A method for optimizing parameters of high-purity molybdenum trioxide sublimation purification

Abstract

The application discloses a high-purity molybdenum trioxide sublimation purification parameter optimization method, relates to the technical field of high polymer compound purification and crystal growth, and realizes digital evaluation of the particle size distribution of raw materials by constructing a raw material particle size state factor Gsize through quantitative analysis of a particle size distribution skewness value S and a particle size distribution kurtosis value K, effectively identifies whether the particle size distribution is concentrated and reasonable before sublimation, screens out abnormal raw materials with excessively large or small particle sizes, and promotes reduction of uneven vapor release caused by particle size differences in the sublimation process. By introducing image recognition and a KNN classification model, the proportion of spherical particles is further screened, and agglomerates are removed. Through proportion calculation of spherical particles, flaky particles, needle-like particles and agglomerated particles, the proportion of spherical particles is maintained at a high level, while the proportion of agglomerated particles is controlled to be below a preset threshold, local disturbance during vapor release is reduced, and the consistency of crystal nucleation in the mass transfer process is improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of high-molecular compound purification and crystal growth, and particularly relates to a high-purity molybdenum trioxide sublimation purification parameter optimization method. BACKGROUND

[0002] Molybdenum trioxide is an important inorganic compound and is widely used in the fields of photoelectric devices, catalytic materials and high-end alloy additives, and has extremely high requirements for its purity and crystal structure. As a key process for preparing high-purity molybdenum trioxide, the sublimation purification technology can effectively remove metal impurities and non-volatile impurities by using the low sublimation temperature to perform gas-solid phase conversion on raw materials.

[0003] In the traditional molybdenum trioxide sublimation purification process, the unevenness of the particle size distribution of the raw materials is one of the important factors affecting the sublimation quality and purification efficiency. Due to the lack of systematic raw material pretreatment and quantitative analysis mechanism, the original powder with mixed particle sizes is often directly put into the sublimation cavity. The particles with large particle sizes will lag behind in releasing steam due to the slow sublimation rate, causing the local shortage of steam concentration in the sublimation area; and the particles with small particle sizes will release a large amount of steam quickly due to the easy gasification, forming a local high-concentration area. The steam release rate difference caused by the uneven particle size distribution can easily cause airflow disturbance and temperature fluctuation in the sublimation area, and then affect the stable mass transfer conditions between the steam and the crystal surface, resulting in unstable crystal nucleation position, uneven crystal size and disordered crystallization direction.

[0004] In addition, if a large number of agglomerates or irregular flaky and needle-shaped particles are contained in the particle size distribution, the heat field distribution and the surface sublimation uniformity will be further disturbed. In particular, the agglomerated particles release steam unevenly during sublimation, and even cause the "false volatilization" phenomenon, which interferes with the integrity of the subsequent crystalline structure. Therefore, in the traditional technology, no clear particle size state evaluation index is introduced, and there is also a lack of mechanism for guiding the setting of sublimation conditions based on the particle size distribution characteristics, which seriously restricts the consistency of the sublimation purification and the repeatability of the high-purity product. SUMMARY

[0005] In view of the deficiencies of the prior art, the application provides a high-purity molybdenum trioxide sublimation purification parameter optimization method to solve the problems mentioned in the background.

[0006] To achieve the above purpose, the application is implemented by the following technical scheme: a high-purity molybdenum trioxide sublimation purification parameter optimization method, comprising the following steps:

[0007] Step one, the particle size distribution of the molybdenum trioxide raw material before sublimation is tested, a laser particle size analyzer, image particle size analysis or SEM image analysis method is used to obtain the particle size distribution spectrum and the particle size distribution skewness value S and the particle size distribution kurtosis value K of the raw material, and an initial particle size state factor Gsize is constructed; when Gsize≥0.7 and K≥3, it indicates that the current molybdenum trioxide raw material is pre-monitored qualified, and is summarized as the first pre-qualified raw material, which enters step two;

[0008] Step two, the first pre-qualified raw material is photographed to obtain the raw material crystal image, the feature vector of the particle is constructed after image preprocessing of the raw material crystal image, and the category label is output through the trained classifier KNN, and the percentage of the kth category label in the total particle number is calculated to obtain the proportion of spherical particles, the proportion of flaky particles, the proportion of needle / cylindrical particles and the proportion of agglomerated particles; when the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of agglomerated particles is <10%, it indicates that it is qualified and enters step three;

[0009] Step three, in the initial stage of sublimation, the temperature field distribution, gas phase concentration distribution and sublimation wall surface crystal face structure in the sublimation cavity are collected to construct a boundary layer structure model under the initial sublimation condition; combined with three-dimensional particle tracking technology, the near-wall diffusion rate curve of molybdenum trioxide vapor is obtained to construct the concentration gradient of the sublimation wall boundary layer and the corresponding vapor diffusion flux J to determine whether there is vapor accumulation in the boundary layer, whether there is mass transfer bottleneck risk, if there is, generate the first micro dynamic correction strategy and instruction, and enter step four after there is no mass transfer bottleneck risk;

[0010] Step four, using thermal imaging matrix, flow field tracking and sublimation vapor concentration inversion technology, the thermal capillary convection path and intensity change in the sublimation cavity are captured, a thermal convection offset vector diagram is constructed, and a flow field stability index F TCV is calculated; a preset stability threshold F is set, when F TCV >F, the system executes a local temperature zone fine tuning instruction, and then enters step five;

[0011] Step five, in the process of condensation and deposition, the sublimation vapor deposition non-uniformity characteristics are identified through condensation surface temperature field scanning and morphology evolution spectrum monitoring, and a sublimation-condensation asymmetry factor ASC is constructed, and a symmetry critical threshold X is preset, when ASC>X, the condensation symmetry is corrected by dynamically adjusting the condensation surface temperature distribution and the crystal face guiding direction, and then the crystallization is completed.

[0012] Preferably, step one comprises:

[0013] S1, test the particle size distribution of the molybdenum trioxide raw material before sublimation, use a laser particle size analyzer, image particle size analysis or SEM image analysis method to obtain the particle size distribution spectrum and skewness, kurtosis statistical characteristics of the raw material, and construct an initial particle size state factor Gsize;

[0014] The specific steps for obtaining the initial particle size state factor Gsize are:

[0015] S11, the particle size distribution spectrum and skewness, kurtosis statistical characteristics of the raw material are obtained to obtain 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S and particle size distribution kurtosis value K;

[0016] S121, the particle size distribution skewness value S is obtained in the following manner:

[0017]

[0018] wherein n represents the total number of particle size data points, represents the i-th particle size center value, the representative particle size of each particle size interval, including 1.2 μm, 1.6 μm; derived from the particle size analysis equipment; represents the average particle size, represents the standard deviation, represents the particle size the corresponding volume fraction is a normalized value;

[0019] S122, evaluate the particle size distribution skewness value S to obtain a first evaluation result, including:

[0020] When S = 0, it indicates that the distribution is completely symmetric with respect to the average value;

[0021] When S > 0, it indicates that the distribution is right-skewed with respect to the average value;

[0022] When S < 0, it indicates that the distribution is left-skewed with respect to the average value.

[0023] Preferably, step one further comprises:

[0024] S123, the particle size distribution kurtosis value K is obtained in the following manner:

[0025]

[0026] The formula means that the fourth order central moment is a normalized value, which reflects the peak sharpness and tail thickness of the distribution; the particle size distribution kurtosis value K is the fourth power of the distance, and the influence of "extreme values far from the average value" in the distribution is greater, which can determine whether the distribution is sharp or flat;

[0027] S124, evaluate the particle size distribution kurtosis value K to obtain a second evaluation result, including:

[0028] When K = 3, it means that the normal distribution is moderate, the tail thickness is moderate, and there is no obvious central or dispersion deviation;

[0029] When K > 3, it means that the distribution is relative to the average value, the concentration is high, the tail is heavier, and a sharp peak distribution is formed, which is good for controlling the uniformity of sublimation;

[0030] When K < 3, it means that the distribution is relative to the average value, the concentration is low, the tail is more dispersed, and a flat-top distribution is formed; the distribution is flat, and the mass transfer in the sublimation stage is limited and fluctuates greatly;

[0031] S125, in combination with the 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S and particle size distribution kurtosis value K obtained in S121-S123, after normalization processing, the initial particle size state factor Gsize is obtained by weighted combination:

[0032]

[0033] In the formula, The distribution width of the particle size or the particle size range reflects the dispersity; 、 And The weight is represented as , , , and ;

[0034] S126, the initial particle size state factor Gsize is evaluated, if the following two judgment conditions are met at the same time, including: if Gsize < 0.7, and K < 3, it means that the distribution is unqualified, which will cause abnormal risk in the subsequent sublimation stage, then trigger the first early warning instruction, including: screening, adjusting the ratio, ultrasonic vibration or mechanical crushing of the raw material, to promote the deaggregation of the raw material particles;

[0035] When Gsize ≥ 0.7, and K ≥ 3, it means that the current molybdenum trioxide raw material pre-monitoring is qualified, and it is summarized as the first pre-qualified raw material, and enters step two.

