A method and system for intelligent temperature field control based on HVPE equipment

Abstract

This invention discloses an intelligent temperature field control method and system based on HVPE equipment. This invention relates to the field of temperature field control. The method comprises: S1: Multi-source heterogeneous data acquisition and spatiotemporal fusion; S2: Quantitative evaluation and distributed collaborative control of temperature field equilibrium, using the data acquired in S1 to calculate the temperature field equilibrium degree, and completing the judgment of temperature field imbalance and the collaborative control of the heater through a six-dimensional fuzzy PID controller and a consistency algorithm to improve temperature field uniformity and stability; S3: Dynamic compensation of four coupled fields (heat, flow, deposition, and pressure); S4: Dynamic learning optimization based on digital twins. The intelligent temperature field control method and system based on HVPE equipment described in this invention, based on multi-source heterogeneous data acquisition, temperature field equilibrium control, four-field coupling compensation, and digital twin optimization, achieves a comprehensive improvement in temperature field control accuracy, adaptability to complex operating conditions, and equipment operation and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of temperature field control, and in particular to an intelligent temperature field regulation method and system based on HVPE equipment. Background Technology

[0002] HVPE equipment is a key piece of equipment for growing high-quality semiconductor single-crystal thin films, especially in the preparation of wide-bandgap semiconductors such as gallium nitride and zinc oxide.

[0003] Traditional HVPE equipment relies solely on single-point thermocouple temperature measurement, which fails to capture micro-regional temperature differences, easily leading to distorted temperature field assessment and delayed control commands. Furthermore, most of these equipment employ single-zone PID control, neglecting the thermal conduction coupling between the sample zone, transition zone, and exhaust gas zone within the HVPE reaction chamber. This results in cross-zone interference causing temperature field oscillations. Moreover, temperature feedback adjustment easily overlooks the cross-influence of flow field turbulence, deposited byproducts, and pressure fluctuations on the temperature field, leading to decreased control accuracy under complex operating conditions.

[0004] Therefore, it is necessary to propose an intelligent temperature field control method and system based on HVPE equipment to solve the above problems. Summary of the Invention

[0005] The main objective of this invention is to provide an intelligent temperature field control method and system based on HVPE equipment, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for intelligent temperature field control based on HVPE equipment includes the following steps:

[0008] S1: Multi-source heterogeneous data acquisition and spatiotemporal fusion: Data is collected by deploying data acquisition devices in different areas of the HVPE equipment, and then processed through a spatiotemporal fusion model to complete the integration of multi-source heterogeneous data, which provides a precise data foundation for subsequent temperature field control.

[0009] S2: Temperature field balance quantification assessment and distributed collaborative control. The temperature field balance is calculated using the data collected by S1. Through a six-dimensional fuzzy PID controller and consistency algorithm, the temperature field imbalance judgment and the collaborative control of the heater are completed to improve the temperature field uniformity and stability.

[0010] S3: Dynamic compensation of four fields of heat, flow, deposition and pressure. Based on the four-field coupling model, the compensation amount is calculated to control the heater power and carrier gas flow rate, and to complete the dynamic compensation of HVPE equipment to eliminate the interference of multiple factors on the temperature field.

[0011] S4: Dynamic learning optimization based on digital twins. By using LSTM and PPO algorithms and combining digital twin simulation, temperature field trend prediction and control parameter optimization are completed, which is used to realize the autonomous optimization of HVPE equipment control strategy.

[0012] Preferably, step S1 specifically includes the following steps:

[0013] S101: Multi-dimensional acquisition of multi-source heterogeneous data divides the reaction chamber of the HVPE equipment into a sample area, a transition area, and an exhaust gas area. By arranging 25 infrared thermocouples in the sample area, the maximum allowable temperature difference between the 25 infrared thermocouples is 2℃.

[0014] Twenty-four fiber optic temperature sensors are arranged in the transition zone, which is the wall of the reaction chamber.

[0015] A contact thermocouple is installed in the exhaust gas area to collect temperature data.

[0016] Data on overall pressure fluctuations are collected using a laser particle imager, a first CCD camera, and a pressure sensor.

[0017] Data on the morphology and roughness of the deposited surface were acquired using an X-ray diffractometer and a second CCD camera.

[0018] S102: Verification of multi-source heterogeneous data. When the HVPE device starts, the device used in S101 starts synchronously and begins data acquisition. When any measurement value of the device used in S1 changes abruptly, i.e., temperature > ±5℃, flow rate > ±2m / s, pressure > ±10Pa, a triple verification mechanism is triggered: First, two repeated samplings are performed within 100ms. If the standard deviation of the three measurements exceeds twice that of the device, the data point is determined to be invalid and marked as abnormal, and a warning signal is issued at the same time.

[0019] S103: Spatiotemporal fusion of multi-source heterogeneous data. The validated multi-source heterogeneous data is stored on a local SSD and preprocessed using a spatiotemporal fusion model. The formula is:

[0020] ;

[0021] in This is the real-time temperature value obtained by fusing the temperature data collected by S101; The average temperature value collected by the infrared thermocouple; The weighted average value of the temperature field collected by the fiber optic temperature sensor; To collect temperature data for contact thermocouples; The epitaxial layer deposition rate is determined by X-ray diffraction. The scanning pattern is calculated. The difference between the real-time pressure in the reaction chamber of the HVPE equipment and the process set pressure; , , , These are the model weight coefficients, with initial values ​​of 0.4, 0.3, 0.1, 0.15, and 0.05, respectively. .

