Coaxial open load design method and device based on air medium

Abstract

This invention discloses a design method and apparatus for coaxial open-circuit loads based on air dielectric, belonging to the field of electrical digital data processing technology. It includes three steps: topology generation, degradation monitoring, and feedback correction. First, a conditional diffusion model is constructed to generate a candidate set of irregularly shaped air cavity topologies. After extracting parameters, the shape gain ratio is calculated, and the optimal structure is selected. Next, a digital twin model is constructed, the real part of the input impedance is measured to calculate impedance drift, and the degradation deviation is calculated by combining the temperature cycling fatigue amplification factor. Finally, the degradation deviation is compared with a threshold, and the feedback trigger coefficient is calculated by combining edge capacitance and resonant frequency shift. Based on the current range, calibration parameters are updated or the model is incrementally retrained. This invention constructs a closed-loop system through three steps, employs multi-dimensional comprehensive evaluation to avoid misjudgments, and achieves a shift from passive maintenance to proactive prevention. The optimized combination of air dielectric and irregularly shaped cavities improves electrical performance stability and calibration reliability.

Description

Technical Field

[0001] This invention relates to the field of electrical digital data processing technology, and in particular to a method and apparatus for designing coaxial open-circuit loads based on air medium. Background Technology

[0002] Coaxial open-circuit loads are one of the core standard components of the SOLT calibration method in the field of radio frequency and microwave measurement. Their function is to provide known open-circuit reflection parameters for test equipment such as vector network analyzers to eliminate systematic errors and improve measurement accuracy. The reflection coefficient and phase linearity of the open-circuit load directly affect the magnitude of the residual of the polynomial fitting, thus determining the quality of calibration. Currently, coaxial open-circuit loads typically use insulating dielectric materials to simultaneously support the inner conductor and perform the open-circuit load function. Cross-linked polystyrene is widely used due to its low relative permittivity and extremely low loss tangent. This structure forms a series load by inserting dielectric pillars into the slotted holes at the tail of the inner conductor. The mechanical strength of the dielectric material is used to maintain the axial positioning of the inner conductor, while its dielectric properties are used to achieve the open-circuit reflection function. This ensures that the transmission line between the connector and the load end face maintains a uniform air dielectric structure, making it easy to directly extract the transmission line delay parameters through physical length.

[0003] As radio frequency measurement frequencies continue to extend into the millimeter-wave band, existing technologies for coaxial open-circuit loads face multiple constraints. In terms of structural manufacturing, the radial dimension of the open-circuit load decreases synchronously with the increase in operating frequency. The processing accuracy and assembly process of the dielectric pillar are extremely demanding. If the material stiffness is too low, the axial support will be insufficient; if the stiffness is too high, it will be prone to breakage under radial stress. It is difficult to achieve both. In terms of electrical performance, the impedance interface discontinuity at the connection between the dielectric pillar and the inner conductor, as well as the step transformation structure, will introduce parasitic resonances in the operating frequency band, destroying the frequency linearity of the reflection coefficient phase, leading to an increase in polynomial fitting residuals and reducing calibration quality. Summary of the Invention

[0004] To address the shortcomings mentioned in the background technology, we propose a design method and apparatus for coaxial open-circuit loads based on air medium.

[0005] The technical solution mainly involves: a coaxial open-circuit load design method based on air medium, including the following steps: Topology generation: Construct a conditional diffusion model, using connector interface standard, target maximum operating frequency, and target edge capacitance upper limit as conditional inputs to generate a candidate set of irregular air cavity cross-section topologies. Perform full-wave electromagnetic simulation on each candidate topology to extract high-order mode cutoff frequency and edge capacitance parameters. Calculate the shape gain ratio based on the corresponding parameters of a standard circular cavity, and select candidate topologies with shape gain ratios higher than a preset threshold as the optimal cavity structure. Degradation monitoring: Construct individualized digital twin models for open-circuit loads, collect data on geometric deviations, material parameters, insertion and removal times, and temperature cycles, select multiple monitoring frequency points within the target frequency band to measure the real part of the input impedance at each frequency point, calculate the relative impedance drift amplitude at each frequency point based on the factory-calibrated real impedance, and calculate the degradation deviation by combining the temperature cycle fatigue amplification factor. Feedback correction: The degradation deviation is compared with the preset trigger threshold. The deviation of the posterior mean of the edge capacitance output by the digital twin model from the factory reference value and the deviation of the posterior mean of the resonant frequency from the factory reference value are combined to calculate the feedback trigger coefficient. Based on the numerical range of the feedback trigger coefficient, the calibration parameters are updated and the conditional diffusion model is incrementally retrained.

[0006] Preferably, the specific steps for generating the candidate set of irregular air cavity cross-section topologies in the topology generation step are as follows: A reverse denoising network for the conditional diffusion model is constructed. The outer conductor inner diameter, inner conductor outer diameter, target maximum operating frequency, and target edge capacitance upper limit corresponding to the connector interface standard are encoded as conditional embedding vectors. The random Gaussian noise tensor is used as the initial generation starting point. Through the iterative denoising process of the conditional diffusion model, the geometric contour data of the air cavity cross-section topology are gradually restored. The generated geometric contour data is morphologically classified and identified. When the cross-sectional contour exhibits periodic rotational gradient characteristics along the axial direction, it is marked as a spiral cavity. When the cross-sectional contour exhibits self-similar recursive nesting characteristics, it is marked as a fractal cavity. The full-wave electromagnetic simulation engine is called sequentially for the marked candidate topologies to extract the high-order mode cutoff frequency and edge capacitance parameters of each candidate topology. The extraction results are then output to the shape gain ratio calculation step for screening.

