Intelligent optimization method for annealing temperature curve of high-silica glass fiber

Abstract

The invention discloses an intelligent optimization method for an annealing temperature curve of a high-silica glass fiber, and relates to the technical field of high-silica glass fiber processing, which thoroughly gets rid of the dependence on artificial experience, can calculate a theoretical optimal annealing curve under a specific working condition by constructing a data-driven prediction model and utilizing an intelligent algorithm for optimization, and improves the processing accuracy of the high-silica glass fiber. Accurate temperature control is achieved, changes of working conditions such as the wire drawing speed and the fiber diameter can be sensed in real time, an annealing temperature curve is dynamically adjusted when the working conditions fluctuate, the uniformity and stability of product quality under different working conditions are ensured, and meanwhile through collaborative optimization of multiple targets such as residual stress and mechanical properties, the product quality is improved. According to the method, the residual stress of the high-silica glass fiber can be remarkably reduced, the strength and flexibility of the fiber can be improved, so that the yield is improved, the energy consumption is reduced, the method has a model self-adaptive updating capability, and through a production-detection-feedback-learning closed loop, the system can adapt to long-term working condition changes, and accumulation and iterative optimization of process knowledge are realized.

Description

Technical Field

[0001] This invention relates to the field of high-silica glass fiber processing technology, and in particular to a method for intelligent optimization of the annealing temperature profile of high-silica glass fiber. Background Technology

[0002] High-silica glass fiber is widely used in cutting-edge fields such as aerospace, high-temperature insulation, and special composite materials due to its excellent high-temperature resistance, dielectric properties, and chemical stability. However, during the drawing and forming process of high-silica glass fiber, the rapid temperature change generates large residual stress inside, resulting in brittleness, reduced strength, and poor flexibility.

[0003] Annealing is a key process for eliminating or reducing residual stress in high-silica glass fibers and improving their mechanical properties. The core of the annealing process lies in the annealing temperature curve, which is the temperature change over time or location of the fiber during the "heating-holding-cooling" process in the annealing furnace.

[0004] In existing technologies, the setting of annealing temperature profiles mainly relies on the following methods: Fixed process method: This method uses one or several fixed temperature profiles that have been verified through long-term production. However, this method cannot adapt to dynamic changes in the production process, such as fluctuations in drawing speed, uneven fiber diameter, and slight differences in raw material composition. Manual experience adjustment: Relying on the experience of senior engineers, the set temperature of each temperature zone of the annealing furnace is manually adjusted based on the test results of the final product. This method is lagging, has poor accuracy, is highly subjective, and is difficult to achieve optimal control. The aforementioned existing technologies all have a core flaw: they cannot dynamically and accurately adjust and optimize the annealing temperature curve according to real-time changes in production conditions. This leads to large fluctuations in the product quality of high silica glass fiber, making it difficult to improve the yield, and may also cause unnecessary energy waste. To address the aforementioned technical deficiencies, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to: construct a data-driven predictive model and utilize intelligent algorithms for optimization, thereby calculating the theoretically optimal annealing curve under specific working conditions, achieving precise temperature control, and enabling real-time sensing of changes in working conditions such as drawing speed and fiber diameter. The annealing temperature curve is dynamically adjusted when working conditions fluctuate, ensuring the uniformity and stability of product quality under different working conditions. Simultaneously, through the synergistic optimization of multiple objectives such as residual stress and mechanical properties, the residual stress of high-silica glass fiber can be significantly reduced, improving fiber strength and flexibility, thereby increasing yield and reducing energy consumption. The system possesses adaptive model updating capabilities, and through a closed loop of production-detection-feedback-learning, it can adapt to long-term changes in working conditions, achieving the accumulation and iterative optimization of process knowledge.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for intelligent optimization of the annealing temperature curve of high-silica glass fiber, comprising the following steps: S1. Real-time acquisition of multi-dimensional data and database construction of historical data: Acquire and integrate multi-source data in the production process of high silica glass fiber, including real-time process parameters, historical process parameters and product quality indicators; S2. Constructing an annealing process-fiber quality prediction model f: Based on the historical process parameters and corresponding product quality indicators obtained in S1, a quality prediction model is trained and constructed using machine learning algorithms. The quality prediction model is used to characterize the mapping relationship between key parameters of the annealing temperature curve and real-time process parameters to the final product quality indicators. S3. Set the multi-objective optimization function J: Set the multi-objective optimization function according to product requirements and production goals; S4. Intelligent optimization of annealing temperature curve: The real-time process parameters collected in S1 are used as the input of the current working condition. The intelligent optimization algorithm is used with the optimization function J set in S3 as the optimization target and the prediction model f constructed in S2 as the fitness function. The global search is performed in the feasible region of the key parameters of the annealing temperature curve to obtain the optimal key parameters of the annealing temperature curve under the current working condition. S5. Optimal Curve Analysis and Closed-Loop Execution: The key parameters of the optimal annealing temperature curve obtained in S4 are analyzed into a sequence of specific temperature setpoints for the internal zones of the annealing furnace. This setpoint sequence is then sent to the underlying control system of the annealing furnace. The actual temperature of each zone is monitored in real time by a high-precision temperature sensor. A PID control algorithm is used for high-speed closed-loop adjustment to ensure that the actual temperature curve accurately tracks the optimal setpoint curve.

