Resonator parameter analysis method and system based on artificial intelligence

Abstract

The invention relates to the technical field of resonators, in particular to a resonator parameter analysis method and system based on artificial intelligence, and the method comprises the steps: generating an ideal coupling matrix of a target resonator according to the center frequency, the passband bandwidth and the in-passband return loss; acquiring a corresponding relationship among an external quality factor, a coupling coefficient between resonant cavities, a coupling window width and a resonant cavity length through electromagnetic simulation or theoretical calculation, constructing a training data set, and training a regression model to predict the coupling window width; obtaining an external quality factor and a coupling coefficient of a target frequency band by utilizing baseband-passband transformation, solving the width of each coupling window through a prediction model, calculating the length of a resonant cavity by combining interpolation and scale transformation to form an initial size, establishing a three-dimensional simulation model to extract a coupling matrix, and iteratively correcting the length of the resonant cavity according to diagonal elements, so as to obtain an initial size; and obtaining a final physical size meeting the index. According to the method, the problems that an existing resonator design and parameter analysis process depends on experience, the number of simulation iterations is large, and the parameter reverse solving efficiency is low can be solved.

Description

Technical Field

[0001] This invention relates to the field of resonator technology, and in particular to a resonator parameter analysis method and system based on artificial intelligence. Background Technology

[0002] Microwave waveguide resonators and filters constructed from them are widely used in radar, satellite communications, and 5G / millimeter-wave communications systems. Traditional design processes typically start with a given center frequency, passband bandwidth, and in-band specifications. The target response is synthesized using coupling matrix theory, and then full-wave electromagnetic simulation is used to perform multiple scans and manual adjustments to the resonator dimensions and coupling window sizes to approximate the ideal response. Because electromagnetic simulation itself involves high computational complexity and a high parameter space dimension, designers often rely on experience to select the initial structure and tuning path. The design cycle for high-order, broadband, or multi-frequency resonators is generally long, and the consistency of results among different engineers is poor. Summary of the Invention

[0003] In view of the above technical problems, the present invention provides a resonator parameter analysis method and system based on artificial intelligence, which solves the problems of relying on experience, having a large number of simulation iterations, and low efficiency of parameter inversion in the existing resonator design and parameter analysis process.

[0004] Other features and advantages of this disclosure will become apparent from the following detailed description, or may be learned in part from practice of this disclosure.

[0005] According to one aspect of the present invention, an artificial intelligence-based method for resonator parameter analysis is proposed, the method comprising: Obtain the performance specifications of the target resonator, including the center frequency, passband bandwidth, and passband return loss index, and generate the ideal coupling matrix of the target resonator based on the performance specifications. For standard waveguide resonant cavity structures with multiple preset operating frequencies, the correspondence between external coupling parameters, inter-cavity coupling parameters and physical dimensions is collected through electromagnetic simulation and / or theoretical calculation, forming a training dataset that includes external quality factor, inter-cavity coupling coefficient, coupling window width and resonant cavity length. A regression-based machine learning model is used to train the training dataset, so that the machine learning model takes the operating frequency, the external quality factor and / or the coupling coefficient between resonant cavities as input and the coupling window width as output, to obtain a prediction model for predicting the coupling window width. Based on the ideal coupling matrix and the center frequency and passband bandwidth of the target resonator, a preset baseband-passband transformation relationship is used to convert the external coupling parameters in the ideal coupling matrix into the corresponding external quality factors, and the inter-cavity coupling parameters into the corresponding inter-cavity coupling coefficients. The coupling window widths of the input / output ports of the target resonator and the coupling window widths between the resonators are then predicted by the prediction model. Based on the obtained coupling window widths and the center frequency of the target resonator, and combined with the relationship between the coupling window width and the cavity length in the training dataset, the cavity lengths of each cavity of the target resonator are calculated through interpolation and scaling transformation, thus forming the initial physical dimensions of the target resonator. A three-dimensional electromagnetic simulation model of the target resonator is established based on the initial physical dimensions. The simulated scattering parameters of the target resonator are calculated at preset frequency sampling points. Based on the simulated scattering parameters and the ideal coupling matrix, the parameters of the coupling matrix are adjusted to minimize the objective function of the error between the scattering parameters synthesized based on the current coupling matrix and the simulated scattering parameters, thereby obtaining the extraction coupling matrix corresponding to the initial physical dimensions. Based on the extracted diagonal elements of the coupling matrix, the resonant cavity length of the target resonator is iteratively corrected to obtain the final physical dimensions of the target resonator that meet the performance specifications.

[0006] Furthermore, the external coupling parameters of the training dataset are obtained in the following manner: For a waveguide structure containing only a single resonant cavity and a coupling window on one side, electromagnetic simulation is performed under multiple coupling window widths. The phase change curve of the reflection parameter with frequency is extracted. The group delay near the resonant frequency is calculated based on the derivative of the phase with respect to frequency. The external quality factor is obtained based on the proportional relationship between the group delay and the corresponding resonant angular frequency. Each external quality factor is associated with the corresponding coupling window width and operating frequency and stored in the training dataset.

[0007] Furthermore, the inter-cavity coupling parameters of the training dataset are obtained in the following manner: For a waveguide structure containing two identical resonant cavities connected to each other through a coupling window, electromagnetic simulation is performed at multiple coupling window widths. The two resonant peak frequencies generated by coupling are extracted. The ratio of the square difference to the sum of the squares of the two resonant peak frequencies is determined as the coupling coefficient between the resonant cavities at the corresponding coupling window width. The coupling coefficient between the resonant cavities is then associated with the corresponding coupling window width and operating frequency and stored in the training dataset.

[0008] Furthermore, the resonant cavity length of the training dataset is obtained in the following manner: With a fixed waveguide cross-sectional dimension, a resonant cavity model is established for different combinations of resonant cavity length, coupling window width, and operating frequency. The corresponding resonant frequencies are obtained through simulation. For a given operating frequency, interpolation is performed on the two-dimensional graph of the resonant cavity length and coupling window width to make the resonant frequency equal to the resonant cavity length of the operating frequency. The resonant cavity length, the corresponding coupling window width, and the operating frequency are then associated and stored in the training dataset.

[0009] Furthermore, after obtaining the training dataset, the training dataset is preprocessed, the preprocessing including: The coupling window width is normalized using the width of the standard waveguide in the corresponding frequency band, and the resonant cavity length is normalized using the width or height of the standard waveguide in the corresponding frequency band. Each element of the coupling matrix is ​​scaled according to its corresponding standard waveguide operating frequency to reduce the impact of frequency band differences on machine learning model training.