[0007] Preferably, in the degradation monitoring step, the specific steps for constructing the individualized digital twin model are as follows: High-precision CT scanning and optical 3D measurement are performed on the open-circuit load before leaving the factory to obtain the actual geometric dimensions and manufacturing deviation data of the air cavity, read the actual dielectric constant and thermal expansion coefficient of the supporting medium material, and write the geometric data and material data into the static parameter layer of the digital twin model. Temperature and humidity sensors are integrated inside the open-circuit load to collect ambient temperature and humidity data in real time. A mechanical counter is integrated on the insertion and removal mechanism of the open-circuit load to collect cumulative insertion and removal counts and insertion and removal torque data in real time. The dynamically collected data is written into the dynamic parameter layer of the digital twin model. The static parameter layer data and the dynamic parameter layer data are input into the long short-term memory network. The network outputs the posterior probability distribution of the edge capacitance and the posterior probability distribution of the resonant frequency at the current time. The expected value of the probability distribution is taken as the posterior mean of the edge capacitance and the posterior mean of the resonant frequency. Multiple monitoring frequency points are selected within the target frequency band to measure the real part of the input impedance at each frequency point. The absolute value of the difference between the measured value and the reference value at each frequency point is calculated based on the factory-calibrated real part of the impedance. The relative drift amplitude of the impedance at each frequency point is obtained by dividing the absolute value of the difference between the measured value and the reference value by the reference value. The cumulative number of temperature cycles and the design allowable number of temperature cycles are calculated to obtain the temperature cycle fatigue amplification factor. The relative drift amplitude of the impedance at each frequency point is multiplied by the corresponding temperature cycle fatigue amplification factor and the average value is taken for all monitoring frequency points to obtain the degradation deviation. The posterior mean of the edge capacitance, the posterior mean of the resonant frequency, and the degradation deviation are output to the feedback correction step.

[0008] Preferably, in the feedback correction step, calibration parameter updates and incremental retraining of the conditional diffusion model are performed according to the numerical range of the feedback trigger coefficient. The specific steps are as follows: When the feedback trigger coefficient is lower than the first preset threshold, it is determined that the degree of open-circuit load degradation is within an acceptable range, and no correction action is triggered. The digital twin model continues to perform degradation monitoring at the normal sampling frequency. When the feedback trigger coefficient is between the first preset threshold and the second preset threshold, it is determined that the open-circuit load has moderate degradation. The posterior mean of the edge capacitance and the posterior mean of the resonant frequency output by the digital twin model are pushed to the calibration software. The calibration software automatically updates the polynomial fitting coefficients to complete the calibration parameter compensation. When the feedback trigger coefficient is higher than the second preset threshold, it is determined that the open circuit load has seriously degraded. While completing the calibration parameter compensation, the degraded data is encapsulated as negative samples according to the timestamp. The weight parameters of the currently deployed conditional diffusion model are read from the cloud model repository. The negative sample data is mixed with the original training data to construct an incremental training dataset. The conditional diffusion model is fine-tuned to make it tend to generate anti-degradation cavity topology. After the fine-tuning is completed, the updated model weights are pushed to the cloud model repository to replace the original version.

[0009] Preferably, when the feedback trigger coefficient is higher than the second preset threshold, the specific steps for encapsulating the degraded data into negative samples are as follows: The degradation deviation sequence, edge capacitance drift sequence, resonant frequency drift sequence, cumulative insertion / removal count sequence, and cumulative temperature cycle sequence within a preset time window before the trigger moment are read from the digital twin model. All sequences are then aligned according to timestamps and concatenated into a multidimensional degradation feature vector. Read the initial geometric parameters and initial material parameters of the open-circuit load at the time of manufacture, and associate and encapsulate the initial parameters with the multidimensional degradation feature vector to form a complete negative sample record. Add a degradation label field to the negative sample record. The value of the label field is obtained by normalizing the feedback trigger coefficient and represents the degradation severity level of the negative sample. The encapsulated negative sample records are written to the cloud negative sample database for random sampling and use during incremental training of the conditional diffusion model at a preset ratio.

[0010] The present invention also provides a coaxial open-circuit load design system based on air medium, including a topology generation module, a degradation monitoring module, and a feedback correction module; The topology generation module is configured to generate a candidate set of irregular air cavity cross-section topologies using the running condition diffusion model, call the full-wave electromagnetic simulation engine to extract the high-order mode cutoff frequency and edge capacitance, calculate the shape gain ratio based on a standard circular cavity, and filter and output the optimal cavity structure to the manufacturing execution system. The degradation monitoring module is configured to build an individualized digital twin model, collect geometric deviation data, material parameter data, insertion and removal number data, and temperature cycle data in real time, and output the degradation deviation to the feedback correction module after calculating the degradation deviation. The feedback correction module is configured to receive degradation deviation data, calculate the feedback trigger coefficient, and trigger calibration parameter update instructions and conditional diffusion model incremental retraining instructions according to the numerical range of the feedback trigger coefficient.

[0011] Preferably, the topology generation module consists of a condition generation unit, a simulation filtering unit, and a structure output unit; The condition generation unit connects to the connector interface standard database and design index input interface, reads the outer conductor inner diameter, inner conductor outer diameter, target maximum operating frequency, and target edge capacitance upper limit as condition inputs, runs the condition diffusion model to generate a candidate set of irregular air cavity cross-section topology, and then outputs it to the simulation screening unit. The simulation screening unit is connected to the full-wave electromagnetic simulation engine. For each candidate topology, the high-order mode cutoff frequency and edge capacitance are simulated and extracted sequentially. The shape gain ratio is calculated based on the corresponding parameters of the standard circular cavity. The shape gain ratio calculation results of each candidate topology are output to the structure output unit. The structure output unit receives the shape gain ratio calculation results of each candidate topology, filters out the candidate topologies with shape gain ratios higher than a preset threshold, sorts them from high to low according to shape gain ratio, and outputs the optimal cavity structure to the manufacturing execution system.

