Data enhancement prediction method for thermoelastic quality factor of micro-hemispherical resonator

Abstract

The invention relates to the technical field of data processing, in particular to a data enhancement prediction method for the thermoelastic quality factor of a micro-hemispherical resonator, which comprises the following steps: S1, determining the radius size of the manufactured micro-hemispherical resonator, determining the simulation value range of the radius and thickness of an anchor point of the resonator, and determining the simulation value range of the thickness of the resonator; harmonic oscillators with different geometric boundary dimensions are established through high-temperature blow molding simulation; s2, thermodynamic simulation is carried out on the harmonic oscillator appearance obtained through blow molding simulation, and a harmonic oscillator thermoelastic quality factor data set is established; s3, performing data enhancement on the thermoelastic quality factor data set by adopting a cubic polynomial fitting method; and S4, training the enhanced thermoelastic quality factor data set by using a deep learning model to obtain a harmonic oscillator thermoelastic quality factor prediction model. According to the method, the bottleneck of scarcity of high-precision simulation data is broken through, unification of high precision and high efficiency is realized, the robustness and reliability of a prediction model are improved, and the method has wide engineering applicability and popularization value.

Description

Technical Field

[0001] This invention relates to the field of data prediction and processing technology, and in particular to a data enhancement prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator. Background Technology

[0002] As a high-precision inertial sensor based on the Coriolis effect, the core performance of a micro-hemispherical resonator gyroscope largely depends on the thermoelastic quality factor of the micro-hemispherical resonator. This parameter directly characterizes the energy retention capability of the resonator under thermoelastic dissipation mechanisms and is a key indicator determining the gyroscope's limiting sensitivity. The geometric parameters of the resonator, such as the anchor radius, thickness, and height, significantly affect its thermoelastic quality factor. However, in actual manufacturing processes, due to factors such as material anisotropy and high-temperature blow molding tolerances, the resonator's geometry deviates from the ideal state, making it difficult for traditional analytical models or simplified equivalent models to accurately predict the energy dissipation characteristics of actual resonators. Currently, a fast and universal prediction theory applicable to arbitrary geometric parameters is lacking.

[0003] In existing technologies, although high-performance finite element simulation can obtain accurate solutions through multi-physics coupling calculations, it has inherent limitations: First, parametric modeling and single-step solution are too time-consuming, resulting in low efficiency of large-scale parameter scanning and severely restricting the breadth of exploration of the design space; second, simulation results are easily affected by factors such as mesh generation, introducing data noise and affecting model accuracy; finally, the cost of obtaining a sufficient number of high-precision simulation samples is extremely high, causing prediction models based on traditional deep learning methods to suffer from insufficient learning and poor generalization ability due to a lack of training data, making it difficult to meet the reliability requirements of precision design.

[0004] Furthermore, current technological development in the field of micro-hemispherical harmonic oscillators is mostly focused on processing technology (such as the publicly disclosed technologies CN117142749B and CN114105075B), without addressing the issue of rapid prediction of the key performance indicator, the thermoelastic quality factor.

[0005] Therefore, there is an urgent need in this field for an intelligent method that can combine the geometric characteristics of the harmonic oscillator to achieve high-precision and low-time prediction of the thermoelastic quality factor, so as to break through the bottleneck of traditional reliance on high-cost simulation and experience-based design. Summary of the Invention

[0006] The main objective of this invention is to overcome the shortcomings of the prior art and provide a data-enhanced prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator.

[0007] The technical solution adopted by this invention to achieve its technical objective is: a data enhancement prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator, comprising the following steps: S1. Determine the radius of the manufactured micro-hemispherical resonator, determine the simulation range of the anchor point radius and thickness of the resonator, and establish resonators with different geometric dimensions through high-temperature blow molding simulation; S2. Perform thermodynamic simulation on the shape of the harmonic oscillator obtained by blow molding simulation, and establish a dataset of the thermoelastic quality factor of the harmonic oscillator. S3. Data augmentation of the thermoelasticity quality factor dataset is performed using a cubic polynomial fitting method. S4. A deep learning model is used to train the enhanced thermoelastic quality factor dataset to obtain a prediction model for the thermoelastic quality factor of the harmonic oscillator.

