CN-121997713-A - Intelligent optimization method for annealing temperature curve of high silica glass fiber

CN121997713ACN 121997713 ACN121997713 ACN 121997713ACN-121997713-A

Abstract

The invention discloses an intelligent optimization method for an annealing temperature curve of a high silica glass fiber, which relates to the technical field of processing of the high silica glass fiber, thoroughly gets rid of dependence on manual experience, can calculate a theoretical optimal annealing curve under a specific working condition by constructing a data-driven prediction model and optimizing by utilizing an intelligent algorithm, realizes accurate control of temperature, can sense working condition changes such as wire drawing speed, fiber diameter and the like in real time, dynamically adjusts the annealing temperature curve when the working condition fluctuates, ensures uniformity and stability of product quality under different working conditions, can obviously reduce the residual stress of the high silica glass fiber through cooperative optimization of multiple targets such as residual stress, mechanical property and the like, improves fiber strength and flexibility, thereby improving yield and reducing energy consumption, has model self-adaptive updating capability, and can adapt to long-term working condition changes through closed loop production-detection-feedback-learning so as to realize accumulation and iterative optimization of process knowledge.

Inventors

YOU FUCAI
ZHANG DONG
LI YUCAI
SONG SHIWEI
YU SHICHUN
DING BO
WANG HAN
WANG JIAN
ZHAO DEPENG

Assignees

沈阳工程学院

Dates

Publication Date: 20260508
Application Date: 20251226

Claims (6)

1. The intelligent optimization method for the annealing temperature curve of the high silica glass fiber is characterized by comprising the following steps of: S1, acquiring multi-dimensional data in real time and establishing a historical data database, wherein the multi-source data in the production process of the high silica glass fiber is acquired and integrated, and the multi-source data comprises real-time technological parameters, historical technological parameters and product quality indexes; S2, constructing an annealing process-fiber quality prediction model f, namely training and constructing a quality prediction model by utilizing a machine learning algorithm based on the historical process parameters and the corresponding product quality indexes obtained in the S1, wherein the quality prediction model is used for representing the mapping relation between the key parameters of an annealing temperature curve and the real-time process parameters to the quality indexes of the final product; s3, setting a multi-objective optimization function J, namely setting the multi-objective optimization function according to the product demand and the production objective; S4, intelligent optimizing of an annealing temperature curve, namely inputting the real-time process parameters acquired in the S1 as current working conditions, adopting an intelligent optimizing algorithm, taking the optimizing function J set in the S3 as an optimizing target, taking the prediction model f constructed in the S2 as an fitness function, carrying out global searching in a feasible region of the key parameters of the annealing temperature curve, and solving to obtain the key parameters of the optimal annealing temperature curve under the current working conditions; S5, analyzing and executing an optimal curve, namely analyzing the key parameters of the optimal annealing temperature curve obtained in the S4 into a specific temperature set value sequence of the internal partition of the annealing furnace, transmitting the set value sequence to a bottom layer control system of the annealing furnace, monitoring the actual temperature of each temperature region in real time through a high-precision temperature sensor, and performing high-speed closed-loop adjustment by utilizing a PID control algorithm to ensure that the actual temperature curve accurately tracks the optimal set curve.
2. The intelligent optimization method for the annealing temperature curve of the high silica glass fiber according to claim 1, wherein the real-time process parameters comprise wire drawing speed v, filament diameter d and actual temperature T of each temperature zone of an annealing furnace; The product quality index is the mechanical property of the finished fiber corresponding to the standard product in the historical production period, and specifically comprises tensile strength, elongation at break and residual stress level; The key parameters of the annealing temperature curve comprise a heating rate, a heat preservation peak temperature, a heat preservation time and a cooling rate.
3. The intelligent optimization method for the annealing temperature curve of the high silica glass fiber according to claim 1, wherein the specific process for constructing the quality prediction model is as follows: S201, acquiring real-time process parameters and key parameters of an annealing temperature curve, integrating the parameters into input features, wherein vectors of the parameters are expressed as follows: Wherein v is wire drawing speed, d is wire diameter, R UP is heating rate, tpak is heat preservation peak temperature, t So that is effective heat preservation time, and R Chief operators is cooling rate; S202, determining an output label of a model as follows: Wherein sigma is tensile strength, S resolution is residual stress level, epsilon is elongation at break; S203, acquiring historical process parameters, wherein the historical process parameters are process parameter sets of standard products in a historical production period, performing data cleaning on the historical process parameters, performing normalization processing on the historical process parameters to obtain training samples, and dividing the training samples into training sets, verification sets and test sets according to the ratio of 8:1:1; S204, constructing a feedforward neural network, which comprises the following steps: The node number is the same as the dimension of the input feature; Setting 2 to 5 fully connected hidden layers; Wherein the number of hidden layer nodes and the number of layers are selected and optimized through the performance on the verification set; Activating function, wherein the hidden layer preferably adopts a ReLU activating function to increase nonlinear expression capability and accelerate convergence; The node number is the same as the dimension of the output label; And S205, randomly and repeatedly extracting small batches of samples from the training set to train in the training process, wherein all the extracted training samples are used as a training period, and the training is completed by iterating to a certain period to obtain a quality prediction model.
4. The intelligent optimization method for the annealing temperature curve of the high silica glass fiber according to claim 1, wherein the specific process of setting the multi-objective optimization function is as follows: S301, a multi-objective optimization function is expressed as follows: Wherein X represents an annealing temperature curve key parameter being searched for by S4, ω1, ω2, ω3 are preset weight coefficients, ω1+ω2+ω3=1, and f1, f2, f3 are normalized cost functions of tensile strength, resolution being residual stress level and elongation at break, respectively; s302, defining f1, f2 and f3 respectively: setting an ideal target intensity sigma target, wherein the cost function f1 is the square of deviation from a target value: wherein σ Perd =fσ (X), i.e. the predicted output value of the mass prediction model for tensile strength; Setting a minimum resolution S resolution ratio , outputting a resolution predicted value S res-p Reply to d through a quality prediction model, and constructing an objective function as follows: ; setting the energy consumption E of the minimum process, and constructing an objective function as follows: t peak is the soak peak temperature, T is therefore the effective soak time, and c1 and c2 are coefficients calibrated according to thermal engineering.
5. The intelligent optimization method for the annealing temperature curve of the high silica glass fiber according to claim 1, wherein the specific process of analyzing the optimal curve is as follows: S401, acquiring a physical model and real-time working condition parameters of an annealing furnace in the annealing process of the high silica glass fiber, wherein the physical model of the annealing furnace comprises the number n of temperature areas, the physical degree L I am of each temperature area, the in-furnace coordinate x I am of the center point of each temperature area, and constructing the relation between the in-furnace position x and the time t of the high silica glass fiber, wherein v is the real-time wire drawing speed; s402, constructing an ideal time and temperature function Targte (t) according to the key parameters of the optimal annealing temperature curve output by the S4; s403, substituting the time point t I am of each temperature zone obtained in S401 into Targte (t) to obtain the target temperature of the high silica glass fiber in the first temperature zone: 。
6. The intelligent optimization method for the high silica glass fiber annealing temperature curve according to claim 1 is characterized by further comprising model self-adaptive updating and iteration, wherein in the production process, quality detection is carried out on finished fibers regularly to obtain actual product quality indexes, the data pair is used as a new sample, and the new sample is fed back into a prediction model of S2 to update a model f on line.