[0012] Preferably, the degradation monitoring module consists of a digital twin construction unit, a data acquisition unit, and a degradation accounting unit; The digital twin building block interfaces with CT scanning equipment and optical 3D measurement equipment, reads the geometric deviation data and material parameter data of open circuit load and writes them into the static parameter layer of the digital twin model, and at the same time interfaces with the long short-term memory network to complete the model initialization and deployment; The data acquisition unit interfaces with temperature sensors, humidity sensors, and mechanical counters to collect ambient temperature data, ambient humidity data, cumulative insertion and removal count data, and insertion and removal torque data in real time. After writing these data into the dynamic parameter layer of the digital twin model, the data is output to the degradation calculation unit. The degradation calculation unit receives static parameter layer data and dynamic parameter layer data, selects multiple monitoring frequency points within the target frequency band to complete the measurement of the real part of the input impedance, calculates the relative impedance drift amplitude at each frequency point based on the factory calibration data, calculates the degradation deviation by combining the temperature cycle fatigue amplification factor, and outputs the calculation results to the feedback correction module.

[0013] Preferably, the feedback correction module consists of a trigger determination unit, a calibration compensation unit, and a model iteration unit; The trigger determination unit receives the degradation deviation degree output by the degradation monitoring module, reads the preset trigger threshold, compares the degradation deviation degree with the preset trigger threshold, reads the deviation degree of the posterior mean of the edge capacitance and the factory reference value output by the digital twin model and the deviation degree of the posterior mean of the resonant frequency and the factory reference value, comprehensively calculates the feedback trigger coefficient, and sends trigger commands to the calibration compensation unit and the model iteration unit respectively according to the numerical range of the feedback trigger coefficient. After receiving the trigger command, the calibration compensation unit reads the posterior mean of the edge capacitance and the posterior mean of the resonant frequency output by the digital twin model, and pushes the posterior mean data to the calibration software. The calibration software automatically updates the polynomial fitting coefficients to complete the calibration parameter compensation. After receiving the trigger command, the model iteration unit reads the degenerate data from the digital twin model and encapsulates it into negative samples. It reads the weight parameters of the currently deployed conditional diffusion model from the cloud model repository, mixes the negative sample data with the original training data to construct an incremental training dataset, performs fine-tuning training on the conditional diffusion model, and then pushes the updated model weights to the cloud model repository.

[0014] The present invention also provides a coaxial open-circuit load device based on air medium, comprising an inner conductor part, an outer conductor part, a dielectric support structure, and an air medium cavity; The outer conductor is the outer metal tube wall that constitutes the coaxial transmission line structure, and the inner conductor is the central metal conductor that constitutes the coaxial transmission line structure. One end of the inner conductor is connected to the connector interface, and the other end of the inner conductor extends into the air medium cavity. The air medium cavity is an open cavity structure in which air medium is filled between the end of the inner conductor and the outer conductor. The radial cross-sectional shape of the air medium cavity is the irregular cross-sectional shape corresponding to the optimal cavity structure selected by the topology generation step. The dielectric support structure is located between the inner conductor and the outer conductor. The dielectric support structure is made of solid insulating material and only undertakes the mechanical support function of the inner conductor. It does not participate in the electromagnetic function of the open circuit load.

[0015] Compared with the prior art, the beneficial effects of the present invention are: In this invention, by organically integrating three steps—topology generation, degradation monitoring, and feedback correction—a complete closed-loop system is constructed, encompassing intelligent cavity structure design, real-time performance tracking, and degradation data-driven model iteration. The topology generation step utilizes a conditional diffusion model to automatically explore the cross-sectional topology of irregularly shaped air cavities, overcoming the limitations of traditional designs that rely on engineers' experience for regular shape design. This allows for the selection of cavity structures with optimal overall performance within a broader design space. The degradation monitoring step uses an individualized digital twin model to perceive the device's performance degradation status in real time, achieving continuous quantitative assessment of the open-circuit load's health status. The feedback correction step executes graded responses based on the degree of degradation, ensuring that the design model continuously evolves with the accumulation of usage data. The synergistic effect of these three steps significantly improves the performance consistency and calibration reliability of open-circuit loads throughout their entire lifecycle.

[0016] In this invention, a multi-dimensional comprehensive evaluation mechanism is adopted to determine the degradation state of open-circuit loads. The evaluation is carried out from three key dimensions: electrical performance drift, parasitic parameter changes, and resonant characteristic shift. This avoids misjudgment or omission that may be caused by a single-dimensional judgment. At the same time, the graded response strategy matches differentiated correction actions according to the degree of degradation. In the case of mild degradation, routine monitoring is maintained; in the case of moderate degradation, calibration parameter compensation is automatically completed; and in the case of severe degradation, generative model evolution is driven to generate anti-degradation cavity topology. This shift from passive maintenance to active prevention enables the system to identify potential risks and actively intervene in the early stage of degradation, effectively extending the effective service life of open-circuit loads and maintaining the long-term stability of calibration accuracy.

[0017] In this invention, by separating the support function from the load function and using air as the open-circuit load medium, the edge capacitance at the inner conductor connection end face is effectively reduced, while the high-order mode cutoff frequency and resonant frequency of the cavity are improved. This significantly improves the phase linearity of the reflection coefficient of the open-circuit load in the operating frequency band. The low dielectric constant and low loss characteristics of air, combined with the topology optimization of the irregular cross-section cavity, enable the device to maintain stable electrical performance throughout its entire life cycle. The modular system architecture allows each functional unit to be developed and deployed independently, reducing system coupling and maintenance costs, and providing a reliable hardware foundation and intelligent management means for high-precision RF calibration. Attached Figure Description

[0018] Figure 1 This is a structural diagram of the coaxial open-circuit load device based on air medium in this invention; Figure 2 This is a flowchart illustrating the overall operation of the present invention.

[0019] In the figure, 1 is the inner conductor section; 2 is the outer conductor section; 3 is the dielectric support structure; and 4 is the air dielectric cavity. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings and preferred embodiments.

