Continuous basalt fiber pool furnace drawing production process and system based on predictive control

Abstract

The application discloses a continuous basalt fiber tank furnace drawing production process and system based on predictive control, which synchronously acquires temperature and viscosity double feedback signals through a sensor immersed in a melt; then the signals are input into a melt rheological digital twin model and a fiber forming process model which are connected in parallel and have a dynamic self-correction function of FeO content, so that a diameter evolution trend is predicted; finally, based on the predicted trend, a model predictive control algorithm is adopted to cooperatively adjust the temperature of a bushing and the drawing speed before the actual diameter deviates, and the corresponding system comprises a sensing unit, a prediction and control unit and an execution unit; the application solves the problems of high breakage rate and uneven product caused by control lag and rigid model in the prior art, and realizes stable, self-adaptive and low-energy-consumption operation of the production process.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, specifically to a continuous basalt fiber pool kiln drawing production process and system based on predictive control. Background Technology

[0002] Continuous basalt fiber has broad application prospects in composite materials, high-temperature filtration, and other fields due to its excellent mechanical properties, high-temperature resistance, and corrosion resistance. The tank furnace drawing method is the mainstream direction for achieving its large-scale, low-cost production. Its core is to melt and homogenize basalt raw materials in a tank furnace and then directly draw them into continuous fibers through a platinum-rhodium alloy spinneret. However, this process is complex and requires high stability. Existing technologies still face a series of long-standing technical bottlenecks in actual industrial production, which restrict the further improvement of fiber quality and the further reduction of production energy consumption.

[0003] At the process control level, existing technologies suffer from severe time lag in regulating key process parameters. From changes in composition and temperature fluctuations in basalt melt to its ultimate impact on fiber diameter, there is an inherent physical transmission and response delay. Currently, most control systems rely on direct measurement of fiber diameter or melt viscosity for feedback regulation. For example, they can predict trends and intervene by monitoring historical melt viscosity data sequences. However, this method is essentially a lag compensation. The response time of its data acquisition, processing, and execution chain often exceeds the effective window period for changes in melt characteristics. This results in the process state having already exceeded the critical point when control commands are issued, causing fiber breakage or diameter exceeding the standard. Production stability is difficult to guarantee, and the post-event remedial mode cannot achieve truly stable fiber drawing. At the model prediction level, existing technologies lack the ability to adapt to the inherent fluctuations of basalt raw materials. As a natural ore, basalt has significant fluctuations in the content of key components such as ferrous oxide, which directly affects the viscosity-temperature characteristics of the melt. Most of the mathematical models currently used for process simulation or control adopt viscosity-temperature relationships with fixed parameters. When the raw material composition deviates from the model calibration benchmark, the prediction accuracy of the model will drop sharply, which weakens the effect of all advanced control strategies based on this model, such as predictive control. This makes the control system vulnerable to fluctuations in major raw materials, and the so-called robustness of intelligent control is insufficient. At the level of control strategies, existing methods are mostly single or decoupled parameter adjustments, such as adjusting only the spinneret temperature or only the drawing speed after detecting diameter deviation. However, fiber forming is a transient result of the coupling effect of melt thermodynamic state and external mechanical stretching. Adjusting a single parameter is often inefficient and may cause new disturbances. The lack of a control mechanism based on a deep understanding of the forming process that can synergistically drive multiple actuators is another key reason why existing technologies cannot quickly and accurately suppress diameter fluctuations and achieve stable drawing of high-quality fibers. In summary, existing continuous basalt fiber pool kiln drawing technology has inherent defects in terms of control timeliness, model adaptability, and strategy synergy. Therefore, there is an urgent need for a new production process and system that can accurately sense the melt state in real time, adaptively compensate for fluctuations in raw material composition, and conduct forward-looking synergistic regulation based on this, so as to fundamentally break through the bottlenecks of production stability and product uniformity and achieve low-energy consumption and high-quality continuous production. Summary of the Invention

[0004] This invention provides a continuous basalt fiber pool kiln drawing production process and system based on predictive control, overcoming the above-mentioned defects; To achieve the above objectives, the present invention adopts the following technical solution: Firstly, a continuous basalt fiber pool kiln drawing process based on predictive control is provided, comprising the following steps: S1, obtain the real-time temperature and real-time apparent viscosity of the basalt melt as dual feedback signals; The real-time apparent viscosity is obtained by inverting the vibration response signal of the sensor element immersed in the melt. S2, synchronously inputs dual feedback signals into the parallel melt rheological digital twin model and fiber forming process model to predict the evolution trend of fiber diameter; Among them, the melt rheological digital twin model is a CFD model based on the dynamic self-correction of FeO content in basalt melt; S3, based on the predicted diameter evolution trend, uses a model predictive control algorithm to coordinate the spinneret temperature and the fiber drawing speed before the fiber diameter deviates from its actual value.

[0005] Preferably, in step S1, the vibration response signal is the damping attenuation spectrum of the sensor element at the excitation frequency; The inversion is accomplished by matching the damping attenuation spectrum with a pre-stored viscosity-damping spectrum database.

[0006] Preferably, in step S1, the real-time temperature is obtained by the sensor element by detecting the thermal radiation spectral shift of the melt.

[0007] Preferably, in step S1, the sensor element is a sapphire optical fiber.

[0008] Preferably, in step S2, the melt rheological digital twin model dynamically corrects the viscosity-temperature characteristic equation of the melt based on the real-time temperature through the embedded FeO activity sub-model.

[0009] Preferably, in step S2, the prediction time domain of the melt rheological digital twin model is 30 to 120 seconds.

[0010] Preferably, in step S3, the coordinated adjustment specifically includes: When the predicted diameter is smaller, increase the spindle temperature and decrease the drawing speed; When the predicted diameter tends to be larger, reduce the spindle temperature and increase the drawing speed.

[0011] Preferably, in step S3, the control cycle of the model predictive control algorithm is 5 to 10 seconds.

[0012] Preferably, in step S3, the magnitude of the coordinated adjustment is dynamically proportionally allocated based on the predicted diameter deviation rate.

[0013] Secondly, a continuous basalt fiber pool kiln drawing production system for implementing the above-mentioned process is provided, comprising: The sensing unit includes sensor elements immersed in the melt in the furnace passage to acquire real-time temperature and real-time apparent viscosity. The prediction and control unit runs the melt rheology digital twin model, the fiber forming process model, and the model predictive control algorithm, and generates control commands based on sensor data. The execution unit synchronously adjusts the operating parameters of the stencil heater and the wire drawing machine according to the control instructions.