[0010] Furthermore, the iterative correction of the resonant cavity length of the target resonator specifically includes: The resonant cavity length correction amount is determined based on the degree to which the resonant frequency deviates from the operating frequency as reflected by the diagonal elements. The resonant cavity length is updated while keeping the coupling window width unchanged, and the extraction coupling matrix is ​​repeatedly calculated until the absolute value of the diagonal elements of the extraction coupling matrix is ​​less than a preset threshold.

[0011] Furthermore, in determining the correction amount for the resonant cavity length, the following is included: Each diagonal element is mapped to the normalized deviation of the resonant frequency of the corresponding resonant cavity relative to the operating frequency. The normalized deviation is then converted into a resonant frequency offset based on the proportional relationship between the operating frequency and the passband bandwidth. This offset is determined according to the monotonic relationship between the resonant frequency and the resonant cavity length, and a pre-obtained length. The correction amount for the length of the resonant cavity is calculated from the frequency-correlation curve.

[0012] According to a second aspect of this disclosure, an artificial intelligence-based resonator parameter analysis system is provided, the system comprising: The specification input module is used to obtain the performance specifications of the target resonator, including the center frequency, passband bandwidth, and passband return loss index, and to generate the ideal coupling matrix of the target resonator based on the performance specifications. The dataset construction module is used to collect the correspondence between external coupling parameters, inter-cavity coupling parameters and physical dimensions for standard waveguide resonant cavity structures with multiple preset operating frequencies through electromagnetic simulation and / or theoretical calculation, forming a training dataset that includes external quality factor, inter-cavity coupling coefficient, coupling window width and resonant cavity length. The model training module is used to train the training dataset using a regression machine learning model, so that the machine learning model takes the operating frequency, the external quality factor and / or the coupling coefficient between resonant cavities as input and the coupling window width as output to obtain a prediction model for predicting the coupling window width. The coupling size prediction module is used to convert the external coupling parameters in the ideal coupling matrix into the corresponding external quality factor and the inter-cavity coupling parameters into the corresponding inter-cavity coupling coefficients based on the ideal coupling matrix, the center frequency and the passband bandwidth of the target resonator, and a preset baseband-passband transformation relationship. The prediction model is then used to predict the coupling window width of the input / output port of the target resonator and the coupling window width between the resonators. The initial size calculation module is used to calculate the cavity length of each cavity of the target resonator based on the obtained coupling window widths and the center frequency of the target resonator, combined with the relationship between the coupling window widths and the cavity lengths in the training dataset, through interpolation and scaling transformation, thereby forming the initial physical size of the target resonator. The simulation module is used to establish a three-dimensional electromagnetic simulation model of the target resonator based on the initial physical dimensions, calculate the simulated scattering parameters of the target resonator at preset frequency sampling points, and adjust the parameters of the coupling matrix based on the simulated scattering parameters and the ideal coupling matrix to minimize the objective function of the error between the scattering parameters synthesized based on the current coupling matrix and the simulated scattering parameters, thereby obtaining the extraction coupling matrix corresponding to the initial physical dimensions. The iterative correction module is used to iteratively correct the resonant cavity length of the target resonator based on the diagonal elements of the extracted coupling matrix, so as to obtain the final physical size of the target resonator that meets the performance specifications.

[0013] The technical solution disclosed herein has the following beneficial effects: Compared with existing design methods that rely on human experience and multiple rounds of simulation scanning, this invention constructs data samples under multiple frequency bands, different coupling strengths, and different combinations of physical dimensions. It introduces a regression-type artificial intelligence model to learn the nonlinear mapping relationship between the external quality factor, the coupling coefficient between resonators, and the geometry of the coupling window. After obtaining the electromagnetic coupling parameters of the target resonator, the corresponding coupling window size can be directly predicted, which greatly reduces the number of initial structure guesses and parameter trials, and significantly improves the automation level of resonator parameter analysis and structure synthesis. Attached Figure Description

[0014] Figure 1 This is a flowchart of an artificial intelligence-based resonator parameter analysis method as described in the embodiments of this specification; Figure 2 This is a block diagram of an artificial intelligence-based resonator parameter analysis system as described in the embodiments of this specification. Detailed Implementation

[0015] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, apparatus, steps, etc., can be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0016] Furthermore, the accompanying drawings are merely illustrative of this disclosure. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0017] This invention provides a resonator parameter analysis method based on artificial intelligence for products. (Refer to...) Figure 1 The diagram shown is a flowchart illustrating an artificial intelligence-based resonator parameter analysis method according to an embodiment of the present invention. This method can be applied to devices such as personal computers, servers, tablets, and mobile phones. The method can be executed by a device, which can be implemented by software and / or hardware. Specifically, the method may include the following steps S101-S106: In step S101, the performance specifications of the target resonator are obtained, including the center frequency, passband bandwidth, and passband return loss index. The ideal coupling matrix of the target resonator is generated based on the performance specifications.

[0018] Among them, the center frequency is used to determine the frequency band in which the resonator operates, the passband bandwidth is used to limit the frequency range that signals are allowed to pass through, and the return loss index within the passband reflects the matching level and standing wave ratio within that frequency band, usually using a certain minimum return loss value (e.g., The parameters are given as 15 dB. Based on these parameters, the corresponding low-pass prototype network parameters can be determined according to existing filter synthesis methods, and the ideal coupling relationship of the target resonator in the baseband domain can be obtained on this basis. Specifically, by synthesizing the expected amplitude response (including passband ripple or equivalent return loss requirements), the coupling strength between each resonator and the external coupling strength between the input / output port and adjacent resonators can be calculated. These coupling relationships are represented in matrix form, constituting the ideal coupling matrix M of the target resonator, where the diagonal elements of M characterize the self-coupling characteristics of each resonator, and the off-diagonal elements... The matrix elements representing the normalized coupling strength between the i-th and j-th resonant cavities and connected to the input / output ports. , Equals are used to characterize the strength of external coupling.