[0036] Preferably, step two includes:

[0037] S21, the first pre-qualified raw material is photographed to obtain a raw material crystal image, and the raw material crystal image is pre-processed, including: after the raw material crystal image is grayed, binarized, background noise is removed and the edge contrast is enhanced, the single particle boundary contour is extracted, and the particle segmentation is completed; and each particle contour is closed processed, and the non-target area is removed;

[0038] The circularity feature, the aspect ratio feature, the boundary irregularity feature and the area proportion feature are extracted for each particle contour to obtain the circularity of the particle , the aspect ratio , the boundary irregularity and the ratio of the projection area and the convex hull area ;

[0039] S2101, the circularity of the particle The calculation expression is:

[0040]

[0041] Wherein, A is the projection area of the particle, P is the boundary perimeter of the particle, and the circularity value closer to 1 indicates that the particle is more regular and close to spherical;

[0042] S2102, the aspect ratio of the particle The calculation expression is:

[0043]

[0044] Wherein, L and W are the length of the long axis and the length of the short axis of the fitted circumscribed rectangle of the particle, respectively, reflecting the stretching degree of the particle;

[0045] S2103, the boundary irregularity of the particle The calculation expression is:

[0046]

[0047] Wherein, the greater the value, the more complex the boundary and the more irregular the morphology;

[0048] S2104, the ratio of the projection area and the convex hull area of the particle The calculation expression is:

[0049]

[0050] Wherein, represents the convex hull area of the particle, used to judge whether there is agglomeration or adhesion;

[0051] S22, the circularity , the aspect ratio , the boundary irregularity and the ratio of the projection area and the convex hull area of the particle are sequentially composed into the feature vector of the particle:

[0052]

[0053] The feature vector of each particle In the input trained classifier KNN, the feature vector of the particle is calculated by the trained classifier KNN The Euclidean distance between the particle and all training sample feature vectors is calculated, the nearest k neighbors are found, and the class with the most occurrences in the k neighbors is counted to obtain the corresponding class, and the class label is output.

[0054] The class label includes: 001 spherical, 002 flaky, 003 acicular / cylindrical, and 004 agglomerated.

[0055] The total number of particles of the kth class label in the first prequalified raw material is counted, and the percentage of particles of the kth class label in the total number of particles is calculated to obtain the spherical particle ratio, flaky particle ratio, acicular / cylindrical ratio, and agglomerated particle ratio.

[0056] When the spherical particle ratio in the first prequalified raw material is ≥50% and the agglomerated particle ratio is <10%, it indicates that it is qualified and enters step three; otherwise, it indicates that it is unqualified, i.e., "spherical particle deficiency" or "excessive agglomeration", which indicates that it needs to be returned to the re-screening process.

[0057] Preferably, step three includes:

[0058] S311, in the sublimation initiation stage, the temperature field distribution in the sublimation cavity is collected by using a temperature sensor array and an infrared thermal imaging system.

[0059] S312, the gas phase molybdenum trioxide concentration distribution in the sublimation cavity is collected in combination with a gas phase concentration sensor or a Raman / infrared spectrometer.

[0060] S313, the crystal face structure of the sublimation wall surface is obtained by in-situ surface diffraction or reflection electron diffraction means.

[0061] S314, the motion trajectory of the molybdenum trioxide vapor particles in the sublimation process is tracked by combining three-dimensional particle tracking velocimetry 3D-PTV, and the velocity field, diffusion path and time sequence of the boundary layer vapor particles are extracted.

[0062] S315, in the molybdenum trioxide sublimation process, the raw material solid is directly converted into vapor at high temperature, and the thin layer region between the sublimation wall surface and the main gas flow is marked as the boundary layer, in which the concentration, temperature and flow rate will change significantly, and it is also the most "difficult" place for gas phase material diffusion.

[0063] S316, the vapor diffusion rate curve of the boundary layer is constructed, and the concentration gradient of the sublimation wall surface boundary layer is calculated according to the vapor diffusion rate curve of the boundary layer.

[0064]

[0065] ​where C is the concentration of the molybdenum trioxide vapor in the boundary layer, and x is the direction perpendicular to the sublimation wall surface; represents the derivative of the concentration with respect to the position, and reflects how fast the vapor concentration drops "from the wall surface outward". represents the numerical value of the concentration gradient of the sublimation wall boundary layer.

[0066] Preferably, step three further comprises:

[0067] S317, according to Fick's first law, the vapor diffusion flux J is proportional to the concentration gradient of the sublimation wall boundary layer, and the expression is as follows:

[0068]

[0069] where D represents the diffusion coefficient, including:

[0070] Under the condition of sublimation in normal pressure air, the sublimation temperature range of molybdenum trioxide is 700-850°C, the corresponding temperature is 973-1123K, and the pressure is 1.0atm. The diffusion coefficient D under this condition is 0.08-0.12cm² / s;

[0071] Under the condition of sublimation in reduced pressure air, the sublimation temperature range of molybdenum trioxide is 600-750°C, the corresponding temperature is 873-1023, and the pressure is 0.1atm. The diffusion coefficient D under this condition is 0.60-0.90cm² / s;

[0072] Under the condition of sublimation in Ar or N2 inert gas protection, the sublimation temperature range of molybdenum trioxide is 650-800°C, the corresponding temperature is 923-1073, and the pressure is 1.0atm. The diffusion coefficient D under this condition is 0.10-0.15cm² / s;

[0073] Under the condition of sublimation in vacuum, the sublimation temperature range of molybdenum trioxide is 550-700°C, the corresponding temperature is 823-973K, and the pressure is 10 -4 atm. The diffusion coefficient D under this condition is 10-25cm² / s, and increases with temperature.

[0074] The negative sign indicates that the substance moves from a place with high concentration to a place with low concentration.

[0075] Preferably, step three further comprises:

[0076] S318, presetting a critical threshold G1 and a diffusion flux efficiency threshold J1, when the concentration gradient of the sublimation wall boundary layer When the critical threshold G1 is set, and the vapor diffusion flux J < diffusion flux efficiency threshold J1, it indicates that the vapor is accumulated in the boundary layer, and there is a risk of mass transfer bottleneck, and a first micro dynamic correction strategy is generated, including: the change amplitude of temperature gradient is adjusted by 0.5℃-2℃ each time, and the vibration amplitude of the boundary layer is increased by 1µm-2µm each time;

[0077] After the first micro dynamic correction strategy is executed, the concentration gradient of the sublimation wall boundary layer and the vapor diffusion flux are re-detected after stabilization, until ≤G1, and the vapor diffusion flux J≥J1, then step four is entered; when ≤G1, and the vapor diffusion flux J≥J1, it indicates that the mass transfer condition is normal, and the vapor diffusion flux efficiency meets the standard, and the sublimation state is continuously monitored.

[0078] Preferably, the step four includes:

[0079] S41, using thermal imaging and airflow sensors, monitoring the temperature and airflow changes in the sublimation cavity, obtaining thermal field data and flow field data in the cavity;

[0080] S42, according to the obtained thermal field data and flow field data in the cavity, combining the sublimation vapor concentration inversion information, constructing a thermal convection displacement vector diagram in the sublimation cavity, and calculating a flow field stability index F TCV :

[0081]

[0082] Wherein, N represents the total number of sampling points, represents the airflow velocity vector of the jth point, represents the average velocity vector, represents the temperature of the jth point, represents the average temperature, represents the weight coefficient of temperature fluctuation; The setting range of includes:

[0083] In the initial coarse adjustment stage, α=1.0, both the speed and the temperature fluctuation are considered;

[0084] Fine monitoring stage: α=1.5-2.0, strengthen the temperature disturbance detection ability;

[0085] Local condensation and re-crystallization stage, α=2.5;

[0086] represents the dimensionless relative speed fluctuation, represents the dimensionless relative temperature fluctuation;

[0087] S43, a stability threshold F is preset, and the flow field stability index F TCVwhether the set stability threshold F is exceeded, if exceeded → trigger local temperature zone fine-tuning instructions to suppress abnormal heat convection, the local temperature zone fine-tuning instructions comprising: arranging a cold source on the back or side of the temperature zone to perform directional heat adjustment of 0.2-0.8°C, and disturbing the local airflow direction in the temperature zone ring to trigger control of the guide vane adjustment to close the direction ±5°-20°;

[0088] If the flow field stability index F TCV If the stability threshold F is not exceeded → maintain the current parameters, continue monitoring, and enter step five.

[0089] Preferably, the step five comprises:

[0090] S51, using a high-resolution infrared thermal imager and a morphology evolution map acquisition device, continuously scanning the condensing surface to obtain the following parameters and construct a condensing surface deposition uniformity dataset, including: condensing surface temperature distribution standard deviation, condensing crystal thickness distribution, and condensing crystal deposition morphology image;

[0091] S52, constructing a sublimation-condensation asymmetry factor ASC according to the condensing surface deposition uniformity dataset, for quantifying the imbalance degree of the deposition distribution on the condensing surface and the thermal field distribution, the expression being:

[0092]

[0093] In the formula, denotes the standard deviation of the crystal thickness, denotes the average value of the crystal thickness, denotes the standard deviation of the condensing surface temperature distribution, denotes the average value of the temperature field during condensation, denotes the morphology image symmetry deviation degree; 、 and denote weights, , , , and ;

[0094] wherein, the acquisition method is:

[0095] For the condensing crystal deposition morphology image, mirror image is performed on the left and right or up and down according to the center axis of the condensing surface, the image difference is calculated using the mean square error MSE method, and the morphology image symmetry deviation degree is obtained :

[0096]

[0097] wherein, M denotes the total number of pixel points in the condensing crystal deposition morphology image, Pixel value of the kth pixel point in the image of the deposition morphology of the condensed crystal, Pixel value corresponding to the kth pixel point after mirroring left and right or up and down along the central axis of the condensation surface;

[0098] If the two images are completely symmetrical, the pixel values at the symmetrical positions are completely the same, then all Therefore , indicating perfect symmetry, if the difference between the symmetrical pixel values is large, then The larger the value is, the more asymmetric the morphology is.

