A control method for preparing alumina crystals using microwave plasma

Abstract

The application provides a control method for preparing an alumina crystal by using a microwave plasma, and specifically comprises the following steps: step one, collecting real-time spectral data of a plasma glow region by a multi-spectral sensor array integrated in a microwave plasma reaction chamber; synchronously collecting temperature distribution data of a reaction chamber wall surface and tail gas component data; step two, pre-processing the real-time spectral data to extract a characteristic spectral intensity vector; step three, calculating a gas phase stoichiometric ratio balance index, an electron temperature and an electron density in the current reaction chamber based on the characteristic spectral intensity vector; step four, adjusting microwave input power and reaction gas flow by using an adaptive PID controller to generate a power adjustment instruction and a flow adjustment instruction; and step five, driving a microwave source to adjust output power by using the power adjustment instruction.

Description

Technical Field

[0001] This invention relates to the field of alumina crystal preparation, and more particularly to a controlled method for preparing alumina crystals using microwave plasma. Background Technology

[0002] Alumina crystals possess excellent optical properties, mechanical strength, and chemical stability, and are widely used in optical windows, special components, semiconductor substrates, and other fields. Current technologies primarily employ microwave plasma methods to prepare alumina crystals. Microwave plasma methods have become one of the mainstream processes for preparing high-quality alumina crystals due to their controllable reaction temperature, high degree of ionization, and uniform reaction atmosphere.

[0003] However, most existing microwave plasma processes for preparing alumina crystals employ open-loop control, relying solely on manual or simple fixed-value adjustments to operating parameters such as microwave power and reactant gas flow rate. This fails to allow for real-time monitoring of changes in the plasma's internal state. Key parameters such as plasma electron temperature, electron density, and gas phase composition ratios fluctuate continuously throughout the reaction, directly causing instability in alumina precursor formation efficiency. This, in turn, leads to fluctuating crystal growth rates and inconsistent crystal quality. Summary of the Invention

[0004] The main objective of this invention is to provide a control method for preparing alumina crystals using microwave plasma, which introduces closed-loop control, enabling more precise control of the preparation process and resulting in higher quality crystals.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a method for controlling the preparation of alumina crystals using microwave plasma, specifically including the following steps: Step 1: Acquire real-time spectral data of the plasma glow region using a multispectral sensor array integrated into the microwave plasma reaction chamber. Real-time spectral data It contains characteristic emission spectral information of aluminum atoms, oxygen atoms, and aluminum oxide molecules, among which, For wavelength, For time; synchronously acquire temperature distribution data of the reaction chamber wall. Exhaust gas composition data , These are the axial position coordinates; Step 2: Process the real-time spectral data Preprocessing is performed to extract the characteristic spectral intensity vector.

[0006] Step 3: Based on the characteristic spectral intensity vector Calculate the gas phase stoichiometric equilibrium index in the current reaction chamber. Electronic temperature and electron density ; Step 4: Based on the calculations obtained in Step 3 Combined with the preset target crystal growth rate and current chamber temperature Microwave input power is adjusted by an adaptive PID controller. and the flow rate of the reacting gas Generate power regulation command and flow regulation commands ; For all at the current moment The average value; Step 5: Utilize the power adjustment command The microwave source is driven to adjust its output power, while the flow regulation command is used simultaneously. The mass flow controller adjusts the mixing ratio of oxygen and inert carrier gas.

[0007] Preferably, the method further includes the following steps: Step Six, Based on , , , , A crystal quality prediction model is constructed using historical crystal growth data to evaluate the crystal quality index in real time. and adjust the quality feedback signal Feedback is sent to step four to form a closed-loop control of crystal quality; Step 7: Establish a multi-objective optimization function to collaboratively optimize crystal growth rate, energy consumption, and gas utilization rate, and generate the optimal control strategy.