[0019] After generating the ideal coupling matrix, it can be viewed as a normalized representation of the coupling relationship in the baseband domain, providing a foundation for establishing a connection with actual operating frequency band parameters. To further explain the physical meaning of each element in the matrix from a performance specification perspective, the center frequency and passband bandwidth can be considered together as the fractional bandwidth (FBW), meaning that the passband parameters, including the center frequency and bandwidth, are collectively referred to as the fractional bandwidth. Based on this, the external coupling elements in the ideal coupling matrix... External quality factors The following relationship exists between them: ; This relation gives the condition of known fractional bandwidth and ideal coupling matrix elements. In this case, how to calculate the external quality factor corresponding to that port illustrates the relationship between external coupling elements and performance specifications in the ideal coupling matrix. Similarly, any off-diagonal element in the ideal coupling matrix... With the actual coupling coefficient in the passband The following conditions must be met: ; This formula gives the transformation relationship between the baseband coupling matrix and the passband coupling coefficient, reflecting how the ideal coupling matrix is ​​mapped to the actual coupling strength in the operating frequency band under a given center frequency and passband bandwidth. Through the above relationship, the ideal coupling matrix generated in step S101 based on the center frequency, passband bandwidth, and return loss index in the passband not only fully characterizes the coupling topology and coupling magnitude of the target resonator in the baseband domain, but also lays the foundation for further deriving the external quality factor and inter-cavity coupling coefficient based on this matrix, and for carrying out parameter analysis and size calculation accordingly.

[0020] In step S102, for standard waveguide resonant cavity structures with multiple preset operating frequencies, the correspondence between external coupling parameters, inter-cavity coupling parameters and physical dimensions is collected through electromagnetic simulation and / or theoretical calculation, forming a training dataset containing external quality factor, inter-cavity coupling coefficient, coupling window width and resonant cavity length.

[0021] As an explanation, in step S102, for multiple preset operating frequency standard waveguide resonant cavity structures, a data sample set related to external coupling, inter-cavity coupling, and cavity length is established. The standard waveguide resonant cavity structure mentioned here can be understood as a standard rectangular waveguide corresponding to different frequency bands (e.g., waveguide cross-sectional dimensions selected according to standard models for different frequency bands), with a single-cavity or dual-cavity resonant structure arranged inside, and rectangular coupling windows opened between cavities or between cavities and ports. By performing electromagnetic simulations and / or theoretical calculations on the above structure at several preset operating frequencies under different combinations of coupling window widths and resonant cavity lengths, the corresponding relationships between parameters such as the external quality factor, inter-cavity coupling coefficient, coupling window width, and resonant cavity length can be obtained point by point. These data are then organized to form a training dataset. Each sample in the training dataset corresponds to a set of correlation values ​​between a specific physical dimension (including coupling window width and resonant cavity length) and electromagnetic parameters (including external quality factor and inter-cavity coupling coefficient) at a specific operating frequency. This provides a data foundation for subsequently using a regression model to learn the mapping relationship between physical dimensions and coupling characteristics.

[0022] The external coupling parameters of the training dataset are obtained as follows: For a waveguide structure containing only a single resonant cavity and a coupling window set on one side, electromagnetic simulation is performed at multiple coupling window widths. The phase-frequency variation curve of the reflection parameters is extracted. The group delay near the resonant frequency is calculated based on the derivative of the phase with respect to the frequency. The external quality factor is obtained based on the proportional relationship between the group delay and the corresponding resonant angular frequency. Each external quality factor is associated with the corresponding coupling window width and operating frequency and stored in the training dataset.

[0023] To obtain the external coupling parameters in the training dataset, a waveguide structure containing only a single resonant cavity and a coupling window on one side can be selected. Electromagnetic simulations can be performed on this structure with multiple different coupling window widths to obtain the corresponding scattering parameters, with a focus on the phase curve of the reflection parameters as a function of frequency. By calculating the derivative of the reflection phase with respect to frequency, the group delay value is obtained near the resonant frequency. Combined with the resonant angular frequency, the external quality factor corresponding to the coupling window width and operating frequency can be calculated. In this embodiment, the external quality factor and group delay are... and resonant angular frequency The following relationship exists between them: ; in, This is the resonant angular frequency of the single-cavity structure at a given coupling window width. This is the group delay obtained from the reflection phase near the resonant frequency. By repeating the above process at multiple preset operating frequencies and multiple combinations of different coupling window widths, a series of... The corresponding data is obtained, and each external quality factor is associated with its corresponding coupling window width and operating frequency and stored in the training dataset, thereby constructing a data subset to describe the variation of the external coupling strength of the input / output ports with the coupling window width and frequency.

[0024] The inter-cavity coupling parameters in the training dataset are obtained as follows: For a waveguide structure containing two identical resonant cavities connected to each other through a coupling window, electromagnetic simulation is performed at multiple coupling window widths. Two resonant peak frequencies generated by coupling are extracted. The ratio of the square difference to the sum of the squares of the two resonant peak frequencies is determined as the inter-cavity coupling coefficient at the corresponding coupling window width. The inter-cavity coupling coefficient is then associated with the corresponding coupling window width and operating frequency and stored in the training dataset.

[0025] To obtain the coupling parameters between resonant cavities in the training dataset, a waveguide structure containing two identical resonant cavities connected to each other through a coupling window can be used, and electromagnetic simulations can be performed with multiple different coupling window widths. Due to the coupling between the two resonant cavities, a pair of split resonance peaks will appear in the frequency response, allowing the extraction of higher resonant frequencies from the simulation results. and lower resonant frequency Inter-cavity coupling coefficient The resonant frequencies can be calculated from these two frequencies according to the following relationship: ; in, The normalized coupling strength between the i-th and j-th resonant cavities is closely related to the coupling window width between the two cavities. Repeating the above simulations and calculations under multiple preset operating frequencies and multiple coupling window widths yields a series of... Data points are collected, and each inter-cavity coupling coefficient is associated with the corresponding coupling window width and operating frequency and stored in the training dataset. In this way, another subset of data is formed in the training dataset, describing the change of inter-cavity coupling strength with coupling window width and frequency, providing a basis for subsequent prediction of the inter-cavity coupling window width based on the target coupling coefficient.

[0026] The resonant cavity length in the training dataset is obtained as follows: with a fixed waveguide cross-sectional size, a resonant cavity model is established for different combinations of resonant cavity length, coupling window width, and operating frequency. The corresponding resonant frequency is obtained through simulation. For a given operating frequency, interpolation is performed on the two-dimensional graph of the resonant cavity length and coupling window width to make the resonant frequency equal to the resonant cavity length of the operating frequency. The resonant cavity length, the corresponding coupling window width, and the operating frequency are then associated and stored in the training dataset.