[0099] Preferably, step five further comprises:

[0100] S53, a preset symmetry critical threshold X is set, when >X, it indicates that there is an asymmetric risk in the current condensation deposition, triggering a second warning instruction; when ≤X, it indicates that the current condensation deposition process is qualified, and the crystal structure is uniform and has no back condensation and backflow signs, then mark the current batch of crystallization process as qualified, and continue to monitor;

[0101] S54, after receiving the second warning instruction, a second slight dynamic correction strategy is generated, the temperature in the asymmetric area is reduced by 0.5-2℃, the deposition is promoted to move to the symmetric area, and the crystal seed direction is rotated or shifted, guiding the crystal to grow along a more stable symmetric direction, the sublimation-condensation asymmetry factor ASC is recalculated after each adjustment to determine whether it falls back to ≤X, if not, continue the second slight dynamic correction strategy until the deposition returns to the symmetric state.

[0102] The application provides a high-purity molybdenum trioxide sublimation purification parameter optimization method.

[0103] (1) The high-purity molybdenum trioxide sublimation purification parameter optimization method realizes digital evaluation of the particle size distribution of the raw material by quantitative analysis of the particle size distribution skewness value S and the particle size distribution kurtosis value K, realizes digital evaluation of the particle size distribution of the raw material, effectively identifies whether the particle size distribution is concentrated and reasonable before sublimation, screens out abnormal raw materials with excessively large or small particle sizes, significantly reduces the uneven vapor release phenomenon caused by particle size difference in the sublimation process, and improves the sublimation stability. By introducing image recognition and KNN classification model, the proportion of spherical particles is further screened and agglomerates are removed, and by calculating the proportions of spherical particles, flaky particles, needle-shaped particles and agglomerated particles, the proportion of spherical particles is maintained at a high level, and the proportion of agglomerated particles is controlled below a preset threshold, which can reduce local disturbance during vapor release and improve the stability and consistency of the crystal nucleation in the mass transfer process.

[0104] (2) The high-purity molybdenum trioxide sublimation purification parameter optimization method can effectively identify the accumulation trend of the vapor in the boundary layer and timely adjust the sublimation conditions in the initial stage of sublimation, so as to avoid the sublimation rate reduction and recondensation caused by the mass transfer bottleneck. During the sublimation process, by monitoring the change of the heat convection path and the flow field stability index F TCV size, the sublimation zone temperature can be adjusted before the heat convection appears drift and disorder trend, so as to ensure the stability of the vapor atmosphere and reduce the risk of defects in the crystallization process. In the condensation stage, by analyzing the spatial distribution of the sublimation vapor deposition and the temperature gradient of the condensation surface, the sublimation-condensation asymmetry factor ASC is quantified, and when the ASC exceeds the critical value, the condensation surface temperature and the crystal surface orientation are corrected, effectively improving the grain size consistency and structural integrity.

[0105] (3) The present application further introduces KNN (K-Nearest Neighbor) classifier as an intelligent classification tool for the condensation surface deposition symmetry image. KNN is a supervised learning algorithm based on sample similarity, which belongs to the basic classification model in the field of artificial intelligence. Without complex model training, the trained classifier KNN can output class labels and calculate the percentage of the kth class label in the total particle number, so as to obtain the proportion of spherical particles, flaky particles, needle / cylinder particles and agglomerated particles, which significantly improves the accuracy and efficiency of particle morphology recognition and reduces the probability of human intervention and misjudgment. BRIEF DESCRIPTION OF DRAWINGS

[0106] Figure 1 The present application is an artificial intelligence high-purity molybdenum trioxide sublimation purification parameter optimization method. DETAILED DESCRIPTION

[0107] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0108] Secondly, during the sublimation process, the vapor molecules near the sublimation wall form a high-concentration boundary layer. If the vapor accumulation in the boundary layer is not timely identified and intervened, the "mass transfer bottleneck" phenomenon is likely to occur, which leads to a decrease in the gas phase concentration gradient, a decrease in the sublimation rate, and the induction of local recondensation or crystallization defects. This problem is often ignored in traditional processes, and there is a lack of quantitative monitoring of the concentration distribution and diffusion flux of the boundary layer.

[0109] In addition, complex thermal convection phenomena are often formed in the sublimation cavity due to temperature difference driving, especially when the steam concentration is high and the gas flow viscosity changes significantly. Local thermal convection paths may drift or reverse flow, destroying the stable atmosphere environment for crystal growth and reducing the crystallization quality. Traditional sublimation systems mostly use static thermal field control, which is difficult to respond to the dynamic changes of convection structure in real time.

[0110] Finally, in the condensation deposition stage, the deposition behavior of sublimation steam is significantly affected by the temperature distribution of the condensation surface and the control of the crystal plane orientation. If the condensation condition is asymmetric, it will lead to non-uniform grain size, unbalanced stress of the deposition layer, and even cause "backflow" phenomenon, which destroys the crystallization homogeneity. However, the existing technology still lacks effective characterization and feedback control methods for condensation symmetry.

[0111] Embodiment 1

[0112] A method for optimizing sublimation purification parameters of high-purity molybdenum trioxide, comprising the following steps:

[0113] Step one, test the particle size distribution of the molybdenum trioxide raw material before sublimation, use laser particle size analyzer, image particle size analysis or SEM image analysis method to obtain the particle size distribution spectrum, particle size distribution skewness value S and particle size distribution kurtosis value K of the raw material, and construct the initial particle size state factor Gsize; when Gsize≥0.7 and K≥3, it means that the current molybdenum trioxide raw material is pre-monitored qualified, and it is summarized as the first pre-qualified raw material, which enters step two;

[0114] Step two, take a picture of the first pre-qualified raw material, obtain the crystal image of the raw material, and construct the feature vector of the particle after image preprocessing of the crystal image of the raw material, and output the class label through the trained classifier KNN, and calculate the percentage of the kth class label particles in the total particle number, obtain the proportion of spherical particles, flaky particles, needle / cylindrical particles and agglomerated particles, when the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of agglomerated particles is <10%, it means that it is qualified, which enters step three;

[0115] Step three, in the initial stage of sublimation, collect the temperature field distribution, gas phase concentration distribution and crystal plane structure of the sublimation wall in the sublimation cavity, and construct the boundary layer structure model under the initial sublimation condition; combined with three-dimensional particle tracking technology, obtain the near-wall diffusion rate curve of molybdenum trioxide vapor, and construct the concentration gradient and the corresponding steam diffusion flux J of the sublimation wall boundary layer to judge whether there is steam accumulation in the boundary layer, whether there is mass transfer bottleneck risk, if there is, generate the first micro dynamic correction strategy and instruction, until there is no mass transfer bottleneck risk, enter step four;

[0116] Step four, using thermal imaging matrix, flow field tracking and sublimation vapor concentration inversion technology, capture the change of heat convection path and intensity in the sublimation cavity, construct the thermal convection displacement vector map, and calculate the flow field stability index F TCV ; preset stability threshold F, when F TCV > F, the system executes local temperature zone fine-tuning instructions, and then enters step five;

[0117] Step five, during the condensation deposition stage, through condensation surface temperature field scanning and morphology evolution map monitoring, identify the sublimation vapor deposition non-uniformity characteristics, and construct the sublimation-condensation asymmetry factor ASC, and preset the symmetry critical threshold X, when ASC > X, adjust the condensation symmetry by dynamically adjusting the condensation surface temperature distribution and crystal face guiding direction, until crystallization.

[0118] In this embodiment, the method realizes digital evaluation of the particle size distribution of the raw material by constructing the raw material particle size state factor Gsize through quantitative analysis of the particle size distribution skewness value S and the particle size distribution kurtosis value K, which can effectively identify whether the particle size distribution is concentrated and reasonable before sublimation, and screen out abnormal raw materials with excessively large or small particle sizes, thereby significantly reducing the phenomenon of uneven vapor release caused by particle size difference in the sublimation process and improving the sublimation stability. By introducing image recognition and KNN classification model, the proportion of spherical particles is further screened and agglomerates are removed. By calculating the proportion of spherical particles, flaky particles, needle-shaped particles and agglomerated particles, the proportion of spherical particles is maintained at a high level, and the proportion of agglomerated particles is controlled below the preset threshold, which can reduce local disturbance during vapor release and improve the stability and consistency of crystal nucleation in the mass transfer process.

[0119] In the initial stage of sublimation, by monitoring the vapor concentration gradient and vapor diffusion flux J in the near-wall region, the accumulation trend of vapor in the boundary layer can be effectively identified, and the sublimation conditions can be adjusted in time, so as to avoid the problems of reduced sublimation rate and recondensation caused by mass transfer bottleneck. During the sublimation process, by monitoring the change of the heat convection path and the flow field stability index F TCV , the size can be adjusted before the heat convection appears drift and turbulence trend, so as to ensure the stability of the vapor atmosphere and reduce the risk of defects in the crystallization process. In the condensation stage, by analyzing the spatial distribution of sublimation vapor deposition and the temperature gradient of the condensation surface, the sublimation-condensation asymmetry factor ASC is quantified, and when ASC exceeds the critical value, the condensation surface temperature and crystal face orientation are corrected, which effectively improves the consistency of grain size and structural integrity.