[0008] Preferably, step two specifically includes the following sub-steps: Step 21: Use a bandpass filter circuit to filter out the... The frequency in the middle is higher than the preset cutoff frequency. The noise components are removed to obtain the denoised spectral signal. ; Step 22, the output of step 21 Peak finding was performed to locate the center wavelengths of the characteristic spectral lines of aluminum atoms at two different upper energy levels. , Central wavelength of oxygen atom characteristic spectral lines Central wavelength of characteristic spectral lines of alumina molecules and the center wavelength of the characteristic spectral lines of aluminum single ionization ions Among them, two aluminum atom spectral lines satisfy , , These are the corresponding upper energy levels; Step 23, based on the obtained , , , and Within the preset integration window width Integrating the spectral intensity, the integrated intensity of the two aluminum atom characteristic spectral lines was calculated. , Intensity of characteristic spectral lines of oxygen atoms Intensity of characteristic spectral lines of alumina molecules and the intensity of characteristic spectral lines of aluminum single ionization ions , will get , , , and The characteristic spectral intensity vector is obtained by combining them. , It is expressed as follows: (1) Preferably, step three specifically includes the following steps: Step 31: Calculate the basic stoichiometric equilibrium index. The formula is as follows: , (2) in, and The preset proportional weighting coefficient is used to characterize the relative concentration relationship between aluminum and oxygen elements and the formation efficiency of alumina precursor. Step 32: Calculate the plasma electron temperature using the modified Boltzmann slope method. The calculation formula is as follows: , (3) in, Boltzmann's constant, , These are the statistical weights of the energy levels on the two spectral lines, respectively. , The spontaneous emission transition probabilities of the two spectral lines can be obtained by looking up tables in the NIST atomic spectroscopy database; Step 33: Utilize the aforementioned electronic temperature , combined and Electron density was calculated by modifying the Saha ionization equation. The calculation formula is as follows: , (4) in, The rest mass of the electron. It is Planck's constant; Let be the partition function of aluminum atoms. The partition function of aluminum monoionized ions can be obtained by looking up tables in atomic spectroscopy databases; The degeneracy corresponding to the characteristic spectral lines of aluminum atoms. The degeneracy corresponding to the characteristic spectral lines of aluminum single ion can be obtained by looking up a table in the NIST atomic spectroscopy database; The central wavelength of the characteristic spectral lines of aluminum atoms. The center wavelength of the characteristic spectral line of aluminum single ion is determined by the spectral peak-finding process in step 22; The spontaneous emission transition probabilities of the characteristic spectral lines of aluminum single ionization ions can be obtained by looking up tables in the NIST atomic spectroscopy database. The integral intensity of the characteristic spectral line of aluminum single ionization ions at time t; This is the first ionization energy of an aluminum atom; The upper energy level corresponding to the characteristic spectral line of aluminum monoionization ions can be obtained by looking up a table in the NIST atomic spectroscopy database; Step 34: Set the electron temperature and the electron density As a correction parameter, for and Dynamic adjustments are made to the basic stoichiometric balance index. After correction, the final gas-phase stoichiometric equilibrium index is obtained. The calculation formula is as follows: , (5) in, and This is a correction function based on changes in electron temperature and density.

[0009] Preferably, step four specifically includes the following steps: Step 41: Adjust the gas phase stoichiometric equilibrium index. The ideal stoichiometric ratio setpoint for the current cycle By comparison, the deviation value is obtained. ; Step 42, based on the obtained and Update the proportional coefficient of the adaptive PID controller Integral coefficient and differential coefficients The formula is as follows: , (6) in, , , These are the three initial parameters for the PID controller. , For the preset ideal plasma state parameters, , , Adjust the sensitivity coefficient for the parameter; Step 43: The deviation value obtained in step 41 Perform proportional, integral, and derivative calculations to obtain the basic control quantity. : ; (7) Step 44: Based on the aforementioned basic control quantity And combined with the current average temperature of the chamber With target temperature The difference Calculate the final power regulation command : , (8) in, This is the temperature compensation coefficient. This is the temperature damping coefficient; For all at time t The average value; Step 45, from Obtain the residual oxygen concentration at the outlet of the reaction chamber. The obtained With preset safety threshold If a comparison is made, Then an initial flow regulation command is generated. : , (9) in, For flow rate adjustment sensitivity, This is the concentration change rate compensation coefficient; when ,but ; Step 46, to Amplitude limiting is performed to obtain the final flow regulation command. : , in, It is a saturation function. , Adjust the upper and lower limits for flow rate.