[0008] Preferably, step S1 includes: S101. Determine the design radius of the target micro-hemispherical harmonic oscillator; S102. Determine the simulation value range for the anchor point radius and thickness, where the total number of possible anchor point radius values ​​is... The total number of thickness values ​​is ,get Group geometric parameters; S103. Perform high-temperature blow molding simulation on the ANSYS POLYFLOW platform, set pressure load, heat flow load, contact interface conditions, material parameters, and use the Fulcher equation to model the viscosity of molten silica. S104. For each combination of anchor point radius and thickness, extract... The geometric shapes of harmonic oscillators at different heights were obtained. The geometry of a harmonic oscillator.

[0009] Preferably, step S2 includes: S201, will The geometry of each harmonic oscillator was imported into COMSOL software. Configurations with incomplete anchor points or distorted boundaries were eliminated, and the remaining configurations were preserved. Effective geometric shape of the group; S202. Establish a multiphysics coupling model of solid mechanics and solid heat transfer in COMSOL, setting the anchor point as a fixed constraint and the room temperature as 293.15 K. Simulation calculations yield... Group thermoelasticity quality factor data.

[0010] Preferably, step S3 includes: S301. For each combination of anchor point radius and thickness, if its effective number of harmonic oscillators... Then perform data augmentation; if If so, the original data will be retained; S302, Regarding the number of effective harmonic oscillators The combination of these factors was used to fit the relationship between the height of the harmonic oscillator and the thermoelastic quality factor using a cubic polynomial, and 100 fitting data points were generated at equal intervals between the minimum and maximum height values. S303, will The original data and The fitted data were merged to construct a data-enhanced set of thermoelastic quality factors.

[0011] Preferably, in step S4, the input of the deep learning model is the anchor point radius, thickness, and height of the harmonic oscillator, and the output is the thermoelastic quality factor.

[0012] Preferably, the deep learning model is a CNN-Transformer model.

[0013] Preferably, the particle swarm optimization algorithm is used to optimize the hyperparameters of the CNN-Transformer model.

[0014] Preferably, the Fulcher equation is: ; in, The viscosity coefficient of fused silica is . It is the thermodynamic temperature.

[0015] Preferably, the heat exchange boundary conditions in the blow molding simulation are as follows: ; in, For interfacial heat flux density, For mold temperature, The interfacial heat transfer coupling coefficient ranges from 400 to 800 W / (m²). 2 ·K).

[0016] Preferably, the thermoelastic quality factor characterizes the energy retention capability of the harmonic oscillator under the dominant thermoelastic dissipation mechanism.

[0017] The working principle of this data augmentation prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator is as follows: Harmonic oscillator shapes of different geometric dimensions are generated through high-temperature blow molding simulation, and an initial thermoelastic quality factor dataset is constructed by combining this with thermodynamic simulation. To address the issues of small sample size and high noise in finite element simulation, cubic polynomial fitting is used to augment the effective samples, expanding the sample size and smoothing data noise. Based on the augmented dataset, a CNN-Transformer deep learning model is used to establish a nonlinear mapping relationship from the harmonic oscillator's geometric parameters (anchor radius, thickness, and height) to the thermoelastic quality factor, achieving high-precision and high-efficiency prediction of the quality factor.

[0018] Compared with the prior art, the beneficial effects of the present invention are: This data augmentation prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator breaks through the bottleneck of scarce high-precision simulation data. Through the innovative strategy of "finite element simulation + data augmentation", it effectively expands and smooths the initial small sample dataset by using cubic polynomial fitting, generating large-scale, high-quality training samples at extremely low computational cost. This fundamentally solves the core problem of overfitting and weak generalization ability of deep learning models due to insufficient samples.