[0120] Embodiment 2

[0121] This embodiment is an explanation and description in embodiment 1, specifically, step one includes:

[0122] S1, test the particle size distribution of the molybdenum trioxide raw material before sublimation, use a laser particle size analyzer, image particle size analysis or SEM image analysis method to obtain the particle size distribution spectrum and skewness, kurtosis statistical characteristics of the raw material, and construct the initial particle size state factor Gsize;

[0123] The specific steps for obtaining the initial particle size state factor Gsize are as follows:

[0124] S11, the particle size distribution spectrum and skewness, kurtosis statistical characteristics of the raw material are obtained to obtain the 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S and particle size distribution kurtosis value K;

[0125] S121, the particle size distribution skewness value S is obtained in the following manner:

[0126]

[0127] wherein n represents the total number of particle size data points, represents the i-th particle size center value, the representative particle size of each particle size interval, including 1.2 μm and 1.6 μm, which is derived from the particle size analysis equipment; represents the average particle size, represents the standard deviation, represents the particle size The corresponding volume fraction is a normalized value, for example, the particles of 1.2 μm account for 5.3% of the total volume → f i = 0.053; if the volume fraction of 1.6 μm particles is 8.2% → f i = 0.082;

[0128] S122, evaluate the particle size distribution skewness value S to obtain a first evaluation result, including:

[0129] When S = 0, it indicates that the distribution is completely symmetrical with respect to the average value, and the left and right sides are mirror images.

[0130] When S > 0, it indicates that the distribution is right-biased with respect to the average value, and the probability of large particle size is high, with a "tail" on the right side.

[0131] When S < 0, it indicates that the distribution is left-biased with respect to the average value, and the proportion of small particle size is large, with a "tail" on the left side.

[0132] Because the third central moment of the normal distribution is 0, 0 is taken as the symmetrical reference benchmark.

[0133] In this embodiment, the introduction of the skewness value S of the particle size distribution quantitatively judges the asymmetry of the particle size distribution of the raw material, which can effectively identify whether the small or large particles in the raw material dominate, and help to avoid the problem of imbalance of steam release rate caused by extreme deviation of particle size in the sublimation process. By introducing the skewness value S, a third-order statistical quantity, the accuracy of particle size evaluation is further improved based on the traditional D50, which can screen the sublimation raw material from the distribution form level, reduce the disturbance of abnormal distribution on the subsequent sublimation thermal field and gas field, and improve the consistency of crystallization.

[0134] Embodiment 3

[0135] This embodiment is an explanation and description in embodiment 2. Specifically, step one further includes:

[0136] S123, the acquisition method of the kurtosis value K of the particle size distribution is:

[0137]

[0138] The meaning of the formula is that the standardized value of the fourth central moment reflects the peak sharpness and tail thickness of the distribution; the kurtosis value K of the particle size distribution is the fourth power of the distance, and the influence of "extreme values far from the average value" in the distribution is greater, which can judge whether the distribution is sharp or flat;

[0139] S124, evaluating the kurtosis value K of the particle size distribution to obtain a second evaluation result, including:

[0140] The kurtosis value K of the particle size distribution is the fourth power of the distance, and the influence of "extreme values far from the average value" in the distribution is greater, which can judge whether the distribution is sharp or flat. Taking normal distribution (i.e. ideal bell curve) as a reference, the kurtosis value is fixed at 3, which is also the standard distribution reference line commonly used in the industry;

[0141] When K=3, it means that the normal distribution is moderate, the tail thickness is moderate, and there is no obvious concentration or dispersion deviation;

[0142] When K>3, it means that the distribution is concentrated around the average value, the tail is heavier, and the distribution is sharp, which is good for controlling the uniformity of sublimation;

[0143] When K<3, it means that the distribution is dispersed around the average value, the tail is more dispersed, and the distribution is flat; the distribution is flat, and the mass transfer is limited in the sublimation stage;

[0144] The 1st-4th particle size distributions are shown in Table 1:

[0145]

[0146] S125, combine the 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S and particle size distribution kurtosis value K obtained in S121-S123, after normalization processing, obtain the initial particle size state factor Gsize by weighted combination:

[0147]

[0148] In the formula, indicates the distribution width or particle size range of the particle size, reflecting the dispersibility; , and indicated as a weight, , , , and ;

[0149] S126, evaluate the initial particle size state factor Gsize, if the following two judgment conditions are met at the same time, including: if Gsize<0.7, and K<3, indicating that the distribution is unqualified, which will cause the abnormal risk of subsequent sublimation stage, then trigger the first early warning instruction, including: screening, adjusting the ratio, ultrasonic vibration or mechanical crushing of raw materials, to promote the deagglomeration of raw material particles;

[0150] When Gsize≥0.7, and K≥3, it means that the current molybdenum trioxide raw material pre-monitoring is qualified, and it is summarized as the first pre-qualified raw material and enters step two. K=3 is the standard kurtosis of normal distribution, which is a general reference value in statistics and particle engineering, representing moderate particle size concentration, moderate tail thickness, and no skewness influence. When: K>3: the particle size is highly concentrated, with a sharp peak distribution, indicating that most of the particle sizes are close to the average value. This distribution is more conducive to sublimation uniformity and mass transfer stability; K<3: indicates that the particle size distribution is flat, and there are more extreme values away from the mean value, indicating that the raw material dispersibility is too high, and the particle inconsistency is strong, which is not conducive to the gasification consistency in the sublimation process, and is easy to appear "false sublimation" or local incomplete sublimation phenomenon.

[0151] Gsize is a normalized index obtained by D10, D50, D90, skewness S, kurtosis K weighted combination, which represents the overall rationality and concentration of the particle size distribution. Set 0.7 as the judgment line, the Gsize normalization value range is generally [0, 1] or close to the range, 0.7 is a high intermediate critical point, which represents the overall rationality of the particle size structure and has a certain concentration; engineering experience value and test verification results show that when Gsize is less than 0.7, the particle size triternate interval (D90-D10) is usually large, the distribution width is high, which indicates that the particle size span of the raw material particles is large and the particle distribution is chaotic; below 0.7 usually means that the raw material particles have obvious agglomeration, uneven mixing or incomplete crushing during preparation, which will directly affect the steam release rate and gasification interface stability in the sublimation process; 0.7 as a fault tolerance boundary can be adjusted according to the tolerance of different equipment, but in most industrial applications, Gsize≥0.7 is the basic sublimable raw material particle size requirement threshold.

[0152] In this embodiment, by introducing the joint evaluation mechanism of particle size distribution skewness value S and kurtosis value K, the symmetry and concentration of the particle size distribution in the raw material can be quantitatively identified, the uneven steam release phenomenon caused by particle size right bias (S>0) or flat distribution (K<3) can be avoided, and the steam stability of the sublimation area can be effectively improved. By setting the pre-screening threshold of Gsize≥0.7 and K≥3, the quantitative grading of the sublimation adaptability of the raw material can be realized, the raw material batches with extremely uneven particle size distribution can be excluded in advance, and the controllability and uniformity of the mass transfer process in the sublimation process can be ensured. The skewness S, kurtosis K, and particle size triternate D10 / D50 / D90 are normalized and an initial particle size state factor Gsize is constructed, which can systematically reflect the overall state of the particle size distribution, replace the single particle size judgment of the traditional experience method, and improve the intelligence and consistency of the quality control link before sublimation. When Gsize<0.7 and K<3, the first warning instruction is triggered, and operations such as ultrasonic vibration and mechanical crushing can be taken to scatter the agglomerates and improve the particle size structure, so as to avoid the "false volatilization" phenomenon caused by large particle size or irregular agglomeration and improve the gasification efficiency in the sublimation stage.

[0153] Embodiment 4

[0154] This embodiment is an explanation and description in embodiment 1. Specifically, step two includes:

[0155] S21, photographing the first prequalified raw material to obtain a raw material crystal image, image preprocessing the raw material crystal image, including: after the raw material crystal image is grayed, binarized, background noise is removed, and the edge contrast is enhanced, the single particle boundary contour is extracted, and the particle segmentation is completed; and each particle contour is closed, and non-target areas are removed;

[0156] The circularity feature, the aspect ratio feature, the boundary irregularity feature and the area ratio feature are extracted for each particle contour to obtain the circularity of the particle , the aspect ratio , the boundary irregularity , and the ratio of the projection area and the convex hull area ;

[0157] S2101, the circularity of the particle The calculation expression is:

[0158]

[0159] Wherein, A is the projection area of the particle, P is the boundary perimeter of the particle, and the circularity value closer to 1 indicates that the particle is more regular and close to spherical;

[0160] S2102, the aspect ratio of the particle The calculation expression is:

[0161]

[0162] Wherein, L and W are the length of the long axis and the length of the short axis of the fitted circumscribed rectangle of the particle, respectively, reflecting the stretching degree of the particle;

[0163] S2103, the boundary irregularity of the particle The calculation expression is:

[0164]

[0165] Wherein, the greater the value, the more complex the boundary and the more irregular the morphology;

[0166] S2104, the ratio of the projection area and the convex hull area of the particle The calculation expression is:

[0167]

[0168] Wherein, The convex hull area of the particle is used to judge whether there is agglomeration or adhesion;

[0169] S22, for example, the four characteristic values of a certain particle are:

[0170] Circularity = 0.95;

[0171] Aspect ratio = 1.2;

[0172] Boundary irregularity = 1.3;

[0173] The ratio of the projection area and the convex hull area of the particle = 0.92;

[0174] The characteristic vector of the particle is :

[0175]

[0176] The feature vector of each particle is calculated Input into the trained classifier KNN, and the feature vector of the particle is calculated by the trained classifier KNN The Euclidean distance between all training sample feature vectors is calculated, the first k nearest neighbors are found, and the class that appears most frequently in the k neighbors is counted to obtain the corresponding class, and the class label is output.