[0010] Preferably, step five specifically includes the following steps: Step 51: The power adjustment command generated in step 44... The input is sent to the microwave power source controller, which, according to... Adjusting the anode current of the magnetron changes the microwave input power. , (10) in, Let be the microwave power at time t; Step 52: Apply the flow adjustment command generated in step 46. The input is sent to the mass flow controller, which, according to... Adjust the oxygen flow rate and the inert carrier gas flow rate separately: , (11) in, Let be the oxygen flow rate at time t. Let be the inert gas flow rate at time t. The limiting function restricts the oxygen flow rate to a certain value. and Within a certain range, the flow rate of the inert gas is limited to... Within the specified range, the inert carrier gas is argon or helium; Step 53: Constructing plasma state stability criteria A comprehensive evaluation of the plasma state after power and flow rate regulation was conducted. , (12) in, , , For the stability criterion, the weighting coefficients are... ; Step 54, if ,in If the preset stability threshold is not met, the plasma state is determined to be unstable, and the process returns to step one to re-acquire data and perform adjustments; if... If the plasma state is stable, proceed to step six to predict the crystal quality.

[0011] Preferably, step six specifically includes the following steps: Step 61, based on and historical crystal quality data Build the training dataset: , (13) in, Including X-ray diffraction full width at half maximum (FWHM), surface roughness, and optical transmittance. Indicates the first Historical crystal quality data; Step 62: Build a crystal quality prediction model based on a long short-term memory network: , (14) In the formula, It is the crystal mass at time t predicted by the crystal quality prediction model. For network parameters, For model residuals; Step 63: Utilize the training dataset The mean squared error loss function and Adam optimizer are used to iteratively train the long short-term memory network through backpropagation algorithm to minimize the error between the predicted crystal quality index and historical crystal quality detection data, until the loss function converges or the preset maximum number of iterations is reached, thus obtaining the trained network parameters. ; Step 64: The crystal quality prediction model obtained using the pre-trained crystal quality prediction model... With the preset quality target By comparison, the crystal quality deviation at time t is calculated: , (15) when At that time, a feedback signal is generated. , This is the quality deviation threshold. For feedback coefficients; Step 65: Using the mass-to-stoichiometric ratio conversion factor The quality feedback signal is mapped to the stoichiometric ratio target correction amount. : ; (16) Step 66: Update the ideal stoichiometric ratio setpoint for the next control cycle: , (17) The updated The ideal stoichiometric ratio setpoint of the current cycle is used as the basis for step 41 of the next control cycle to form the outer slow closed-loop control of crystal quality.

[0012] Preferably, step seven specifically includes the following steps: Step 71: Construct a multi-objective optimization function Taking into account three optimization objectives: crystal growth rate, energy efficiency, and gas utilization rate: (18) in, This represents the actual crystal growth rate. Actual energy consumption For maximum allowable energy consumption, This represents the cumulative gas consumption. For the maximum permissible gas consumption, , , These are the weighting coefficients for growth rate, energy efficiency, and gas utilization rate, respectively. ; Step 72: Use the crystal quality prediction model trained in steps 61 to 63 as a constraint condition. The constraint condition requires that the optimized control parameters must guarantee... The particle swarm optimization algorithm is used to solve the problem. Minimize the optimal combination of control parameters The number of particles and the number of iterations in the particle swarm optimization algorithm are predetermined; Step 73: Combine the optimal control parameters As a reference setting value for steps four and five, the reference setting value takes effect in the next control cycle.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This control method enables more accurate control of microwave plasma during the preparation of alumina crystals, thereby improving the quality of alumina crystal preparation. Attached Figure Description

[0014] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0015] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0016] A method for controlling the preparation of alumina crystals using microwave plasma specifically includes the following steps: Step 1: Acquire real-time spectral data of the plasma glow region using a multispectral sensor array integrated into the microwave plasma reaction chamber. Real-time spectral data It contains characteristic emission spectral information of aluminum atoms, oxygen atoms, and aluminum oxide molecules, among which, For wavelength, For time; synchronously acquire temperature distribution data of the reaction chamber wall. Exhaust gas composition data , Here are the axial position coordinates. Data was collected by an array of thermocouples arranged at equal intervals along the axial direction of the chamber. Mass spectrometer data collected at the exhaust outlet will be used to collect data. , and Perform timestamp alignment and construct multi-source heterogeneous data frames: ; Step 2: Process the real-time spectral data Preprocessing is performed to extract the characteristic spectral intensity vector.