[0007] Furthermore, the real-time process parameters specifically include drawing speed v, raw wire diameter d, and actual temperature T of each temperature zone in the annealing furnace; The product quality indicators are the finished fiber mechanical properties of standard products within the historical production cycle, specifically including tensile strength, elongation at break and residual stress level. The key parameters of the annealing temperature curve include the heating rate, peak holding temperature, holding time, and cooling rate.

[0008] Furthermore, the specific process of constructing the quality prediction model is as follows: S201. Obtain real-time process parameters and key parameters of the annealing temperature curve, and integrate them into input features, the vector representation of which is as follows: Where v is the drawing speed, d is the diameter of the raw wire, and R is the wire diameter. UP T represents the heating rate, Tpeak represents the peak holding temperature, and t represents the heating rate. 所以 For effective heat preservation time, R 首席运营官 Cooling rate; S202. Determine the output labels of the model as follows: , where σ is the tensile strength, S is the residual stress level, and ε is the elongation at break; S203. Obtain historical process parameters. The historical process parameters are a set of process parameters for standard products within a historical production cycle. After cleaning and normalizing the historical process parameters, training samples are obtained. The training samples are divided into training set, validation set and test set in a ratio of 8:1:1. S204. Construct a feedforward neural network, including: Input layer: The number of nodes is the same as the dimension of the input features; Hidden layers: Set 2 to 5 fully connected hidden layers; The selection of the number of hidden layer nodes and the number of layers is optimized by tuning the performance on the validation set; Activation function: The hidden layer preferably uses the ReLU activation function to increase nonlinear expressiveness and accelerate convergence; Output layer: The number of nodes is the same as the dimension of the output labels; S205. During the training process, small batches of samples are randomly and non-repeatedly drawn from the training set for training. The training cycle is completed after all training samples are drawn. The training is completed after a certain number of cycles, and the quality prediction model is obtained.

[0009] Furthermore, the specific process for defining the multi-objective optimization function is as follows: S301, The multi-objective optimization function is expressed as follows: , where X represents the key parameters of the annealing temperature curve that S4 is searching for, ω1, ω2, ω3 are preset weight coefficients, and ω1 + ω2 + ω3 = 1, f1, f2, f3 are the normalized cost functions of tensile strength, residual stress level and elongation at break, respectively. S302, respectively, limit f1, f2, and f3: If we set an ideal target intensity σtarget, then the cost function f1 is the squared deviation from the target value: , where σPerd=fσ(X), which is the predicted output value of the mass prediction model for tensile strength; Set the minimum resolution S 分辨率 The resolution prediction value S is output through the quality prediction model. res-p回复d The objective function is constructed as follows: ; To minimize the process energy consumption E, the objective function is constructed as follows: T peak t represents the peak insulation temperature, t represents the effective insulation time, and c1 and c2 are coefficients calibrated according to thermodynamics.

[0010] Furthermore, the specific process of analyzing the optimal curve is as follows: S401. Obtain the physical model and real-time operating parameters of the annealing furnace during the high-silica glass fiber annealing process. The physical model of the annealing furnace includes the number of temperature zones n and the physical degree L of each temperature zone. 我 The furnace coordinates x at the center point of each temperature zone 我 The relationship between the position x of high silica glass fiber in the furnace and time t was constructed: t 我 =x 我 / v, where v is the real-time wire drawing speed; S402. Based on the key parameters of the optimal annealing temperature curve output by S4, construct the ideal time and temperature function Targte(t). S403, Obtain the time point t for each temperature zone from S401. 我 Substituting into Targte(t), we obtain the target temperature of the high-silica glass fiber in the i-th temperature zone: .