[0177] The feature vector of each particle is calculated Input into the trained classifier KNN, and the feature vector of the particle is calculated by the trained classifier KNN

[0178] Sample 1: Category = Spherical; Sample 2: Category = Flaky;

[0179] Sample 3: Category = Needle-shaped / Cylindrical; Sample 4: Category = Agglomerated mass;

[0180] The feature vector of each particle is calculated by the trained classifier KNN The Euclidean distance between all training sample feature vectors is calculated, the first k nearest neighbors are found, and the class that appears most frequently in the k neighbors is counted to obtain the corresponding class, and the class label is output.

[0181] The class label includes: 001 Spherical, 002 Flaky, 003 Needle-shaped / Cylindrical, and 004 Agglomerated;

[0182] The total number of particles of the kth class label in the first prequalified raw material is counted, and the percentage of particles of the kth class label in the total number of particles is calculated to obtain the spherical particle ratio, flaky particle ratio, needle-shaped / cylindrical ratio, and agglomerated particle ratio.

[0183] When the spherical particle ratio in the first prequalified raw material is ≥ 50% and the agglomerated particle ratio is < 10%, it is qualified and enters step three; otherwise, it is unqualified, i.e. "insufficient spherical" or "too much agglomeration", indicating that it needs to be returned and re-screened.

[0184] Spherical particles have the smallest specific surface area and the optimal spatial packing characteristics, which can form a dense and uniform raw material bed during the packing process, thereby promoting uniform heat conduction in the bed and reducing unstable phenomena such as crystal crust or explosion caused by local overheating. Setting the proportion of spherical particles ≥ 50% helps to ensure smooth and stable sublimation interface and improve sublimation yield and crystal quality. Agglomerated particles often have a large volume and complex morphology, which can easily form internal incomplete sublimation, uneven surface vaporization or structural collapse during sublimation, and in severe cases, can cause sublimation path blockage or local overheating. Setting the proportion of agglomerated particles < 10% is to control the consistency of raw materials during mass transfer and vaporization, reduce the risk of abnormal sublimation, and ensure continuous and stable sublimation process. According to statistical analysis of the morphology distribution data of molybdenum trioxide and other sublimation materials, the proportion of spherical particles of general production grade raw materials is usually between 40% and 70% after one screening and particle regulation. Setting the proportion ≥ 50% is to ensure that the median level of morphology deviation reaches the controllable region, rather than occasional or local improvement. According to the comparative experiments of multiple batches of raw materials sublimation, when the proportion of spherical particles exceeds 50%, the sublimation temperature control curve is more stable, the average particle size fluctuation of the finished crystal is reduced by 10% to 15%, and the sublimation yield is increased by about 5% on average; while the proportion of agglomerated particles exceeds 10%, abnormal phenomena such as bed deformation and product mixing are easily occurred in the middle and late stages of sublimation. Therefore, this determination method is derived from multiple rounds of process verification results, and is an effective strategy for quantitative morphology screening before sublimation.

[0185] The following is an example of particle sample label output category marking, as shown in Table 2 below:

[0186]

[0187] Spherical samples have a higher roundness (close to 1), a long-diameter ratio close to 1, a lower boundary irregularity, and a high area ratio. The roundness of flaky samples is medium to low, the long-diameter ratio is slightly larger, the boundary irregularity is medium, and the area ratio is low. The roundness of needle-shaped / cylindrical samples is slightly high, but the long-diameter ratio is significantly larger, the boundary irregularity is higher, and the area ratio is moderate. The roundness of agglomerated mass samples is lower, the long-diameter ratio is not too high, but the boundary irregularity is the largest, and the area ratio is the lowest.

[0188] In this embodiment, the image processing and feature extraction method is adopted to avoid the subjective errors caused by manual judgment, and can objectively and accurately identify typical particle morphologies such as spherical, flaky, acicular and agglomerated. By calculating the proportion of particles with different morphologies, a standardized morphology evaluation mechanism is established, especially for setting double constraint conditions for the proportion of spherical particles ≥ 50% and the proportion of agglomerated particles < 10%, to ensure that the raw material has good morphology uniformity and dispersity during sublimation, which helps to control the stability of sublimation interface and the uniformity of sublimation rate. If the particle morphology does not meet the standard, such as "insufficient spherical" or "too much agglomeration", the screening or deagglomeration process can be triggered before sublimation to avoid problems such as local fusion, crust or uneven sublimation caused by abnormal raw material morphology.

[0189] Embodiment 5

[0190] Secondly, in the sublimation process, when the vapor molecules approach the sublimation wall, a high concentration boundary layer will be formed. If the vapor accumulation in the boundary layer is not identified and intervened in time, the "mass transfer bottleneck" phenomenon will easily occur, which will lead to the decrease of gas phase concentration gradient and sublimation rate, and induce local recondensation or crystallization defects. This problem is often ignored in traditional processes, and there is a lack of quantitative monitoring of the concentration distribution and diffusion flux of the boundary layer.

[0191] This embodiment is an explanation and description in embodiment 1. Specifically, step three includes:

[0192] S311, in the initial stage of sublimation, the temperature field distribution in the sublimation cavity is collected by using a temperature sensor array and an infrared thermal imaging system;

[0193] S312, the gas phase molybdenum trioxide concentration distribution in the sublimation cavity is collected in combination with a gas phase concentration sensor or a Raman / infrared spectrometer;

[0194] S313, the crystal face structure of the sublimation wall is obtained by in-situ surface diffraction or reflection electron diffraction means;

[0195] S314, the motion trajectory of molybdenum trioxide vapor particles in the sublimation process is tracked by combining three-dimensional particle tracking velocimetry 3D-PTV, and the velocity field, diffusion path and time sequence of the vapor particles in the boundary layer are extracted;

[0196] S315, in the sublimation process of molybdenum trioxide, the raw material solid is directly converted into vapor at high temperature, and the thin layer region between the sublimation wall and the main gas flow is marked as the boundary layer, in which the concentration, temperature and flow rate will change significantly, and it is also the most "difficult" place for gas phase material diffusion;

[0197] S316, construct the vapor diffusion rate curve of the boundary layer, and calculate the concentration gradient of the sublimation wall boundary layer according to the vapor diffusion rate curve of the boundary layer :

[0198]

[0199] Wherein, C is the concentration of molybdenum trioxide vapor in the boundary layer, and x is the direction perpendicular to the sublimation wall surface; The derivative of concentration with respect to position reflects "how fast the vapor concentration drops from the wall outward"; The numerical value of the concentration gradient of the sublimation wall boundary layer.

[0200] S317, according to Fick's first law, the vapor diffusion flux J is proportional to the concentration gradient of the sublimation wall boundary layer, and the expression is as follows:

[0201]

[0202] In the formula, D represents the diffusion coefficient, including:

[0203] Under the condition of sublimation in normal pressure air, the sublimation temperature range of molybdenum trioxide is 700-850°C, the corresponding temperature is 973-1123K, and the pressure is 1.0atm. The diffusion coefficient D of this condition is: 0.08-0.12cm² / s;

[0204] Under the condition of sublimation in reduced pressure air, the sublimation temperature range of molybdenum trioxide is 600-750°C, the corresponding temperature is 873-1023, and the pressure is 0.1atm. The diffusion coefficient D of this condition is: 0.60-0.90cm² / s;

[0205] Under the condition of sublimation in Ar or N2 inert gas protection, the sublimation temperature range of molybdenum trioxide is 650-800°C, the corresponding temperature is 923-1073, and the pressure is 1.0atm. The diffusion coefficient D of this condition is: 0.10-0.15cm² / s;

[0206] Under the condition of sublimation in vacuum, the sublimation temperature range of molybdenum trioxide is 550-700°C, the corresponding temperature is 823-973K, and the pressure is 10 -4 atm. The diffusion coefficient D of this condition is: 10-25cm² / s, and increases with temperature;

[0207] The negative sign indicates that the substance moves from a place with high concentration to a place with low concentration.

[0208] S318, preset critical threshold G1 and diffusion flux efficiency threshold J1, when the concentration gradient of the sublimation wall boundary layer When the critical threshold G1 is set, and the vapor diffusion flux J < the diffusion flux efficiency threshold J1, it indicates that the vapor is accumulated in the boundary layer, and there is a risk of mass transfer bottleneck. The first micro-dynamic correction strategy is generated, including: the change amplitude of temperature gradient is adjusted by 0.5-2°C each time, and the vibration amplitude of the boundary layer is increased by 1-2µm each time, the local surface temperature is increased to enhance the evaporation speed and diffusion rate of the vapor, or the temperature distribution uniformity is adjusted to reduce the accumulation of local high concentration; the purpose of vibration is to disturb the boundary layer gas, reduce the concentration accumulation, promote gas diffusion and ventilation; and after stabilization, the concentration gradient of the sublimation wall boundary layer and the vapor diffusion flux are re-detected until ≤G1, and the vapor diffusion flux J≥J1, then step four is entered; when ≤G1, and the vapor diffusion flux J≥J1, it indicates that the mass transfer condition is normal, and the vapor diffusion flux efficiency meets the standard, and the sublimation state is continuously monitored.