[0022] Preferably, step S2 specifically includes the following steps:

[0023] S201: The real-time temperature data collected in step S101 by 25 infrared thermocouples Standard deviation of flow field velocity measured by laser particle imager Temperature difference of infrared thermocouples Flow field influencing factors The dynamic temperature field uniformity calculation formula is used to complete the quantitative evaluation of temperature field uniformity. The calculation period is 100ms per sampling period. The formula is:

[0024] ;

[0025] in For temperature field uniformity; here It is one of 25 infrared thermocouples; This represents the real-time average temperature of the sample area. Set the velocity for the flow field; for Standard deviation of flow field velocity measured by a laser particle imager at any given time;

[0026] S202: The sample area, transition area, and exhaust gas area are divided into three independent temperature control zones, and the following devices are deployed in each zone to achieve distributed control:

[0027] Sample area: An RF heater is used to directly heat the substrate tray placed in the sample area via a PWM signal;

[0028] Transition zone: A resistance heater is used, which is installed around the upper middle part of the outer wall of the transition zone;

[0029] Exhaust gas area: A resistance heater is used, which is installed 100mm outside the exhaust gas pipe inlet;

[0030] S203: When S201 is calculated <When the preset threshold is reached, temperature control is achieved through a six-dimensional fuzzy PID controller in coordination with temperature control:

[0031] Input variables include real-time temperature difference Temperature change rate Deposition thickness deviation Pressure deviation It is divided into seven fuzzy subsets: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large. Through 81 core fuzzy rules, the output variable is the PID parameter adjustment coefficient. , , The adjustment range is 50%-200% of the baseline value. The adjusted parameters directly affect the heaters in the sample area, transition area, and exhaust gas area.

[0032] S204: When the temperature difference between adjacent temperature zones exceeds ±5℃, a distributed cooperative control mechanism is triggered via the CAN bus: a consensus algorithm is used, and the formula is as follows:

[0033] ;

[0034] in For the region The current power; , where is the synergy coefficient; middle Indicates the region The adjacent region numbers belong to the set. ; For the region At the next discrete time Heating power; For the region At the current discrete time Heating power; Adjacent areas With the current region At discrete time The power difference will be calculated. The power is then adjusted in the corresponding regions.

[0035] Preferably, in step S202, the temperature control target for the sample area is ±0.5℃; the temperature control target for the transition area is ±1℃; and the control target for the exhaust gas area is ±2℃.

[0036] Preferably, step S3 specifically includes the following steps:

[0037] S301: Establish a coupled compensation model, where the compensation formula for the thermal field-flow field-pressure coupling is:

[0038] ;

[0039] in This is the heating power compensation amount, used to correct the interference of flow field, deposition, and pressure changes on the temperature field; Real-time flow field status measured by laser particle imager; Let be the coupling coefficient, where =0.8 W·s / m; =50W·s / nm; =200W / Pa;

[0040] The formula for coordinated adjustment of the process and thermal field is:

[0041] ;

[0042] in This is the carrier gas flow rate adjustment amount, used to balance the flow field fluctuations caused by thermal field adjustments; =0.1sccm / W, which is the power-to-flow conversion factor; =5sccm / (m / s), which is the flow velocity stability compensation coefficient. Increasing the size increases flow stability; The standard deviation of the flow field velocity measured by the laser particle imager;

[0043] S302: Dynamically supplement execution steps:

[0044] Heating power adjustment, including the sample area: according to Calculate the PWM duty cycle adjustment amount, and output it in real time through the RF heater at a modulation frequency of 100Hz, with a power regulation accuracy of ±0.5W; Transition zone: based on the required cross-zone heat conduction compensation value in the transition zone and... The multiplication factor is used as the power compensation value, and the adjustment is performed by the resistance heater, with a power change rate ≤10W / s; Exhaust gas zone: only when At that time, As an auxiliary adjustment, it is used to prevent drastic fluctuations in exhaust gas temperature and is adjusted by the heater;

[0045] Carrier gas flow rate adjustment is performed using a Brooks 5850S ​​mass flow controller, adjusting the required carrier gas flow rate during HVPE equipment operation according to the formula: [Formula omitted for brevity]

[0046] ;

[0047] in To adjust the target flow rate setting of the rear carrier gas; To adjust the original flow rate setting of the front carrier gas.

[0048] Preferably, when any data fluctuation in S301 exceeds a self-set threshold, a data filtering algorithm is activated to eliminate the interference of high-frequency noise on the compensation calculation. The adjustment range of the heating power of a single temperature zone does not exceed 30% of the current power to prevent current overload; the carrier gas flow rate does not exceed 90% of the range of the mass flow controller after adjustment to avoid gas pressure exceeding the limit.

[0049] Preferably, in S3, when the following combined conditions are detected: , When the preset threshold and surface roughness are greater than 5nm, the following operations are triggered:

[0050] A six-axis robotic arm with a silicon carbide scraper at the end is used to clean the byproducts on the reaction chamber wall at a speed of 5 mm / s along a preset path; at the same time, a 254 nm ultraviolet lamp is turned on and continuously irradiated for 30 seconds to enhance the desorption of adsorbed atoms on the wall through photochemical action, thereby reducing particle redeposition.

[0051] Preferably, step S4 specifically includes the following steps:

[0052] S401: The data collected by S1 is input into the LSTM neural network model, which captures the dynamic changes in the temperature field during the operation of the HVPE device through two layers of 128-dimensional LSTM units. Finally, the output layer predicts the operating parameters of the sample area, transition area and exhaust gas area in the next 10 minutes. When the LSTM neural network model predicts that the temperature field fluctuation exceeds ±1℃, the power of the HVPE device is adjusted through S2, and an early warning is issued at the same time.

[0053] S402: The control strategy of the HVPE equipment is optimized through the PPO reinforcement learning model. Every hour, the PPO reinforcement learning model generates a set of optimization schemes including heating power, carrier gas flow rate, and fuzzy PID parameters based on the current temperature field uniformity, deposition thickness uniformity, and energy consumption. The schemes are first verified by 100 process cycle simulations on the digital twin simulation platform to ensure that there are no temperature overshoots or deposition anomalies. After being encrypted by the industrial host, the schemes are sent to S2 and S3 via the CAN bus for execution of the control strategy.