[0027] To obtain the resonant cavity length in the training dataset, under the premise of fixed waveguide cross-sectional dimensions, corresponding resonant cavity models are established for combinations of different resonant cavity lengths, different coupling window widths, and different operating frequencies. The resonant frequencies under each combination are obtained through electromagnetic simulation. Since the resonant cavity length and operating frequency have a monotonically changing relationship in a standard waveguide, and the presence of the coupling window has a certain influence on the effective resonant length, the curve corresponding to the resonant cavity length and resonant frequency can be regarded as an interpolable function curve under a specific waveguide cross-section and coupling window width. For a given operating frequency, several sampling points are first selected on the two-dimensional plane formed by the resonant cavity length and the coupling window width. The corresponding resonant frequency is obtained through simulation. Then, the length-frequency relationship is interpolated in this two-dimensional graph to find the resonant cavity length value corresponding to the target operating frequency. Based on this, and combined with different coupling window width conditions, a... The training dataset contains data points (width, cavity length, etc.). These cavity lengths are correlated with their corresponding coupling window widths and operating frequencies. This correlation is then stored in the training dataset to obtain the data portion characterizing the cavity length's variation with frequency and coupling window geometry. This provides data support for subsequently determining the cavity length using interpolation and scaling transformations, given a target coupling window width and center frequency.

[0028] In step S103, a regression-based machine learning model is used to train the training dataset, so that the machine learning model takes the operating frequency, the external quality factor and / or the coupling coefficient between resonant cavities as input and the coupling window width as output to obtain a prediction model for predicting the coupling window width.

[0029] In this embodiment, the training dataset constructed in step S102 is used to train a regression-based machine learning model, enabling the model to establish a nonlinear mapping relationship between the operating frequency, external quality factor, and / or inter-cavity coupling coefficient and the corresponding coupling window width. To better model different coupling types, this embodiment splits the regression modeling process into two independent regression models: one model is trained on external coupling data, with the operating frequency and external quality factor as input features and the coupling window width at the input or output port as output; the other model is trained on inter-cavity coupling data, with the operating frequency and inter-cavity coupling coefficient as input features and the coupling window width between adjacent resonators as output. In this way, the mapping relationships from frequency and external quality factor to external coupling window width, and from frequency and coupling coefficient to inter-cavity coupling window width, can be learned separately, allowing the two types of coupling to be handled independently in the model and reducing the complexity of a single model.

[0030] In step S104, based on the ideal coupling matrix and the center frequency and passband bandwidth of the target resonator, a preset baseband-passband transformation relationship is used to convert the external coupling parameters in the ideal coupling matrix into the corresponding external quality factors, and the inter-cavity coupling parameters into the corresponding inter-cavity coupling coefficients. The coupling window widths of the input / output ports of the target resonator and the coupling window widths between the resonators are then predicted by the prediction model.

[0031] As described in step S101 above, based on the obtained ideal coupling matrix of the target resonator, this matrix is ​​considered as a normalized coupling representation in the baseband domain. Then, combining the center frequency and passband bandwidth of the target resonator, the baseband coupling parameters are converted into the external quality factor and inter-cavity coupling coefficient operating under actual passband conditions. To this end, firstly based on the center frequency... The fractional bandwidth (FBW) is determined by the upper and lower edge frequencies of the passband, meaning that the center frequency and bandwidth are considered together as a fractional bandwidth, thus providing a bandpass scale for baseband-passband conversion. Under this fractional bandwidth, for external coupling elements connected to the input or output ports in an ideal coupling matrix... Using a preset baseband-passband conversion relationship, it is converted into an external quality factor corresponding to the external coupling of the port. Given the external quality factor and baseband coupling element. The fractional bandwidth (FBW) and its corresponding function satisfy the following relationship: ; This formula allows for the direct calculation of the required external quality factor for each input / output port, given the fractional bandwidth and the external coupling elements of the ideal coupling matrix. This provides the external coupling parameters in the passband domain for subsequent inverse calculation of the corresponding coupling window width using the prediction model.

[0032] For any two resonant cavities in the ideal coupling matrix, the off-diagonal elements between them... This element characterizes the normalized coupling strength between the i-th and j-th resonators in the baseband domain, and similarly needs to be mapped to the coupling coefficient in the passband domain under the conditions of the target center frequency and passband bandwidth. Therefore, a preset baseband-passband transformation relationship is used to... Convert to passband coupling coefficient The relationship is as follows: ; Through the above transformation, while maintaining the ideal coupling matrix topology, all inter-cavity baseband coupling parameters can be uniformly converted into passband coupling coefficients consistent with the target center frequency and passband bandwidth. Subsequently, the "target operating frequency and external quality factor" are then... The features are input to the regression prediction model trained for external coupling in the aforementioned steps, and the model outputs the input / output coupling window width that matches the external quality factor of each port; simultaneously, the target operating frequency and the coupling coefficient between the resonant cavities are... The features are respectively input into the regression prediction model obtained by training the inter-cavity coupling in the aforementioned steps, and the model outputs the features corresponding to each feature. Matching coupling window widths between resonant cavities. In accordance with the conventions for standard waveguide normalization in the manual, when the prediction model outputs a normalized coupling window width, it can also be inversely normalized according to the standard waveguide width corresponding to the target operating frequency band. This allows obtaining the actual coupling window widths between each input / output port and adjacent resonant cavities in the target resonator structure, realizing the process of calculating the physical coupling structure dimensions from the ideal coupling matrix and target frequency band parameters via baseband-passband transformation and inverse calculation using the prediction model.

[0033] In step S105, based on the obtained coupling window widths and the center frequency of the target resonator, and combined with the relationship between the coupling window width and the resonant cavity length in the training dataset, the resonant cavity length corresponding to each resonant cavity of the target resonator is calculated through interpolation and scaling transformation, thus forming the initial physical dimensions of the target resonator.

[0034] For the standard waveguide resonant cavity structure selected during the training phase, simulations can be performed on combinations of different operating frequencies, coupling window widths, and resonant cavity lengths, while keeping the waveguide cross-sectional dimensions fixed. This yields a series of relationships between the resonant cavity length and the resonant frequency at a given operating frequency and coupling window width. Based on these data, a curve showing the relationship between the resonant cavity length and the coupling window width for a specific reference standard waveguide (e.g., a standard rectangular waveguide of a certain type) can be plotted. The cavity length can be determined based on the operating frequency, while the waveguide height and width are determined by the operating frequencies of the standard waveguides in each frequency band. Therefore, given a standard waveguide cross-section, the resonant cavity length primarily varies with the operating frequency and the coupling window width.