[0209] The critical upper limit G1 of the concentration gradient indicates that when it exceeds this value, gas phase accumulation or uneven sublimation may occur in the system. If the concentration gradient in the boundary layer is too large, it means that the vapor concentration changes sharply per unit length, and the vapor is accumulated in the boundary layer, which has the risk of mass transfer "blockage". After statistical analysis of multiple groups of data, it is found that: for example, G1=0.25-0.35 mol / cm 4 , crystal segregation, agglomeration, and irregular growth phenomena often occur, indicating that gas diffusion is limited and accumulation is obvious;

[0210] Under different atmosphere conditions (vacuum, inert gas, normal pressure), combined with the typical value range of the diffusion coefficient D (such as 0.1-25 cm² / s), the J value under each working condition is simulated / measured; at the same time, the corresponding product crystallization situation (whether there are defects, agglomeration, wall sticking, etc.) is evaluated to determine the "minimum effective diffusion flux"; for example, under vacuum sublimation conditions (D is larger), the corresponding effective minimum diffusion flux J1≈0.02-0.05 mol / (cm2⋅s); under normal pressure inert gas conditions, due to smaller D, the effective J1 will be lower;

[0211] In this embodiment, in the sublimation starting heating stage: heating causes the molybdenum trioxide solid to sublimate into steam, and the steam migrates to the condensation area. Through the temperature sensor array, the infrared thermal imaging system, the gas phase concentration sensor, and the in-situ crystal face detection means, the temperature field, the gas phase concentration field, and the crystal face structure distribution inside the sublimation cavity are systematically constructed, and the dynamic analysis of the steam diffusion path and the velocity field is realized by combining three-dimensional particle tracking velocimetry (3D-PTV), so as to accurately extract the concentration gradient of the boundary layer of the sublimation wall surface. This enables the system to timely identify the mass transfer bottleneck risk caused by steam accumulation, and solves the problem of uncontrollable "black box" state of the boundary layer in the traditional sublimation process. Using the concentration gradient threshold (G1) and the diffusion flux efficiency threshold (J1) as the dual judgment conditions for the evaluation of the mass transfer efficiency, and combining the Fick's first law principle for dynamic judgment, the non-ideal diffusion state of "high concentration gradient but insufficient diffusion flux" can be effectively identified. This mechanism avoids the misjudgment problem caused by single parameter judgment, and makes the judgment of the mass transfer state more scientific and stable. When the system detects that the boundary layer appears mass transfer bottleneck, the first micro dynamic correction strategy is generated, which adjusts the temperature gradient of the sublimation wall surface by a small amplitude, including the change amplitude of the temperature gradient of each adjustment is 0.5℃~2℃ and the vibration amplitude of the boundary layer is increased by 1µm~2µm each time, which effectively promotes the diffusion of gas molecules and the gas exchange efficiency of the boundary layer without destroying the overall sublimation thermal field structure. After the implementation of the first micro dynamic correction strategy, the concentration gradient and diffusion flux are re-acquired and judged to form a complete closed-loop control logic, which ensures that each correction has monitoring basis and stability confirmation, thereby effectively preventing process deviation caused by over-correction or misjudgment. Compared with the traditional large-amplitude temperature jump or mechanical vibration method, this strategy has the advantages of small perturbation range, high control precision, and not easy to cause new thermal inhomogeneity or crystal segregation.

[0212] The diffusion coefficient D constructed in the present application covers various typical sublimation environments such as normal pressure, reduced pressure, inert gas protection and vacuum, and by setting different diffusion coefficient ranges D under different atmosphere conditions, the ability of the system to adapt to different sublimation process scenarios is improved, and the system has wide engineering applicability and universality.

[0213] Embodiment 6

[0214] In addition, complex thermal capillary convection phenomena are often formed in the sublimation cavity due to temperature difference, especially when the steam concentration is high and the gas flow viscosity changes significantly, local thermal convection paths may drift or reverse flow, which destroys the stable atmosphere environment for crystal growth and reduces the crystallization quality. Traditional sublimation systems mostly use static thermal field control, which is difficult to realize real-time sensing and response to the dynamic changes of the convection structure.

[0215] The present embodiment is an explanation and description in embodiment 1, specifically, the step four comprises:

[0216] S41, monitor the temperature and airflow changes in the sublimation cavity using thermal imaging and airflow sensors to obtain thermal field data and flow field data in the cavity;

[0217] S42, construct a thermal convection offset vector map in the sublimation cavity and calculate a flow field stability index F based on the obtained thermal field data and flow field data in the cavity, combined with sublimation vapor concentration inversion information. TCV :

[0218]

[0219] wherein N represents the total number of sampling points, represents the airflow velocity vector of the jth point, represents the average velocity vector, represents the temperature of the jth point, represents the average temperature, represents the weight coefficient of temperature fluctuation; The setting range of includes:

[0220] In the initial coarse adjustment stage, a = 1.0, considering both speed and temperature fluctuation;

[0221] In the fine adjustment monitoring stage: a = 1.5 ~ 2.0, strengthening the temperature disturbance detection capability;

[0222] In the local condensation and re-crystallization stage, a = 2.5;

[0223] represents the dimensionless relative speed fluctuation, represents the dimensionless relative temperature fluctuation;

[0224] S43, preset a stability threshold F, and judge whether the flow field stability index F TCV exceeds the set stability threshold F, if it does, trigger a local temperature zone fine adjustment instruction to suppress abnormal thermal capillary convection, the local temperature zone fine adjustment instruction includes: arranging a cold source on the back or side of the temperature zone, performing directional thermal adjustment of 0.2 ~ 0.8 ℃, and according to the local airflow direction disturbance, triggering control of the guide vane adjustment to close the direction ± 5° ~ 20°; to break the local circulation or deflection path, so that the airflow restores symmetry and straightness;

[0225] If the flow field stability index F TCV does not exceed the stability threshold F, maintain the current parameters, continue to monitor, and enter step five.

[0226] When using tools such as Ansys Fluent, COMSOL, etc. to perform CFD numerical simulation on the sublimation cavity, through the operation of multiple thermal flow coupling models, the following rules can be obtained: F TCV≤ stable threshold F, the flow field is stable and symmetric, and the sublimation vapor is uniformly distributed; F TCV > stable threshold F, the cavity appears obvious thermal convection, boundary layer disturbance, and is easy to induce condensation, flow deviation or re-crystallization. In a typical sublimation cavity size (such as diameter 100 mm, height 200 mm), center temperature 750℃, wall temperature 700℃, the stable threshold F ≈ 0.35~0.45;

[0227] F TCV Exceeding threshold F represents that the local flow field of the cavity is unstable, in order to avoid vapor accumulation, flow deviation and re-crystallization, triggering local temperature area fine tuning instructions: heat adjustment (0.2~0.8℃): slowly breaking thermal convection driving flow; guide vane adjustment (±5°~20°): reshaping airflow direction, guiding its symmetric backflow, and inhibiting vortex or short circuit path; temperature reduction area control cold source: forming temperature difference wall, blocking convection path.

[0228] In this embodiment, in the thermal convection and flow field stability stage: there is convection in the cavity due to temperature difference, which affects the flow and distribution of vapor. Based on the flow field stability index F TCV , by introducing temperature fluctuation weight regulation mechanism, real-time discrimination of thermal convection and flow deviation trend of sublimation cavity is realized. When F TCV exceeds the stable threshold F, local heat regulation and guide path correction are automatically triggered, which effectively suppresses the vapor flow deviation and condensation re-crystallization problem, and significantly improves the mass transfer efficiency and crystal morphology uniformity of the sublimation process.

[0229] Embodiment 7

[0230] Finally, in the condensation deposition stage, the deposition behavior of sublimation vapor is significantly affected by the temperature distribution of the condensation surface and the control of the crystal face orientation. If the condensation condition is asymmetric, it will lead to non-uniform grain size, unbalanced stress of the deposition layer, and even induce "re-condensation backflow" phenomenon, which destroys the crystalline homogeneity. However, the existing technology still lacks effective characterization and feedback control methods for condensation symmetry.

[0231] This embodiment is an explanation and description in embodiment 1. Specifically, the step five comprises:

[0232] S51, using a high-resolution infrared thermal imager and a morphology evolution atlas acquisition device, continuously scanning the condensation surface to obtain the following parameters, and constructing a condensation surface deposition uniformity data set, including: condensation surface temperature distribution standard deviation, condensation crystal thickness distribution, and condensation crystal deposition morphology image;

[0233] S52, according to the condensation surface deposition uniformity data set, constructing a sublimation-condensation asymmetry factor ASC, which is used to quantify the imbalance degree of the deposition distribution and the thermal field distribution on the condensation surface, and the expression is:

[0234]

[0235] In the formula, standard deviation of the thickness of the crystal, average value of the thickness of the crystal, standard deviation of the temperature distribution of the condensation surface, average value of the temperature field during condensation, symmetry deviation of the topographic image;

[0236] wherein, the acquisition method is:

[0237] the symmetry deviation of the topographic image is obtained by calculating the image difference using the mean square error (MSE) method after mirroring the condensation crystal deposition topographic image left and right or up and down along the center axis of the condensation surface

[0238]

[0239] wherein, M represents the total number of pixel points in the condensation crystal deposition topographic image, represents the pixel value of the kth pixel point in the condensation crystal deposition topographic image, represents the pixel value corresponding to the kth pixel point after mirroring left and right or up and down along the center axis of the condensation surface; 、 and are weights, , , , and ;

[0240] if the two images are completely symmetrical, the pixel values at the symmetrical positions are completely the same, and all , , indicating perfect symmetry, if the difference between the symmetrical pixel values is large, the greater the value, the more asymmetrical the topography.