[0054] An intelligent temperature field control system based on HVPE equipment includes a multi-source heterogeneous data acquisition module, a data processing and analysis module, a temperature field equalization control module, a four-field coupling dynamic compensation module, and a digital twin optimization module. The multi-source heterogeneous data acquisition module acquires temperature data through infrared thermocouples, fiber optic temperature sensors, and contact thermocouples placed in the sample area, transition area, and exhaust gas area of ​​the HVPE equipment; acquires flow field velocity and pressure data through a laser particle imager and a pressure sensor; and acquires deposition thickness and surface morphology data through an X-ray diffractometer and a CCD camera.

[0055] The data processing and analysis module is used to transmit the data collected by the multi-source heterogeneous data acquisition module to the industrial host. The multi-source data is preprocessed through a spatiotemporal fusion model. When the system starts, the acquisition devices used by the multi-source heterogeneous data acquisition module collect data synchronously at a frequency of 10Hz. When the data is abnormal, three repeated sampling verifications are performed. Invalid data is marked and an alarm is triggered. The fused data is used to calculate the temperature field uniformity, drive four-field coupling compensation, and support the training of the digital twin model.

[0056] The temperature field equalization control module evaluates the temperature field uniformity through the equalization degree calculation formula. When the equalization degree is less than the preset threshold, the six-dimensional fuzzy PID controller is activated to adjust the heater power of the sample area, the transition area and the exhaust gas area. When the temperature difference between adjacent temperature areas exceeds ±0.5℃, the consensus algorithm is used to realize distributed collaborative control to maintain the temperature field equalization.

[0057] Preferably, the four-field coupling dynamic compensation module calculates the heating power compensation amount and the carrier gas flow rate adjustment amount according to the heat-flow-deposition-pressure coupling compensation formula, and controls the heater power and mass flow controller of the sample area, transition area and exhaust gas area. When the deposition rate is >1nm / s and the temperature field uniformity is <preset threshold, the robotic arm cleaning and ultraviolet light irradiation are triggered to eliminate the influence of multi-field interference on the temperature field.

[0058] The digital twin optimization module uses an LSTM model to predict temperature field parameters for the next 10 minutes. When the parameters exceed a preset threshold, it sends pre-compensation parameters to the temperature field equalization control module. The PPO reinforcement learning algorithm generates optimization strategies every hour. After verification by digital twin simulation, the model is updated based on the new data to achieve autonomous optimization of the control strategy.

[0059] Compared with the prior art, the present invention provides an intelligent temperature field control method and system based on HVPE equipment, which has the following beneficial effects:

[0060] 1. This intelligent temperature field control method and system based on HVPE equipment, through the arrangement of multiple infrared thermocouples and fiber optic sensors, can accurately sense the three-dimensional temperature field of HVPE equipment, effectively improving the real-time temperature monitoring accuracy. Compared with single-point temperature measurement, the micro-area temperature difference identification capability can be effectively improved, avoiding the hidden problem of epitaxial layer defects caused by local overheating even when the overall temperature meets the standard. At the same time, it can solve the problem of heat conduction lag caused by measuring only the temperature of the sample area. By collecting the flow field velocity distribution and identifying turbulent areas, it can avoid temperature field deflection caused by high-speed airflow. Combined with X-ray diffractometer and CCD camera, it can monitor the deposition rate and surface morphology in real time, solving the problem of local thermal resistance changes caused by the accumulation of deposition by-products.

[0061] 2. This intelligent temperature field control method and system based on HVPE equipment divides the reaction chamber of the HVPE equipment into three temperature control zones: a sample zone, a transition zone, and an exhaust gas zone. Heating devices of varying precision are deployed, and a distributed collaborative control network is constructed. The sample zone uses a high-frequency response radio frequency heater to achieve rapid and precise temperature adjustment in the micro-area. The transition zone uses a resistance heater to buffer cross-zone heat conduction. The exhaust gas zone maintains overall thermal balance. Through quantitative evaluation of the equilibrium, and by integrating temperature uniformity and flow field stability parameters, the temperature field state can be dynamically determined. When the equilibrium is not met, a six-dimensional fuzzy PID controller is used to convert temperature difference, temperature difference rate of change, deposition thickness deviation, and pressure deviation into fuzzy language variables. Through 81 core rules, the proportional, integral, and derivative parameters are dynamically adjusted, effectively solving the problem of insufficient control precision of traditional PID in multi-variable coupled scenarios. Regarding the problem of cross-zone temperature difference exceeding limits, a consensus algorithm is used to dynamically allocate heating power between adjacent areas, forming a rapid response, buffer compensation, and global balance control system, which can effectively suppress temperature field drift caused by heat conduction.

[0062] 3. This intelligent temperature field control method and system based on HVPE equipment constructs a four-field coupling model of heat-flow-deposition-pressure, incorporating flow rate deviation, deposition rate, and pressure fluctuations into the compensation calculation of heating power and carrier gas flow. This enables active identification and dynamic cancellation of multi-physics field interference. When flow field turbulence or deposition anomalies are detected, the system automatically triggers targeted compensation measures: a robotic arm equipped with a silicon carbide scraper cleans byproducts on the reaction chamber wall along a preset path, combined with ultraviolet light irradiation to enhance the desorption of adsorbed atoms on the wall, reducing the impact of particle redeposition on local thermal resistance; the carrier gas flow and heating power are adjusted in real time according to the coupling formula to balance the temperature field fluctuations caused by flow rate changes and the heat consumption changes caused by deposition rate anomalies. Strict hardware safety boundaries are set during the compensation process to avoid excessive adjustment of power and flow leading to equipment overload, ensuring the accuracy and safety of the executed actions.