[0035] Based on this, for each target resonant cavity, its corresponding coupling window width and the center frequency of the target resonator are first taken, and then mapped onto the "length-width" curve or table formed by the training data for interpolation calculation. Specifically, a set of data with the same or similar center frequency as the target can be selected from the training data. The coupling window width is regarded as the independent variable, and interpolation is performed on the length-width curve corresponding to that frequency according to the position of the coupling window width to obtain the resonant cavity length matching the target center frequency. For coupling window widths that do not fall completely on the existing sampling points, linear interpolation or higher-order interpolation can be used to interpolate between adjacent samples to obtain a continuous resonant cavity length estimate. Another equivalent implementation is to regard the resonant cavity length as a two-dimensional function of the operating frequency and the coupling window width, and perform two-dimensional interpolation in the three-dimensional data space of length-frequency-width. That is, given the center frequency and the coupling window width, the corresponding resonant cavity length is solved by interpolation based on the neighboring data points. Through the above process of "interpolating the curve based on the coupling window width and determining the cavity length in combination with the center frequency", an initial length value can be given for each resonant cavity of the target resonator.

[0036] Considering that training data is typically acquired on several reference frequency bands or reference standard waveguides, while the center frequency of the actual target resonator may differ from the reference frequency bands, a scaling transformation (frequency scaling) can be performed on the interpolated cavity length to adapt to the target center frequency. For example, the training data can be grouped according to the operating frequency of the standard waveguide, and the relationship between length and coupling window width can be interpolated within each group. Then, combined with the proportional relationship between the target center frequency and the reference frequency band, the interpolation result can be applied with appropriate frequency scaling, so that the cavity lengths obtained in different frequency bands have consistent electrical behavior. In practical applications, the normalized cavity length can be obtained first on the reference waveguide, and then the length can be scaled according to the variation of the standard waveguide size with the operating frequency and the target center frequency to adapt to the target frequency band of the current design. Through the comprehensive processing of the above interpolation and scaling transformation, the lengths corresponding to each stage of the target resonator cavity can be calculated based on the predicted coupling window widths and the center frequency of the target resonator. Together with the corresponding coupling window widths, these lengths constitute the initial physical dimensions of the target resonator, providing a reasonable starting structure for subsequent three-dimensional electromagnetic simulation and iterative optimization.

[0037] In step S106, a three-dimensional electromagnetic simulation model of the target resonator is established based on the initial physical dimensions. The simulated scattering parameters of the target resonator are calculated at preset frequency sampling points. Based on the simulated scattering parameters and the ideal coupling matrix, the parameters of the coupling matrix are adjusted to minimize the objective function of the error between the scattering parameters synthesized based on the current coupling matrix and the simulated scattering parameters, thereby obtaining the extraction coupling matrix corresponding to the initial physical dimensions.

[0038] First, using the initial physical dimensions of the target resonator obtained in step S105, a corresponding electromagnetic simulation model is built in a three-dimensional full-wave simulation environment. Based on the initial physical dimensions, parameters such as the length of each resonator, the width of the coupling window between each input / output port and adjacent resonators, the waveguide cross-section size, and the excitation port are set. Several preset frequency sampling points are selected within a certain range including the center frequency and its vicinity. The three-dimensional model is then solved to obtain the simulated scattering parameters of the target resonator at these sampling frequencies, including at least reflection and transmission parameters. To facilitate comparison with the scattering parameters subsequently synthesized from the coupling matrix, in this embodiment, the scattering parameters synthesized from the current coupling matrix are denoted as... The scattering parameters obtained from the three-dimensional electromagnetic simulation are denoted as... ,in Let i be the i-th frequency sampling point.

[0039] After obtaining the simulated scattering parameters, an objective function is constructed based on the ideal coupling matrix generated in step S101. The elements of the coupling matrix are adjusted to make the scattering parameters synthesized from the current coupling matrix approximate the three-dimensional simulation results as closely as possible. Specifically, this embodiment uses an objective function with reflection parameters and transmission parameter errors as components, and its form is: ; ; ; in, Used to measure the mean squared error between the reflection parameters obtained from the ideal coupling matrix synthesis and the reflection parameters from the 3D simulation. Used to measure the mean squared error between the transmission parameters obtained from the ideal coupling matrix synthesis and the transmission parameters from the 3D simulation. Let be the objective function for the total error. By minimizing this objective function, the coupling matrix corresponding to the current physical size can be extracted from the scattering parameters obtained from the simulation.

[0040] To find the minimum value of the objective function, this embodiment employs an annealing simulation algorithm to optimize the coupling matrix. Specifically, using the ideal coupling matrix as the initial solution, each element (including external coupling elements and inter-cavity coupling elements) is treated as a variable to be optimized. Under a given initial temperature, a small random perturbation is applied to these matrix elements to obtain a new candidate coupling matrix. The scattering parameters are then resynthesized from this candidate matrix, and the new objective function value is calculated. If the new objective function value is less than the current objective function value, the candidate matrix is ​​unconditionally accepted as the new current solution. If the new objective function value is greater than the current value, the poor solution is accepted with a certain probability. This acceptance probability decreases as the current temperature decreases, thus allowing escape from local minima in the early stages of optimization and gradually converging to the global or near-global optimum as the temperature gradually decreases. As iterations proceed, the temperature is gradually reduced using a preset cooling strategy. At each temperature level, the process of "random perturbation—calculating scattering parameters—evaluating the objective function—accepting or rejecting according to the annealing criterion" is repeated. When the temperature decreases to a set lower limit or the objective function change is lower than a preset threshold for several consecutive iterations, the optimization process is considered to have converged. The coupling matrix obtained at this time is the extracted coupling matrix.

[0041] Through the optimization process described above using the combined annealing simulation algorithm, the scattering parameters synthesized from the coupling matrix can be highly consistent with the scattering parameters of the three-dimensional electromagnetic simulation within a few iterations. This ensures that the extracted coupling matrix has high accuracy in reflecting the coupling characteristics of the actual physical structure. This extracted coupling matrix corresponds one-to-one with the initial physical dimensions used in step S105, accurately characterizing the electromagnetic coupling relationships between each resonator and with external ports under these physical dimensions. This provides a reliable equivalent parameter basis for subsequent iterative correction and performance optimization of the resonator length based on the extracted coupling matrix.

[0042] In step S107, the resonant cavity length of the target resonator is iteratively corrected according to the diagonal elements of the extracted coupling matrix to obtain the final physical size of the target resonator that meets the performance specifications.

[0043] The process of iteratively correcting the resonant cavity length of the target resonator specifically includes: determining the resonant cavity length correction amount based on the degree to which the resonant frequency deviates from the operating frequency as reflected by the diagonal elements; updating the resonant cavity length while keeping the coupling window width unchanged; and repeatedly calculating the extracted coupling matrix until the absolute value of the diagonal elements of the extracted coupling matrix is ​​less than a preset threshold.