[0021] Example 1, refer to Figure 1-2 As shown, the coaxial open-circuit load design method based on air medium includes three steps: topology generation, degradation monitoring, and feedback correction. Specifically, the topology generation step is performed first to construct a conditional diffusion model. The connector interface standard, the target maximum operating frequency, and the target edge capacitance upper limit are used as conditional inputs to drive the model to generate candidate topology structures for irregular air cavity cross sections. Full-wave electromagnetic simulation is performed on each candidate topology to extract its higher-order mode cutoff frequency and edge capacitance parameters. The shape gain ratio is calculated based on the corresponding parameters of the standard circular cavity. Candidate topologies with a shape gain ratio higher than a preset threshold are selected as the optimal cavity structure. Next, a degradation monitoring step is performed. An individualized digital twin model is built for each open-circuit load. Multi-dimensional data such as geometric deviation, material parameters, insertion and removal times, and temperature cycles are collected. Multiple monitoring frequency points are selected within the target frequency band. The real part of the input impedance at each frequency point is measured. The relative impedance drift amplitude at each frequency point is calculated based on the factory-calibrated real part of the impedance. The degradation deviation is calculated in combination with the temperature cycle fatigue amplification factor, so as to realize the real-time quantitative assessment of the performance degradation state of the open-circuit load. Finally, a feedback correction step is performed, comparing the degradation deviation with the preset trigger threshold. At the same time, the deviation of the posterior mean of the edge capacitance output by the digital twin model from the factory reference value and the deviation of the posterior mean of the resonant frequency from the factory reference value are combined to comprehensively calculate the feedback trigger coefficient. Based on the numerical range of the feedback trigger coefficient, calibration parameter updates or conditional diffusion model incremental retraining are performed respectively, forming a complete closed loop from design to use to feedback optimization.

[0022] This embodiment organically connects the three steps of topology generation, degradation monitoring, and feedback correction to construct a complete link from intelligent cavity structure design to real-time performance tracking and degradation data-driven model iteration, realizing the transformation of the open-loop load design method from an open-loop static mode to a closed-loop intelligent mode. In this embodiment, the limitations of traditional regular shape design are broken. The digital twin model realizes real-time perception of device performance degradation, and the feedback correction mechanism ensures that the design model continues to evolve with the accumulation of usage data. The synergistic effect of the three significantly improves the performance consistency and calibration reliability of open circuit loads throughout their entire life cycle.

[0023] Example 2, refer to Figure 1-2 As shown, in the topology generation step, the generation and filtering of the candidate set of topologies for the irregular air cavity cross-section are performed first. The specific process is as follows: First, an inverse denoising network for the conditional diffusion model is constructed. The inner diameter of the outer conductor, the outer diameter of the inner conductor as specified in the connector interface standard, as well as the pre-set target maximum operating frequency and target edge capacitance upper limit are jointly encoded into a conditional embedding vector. Starting with the random Gaussian noise tensor, the geometric contour data of the air cavity cross-section topology are gradually restored through iterative denoising. After generation, the geometric contours are morphologically classified: axial periodic rotational gradient structures are classified as spiral cavities, and self-similar recursive nested structures are classified as fractal cavities. Then, full-wave electromagnetic simulation is performed on each candidate topology to extract two key parameters: high-order mode cutoff frequency and edge capacitance. Since the higher-order mode cutoff frequency and edge capacitance belong to different physical dimensions, they are first processed to be dimensionless before entering the ratio calculation. Then, the extreme value method is used to normalize them to the 0-1 range. The reference value is taken as the theoretical limit value under the target connector interface standard to ensure the consistency of dimensions in subsequent calculations. After dimensionless transformation, the extracted parameters are substituted into the shape gain ratio calculation logic. Using the high-order mode cutoff frequency of a standard circular cavity as a reference, the deviation ratio of the candidate cavity cutoff frequency from the reference value is calculated to obtain the relative improvement rate of the cutoff frequency. Using the edge capacitance of a standard circular cavity as a reference, the ratio of the reference value to the candidate value is calculated to obtain the capacitance reduction factor. Multiplying the two yields the shape gain ratio. The specific calculation formula is as follows: ; In the formula, G is the calculated shape gain ratio of the candidate irregular cavity relative to the reference circular cavity, which is the core indicator used to quantitatively evaluate the comprehensive performance of the candidate cavity. The larger the value, the better the comprehensive performance of the candidate cavity in terms of broadband performance and parasitic parameter control. f c The high-order mode cutoff frequency of the candidate irregular cavity is obtained by performing full-wave electromagnetic simulation on the candidate cavity: In the simulation environment, a broadband excitation signal is applied to the cavity and its transmission characteristics are scanned. The frequency corresponding to when the first high-order mode begins to transmit is this value, which reflects the highest frequency at which the cavity can maintain the main mode single-mode transmission. The higher this value, the wider the single-mode working bandwidth of the cavity. f c0 The high-order mode cutoff frequency of a standard circular cavity of the same size is extracted by performing the same full-wave electromagnetic simulation on a standard circular cavity with the same outer conductor inner diameter and inner conductor outer diameter as the candidate cavity. It is used as a benchmark reference value to measure the degree of improvement of the candidate cavity in high-order mode suppression. C d The edge capacitance at the inner conductor connection end face of the candidate irregular cavity is extracted by quasi-static field simulation of the inner conductor end region of the candidate cavity through electrostatic field solver. It reflects the parasitic capacitance value between the inner conductor end face and the cavity wall due to the electric field edge effect. The smaller the value, the less severe the deterioration of the high-frequency phase linearity of the open circuit load. C d0 The edge capacitance of a standard circular cavity of the same size is extracted by performing the same electrostatic field simulation on the standard circular cavity and is used as a benchmark reference value to measure the degree of improvement of the candidate cavity in reducing parasitic capacitance. After the calculation is completed, candidate topologies with a shape gain ratio higher than the preset threshold are selected as the optimal cavity structure. The higher the shape gain ratio, the better the overall balance between improving the high-order mode cutoff frequency and reducing the edge capacitance of the candidate cavity, and the more suitable it is as the final open-circuit load air cavity structure. In this embodiment, the combination of automatic generation of irregular cavities driven by the conditional diffusion model and shape gain ratio quantization screening enables efficient identification of the optimal structure from a massive number of candidate topologies. The shape gain ratio simultaneously takes into account the increase in the cutoff frequency of higher-order modes and the decrease in edge capacitance. The two are multiplied to form a comprehensive score, ensuring that the screening results achieve the optimal balance in both broadband performance and parasitic parameter control, providing a clear design basis for subsequent manufacturing processes.