[0017] Step 3: Based on the characteristic spectral intensity vector Calculate the gas phase stoichiometric equilibrium index in the current reaction chamber. Electronic temperature and electron density ; Step 4: Based on the calculations obtained in Step 3 Combined with the preset target crystal growth rate and current chamber temperature Microwave input power is adjusted by an adaptive PID controller. and the flow rate of the reacting gas Generate power regulation command and flow regulation commands ; For all at the current moment The average value; Step 5: Utilize the power adjustment command The microwave source is driven to adjust its output power, while the flow regulation command is used simultaneously. The mass flow controller adjusts the mixing ratio of oxygen and inert carrier gas to maintain a constant total inlet flow rate, thereby stabilizing the plasma state in the reaction chamber within the preset crystal growth window and controlling the crystallization quality of alumina crystals. Step Six, Based on , , , , A crystal quality prediction model is constructed using historical crystal growth data to evaluate the crystal quality index in real time. and adjust the quality feedback signal Feedback is sent to step four to form a closed-loop control of crystal quality; Step 7: Establish a multi-objective optimization function to collaboratively optimize crystal growth rate, energy consumption, and gas utilization rate, and generate the optimal control strategy.

[0018] Step two specifically includes the following sub-steps: Step 21: Use a bandpass filter circuit to filter out the... The frequency in the middle is higher than the preset cutoff frequency. The noise components are removed to obtain the denoised spectral signal. The preset cutoff frequency Determined based on the wavelength range of the characteristic spectral lines and the temporal noise characteristics of the spectral acquisition; Step 22, the output of step 21 Peak finding was performed, and a Gaussian fitting algorithm was used to locate the center wavelengths of two aluminum atom characteristic spectral lines at different upper energy levels. , Central wavelength of oxygen atom characteristic spectral lines Central wavelength of characteristic spectral lines of alumina molecules and the center wavelength of the characteristic spectral lines of aluminum single ionization ions Among them, two aluminum atom spectral lines satisfy , , These are the corresponding upper energy levels; Step 23, based on the obtained , , , and Within the preset integration window width Integrating the spectral intensity, the integrated intensity of the two aluminum atom characteristic spectral lines was calculated. , Intensity of characteristic spectral lines of oxygen atoms Intensity of characteristic spectral lines of alumina molecules and the intensity of characteristic spectral lines of aluminum single ionization ions , will get , , , and The characteristic spectral intensity vector is obtained by combining them. , It is expressed as follows: ; (1) Step three specifically includes the following steps: Step 31: Calculate the basic stoichiometric equilibrium index. The formula is as follows: , (2) in, and The preset proportional weighting coefficient is used to characterize the relative concentration relationship between aluminum and oxygen elements and the formation efficiency of alumina precursor. Step 32: Calculate the plasma electron temperature using the modified Boltzmann slope method. The calculation formula is as follows: , (3) in, Boltzmann's constant, , These are the statistical weights of the energy levels on the two spectral lines, respectively. , The spontaneous emission transition probabilities of the two spectral lines can be obtained by looking up tables in the NIST atomic spectroscopy database; Step 33: Utilize the aforementioned electronic temperature , combined and Electron density was calculated by modifying the Saha ionization equation. The calculation formula is as follows: , (4) in, The rest mass of the electron. It is Planck's constant; Let be the partition function of aluminum atoms. The partition function of aluminum monoionized ions can be obtained by looking up tables in atomic spectroscopy databases; The degeneracy corresponding to the characteristic spectral lines of aluminum atoms. The degeneracy corresponding to the characteristic spectral lines of aluminum single ion can be obtained by looking up a table in the NIST atomic spectroscopy database; The central wavelength of the characteristic spectral lines of aluminum atoms. The center wavelength of the characteristic spectral line of aluminum single ion is determined by the spectral peak-finding process in step 22; The spontaneous emission transition probabilities of the characteristic spectral lines of aluminum single ionization ions can be obtained by looking up tables in the NIST atomic spectroscopy database. The integral intensity of the characteristic spectral line of aluminum single ionization ions at time t; This is the first ionization energy of an aluminum atom; The upper energy level corresponding to the characteristic spectral line of aluminum monoionization ions can be obtained by looking up a table in the NIST atomic spectroscopy database; It should be noted that It is actually a constant, which can be obtained by looking up a table.