[0063] 4. The intelligent temperature field control method and system based on HVPE equipment can provide real-time operating condition information for equalization control and coupling compensation through the set multi-source data acquisition module. The temperature field equalization control module generates basic adjustment commands based on the evaluation results. The four-field coupling compensation module can dynamically correct complex disturbances. The digital twin optimization module continuously improves the accuracy of the control strategy through historical data and real-time feedback.

[0064] 5. The intelligent temperature field control method and system based on HVPE equipment can effectively suppress the temperature difference in the micro-region of the sample area through the deployment of high-density temperature sensors and distributed collaborative control, and avoid epitaxial layer defects caused by local overheating or undercooling; the multi-field coupling compensation mechanism can offset the interference of flow field, deposition and other factors in real time, and ensure that the temperature field remains stable under complex working conditions, providing a uniform thermal environment for high-quality crystal growth.

[0065] 6. The intelligent temperature field control method and system based on HVPE equipment can quickly respond through multi-field data fusion analysis and intelligent algorithms when facing harsh conditions such as flow rate fluctuations, pressure changes, and abnormal deposition rates. It can automatically adjust control parameters and execute actions to maintain the temperature field balance within the ideal range, thereby significantly improving the reliability of HPE equipment under extreme process conditions. Attached Figure Description

[0066] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0067] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0068] Example 1:

[0069] like Figure 1 As shown, a smart temperature field control method based on HVPE equipment includes the following steps:

[0070] S1: Multi-source heterogeneous data acquisition and spatiotemporal fusion. Data is collected by deploying data acquisition devices in different areas of the HVPE equipment, and then processed through a spatiotemporal fusion model to integrate multi-source heterogeneous data. This provides a precise data foundation for subsequent temperature field control. The specific steps include the following:

[0071] S101: Multi-dimensional acquisition of multi-source heterogeneous data divides the reaction chamber of the HVPE equipment into a sample area, a transition area, and an exhaust gas area. By arranging 25 infrared thermocouples in the sample area, the maximum allowable temperature difference between the 25 infrared thermocouples is 2℃.

[0072] Twenty-four fiber optic temperature sensors are arranged in the transition zone, which is the wall of the reaction chamber.

[0073] A contact thermocouple is installed in the exhaust gas area to collect temperature data.

[0074] Data on overall pressure fluctuations are collected using a laser particle imager, a first CCD camera, and a pressure sensor.

[0075] Data on the morphology and roughness of the deposited surface were acquired using an X-ray diffractometer and a second CCD camera.

[0076] S102: Verification of multi-source heterogeneous data. When the HVPE device starts, the devices used in S101 start synchronously and begin data acquisition at preset frequencies: infrared thermocouple, fiber optic temperature sensor, contact thermocouple 10Hz, laser particle imager 10Hz, pressure sensor 20Hz, X-ray diffractometer 1 time / minute, first CCD camera and second CCD camera 15fps. When any measurement value of any device used in S1 changes abruptly, i.e., temperature > ±5℃, flow rate > ±2m / s, pressure > ±10Pa, a triple verification mechanism is triggered: firstly, two repeated samplings are performed within 100ms. If the standard deviation of the three measurements exceeds twice that of the device, the data point is determined to be invalid and marked as abnormal, and a warning signal is issued simultaneously.

[0077] S103: Spatiotemporal fusion of multi-source heterogeneous data. The validated multi-source heterogeneous data is stored on a local SSD and preprocessed using a spatiotemporal fusion model. The formula is:

[0078] ;

[0079] in This is the real-time temperature value obtained by fusing the temperature data collected by S101; The average temperature value collected by the infrared thermocouple; The weighted average value of the temperature field collected by the fiber optic temperature sensor; To collect temperature data for contact thermocouples; The epitaxial layer deposition rate is determined by X-ray diffraction. The scanning pattern is calculated. The difference between the real-time pressure in the reaction chamber of the HVPE equipment and the process set pressure; , , , These are the model weight coefficients, with initial values ​​of 0.4, 0.3, 0.1, 0.15, and 0.05, respectively. ;

[0080] The particle swarm optimization algorithm is used to adjust the weight coefficients in real time. The objective function is to minimize the root mean square error between the fused temperature and the measured temperature. The parameters of the particle swarm optimization algorithm are set as follows: the number of particles is fixed at 30, the number of iterations is 50, and the inertia weight is... The acceleration factor decreases linearly from 0.8 to 0.4. The calibration cycle is once every 10 minutes. During the calibration process, data acquisition and basic control functions continue to operate, and the optimized weight parameters take effect immediately after the iteration ends.

[0081] S2: Temperature field uniformity quantification assessment and distributed collaborative control. Using the data collected by S1, the temperature field uniformity is calculated. A six-dimensional fuzzy PID controller and consistency algorithm are used to determine temperature field imbalance and perform collaborative control of the heater, thereby improving temperature field uniformity and stability. Specific operation steps include:

[0082] S201: The real-time temperature data collected in step S101 by 25 infrared thermocouples Standard deviation of flow field velocity measured by laser particle imager Temperature difference of infrared thermocouples Flow field influencing factors The dynamic temperature field uniformity calculation formula is used to complete the quantitative evaluation of temperature field uniformity. The calculation period is 100ms per sampling period. The formula is:

[0083] ;

[0084] in For temperature field uniformity; here It is one of 25 infrared thermocouples; This represents the real-time average temperature of the sample area. Set the velocity for the flow field; for Standard deviation of flow field velocity measured by a laser particle imager at any given time;

[0085] S202: The sample area, transition area, and exhaust gas area are divided into three independent temperature control zones, and the following devices are deployed in each zone to achieve distributed control:

[0086] Sample area: The substrate tray placed in the sample area is directly heated by the MKS247B RF heater (power range 0-500W, pulse width modulation frequency 100Hz, power adjustment accuracy ±0.5W) through PWM signal (duty cycle 0-100%).