[0019] This data-augmented prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator achieves a balance between high accuracy and high efficiency. Leveraging the powerful nonlinear mapping capabilities of deep neural networks (such as CNN-Transformer), it establishes an end-to-end rapid prediction model from geometric parameters to the quality factor. Once trained, this model can complete predictions within milliseconds, replacing finite element simulations that can take hours or even days, significantly shortening the design cycle and improving R&D efficiency.

[0020] This data augmentation prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator improves the robustness and reliability of the prediction model. By introducing a data augmentation strategy, not only is the number of samples increased, but the inherent data noise in finite element simulation is also effectively suppressed by fitting physical laws. This makes the prediction results of the trained deep learning model more stable and reliable, and can accurately capture the complex physical relationship between geometric parameters and the thermoelastic quality factor.

[0021] This data-enhanced prediction method for the thermoelastic quality factor of microhemispherical resonators possesses broad engineering applicability and promotional value. It provides an intelligent design tool for performance prediction and structural optimization of microhemispherical resonators, enabling rapid evaluation of the thermoelastic quality factor under different design parameters and guiding the design of low-loss resonators. This method exhibits high modeling accuracy, strong generalization ability, and excellent computational efficiency, making it particularly suitable for rapid performance evaluation and optimization of MEMS devices under multi-parameter constraints. It is of great significance for improving the R&D efficiency and engineering application level of the entire MEMS field. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating the operational steps of a data-enhanced prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator.

[0024] Figure 2A system flowchart for constructing a dataset to enhance the thermoelasticity quality factor.

[0025] Figure 3 A comparison chart showing the effects of different fitting methods for data augmentation.

[0026] Figure 4 This is a comparison chart of the predictions made by the augmented data model and the unaugmented data model. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. However, it should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0028] In the description of this invention, it should be noted that when an element is referred to as being "fixed to" or "set on" another element, it can be directly on or indirectly on the other element. When an element is referred to as being "connected to" another element, it can be directly connected to or indirectly connected to the other element.

[0029] In the description of this invention, it should be noted that the terms "center," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. "Several" means one or more, unless otherwise explicitly specified.

[0030] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0031] Example 1: Please see the appendix Figures 1-2 A data-enhanced prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator, the specific steps of which are as follows: S1. Determine the radius of the manufactured micro-hemispherical resonator, determine the simulation range of the anchor point radius and thickness of the resonator, and establish resonators with different geometric dimensions through blow molding simulation.

[0032] Furthermore, in this embodiment, the specific solution for step S1 is as follows: S101. Determine the design radius of the target micro-hemispherical harmonic oscillator.

[0033] S102. Determine the simulation range of the anchor point radius and thickness of the micro-hemispherical resonator in the critical height simulation experiment. The total number of possible anchor point radius values ​​is... The total number of thickness values ​​is You can get A set of geometric parameters that satisfy: .

[0034] S103. High-Temperature Blow Molding Simulation Parameter Settings. The numerical simulation of the blow molding process is based on the ANSYS POLYFLOW platform, using a two-dimensional axisymmetric model. In the simulation settings, the upper surface of the fused silica sheet must simultaneously bear the downward pressure load and the heat flow load; its lower surface is defined as the contact interface with the graphite mold surface to simulate the contact behavior between the two during the actual molding process. To reproduce the clamping conditions in real tooling, fixed displacement constraints are applied to the lateral boundaries of the fused silica sheet and the corresponding areas of the graphite mold. In addition, to ensure the accuracy of the simulation results, the constitutive parameters of various materials of the fused silica and the graphite mold are clearly defined. These parameters cover key physical properties such as density, elastic modulus, thermal conductivity, Poisson's ratio, and coefficient of thermal expansion.