[0019] Step 34: Set the electron temperature and the electron density As a correction parameter, for and Dynamic adjustments are made to the basic stoichiometric balance index. After correction, the final gas-phase stoichiometric equilibrium index is obtained. The calculation formula is as follows: , (5) in, and This is a correction function based on changes in electron temperature and density. , , These are preset parameters.

[0020] Step four specifically includes the following steps: Step 41: Adjust the gas phase stoichiometric equilibrium index. The ideal stoichiometric ratio setpoint for the current cycle By comparison, the deviation value is obtained. ; Step 42, based on the obtained and Update the proportional coefficient of the adaptive PID controller Integral coefficient and differential coefficients The formula is as follows: , (6) in, , , These are the three initial parameters for the PID controller. , For the preset ideal plasma state parameters, , , Adjust the sensitivity coefficient for the parameter; Step 43: The deviation value obtained in step 41 Perform proportional, integral, and derivative calculations to obtain the basic control quantity. : ; (7) Step 44: Based on the aforementioned basic control quantity And combined with the current average temperature of the chamber With target temperature The difference Calculate the final power regulation command : , (8) in, This is the temperature compensation coefficient. This is the temperature damping coefficient; For all at time t The average value; Step 45, from Obtain the residual oxygen concentration at the outlet of the reaction chamber. The obtained With preset safety threshold If a comparison is made, Then an initial flow regulation command is generated. : , (9) in, For flow rate adjustment sensitivity, This is the concentration change rate compensation coefficient; when ,but ; Step 46, to Amplitude limiting is performed to obtain the final flow regulation command. : , in, It is a saturation function. , Adjust the upper and lower limits for flow rate.

[0021] Step five specifically includes the following steps: Step 51: The power adjustment command generated in step 44... The input is sent to the microwave power source controller, which, according to... Adjusting the anode current of the magnetron changes the microwave input power. , (10) in, Let t be the microwave power at time t. The microwave power source controller can respond quickly, with a response time of less than 10 milliseconds. Step 52: Apply the flow adjustment command generated in step 46. The input is sent to the mass flow controller, which, according to... Adjust the oxygen flow rate and the inert carrier gas flow rate separately: , (11) in, Let be the oxygen flow rate at time t. Let be the inert gas flow rate at time t. The limiting function restricts the oxygen flow rate to a certain value. and Within a certain range, the flow rate of the inert gas is limited to... Within the specified range, the inert carrier gas is argon or helium; Step 53: Constructing plasma state stability criteria A comprehensive evaluation of the plasma state after power and flow rate regulation was conducted. , (12) in, , , For the stability criterion, the weighting coefficients are... ; Step 54, if ,in If the preset stability threshold is not met, the plasma state is determined to be unstable, and the process returns to step one to re-acquire data and perform adjustments; if... If the plasma state is stable, proceed to step six to predict the crystal quality.

[0022] Preferably, step six specifically includes the following steps: Step 61, based on and historical crystal quality data Build the training dataset: , (13) in, Including X-ray diffraction full width at half maximum (FWHM), surface roughness, and optical transmittance. Indicates the first Historical crystal quality data; Step 62: Build a crystal quality prediction model based on a long short-term memory network: , (14) In the formula, It is the crystal mass at time t predicted by the crystal quality prediction model. For network parameters, For model residuals; Step 63: Utilize the training dataset The mean squared error loss function and Adam optimizer are used to iteratively train the long short-term memory network through backpropagation algorithm to minimize the error between the predicted crystal quality index and historical crystal quality detection data, until the loss function converges or the preset maximum number of iterations is reached, thus obtaining the trained network parameters. .