[0087] Transition zone: A Watlow F4T resistance heater (power range 0-1000W, equipped with a thyristor power regulator, power adjustment resolution 1W) is used and installed around the upper middle part of the outer wall of the transition zone.

[0088] Exhaust gas area: An Omega CN7600 resistance heater (power range 0-500W, control accuracy ±2℃) is used and installed 100mm outside the exhaust gas pipe inlet.

[0089] The temperature control target for the sample area is ±0.5℃; the temperature control target for the transition area is ±1℃; and the control target for the exhaust gas area is ±2℃.

[0090] S203: When S201 is calculated <When the preset threshold is reached, temperature control is achieved through a six-dimensional fuzzy PID controller in coordination with temperature control:

[0091] Input variables include real-time temperature difference Temperature change rate Deposition thickness deviation Pressure deviation It is divided into seven fuzzy subsets: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large. Through 81 core fuzzy rules, the output variable is the PID parameter adjustment coefficient. , , The adjustment range is 50%-200% of the baseline value. The adjusted parameters directly affect the heaters in the sample area, transition area, and exhaust gas area. The six-dimensional fuzzy PID controller has four key input variables, namely the real-time temperature difference. Temperature change rate Deposition thickness deviation Pressure deviation Real-time temperature difference Temperature change rate Deposition thickness deviation Pressure deviation The fuzzy rule table contains 7×7×7×7=2401 rules, which are simplified to 81 core rules through orthogonal experiments in practical applications. Set the process temperature for the sample area; The target thickness for thin film deposition; The deposition thickness is measured in real time using an online ellipsometry. Set the pressure for the reaction chamber process; This is the real-time measurement value from the pressure sensor. This is the adjustment amount for the proportional coefficient; This is the adjustment amount for the integral coefficient; For the adjustment of the differential coefficients, the core rule is as follows:

[0092] when For a large negative For the sake of righteousness, For a large negative When the value is positive, the output PID parameter adjustment coefficient is: Increase by 40% Reduce by 20% The 15% increase rule is used to rapidly increase the temperature while avoiding overshoot due to excessively high deposition rates.

[0093] like Zero Zero For positive small, When it is a negative hour, then Reduce by 10% An additional 5% is added to fine-tune the heating power, balancing the effects of a slight increase in deposition thickness and a slight decrease in pressure on the temperature field.

[0094] The 81 rules are based on multi-source data collected by S1, combined with the simulation results of the S3 four-field coupling model, and the minimum complete set formed after eliminating redundant and low-impact combinations, to ensure coverage of more than 95% of typical working conditions.

[0095] S204: When the temperature difference between adjacent temperature zones exceeds ±5℃, a distributed cooperative control mechanism is triggered via the CAN bus: a consensus algorithm is used, and the formula is as follows:

[0096] ;

[0097] in For the region The current power; , where is the synergy coefficient; middle Indicates the region The adjacent region numbers belong to the set. ; For the region At the next discrete time Heating power; For the region At the current discrete time Heating power; Adjacent areas With the current region At discrete time The power difference is calculated, with each collaborative calculation cycle lasting 500ms. The power adjustment step size does not exceed 5% of the current power to avoid temperature overshoot caused by sudden power changes. The power is adjusted in the corresponding areas. The sample area controller uses high-frequency PWM modulation to achieve fast response, while the controllers in the transition area and exhaust gas area use smooth power adjustment to prevent thermal stress concentration. When the actual temperature of any temperature area deviates from the set value by more than twice the control target (i.e., sample area > ±1℃, transition area > ±2℃, exhaust gas area > ±4℃), hardware redundancy protection is automatically triggered: the power supply to the heater in the corresponding area is cut off, and the liquid nitrogen cooling valve is activated to limit the temperature fluctuation within a safe range: sample area ≤ ±1.5℃, transition area ≤ ±3℃, exhaust gas area ≤ ±6℃.

[0098] S3: Dynamic compensation based on the coupling of four fields: heat, flow, deposition, and pressure. This involves calculating compensation amounts based on a four-field coupling model to control heater power and carrier gas flow rate, achieving dynamic compensation for the HVPE equipment and eliminating interference from multiple factors on the temperature field. Specific operational steps include:

[0099] S301: Establish a coupled compensation model, where the compensation formula for the thermal field-flow field-pressure coupling is:

[0100] ;

[0101] in This is the heating power compensation amount, used to correct the interference of flow field, deposition, and pressure changes on the temperature field; Real-time flow field status measured by laser particle imager; Let be the coupling coefficient, where =0.8 W·s / m; =50W·s / nm; =200W / Pa;

[0102] The formula for coordinated adjustment of the process and thermal field is:

[0103] ;

[0104] in This is the carrier gas flow rate adjustment amount, used to balance the flow field fluctuations caused by thermal field adjustments; =0.1sccm / W, which is the power-to-flow conversion factor; =5sccm / (m / s), which is the flow velocity stability compensation coefficient. Increasing the size increases flow stability; The standard deviation of the flow field velocity measured by the laser particle imager;

[0105] When any data fluctuation in S301 exceeds the self-set threshold, the data filtering algorithm is activated to eliminate the interference of high-frequency noise on the compensation calculation. The adjustment range of the heating power of a single temperature zone shall not exceed 30% of the current power to prevent current overload; the carrier gas flow rate shall not exceed 90% of the range of the mass flow controller after adjustment to avoid gas pressure exceeding the limit.

