Feeding control method, system, equipment and medium in monocrystalline silicon growth

Abstract

The invention discloses a feeding control method, system, equipment and medium in monocrystalline silicon growth, and relates to the technical field of feeding control, the method comprises the following steps: temperature acquisition is carried out through a temperature sensing array deployed in a reaction furnace, and temperature distribution in the furnace is obtained; raw material attribute characteristics and equipment state characteristics are read, the in-furnace temperature distribution, the raw material attribute characteristics and the equipment state characteristics are input into a charging analysis model, and an optimized charging scheme is output; and through a preset feeding device, silicon material feeding is executed according to the optimal feeding speed, the optimal feeding angle and the optimal feeding amount. The technical problems of lack of accuracy, stability and efficiency of temperature control during feeding in the prior art are solved, and the technical effects of feeding precision, temperature stabilization and stable and reliable material conveying in the monocrystalline silicon growth process are achieved, so that the production efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of feeding control, and in particular to a feeding control method, system, equipment and medium in the growth of single crystal silicon. Background Art

[0002] As a key basic material for the modern electronic information industry and photovoltaic power generation industry, process control during the growth of single crystal silicon is crucial to product quality and production efficiency. In the single crystal silicon growth process, the traditional feeding method has many problems and it is difficult to meet the growing demand for high-quality and high-efficiency production. At present, the common traditional feeding methods mostly rely on manual experience operation or adopt relatively simple automatic control methods. There is a significant inaccuracy in the manual control method, and it is difficult for operators to accurately grasp the feeding speed, angle and amount, which can easily lead to uneven distribution of silicon materials in the furnace, and then cause a series of problems. For example, too much local silicon material may cause the reaction in this area to be too violent, causing the silicon material to splash out of the melt, which not only wastes raw materials, but also may damage the furnace equipment and affect production safety and continuity; while the area with too little silicon material may not be able to fully participate in the reaction, affecting the integrity and quality uniformity of the growth of single crystal silicon. The unevenness of the temperature field in the reactor is also an important factor leading to sputtering of raw materials.

[0003] The prior art has the technical problems of lack of accuracy, stability and efficiency in temperature control during feeding. Summary of the invention

[0004] The present application provides a method, system, device and medium for controlling feeding in the growth of single crystal silicon, which are used to solve the technical problems in the prior art of lack of accuracy, stability and efficiency of temperature control during feeding.

[0005] In view of the above problems, the present application provides a method, system, device and medium for controlling feeding in the growth of single crystal silicon.

[0006] In a first aspect of the present application, a method for controlling feeding in single crystal silicon growth is provided, the method comprising:

[0007] Temperature is collected by a temperature sensor array deployed in the reaction furnace to obtain the temperature distribution in the furnace; raw material property characteristics and equipment status characteristics are read, and the temperature distribution in the furnace, raw material property characteristics and equipment status characteristics are input into a feeding analysis model to output an optimized feeding plan, wherein the optimized feeding plan includes an optimal feeding speed, an optimal feeding angle and an optimal feeding amount; silicon material feeding is performed according to the optimal feeding speed, the optimal feeding angle and the optimal feeding amount through a predetermined feeding device.

[0008] A second aspect of the present application provides a feeding control system for growing single crystal silicon, the system comprising:

[0009] A furnace temperature distribution acquisition module, the furnace temperature distribution acquisition module is used to collect temperature through a temperature sensor array deployed in the reaction furnace to obtain the temperature distribution in the furnace; an optimized feeding plan output module, the optimized feeding plan output module is used to read the raw material property characteristics and equipment status characteristics, input the furnace temperature distribution, raw material property characteristics and equipment status characteristics into a feeding analysis model, and output an optimized feeding plan, wherein the optimized feeding plan includes an optimal feeding speed, an optimal feeding angle and an optimal feeding amount; a silicon material feeding execution module, the silicon material feeding execution module is used to execute silicon material feeding according to the optimal feeding speed, the optimal feeding angle and the optimal feeding amount through a predetermined feeding device.

[0010] In a third aspect of the present application, the present application provides an electronic device, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is used to execute a method for controlling feeding in single crystal silicon growth provided in the present application.

[0011] In a fourth aspect of the present application, the present application provides a computer-readable storage medium storing a computer program, wherein the computer program is used to execute a method for controlling feeding in single crystal silicon growth provided by the present application.

[0012] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0013] The temperature sensor array deployed in the reactor is used to collect temperature and obtain the temperature distribution in the furnace; the raw material attribute characteristics and equipment status characteristics are read, and the temperature distribution in the furnace, the raw material attribute characteristics and the equipment status characteristics are input into the feeding analysis model to output the optimized feeding plan; the silicon material is fed according to the optimal feeding speed, the optimal feeding angle and the optimal feeding amount through the predetermined feeding device. The technical effect of accurate feeding, stable temperature and stable and reliable material transportation in the process of single crystal silicon growth is achieved, thereby improving production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 A schematic flow chart of a method for controlling feeding in the growth of single crystal silicon provided in an embodiment of the present application;

[0016] Figure 2A schematic diagram of the structure of a feeding control system for single crystal silicon growth provided in an embodiment of the present application.

[0017] Figure 3 A schematic diagram of the structure of an electronic device provided in this application.

[0018] Explanation of the reference numerals: furnace temperature distribution acquisition module 10 , optimized feeding scheme output module 20 , silicon material feeding execution module 30 , processor 21 , memory 22 , input device 23 , output device 24 . DETAILED DESCRIPTION

[0019] The present application provides a method, system, device and medium for controlling feeding in single crystal silicon growth, which is used to solve the technical problems in the prior art of lack of accuracy, stability and efficiency of temperature control during feeding.

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] Embodiment 1, as Figure 1 As shown, the present application provides a method for controlling feeding in single crystal silicon growth, the method comprising:

[0022] Step S100: collecting temperature through a temperature sensor array deployed in the reaction furnace to obtain the temperature distribution in the furnace.