[0241] S53, a preset symmetry critical threshold X, when >X, indicating that the current condensation deposition has an asymmetry risk, triggering a second warning instruction; when ≤X, indicating that the current condensation deposition process is qualified, and the crystal structure is uniform and has no back condensation and backflow signs, then marking the current batch of crystallization process as qualified, and continuing to monitor;

[0242] Through analysis of >200 groups of crystal deposition process data, combined with the measured parameters of crystal products and failure analysis, it is found that:

[0243] ​When ASC≤0.15, the crystal grains are uniform, stable and highly symmetrical, the product is excellent and can be passed in batches;

[0244] When 0.15<ASC≤0.25, the local crystal grains are uneven, slightly stressed, and moderately deviated in symmetry;

[0245] When ASC>0.25, the crystal grains are obviously recondensed, the structure is distorted and the symmetry is seriously unbalanced, and the batch is determined to be unqualified;

[0246] Therefore, the crystal structure instability and recondensation phenomenon mainly occur when ASC>0.25;

[0247] In actual industrial applications, too sensitive threshold values (such as X=0.15) can cause frequent false alarms and affect process efficiency, so a compromise principle of stability and response sensitivity is combined;

[0248] X=0.20~0.25 is set as the risk control critical value interval.

[0249] S54, after receiving the second early warning instruction, a second micro dynamic correction strategy is generated, the temperature in the asymmetric region is reduced by 0.5~2℃, the deposition is promoted to move to the symmetric region, and the crystal seed direction is rotated or offset, the crystal is guided to grow along a more stable symmetric direction, the sublimation- condensation asymmetry factor ASC is recalculated after each adjustment, and it is judged whether it falls back to ≤X. If not, continue the second micro dynamic correction strategy until the deposition returns to the symmetric state.

[0250] In this embodiment, in the condensation deposition stage: the vapor is cooled and deposited into a solid on the condensation surface to form crystal grains. The sublimation-condensation asymmetry factor ASC is a comprehensive index for measuring the symmetry of the heat-mass distribution of the condensation surface, which can effectively identify the symmetry problems caused by heat field deviation, uneven crystal morphology, etc. in the crystal growth process. When ASC exceeds the preset threshold value X, the second early warning instruction is automatically triggered, and the local temperature reduction and crystal seed orientation fine adjustment strategy is used to realize the dynamic repair of the condensation behavior, thereby significantly improving the uniformity and structural stability of the crystal deposition and reducing the occurrence of abnormal conditions such as “recondensation backflow”.

[0251] The size of the threshold value is set for easy comparison. The size of the threshold value depends on the amount of sample data and the base number set by the person skilled in the art for each group of sample data; as long as it does not affect the proportional relationship of the parameters and the quantized numerical values.

[0252] The above formulas are obtained by collecting a large amount of data for software simulation and selecting a formula close to the true value, and the coefficients in the formula are set by the person skilled in the art according to the actual situation. The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can make equivalent replacement or change according to the technical scheme and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for optimizing parameters in the sublimation purification of high-purity molybdenum trioxide, characterized in that, Includes the following steps: Step 1: Perform particle size distribution testing on the molybdenum trioxide raw material before sublimation. Use laser particle size analyzer, image particle size analysis, or SEM image analysis to obtain the particle size distribution spectrum, particle size distribution skewness value S, and particle size distribution kurtosis value K of the raw material, and construct and obtain the initial particle size state factor Gsize; when Gsize≥0.7 and K≥3, it means that the current molybdenum trioxide raw material is qualified for pre-monitoring, and is summarized as the first pre-qualified raw material, and proceed to Step 2; Step 2: Take a picture of the first pre-qualified raw material to obtain the raw material crystal image. After image preprocessing of the raw material crystal image, construct the feature vector of the particles, and output the category label through the trained classifier KNN. Calculate the percentage of particles with the kth category label to the total number of particles, and obtain the proportion of spherical particles, plate-like particles, needle-like / columnar particles, and aggregated particles. When the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of aggregated particles is <10%, it is considered qualified, and proceed to Step 3. Step 3: In the initial stage of sublimation, the temperature field distribution, gas phase concentration distribution, and crystal plane structure of the sublimation wall are collected within the sublimation chamber to construct a boundary layer structure model under initial sublimation conditions. Using 3D particle tracking technology, the near-wall diffusion rate curve of molybdenum trioxide vapor is obtained to construct the concentration gradient of the sublimation wall boundary layer. And the corresponding steam diffusion flux J, to determine whether there is steam accumulation in the boundary layer and the risk of mass transfer bottleneck. If there is, generate the first micro-dynamic correction strategy and give instructions until there is no risk of mass transfer bottleneck, and then proceed to step four. Step 4: Using thermal imaging matrix, flow field tracking, and sublimation vapor concentration inversion techniques, capture the thermocapillary convection path and intensity changes within the sublimation cavity, construct a thermal convection migration vector map, and calculate the flow field stability index F. TCV A preset stability threshold F is established when F... TCV >F, the system executes the local temperature zone fine-tuning command, and then proceeds to step five; Step 5: During the condensation deposition stage, the non-uniformity of sublimation vapor deposition is identified by scanning the temperature field of the condensation surface and monitoring the morphological evolution spectrum. The sublimation-condensation asymmetry factor (ASC) is constructed, and a symmetry critical threshold X is preset. When ASC > X, the condensation symmetry is corrected by dynamically adjusting the temperature distribution of the condensation surface and the crystal plane guidance direction until crystallization occurs.

2. The method for optimizing parameters of high-purity molybdenum trioxide sublimation purification according to claim 1, characterized in that, Step one includes: S1. The particle size distribution of the molybdenum trioxide raw material before sublimation is tested. The particle size distribution spectrum of the raw material and its skewness and kurtosis statistical characteristics are obtained by using a laser particle size analyzer, image particle size analysis or SEM image analysis method, and the initial particle size state factor Gsize is constructed. The specific steps for obtaining the initial particle size state factor Gsize are as follows: S11. The particle size distribution spectrum of the raw material and its skewness and kurtosis statistical characteristics, to obtain the 10% cumulative volume particle size D10, 50% cumulative volume particle size D50, 90% cumulative volume particle size D90, particle size distribution skewness value S and particle size distribution kurtosis value K. S121, The method for obtaining the particle size distribution skewness value S is as follows: Where n represents the total number of granular data points, This represents the center value of the i-th particle size, and the representative particle size for each particle size range, including 1.2 μm and 1.6 μm; derived from particle size analysis equipment. Indicates the average particle size. Indicates standard deviation, Indicates particle size The corresponding volume fraction is the normalized value; S122. Evaluate the particle size distribution skewness value S to obtain the first evaluation result, including: When S = 0, it means that the distribution is completely symmetrical with respect to the mean; When S > 0, it indicates that the distribution is right-skewed relative to the mean; When S < 0, it indicates that the distribution is skewed to the left relative to the mean.

3. The method for optimizing parameters of high-purity molybdenum trioxide sublimation purification according to claim 2, characterized in that, Step one also includes: S123, The method for obtaining the kurtosis value K of the particle size distribution is as follows: The formula means that the standardized value of the fourth central moment reflects the peakness and tail thickness of the distribution; the kurtosis value K of the particle size distribution is the fourth power of the distance, and the "extreme values ​​far from the mean" in the distribution have a greater impact, which is used to determine whether the distribution is peaked or flat. S124. Evaluate the kurtosis value K of the particle size distribution to obtain a second evaluation result, including: When K=3, it indicates that the normal distribution is moderate, the tail thickness is moderate, and there is no concentrated or dispersed shift. When K > 3, it indicates that the distribution is highly concentrated near the average value, with a heavier tail, forming a peak distribution. The particle size uniformity is acceptable, which is beneficial for controlling the sublimation uniformity. When K < 3, it indicates that the distribution is less concentrated near the mean, with more dispersed tails, forming a flat-topped distribution; the distribution is flat, and mass transfer is limited and fluctuates greatly during the sublimation stage. S125. Combining the 10% cumulative volumetric particle size D10, 50% cumulative volumetric particle size D50, 90% cumulative volumetric particle size D90, particle size distribution skewness value S, and particle size distribution kurtosis value K obtained from S121-S123, after normalization, the initial particle size state factor Gsize is obtained through weighted combination: In the formula, It indicates the width or range of particle size distribution, reflecting dispersibility; , and Represented as weights, , , ,and ; S126. Evaluate the initial particle size state factor Gsize. If the following two judgment conditions are met at the same time, including: if Gsize < 0.7 and K < 3, it means that the distribution is unqualified, which will lead to abnormal risks in the subsequent sublimation stage. Then, the first warning instruction is triggered, including: screening the raw materials, adjusting the ratio, ultrasonic vibration or mechanical crushing to promote the deagglomeration of raw material particles. When Gsize≥0.7 and K≥3, it indicates that the current molybdenum trioxide raw material pre-monitoring is qualified, and it is summarized as the first pre-qualified raw material, proceeding to step two.