[0106] S302: Dynamically supplement execution steps:

[0107] Heating power adjustment, including the sample area: according to Calculate the PWM duty cycle adjustment amount and output it in real time at a modulation frequency of 100Hz through the MKS247B RF heater, with a power regulation accuracy of ±0.5W; Transition zone: based on the required cross-zone heat conduction compensation value in the transition zone and... The multiplication factor serves as the power compensation value, adjusted via the Watlow F4T resistance heater, with a power change rate ≤10W / s; Exhaust gas zone: only when At that time, As an auxiliary adjustment, it is used to prevent drastic fluctuations in exhaust gas temperature and is adjusted by the OmegaCN7600 heater;

[0108] Carrier gas flow rate adjustment is performed using a Brooks 5850S ​​mass flow controller (adjustment step size 0.1 sccm, response time < 500 ms). The carrier gas flow rate for the required gas during HVPE equipment operation is adjusted according to the formula to maintain process stability. The formula is:

[0109] ;

[0110] in To adjust the target flow rate setting of the rear carrier gas; To adjust the original flow rate setting of the front carrier gas

[0111] When the following joint conditions are detected: , When the preset threshold and surface roughness are greater than 5nm, the following operations are triggered:

[0112] A six-axis robotic arm with a silicon carbide scraper at the end is used to clean the byproducts on the reaction chamber wall at a speed of 5 mm / s along a preset path; at the same time, a 254 nm ultraviolet lamp with a power of 500 mW / cm² is turned on, covering 80% of the reaction chamber wall area and continuously irradiating for 30 seconds. The photochemical effect enhances the desorption of adsorbed atoms on the wall surface, which is used to reduce particle redeposition.

[0113] S4: Dynamic learning optimization based on digital twins, using LSTM and PPO algorithms combined with digital twin simulation, to complete temperature field trend prediction and control parameter optimization, thereby realizing autonomous optimization of HVPE equipment control strategy. Specific steps include:

[0114] S401: The data collected by S1 is input into the LSTM neural network model, which captures the dynamic changes in the temperature field during the operation of the HVPE device through two layers of 128-dimensional LSTM units. Finally, the output layer predicts the operating parameters of the sample area, transition area and exhaust gas area in the next 10 minutes. When the LSTM neural network model predicts that the temperature field fluctuation exceeds ±1℃, the power of the HVPE device is adjusted through S2, and an early warning is issued at the same time.

[0115] S402: The control strategy of HVPE equipment is optimized through the PPO reinforcement learning model. Every hour, the PPO reinforcement learning model generates a set of optimization schemes including heating power, carrier gas flow rate and fuzzy PID parameters based on the current temperature field uniformity, deposition thickness uniformity and energy consumption. The schemes are first verified by 100 process cycle simulations on the digital twin simulation platform to ensure that there are no temperature overshoots or deposition abnormalities. After being encrypted by the industrial host, the schemes are sent to S2 and S3 via the CAN bus for execution of the control strategy.

[0116] If the parameters generated by the PPO reinforcement learning model for three consecutive times cause a decrease in the temperature field uniformity of the HVPE device, the model will automatically revert to the three valid parameters from the previous three times and mark the model training as abnormal, pending review and confirmation by engineers.

[0117] Example 2:

[0118] An intelligent temperature field control system based on HVPE equipment includes a multi-source heterogeneous data acquisition module, a data processing and analysis module, a temperature field equalization control module, a four-field coupling dynamic compensation module, and a digital twin optimization module. The multi-source heterogeneous data acquisition module acquires temperature data through infrared thermocouples, fiber optic temperature sensors, and contact thermocouples placed in the sample area, transition area, and exhaust gas area of ​​the HVPE equipment; a laser particle imager and a pressure sensor acquire flow field velocity and pressure data; and an X-ray diffractometer and a CCD camera acquire deposition thickness and surface morphology data.

[0119] The data processing and analysis module is used to transmit the data collected by the multi-source heterogeneous data acquisition module to the industrial host. The multi-source data is preprocessed through a spatiotemporal fusion model. When the system starts, the acquisition devices used by the multi-source heterogeneous data acquisition module collect data synchronously at a frequency of 10Hz. When the data is abnormal, three repeated sampling verifications are performed. Invalid data is marked and an alarm is triggered. The fused data is used to calculate the temperature field uniformity, drive four-field coupling compensation, and support the training of the digital twin model.

[0120] The temperature field equalization control module evaluates the temperature field uniformity through the equalization degree calculation formula. When the equalization degree is less than the preset threshold, the six-dimensional fuzzy PID controller is activated to adjust the heater power of the sample area, transition area and exhaust gas area. When the temperature difference between adjacent temperature areas exceeds ±0.5℃, the consensus algorithm is used to realize distributed collaborative control to maintain the temperature field equalization.

[0121] The four-field coupling dynamic compensation module calculates the heating power compensation amount and carrier gas flow rate adjustment amount based on the heat-flow-deposition-pressure coupling compensation formula, and controls the heater power and mass flow controller of the sample area, transition area and exhaust gas area. When the deposition rate is >1nm / s and the temperature field uniformity is <preset threshold, it triggers robotic arm cleaning and ultraviolet light irradiation to eliminate the influence of multi-field interference on the temperature field.

[0122] The digital twin optimization module uses an LSTM model to predict temperature field parameters for the next 10 minutes. When the parameters exceed a preset threshold, pre-compensation parameters are sent to the temperature field equalization control module. The PPO reinforcement learning algorithm generates optimization strategies every hour. After verification by digital twin simulation, the model is updated based on the new data to achieve autonomous optimization of the control strategy.