[0023] Specifically, the temperature is collected by the temperature sensor array deployed in the reactor. The temperature sensor array is composed of high-precision temperature sensors. The deployment positions of these sensors have been carefully designed and are distributed in multiple key positions of the furnace cavity, including the bottom, side walls, top and near the feeding area. The sensors deployed at the bottom can monitor the temperature of the bottom area of ​​the melt in the furnace, which is crucial for understanding the initial state of silicon melting and the transfer and accumulation of heat at the bottom. The change in the bottom temperature will affect the melting speed and uniformity of the silicon material, and then affect the fusion effect of the silicon material and the melt during the subsequent feeding process. The sensors on the side walls can capture the temperature gradient near the furnace wall. Due to the heat exchange between the furnace wall and the space inside the furnace, the change in the side wall temperature can reflect the process of heat transfer in the furnace horizontally, which helps to evaluate the uniformity of the temperature field in the entire furnace. The upper sensor is responsible for monitoring the temperature of the upper space in the furnace, which is of great significance for understanding the temperature state of the gas environment in the furnace and the distribution of heat in the vertical direction of the furnace. It can also provide a basis for judging the heating condition of the silicon material during the rising process in the furnace. The sensors near the feeding area are of utmost importance, because this area is the entrance for silicon materials to enter the furnace, and its temperature changes directly affect the initial state of silicon materials when they enter the melt and the physical and chemical reactions of silicon materials and melts at the moment of contact. By real-time monitoring of the temperature near the feeding area, the temperature factor can be accurately controlled at the moment of feeding, providing key data support for reducing silicon splashing. These high-precision temperature sensors work together to ensure the accuracy and comprehensiveness of data collection, and can quickly and accurately sense the temperature changes at various locations in the furnace, and transmit these data to the control system in a timely manner. Based on these comprehensive and accurate temperature distribution data, the control system can understand the dynamic changes of the temperature field in the furnace in real time, providing a solid data foundation for subsequent feeding decisions. Whether it is to judge whether the current furnace temperature is suitable for feeding, determine the optimal parameters when feeding, or make real-time adjustments according to temperature changes during the feeding process, this accurate and comprehensive furnace temperature distribution information is indispensable. Only through such meticulous temperature monitoring and data collection can we effectively control the feeding process in the complex process of single crystal silicon growth, improve raw material utilization, reduce energy consumption, increase product yield, and ensure the stability and efficiency of the entire production process.

[0024] Step S200: Read the raw material property characteristics and equipment status characteristics, input the temperature distribution in the furnace, the raw material property characteristics and the equipment status characteristics into the feeding analysis model, and output an optimized feeding plan, wherein the optimized feeding plan includes an optimal feeding speed, an optimal feeding angle and an optimal feeding amount.

[0025] Specifically, the raw material attribute characteristics and equipment status characteristics are first read, which is an important basic data preparation link for generating an optimized feeding plan. The raw material attribute characteristics cover key parameters such as silicon material purity, particle size distribution and shape factor. The purity of silicon material is directly related to the quality of the final single crystal silicon product. High-purity silicon material helps to reduce the introduction of impurities and ensure the crystal structure integrity of single crystal silicon. The particle size distribution has a significant effect on the melting behavior of silicon material in the furnace. The appropriate particle size distribution can ensure that the silicon material is evenly heated during the heating process, and a stable and efficient melting process is achieved. The shape factor determines the accumulation and flow characteristics of silicon material in the furnace. Silicon materials of different shapes have different movement trajectories and contact methods with the melt during the feeding process. Equipment status characteristics should not be ignored either, which at least include aspects such as heater efficiency, stirring rate and feeding nozzle status. The heater efficiency directly affects the temperature increase rate and stability in the furnace. An efficient heater can more accurately control the temperature in the furnace and provide a suitable thermal environment for the growth of single crystal silicon. The stirring rate determines the uniformity of the mixing of silicon material and melt in the furnace. Proper stirring helps heat transfer and material diffusion, avoiding local overheating or uneven composition. The condition of the feeding nozzle is crucial to the feeding process of silicon material. Whether the nozzle is blocked, worn, and the integrity of its internal structure will affect the stability of the ejection speed, angle, and flow rate of the silicon material. The read raw material attribute characteristics and equipment status characteristics are input into the carefully constructed feeding analysis model together with the obtained furnace temperature distribution as input data. The feeding analysis model is obtained by using the sample feeding data set collected based on historical production records to conduct in-depth supervised training of the BP neural network until convergence. During the training process, the sample feeding data set is fully collected based on the historical production records, which includes sample temperature distribution, sample raw material attribute characteristics, and sample equipment status characteristics. After randomly selecting the first sample feeding data, a detailed search is performed based on this to determine the first sample feeding scheme set and the first sample feeding evaluation parameter set related to it. The first sample feeding evaluation parameters here focus on key indicators such as silicon splashing coefficient and feeding time. The silicon splashing coefficient directly reflects the degree of silicon splashing during the feeding process and is one of the important bases for measuring the effect of feeding. The feeding time is closely related to the production efficiency. A shorter feeding time helps to improve the overall efficiency of monocrystalline silicon production. Then, a feeding fitness evaluation function closely related to these evaluation parameters is constructed, which clearly stipulates that the feeding fitness is negatively correlated with the silicon splashing coefficient and the feeding time, that is, the less silicon splashing and the shorter the feeding time, the higher the feeding fitness. According to this function, the first sample feeding evaluation parameter set is rigorously evaluated, and the first sample feeding scheme corresponding to the maximum feeding fitness is carefully selected as the first sample optimal feeding scheme. Then, the first sample feeding data and the first sample optimal feeding scheme are combined into the first training data and added to the training data set. This cycle is repeated to continuously optimize the training data set, and finally a trained feeding analysis model with powerful analysis capabilities is obtained.When the temperature distribution in the furnace, the raw material property characteristics and the equipment status characteristics in actual production are input into the model, the model undergoes complex calculations and analysis, comprehensively considers the interrelationships and influences between various factors, and outputs an optimized feeding plan. The optimized feeding plan includes key elements such as the optimal feeding speed, the optimal feeding angle and the optimal feeding amount. The determination of the optimal feeding speed requires weighing many factors such as the melting speed of the silicon material in the furnace, the flow state of the melt and the avoidance of silicon material splashing. If the feeding speed is too fast, the silicon material may accumulate on the melt surface before it is fully melted, causing silicon splashing, and may also destroy the temperature field and composition uniformity of the melt; while the feeding speed is too slow, it will affect production efficiency and extend the growth cycle of single crystal silicon. The setting of the optimal feeding angle should take into account the temperature field distribution in the furnace, the flow trajectory of the silicon material and the contact method with the melt surface. The appropriate feeding angle can enable the silicon material to enter the melt in the most favorable way, reduce the impact on the melt surface, reduce the risk of silicon splashing, and promote the uniform distribution of silicon material in the melt. The calculation of the optimal feeding amount is based on factors such as the remaining amount of silicon material in the current furnace, the target specifications for single crystal silicon growth, and the reaction state in the furnace. Accurately controlling the feeding amount can ensure the reasonable supply of raw materials during the growth of single crystal silicon, and avoid problems such as abnormal crystal growth or waste of resources caused by excessive or insufficient feeding. The optimized feeding scheme obtained through such a comprehensive and scientific model calculation can effectively reduce silicon splashing in actual feeding operations, improve raw material utilization, and ensure that the feeding process is highly consistent with the complex needs of single crystal silicon growth, thereby improving the quality and efficiency of single crystal silicon growth, reducing production costs, and providing strong support for the optimization and improvement of the entire single crystal silicon production process.

[0026] Step S300: feeding the silicon material through a predetermined feeding device according to the optimal feeding speed, optimal feeding angle and optimal feeding amount.

[0027] Specifically, in the single crystal silicon growth process, through a carefully designed predetermined feeding device, the silicon material feeding operation is performed accurately and stably according to the optimized optimal feeding speed, optimal feeding angle and optimal feeding amount.