[0024] Example 3, refer to Figure 1-2 As shown, in the degradation monitoring step, an individualized digital twin model is constructed and the degradation deviation is calculated. The specific process is as follows: First, the digital twin model is initialized and constructed. High-precision CT scans and optical 3D measurements are performed on the open-circuit load before it leaves the factory to obtain the actual geometric dimensions and manufacturing deviation data of the air cavity. At the same time, the actual dielectric constant and thermal expansion coefficient of the supporting medium material are read. These inherent attribute data are written into the static parameter layer of the digital twin model. Temperature and humidity sensors are integrated inside the open-circuit load to collect ambient temperature and humidity data in real time. A mechanical counter is integrated on the insertion and removal mechanism to collect the cumulative number of insertions and removals and the insertion and removal torque data in real time. These dynamic change data are written into the dynamic parameter layer of the digital twin model. The data from the static parameter layer and the dynamic parameter layer are input into a long short-term memory network. This network outputs the posterior probability distribution of the edge capacitance and the posterior probability distribution of the resonant frequency at the current moment. The expected value of the probability distribution is taken as the posterior mean of the edge capacitance and the posterior mean of the resonant frequency.

[0025] Furthermore, before entering the ratio calculation, dimensionless processing is performed, and then the extreme value method is used to normalize the two to the 0-1 interval. The benchmark value is taken as the theoretical limit value or design allowable value of the corresponding parameter to ensure the dimensional consistency of subsequent calculations. After unifying the dimensions, multiple monitoring frequency points are selected within the target frequency band. The real part of the input impedance at each frequency point is measured. Using the factory-calibrated real part of the impedance as a reference, the deviation between the measured value and the reference value at each frequency point is calculated. At the same time, the ratio of the cumulative number of temperature cycles to the design allowable number of cycles is calculated and superimposed with the reference value to obtain the temperature cycle fatigue amplification factor. The impedance deviation at each frequency point is multiplied by the corresponding fatigue amplification factor, and then the average value is taken for all frequency points to obtain the degradation deviation. The specific calculation formula is as follows: ; In the formula, D is the calculated degradation deviation of the open-circuit load at the current monitoring time. The larger the value, the more severe the drift of the device performance relative to the factory condition. M represents the total number of monitoring frequency points selected within the target frequency band, which is determined by the frequency sampling settings of the test system. Z j The real part of the open-circuit load input impedance measured at the j-th monitoring frequency point up to the current time is obtained by single-frequency measurement at the corresponding frequency point using a vector network analyzer, reflecting the impedance characteristics of the open-circuit load at the current frequency point. Z0 is the reference value of the real part of the input impedance at the same frequency point under the factory calibration state of the open-circuit load. It is obtained by full-band calibration measurement using a vector network analyzer before the open-circuit load leaves the factory. It is stored as a reference value in the static parameter layer of the digital twin model for reference comparison during subsequent degradation assessment. N jThe cumulative number of temperature cycles at the j-th monitoring frequency point up to the current moment is obtained by counting the heating-cooling cycle events recorded by the temperature sensor integrated near the open-circuit load. It reflects the degree of thermomechanical stress accumulation experienced by the device. The more temperature cycles, the more severe the material fatigue accumulation. N0 represents the design allowable number of temperature cycles for open-circuit loads, given in the product reliability specification and determined based on the material properties and life test data of the device. It indicates the upper limit of the number of temperature cycles that the device can withstand under normal operating conditions and serves as a normalized benchmark for temperature cycling fatigue amplification.

[0026] After completing the degradation deviation calculation, the posterior mean of the edge capacitance, the posterior mean of the resonant frequency, and the degradation deviation are output to the feedback correction step as input parameters for subsequent feedback trigger coefficient calculation. In this embodiment, the separate design of the static parameter layer and the dynamic parameter layer enables the digital twin model to independently model the inherent properties of the device and the changes in the usage environment. The application of the long short-term memory network allows the degradation prediction results to be presented in the form of a probability distribution, quantifying the uncertainty of the prediction results. The degradation deviation calculation logic integrates the impedance drift amplitude and the temperature cycle fatigue effect. The relative impedance drift amplitude reflects the degree of degradation of electrical performance, and the temperature cycle fatigue amplification factor characterizes the accelerating effect of mechanical thermal stress on the degradation rate. The two are multiplied to form a composite degradation index, which can more comprehensively and accurately reflect the actual health status of the open circuit load.