[0023] Step 64: The crystal quality prediction model obtained using the pre-trained crystal quality prediction model... With the preset quality target By comparison, the crystal quality deviation at time t is calculated: , (15) when At that time, a feedback signal is generated. , This is the quality deviation threshold. For feedback coefficients; otherwise, ; Step 65: Using the mass-to-stoichiometric ratio conversion factor The quality feedback signal is mapped to the stoichiometric ratio target correction amount. : ; (16) Step 66: Update the ideal stoichiometric ratio setpoint for the next control cycle: , (17) The updated The ideal stoichiometric ratio setpoint of the current cycle is used as the basis for step 41 of the next control cycle to form the outer slow closed-loop control of crystal quality.

[0024] Step seven specifically includes the following steps: Step 71: Construct a multi-objective optimization function Taking into account three optimization objectives: crystal growth rate, energy efficiency, and gas utilization rate: (18) in, This represents the actual crystal growth rate. Actual energy consumption For maximum allowable energy consumption, This represents the cumulative gas consumption. For the maximum permissible gas consumption, , , These are the weighting coefficients for growth rate, energy efficiency, and gas utilization rate, respectively. In practice, the actual crystal growth rate is determined by real-time monitoring of the crystal's weight using a weight sensor, and the actual crystal growth rate is obtained based on the weight change per unit time. Actual energy consumption refers to the electrical energy consumed in the circuit.

[0025] Step 72: Use the crystal quality prediction model trained in steps 61 to 63 as a constraint condition. The constraint condition requires that the optimized control parameters must guarantee... The particle swarm optimization algorithm is used to solve the problem. Minimize the optimal combination of control parameters The number of particles and the number of iterations in the particle swarm optimization algorithm are predetermined.

[0026] Step 73: Combine the optimal control parameters As reference setpoints for steps four and five, these setpoints take effect in the next control cycle, achieving multi-objective collaborative optimization control of the crystal growth process. The control cycle is dynamically adjusted according to the plasma state change rate. Specifically, [the following text is missing from the original extract]. Assign the value to the target temperature in step 44 ,Will Assign the value to the microwave power in step 51 ,Will According to the allocation weight of oxygen and inert carrier gas and Decomposed into: , (19) , (20) Will Assigned to step 52 ,Will Assign to ,in, , The weightings for oxygen and inert carrier gas are respectively. + .

[0027] 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 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 claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for controlling the preparation of alumina crystals using microwave plasma, characterized in that, Specifically, the steps include the following: Step 1: Acquire real-time spectral data of the plasma glow region using a multispectral sensor array integrated into the microwave plasma reaction chamber. Real-time spectral data It contains characteristic emission spectral information of aluminum atoms, oxygen atoms, and aluminum oxide molecules, among which, For wavelength, For time; synchronously acquire temperature distribution data of the reaction chamber wall. Exhaust gas composition data , These are the axial position coordinates; Step 2: Process the real-time spectral data Preprocessing is performed to extract the characteristic spectral intensity vector. ; Step 3: Based on the characteristic spectral intensity vector Calculate the gas phase stoichiometric equilibrium index in the current reaction chamber. Electronic temperature and electron density ; Step 4: Based on the calculations obtained in Step 3 Combined with the preset target crystal growth rate and current chamber temperature Microwave input power is adjusted by an adaptive PID controller. and the flow rate of the reacting gas Generate power regulation command and flow regulation commands ; For all at the current moment The average value; Step 5: Utilize the power adjustment command The microwave source is driven to adjust its output power, while the flow regulation command is used simultaneously. The mass flow controller adjusts the mixing ratio of oxygen and inert carrier gas.

2. The control method according to claim 1, characterized in that, It also includes the following steps: Step Six, Based on , , , , A crystal quality prediction model is constructed using historical crystal growth data to evaluate the crystal quality index in real time. and adjust the quality feedback signal Feedback is sent to step four to form a closed-loop control of crystal quality; Step 7: Establish a multi-objective optimization function to collaboratively optimize crystal growth rate, energy consumption, and gas utilization rate, and generate the optimal control strategy.