[0028] As the primary component of the device, the intelligent rotating nozzle uses wear-resistant and high-temperature resistant materials, such as advanced ceramic-based composite materials, which can withstand the high temperature environment in the furnace (up to 1500°C or more) and the wear caused by the high-speed flow of silicon materials. The internally integrated micro-motor drives the rotating mechanism. The motor uses a high-precision brushless DC motor with excellent speed control accuracy (accurate to ±10rpm) and torque output stability. During the feeding process, according to the pre-set optimal feeding speed, the motor's rotation speed (ranging from 50rpm to 500rpm) and direction are precisely adjusted to use the centrifugal force generated by the rotation to change the trajectory of the silicon material falling. From the analysis of fluid mechanics principles, reasonable control of the rotation speed can make the silicon material form a uniformly dispersed material flow, reduce the agglomeration between silicon material particles, and thus reduce its impact on the melt surface. For example, when the rotation speed is 200rpm, the silicon material can be evenly dispersed and fall in an umbrella shape. Compared with the vertical falling method without rotation, the impact force can be reduced by about 30%-50%, effectively reducing the sputtering phenomenon and creating favorable conditions for the smooth fusion of silicon materials in the melt.

[0029] The adjustable angle feeding arm is designed with a multi-axis linkage mechanical structure, and the angle adjustment is achieved by a high-precision servo motor driving a ball screw, with an angle adjustment accuracy of ±0.1°. It is connected in real time with the furnace temperature field monitoring system (temperature data is obtained through thermocouples arranged at different positions in the furnace, with an accuracy of ±1°C) and material status monitoring sensors (such as monitoring the melt level height using the laser ranging principle, with an accuracy of ±0.5mm). Based on the data fed back by these sensors, the control chip (such as a high-performance programmable logic controller PLC) calculates the optimal feeding angle under the current furnace environment according to the preset algorithm. For example, when the temperature in a certain area of ​​the furnace is high and the melt level is low, the feeding arm will automatically adjust to a smaller tilt angle (such as 30°), so that the silicon material can flow smoothly along the surface of the melt, avoiding splashing caused by the silicon material directly impacting the deep melt due to excessive angles, ensuring that the raw materials enter the melt at the most suitable angle, and promoting the uniform distribution and efficient fusion of the silicon material in the melt.

[0030] The high-precision weight sensor (for example, a strain gauge sensor with a measurement accuracy of ±0.05% of the full scale) integrated in the precision metering and distribution system monitors the weight of silicon materials in real time. The precision valve (such as a high-precision electric butterfly valve with a flow regulation accuracy of ±1%) uses the proportional integral differential PID algorithm to accurately control the valve opening according to the sensor data and the preset optimal feeding amount, and the control system (microprocessor-based intelligent control unit) uses the proportional integral differential PID algorithm to achieve precise control of the volume or mass of each feeding. For example, when the optimal feeding amount is set to 100g, the system can control the actual feeding amount between 99.95g and 100.05g, ensuring accurate feeding on demand, avoiding the melt composition fluctuation caused by excessive feeding, uneven crystal growth, or reduced production efficiency caused by insufficient feeding, and maintaining the stability and consistency of the feeding process.

[0031] The thermal insulation layer is a high-efficiency thermal insulation material added to the outside of the nozzle and the feeding arm, such as nano aerogel felt, which has an extremely low thermal conductivity (as low as 0.02W / (m·K)). It is tightly attached to the surface of the mechanical structure through a special coating process to form an effective thermal insulation barrier. In the high-temperature furnace environment, the thermal insulation layer effectively prevents the transfer of heat to the mechanical parts, ensuring that the temperature of the mechanical structure is maintained at a low level (such as below 200°C). From the perspective of material mechanics, this avoids problems such as thermal expansion of metal parts and decreased material strength caused by high temperatures, thereby reducing operational errors caused by thermal deformation (such as the feeding angle deviation can be controlled within ±0.05°), ensuring the stable operation of mechanical parts in high-temperature environments, extending the service life of the device, and providing reliable guarantees for long-term stable feeding operations.

[0032] The pressure sensor in the dynamic pressure balance device (such as a high-precision piezoresistive sensor with a pressure measurement accuracy of ±0.1kPa) monitors the raw material pressure in the feeding pipeline in real time, and the regulating valve (such as an electric proportional regulating valve with a regulating response time of less than 1s) is adjusted by the control system (microprocessor-based intelligent control module) according to the preset pressure control strategy based on the sensor feedback signal. For example, when the stable pressure before the raw material enters the furnace chamber is set to 50kPa, the control system continuously adjusts the regulating valve opening through the PID algorithm to ensure that the actual pressure fluctuates between 49.95kPa-50.05kPa, avoiding the splashing of raw materials due to sudden pressure changes (such as pressure fluctuations exceeding ±5kPa caused by pipeline blockage or flow changes). This real-time monitoring and precise adjustment mechanism ensures the stability of the pressure of the raw materials during transportation and ensures the stability and safety of the feeding process.

[0033] The intelligent feedback and adaptive control interface realizes real-time bidirectional data transmission between the feeding device and the main control system based on a high-speed and reliable data transmission protocol (such as Ethernet communication protocol, with a data transmission rate of up to 100Mbps). The main control system (such as an industrial computer or a dedicated controller) receives data from multiple sensors such as the temperature sensor array and material state sensor in the furnace, and uses complex intelligent algorithms (such as prediction algorithms based on neural networks) to analyze and calculate the optimal feeding parameters under the current working conditions in the furnace. Then these parameter instructions are sent to the feeding device through the interface. After receiving the instructions, the feeding device immediately adjusts the feeding parameters such as the rotation speed and direction of the intelligent rotating nozzle, the angle of the adjustable angle feeding arm, and the valve opening of the precision metering and distribution system. At the same time, the feeding device feeds back its own operating status data (such as motor speed, valve position, sensor measurement value, etc.) to the main control system in real time. This closed-loop feedback adjustment mechanism enables the feeding device to dynamically adjust the feeding parameters in real time according to changes in the working conditions in the furnace (such as temperature fluctuations, melt level changes, etc.) to maintain the optimal feeding state. For example, when the temperature in the furnace suddenly rises by 10°C, the main control system calculates through an algorithm that the feeding speed needs to be reduced by 10% and the feeding angle needs to be adjusted by 5°. The feeding device responds quickly and ensures that the silicon material feeding always adapts to the dynamically changing environment in the furnace, effectively improving the quality and efficiency of single crystal silicon growth, reducing production costs, and increasing product yield.

[0034] In a possible implementation, step S300 further includes:

[0035] Step S310: according to a predetermined monitoring plan, the temperature sensor array is used to collect real-time temperature distribution in the furnace and real-time silicon material status at fixed points.

[0036] Step S320: performing deviation analysis on the real-time furnace temperature distribution and the real-time silicon material state based on the reference furnace temperature distribution and the reference silicon material state, and determining the temperature distribution deviation and the silicon material state deviation.

[0037] Step S330: Based on the temperature distribution deviation and the silicon material state deviation, optimizing and compensating the optimal feeding speed and the optimal feeding angle.