[0027] Example 4, refer to Figure 1-2 As shown, in the feedback correction step, based on the degradation deviation output from the degradation monitoring step and the posterior information of edge capacitance and resonant frequency provided by the digital twin model, the severity of degradation is comprehensively determined and a graded response is executed. The specific process is as follows: First, the degradation deviation calculated by the degradation monitoring step is obtained. This value is compared with a pre-set trigger threshold, and the proportional relationship between the two is calculated to obtain the degradation severity multiple. Simultaneously, the posterior mean of the edge capacitance output by the digital twin model is read and compared with the baseline value of the edge capacitance under factory calibration. The offset ratio between the two is calculated to obtain the capacitance drift ratio. The posterior mean of the resonant frequency output by the digital twin model is read and compared with the baseline value of the resonant frequency under factory calibration. The offset ratio between the two is calculated to obtain the frequency drift ratio. The three dimensionless quantities—degradation severity multiple, capacitance drift ratio, and frequency drift ratio—are multiplied together to obtain the feedback trigger coefficient. The specific calculation formula is as follows: ; In the formula, K is the feedback triggering coefficient of degraded data, which is the core indicator used to comprehensively determine the urgency of open-circuit load degradation. The larger the value of the feedback triggering coefficient K, the higher the urgency of the degraded data to correct the model retraining. D represents the degree of degradation deviation. t The degradation deviation threshold that triggers feedback is a fixed value, determined statistically based on a large amount of accelerated life test data with open-circuit load. It serves as a baseline for distinguishing whether the degree of degradation has entered the range that needs attention. A value below this threshold indicates that the degradation is within an acceptable range. μ C The posterior mean of the edge capacitance output by the digital twin model under the current degradation state is obtained by performing time-series modeling of the static parameter layer and dynamic parameter layer data by the long short-term memory network, outputting the posterior probability distribution of the edge capacitance, and then taking the expected value of the probability distribution, which reflects the parasitic capacitance state of the device at the inner conductor connection end face. C b The edge capacitance reference value under the factory calibration state of open-circuit load is obtained by precise calibration measurement through electrostatic field solver before the device leaves the factory, and is stored in the static parameter layer of the digital twin model as the reference value for capacitance drift determination; μ f The posterior mean of the resonant frequency output by the digital twin model under the current degradation state is obtained by using a Bayesian long short-term memory network to perform time-series modeling of the static and dynamic parameter layer data, outputting the posterior probability distribution of the resonant frequency, and then taking the expected value of the probability distribution, which reflects the current resonant characteristic state of the device. f r0 The resonant frequency reference value under open-circuit load factory calibration is obtained through full-wave electromagnetic simulation combined with actual measurement calibration before the device leaves the factory, and is stored in the static parameter layer of the digital twin model as a reference value for frequency drift determination.

[0028] After the feedback trigger coefficient is calculated, a graded response is executed according to the numerical range of the coefficient. When the feedback trigger coefficient is lower than the first preset threshold, it is determined that the open-circuit load degradation is slight and no correction action is triggered. The digital twin model continues to perform routine degradation monitoring according to the predetermined sampling frequency. When the feedback trigger coefficient is between the first preset threshold and the second preset threshold, it is determined that the open-circuit load has a moderate degree of degradation. The posterior mean of the edge capacitance and the posterior mean of the resonant frequency output by the digital twin model are pushed to the calibration software, and the calibration software automatically updates the polynomial fitting coefficients to complete the calibration parameter compensation. When the feedback trigger coefficient is higher than the second preset threshold, it is determined that the open circuit load has seriously degraded. While completing the calibration parameter compensation, the degraded data is encapsulated as negative samples according to the timestamp. The weight parameters of the currently deployed conditional diffusion model are read from the cloud model repository. The negative sample data and the original training data are mixed in proportion to construct an incremental training dataset. The conditional diffusion model is fine-tuned to make it tend to generate anti-degradation cavity topology. After the fine-tuning is completed, the updated model weights are pushed to the cloud model repository to replace the original version. Wherein, the first preset threshold is equal to the degradation deviation threshold D. t The second preset threshold is determined by conducting accelerated life tests on a large number of open-circuit loads and statistically analyzing the distribution boundary of the trigger coefficient K under two scenarios: trigger calibration compensation and trigger model retraining.

[0029] This embodiment, through a multi-dimensional comprehensive evaluation mechanism, can identify potential risks in the early stages of degradation and proactively trigger preventive corrections, significantly improving the timeliness and reliability of open-circuit load lifecycle management. At the same time, the graded response strategy matches differentiated correction actions according to the degree of degradation: maintaining monitoring during mild degradation, automatically calibrating and compensating during moderate degradation, and driving generative model evolution during severe degradation. This realizes the transformation from passive maintenance to proactive prevention, effectively extending the effective service life of open-circuit loads and maintaining the long-term stability of calibration accuracy.