[0014] Beneficial effects Compared with the prior art, the present invention has the following beneficial effects: To improve control timeliness and production stability, sapphire fiber optic sensors are used to synchronously and in-situ acquire melt temperature and apparent viscosity. Combined with a digital twin model with a prediction time domain of 30-120 seconds, the evolution trend of fiber diameter can be predicted in advance. Based on this prediction, the model predictive control algorithm adjusts the spinneret temperature and drawing speed in a coordinated manner before the diameter deviates from the actual value. This changes the control action from lagging remedy to proactive intervention, reduces the fiber breakage rate by orders of magnitude, and improves the continuity and stability of production. To enhance the system's adaptability to raw material fluctuations, an FeO activity sub-model is embedded in the melt rheological digital twin model. This sub-model dynamically corrects the viscosity-temperature characteristic equation based on real-time temperature, enabling the core process model to respond in real time to changes in FeO content in basalt raw materials. This overcomes the defect of inaccurate predictions in fixed parameter models when composition fluctuates, improves the robustness of the control system under raw material fluctuation conditions, and ensures the long-term uniformity of fiber diameter. Achieving precise and energy-saving coordinated control, through the coordinated adjustment rules of temperature and speed coupling, and the strategy of dynamically allocating the adjustment range according to the predicted deviation rate, the forming process can be refined and intelligently controlled. This coordinated control avoids the limitations and oscillations of single parameter adjustment, and can quickly correct deviations with minimal and matched intervention. While improving product accuracy, it reduces energy waste caused by over-adjustment or frequent start-stop, achieving a balance between high quality and low energy consumption. Attached Figure Description

[0015] Figure 1 A flowchart of a continuous basalt fiber pool kiln drawing production process based on predictive control provided in Embodiment 1 of the present invention; Figure 2 The following is a timing diagram of the model predictive control algorithm for a continuous basalt fiber pool kiln drawing production process based on predictive control, provided in Embodiment 1 of the present invention. Detailed Implementation

[0016] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, exemplary embodiments will be described in detail below. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of methods and systems consistent with some aspects of this application as detailed in the appended claims.

[0017] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0018] To better illustrate the purpose, technical solutions, and advantages of this application, the following description, in conjunction with specific embodiments and comparative examples, aims to provide a detailed understanding of the content of this application, rather than limiting it. All other embodiments obtained by those skilled in the art without inventive effort are within the protection scope of this application. Unless otherwise specified, the experimental reagents and instruments involved in the implementation of this application are commonly used reagents and instruments. In this application, the technical features described in an open-ended manner include both closed-ended technical solutions composed of the listed features and open-ended technical solutions that include the listed features.

[0019] The following are preferred embodiments, which provide a detailed description of the specific implementation methods, features, and effects of the present invention.

[0020] Example 1 Please refer to Figure 1 , 2 This embodiment provides a continuous basalt fiber pool kiln drawing production process based on predictive control, including the following steps: S1, obtain the real-time temperature and real-time apparent viscosity of the basalt melt as dual feedback signals; In the homogenization or operation path of the continuous basalt fiber pool kiln drawing production line, a high-temperature resistant sensor element is immersed, preferably sapphire optical fiber. This choice is based on the harsh operating conditions of the process and the measurement requirements. Sapphire optical fiber has excellent high temperature resistance and melt corrosion resistance. This sapphire optical fiber sensor undertakes the dual function of simultaneously acquiring the real-time temperature and real-time apparent viscosity of the melt. The temperature of basalt melt in the furnace passage is usually above 1,300 degrees Celsius, and the melt itself is highly chemically corrosive. Ordinary metal or quartz sensors will be quickly damaged or experience performance drift in this environment. Sapphire single crystals are irreplaceable materials due to their extremely high melting point, excellent high-temperature structural stability, outstanding thermal shock resistance, and excellent chemical corrosion resistance. The optical fibers drawn from sapphire inherit the advantages of these bulk materials and also have flexible functions of light guiding and sensing. In practice, the selected sapphire optical fiber is a single crystal pure alumina structure with a diameter between 0.3 and 1 mm to ensure sufficient mechanical strength. One end of the optical fiber is processed and encapsulated into a sensitive probe that can be directly immersed in the melt, while the other end is connected to a signal demodulation and processing device located outside the kiln. Sapphire fiber optic sensors bear the important responsibility of simultaneously measuring both temperature and viscosity. The realization of this dual function depends on the coordinated interpretation of optical signals and mechanical vibration signals transmitted through the same sapphire fiber. When integrating a sapphire fiber optic sensor into a tank furnace drawing system, it must be installed on the side wall or top of the tank furnace passage using a dedicated high-temperature sealing structure to ensure that its sensitive end is immersed in the melt to a certain depth while maintaining the furnace's airtightness. As the core component of the sensing unit, the sensor's output raw optical and vibration signals are transmitted to the prediction and control unit for processing. During continuous production, sapphire fiber optic sensors continuously provide real-time temperature and viscosity data streams. These data are the fundamental basis for the subsequent dynamic self-calibration of the melt rheology digital twin model and the collaborative adjustment decisions made by the model predictive control algorithm. The long-term stability of the measurement is directly related to the durability and reliability of the entire predictive control system performance. For temperature measurement, sapphire fiber itself is used as a light waveguide for radiation thermometry. The probe end immersed in the melt directly receives the thermal radiation light of the melt. This light is transmitted along the sapphire fiber to the back-end spectral analysis module. Since the peak wavelength of the spectrum will shift as the temperature rises, the real-time temperature value of the melt can be calculated by accurately measuring the amount of spectral shift. This temperature measurement process is continuous, and the data update frequency can reach several times per second. For obtaining real-time apparent viscosity, this invention employs an indirect inversion method based on fluid dynamics, which applies a specific frequency of micromechanical excitation to a sapphire fiber optic sensor immersed in the melt, such as by using a micro piezoelectric ceramic transducer attached to the fiber optic to generate micro-amplitude vibrations. Basalt melt will dampen the vibration of the sensor. By monitoring the damping attenuation spectrum of the sensor vibration, that is, the curve characteristics of the vibration amplitude attenuation with time or period, the viscous resistance of the melt can be reflected. By matching and comparing the vibration damping attenuation spectrum acquired in real time with the viscosity damping spectrum database established in advance through a large number of experiments, the real-time apparent viscosity value of the current melt can be obtained. Temperature and viscosity signals together constitute the dual feedback signals for process control. The fiber segment immersed in the melt vibrates accordingly, and its vibration damping characteristics are directly affected by the viscosity of the melt. This vibration signal is also transmitted to the external photoelectric detection system through the sapphire fiber itself, and is converted into a vibration damping attenuation spectrum, from which the viscosity is inverted.