[0035] To accurately model the nonlinear viscosity of fused silica with temperature, the Fulcher equation is used, and the relationship is as follows: ; in, Let be the viscosity coefficient of fused silica, with dimensions in Pascals. Seconds (Pa·s); This is the thermodynamic temperature of the system, expressed in Kelvin (K).

[0036] The heat transfer behavior between the two is simulated by defining heat exchange boundary conditions at the contact interface between the fused silica and the graphite mold: ; Among them, parameters , , These correspond to the interfacial heat flux density, mold temperature, and interfacial heat transfer coupling coefficient, respectively. The value needs to be determined in conjunction with the specific process, and its conventional range is 400W / (m³). 2 ·K) to 800W / (m 2 ·K).

[0037] S104. The geometric shapes of harmonic oscillators with different geometric dimensions were obtained through high-temperature blow molding simulation. For each set of anchor point radii and thickness simulations, since blow molding is a time-series simulation, the height of the harmonic oscillator increases with the blowing time. After each set of anchor point radii and thickness simulations, multiple harmonic oscillators of different heights can be extracted. To facilitate subsequent data enhancement and reduce the time consumption of subsequent thermodynamic finite element simulations, the blow molding simulation of each set of anchor point radii and thickness combinations was extracted. ( ( ) harmonic oscillators at different heights. In the end, a total of The geometric shape of a harmonic oscillator satisfies: .

[0038] S2. Perform thermodynamic simulation on the shape of the harmonic oscillator obtained by blow molding simulation, and establish a dataset of the thermoelastic quality factor of the harmonic oscillator.

[0039] Furthermore, in this embodiment, the specific solution for step S2 is as follows: S201, Cleaning of the harmonic oscillator geometry set. The dataset of harmonic oscillator geometry was imported into COMSOL software. Due to the viscoelastic flow of molten silica at high temperatures, the inclined interface formed in the anchor region caused incomplete anchor points or boundary distortion in some simulated configurations, thus failing to meet the requirements of actual assembly and integration. Therefore, these physically unrealizable data need to be removed to ensure the engineering validity of subsequent datasets. After screening, the remaining usable geometries are... ( )Group.

[0040] S202. Construct a thermoelastic quality factor dataset through thermodynamic simulation. Establish a multiphysics coupled field of solid mechanics and solid heat transfer in COMSOL software, and import the geometric shape dataset into COMSOL software all at once. Set the anchor point of the harmonic oscillator to a fixed constraint and room temperature of 293.15 K, and simulate to calculate the thermoelastic quality factor under the working modes of the harmonic oscillator. The corresponding simulation results are obtained. A dataset of thermoelastic quality factors corresponding to a set of geometric shapes.

[0041] S3. Use cubic polynomial fitting to enhance the thermoelasticity quality factor dataset.

[0042] Furthermore, in this embodiment, the specific solution for step S3 is as follows: S301. Select the objects for which data augmentation is needed. After the blow molding simulation ends, the blow molding simulation for each combination of anchor point radius and thickness is captured. ( There were several resonators at different heights. However, after importing the COMSOL data, some resonators failed to meet the actual assembly and integration requirements and were discarded, resulting in a decrease in the final number of resonators under certain combinations of anchor point radii and thicknesses. .

[0043] like If the sample data is too small, using polynomial fitting will produce a severe "Runge phenomenon," causing the curve to oscillate wildly between points, resulting in complete distortion. Therefore, for The combination of anchor point radius and thickness is not fitted to data augmentation, for Enhance the fitted data.

[0044] S302. Data augmentation is performed using cubic polynomial fitting. For the anchor point radius and thickness combination requiring data augmentation, a cubic polynomial fitting is performed on the harmonic oscillator height and the corresponding thermoelastic quality factor for that group. After fitting, using the minimum and maximum values ​​of the harmonic oscillator height as intervals, 100 harmonic oscillator heights and fitted thermoelastic quality factors are generated at equal intervals based on the results of the cubic fitting, replacing the original values. The height of each harmonic oscillator and the corresponding simulated thermoelastic quality factor.