[0030] Example 5, refer to Figure 1-2 As shown, the present invention also provides a coaxial open-circuit load design system based on air medium, including a topology generation module, a degradation monitoring module and a feedback correction module; Among them, the topology generation module generates a candidate set of irregular air cavity cross-section topologies by running the conditional diffusion model, calls the full-wave electromagnetic simulation engine to extract the high-order mode cutoff frequency and edge capacitance, calculates the shape gain ratio based on the standard circular cavity, and selects and outputs the optimal cavity structure to the manufacturing execution system. The degradation monitoring module constructs an individualized digital twin model, collects geometric deviation, material parameters, insertion and removal times, and temperature cycle data in real time, and outputs the degradation deviation calculation to the feedback correction module. The feedback correction module receives the degradation deviation data, completes the feedback trigger coefficient calculation, and triggers the calibration parameter update command or the conditional diffusion model incremental retraining command according to the numerical range of the coefficient. Furthermore, the topology generation module consists of three units working collaboratively: a condition generation unit, a simulation screening unit, and a structure output unit. The condition generation unit interfaces with the connector interface standard database and the design specification input interface, reading the outer conductor inner diameter, inner conductor outer diameter, target maximum operating frequency, and target edge capacitance upper limit as condition inputs. After running the condition diffusion model to generate a set of topology candidates, it outputs the results to the simulation screening unit. The simulation screening unit interfaces with the full-wave electromagnetic simulation engine, performing simulation extraction of high-order mode cutoff frequency and edge capacitance for each candidate topology. It calculates the shape gain ratio based on the corresponding parameters of the standard circular cavity and outputs the results to the structure output unit. The structure output unit filters out candidate topologies with shape gain ratios higher than a preset threshold, sorts them from high to low, and outputs them to the manufacturing execution system. The degradation monitoring module consists of a digital twin construction unit, a data acquisition unit, and a degradation calculation unit. The digital twin construction unit interfaces with CT scanning equipment and optical 3D measurement equipment, reads geometric deviation data and material parameter data and writes them into the static parameter layer. At the same time, it interfaces with the long short-term memory network to complete the model initialization and deployment. The data acquisition unit interfaces with temperature sensors, humidity sensors, and mechanical counters to collect ambient temperature, ambient humidity, cumulative insertion and removal times, and insertion and removal torque data in real time. After writing the data into the dynamic parameter layer, it outputs the data to the degradation calculation unit. The degradation calculation unit receives the data from the static and dynamic parameter layers, selects multiple monitoring frequency points within the target frequency band to measure the real part of the input impedance, calculates the relative impedance drift amplitude of each frequency point based on the factory calibration data, and calculates the degradation deviation degree by combining the temperature cycle fatigue amplification factor. The result is then output to the feedback correction module. The feedback correction module consists of a trigger determination unit, a calibration compensation unit, and a model iteration unit. The trigger determination unit receives the degradation deviation degree, compares it with the preset trigger threshold, reads the deviation of the posterior mean of the edge capacitance output by the digital twin model from the factory reference value, and reads the deviation of the posterior mean of the resonant frequency from the factory reference value. It comprehensively calculates the feedback trigger coefficient and sends trigger commands to the calibration compensation unit or the model iteration unit according to the numerical range of the coefficient. After receiving the command, the calibration compensation unit pushes the posterior mean data to the calibration software and automatically updates the polynomial fitting coefficients to complete parameter compensation. After receiving the command, the model iteration unit reads the degradation data from the digital twin model and encapsulates it into negative samples. It reads the currently deployed conditional diffusion model weights from the cloud model repository, mixes the negative samples with the original training data to construct an incremental training dataset, performs fine-tuning training on the model, and pushes the updated weights to the cloud model repository. This embodiment constructs a complete open-circuit load design system through the collaborative work of three modules. The internal units of each module have clear division of labor and clear data flow, realizing the full-process automated management from topology generation to degradation monitoring and feedback correction. The modular system architecture allows each functional unit to be developed and deployed independently, reducing system coupling and maintenance costs. The close cooperation between the condition generation unit and the simulation screening unit improves the search efficiency of the optimal topology. The linkage between the digital twin construction unit and the degradation accounting unit enables real-time perception of device performance. The collaboration between the trigger judgment unit, the calibration compensation unit, and the model iteration unit ensures that degradation data can drive model evolution in a timely manner. The overall system operating efficiency and intelligence level are significantly improved.

[0031] Example 6, refer to Figure 1 As shown, the present invention also provides a coaxial open-circuit load device based on air medium, including an inner conductor part 1, an outer conductor part 2, a dielectric support structure 3, and an air medium cavity 4; The outer conductor 2 is the outer metal tube wall that constitutes the coaxial transmission line structure, serving as electromagnetic shielding and signal loop. The inner conductor 1 is the central metal conductor that constitutes the coaxial transmission line structure. One end of the inner conductor 1 is connected to the connector interface for connection to the test system or calibration link. The other end of the inner conductor 1 extends into the air medium cavity 4 as the electromagnetic reflection end face of the open-circuit load. The air dielectric cavity 4 is an open cavity structure in which air dielectric is filled between the end of the inner conductor part 1 and the outer conductor part 2. The radial cross-sectional shape of the air dielectric cavity 4 is the irregular cross-sectional shape corresponding to the optimal cavity structure. This irregular cross-sectional shape is generated by the conditional diffusion model and obtained by shape gain ratio screening. It can effectively improve the high-order mode cutoff frequency and reduce the edge capacitance at the connection end face of the inner conductor part 1 within the target operating frequency band, thereby improving the phase linearity of the open-circuit load reflection coefficient, reducing the polynomial fitting residual, and improving the calibration quality. The dielectric support structure 3 is disposed between the inner conductor part 1 and the outer conductor part 2. It is made of solid insulating material. The dielectric support structure 3 only undertakes the mechanical support and axial positioning functions of the inner conductor part 1 and does not participate in the electromagnetic function realization of the open circuit load. That is, the support medium and the open circuit load medium are completely separated in function, avoiding the impedance discontinuity and additional resonance problems introduced by the support medium simultaneously acting as the open circuit load medium in the traditional scheme.

[0032] This embodiment separates the support and load functions and uses air as the open-circuit load medium, which effectively reduces the edge capacitance at the connection end face of the inner conductor part 1. At the same time, it improves the high-order mode cutoff frequency and resonant frequency of the cavity, which significantly improves the phase linearity of the reflection coefficient of the open-circuit load in the operating frequency band. The low dielectric constant and low loss characteristics of air medium, combined with the topology optimization of the irregular cross-section cavity, enable the device to maintain stable electrical performance throughout its entire life cycle, providing a reliable physical basis for high-precision calibration.

[0033] It should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should also be within the scope of protection of this invention.

Claims

1. A method for designing coaxial open-circuit loads based on air medium, characterized in that: Includes the following steps: Topology generation: Construct a conditional diffusion model, using connector interface standard, target maximum operating frequency, and target edge capacitance upper limit as conditional inputs to generate a candidate set of irregular air cavity cross-section topologies. Perform full-wave electromagnetic simulation on each candidate topology to extract high-order mode cutoff frequency and edge capacitance parameters. Calculate the shape gain ratio based on the corresponding parameters of a standard circular cavity, and select candidate topologies with shape gain ratios higher than a preset threshold as the optimal cavity structure. Degradation monitoring: Construct individualized digital twin models for open-circuit loads, collect data on geometric deviations, material parameters, insertion and removal times, and temperature cycles, select multiple monitoring frequency points within the target frequency band to measure the real part of the input impedance at each frequency point, calculate the relative impedance drift amplitude at each frequency point based on the factory-calibrated real impedance, and calculate the degradation deviation by combining the temperature cycle fatigue amplification factor. Feedback correction: The degradation deviation is compared with the preset trigger threshold. The deviation of the posterior mean of the edge capacitance output by the digital twin model from the factory reference value and the deviation of the posterior mean of the resonant frequency from the factory reference value are combined to calculate the feedback trigger coefficient. Based on the numerical range of the feedback trigger coefficient, the calibration parameters are updated and the conditional diffusion model is incrementally retrained.