[0038] Specifically, in the process of single crystal silicon growth, in order to ensure the continuous optimization and precise control of the feeding process, according to the predetermined monitoring plan, the temperature sensor array previously deployed in the reactor is used for fixed-point data collection. The temperature sensor array consists of multiple high-precision temperature sensors, which are distributed in key positions such as the bottom, side walls, top and feeding area of ​​the furnace chamber to ensure that the temperature changes in the furnace can be fully and accurately sensed. When collecting real-time furnace temperature distribution, these sensors measure the temperature of each area in the furnace at an extremely high frequency (for example, collecting data once every 100 milliseconds), so as to obtain detailed temperature field information, and the measurement accuracy can reach ±0.1°C. At the same time, with the help of specially designed silicon material state monitoring technology, such as the silicon material state monitoring system based on the principle of optical imaging, the high-definition camera is installed at a specific position on the furnace wall, combined with advanced image processing algorithms, the shape, distribution and flow state of the silicon material in the furnace can be monitored and analyzed in real time, so as to obtain real-time silicon material state information. These data will provide a solid foundation for subsequent deviation analysis and optimization of feeding parameters.

[0039] After obtaining the real-time furnace temperature distribution and real-time silicon material state, the real-time collected data is carefully compared with the pre-set reference furnace temperature distribution and reference silicon material state as the reference standard. The reference furnace temperature distribution is determined based on a large amount of historical production data and the ideal temperature curve of single crystal silicon growth. It represents the distribution state of the furnace temperature under the optimal feeding and growth conditions, such as the stable temperature range that a specific area should maintain at different stages of single crystal silicon growth. The reference silicon material state covers the melting degree of silicon material under ideal conditions, the uniform distribution state in the melt, and the interaction state with other substances in the furnace. Through complex data processing algorithms, the difference between the real-time furnace temperature distribution and the reference temperature distribution is calculated to obtain the temperature distribution deviation. For example, in a critical area, if the real-time temperature is 5°C higher than the reference temperature, it indicates that there is a temperature deviation. Similarly, the real-time silicon material state is compared and analyzed with the reference silicon material state to determine the silicon material state deviation, such as the deviation degree of the actual distribution of silicon material in the melt from the ideal uniform distribution, or the difference between the melting speed of silicon material and the expected speed. These deviation values ​​will intuitively reflect the deviation of the current feeding process from the ideal state, providing a clear direction for subsequent optimization and compensation.

[0040] After obtaining the temperature distribution deviation and silicon material state deviation, these data will be used as input for the machine learning compensation model. The compensation model is trained based on a large amount of historical production data, which covers the optimal feeding speed and angle adjustment strategies corresponding to different temperature deviations and silicon material state deviations. For example, the data set contains multiple groups of samples such as when the temperature in a certain area of ​​the furnace is 5°C higher and the silicon material is locally accumulated, the feeding speed should be reduced by 15% and the feeding angle should be adjusted to 35°. The model uses a neural network algorithm to construct a neural network structure including an input layer (temperature deviation and silicon material state deviation), a hidden layer (several neurons, used to extract data features and patterns) and an output layer (optimal feeding speed and angle adjustment amount). During the training process, the connection weights between neurons are continuously adjusted so that the model can accurately predict reasonable feeding speed and angle adjustment values ​​based on the input deviation data. Based on the trained model, when the real-time temperature distribution deviation and silicon material state deviation are input, the model quickly calculates and outputs the corresponding optimal feeding speed and optimal feeding angle optimization compensation amount. Assume that the current temperature distribution deviation shows that the temperature of the feeding area in the furnace is 3°C higher than the reference temperature, and the silicon material state deviation indicates that the silicon material has a slight tendency to accumulate on the surface of the melt. After complex calculations, the model concludes that the optimal feeding speed needs to be reduced by 8% and the optimal feeding angle needs to be adjusted from 40° to 32°. According to the output results of the model, the control system immediately sends instructions to the predetermined feeding device, the micro motor of the intelligent rotating nozzle adjusts the speed to change the falling speed of the silicon material, and the servo motor of the adjustable angle feeding arm accurately adjusts the angle. Through such an optimization and compensation process, it can dynamically adapt to the changes in the working conditions in the furnace in real time, effectively reduce the silicon splashing phenomenon caused by abnormal temperature and silicon material state, improve the uniformity of silicon material in the melt, promote the stable growth of single crystal silicon, improve product quality and production efficiency, and make the entire single crystal silicon production process more intelligent, precise and efficient.

[0041] In a possible implementation, step S100 further includes:

[0042] Step S110: The temperature sensing array includes a plurality of temperature sensors, which are respectively deployed at predetermined positions in the reaction furnace, wherein the predetermined positions at least include a bottom, a side wall, an upper portion, and a feeding area.

[0043] Specifically, in the single crystal silicon growth process, the deployment of the temperature sensor array is the key link to accurately obtain the temperature information in the reactor, and its reasonable layout provides an indispensable data basis for the subsequent feeding control and the entire growth process. The temperature sensor array consists of multiple high-precision temperature sensors, which are carefully deployed at predetermined positions in the reactor according to specific design requirements. Among them, the sensors at the bottom position play a vital role. They can directly contact and monitor the temperature conditions of the bottom area of ​​the melt in the furnace. Since the bottom is the starting position of silicon melting, the temperature change here directly affects the initial melting speed and uniformity of the silicon material, and has a fundamental impact on the stability and quality of the entire single crystal silicon growth process. The side wall sensors are distributed around the reactor, and their main function is to capture the temperature dynamics of the area near the furnace wall. There is heat exchange between the furnace wall and the space inside the furnace. The change in the side wall temperature can reflect the process of heat transfer in the furnace horizontally, which helps to evaluate the uniformity of the temperature field in the entire furnace, and timely discover possible local overheating or overcooling areas, providing a basis for adjusting the heating strategy. The upper sensor is responsible for monitoring the temperature of the upper space in the furnace, which is of great significance for understanding the temperature state of the gas environment in the furnace and the distribution of heat in the vertical direction of the furnace. During the growth of single crystal silicon, factors such as silicon material sublimation and gas convection will cause changes in the temperature of the upper space. These changes are interrelated with the temperature of the lower melt and jointly affect the physical and chemical process of single crystal silicon growth. The feeding area is a key area for silicon materials to enter the furnace, so it is particularly important to deploy sensors here. The temperature of the feeding area directly affects the initial state of the silicon material when it enters the melt, such as the melting rate of the silicon material, the degree of mixing with the melt, etc., and is also closely related to the occurrence of silicon splashing. By deploying sensors in the feeding area, the temperature changes in the area can be monitored in real time, and the temperature factors can be accurately controlled at the moment of feeding, providing key data support for optimizing the feeding strategy and reducing the risk of silicon splashing. By deploying temperature sensors at least at predetermined positions including the bottom, side walls, top and feeding area, it is possible to achieve all-round and multi-angle monitoring of the temperature field in the reactor, ensuring that comprehensive, accurate and representative temperature distribution information is obtained, laying a solid foundation for precise control of the single crystal silicon growth process.

