly select sub-task sets, and perform times of gradient descent on the model, update and record the current model parameters ; is a positive integer less than or equal to ; is any positive integer; Step B4: Perform outer-loop update on the initial model parameters. Calculate the loss value for the validation set corresponding to the subtask set through the loss function, backpropagate the loss value to the initial model parameters, and perform meta-update through the outer-loop update formula to obtain the updated model parameters ; Step B5: Repeating steps B3 and B4 until the average prediction error reaches the preset value, and taking the model parameters updated last time as the corresponding parameters of the finally deployed furnace temperature prediction model.

Images