[0021] S2, after obtaining the real-time temperature and real-time apparent viscosity dual feedback signals, synchronously inputs them into two parallel-running digital models; The first model is a melt rheological digital twin model, which is a simulation model built on the principles of computational fluid dynamics and specially customized. The model incorporates an FeO activity sub-model, because the FeO content in basalt raw materials will passively fluctuate within a certain range, and the activity of FeO significantly affects the viscosity-temperature characteristics of the melt. Based on the real-time collected temperature signal, the sub-model dynamically corrects the relevant parameters in the viscosity-temperature characteristic equation, enabling the entire digital twin model to simulate the flow and heat transfer behavior of the melt in the passage and the baffle plate under the current specific composition, and predict the uniformity of the melt temperature distribution and the flow stability at the baffle plate outlet in the future, such as the next 30 to 120 seconds. The second model is the fiber forming process model. The fiber forming process model receives the prediction results provided by the melt rheological digital twin model and combines them with the physical laws of the drawing process to simulate the complete process of the melt flowing out of the spinneret, being drawn thin by the stretching force, and cooling and solidifying into fibers in the air. Through simulation calculations of this model, the evolution trend of fiber diameter in the future can be predicted in advance, such as predicting whether the diameter will develop in the direction of becoming thinner or thicker and the rate of change.

[0022] S3, based on the predicted diameter evolution trend, uses a model predictive control algorithm to coordinate the spinneret temperature and the fiber drawing speed before the fiber diameter deviates from its actual value. Based on the diameter evolution trend predicted by the fiber forming process model, the control system initiates adjustment actions when it predicts that the fiber diameter is about to deviate but has not yet deviated. This invention adopts model predictive control algorithm as the core control strategy. The algorithm aims to stabilize the fiber diameter and uses the fiber forming process model in step S2 as an internal prediction model for rolling optimization calculation. The specific control cycle of the algorithm can be set to 5 to 10 seconds. In each control cycle, the algorithm will comprehensively consider the predicted future diameter deviation and calculate the optimal adjustment command. This instruction coordinates the stencil temperature and the drawing speed. The specific coordination rules are as follows: When the model predicts that the fiber diameter will tend to become thinner, the control algorithm will generate a combination of instructions to simultaneously increase the spinneret temperature and decrease the drawing speed. Conversely, when the predicted diameter tends to increase, a combination of instructions is generated that simultaneously reduces the spindle temperature and increases the drawing speed. This coordination of temperature and speed, with one rise and one fall or one fall and one rise, can more quickly and smoothly counteract disturbances. Furthermore, the adjustment range is not fixed, but dynamically proportionally allocated based on the predicted diameter deviation rate. A faster deviation rate results in a larger adjustment range, while a slower deviation rate results in a smaller adjustment range, thus achieving fine-grained control.

[0023] Specifically, by innovatively acquiring melt temperature and viscosity signals synchronously and in real time using sapphire fiber optic sensors, a digital twin model of melt rheology and a fiber forming process model with dynamic correction function for FeO activity are constructed. Finally, model predictive control algorithms are used to coordinately adjust the spinneret temperature and drawing speed. The entire process realizes closed-loop advanced control from sensing, prediction to execution, effectively solving the problem of melt rheological property changes caused by fluctuations in basalt raw material composition. It can accurately intervene before the fiber diameter deviates from its actual value, improving the stability of the drawing process and the uniformity of fiber product quality.

[0024] Further, in step S1, the vibration response signal is the damping attenuation spectrum of the sensor element at a specific excitation frequency; In step S1, in order to achieve in-situ inversion of the apparent viscosity of basalt melt, it is necessary to obtain the vibration damping attenuation spectrum of the sensor element in the melt. In practice, sapphire optical fiber is used as the sensor element. A micro piezoelectric ceramic transducer tightly attached to the sapphire optical fiber is used to apply precise and controllable micro-axial excitation to the optical fiber. The excitation is a short-duration, single, specific frequency signal, such as an audio pulse with a center frequency of 5000 Hz. After the pulse ends, the sapphire fiber segment immersed in the melt will generate free decay vibration under the combined action of its natural frequency and the viscous resistance of the melt. This damped vibration is transmitted through the sapphire optical fiber itself to an external photoelectric detection system, where it is recorded in real time and converted into a curve showing the decay of vibration amplitude over time or the number of vibration cycles, i.e., the damped decay spectrum. For example, the time required for the amplitude to decay to a certain proportion of the initial value and the shape of the decay curve directly reflect the melt's ability to dissipate vibrational energy, and this dissipation ability is closely related to the apparent viscosity of the melt. Converting damping attenuation spectra into viscosity values ​​requires a pre-established, highly targeted viscosity-damping spectrum database. The construction of this database is not a simple signal collection, but rather based on a deep physicochemical understanding of the complex rheological properties of basalt melts, especially the nonlinear influence of its FeO content on the melt structure. First, the database's basic data comes from extensive laboratory calibration. In a high-temperature experimental furnace simulating the working conditions of a pool kiln, a series of basalt melt samples with different known formulations and FeO contents were prepared, and their precise apparent viscosity at different temperatures was simultaneously measured using a high-precision rotational viscometer. Meanwhile, in the same melt sample, the above-mentioned sapphire fiber sensor and excitation detection system are used to obtain its complete damping attenuation spectrum. Each set of data on "melt composition, FeO content, temperature, precise viscosity, and damping attenuation spectrum" constitutes the basic data points. Second, a mapping model between viscosity and damping spectrum characteristic parameters is established in the database. The content and oxidation state of iron oxides, such as FeO and Fe2O3, will significantly change the network structure and ion mobility of silicate melts, and have a nonlinear effect on their viscosity. This effect also changes the internal friction characteristics of the melt, i.e., its damping ability against vibration. When constructing the mathematical model of the database, the FeO content must be taken as a key independent variable. For example, the database may contain a series of sub-models, each corresponding to a specific FeO content range. For each sub-model, a multivariate nonlinear mapping relationship between temperature, damping spectrum characteristic parameters, such as logarithmic decay rate, half-life and apparent viscosity is established. Third, considering the volatility of industrial raw materials, the database needs to have a sufficiently broad coverage to accommodate the common fluctuation range of FeO content in basalt raw materials, such as 7% to 16%. Through targeted design and construction, this database can accurately capture the nonlinear drift of the damping spectrum caused by changes in FeO content, providing a reliable basis for inversion, which is different from general signal processing methods that do not consider the specific composition of the melt. The inversion is accomplished by matching the damping attenuation spectrum with a pre-stored viscosity-damping spectrum database; In actual production, the inversion is completed in real time through the following steps: The damping attenuation spectrum generated by sapphire optical fiber in the melt is acquired in real time, and the predetermined characteristic parameters of the spectrum are extracted. Combining the melt temperature data measured in real time by the same sensor with the estimated FeO content of the current batch of basalt material provided by the upstream raw material analysis system, a matching query is performed in the pre-stored viscosity-damping spectrum database. The matching process is essentially about finding the data point in the database that is closest to the current operating conditions or performing interpolation calculations. The system inputs the current temperature, estimated FeO content, and damping spectrum characteristic parameters into the database mapping model, and through interpolation calculations, quickly outputs the corresponding melt apparent viscosity value. This process has extremely low computational latency, enabling real-time viscosity inversion several times per second, providing continuous dual feedback signals for step S2; Specifically, a method for inverting the apparent viscosity of basalt melt using damping attenuation spectrum is employed. Vibration sensing is applied to high-temperature melts to construct a dedicated database that deeply correlates melt chemical composition, especially FeO content, with damping characteristics, ensuring the accuracy and reliability of the inversion results.