[0011] Furthermore, it also includes model adaptive updates and iterations: during the production process, the finished fiber is regularly inspected to obtain actual product quality indicators. This data is used as a new sample and fed back into the prediction model of S2 to update model f online, so as to compensate for the impact of model drift, equipment aging and other factors, and realize the model's self-adaptation and continuous evolution.

[0012] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This intelligent optimization method for the annealing temperature curve of high-silica glass fiber completely eliminates the reliance on manual experience. By constructing a data-driven predictive model and using intelligent algorithms for optimization, it can calculate the theoretically optimal annealing curve under specific working conditions, achieving precise temperature control. It can also sense changes in working conditions such as drawing speed and fiber diameter in real time, dynamically adjusting the annealing temperature curve when working conditions fluctuate, ensuring the uniformity and stability of product quality under different working conditions. At the same time, through the synergistic optimization of multiple objectives such as residual stress and mechanical properties, it can significantly reduce the residual stress of high-silica glass fiber, improve fiber strength and flexibility, thereby increasing yield and reducing energy consumption. It has the ability to adaptively update the model. Through a closed loop of production-detection-feedback-learning, the system can adapt to long-term changes in working conditions, realizing the accumulation and iterative optimization of process knowledge. Attached Figure Description

[0013] Figure 1A schematic diagram of the overall method flow of the present invention is shown. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0015] like Figure 1 As shown, a method for intelligent optimization of the annealing temperature profile of high-silica glass fiber includes the following steps: S1. Real-time acquisition of multi-dimensional data and database construction of historical data: Acquire and integrate multi-source data in the production process of high silica glass fiber, including real-time process parameters, historical process parameters and product quality indicators; Real-time process parameters, specifically including drawing speed v, raw wire diameter d, and actual temperature T of each zone of the annealing furnace; The product quality indicators are the finished fiber mechanical properties of standard products within the historical production cycle, specifically including tensile strength, elongation at break and residual stress level. Key parameters of the annealing temperature profile include heating rate, peak holding temperature, holding time, and cooling rate.

[0016] S2. Constructing an annealing process-fiber quality prediction model f: Based on the historical process parameters and corresponding product quality indicators obtained in S1, a quality prediction model is trained and constructed using machine learning algorithms. The quality prediction model is used to characterize the mapping relationship between key parameters of the annealing temperature curve and real-time process parameters to the final product quality indicators. The specific process of constructing a quality prediction model is as follows: S201. Obtain real-time process parameters and key parameters of the annealing temperature curve, and integrate them into input features, the vector representation of which is as follows: Where v is the drawing speed, d is the diameter of the raw wire, and R is the wire diameter. UP T represents the heating rate, Tpeak represents the peak holding temperature, and t represents the heating rate. 所以 For effective heat preservation time, R 首席运营官 Cooling rate; S202. Determine the output labels of the model as follows: , where σ is the tensile strength, S is the residual stress level, and ε is the elongation at break; S203. Obtain historical process parameters. Historical process parameters are the set of process parameters for standard products within the historical production cycle. After cleaning and normalizing the historical process parameters, training samples are obtained. The training samples are divided into training set, validation set and test set in a ratio of 8:1:1. S204. Construct a feedforward neural network, including: Input layer: The number of nodes is the same as the dimension of the input features; Hidden layers: Set 2 to 5 fully connected hidden layers; The selection of the number of hidden layer nodes and the number of layers is optimized by tuning the performance on the validation set; Activation function: The hidden layer preferably uses the ReLU activation function to increase nonlinear expressiveness and accelerate convergence; Output layer: The number of nodes is the same as the dimension of the output labels; S205. During the training process, small batches of samples are randomly and non-repeatedly drawn from the training set for training. The training cycle is completed after all training samples are drawn. The training is completed after a certain number of cycles, and the quality prediction model is obtained.

[0017] S3. Define the multi-objective optimization function J: Based on product requirements and production goals, define the multi-objective optimization function. The multi-objective optimization function is used to explore the relationship between maximizing key product performance, minimizing defective indicators, minimizing process energy consumption, and production goals. The specific process for setting the multi-objective optimization function is as follows: S301, The multi-objective optimization function is expressed as follows: , where X represents the key parameters of the annealing temperature curve that S4 is searching for, ω1, ω2, ω3 are preset weight coefficients, and ω1 + ω2 + ω3 = 1, f1, f2, f3 are the normalized cost functions of tensile strength, residual stress level and elongation at break, respectively. S302, respectively, limit f1, f2, and f3: If we set an ideal target intensity σtarget, then the cost function f1 is the squared deviation from the target value: , where σPerd=fσ(X), which is the predicted output value of the mass prediction model for tensile strength; Set the minimum resolution S 分辨率 The resolution prediction value S is output through the quality prediction model. res-p回复d The objective function is constructed as follows: ; To minimize the process energy consumption E, the objective function is constructed as follows: Tpeak t represents the peak insulation temperature, t represents the effective insulation time, and c1 and c2 are coefficients calibrated according to thermodynamics.