3. The control method according to claim 2, characterized in that, Step two specifically includes the following sub-steps: Step 21: Use a bandpass filter circuit to filter out the... The frequency in the middle is higher than the preset cutoff frequency. The noise components are removed to obtain the denoised spectral signal. ; Step 22, the output of step 21 Peak finding was performed to locate the center wavelengths of the characteristic spectral lines of aluminum atoms at two different upper energy levels. , Central wavelength of oxygen atom characteristic spectral lines Central wavelength of characteristic spectral lines of alumina molecules and the center wavelength of the characteristic spectral lines of aluminum single ionization ions Among them, two aluminum atom spectral lines satisfy , , These are the corresponding upper energy levels; Step 23, based on the obtained , , , and Within the preset integration window width Integrating the spectral intensity, the integrated intensity of the two aluminum atom characteristic spectral lines was calculated. , Intensity of characteristic spectral lines of oxygen atoms Intensity of characteristic spectral lines of alumina molecules and the intensity of characteristic spectral lines of aluminum single ionization ions , will get , , , and The characteristic spectral intensity vector is obtained by combining them. , It is expressed as follows: （1）。 4. The control method according to claim 3, characterized in that, Step three specifically includes the following steps: Step 31: Calculate the basic stoichiometric equilibrium index. The formula is as follows: ， （2） in, and The preset proportional weighting coefficient is used to characterize the relative concentration relationship between aluminum and oxygen elements and the formation efficiency of alumina precursor. Step 32: Calculate the plasma electron temperature using the modified Boltzmann slope method. The calculation formula is as follows: ， （3） in, Boltzmann's constant, , These are the statistical weights of the energy levels on the two spectral lines, respectively. , The spontaneous emission transition probabilities of the two spectral lines can be obtained by looking up tables in the NIST atomic spectroscopy database; Step 33: Utilize the aforementioned electronic temperature , combined and Electron density was calculated by modifying the Saha ionization equation. The calculation formula is as follows: ，（4） in, The rest mass of the electron. It is Planck's constant; Let be the partition function of aluminum atoms. The partition function of aluminum monoionized ions can be obtained by looking up tables in atomic spectroscopy databases; The degeneracy corresponding to the characteristic spectral lines of aluminum atoms. The degeneracy corresponding to the characteristic spectral lines of aluminum single ion can be obtained by looking up a table in the NIST atomic spectroscopy database; The central wavelength of the characteristic spectral lines of aluminum atoms. The center wavelength of the characteristic spectral line of aluminum single ion is determined by the spectral peak-finding process in step 22; The spontaneous emission transition probability of the characteristic spectral line of aluminum single ion can be obtained by looking up a table in the NIST atomic spectral database. The integral intensity of the characteristic spectral line of aluminum single ionization ions at time t; This is the first ionization energy of an aluminum atom; The upper energy level corresponding to the characteristic spectral line of aluminum monoionization ions can be obtained by looking up a table in the NIST atomic spectroscopy database; Step 34: Set the electron temperature and the electron density As a correction parameter, for and Dynamic adjustments are made to the basic stoichiometric balance index. After correction, the final gas-phase stoichiometric equilibrium index is obtained. The calculation formula is as follows: ，（5） in, and This is a correction function based on changes in electron temperature and density.

5. The control method according to claim 4, characterized in that, Step four specifically includes the following steps: Step 41: Adjust the gas phase stoichiometric equilibrium index. The ideal stoichiometric ratio setpoint for the current cycle By comparison, the deviation value is obtained. ; Step 42, based on the obtained and Update the proportional coefficient of the adaptive PID controller Integral coefficient and differential coefficients The formula is as follows: ， （6） in, , , These are the three initial parameters for the PID controller. , For the preset ideal plasma state parameters, , , Adjust the sensitivity coefficient for the parameter; Step 43: The deviation value obtained in step 41 Perform proportional, integral, and derivative calculations to obtain the basic control quantity. : ； （7） Step 44: Based on the aforementioned basic control quantity And combined with the current average temperature of the chamber With target temperature The difference Calculate the final power regulation command : , （8） in, This is the temperature compensation coefficient. This is the temperature damping coefficient; For all at time t The average value; Step 45, from Obtain the residual oxygen concentration at the outlet of the reaction chamber. The obtained With preset safety threshold If a comparison is made, Then an initial flow regulation command is generated. : ， （9） in, For flow rate adjustment sensitivity, This is the concentration change rate compensation coefficient; when ,but ; Step 46, to Amplitude limiting is performed to obtain the final flow regulation command. : ， in, It is a saturation function. , These are the upper and lower limits for the flow regulation command.