[0045] S303, Employs a data-enhanced set of thermoelasticity quality factors. [This will] include all... Data on the height of the resonator under the combination of anchor point radius and thickness, and the corresponding simulated thermoelastic quality factor (total). (one), with all The heights of 100 harmonic oscillators and the fitted thermoelastic quality factors (total) obtained from the combination of anchor point radius and thickness. (1), constructing a data-enhanced thermoelasticity quality factor set.

[0046] The process for constructing the thermoelasticity quality factor enhancement dataset is detailed in the appendix. Figure 2 .

[0047] S4. A deep learning model is used to train the enhanced thermoelastic quality factor dataset to obtain a prediction model for the thermoelastic quality factor of the harmonic oscillator.

[0048] Furthermore, in this embodiment, the specific solution for step S4 is as follows: A deep learning algorithm is used to train a set of data-enhanced thermoelastic quality factors. The input is the radius, thickness and height of the resonator anchor point, and the output is the thermoelastic quality factor, thus obtaining a predictive model for the thermoelastic quality factor of the resonator.

[0049] The specific application process of the data-enhanced prediction method for the thermoelastic quality factor of the micro-hemispherical harmonic oscillator is as follows: First, a harmonic oscillator shape library with different geometric dimensions (including anchor point radius, thickness, and dynamically changing height) is generated through a parametric high-temperature blow molding simulation system. Subsequently, thermodynamic coupling simulation was used to obtain the initial thermoelastic quality factor dataset; To address the issues of sample scarcity and numerical noise in finite element simulation, an innovative cubic polynomial fitting enhancement strategy based on physical laws is introduced, which significantly expands the sample size and improves data quality while preserving the original data distribution characteristics. Finally, based on the enhanced large-sample dataset, an end-to-end prediction model from multidimensional geometric parameters to thermoelastic quality factors is established using a CNN-Transformer hybrid neural network with strong nonlinear mapping capabilities. The network hyperparameters are optimized through particle swarm optimization to ensure the model's generalization ability and prediction accuracy in complex physical fields.

[0050] The specific usage process includes: (1) Parameter setting and simulation preparation: Determine the design radius of the target harmonic oscillator, set the parameter space of the anchor point radius and thickness, and configure the pressure load, heat flow load, material constitutive parameters and Fulcher viscosity model for blow molding simulation on the ANSYS POLYFLOW platform; (2) Multiphysics simulation and data construction: Perform blow molding simulation to obtain the shape of the harmonic oscillator at different heights, import it into COMSOL for geometric cleaning to remove configurations that do not meet the assembly requirements, and calculate the thermoelastic quality factor of each configuration through solid mechanics and solid heat transfer coupled field simulation. (3) Data augmentation: For parameter combinations with ≥4 effective samples, a cubic polynomial is used to fit the relationship between the height of the harmonic oscillator and the quality factor. 100 augmented data points are generated at equal intervals within the extreme value range of the height, and these data are combined with the original small sample data to construct an augmented dataset. (4) Intelligent model training and deployment: With anchor point radius, thickness and height as input features and thermoelastic quality factor as output target, the CNN-Transformer model is trained using the enhanced dataset. The network hyperparameters are optimized by combining particle swarm optimization algorithm. Finally, an intelligent model that can quickly and accurately predict thermoelastic quality factor is obtained, providing an efficient and reliable solution for the structural design and performance optimization of micro-hemispherical harmonic oscillators.

[0051] Example 2: Please see Figures 3-4 Based on the above embodiments, the data enhancement prediction method for the thermoelastic quality factor of the micro-hemispherical harmonic oscillator is implemented using the following values, as detailed below.

[0052] The critical height and quality factor of a certain type of microhemispherical harmonic oscillator are predicted using a data-enhanced prediction method based on the thermoelastic quality factor of the microhemispherical harmonic oscillator.