Description

Intelligent optimization method for annealing temperature curve of high silica glass fiber Technical Field The invention relates to the technical field of high silica glass fiber processing, in particular to an intelligent optimization method for an annealing temperature curve of a high silica glass fiber. Background The high silica glass fiber is widely applied to the tip fields of aerospace, high-temperature heat insulation, special composite materials and the like due to the excellent high temperature resistance, dielectric property and chemical stability. However, in the process of wiredrawing and forming, high residual stress is generated in the high silica glass fiber due to rapid temperature change, so that the fiber is brittle, the strength is reduced, and the flexibility is poor. The annealing treatment is a key process for eliminating or reducing the residual stress of the high silica glass fiber and improving the mechanical property of the high silica glass fiber. The core of the annealing process is an annealing temperature curve, namely the temperature change law of time or position of the temperature of heating-preserving-cooling' of the fiber in the annealing furnace. In the prior art, the setting of the annealing temperature curve mainly depends on the following ways: The fixing process method adopts one or more fixed temperature curves which are verified by long-term production, and the method cannot adapt to dynamic changes in the production process, such as fluctuation of wire drawing speed, uneven diameter of fibrils, tiny difference of raw material components and the like; Manual experience adjustment, which is to manually adjust the set temperature of each temperature zone of the annealing furnace according to the detection result of the final product by relying on the experience of a senior engineer, and the mode has lag adjustment, poor precision and strong subjectivity, and is difficult to realize optimal control; The prior art has a core defect that the annealing temperature curve cannot be dynamically and accurately adjusted and optimized according to the real-time change of the production working condition, so that the fluctuation of the product quality of the high silica glass fiber is large, the yield is difficult to improve, and unnecessary energy waste is possibly caused; In view of the technical drawbacks described above, solutions are now proposed. Disclosure of Invention The invention aims to construct a data-driven prediction model, optimize by utilizing an intelligent algorithm, calculate a theoretical optimal annealing curve under a specific working condition, realize accurate control of temperature, sense working condition changes such as wire drawing speed, fiber diameter and the like in real time, dynamically adjust an annealing temperature curve when the working conditions fluctuate, ensure uniformity and stability of product quality under different working conditions, obviously reduce the residual stress of high silica glass fiber, improve fiber strength and flexibility through cooperative optimization of multiple targets such as residual stress, mechanical property and the like, improve yield and reduce energy consumption, and have model self-adaptive updating capability, and enable a system to adapt to long-term working condition changes through closed loop of production-detection-feedback-learning, thereby realizing accumulation of process knowledge and iterative optimization. In order to achieve the purpose, the invention adopts the following technical scheme that the intelligent optimization method for the annealing temperature curve of the high silica glass fiber comprises the following steps: S1, acquiring multi-dimensional data in real time and establishing a historical data database, wherein the multi-source data in the production process of the high silica glass fiber is acquired and integrated, and the multi-source data comprises real-time technological parameters, historical technological parameters and product quality indexes; S2, constructing an annealing process-fiber quality prediction model f, namely training and constructing a quality prediction model by utilizing a machine learning algorithm based on the historical process parameters and the corresponding product quality indexes obtained in the S1, wherein the quality prediction model is used for representing the mapping relation between the key parameters of an annealing temperature curve and the real-time process parameters to the quality indexes of the final product; s3, setting a multi-objective optimization function J, namely setting the multi-objective optimization function according to the product demand and the production objective; S4, intelligent optimizing of an annealing temperature curve, namely inputting the real-time process parameters acquired in the S1 as current working conditions, adopting an intelligent optimizing algorithm, taking the optimizing function J set in the S3 as an