[0044] In a possible implementation, step S200 further includes:

[0045] Step S210: reading raw material property characteristics and equipment status characteristics, wherein the raw material property characteristics at least include silicon material purity, particle size distribution and shape factor, and the equipment status characteristics at least include heater efficiency, stirring rate and feeding nozzle status.

[0046] Step S220: Collect a sample feeding data set according to historical production records, and use the sample feeding data set to supervise the BP neural network until convergence, so as to obtain a trained feeding analysis model.

[0047] Step S230: input the temperature distribution in the furnace, the raw material property characteristics and the equipment status characteristics into the charging analysis model, and output the optimized charging plan.

[0048] Specifically, when reading the raw material property characteristics, the purity of silicon material is a key indicator, which is directly related to the quality of single crystal silicon products. High-purity silicon material helps to reduce the introduction of impurities and ensure the integrity of the crystal structure of single crystal silicon, thereby improving its electrical properties. In the semiconductor industry, the requirements for the purity of silicon material are extremely high. Slight differences in purity may cause significant changes in key performance indicators such as resistivity and carrier mobility of single crystal silicon. Particle size distribution should not be ignored either, as it affects the melting behavior of silicon material in the furnace. Silicon materials of different particle sizes have different heating and melting rates during the heating process. Reasonable particle size distribution can ensure that the silicon material is evenly heated in the melt and achieve a stable and efficient melting process. The shape factor determines the accumulation and flow characteristics of silicon material in the furnace. Silicon materials with regular and uniform shapes are more likely to form a stable material flow during the feeding process, which is conducive to precise control of the feeding process and reducing the problem of uneven feeding caused by unstable silicon material flow. At the same time, the reading of equipment status characteristics is also crucial. The efficiency of the heater directly affects the rate and stability of the temperature rise in the furnace. An efficient heater can more accurately control the temperature in the furnace and provide a suitable thermal environment for the growth of single crystal silicon. At different stages of single crystal silicon growth, the temperature in the furnace needs to be accurately controlled within a specific range. The efficiency of the heater directly determines whether this requirement can be met. The stirring rate affects the uniformity of the mixing of silicon material and melt in the furnace. Proper stirring helps heat transfer and material diffusion, avoiding local overheating or uneven composition. For example, during the melting of silicon material, stirring can transfer heat to the inside of the silicon material faster, accelerate the melting process, and make the impurities in the melt more evenly distributed, which is conducive to subsequent crystal growth. The condition of the feeding nozzle plays a key role in the feeding process of silicon material. Whether the nozzle is blocked, worn, and the integrity of its internal structure will affect the stability of the ejection speed, angle, and flow rate of the silicon material. A feeding nozzle in good condition can ensure that the silicon material accurately enters the melt in the furnace at a stable speed, angle, and flow rate, achieving accurate feeding and reducing silicon splashing.

[0049] Collecting sample feeding data sets based on historical production records is an important basis for building an accurate feeding analysis model. During the collection process, sample data under various working conditions are comprehensively collected, including information such as sample temperature distribution, raw material attribute characteristics and equipment status characteristics under different silicon material purities, particle size distributions, shape factors, and different heater efficiencies, stirring rates, and feeding nozzle condition combinations. These rich sample feeding data sets are used to supervise the training of the BP neural network. The BP neural network has a strong nonlinear mapping ability and can learn the complex nonlinear relationship between input data (sample feeding data) and output data (optimal feeding scheme). During the training process, the first sample feeding data is randomly selected, and a detailed search is performed based on this to determine the first sample feeding scheme set and the first sample feeding evaluation parameter set related to it. The evaluation parameters here focus on key indicators such as the silicon splashing coefficient and the feeding time, because the silicon splashing coefficient directly reflects the degree of silicon material splashing during the feeding process, which is one of the important bases for measuring the quality of the feeding effect; the feeding time is closely related to the production efficiency, and a shorter feeding time helps to improve the overall efficiency of monocrystalline silicon production. Construct a charging fitness evaluation function that is closely related to these evaluation parameters, and clearly stipulate that the charging fitness is negatively correlated with the silicon splashing coefficient and the charging time, that is, the less silicon splashing and the shorter the charging time, the higher the charging fitness. According to this function, the first sample charging evaluation parameter set is rigorously evaluated, and the first sample charging scheme corresponding to the maximum charging fitness is carefully selected as the first sample optimal charging scheme, and then the first sample charging data and the first sample optimal charging scheme are combined into the first training data and added to the training data set. This cycle is repeated to continuously optimize the training data set, adjust the weights and thresholds of the neural network, and make the network's predicted output continuously approach the true optimal charging scheme until the model converges, thereby obtaining a trained charging analysis model with powerful analysis capabilities.

[0050] When the temperature distribution in the furnace, the raw material property characteristics and the equipment status characteristics in actual production are input into the trained feeding analysis model, the model begins to play its core role. The neurons inside the model perform complex calculations and information transmission based on the input data, and comprehensively consider the relationship and influence between various factors. Based on the patterns and rules learned during the training process, the model outputs an optimized feeding plan, which includes key elements such as the optimal feeding speed, the optimal feeding angle and the optimal feeding amount. The determination of the optimal feeding speed requires weighing many factors such as the melting speed of the silicon material in the furnace, the flow state of the melt, and avoiding the splashing of the silicon material. If the feeding speed is too fast, the silicon material may accumulate on the surface of the melt before it is fully melted, causing silicon splashing, and may also destroy the temperature field and composition uniformity of the melt; while the feeding speed is too slow, it will affect production efficiency and extend the growth cycle of single crystal silicon. The setting of the optimal feeding angle should take into account the temperature field distribution in the furnace, the flow trajectory of the silicon material, and the contact method with the melt surface. The appropriate feeding angle can enable the silicon material to enter the melt in the most favorable way, reduce the impact on the melt surface, reduce the risk of silicon splashing, and promote the uniform distribution of silicon material in the melt. The calculation of the optimal feeding amount is based on factors such as the remaining amount of silicon material in the current furnace, the target specifications for single crystal silicon growth, and the current reaction state in the furnace. Accurate control of the feeding amount can ensure the reasonable supply of raw materials during the growth of single crystal silicon, and avoid problems such as abnormal crystal growth or waste of resources caused by excessive or insufficient feeding. The optimized feeding scheme obtained through such a comprehensive and scientific model calculation can effectively reduce the silicon splashing phenomenon in the actual feeding operation, improve the raw material utilization rate, and ensure that the feeding process is highly consistent with the complex needs of single crystal silicon growth, thereby improving the quality and efficiency of single crystal silicon growth, reducing production costs, and providing strong support for the optimization and improvement of the entire single crystal silicon production process.

[0051] In a possible implementation, step S220 further includes:

[0052] Step S221: Collecting a sample feeding data set according to historical production records, wherein the sample feeding data includes sample temperature distribution, sample raw material property characteristics and sample equipment status characteristics.

[0053] Step S222: randomly selecting first sample feeding data, searching based on the first sample feeding data, and determining a first sample feeding scheme set and a first sample feeding evaluation parameter set, wherein the first sample feeding evaluation parameters include a silicon sputtering coefficient and a feeding time.