[0123] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent temperature field regulation method based on an HVPE device, characterized in that: Comprise the following operation steps: S1: Multisource heterogeneous data multi-dimensional collection and space-time fusion, data collection is carried out through the deployment of data collection equipment in different areas of HVPE equipment, and the multisource heterogeneous data integration is completed through the space-time fusion model processing, which is used to provide accurate data basis for subsequent temperature field control; S2: Temperature field balance quantitative evaluation and distributed collaborative control, the data collected by S1 is used to calculate the temperature field balance, and the six-dimensional fuzzy PID controller and consistency algorithm are used to complete the temperature field imbalance judgment and collaborative control of the heater, which is used to improve the uniformity and stability of the temperature field; S3: Four-field coupling dynamic compensation, based on the four-field coupling model, the compensation amount is calculated, which is used to control the heater power and carrier gas flow, and the dynamic compensation of the HVPE equipment is completed, which is used to eliminate the interference of multiple factors on the temperature field, comprising the following operation steps: S301: Establish a coupling compensation model, wherein the compensation formula of thermal field-flow field-pressure coupling is: ； wherein is a heating power compensation amount for correcting the interference of the flow field, deposition, and pressure change on the temperature field; is a real-time state of the flow field measured by the laser particle imager; is a coupling coefficient, wherein = 0.8 W·s / m; = 50 W·s / nm; = 200 W / Pa; The formula for the coordinated adjustment of flow-thermal field is: ； wherein is the carrier gas flow adjustment for balancing the flow field fluctuation caused by thermal field adjustment; = 0.1 sccm / W, is the power-flow conversion factor; = 5 sccm / (m / s), is the flow rate stability compensation factor, when increases, the flow stability increases; is the standard deviation of the flow field velocity measured by the laser particle imager; S302: Dynamic compensation execution step: Heating power adjustment, including sample area: according to The PWM duty cycle adjustment amount is calculated and output in real time by the radio frequency heater at a modulation frequency of 100 Hz, and the power adjustment accuracy is ±0.5 W; Transition zone: the adjustment is performed by the resistive heater as a power compensation value, with a power rate of change ≤ 10 W / s; tail gas zone: only when the adjustment is performed by the heater as an auxiliary adjustment value, to prevent sharp fluctuations in tail gas temperature. ​​ Carrier gas flow adjustment, the required carrier gas flow of the HVPE equipment during operation is adjusted according to the formula through the Brooks5850S mass flow controller, which is used to maintain the stability of the flow, and the formula is: ； wherein a target flow rate set value of the adjusted carrier gas; an original flow rate set value of the adjusted carrier gas; S4: Dynamic learning optimization based on digital twinning, with the help of LSTM and PPO algorithm, combined with digital twinning simulation, the temperature field trend prediction and control parameter optimization are completed, which is used to realize the autonomous optimization of the HVPE equipment control strategy. 2.The smart temperature field regulation method based on an HVPE device according to claim 1, wherein: The S1 specifically comprises the following operation steps: S101: Multidimensional collection of multisource heterogeneous data, the reaction chamber of the HVPE equipment is divided into a sample area, a transition area and a tail gas area, 25 infrared thermocouples are arranged in the sample area, and the maximum temperature difference allowed value between the 25 infrared thermocouples is 2℃; 24 optical fiber temperature sensors are arranged in the transition area, wherein the transition area is the reaction chamber wall of the reaction chamber; One contact thermocouple is arranged in the tail gas area for temperature data collection; The overall pressure fluctuation data is collected by a laser particle imager, a first CCD camera and a pressure sensor; The data of deposition surface topography and roughness is collected by an X-ray diffractometer and a second CCD camera; S102: Verification of multisource heterogeneous data, when the HVPE equipment starts, the devices used in S101 start synchronously to start data collection;When the measured value of any S1 used device jumps, i.e. temperature>±5℃, flow rate>±2m / s, pressure>±10Pa, a three-way verification mechanism is triggered: first, repeat sampling within 100ms twice, if the standard deviation of three measurements exceeds twice the value of the device, the data point is invalid and marked as abnormal, and a warning signal is sent; S103: Space-time fusion of multisource heterogeneous data, the verified multisource heterogeneous data is stored in the local SSD, and preprocessed through the space-time fusion model, the formula is: ； wherein is the real-time temperature value after fusion of the temperature data collected in S101; is the average temperature value collected by the infrared thermocouple; is the weighted average value of the temperature field collected by the optical fiber temperature sensor; is the temperature data collected by the contact thermocouple; is the epitaxial layer deposition rate, calculated by the X-ray diffractometer through the scanning mode; is the difference between the real-time pressure in the reaction chamber of the HVPE device and the process set pressure; , , , is the model weight coefficient, the initial values are 0.4, 0.3, 0.1, 0.15, 0.05, wherein .

3. The method of claim 1, wherein the method comprises: The S2 specifically comprises the following operation steps: S201: the real-time temperature data collected by the 25 infrared thermocouples in step S101 ; standard deviation of flow field velocity measured by laser particle imager ; temperature difference of infrared thermocouples ; flow field influence factor , the dynamic temperature field uniformity calculation formula is completed to quantify the evaluation of temperature field uniformity, and the calculation period is 100 ms for each sampling period, and the formula is: ； wherein is the temperature field uniformity; here is one of 25 infrared thermocouples; is the real-time average temperature of the sample area; is the flow field self-set velocity; is is the standard deviation of the flow field velocity measured by the laser particle imager at that moment; S202: Divide the sample area, the transition area and the tail gas area into three independent temperature control areas, and deploy the following devices to realize distributed control: Sample zone: using radio frequency heater, through PWM signal directly heating the substrate tray placed in the sample zone; Transition zone: using resistance heater, which is installed in the middle and upper part of the outer wall of the transition zone; Tail gas zone: using resistance heater, which is installed 100mm outside the inlet of the tail gas pipeline; S203: When the S201 calculated When the preset threshold is reached, the temperature control is performed by the six-dimensional fuzzy PID controller. Input variables include real-time temperature difference , temperature change rate , deposition thickness deviation , pressure deviation , which is divided into seven fuzzy subsets of negative large, negative medium, negative small, zero, positive small, positive medium and positive large, through 81 core fuzzy rules, and the output variable is the PID parameter adjustment coefficient 、 、 , and the adjustment range is 50%-200% of the reference value. The adjusted parameters directly act on the heaters of the sample area, the transition area and the tail gas area. S204: When the temperature difference between adjacent temperature zones exceeds ±5℃, trigger the distributed collaborative control mechanism through CAN bus: use the consensus algorithm, the formula is: ； wherein is the current power of the region ; is the coordination coefficient ; is the adjacent region number of the region , belonging to the set ; is the heating power of the region at the next discrete time ; is the heating power of the region at the current discrete time ; is the power difference value of the adjacent region and the current region at the discrete time , and the calculated is issued to the corresponding region for power adjustment.