6. The control method according to claim 5, characterized in that, Step five specifically includes the following steps: Step 51: The power adjustment command generated in step 44... The input is sent to the microwave power source controller, which, according to... Adjusting the anode current of the magnetron changes the microwave input power. ， （10） in, Let be the microwave power at time t; Step 52: Apply the flow adjustment command generated in step 46. The input is sent to the mass flow controller, which, according to... Adjust the oxygen flow rate and the inert carrier gas flow rate separately: ， （11） in, Let be the oxygen flow rate at time t. Let be the inert gas flow rate at time t. The limiting function restricts the oxygen flow rate to a certain value. and Within a certain range, the flow rate of the inert gas is limited to... Within the specified range, the inert carrier gas is argon or helium; Step 53: Constructing plasma state stability criteria A comprehensive evaluation of the plasma state after power and flow rate regulation was conducted. ，（12） in, , , For the stability criterion, the weighting coefficients are... ; Step 54, if ,in If the preset stability threshold is not met, the plasma state is determined to be unstable, and the process returns to step one to re-acquire data and perform adjustments; if... If the plasma state is stable, proceed to step six to predict the crystal quality.

7. The control method according to claim 6, characterized in that, Step six specifically includes the following steps: Step 61, based on and historical crystal quality data Build the training dataset: ， （13） in, Including X-ray diffraction full width at half maximum (FWHM), surface roughness, and optical transmittance. Indicates the first Historical crystal quality data; Step 62: Build a crystal quality prediction model based on a long short-term memory network: ，（14） In the formula, It is the crystal mass at time t predicted by the crystal quality prediction model. For network parameters, For model residuals; Step 63: Utilize the training dataset The mean squared error loss function and Adam optimizer are used to iteratively train the long short-term memory network through backpropagation algorithm to minimize the error between the predicted crystal quality index and historical crystal quality detection data, until the loss function converges or the preset maximum number of iterations is reached, thus obtaining the trained network parameters. ; Step 64: The crystal quality prediction model obtained using the pre-trained crystal quality prediction model... With the preset quality target By comparison, the crystal quality deviation at time t is calculated: ， （15） when At that time, a feedback signal is generated. , This is the quality deviation threshold. For feedback coefficients; Step 65: Using the mass-to-stoichiometric ratio conversion factor The quality feedback signal is mapped to the stoichiometric ratio target correction amount. : ； （16） Step 66: Update the ideal stoichiometric ratio setpoint for the next control cycle: ， （17） The updated The ideal stoichiometric ratio setpoint of the current cycle is used as the basis for step 41 of the next control cycle to form the outer slow closed-loop control of crystal quality.

8. The control method according to claim 7, characterized in that, Step seven specifically includes the following steps: Step 71: Construct a multi-objective optimization function Taking into account three optimization objectives: crystal growth rate, energy efficiency, and gas utilization rate: (18) in, This represents the actual crystal growth rate. Actual energy consumption For maximum allowable energy consumption, This represents the cumulative gas consumption. For the maximum permissible gas consumption, , , These are the weighting coefficients for growth rate, energy efficiency, and gas utilization rate, respectively. ; Step 72: Use the crystal quality prediction model trained in steps 61 to 63 as a constraint condition. The constraint condition requires that the optimized control parameters must guarantee... The particle swarm optimization algorithm is used to solve the problem. Minimize the optimal combination of control parameters The number of particles and the number of iterations in the particle swarm optimization algorithm are predetermined; Step 73: Combine the optimal control parameters As a reference setting value for steps four and five, the reference setting value takes effect in the next control cycle.

Images