[0044] As a supplement, determining the correction amount for the resonant cavity length includes: mapping each diagonal element to a normalized deviation of the resonant frequency of the corresponding resonant cavity relative to the operating frequency; then converting the normalized deviation into a resonant frequency offset based on the proportional relationship between the operating frequency and the passband bandwidth; and finally, based on the monotonic relationship between the resonant frequency and the resonant cavity length and the pre-obtained length... The correction amount for the length of the resonant cavity is calculated from the frequency-correlation curve.

[0045] After obtaining the extraction coupling matrix corresponding one-to-one with the initial physical dimensions, the diagonal elements of this coupling matrix are used to evaluate the resonant frequency deviation of each stage of the resonant cavity, and the cavity length is iteratively corrected accordingly. For resonators synthesized based on coupling matrix theory, ideally, the resonant frequency of each cavity should be consistent with the target operating frequency. In this case, the diagonal elements of the normalized coupling matrix are theoretically zero. However, when there are processing errors in the actual three-dimensional structure or the initial dimensions are only approximate solutions, the coupling matrix extracted from the simulation scattering parameters will have non-zero elements on the diagonal. The magnitude and sign of these elements reflect the degree of deviation of the corresponding resonant frequency from the operating frequency: the larger the absolute value of the diagonal element, the more severe the resonant frequency deviation; the sign indicates whether the resonant frequency is above or below the target frequency. Therefore, in this embodiment, each diagonal element is first regarded as the normalized frequency deviation of the resonant cavity. It is mapped to the normalized deviation of the resonant frequency relative to the operating frequency through a pre-set proportional relationship. Then, combined with the operating frequency and passband bandwidth, the normalized deviation is converted into the actual resonant frequency offset. In this way, without directly performing frequency scanning, the offset of the current resonant frequency of each stage resonator relative to the target operating frequency can be quantitatively estimated simply by extracting the diagonal elements of the coupling matrix, providing a basis for length correction.

[0046] After obtaining the resonant frequency offset of each stage of the resonant cavity, the frequency offset is converted into a correction amount for the resonant cavity length using the length-frequency correspondence obtained through simulation or calibration. Since the resonant cavity length and resonant frequency have a monotonic relationship under fixed waveguide cross-section and fixed coupling window conditions, this embodiment can treat the length-frequency curve as a strictly monotonic function: when the resonant frequency is higher than the target operating frequency, the resonant frequency can be lowered by appropriately increasing the resonant cavity length; conversely, the resonant frequency can be raised by decreasing the resonant cavity length. In specific implementation, the length-frequency curve obtained from multiple sets of length and frequency sampling points can be used to interpolate the difference between the current resonant frequency and the target frequency on this curve, calculating the amount of length that needs to be increased or decreased, thereby obtaining the length correction amount for each resonant cavity. In a preferred embodiment, the length correction amount is not directly taken as the full compensation given by the length-frequency curve, but is appropriately scaled according to the size and sign of the diagonal elements to avoid excessive single correction leading to new overcompensation, thus making the length adjustment process smoother and more stable.

[0047] After determining the length corrections for each stage of the resonator, the resonator length of the target resonator is updated while keeping the coupling window width constant. This involves modifying the length of each stage of the resonator in the 3D geometric model. Subsequently, a 3D electromagnetic simulation is performed on the updated structure, and a new coupling matrix is ​​extracted again following the aforementioned steps. The new coupling matrix also contains the updated diagonal elements, which can be used again to evaluate the resonant frequency deviation of each stage of the resonator and calculate the next round of length corrections. This forms a closed-loop iterative optimization process: using the currently extracted diagonal elements of the coupling matrix as input, the length correction is calculated, the resonator length is updated, the simulation is repeated, and the coupling matrix is ​​extracted again, and so on. Convergence criteria can be set during the iteration process. For example, when the absolute values ​​of all diagonal elements are less than a preset threshold, the resonant frequency of each stage of the resonator is considered sufficiently close to the target operating frequency, or the iteration is terminated when the improvement of the objective function is no longer significant after multiple iterations. When this convergence condition is met, the resonator length and coupling window width together constitute the final physical dimensions of the target resonator, resulting in a target resonator structure that meets the predetermined center frequency, passband bandwidth, and passband return loss specifications.

[0048] In this embodiment, the length iterative correction mechanism based on extracting the diagonal elements of the coupling matrix, as described above, allows for targeted compensation of resonant frequency deviations while maintaining the coupling structure unchanged, avoiding repeated searches for coupling window size and global parameter scanning. Simultaneously, by combining the aforementioned length-frequency curve and normalized frequency deviation mapping relationship, each correction step is based on a physically reasonable monotonic relationship, ensuring the convergence and predictability of the iterative process. Thus, without adding additional training models, the initial physical dimensions obtained based on artificial intelligence, the coupling matrix extracted from 3D simulation, and the length correction guided by diagonal elements are organically combined. This achieves a complete closed-loop design process from performance specifications, through coupling parameter prediction, initial size generation, coupling matrix extraction, to length iterative correction, ultimately obtaining the final physical dimensions of the target resonator that meet the performance requirements.

[0049] In one embodiment, after obtaining the training dataset, the training dataset is preprocessed, the preprocessing including: The coupling window width is normalized using the width of the standard waveguide in the corresponding frequency band, and the resonant cavity length is normalized using the width or height of the standard waveguide in the corresponding frequency band. Each element of the coupling matrix is ​​scaled according to its corresponding standard waveguide operating frequency to reduce the impact of frequency band differences on machine learning model training.

[0050] After obtaining the training dataset, the geometric dimensions and equivalent electromagnetic parameters in the training data are first normalized and scaled to map samples from different frequency bands and standard waveguides into a unified feature space, reducing the impact of frequency band differences on subsequent machine learning model training. Each sample in the training dataset contains information such as the coupling window width, resonant cavity length, and the standard waveguide model for the corresponding frequency band. Since standard rectangular waveguides for different frequency bands have different cross-sectional widths and heights, if absolute dimensions in millimeters are used directly for learning, there will be significant differences in the numerical range of samples from different frequency bands, which may cause the model to favor samples from a certain frequency band and reduce the overall generalization ability. Therefore, in this embodiment, it is preferable to normalize the coupling window width using the width of the standard waveguide for the corresponding frequency band. That is, the coupling window width of each sample is dimensionless relative to the width of the standard waveguide in its frequency band, so that the coupling openings under different frequency bands are mapped to similar normalized intervals in the model's view. This allows the model to focus more on "relative dimensions" rather than "absolute dimensions," enhancing the comparability between different operating frequency bands.