4. The method of claim 3, wherein the method further comprises: In the S202, the temperature control target of the sample zone is ±0.5℃; the temperature control target of the transition zone is ±1℃; the control target of the tail gas zone is ±2℃.

5. The method of claim 1, wherein the method is based on an HVPE apparatus. When any data fluctuation in the S301 exceeds the self-set threshold, start the data filtering algorithm to eliminate the interference of high-frequency noise on compensation calculation, the adjustment range of the heating power of a single temperature zone does not exceed 30% of the current power, which is used to prevent current overload; after adjusting the carrier gas flow, it does not exceed 90% of the range of the mass flow controller, which is used to avoid gas path pressure overrun.

6. The method of claim 1, wherein the method is based on an HVPE apparatus. In the S3, when the following combined conditions are detected: , When the surface roughness < preset threshold, surface roughness > 5 nm, the following operations are triggered: Use the six-axis mechanical arm with a carbonized silicon scraper at the end to clean the reaction chamber wall surface at a speed of 5mm / s along the preset path; simultaneously turn on the 254nm ultraviolet light lamp, continuously irradiate for 30s, enhance the wall adsorption atom desorption through photochemical action, which is used to reduce particle redeposition.

7. The method of claim 1, wherein the method is based on an HVPE apparatus. The S4 specifically includes the following operation steps: S401: input the data collected in S1 into the LSTM neural network model, capture the temperature field dynamic change rule of the HVPE equipment during operation through 2 layers of 128-dimensional LSTM units; finally predict the working parameters of the sample zone, transition zone and tail gas zone in the next 10 minutes in the output layer, when the temperature field fluctuation predicted by the LSTM neural network model exceeds ±1℃, adjust the power of the HVPE equipment through S2, and at the same time, give a warning; S402: optimize the control strategy of the HVPE equipment through the PPO reinforcement learning model, every hour, the PPO reinforcement learning model generates an optimization scheme containing heating power, carrier gas flow and fuzzy PID parameters according to the current temperature field balance degree, deposition thickness uniformity and energy consumption, the scheme is preferentially verified through 100 process cycle simulations in the digital twin simulation platform, which is used to ensure that there is no temperature overshoot and deposition abnormality, and after encryption processing by the industrial host, it is sent to S2 and S3 for execution of the control strategy.

8. An intelligent temperature field regulation system based on an HVPE device, which adopts an intelligent temperature field regulation method based on an HVPE device according to any one of claims 1-7, comprising a multi-source heterogeneous data acquisition module, a data processing and analysis module, a temperature field balance control module, a four-field coupling dynamic compensation module, and a digital twin optimization module, characterized in that: The multi-source heterogeneous data acquisition module collects temperature data through infrared thermocouples, optical fiber temperature sensors and contact thermocouples placed in the sample zone, transition zone and tail gas zone of the HVPE equipment; collects flow field speed and pressure data through laser particle imagers and pressure sensors; collects deposition thickness and surface morphology data through X-ray diffractometers and CCD cameras; The data processing and analysis module is used for transmitting the data collected by the multi-source heterogeneous data collection module to an industrial host, pre-processing the multi-source data through a space-time fusion model, when the system starts, the collection equipment used by the multi-source heterogeneous data collection module synchronously collects at a frequency of 10 Hz; when the data is abnormal, three repeated sampling verifications are performed, invalid data is marked and an alarm is given, and the fused data is used for calculating the temperature field uniformity, driving the four-field coupling compensation, and supporting the training of the digital twin model; The temperature field uniformity control module evaluates the temperature field uniformity through a uniformity calculation formula, when the uniformity is less than a preset threshold, a six-dimensional fuzzy PID controller is started to adjust the heater power of the sample area, the transition area and the tail gas area; when the temperature difference between adjacent temperature areas is more than ±0.5℃, a consistency algorithm is used to realize distributed collaborative control for maintaining the temperature field uniformity.

9. The intelligent temperature field regulation system based on an HVPE apparatus according to claim 8, characterized in that: The four-field coupling dynamic compensation module calculates the heating power compensation amount and the carrier gas flow adjustment amount according to a heat-flow-deposition-pressure coupling compensation formula, controls the heater power and the mass flow controller of the sample area, the transition area and the tail gas area, when the deposition rate is more than 1 nm / s and the temperature field uniformity is less than a preset threshold, triggers the mechanical arm cleaning and ultraviolet irradiation to eliminate the influence of multi-field interference on the temperature field; The digital twin optimization module uses an LSTM model to predict the temperature field parameters in the next 10 minutes, sends the pre-compensation parameters to the temperature field uniformity control module when the parameters exceed the preset threshold, and a PPO reinforcement learning algorithm generates an optimization strategy every hour, the model is updated according to new data after digital twin simulation verification, and is used to realize autonomous optimization of the control strategy.

Images