[0025] Furthermore, in step S1, the real-time temperature is obtained by the sensor element by detecting the thermal radiation spectral shift of the melt; In step S1, in order to achieve accurate, in-situ and non-contact measurement of the temperature of high-temperature basalt melt, this embodiment adopts a temperature measurement method based on thermal radiation spectral shift, the core physical principle of which is Planck's blackbody radiation law. This law states that any object above absolute zero will radiate electromagnetic waves, and the characteristic curve of its radiated energy distribution with wavelength depends solely on the object's own temperature. Specifically, as the object's temperature increases, the peak wavelength of its radiation spectrum will shift towards shorter wavelengths. This shift is called spectral shift or Wien shift. The true temperature of the basalt melt can be directly determined by accurately measuring the peak wavelength or distribution characteristics of the thermal radiation spectrum emitted by the basalt melt. The sensor element is a sapphire optical fiber. In this embodiment, the same sapphire optical fiber is used as a waveguide for transmitting thermal radiation optical signals. Specifically, the end face of the sapphire optical fiber probe immersed in the melt directly receives the infrared and visible light signals radiated by the melt in front. Since the sapphire material has stable physicochemical properties at high temperatures and good light transmittance in the required wavelength band, it can efficiently transmit the thermal radiation light of the melt to the analysis system at the back end. At the rear end of the optical fiber, there is a spectral analysis module consisting of a collimating lens, a grating beam splitting system, and a photodetector array. The melt thermal radiation light transmitted from the sapphire optical fiber is collimated, dispersed by the grating to form a continuous spectrum, and captured by the detector array to obtain a spectral map of the intensity distribution of the current melt thermal radiation with wavelength. After obtaining the spectrum, the system's signal processor will analyze the characteristics of the spectrum in real time, and accurately determine the wavelength corresponding to the peak of the spectral energy distribution curve through the peak-finding algorithm. The processor will then substitute the currently measured peak wavelength into the calculation model based on Planck's law and Wien's displacement law. The model has been rigorously calibrated at known temperature points and can directly convert wavelength shift into corresponding temperature values. This calculation process is completed at high speed in the embedded processor, enabling continuous measurement of the melt temperature in real time. The data update frequency is synchronized with the spectral acquisition frequency, reaching several times per second. The real-time temperature and real-time apparent viscosity signals obtained by the above method constitute a complete dual feedback signal. In practice, the dual feedback signal is synchronously input and executes step S2, that is, it is input into the parallel-running melt rheology digital twin model and fiber forming process model to start the subsequent prediction process. Temperature signals play a dual crucial role here: as a fundamental parameter reflecting the current thermal state of the melt, they directly participate in process monitoring; as a key real-time input boundary condition for the melt rheological digital twin model, they are also a necessary input variable for the dynamic correction of the viscosity-temperature characteristic equation of the FeO activity sub-model in the model, ensuring that the model can adaptively simulate the actual working conditions of the melt. Specifically, based on the strict laws of blackbody radiation physics, by combining the superior light transmission characteristics of sapphire optical fiber with high-resolution spectral analysis technology, accurate, rapid and stable in-situ temperature measurement can be achieved inside high-temperature and highly corrosive basalt melts, avoiding the maintenance difficulties and signal delay problems caused by traditional thermocouple contact temperature measurement.

[0026] Furthermore, in step S2, the melt rheological digital twin model dynamically corrects the viscosity-temperature characteristic equation of the melt based on the real-time temperature through the embedded FeO activity sub-model. In step S2, the melt rheological digital twin model does not use a viscosity-temperature relationship with fixed parameters. Its prediction accuracy is embedded with a dedicated FeO activity sub-model. This sub-model combines the static data of the raw material chemical composition with the real-time thermal state of the melt to dynamically calculate the viscosity-temperature characteristic equation parameters that best reflect the current true rheological properties of the melt. The sub-model receives two crucial input data points. The first is the FeO content of the current production batch of basalt feedstock, expressed as a weight percentage. This data is derived from quantitative analysis performed by X-ray fluorescence spectrometry when the feedstock enters the plant, and its value is within a typical fluctuation range, such as between 7% and 16%. The second input is the real-time temperature of the melt, which is measured and transmitted in real time by a sapphire fiber optic sensor from step S1. Temperature is a key variable that determines the thermodynamic activity and behavior of FeO ions in the melt structure. The dynamic correction process is a multi-step calculation based on physicochemical principles, rather than a simple table lookup or linear interpolation. The dynamic correction process is as follows: The real-time thermodynamic activity of FeO is calculated. The FeO activity sub-model calculates the instantaneous thermodynamic activity of the FeO component under the current thermal state based on the received real-time temperature. This calculation is based on a regular solution model applicable to silicate melts or an empirical correlation calibrated by experiments. The effective effect of FeO on melt viscosity is not directly equal to its content percentage, but rather its effective concentration or chemical potential at the current temperature, i.e., activity. The activity is exponentially related to the reciprocal of the absolute temperature, which means that for the same FeO content, the chemical activity and the ability to destroy the network structure are different in the high-temperature zone of the homogenization section of the pool kiln and the low-temperature zone of the perforated plate. The FeO activity sub-model uses this step to integrate static content and dynamic temperature into a physical quantity that changes in real time with production conditions, namely the FeO thermodynamic activity value. After applying a nonlinear function to correct the viscosity-temperature characteristic equation and obtaining the real-time FeO activity value, the sub-model substitutes it into a preset correction function with a specific mathematical form. The function outputs the adjustment amount of the key parameters of the viscosity-temperature characteristic equation. The viscosity-temperature characteristic equation is used to describe the constitutive relationship of melt viscosity as a function of temperature, and it is often in the form of the Arrhenius equation. The key parameters in this correction function, including the coupling coefficient and the exponent, were determined through the following experimental modeling process: A series of basalt standard samples with FeO content ranging from 7% to 16% and other components in proportion were prepared. The viscosity of each sample at different temperatures was accurately measured using a high-temperature viscometer, and their respective baseline activation energy and viscosity-temperature characteristic equations were obtained by fitting. Meanwhile, thermodynamic software was used to calculate the FeO activity of each sample at different temperatures; finally, nonlinear regression analysis was performed on the corresponding data of activation energy change and FeO activity to determine the specific values ​​of coupling coefficient and exponent. In a specific implementation, the modification is applied to the activation energy term in the Arrhenius equation, and the modified effective activation energy is dynamically determined by the following relationship:

[0027] Among them, the reference activation energy is the reference activation energy assuming that the melt does not contain FeO active components, and its value is determined through basic formulation experiments; The coupling coefficient is a physical parameter characterizing the intensity of the effect of FeO unit activity on activation energy; An exponent is a specific constant greater than 1; The mathematical form with an exponent greater than 1 is crucial to ensure that the effect of FeO activity on viscosity is non-linear, showing that as activity increases, its destructive effect on the silicon-oxygen network structure and the resulting viscosity reduction trend gradually approach saturation. This nonlinear relationship cannot be obtained by simple linear scaling adjustment. It originates from the deep data fitting and physical understanding of the measured viscous flow activation energy of a large number of basalt melts with different FeO contents at high and low temperatures, describing the real behavior of FeO in complex silicate melts. After completing the equation correction and updating the model, and obtaining the effective activation energy through the above calculations, the viscosity-temperature characteristic equation, namely the Arrhenius equation, is updated in real time. Combined with the current real-time temperature, this equation can immediately calculate an apparent viscosity reference value that is closer to the actual state of the melt. The viscosity-temperature characteristic equation, which is dynamically corrected and updated in real time, is then synchronized to the material property library of the melt rheological digital twin model, replacing the old equation from the previous calculation cycle.

[0028] The melt rheological digital twin model integrating the above FeO activity sub-model performs the above correction process once in each control cycle, if synchronized with the model predictive control algorithm cycle. The model uses real-time temperature and viscosity feedback signals as boundary conditions and performs computational fluid dynamics simulation based on the rheological laws described by the latest revised viscosity-temperature characteristic equation. The result is that the digital twin model can simulate in real time the flow, heat transfer and homogenization process of the melt in the passage and the sprue under the current specific raw material composition and real-time temperature field, and predict the rheological uniformity of the melt at the sprue outlet at a high accuracy in the future. This prediction result is the direct input on which the subsequent fiber forming process model performs simulation of the fiber diameter evolution trend.

[0029] Furthermore, in step S2, the prediction time domain of the melt rheological digital twin model is 30 to 120 seconds; In step S2, one of the core tasks of the melt rheological digital twin model is to predict the evolution trend of melt rheological uniformity and fiber diameter over a future period of time. The length of this predicted future period of time is called the prediction time domain. Limiting the prediction time domain to 30 to 120 seconds is a reasonable and optimized range determined after theoretical and practical verification based on the physical process time scale and control requirements of the continuous basalt fiber pool kiln drawing process. The lower limit of this range is set at 30 seconds, mainly considering the inherent physical delay of the process and the necessary response time of the control system. From the detection of the change in the melt state to the transmission of the change through the pathway and its eventual impact on the fiber formation at the outlet of the spinneret, there is a non-negligible transmission and response process. In addition, it takes time for the model to complete a simulation calculation, for the control algorithm to complete the optimization solution and issue instructions, and for the actuators such as heaters and drawing machines to respond. If the prediction time is too short, such as less than 30 seconds, the control system will be equivalent to providing real-time feedback, which will not allow enough lead time for the above series of delays, thus losing the meaning of advance control and failing to effectively cope with fluctuations. The upper limit of this range is set at 120 seconds, mainly based on the balance between prediction accuracy and computing resources. On the one hand, the state of basalt melt is affected by many factors, and the uncertainty of model prediction will increase significantly over time. For long-term forecasts exceeding 120 seconds, the confidence level drops sharply, and control decisions based on low-confidence forecasts may introduce erroneous disturbances. On the other hand, the longer the prediction time domain, the more computational resources required for the digital twin model and model predictive control algorithm increase non-linearly. Setting the upper limit of the time domain to 120 seconds can ensure that the prediction results remain highly reliable for a sufficiently long time within the industrially acceptable cost of computing hardware, thus achieving a balance between foresight and accuracy.

[0030] In practice, the prediction time domain is a parameter that can be set and adjusted within the range of 30 to 120 seconds according to actual production conditions and control objectives; If production stability requirements are extremely high and computing resources are sufficient, the prediction time domain can be set to a longer value, such as 90 seconds or 120 seconds, to obtain a longer warning and adjustment window. In Example 2, namely the FeO activity sub-model verification experiment, in order to comprehensively evaluate the predictive stability of the model over a longer period of time, a typical value of 90 seconds, which is in the middle of the range, was selected for testing. Under certain specific operating conditions, a shorter prediction time domain can also be used. For example, when the raw material composition is very stable and the production disturbance is small, the prediction time domain can be set to 40 seconds or 50 seconds to further reduce the real-time calculation load while ensuring the control effect. For example, when dealing with sudden and rapid disturbances, a shorter prediction time-domain model can be temporarily activated for rapid calculation to prioritize ensuring the timeliness of control. Regardless of the specific value set, it falls within the verified effective range of 30 to 120 seconds. The setting of the prediction time domain directly affects steps S2 and S3. In step S2, the digital twin model will simulate and calculate the melt and fiber state from the current moment to the next 30-120 seconds according to the set prediction time domain length. In step S3, the model predictive control algorithm uses the predicted evolution trend in the time domain as the basis for optimization, and calculates the control commands for multiple future steps in a rolling manner to achieve coordinated adjustment before the fiber diameter actually deviates. Specifically, this range is determined after comprehensive consideration and optimization of multiple factors such as the transmission delay of the pool furnace drawing process, system response time, model prediction confidence, and industrial computing costs, to ensure that the control system has sufficient look-ahead time to cope with disturbances and maintain the necessary prediction accuracy and real-time performance.