[0054] Step S223: constructing a charging fitness evaluation function, wherein the charging fitness is negatively correlated with the silicon sputtering coefficient and negatively correlated with the charging time.

[0055] Step S224: Evaluate the first sample feeding evaluation parameter set according to the feeding fitness evaluation function, and select the first sample feeding scheme corresponding to the maximum feeding fitness as the first sample optimal feeding scheme.

[0056] Step S225: forming first training data according to the first sample feeding data and the first sample optimal feeding scheme, and adding the first training data to the training data set.

[0057] Step S226: Use the training data set to perform supervised training on the BP neural network until convergence, and obtain a trained feed analysis model.

[0058] Specifically, by deeply mining the detailed records of a large number of past single crystal silicon production processes, sample data covering a variety of working conditions are collected. The sample feeding data comprehensively includes the sample temperature distribution, sample raw material property characteristics and sample equipment status characteristics. The sample temperature distribution data accurately records the temperature changes at different locations in the reactor at various production stages. These data reflect the complex thermal field environment in the furnace, which is crucial for understanding the heat transfer laws and the influence of temperature on the state of silicon materials during the growth of single crystal silicon. The sample raw material property characteristics record in detail information such as silicon material purity, particle size distribution and shape factor. Silicon materials of different purities have different chemical activities and physical properties during the reaction process. The particle size distribution affects the heating area and melting rate of the silicon material, and the shape factor determines the accumulation and flow characteristics of the silicon material in the furnace. These factors work together on the feeding process and the growth results of single crystal silicon. The sample equipment status characteristics record equipment-related information such as heater efficiency, stirring rate and feeding nozzle status. The heater efficiency directly affects the control accuracy of the temperature in the furnace. The stirring rate affects the mixing uniformity of silicon material and melt. The feeding nozzle status determines the feeding method and effect of silicon material. These changes in equipment status have a significant impact on the production process.

[0059] After obtaining the sample feeding data set, the first sample feeding data is randomly selected, and subsequent operations are carried out based on this data. Based on the first sample feeding data, the first sample feeding scheme set and the first sample feeding evaluation parameter set related to it are determined through the data retrieval and analysis mechanism. In actual production, different feeding strategies will be adopted for each specific set of sample temperature distribution, raw material property characteristics and equipment status characteristics. These strategies constitute the first sample feeding scheme set, and the first sample feeding evaluation parameter set focuses on the key measurement indicators of the effects of these feeding schemes, among which the silicon splashing coefficient and the feeding time are particularly important. The silicon splashing coefficient directly reflects the degree of silicon material splashing during the feeding process. This is a key parameter closely related to product quality and production efficiency. Silicon material splashing not only causes raw material loss, but also may affect the furnace environment and the stability of single crystal silicon growth. The feeding time is directly related to production efficiency. A shorter feeding time helps to produce more single crystal silicon products per unit time, reduce production costs, and improve production efficiency.

[0060] A feeding fitness evaluation function is constructed, which clearly stipulates the negative correlation between the feeding fitness and the silicon splashing coefficient and the feeding time. That is, the smaller the silicon splashing coefficient and the shorter the feeding time, the higher the feeding fitness. From the perspective of mathematical models, this negative correlation is expressed in the form of multiple functions, in the form of weighted summation, in which the silicon splashing coefficient and the feeding time are multiplied by the corresponding negative weight coefficients, and then the sum is obtained to obtain the feeding fitness. This construction method enables the evaluation function to accurately reflect the comprehensive performance of the feeding scheme in reducing silicon material waste and improving production efficiency, and provides a scientific and objective evaluation standard for the subsequent screening of the optimal feeding scheme.

[0061] According to the constructed feeding fitness evaluation function, the first sample feeding evaluation parameter set is rigorously evaluated, and the silicon splash coefficient and feeding time corresponding to each group of first sample feeding schemes are substituted into the evaluation function to calculate the respective feeding fitness values. Then, by comparing these fitness values, the first sample feeding scheme corresponding to the maximum feeding fitness is selected and set as the first sample optimal feeding scheme. This process is similar to finding the optimal solution among many candidate schemes. Through quantitative evaluation indicators, the most suitable feeding scheme under specific sample conditions is determined. This scheme performs best in reducing silicon splash and shortening feeding time, providing an ideal reference standard for subsequent model training.

[0062] After determining the optimal feeding scheme for the first sample, the first sample feeding data and the first sample optimal feeding scheme are combined into the first training data and added to the training data set. This operation continuously expands the training data set to include more representative sample data and the corresponding optimal feeding scheme. As this process is repeated, the training data set is gradually enriched and improved, providing sufficient data support for the training of the BP neural network, which helps to improve the accuracy and generalization ability of the model, so that it can better adapt to various actual production situations and accurately predict the optimal feeding scheme.

[0063] The BP neural network is supervised and trained using the continuously expanding training data set. The BP neural network consists of an input layer, a hidden layer, and an output layer. During the training process, the input layer receives data such as sample temperature distribution, raw material attribute characteristics, and equipment status characteristics. Through the complex calculation and information transmission of neurons in the hidden layer, the output layer attempts to predict the corresponding optimal feeding scheme (including optimal feeding speed, optimal feeding angle, and optimal feeding amount, etc.). Then, the predicted result is compared with the actual optimal feeding scheme of the first sample, and the error is calculated. According to the error back propagation algorithm, the error is back propagated from the output layer to the input layer layer by layer, and the connection weights and thresholds between neurons are adjusted to reduce the error. This process is repeated continuously. As the number of training rounds increases, the prediction ability of the model gradually improves and the error gradually decreases. Until the model converges, that is, the error reaches a stable minimum value, the trained feeding analysis model is obtained at this time. After training and optimization with a large amount of data, the model can accurately output the optimized feeding scheme according to various input conditions in actual production, providing strong support for precise feeding control in the growth process of single crystal silicon, effectively improving production efficiency, reducing costs, and improving product quality.

[0064] In a possible implementation, step S300 further includes:

[0065] Step S310: The predetermined feeding device includes an intelligent rotating nozzle, an adjustable angle feeding arm, a precision metering and distribution system, a thermal insulation protection layer, a dynamic pressure balance device, an intelligent feedback and adaptive control interface, and a closed-loop feedback control system.

[0066] Specifically, the predetermined feeding device involved in the single crystal silicon growth process is a highly integrated and coordinated system. The intelligent rotating nozzle is made of wear-resistant and high-temperature resistant materials. The internal micro-motor drives the rotating mechanism, which can accurately control the trajectory and speed of the silicon material to reduce the risk of silicon splashing. The adjustable angle feeding arm has a multi-axis flexible adjustment function, and refers to the real-time temperature field and material state in the furnace to accurately adjust the feeding angle to ensure that the silicon material is suitable for entering the melt. The precision metering and distribution system uses high-precision weight sensors and precision valves, and uses advanced algorithms to achieve precise metering and control to maintain stable and consistent feeding. The thermal insulation layer uses high-efficiency thermal insulation materials to prevent heat transfer, reduce thermal deformation errors, and ensure stable operation of mechanical parts. The dynamic pressure balance device monitors and adjusts the raw material pressure in real time through pressure sensors and regulating valves to avoid splashing caused by sudden pressure changes. The intelligent feedback and adaptive control interface realizes data interaction and command transmission based on high-speed transmission protocols, so that the device can dynamically adjust parameters according to the analysis results of the main control system algorithm. The closed-loop feedback control system integrates sensor data and algorithm analysis to form a real-time monitoring and automatic adjustment mechanism to cope with changes in working conditions in the furnace and maintain the best feeding state. The various components cooperate with each other to promote the precision, intelligence and efficiency of silicon material feeding during the growth of single crystal silicon.