[0053] The radius of the microhemispherical resonator is set to 5 mm. The geometric parameters of the microhemispherical resonator in the thermoelastic quality factor simulation experiment are determined: the simulated value ranges for the anchor point radius, thickness, and height. Specifically, the total number of possible anchor point radius values ​​is... Total number of thickness values The range of values ​​can be found in Table 1 below. Group of geometric parameters.

[0054] Table 1: Simulation Value Range of Geometric Parameters for the Microhemispherical Resonator

[0055] During the blow molding simulation, a 1.5 × 10⁻⁶ layer was applied to the upper surface of the fused silica sheet. 6 A heat flux density of W / m² is assumed, and the surface is designed to withstand a pressure of 50 kPa. A heat exchange coupling coefficient of 500 W / (m²) is selected. K), and the relevant physical parameters are detailed in Table 2 below.

[0056] Table 2: Material Parameters for Finite Element Blow Molding

[0057] Simulations were performed using the ANSYS POLYFLOW module, capturing blow molding simulations for each combination of anchor point radii and thicknesses. Several harmonic oscillators at different heights. In the end, a total of [number] can be obtained. The geometry of a harmonic oscillator.

[0058] These models were then imported into COMSOL for further analysis. It was found that 29 models had actual anchor point radii that were too small or had missing anchor point structures, failing to meet the requirements of actual assembly processes. Therefore, these 29 invalid data sets were removed, and the remaining models were retained. A set of valid samples. The anchor point of the harmonic oscillator was set to a fixed constraint and room temperature of 293.15 K. The quality factor of the harmonic oscillator under the working mode was calculated by simulation, and the quality factor dataset was obtained.

[0059] Removing 29 sets of invalid data resulted in a decrease in the final number of resonators under certain combinations of anchor point radii and thicknesses. .like The radius of the anchor point and thickness Combined data does not fit data augmentation, for Augmentation was applied to the fitted data. (Attached) Figure 3 The height of the resonator is shown under four sets of anchor point radii and thicknesses. With thermoelasticity quality factor The data fitting results clearly show that the cubic polynomial fitting yields the highest accuracy. After fitting, using the minimum and maximum values ​​of the harmonic oscillator heights as intervals, 100 harmonic oscillator heights and fitted thermoelastic quality factors are generated at equal intervals based on the cubic fitting results, replacing the original values. The height of each harmonic oscillator and the corresponding simulated thermoelastic quality factor. Finally, if... Unfitted data and The fitted data were merged, resulting in 4107 sets of sample data, thus achieving data augmentation.

[0060] This example uses a CNN-Transformer model to train the enhanced quality factor dataset. The input is... , and The output is Simultaneously, a particle swarm optimization algorithm is used to optimize hyperparameters, ensuring the CNN-Transformer model achieves the best possible training performance. The trained model is then exported, and the unenhanced dataset is used... , and As input, the output results are compared with the finite element results in the unenhanced dataset. For comparison, we trained the CNN-Transformer model on an unenhanced quality factor dataset, also using particle swarm optimization, to derive the trained model's performance on the unenhanced dataset. , and As input, the output results are compared with the finite element results in the unenhanced dataset. Compare them. See appendix. Figure 4 As can be clearly seen from the figure, the enhanced quality factor dataset model outperforms the unenhanced quality factor dataset model in terms of prediction accuracy, proving the effectiveness of the method proposed in this invention.

[0061] The solution in this embodiment can be selectively combined with solutions in other embodiments.

[0062] It should be noted that although the above embodiments have been described herein, this does not limit the scope of patent protection of this invention. Therefore, any changes and modifications made to the embodiments described herein based on the innovative concept of this invention, or equivalent structural, procedural, or functional transformations made using the description and drawings of this invention, directly or indirectly applying the above technical solutions to other related technical fields, are all included within the scope of protection of this invention.