[0031] Furthermore, in step S3, the coordinated adjustment specifically involves: when the predicted diameter tends to be smaller, increasing the spinneret temperature and decreasing the drawing speed; when the predicted diameter tends to be larger, decreasing the spinneret temperature and increasing the drawing speed. In step S3, based on the decision output by the model predictive control algorithm, a bidirectional and coupled adjustment is made to the spinneret temperature and the fiber drawing speed. The specific rules are directly determined by the process principle of fiber forming. When the fiber forming process model predicts that the fiber diameter will tend to become thinner in the future, the control algorithm performs two operations simultaneously: increasing the spinneret temperature and reducing the fiber drawing speed. Increasing the temperature of the sprue can reduce the viscosity of the melt at the sprue outlet, making the melt more fluid and stretchable, and resisting the tendency for the diameter to become thinner. At the same time, reducing the drafting speed directly reduces the tensile force acting on the fiber, thus inhibiting the fiber from being further thinned from a mechanical perspective. These two adjustments work simultaneously from both thermodynamic and kinetic dimensions, in the same direction, and together promote the fiber diameter to stabilize back towards the target value. Conversely, when the model predicts that the fiber diameter tends to increase, the control algorithm simultaneously performs a coordinated operation of reducing the spinneret temperature and increasing the drawing speed. Reducing the spinneret temperature will increase the melt viscosity, decrease the fluidity and stretchability, and suppress the increase in diameter caused by excessive melt flow or excessively low viscosity. At the same time, increasing the drawing speed increases the tensile force, which helps to draw the coarser fibers to the standard specifications in a timely manner. This intervention addresses the two fundamental factors of thermodynamic state and external mechanical conditions, thereby efficiently correcting diameter deviations. The specific implementation of the coordinated adjustment rule in the control loop is as follows: The coordinated adjustment rule is encoded as a constraint and guide in the objective function of the model predictive control algorithm. In each control cycle, when the algorithm performs optimization calculations in the rolling time domain, the adjustment amounts of both are used as a joint decision vector for solution. The algorithm first receives the predicted curve of the future evolution trend of fiber diameter from step S2; If the algorithm determines that the predicted curve indicates that the diameter will continue to be below the target lower limit, then in the optimization problem it solves, it will treat increasing the setpoint of the sprue temperature and decreasing the setpoint of the drawing speed as a pair of strongly correlated expected actions, and predict the dynamic effects of these two actions on the diameter based on the process model, and calculate the optimal temperature adjustment amount and speed adjustment amount that can make the diameter return to the target range the fastest and most smoothly. These adjustments are precisely arranged in a time series to form a sequence of control commands that coordinate with each other. The first instruction in the control instruction sequence is sent to the execution unit. The sluice plate temperature controller in the execution unit receives the new temperature setpoint and adjusts the power of the heating element to bring the sluice plate temperature toward the setpoint. At the same time, the frequency converter of the wire drawing machine receives the new speed setpoint and adjusts the motor speed to change the drawing speed. Through a further defined dynamic proportional distribution method, the magnitude of each coordinated adjustment is matched with the predicted severity of the diameter deviation.

[0032] Furthermore, in step S3, the control cycle of the model predictive control algorithm is 5 to 10 seconds; In step S3, the model predictive control algorithm makes decisions and executes according to a fixed, periodic rhythm. The time interval between two adjacent, complete cycles from sensor data acquisition, model prediction, optimization solution to command output is called the control cycle of the algorithm. The control cycle is limited to 5 to 10 seconds, which is an optimization range determined after comprehensive consideration of the dynamic characteristics of the continuous basalt fiber pool kiln drawing process, the system hardware performance and control effect requirements. The lower limit of this range was set at 5 seconds, mainly considering the computation time required for the algorithm to complete one full rolling time-domain optimization, and to avoid unnecessary wear and tear on the actuator caused by too frequent adjustments. The model predictive control algorithm requires multiple forward simulations of the melt rheological digital twin model and the fiber forming process model in each cycle, and solving the constrained optimization problem, which requires a considerable amount of real-time computation. Meanwhile, as physical devices, the stencil heating system and the wire drawing machine have a certain inertia in their response. An excessively short cycle may result in the previous adjustment not being fully effective before the next adjustment is issued, which not only fails to improve control accuracy but may also cause system oscillation. The upper limit of this range is set at 10 seconds, mainly based on the dynamic speed of process disturbances and the timeliness requirements of control response. The state of basalt melt and fiber forming process may change rapidly due to factors such as raw materials and temperature. If the control cycle is too long, for example, more than 10 seconds, the system will have a significant lag in detecting and intervening in the disturbance, which may miss the best adjustment opportunity and make it difficult to achieve the goal of coordinated adjustment in step S3 before the fiber diameter actually deviates. In practice, the control cycle is set as a fixed parameter that is adjustable within the range of 5 to 10 seconds. This cycle is coordinated with the data acquisition frequency of step S1 and the model calculation speed of step S2. In a specific configuration, the control cycle can be set to 8 seconds, and the sensor data acquisition cycle of step S1 can be set to 2 seconds accordingly, so as to ensure that each control cycle can make decisions based on the latest multiple sets of process data. The start of each control cycle is triggered by the system clock or a dedicated timer. After the cycle begins, the real-time melt temperature and real-time apparent viscosity data provided in step S1 are collected first. The latest data, along with the historical data from the previous period, are input into the prediction model in step S2. Based on this data, the digital twin model re-simulates within its prediction time domain, for example, within 90 seconds, to update the prediction of the future fiber diameter evolution trend. The optimizer of the model predictive control algorithm is based on the latest prediction trend and aims to stabilize the fiber diameter. It then re-solves the optimal control sequence for multiple future control cycles. The control command corresponding to the current moment in the optimal sequence, namely the coordinated adjustment amount of the stencil temperature and the wire drawing speed, is output to the execution unit. The execution unit drives the heater and motor to operate according to the command until the next control cycle begins, and repeats the above process. This periodic operation mechanism ensures that control decisions are always based on the latest process status and are continuously optimized and corrected at a pace that matches the process dynamics. This enables the control system to periodically assess future trends and output adjustment commands in a timely manner when it predicts that the diameter is about to deviate but has not yet actually deviated.