4. The method for optimizing parameters of high-purity molybdenum trioxide sublimation purification according to claim 3, characterized in that, Step two includes: S21. Take a picture of the first pre-qualified raw material to obtain a raw material crystal image, and perform image preprocessing on the raw material crystal image, including: converting the raw material crystal image to grayscale, binarizing, removing background noise and enhancing edge contrast, extracting the single particle boundary contour, and completing particle segmentation; and performing closure processing on each particle contour to remove non-target areas. For each particle's outline, extract roundness features, aspect ratio features, boundary irregularity features, and area ratio features to obtain the particle's roundness. Aspect Ratio Boundary irregularity and the ratio of projected area to convex hull area ; S2101, Particle roundness The calculation expression is: Where A is the projected area of ​​the particle, P is the perimeter of the particle's boundary, and the roundness value closer to 1 indicates that the particle is more regular and closer to a sphere. S2102, Aspect Ratio of Particles The calculation expression is: Where L and W are the major and minor axis lengths of the bounding rectangle fitted to the particle, respectively, reflecting the degree of stretching of the particle; S2103, Particle boundary irregularity The calculation expression is: S2104, the ratio of the particle's projected area to its convex hull area. The calculation expression is: in, This represents the convex hull area of ​​the particle, used to determine whether there is aggregation or adhesion; S22, the roundness of the particles Aspect Ratio Boundary irregularity and the ratio of projected area to convex hull area The eigenvectors of particles are composed of the following in sequence: The feature vector of each particle The particle's feature vector is calculated using the trained classifier KNN. The Euclidean distance between the feature vectors of all training samples is used to find the k nearest neighbors, and the category that appears most frequently among these k neighbors is counted to obtain the corresponding category and output the category label. Category labels include: 001 spherical, 002 flaky, 003 needle-like / columnar, and 004 aggregated. Count the total number of particles in the kth category label in the first pre-qualified raw material, and calculate the percentage of particles in the kth category label to the total number of particles, to obtain the proportion of spherical particles, the proportion of sheet-like particles, the proportion of needle-like / columnar particles, and the proportion of agglomerated particles. If the proportion of spherical particles in the first pre-qualified raw material is ≥50% and the proportion of agglomerated particles is <10%, it is considered qualified and proceeds to step three; if the conditions are not met, it is considered unqualified, i.e., "insufficient spherical particles" or "excessive agglomeration", which means that it needs to be returned for re-screening.

5. The method for optimizing parameters of high-purity molybdenum trioxide sublimation purification according to claim 4, characterized in that, Step three includes: S311. In the initial stage of sublimation, the temperature field distribution inside the sublimation cavity is collected using a temperature sensor array and an infrared thermal imaging system. S312. Combine with a gas phase concentration sensor or Raman / infrared spectrometer to collect the concentration distribution of gaseous molybdenum trioxide in the sublimation chamber. S313. Obtain the sublimation wall crystal structure through in-situ surface diffraction or reflected electron diffraction. S314. Combining three-dimensional particle tracking and velocimetry (3D-PTV), the motion trajectory of molybdenum trioxide vapor particles during sublimation is tracked, and the velocity field, diffusion path, and time series of boundary layer vapor particles are extracted. S315. During the sublimation of molybdenum trioxide, the raw material solid is directly converted into vapor at high temperature, and the thin layer between the wall and the main gas flow is sublimated. The thin layer is marked as the boundary layer. S316. Construct the vapor diffusion rate curve of the boundary layer, and calculate the concentration gradient of the sublimation wall boundary layer based on the vapor diffusion rate curve. : Where C is the concentration of molybdenum trioxide vapor in the boundary layer, and x is the direction perpendicular to the sublimation wall; It represents the derivative of concentration with respect to position, reflecting "how quickly the steam concentration decreases from the wall outwards"; This represents the numerical value of the concentration gradient in the sublimation wall boundary layer.

6. The method for optimizing parameters of high-purity molybdenum trioxide sublimation purification according to claim 5, characterized in that, Step three also includes: S317. According to Fick's first law, the vapor diffusion flux J is proportional to the concentration gradient of the sublimation wall boundary layer, as expressed below: In the formula, D represents the diffusion coefficient, which includes: Under normal atmospheric pressure and air sublimation conditions, the sublimation temperature range of molybdenum trioxide is 700-850°C, corresponding to a temperature of 973–1123 K. At a pressure of 1.0 atm, the diffusion coefficient D under these conditions is 0.08-0.12 cm² / s. Under reduced pressure air sublimation conditions, the sublimation temperature range of molybdenum trioxide is 600-750°C, corresponding to temperatures of 873–1023°C. At a pressure of 0.1 atm, the diffusion coefficient D under these conditions is 0.60-0.90 cm² / s. Under Ar or N2 inert gas protection sublimation conditions, the sublimation temperature range of molybdenum trioxide is 650-800°C, corresponding to temperatures of 923–1073°C. At a pressure of 1.0 atm, the diffusion coefficient D under these conditions is 0.10-0.15 cm² / s. Under vacuum sublimation conditions, molybdenum trioxide sublimates at a temperature range of 550–700°C, corresponding to temperatures of 823–973 K, at a pressure of 10. -4 Atm, the diffusion coefficient D under this condition is 10–25 cm² / s, and it increases with increasing temperature; A negative sign indicates that a substance moves from a place with a high concentration to a place with a low concentration.

7. The method for optimizing parameters of high-purity molybdenum trioxide sublimation purification according to claim 6, characterized in that, Step three also includes: S318, preset critical threshold G1 and diffusion flux efficiency threshold J1, when the concentration gradient of the sublimation wall boundary layer... >When the critical threshold G1 is set, and the steam diffusion flux J < the diffusion flux efficiency threshold J1, it indicates that steam is accumulating in the boundary layer and there is a risk of mass transfer bottleneck. The first micro-dynamic correction strategy is generated, including: adjusting the temperature gradient by 0.5℃~2℃ each time and increasing the vibration amplitude of the boundary layer by 1µm~2µm each time. After the first micro-amplitude dynamic correction strategy is implemented, and after stabilization is achieved, the concentration gradient and vapor diffusion flux of the sublimation wall boundary layer are re-detected until... When ≤G1, and the steam diffusion flux J≥J1, then proceed to step four; when When ≤G1 and the steam diffusion flux J≥J1, it indicates that the mass transfer conditions are normal, the steam diffusion flux efficiency meets the standard, and the sublimation state is continuously monitored.

8. The method for optimizing parameters of high-purity molybdenum trioxide sublimation purification according to claim 7, characterized in that, Step four includes: S41. Using thermal imaging and airflow sensors, monitor the temperature and airflow changes inside the sublimation cavity to obtain thermal field data and flow field data inside the cavity. S42. Based on the obtained thermal and flow field data within the cavity, and combined with the sublimation vapor concentration inversion information, construct a thermal convection migration vector diagram within the sublimation cavity, and calculate the flow field stability index F. TCV : Where N represents the total number of sampling points, This represents the airflow velocity vector at point j. Represents the average velocity vector. This represents the temperature at point j. Indicates average temperature. Weighting coefficients representing temperature fluctuations; The settings range includes: This represents a dimensionless relative velocity fluctuation. This represents dimensionless relative temperature fluctuations. S43. Preset stability threshold F to determine the flow field stability index F TCV Whether the set stable threshold F is exceeded, if it is exceeded → trigger the local temperature zone fine-tuning command to suppress abnormal thermocapillary convection. The local temperature zone fine-tuning command includes: deploying a cold source on the back or side of the temperature zone, performing directional thermal adjustment of 0.2~0.8℃, and triggering the control guide vane to adjust the closing direction ±5°~20° according to the local airflow direction disturbance in the temperature zone circle. If the flow field stability index F TCV If the stable threshold F is not exceeded, maintain the current parameters, continue monitoring, and proceed to step five.

9. The method for optimizing parameters of high-purity molybdenum trioxide sublimation purification according to claim 8, characterized in that, Step five includes: S51. Using a high-resolution infrared thermal imager and a morphology evolution map acquisition device, the condensation surface is continuously scanned to obtain the following parameters and construct a uniform deposition dataset of the condensation surface, including: standard deviation of the temperature distribution of the condensation surface, thickness distribution of the condensation crystals, and morphology images of the condensation crystal deposition. S52. Based on the uniform deposition dataset on the condensation surface, a sublimation-condensation asymmetry factor (ASC) is constructed to quantify the degree of imbalance between the deposition distribution and the thermal field distribution on the condensation surface. The expression is: In the formula, The standard deviation of crystal thickness This represents the average thickness of the crystal. The standard deviation of the temperature distribution on the condensation surface. This represents the average value of the temperature field during the condensation process. Indicates the degree of deviation of the symmetry of the topographic image; , and Represented as weights, , , ,and ; in, The method of obtaining it is: For images of condensed crystal deposition morphology, mirror them horizontally or vertically along the central axis of the condensation surface, and use the mean square error (MSE) method to calculate the image differences and obtain the symmetry deviation of the morphology image. : Where M represents the total number of pixels in the image of condensed crystal deposition morphology. This represents the pixel value of the k-th pixel in the image of condensed crystal deposition morphology. This represents the pixel value corresponding to the kth pixel after mirroring the image horizontally or vertically along the central axis of the condensation surface. If two images are perfectly symmetrical, and the pixel values ​​at symmetrical positions are exactly the same, then all ,so This indicates perfect symmetry. If the difference in pixel values ​​between symmetrical pixels is large, then... The larger the value, the more asymmetrical the shape.

10. The method for optimizing parameters of high-purity molybdenum trioxide sublimation purification according to claim 9, characterized in that, Step five also includes: S53. Preset a symmetrical critical threshold X, when... If the value is greater than X, it indicates that there is an asymmetric risk in the current condensation deposition, triggering the second early warning instruction; when If the value is ≤X, it indicates that the current condensation and deposition process is qualified, and the crystal structure is uniform with no signs of backflow. The current batch crystallization process is then marked as qualified and will be continuously monitored. S54. After receiving the second early warning command, generate a second micro-dynamic correction strategy, reduce the temperature by 0.5 to 2°C in the asymmetric region to promote the migration of the deposit to the symmetric region, and guide the crystal to grow along a more stable symmetric direction by rotating or shifting the seed direction. After each adjustment, recalculate the sublimation-condensation asymmetry factor ASC to determine whether it has fallen back to ≤X. If it has not fallen back, continue the second micro-dynamic correction strategy until the deposition returns to a symmetric state.

Images