[0067] In a possible implementation, step S226 further includes:

[0068] Step S2261: After the current stage of feeding is completed, the temperature sensor array is used to collect and obtain the current temperature distribution in the furnace.

[0069] Step S2262: Adaptively adjust the feeding scheme according to the current temperature distribution in the furnace through the feeding analysis model.

[0070] Specifically, during the growth of single crystal silicon, after the feeding of each stage is completed, the temperature sensor array previously deployed in the reactor is immediately started to perform key temperature distribution acquisition work. The temperature sensor array is composed of multiple high-precision temperature sensors, which are widely distributed in various key positions in the reactor, including the bottom, side walls, top and feeding area. At the moment when the feeding is completed in the current stage, the sensor quickly captures the temperature changes at various positions in the furnace at an extremely high frequency (for example, collecting data every 100 milliseconds) to obtain detailed and accurate information on the current temperature distribution in the furnace. These data will become key basic data for evaluating the current state of the furnace and providing a basis for subsequent adjustments to the feeding strategy. For example, if the temperature in a certain area of ​​the furnace rises or falls abnormally after the feeding is completed, it may affect the further reaction of the silicon material and the growth quality of the single crystal silicon. Therefore, accurately obtaining the temperature distribution at this time is crucial for timely detection of problems and taking corresponding measures.

[0071] Based on the collected data of the current temperature distribution in the furnace, the feeding scheme is adaptively adjusted through the constructed advanced feeding analysis model. The feeding analysis model is an intelligent model trained with a large amount of historical production data. It can understand the complex relationship between the temperature distribution in the furnace and the optimal feeding strategy. When the current temperature distribution data in the furnace is input, the model quickly calculates and determines the optimal feeding strategy that adapts to the current temperature state in the furnace according to its internal complex algorithms and learned patterns. If the current temperature distribution in the furnace shows that the temperature in a certain area is too high, according to the principles of heat conduction and silicon material reaction kinetics, the model will output adjustment strategies such as reducing the feeding speed and adjusting the feeding angle to avoid the high temperature area to prevent the silicon material from over-reacting at high temperature or causing silicon splashing, ensuring that the silicon material can be smoothly integrated into the melt under a suitable temperature environment to promote the stable growth of single crystal silicon. On the contrary, if the temperature is low, the model will recommend measures such as increasing the feeding speed or adjusting the heating power to restore the temperature in the furnace to a range that is conducive to the growth of single crystal silicon as soon as possible. Through this adaptive adjustment mechanism based on real-time temperature data, it can ensure that the entire feeding process is closely aligned with the dynamic needs of single crystal silicon growth, effectively improve the quality and efficiency of single crystal silicon growth, reduce the risk of growth defects caused by temperature fluctuations, and achieve precise control and optimization of the single crystal silicon production process.

[0072] Embodiment 2 is based on the same inventive concept as the method for controlling feeding in single crystal silicon growth in the above embodiment. Figure 2 As shown, the present application provides a feeding control system for single crystal silicon growth, and the system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0073] The furnace temperature distribution acquisition module 10 is used to acquire the temperature distribution in the furnace by collecting temperature through a temperature sensor array deployed in the reaction furnace.

[0074] The optimized feeding scheme output module 20 is used to read the raw material property characteristics and equipment status characteristics, input the temperature distribution in the furnace, the raw material property characteristics and the equipment status characteristics into the feeding analysis model, and output the optimized feeding scheme, wherein the optimized feeding scheme includes the optimal feeding speed, the optimal feeding angle and the optimal feeding amount.

[0075] The silicon material feeding execution module 30 is used to execute silicon material feeding according to the optimal feeding speed, optimal feeding angle and optimal feeding amount through a predetermined feeding device.

[0076] Furthermore, the silicon material feeding execution module 30 also includes:

[0077] The temperature distribution acquisition unit is used to collect the real-time temperature distribution in the furnace and the real-time silicon material status at a fixed point using the temperature sensor array according to a predetermined monitoring plan.

[0078] The silicon material state deviation determining unit is used to perform deviation analysis on the real-time furnace temperature distribution and the real-time silicon material state based on the reference furnace temperature distribution and the reference silicon material state, and determine the temperature distribution deviation and the silicon material state deviation.

[0079] An optimization compensation unit is used to optimize and compensate the optimal feeding speed and the optimal feeding angle based on the temperature distribution deviation and the silicon material state deviation.

[0080] Furthermore, the furnace temperature distribution acquisition module 10 further includes:

[0081] The temperature sensing array includes a plurality of temperature sensors respectively disposed at predetermined positions in the reaction furnace, wherein the predetermined positions at least include a bottom, a side wall, an upper portion and a feeding area.

[0082] Furthermore, the optimized feeding scheme output module 20 also includes:

[0083] A feature reading unit, wherein the feature reading unit is used to read raw material property features and equipment status features, wherein the raw material property features at least include silicon material purity, particle size distribution and shape factor, and the equipment status features at least include heater efficiency, stirring rate and feeding nozzle status.

[0084] The supervised training unit is used to collect a sample feeding data set according to historical production records, and use the sample feeding data set to perform supervised training on the BP neural network until convergence, so as to obtain a trained feeding analysis model.

[0085] The optimized feeding scheme output unit is used to input the temperature distribution in the furnace, the raw material property characteristics and the equipment status characteristics into the feeding analysis model, and output the optimized feeding scheme.

[0086] Furthermore, the supervised training unit also includes:

[0087] A feed data set acquisition unit is used to acquire a sample feed data set according to historical production records, wherein the sample feed data includes sample temperature distribution, sample raw material property characteristics and sample equipment status characteristics.

[0088] A feeding evaluation parameter set determination unit is used to randomly select first sample feeding data, perform retrieval based on the first sample feeding data, and determine a first sample feeding scheme set and a first sample feeding evaluation parameter set, wherein the first sample feeding evaluation parameters include a silicon sputtering coefficient and a feeding time.

[0089] A fitness evaluation function construction unit is used to construct a charging fitness evaluation function, wherein the charging fitness is negatively correlated with the silicon sputtering coefficient and negatively correlated with the charging time.

[0090] The optimal feeding scheme acquisition unit is used to evaluate the first sample feeding evaluation parameter set according to the feeding fitness evaluation function, and select the first sample feeding scheme corresponding to the maximum feeding fitness as the first sample optimal feeding scheme.