[0018] S4. Intelligent optimization of annealing temperature curve: The real-time process parameters collected in S1 are used as the input of the current working condition. The intelligent optimization algorithm is used with the optimization function J set in S3 as the optimization target and the prediction model f constructed in S2 as the fitness function. The global search is performed in the feasible region of the key parameters of the annealing temperature curve to obtain the optimal key parameters of the annealing temperature curve under the current working condition. S5. Optimal Curve Analysis and Closed-Loop Execution: The key parameters of the optimal annealing temperature curve obtained in S4 are analyzed into a sequence of specific temperature setpoints for the internal zones of the annealing furnace. This setpoint sequence is then sent to the underlying control system of the annealing furnace. The actual temperature of each zone is monitored in real time by a high-precision temperature sensor. A PID control algorithm is used for high-speed closed-loop adjustment to ensure that the actual temperature curve accurately tracks the optimal setpoint curve.

[0019] The specific process of analyzing the optimal curve is as follows: S401. Obtain the physical model and real-time operating parameters of the annealing furnace during the high-silica glass fiber annealing process. The physical model of the annealing furnace includes the number of temperature zones n and the physical degree L of each temperature zone. 我 The furnace coordinates x at the center point of each temperature zone 我 The relationship between the position x of high silica glass fiber in the furnace and time t was constructed: t 我 =x 我 / v, where v is the real-time wire drawing speed; S402. Based on the key parameters of the optimal annealing temperature curve output by S4, construct the ideal time and temperature function Targte(t). S403, Obtain the time point t for each temperature zone from S401. 我 Substituting into Targte(t), we obtain the target temperature of the high-silica glass fiber in the i-th temperature zone: .

[0020] It also includes model adaptive updates and iterations: During the production process, the finished fiber is regularly inspected to obtain actual product quality indicators. This data is used as a new sample and fed back into the prediction model of S2 to update model f online, so as to compensate for the impact of model drift, equipment aging and other factors, and realize the model's self-adaptation and continuous evolution.

[0021] This invention completely eliminates reliance on manual experience. By constructing a data-driven predictive model and utilizing intelligent algorithms for optimization, it can calculate the theoretically optimal annealing curve under specific working conditions, achieving precise temperature control. Furthermore, it can sense changes in working conditions such as drawing speed and fiber diameter in real time, dynamically adjusting the annealing temperature curve when conditions fluctuate, ensuring the uniformity and stability of product quality under different conditions. Simultaneously, through the synergistic optimization of multiple objectives such as residual stress and mechanical properties, it can significantly reduce the residual stress of high-silica glass fibers, improve fiber strength and flexibility, thereby increasing yield and reducing energy consumption. It possesses adaptive model updating capabilities, and through a closed loop of production-detection-feedback-learning, the system can adapt to long-term changes in working conditions, achieving the accumulation and iterative optimization of process knowledge.

[0022] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value.

[0023] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation. The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent optimization of the annealing temperature profile of high-silica glass fiber, characterized in that, Includes the following steps: S1. Real-time acquisition of multi-dimensional data and database construction of historical data: Acquire and integrate multi-source data in the production process of high silica glass fiber, including real-time process parameters, historical process parameters and product quality indicators; S2. Constructing an annealing process-fiber quality prediction model f: Based on the historical process parameters and corresponding product quality indicators obtained in S1, a quality prediction model is trained and constructed using machine learning algorithms. The quality prediction model is used to characterize the mapping relationship between key parameters of the annealing temperature curve and real-time process parameters to the final product quality indicators. S3. Set the multi-objective optimization function J: Set the multi-objective optimization function according to product requirements and production goals; S4. Intelligent optimization of annealing temperature curve: The real-time process parameters collected in S1 are used as the input of the current working condition. The intelligent optimization algorithm is used with the optimization function J set in S3 as the optimization target and the prediction model f constructed in S2 as the fitness function. The global search is performed in the feasible region of the key parameters of the annealing temperature curve to obtain the optimal key parameters of the annealing temperature curve under the current working condition. S5. Optimal Curve Analysis and Closed-Loop Execution: The key parameters of the optimal annealing temperature curve obtained in S4 are analyzed into a sequence of specific temperature setpoints for the internal zones of the annealing furnace. This setpoint sequence is then sent to the underlying control system of the annealing furnace. The actual temperature of each zone is monitored in real time by a high-precision temperature sensor. A PID control algorithm is used for high-speed closed-loop adjustment to ensure that the actual temperature curve accurately tracks the optimal setpoint curve.