Claims

1. A data-enhanced prediction method for the thermoelastic quality factor of a micro-hemispherical harmonic oscillator, characterized in that, Includes the following steps: S1. Determine the radius of the manufactured micro-hemispherical resonator, determine the simulation range of the anchor point radius and thickness of the resonator, and establish resonators with different geometric dimensions through high-temperature blow molding simulation; S2. Perform thermodynamic simulation on the shape of the harmonic oscillator obtained by blow molding simulation, and establish a dataset of the thermoelastic quality factor of the harmonic oscillator. S3. Data augmentation of the thermoelasticity quality factor dataset is performed using a cubic polynomial fitting method. S4. A deep learning model is used to train the enhanced thermoelastic quality factor dataset to obtain a prediction model for the thermoelastic quality factor of the harmonic oscillator.

2. The data enhancement prediction method for the thermoelastic quality factor of a microhemispherical harmonic oscillator according to claim 1, characterized in that, Step S1 includes: S101. Determine the design radius of the target micro-hemispherical harmonic oscillator; S102. Determine the simulation value range for the anchor point radius and thickness; where the total number of possible anchor point radius values ​​is... The total number of thickness values ​​is ,get Group geometric parameters; S103. Perform high-temperature blow molding simulation on the ANSYS POLYFLOW platform, set pressure load, heat flow load, contact interface conditions, material parameters, and use the Fulcher equation to model the viscosity of molten silica. S104. For each combination of anchor point radius and thickness, extract... The geometric shapes of harmonic oscillators at different heights were obtained. The geometry of a harmonic oscillator.

3. The data enhancement prediction method for the thermoelastic quality factor of a microhemispherical harmonic oscillator according to claim 2, characterized in that, Step S2 includes: S201, will The geometry of each harmonic oscillator was imported into COMSOL software. Configurations with incomplete anchor points or distorted boundaries were eliminated, and the remaining configurations were preserved. Effective geometric shape of the group; S202. Establish a multiphysics coupling model of solid mechanics and solid heat transfer in COMSOL, setting the anchor point as a fixed constraint and the room temperature as 293.15 K. Simulation calculations yield... Group thermoelasticity quality factor data.

4. The data enhancement prediction method for the thermoelastic quality factor of a microhemispherical harmonic oscillator according to claim 3, characterized in that, Step S3 includes: S301. For each combination of anchor point radius and thickness, if its effective number of harmonic oscillators... Then perform data augmentation; if If so, the original data will be retained; S302, Regarding the number of effective harmonic oscillators The combination of these factors was used to fit the relationship between the height of the harmonic oscillator and the thermoelastic quality factor using a cubic polynomial, and 100 fitting data points were generated at equal intervals between the minimum and maximum height values. S303, will The original data and The fitted data were merged to construct a data-enhanced set of thermoelastic quality factors.

5. The data enhancement prediction method for the thermoelastic quality factor of a microhemispherical harmonic oscillator according to claim 4, characterized in that: In step S4, the input to the deep learning model is the anchor point radius, thickness, and height of the harmonic oscillator, and the output is the thermoelastic quality factor.

6. The data enhancement prediction method for the thermoelastic quality factor of a microhemispherical harmonic oscillator according to claim 5, characterized in that: The deep learning model is the CNN-Transformer model.

7. The data enhancement prediction method for the thermoelastic quality factor of a microhemispherical harmonic oscillator according to claim 6, characterized in that: The particle swarm optimization algorithm is used to optimize the hyperparameters of the CNN-Transformer model.

8. The data enhancement prediction method for the thermoelastic quality factor of a microhemispherical harmonic oscillator according to claim 2, characterized in that: The Fulcher equation is: ； in, The viscosity coefficient of fused silica is . It is the thermodynamic temperature.

9. The data enhancement prediction method for the thermoelastic quality factor of a microhemispherical harmonic oscillator according to claim 2, characterized in that: The heat exchange boundary conditions in blow molding simulation are: ； in, For interfacial heat flux density, For mold temperature, denoted as the interface heat transfer coupling coefficient.

Images