[0051] Regarding the resonant cavity length, considering the differences in the cross-sectional dimensions of standard waveguides across different frequency bands, and the inherent relationship between the resonant cavity length and the cross-sectional dimensions of the standard waveguide, this embodiment normalizes the resonant cavity length using the width or height of the standard waveguide for the corresponding frequency band. Specifically, the width dimension of the standard waveguide can be selected, or the height dimension can be used as the normalization reference in some cavity structures. The resonant cavity length is then expressed dimensionlessly using this reference dimension, ensuring that the cavity length data for samples across different frequency bands are distributed within a similar numerical range. Through this normalization process, when the model learns the "relationship between resonant cavity length and frequency, and coupling window width," it is actually learning the "proportional relationship between length and waveguide size," thus reducing frequency band differences to a geometric similarity problem. This facilitates the use of a unified model to cover multiple frequency bands.

[0052] Furthermore, during the training data preprocessing stage, the elements of the coupling matrix were scaled according to the corresponding standard waveguide operating frequency. Because the operating frequencies differ significantly across frequency bands, even with identical coupling topologies and similar normalized coupling structures, the equivalent coupling parameters are numerically affected by the operating frequency. To reduce this inter-band scale difference, this embodiment scales the coupling matrix elements corresponding to each sample according to the operating frequency of the standard waveguide to which that sample belongs, mapping the original coupling matrix elements to a more concentrated numerical range across multiple frequency bands. Through this scaling, when the model subsequently learns and predicts based on the coupling matrix elements, external quality factors, and inter-cavity coupling coefficients, it can process data from different frequency bands under a unified scale, avoiding the dominance of large numerical features from a single frequency band in model parameter updates, thereby improving the stability and accuracy of the model in cross-frequency band applications. Combining the above normalization and scaling preprocessing, the training dataset has already eliminated or weakened the frequency band differences at both the geometric size and coupling parameter levels before being input into the regression machine learning model, enabling the model to effectively learn the mapping relationships between parameters within a unified data space.

[0053] Based on the same line of thought, such as Figure 2 The diagram shown is a structural block diagram of an artificial intelligence-based resonator parameter analysis system according to an embodiment of the present invention. The system includes: The specification input module 201 is used to obtain the performance specifications of the target resonator, including the center frequency, passband bandwidth and passband return loss index, and to generate the ideal coupling matrix of the target resonator based on the performance specifications. The dataset construction module 202 is used to collect the correspondence between external coupling parameters, inter-cavity coupling parameters and physical dimensions for standard waveguide resonant cavity structures with multiple preset operating frequencies through electromagnetic simulation and / or theoretical calculation, forming a training dataset that includes external quality factor, inter-cavity coupling coefficient, coupling window width and resonant cavity length. The model training module 203 is used to train the training dataset using a regression machine learning model, so that the machine learning model takes the operating frequency, the external quality factor and / or the coupling coefficient between resonant cavities as input and the coupling window width as output to obtain a prediction model for predicting the coupling window width. The coupling size prediction module 204 is used to convert the external coupling parameters in the ideal coupling matrix into the corresponding external quality factor and the inter-cavity coupling parameters into the corresponding inter-cavity coupling coefficients based on the ideal coupling matrix and the center frequency and passband bandwidth of the target resonator, using a preset baseband-passband transformation relationship. The prediction model is then used to predict the coupling window width of the input / output port of the target resonator and the coupling window width between the resonators. The initial size calculation module 205 is used to calculate the cavity length of each cavity of the target resonator based on the obtained coupling window widths and the center frequency of the target resonator, combined with the relationship between the coupling window width and the cavity length in the training dataset, through interpolation and scaling transformation, thereby forming the initial physical size of the target resonator. The simulation module 206 is used to establish a three-dimensional electromagnetic simulation model of the target resonator based on the initial physical dimensions, calculate the simulated scattering parameters of the target resonator at preset frequency sampling points, and adjust the parameters of the coupling matrix based on the simulated scattering parameters and the ideal coupling matrix to minimize the objective function of the error between the scattering parameters synthesized based on the current coupling matrix and the simulated scattering parameters, thereby obtaining the extraction coupling matrix corresponding to the initial physical dimensions. The iterative correction module 207 is used to iteratively correct the resonant cavity length of the target resonator based on the diagonal elements of the extracted coupling matrix, so as to obtain the final physical size of the target resonator that meets the performance specifications.

[0054] Compared with existing design methods that rely on human experience and multiple rounds of simulation scanning, this system constructs data samples under multiple frequency bands, different coupling strengths, and different combinations of physical dimensions. It introduces a regression-type artificial intelligence model to learn the nonlinear mapping relationship between the external quality factor, the coupling coefficient between resonators, and the geometry of the coupling window. After obtaining the electromagnetic coupling parameters of the target resonator, it can directly predict the corresponding coupling window size, which greatly reduces the number of initial structural guesses and parameter trials, and significantly improves the automation level of resonator parameter analysis and structural synthesis.

[0055] The specific details of the above system have been described in detail in the method section of the implementation plan. For any undisclosed details, please refer to the implementation plan of the method section, and therefore will not be repeated here.

[0056] The accompanying drawings are merely illustrative of the processes included in the methods according to exemplary embodiments of this disclosure and are not intended to be limiting. It is readily understood that the processes shown in the drawings do not indicate or limit the temporal order of these processes. Furthermore, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0057] It should be noted that although several modules or units of the system have been mentioned in the detailed description above, this division is not mandatory. In fact, according to exemplary embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0058] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

[0059] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A resonator parameter analysis method based on artificial intelligence, characterized in that, The method includes: Obtain the performance specifications of the target resonator, including the center frequency, passband bandwidth, and passband return loss index, and generate the ideal coupling matrix of the target resonator based on the performance specifications. For standard waveguide resonant cavity structures with multiple preset operating frequencies, the correspondence between external coupling parameters, inter-cavity coupling parameters and physical dimensions is collected through electromagnetic simulation and / or theoretical calculation, forming a training dataset that includes external quality factor, inter-cavity coupling coefficient, coupling window width and resonant cavity length. A regression-based machine learning model is used to train the training dataset, so that the machine learning model takes the operating frequency, the external quality factor and / or the coupling coefficient between resonant cavities as input and the coupling window width as output, to obtain a prediction model for predicting the coupling window width. Based on the ideal coupling matrix and the center frequency and passband bandwidth of the target resonator, a preset baseband-passband transformation relationship is used to convert the external coupling parameters in the ideal coupling matrix into the corresponding external quality factors, and the inter-cavity coupling parameters into the corresponding inter-cavity coupling coefficients. The coupling window widths of the input / output ports of the target resonator and the coupling window widths between the resonators are then predicted by the prediction model. Based on the obtained coupling window widths and the center frequency of the target resonator, and combined with the relationship between the coupling window width and the cavity length in the training dataset, the cavity lengths of each cavity of the target resonator are calculated through interpolation and scaling transformation, thus forming the initial physical dimensions of the target resonator. A three-dimensional electromagnetic simulation model of the target resonator is established based on the initial physical dimensions. The simulated scattering parameters of the target resonator are calculated at preset frequency sampling points. Based on the simulated scattering parameters and the ideal coupling matrix, the parameters of the coupling matrix are adjusted to minimize the objective function of the error between the scattering parameters synthesized based on the current coupling matrix and the simulated scattering parameters, thereby obtaining the extraction coupling matrix corresponding to the initial physical dimensions. Based on the extracted diagonal elements of the coupling matrix, the resonant cavity length of the target resonator is iteratively corrected to obtain the final physical dimensions of the target resonator that meet the performance specifications.