2. The intelligent optimization method for the annealing temperature curve of high-silica glass fiber according to claim 1, characterized in that, The real-time process parameters specifically include drawing speed v, raw wire diameter d, and actual temperature T of each temperature zone in the annealing furnace; The product quality indicators are the finished fiber mechanical properties of standard products within the historical production cycle, specifically including tensile strength, elongation at break and residual stress level. The key parameters of the annealing temperature curve include the heating rate, peak holding temperature, holding time, and cooling rate.

3. The intelligent optimization method for the annealing temperature curve of high-silica glass fiber according to claim 1, characterized in that, The specific process of constructing a quality prediction model is as follows: S201. Obtain real-time process parameters and key parameters of the annealing temperature curve, and integrate them into input features, the vector representation of which is as follows: Where v is the drawing speed, d is the diameter of the raw wire, and R is the wire diameter. UP T represents the heating rate, Tpeak represents the peak holding temperature, and t represents the heating rate. 所以 For effective heat preservation time, R 首席运营官 Cooling rate; S202. Determine the output labels of the model as follows: , where σ is the tensile strength, S is the residual stress level, and ε is the elongation at break; S203. Obtain historical process parameters. The historical process parameters are a set of process parameters for standard products within a historical production cycle. After cleaning and normalizing the historical process parameters, training samples are obtained. The training samples are divided into training set, validation set and test set in a ratio of 8:1:

1. S204. Construct a feedforward neural network, including: Input layer: The number of nodes is the same as the dimension of the input features; Hidden layers: Set 2 to 5 fully connected hidden layers; The selection of the number of hidden layer nodes and the number of layers is optimized by tuning the performance on the validation set; Activation function: The hidden layer preferably uses the ReLU activation function to increase nonlinear expressiveness and accelerate convergence; Output layer: The number of nodes is the same as the dimension of the output labels; S205. During the training process, small batches of samples are randomly and non-repeatedly drawn from the training set for training. The training cycle is completed after all training samples are drawn. The training is completed after a certain number of cycles, and the quality prediction model is obtained.

4. The intelligent optimization method for the annealing temperature curve of high-silica glass fiber according to claim 1, characterized in that, The specific process for setting the multi-objective optimization function is as follows: S301, The multi-objective optimization function is expressed as follows: , where X represents the key parameters of the annealing temperature curve that S4 is searching for, ω1, ω2, ω3 are preset weight coefficients, and ω1 + ω2 + ω3 = 1, f1, f2, f3 are the normalized cost functions of tensile strength, residual stress level and elongation at break, respectively. S302, respectively, limit f1, f2, and f3: If we set an ideal target intensity σtarget, then the cost function f1 is the squared deviation from the target value: , where σPerd=fσ(X), which is the predicted output value of the mass prediction model for tensile strength; Set the minimum resolution S 分辨率 The resolution prediction value S is output through the quality prediction model. res-p回复d The objective function is constructed as follows: ； To minimize the process energy consumption E, the objective function is constructed as follows: T peak t represents the peak insulation temperature, t represents the effective insulation time, and c1 and c2 are coefficients calibrated according to thermodynamics.

5. The intelligent optimization method for the annealing temperature curve of high-silica glass fiber according to claim 1, characterized in that, The specific process of analyzing the optimal curve is as follows: S401. Obtain the physical model and real-time operating parameters of the annealing furnace during the high-silica glass fiber annealing process. The physical model of the annealing furnace includes the number of temperature zones n and the physical degree L of each temperature zone. 我 The furnace coordinates x at the center point of each temperature zone 我 The relationship between the position x of high silica glass fiber in the furnace and time t was constructed: t 我 =x 我 / v, where v is the real-time wire drawing speed; S402. Based on the key parameters of the optimal annealing temperature curve output by S4, construct the ideal time and temperature function Targte(t). S403, Obtain the time point t for each temperature zone from S401. 我 Substituting into Targte(t), we obtain the target temperature of the high-silica glass fiber in the i-th temperature zone: 。 6. The intelligent optimization method for the annealing temperature curve of high-silica glass fiber according to claim 1, characterized in that, It also includes model adaptive updates and iterations: during the production process, the finished fiber is regularly inspected to obtain actual product quality indicators. This data is then used as a new sample and fed back into the prediction model of S2 to update model f online.

Images