[0091] The first training data composition unit is used to compose first training data according to the first sample feeding data and the first sample optimal feeding scheme, and add the first training data to the training data set.

[0092] The feed analysis model acquisition unit is used to perform supervised training on the BP neural network using the training data set until convergence, and obtain the trained feed analysis model.

[0093] Furthermore, the optimized feeding scheme output module 20 also includes:

[0094] The predetermined feeding device includes an intelligent rotating nozzle, an adjustable angle feeding arm, a precision metering and distribution system, a heat insulation protection layer, a dynamic pressure balancing device, an intelligent feedback and adaptive control interface, and a closed-loop feedback control system.

[0095] Furthermore, the feeding analysis model acquisition unit also includes:

[0096] The current furnace temperature distribution acquisition unit is used to acquire the current furnace temperature distribution by using the temperature sensor array after the current stage of feeding is completed.

[0097] An adaptive adjustment unit is used to perform adaptive adjustment of the charging scheme according to the current temperature distribution in the furnace through the charging analysis model.

[0098] Embodiment 3, Figure 3 This is a schematic diagram of the structure of an electronic device provided in Example 3 of the present application, and is a block diagram of an exemplary electronic device suitable for implementing an implementation of the present invention. Figure 3The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention. Figure 3 As shown, the electronic device includes a processor 21, a memory 22, an input device 23 and an output device 24; the number of processors 21 in the electronic device can be one or more. Figure 3 Taking a processor 21 as an example, the processor 21, memory 22, input device 23 and output device 24 in the electronic device can be connected through a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0099] In the fourth embodiment, the memory 22 is a computer-readable storage medium that can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to a feeding control method in the growth of single crystal silicon in the embodiment of the present application. The processor 21 executes various functional applications and data processing of the computer device by running the software programs, instructions and modules stored in the memory 22, that is, realizing the feeding control method in the growth of single crystal silicon.

[0100] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0101] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0102] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for controlling feeding in the growth of single crystal silicon, characterized in that: include: The temperature distribution in the furnace is obtained by collecting temperature through a temperature sensor array deployed in the reaction furnace; Reading raw material property characteristics and equipment status characteristics, inputting the temperature distribution in the furnace, raw material property characteristics and equipment status characteristics into a feeding analysis model, and outputting an optimized feeding plan, wherein the optimized feeding plan includes an optimal feeding speed, an optimal feeding angle and an optimal feeding amount; The silicon material is fed according to the optimal feeding speed, the optimal feeding angle and the optimal feeding amount through a predetermined feeding device.

2. The method for controlling feeding in the growth of single crystal silicon according to claim 1, characterized in that: The silicon material is fed according to the optimal feeding speed, the optimal feeding angle and the optimal feeding amount, and then the method further comprises: According to a predetermined monitoring plan, the temperature sensor array is used to collect the real-time temperature distribution in the furnace and the real-time silicon material state at fixed points; Based on the reference furnace temperature distribution and the reference silicon material state, a deviation analysis is performed on the real-time furnace temperature distribution and the real-time silicon material state to determine the temperature distribution deviation and the silicon material state deviation; Based on the temperature distribution deviation and the silicon material state deviation, the optimal feeding speed and the optimal feeding angle are optimized and compensated.

3. The method for controlling feeding in the growth of single crystal silicon according to claim 1, characterized in that: The temperature sensing array includes a plurality of temperature sensors respectively disposed at predetermined positions in the reaction furnace, wherein the predetermined positions at least include a bottom, a side wall, an upper portion and a feeding area.

4. The method for controlling feeding in the growth of single crystal silicon according to claim 1, characterized in that: Read the raw material attribute characteristics and equipment status characteristics, input the temperature distribution in the furnace, the raw material attribute characteristics and the equipment status characteristics into the charging analysis model, and output the optimized charging plan, including: Reading raw material property characteristics and equipment status characteristics, wherein the raw material property characteristics at least include silicon material purity, particle size distribution and shape factor, and the equipment status characteristics at least include heater efficiency, stirring rate and feeding nozzle status; Collecting a sample feeding data set according to historical production records, and using the sample feeding data set to supervise the BP neural network until convergence, to obtain a trained feeding analysis model; The temperature distribution in the furnace, the raw material property characteristics and the equipment status characteristics are input into the charging analysis model, and the optimized charging plan is output.

5. The method for controlling feeding in the growth of single crystal silicon according to claim 4, characterized in that: The BP neural network is supervised and trained using the sample feed data set until convergence, and a trained feed analysis model is obtained, including: Collecting a sample feeding data set according to historical production records, wherein the sample feeding data includes sample temperature distribution, sample raw material attribute characteristics, and sample equipment status characteristics; Randomly selecting first sample feeding data, searching based on the first sample feeding data, and determining a first sample feeding scheme set and a first sample feeding evaluation parameter set, wherein the first sample feeding evaluation parameters include a silicon splashing coefficient and a feeding time; Constructing a charging fitness evaluation function, wherein the charging fitness is negatively correlated with the silicon sputtering coefficient and negatively correlated with the charging time; The first sample feeding evaluation parameter set is evaluated according to the feeding fitness evaluation function, and the first sample feeding scheme corresponding to the maximum feeding fitness is selected as the first sample optimal feeding scheme; According to the first sample feeding data and the first sample optimal feeding scheme, first training data is formed and added to the training data set; The training data set is used to perform supervised training on the BP neural network until convergence, thereby obtaining a trained feed analysis model.

6. The method for controlling feeding in the growth of single crystal silicon according to claim 1, characterized in that: The predetermined feeding device includes an intelligent rotating nozzle, an adjustable angle feeding arm, a precision metering and distribution system, a heat insulation protection layer, a dynamic pressure balancing device, an intelligent feedback and adaptive control interface, and a closed-loop feedback control system.

7. The method for controlling feeding in the growth of single crystal silicon according to claim 1, characterized in that: The method further comprises: After the current stage of feeding is completed, the temperature distribution in the current furnace is acquired by using the temperature sensor array; The charging analysis model is used to adaptively adjust the charging scheme according to the current temperature distribution in the furnace.

8. A feeding control system for growing single crystal silicon, characterized in that: include: A furnace temperature distribution acquisition module, which is used to acquire temperature through a temperature sensor array deployed in the reaction furnace to acquire the temperature distribution in the furnace; An optimized feeding scheme output module, which is used to read the raw material property characteristics and equipment status characteristics, input the temperature distribution in the furnace, the raw material property characteristics and the equipment status characteristics into the feeding analysis model, and output an optimized feeding scheme, wherein the optimized feeding scheme includes an optimal feeding speed, an optimal feeding angle and an optimal feeding amount; A silicon material feeding execution module is used to execute silicon material feeding according to the optimal feeding speed, optimal feeding angle and optimal feeding amount through a predetermined feeding device.

9. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; The processor is used to implement a feeding control method in single crystal silicon growth as described in any one of claims 1-7 when executing the executable instructions stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a method for controlling feeding in single crystal silicon growth as described in any one of claims 1 to 7 is implemented.

Images