2. The resonator parameter analysis method based on artificial intelligence according to claim 1, characterized in that, The external coupling parameters of the training dataset are obtained in the following manner: For a waveguide structure containing only a single resonant cavity and a coupling window on one side, electromagnetic simulation is performed under multiple coupling window widths. The phase change curve of the reflection parameter with frequency is extracted. The group delay near the resonant frequency is calculated based on the derivative of the phase with respect to frequency. The external quality factor is obtained based on the proportional relationship between the group delay and the corresponding resonant angular frequency. Each external quality factor is associated with the corresponding coupling window width and operating frequency and stored in the training dataset.

3. The resonator parameter analysis method based on artificial intelligence according to claim 1, characterized in that, The inter-cavity coupling parameters of the training dataset are obtained in the following manner: For a waveguide structure containing two identical resonant cavities connected to each other through a coupling window, electromagnetic simulation is performed at multiple coupling window widths. The two resonant peak frequencies generated by coupling are extracted. The ratio of the square difference to the sum of the squares of the two resonant peak frequencies is determined as the coupling coefficient between the resonant cavities at the corresponding coupling window width. The coupling coefficient between the resonant cavities is then associated with the corresponding coupling window width and operating frequency and stored in the training dataset.

4. The resonator parameter analysis method based on artificial intelligence according to claim 1, characterized in that, The resonant cavity length of the training dataset is obtained in the following manner: With a fixed waveguide cross-sectional dimension, a resonant cavity model is established for different combinations of resonant cavity length, coupling window width, and operating frequency. The corresponding resonant frequencies are obtained through simulation. For a given operating frequency, interpolation is performed on the two-dimensional graph of the resonant cavity length and coupling window width to make the resonant frequency equal to the resonant cavity length of the operating frequency. The resonant cavity length, the corresponding coupling window width, and the operating frequency are then associated and stored in the training dataset.

5. The resonator parameter analysis method based on artificial intelligence according to claim 1, characterized in that, After obtaining the training dataset, the training dataset is preprocessed, the preprocessing including: The coupling window width is normalized using the width of the standard waveguide in the corresponding frequency band, and the resonant cavity length is normalized using the width or height of the standard waveguide in the corresponding frequency band. Each element of the coupling matrix is ​​scaled according to its corresponding standard waveguide operating frequency to reduce the impact of frequency band differences on machine learning model training.

6. The resonator parameter analysis method based on artificial intelligence according to claim 1, characterized in that, The iterative correction of the resonant cavity length of the target resonator specifically includes: The resonant cavity length correction amount is determined based on the degree to which the resonant frequency deviates from the operating frequency as reflected by the diagonal elements. The resonant cavity length is updated while keeping the coupling window width unchanged, and the extraction coupling matrix is ​​repeatedly calculated until the absolute value of the diagonal elements of the extraction coupling matrix is ​​less than a preset threshold.

7. The resonator parameter analysis method based on artificial intelligence according to claim 6, characterized in that, Determining the correction amount for the resonant cavity length includes: Each diagonal element is mapped to the normalized deviation of the resonant frequency of the corresponding resonant cavity relative to the operating frequency. The normalized deviation is then converted into a resonant frequency offset based on the proportional relationship between the operating frequency and the passband bandwidth. This offset is determined according to the monotonic relationship between the resonant frequency and the resonant cavity length, and a pre-obtained length. The correction amount for the length of the resonant cavity is calculated from the frequency-correlation curve.

8. An artificial intelligence-based resonator parameter analysis system, the system comprising: The specification input module is used to obtain the performance specifications of the target resonator, including the center frequency, passband bandwidth, and passband return loss index, and to generate the ideal coupling matrix of the target resonator based on the performance specifications. The dataset construction module is used to collect the correspondence between external coupling parameters, inter-cavity coupling parameters and physical dimensions for standard waveguide resonant cavity structures with multiple preset operating frequencies through electromagnetic simulation and / or theoretical calculation, forming a training dataset that includes external quality factor, inter-cavity coupling coefficient, coupling window width and resonant cavity length. The model training module is used to train the training dataset using a regression machine learning model, so that the machine learning model takes the operating frequency, the external quality factor and / or the coupling coefficient between resonant cavities as input and the coupling window width as output to obtain a prediction model for predicting the coupling window width. The coupling size prediction module is used to convert the external coupling parameters in the ideal coupling matrix into the corresponding external quality factor and the inter-cavity coupling parameters into the corresponding inter-cavity coupling coefficients based on the ideal coupling matrix, the center frequency and the passband bandwidth of the target resonator, and a preset baseband-passband transformation relationship. The prediction model is then used to predict the coupling window width of the input / output port of the target resonator and the coupling window width between the resonators. The initial size calculation module is used to calculate the cavity length of each cavity of the target resonator based on the obtained coupling window widths and the center frequency of the target resonator, combined with the relationship between the coupling window widths and the cavity lengths in the training dataset, through interpolation and scaling transformation, thereby forming the initial physical size of the target resonator. The simulation module is used to establish a three-dimensional electromagnetic simulation model of the target resonator based on the initial physical dimensions, calculate the simulated scattering parameters of the target resonator at preset frequency sampling points, and adjust the parameters of the coupling matrix based on the simulated scattering parameters and the ideal coupling matrix to minimize the objective function of the error between the scattering parameters synthesized based on the current coupling matrix and the simulated scattering parameters, thereby obtaining the extraction coupling matrix corresponding to the initial physical dimensions. The iterative correction module is used to iteratively correct the resonant cavity length of the target resonator based on the diagonal elements of the extracted coupling matrix, so as to obtain the final physical size of the target resonator that meets the performance specifications.

Images