[0033] Furthermore, the coordinated adjustment rules are further refined. In step S3, the magnitude of the coordinated adjustment is dynamically proportionally allocated based on the predicted diameter deviation rate. In step S3, when making decisions, the model predictive control algorithm not only relies on whether the future diameter trend predicted by the fiber forming process model is to become thinner or thicker, but also makes extensive use of the diameter deviation rate predicted by the model. The predicted diameter deviation rate, specifically, means the instantaneous rate of change of the fiber diameter from its target value within a specific future period, such as the next control cycle, calculated by the fiber forming process model based on the current and historical process conditions. Its unit is usually micrometers per second. This rate is a vector value; it is negative when the diameter tends to narrow and positive when it tends to widen. The magnitude of this rate directly and quantitatively reflects the severity and urgency of the current process state deviating from the steady state. For example, the model may predict that the diameter is thinning at a rate of 0.5 micrometers per second, where 0.5 micrometers per second is the predicted deviation rate at this moment. Dynamic proportional allocation refers to the adjustment range of the spindle temperature and drawing speed, which no longer uses fixed values, but establishes a dynamic relationship with the absolute value of the predicted diameter deviation rate obtained by the real-time calculation above through a preset proportional function. The faster the deviation, the greater the adjustment force; the slower the deviation, the gentler the adjustment force, giving the control system the ability to adapt to disturbance energy. In practice, this mechanism is encoded as a dynamic weight constraint or cost function term in the model predictive control algorithm optimization problem. In specific implementations, the basic adjustment magnitude can be determined through the following exemplary functional relationship:

[0034] The proportional coefficient is a constant parameter calibrated based on the specific characteristics of the pool kiln drawing production line, such as thermal inertia and mechanical response speed. The minimum action threshold is a very small positive number set to avoid unnecessary frequent actions caused by minor noise. After obtaining the basic adjustment range, it is then dynamically allocated to the specific spindle temperature adjustment and drawing speed adjustment according to the rules of coordinated adjustment. For example, when the model predicts that the diameter is rapidly narrowing, the algorithm calculates a larger basic adjustment range. According to the rules of coordinated adjustment, this range will be dynamically calculated based on a preset internal allocation coefficient that considers the temperature and speed control efficiency ratio, such as temperature accounting for 70% and speed accounting for 30%.

[0035] This allocation ensures that the two aspects of the coordinated adjustment are matched in strength, working together to address the current rate of forecast deviation; The dynamic proportional allocation mechanism, as a calculation module, is embedded in each control cycle and runs for 5 to 10 seconds. In each control cycle, the algorithm completes the following sequential actions: Obtain the latest fiber diameter prediction trajectory containing rate information from step S2; Extract the predicted diameter deviation rate at the current moment; Call the above dynamic proportional allocation function to calculate the basic adjustment range required for this rate in this cycle; By combining the coordinated adjustment rules, the temperature is increased and the speed is decreased when the adjustment is more refined, and the temperature is decreased and the speed is increased when the adjustment is more coarse. The magnitude is then allocated into specific temperature and speed adjustment commands. This instruction is used as the starting point or constraint for optimization. After rolling optimization, the final execution instruction is output, so as to output the control intensity that is precisely matched with the severity of the deviation from the prediction at each control moment. Furthermore, a continuous basalt fiber pool kiln drawing production system for realizing the above process includes: The sensing unit includes sensor elements immersed in the melt in the furnace passage to acquire real-time temperature and real-time apparent viscosity. The sensing unit is the material basis for the system to realize step S1. The components of this unit are sensor elements immersed in the melt in the furnace passage. In specific implementation, the sensor element is made of sapphire optical fiber, which has the characteristics of high temperature resistance and corrosion resistance. This sensing unit performs two simultaneous measurement functions through this sapphire fiber optic sensor to acquire the dual feedback signals required for process control. The first function is to obtain the real-time temperature of the melt, which is accomplished by detecting and calculating the thermal radiation spectral shift of the melt using the sensor. The second function is to obtain the real-time apparent viscosity of the melt, which is accomplished by generating vibrations at the excitation frequency using the sensor and monitoring its damping attenuation spectrum, and then matching and inverting this damping attenuation spectrum with a pre-stored viscosity damping spectrum database. The sensing unit contains the necessary signal exciter, photoelectric conversion module and primary signal processor to ensure that stable and reliable real-time temperature and real-time apparent viscosity digital signals are transmitted to the downstream unit. The prediction and control unit runs the melt rheology digital twin model, the fiber forming process model, and the model predictive control algorithm, and generates control commands based on sensor data. The prediction and control unit is the core computing and decision-making hub for the system to implement steps S2 and S3. This unit is physically composed of an industrial control computer or a high-performance programmable logic controller. The unit has three core software models installed and running. The first is the melt rheological digital twin model, which is a CFD model based on the dynamic self-correction of FeO content in basalt melt. It can dynamically correct the viscosity-temperature equation based on the received real-time temperature signal through the embedded FeO activity sub-model. The second is the fiber forming process model, which is used to receive the output of the digital twin model and predict the evolution trend of fiber diameter; The third is the model predictive control algorithm, which starts at regular intervals of five to ten seconds according to the set control cycle. Within each control cycle, this unit receives dual feedback signals from the sensing unit, calls the aforementioned model for joint calculation, and its workflow is as follows: Real-time data is input into the digital twin model and the fiber forming process model to predict the diameter trend in the prediction time domain for the next 30 to 120 seconds. Based on this prediction result, the model predictive control algorithm follows the collaborative adjustment rule and uses the amplitude strategy of dynamic proportional allocation according to the predicted diameter deviation rate to solve for the optimal control command, which includes the specific adjustment amount of the stencil temperature setpoint and the adjustment amount of the wire drawing speed setpoint. The execution unit synchronously adjusts the operating parameters of the spindle heater and the wire drawing machine according to the control instructions; The execution unit is the physical execution terminal that ultimately implements the control action in step S3 of the system. This unit receives digital control commands from the prediction and control unit and converts them into control signals for the actual industrial equipment. The unit mainly consists of two parallel controllers. The first is the slug temperature controller, which is usually a high-precision multi-segment temperature controller. According to the adjustment amount of the slug temperature in the instruction, the slug temperature is precisely changed by adjusting the current or power flowing through the platinum-rhodium alloy slug heating element. The second is the fiber drawing speed controller, which is usually a servo drive or frequency converter. Based on the adjustment amount of the drawing speed in the instruction, the fiber drawing speed is precisely changed by adjusting the speed of the main traction motor of the drawing machine. The two controllers act synchronously after receiving instructions at the same time, physically achieving coordinated operation of increasing the spinneret temperature and decreasing the drawing speed or decreasing the spinneret temperature and increasing the drawing speed.

[0036] Example 2 This embodiment provides an experiment on the predictive control accuracy of the FeO activity sub-model; The experiment was conducted using a combination of computer numerical simulation and a small-scale tank kiln drawing test bench. The numerical simulation was used to accurately compare the predictive capabilities of different models under fully controllable conditions. The small-scale test was used to provide real physical property parameters for the simulation and to verify the reliability of the simulation results.

[0037] The control group model was constructed using a conventional melt rheological CFD model. This model uses a fixed viscosity-temperature characteristic equation, and its parameters are calibrated based on a baseline basalt formulation with an FeO content of 11.5%. This model does not include an FeO activity sub-model and cannot dynamically adjust the rheological properties according to changes in raw material composition. The experimental group model constructs a complete digital twin model of melt rheology. This model embeds an FeO activity sub-model. The sub-model dynamically calculates the FeO activity based on the real-time temperature signal from step S1, and uses this to correct the viscosity-temperature characteristic equation. The correction function adopts the effective activation energy formula of Example 1, where the exponent is 1.8, which is greater than 1, reflecting nonlinearity. The model performs online self-correction based on the melt temperature and viscosity obtained by the method of Example 1 in real time. The simulation was set to an extreme but reasonable disturbance scenario. After the wire drawing process had been running stably for 30 minutes, the FeO content of the basalt raw material increased stepwise from the initial 9.0% to 14.0%. Other process parameters were initially set as follows: The temperature of the homogenization section of the pool kiln is 1450℃, the target temperature of the stencil inlet is 1300℃, the stencil diameter is 3mm, and the initial drawing speed is set to 251.2cm / s. The prediction time domain of the model predictive control algorithm parameters is set to 90 seconds, and the control period is set to 8 seconds. In the laboratory, a high-temperature rotational viscometer was used to measure the viscosity data of two basalt melts with FeO contents of 9.0% and 14.0% in the temperature range of 1250℃-1350℃. This data was used to calibrate the fixed equation of the control group model and the baseline parameters of the experimental group model. At the same time, based on thermodynamic software and referring to the research method of the CaO-FeO-MgO-SiO2 slag system activity model, an FeO activity calculation module suitable for the basalt system was established as part of the experimental group model. The control group and experimental group models constructed above were imported into a general computational fluid dynamics software, such as ANSYS Fluent. The two models were run according to the set working conditions to simulate the dynamic process from the step change in raw material composition to a duration of 300 seconds. In the simulation, the sensor, namely the sapphire fiber, provides a set of real melt temperature and inverted viscosity data every 2 seconds; The model predictive control algorithm runs every 8 seconds. Based on the fiber diameter trend predicted by the model, it generates a collaborative adjustment command according to the collaborative adjustment rules to adjust the set value of the sprue temperature. Throughout the simulation, the following key metrics were continuously monitored and recorded for final comparison: The viscosity uniformity at the outlet of the sprue is expressed by the coefficient of variation of the melt viscosity at the outlet of all sprue holes in the sprue. Fiber diameter prediction error: The model's predicted fiber diameter for the next 90 seconds is recorded every 8 seconds and compared with the actual fiber diameter calculated based on simulated physical laws 90 seconds after that moment. The average absolute percentage error is calculated. Record the fluctuation range of the stencil temperature setting and the frequency of the drawing speed adjustment; After the simulation was completed, the data from the control group and the experimental group were statistically compared, and the results are shown in the table below: Table 1 Experimental Results and Comparative Analysis

[0038] Experimental data clearly show that, compared with the traditional method using a fixed parameter model (control group), the dynamic self-calibrating digital twin model (experimental group) of this invention, which includes the FeO activity sub-model, is more effective when faced with a step change in the FeO content of the raw material: The system can recover stability more quickly, significantly improve the uniformity of melt viscosity at the spinneret outlet, provide a more stable material basis for fiber drawing, achieve an order-of-magnitude improvement in the accuracy of fiber diameter prediction, and make the control process more stable, with potential reductions in energy consumption and mechanical losses.

[0039] Finally, it should be noted that the device models used in this embodiment are only for verification and do not limit the scope of this application. The above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit the scope of protection of this application. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this application without departing from the substance and scope of the technical solutions of this application.

Claims

1. A continuous basalt fiber pool kiln drawing process based on predictive control, characterized in that, Includes the following steps: S1, obtain the real-time temperature and real-time apparent viscosity of the basalt melt as dual feedback signals; The real-time apparent viscosity is obtained by inverting the vibration response signal of the sensor element immersed in the melt; S2, The dual feedback signals are synchronously input into the parallel melt rheological digital twin model and the fiber forming process model to predict the evolution trend of fiber diameter; The melt rheological digital twin model is a CFD model based on the dynamic self-correction of FeO content in basalt melt. S3, based on the predicted diameter evolution trend, uses a model predictive control algorithm to coordinate the spinneret temperature and the fiber drawing speed before the fiber diameter deviates from its actual value.

2. The process according to claim 1, characterized in that, In step S1, the vibration response signal is the damping attenuation spectrum of the sensor element at the excitation frequency; The inversion is performed by matching the damping attenuation spectrum with a pre-stored viscosity-damping spectrum database.

3. The process according to claim 1, characterized in that, In step S1, the real-time temperature is obtained by the sensor element by detecting the thermal radiation spectral shift of the melt.

4. The process according to claim 1, characterized in that, In step S1, the sensor element is a sapphire optical fiber.

5. The process according to claim 1, characterized in that, In step S2, the melt rheological digital twin model dynamically corrects the viscosity-temperature characteristic equation of the melt based on the real-time temperature through the embedded FeO activity sub-model.

6. The process according to claim 1, characterized in that, In step S2, the prediction time domain of the melt rheological digital twin model is 30 to 120 seconds.

7. The process according to claim 1, characterized in that, In step S3, the coordinated adjustment specifically includes: When the predicted diameter is smaller, increase the spindle temperature and decrease the drawing speed; When the predicted diameter tends to be larger, reduce the spindle temperature and increase the drawing speed.

8. The process according to claim 1, characterized in that, In step S3, the control cycle of the model predictive control algorithm is 5 to 10 seconds.

9. The process according to claim 7, characterized in that, In step S3, the magnitude of the coordinated adjustment is dynamically proportionally allocated based on the predicted diameter deviation rate.

10. A continuous basalt fiber pool kiln drawing production system for implementing the process described in any one of claims 1 to 9, characterized in that, include: The sensing unit includes sensor elements immersed in the melt in the furnace passage to acquire the real-time temperature and real-time apparent viscosity. The prediction and control unit runs the melt rheological digital twin model, the fiber forming process model, and the model prediction and control algorithm, and generates control commands based on the sensor data. The execution unit synchronously adjusts the operating parameters of the stencil heater and the wire drawing machine according to the control command.

Images