2. The coaxial open-circuit load design method based on air medium according to claim 1, characterized in that: The specific steps for generating the candidate set of topologies for irregularly shaped air cavity cross-sections in the topology generation step are as follows: A reverse denoising network for the conditional diffusion model is constructed. The outer conductor inner diameter, inner conductor outer diameter, target maximum operating frequency, and target edge capacitance upper limit corresponding to the connector interface standard are encoded as conditional embedding vectors. The random Gaussian noise tensor is used as the initial generation starting point. Through the iterative denoising process of the conditional diffusion model, the geometric contour data of the air cavity cross-section topology are gradually restored. The generated geometric contour data is morphologically classified and identified. When the cross-sectional contour exhibits periodic rotational gradient characteristics along the axial direction, it is marked as a spiral cavity. When the cross-sectional contour exhibits self-similar recursive nesting characteristics, it is marked as a fractal cavity. The full-wave electromagnetic simulation engine is called sequentially for the marked candidate topologies to extract the high-order mode cutoff frequency and edge capacitance parameters of each candidate topology. The extraction results are then output to the shape gain ratio calculation step for screening.

3. The coaxial open-circuit load design method based on air medium according to claim 1, characterized in that: The specific steps for constructing a personalized digital twin model in the degradation monitoring step are as follows: High-precision CT scanning and optical 3D measurement are performed on the open-circuit load before leaving the factory to obtain the actual geometric dimensions and manufacturing deviation data of the air cavity, read the actual dielectric constant and thermal expansion coefficient of the supporting medium material, and write the geometric data and material data into the static parameter layer of the digital twin model. Temperature and humidity sensors are integrated inside the open-circuit load to collect ambient temperature and humidity data in real time. A mechanical counter is integrated on the insertion and removal mechanism of the open-circuit load to collect cumulative insertion and removal counts and insertion and removal torque data in real time. The dynamically collected data is written into the dynamic parameter layer of the digital twin model. The static parameter layer data and the dynamic parameter layer data are input into the long short-term memory network. The network outputs the posterior probability distribution of the edge capacitance and the posterior probability distribution of the resonant frequency at the current time. The expected value of the probability distribution is taken as the posterior mean of the edge capacitance and the posterior mean of the resonant frequency. Multiple monitoring frequency points are selected within the target frequency band to measure the real part of the input impedance at each frequency point. The absolute value of the difference between the measured value and the reference value at each frequency point is calculated based on the factory-calibrated real part of the impedance. The relative drift amplitude of the impedance at each frequency point is obtained by dividing the absolute value of the difference between the measured value and the reference value by the reference value. The cumulative number of temperature cycles and the design allowable number of temperature cycles are calculated to obtain the temperature cycle fatigue amplification factor. The relative drift amplitude of the impedance at each frequency point is multiplied by the corresponding temperature cycle fatigue amplification factor and the average value is taken for all monitoring frequency points to obtain the degradation deviation. The posterior mean of the edge capacitance, the posterior mean of the resonant frequency, and the degradation deviation are output to the feedback correction step.

4. The coaxial open-circuit load design method based on air medium according to claim 1, characterized in that: In the feedback correction step, calibration parameter updates and incremental retraining of the conditional diffusion model are performed according to the numerical range of the feedback trigger coefficient. The specific steps are as follows: When the feedback trigger coefficient is lower than the first preset threshold, it is determined that the degree of open-circuit load degradation is within an acceptable range, and no correction action is triggered. The digital twin model continues to perform degradation monitoring at the normal sampling frequency. When the feedback trigger coefficient is between the first preset threshold and the second preset threshold, it is determined that the open-circuit load has moderate degradation. The posterior mean of the edge capacitance and the posterior mean of the resonant frequency output by the digital twin model are pushed to the calibration software. The calibration software automatically updates the polynomial fitting coefficients to complete the calibration parameter compensation. When the feedback trigger coefficient is higher than the second preset threshold, it is determined that the open circuit load has seriously degraded. While completing the calibration parameter compensation, the degraded data is encapsulated as negative samples according to the timestamp. The weight parameters of the currently deployed conditional diffusion model are read from the cloud model repository. The negative sample data is mixed with the original training data to construct an incremental training dataset. The conditional diffusion model is fine-tuned to make it tend to generate anti-degradation cavity topology. After the fine-tuning is completed, the updated model weights are pushed to the cloud model repository to replace the original version.

5. The coaxial open-circuit load design method based on air medium according to claim 4, characterized in that: When the feedback trigger coefficient is higher than the second preset threshold, the specific steps for encapsulating the degraded data into negative samples are as follows: The degradation deviation sequence, edge capacitance drift sequence, resonant frequency drift sequence, cumulative insertion / removal count sequence, and cumulative temperature cycle sequence within a preset time window before the trigger moment are read from the digital twin model. All sequences are then aligned according to timestamps and concatenated into a multidimensional degradation feature vector. Read the initial geometric parameters and initial material parameters of the open-circuit load at the time of manufacture, and associate and encapsulate the initial parameters with the multidimensional degradation feature vector to form a complete negative sample record. Add a degradation label field to the negative sample record. The value of the label field is obtained by normalizing the feedback trigger coefficient and represents the degradation severity level of the negative sample. The encapsulated negative sample records are written to the cloud negative sample database for random sampling and use during incremental training of the conditional diffusion model at a preset ratio.

6. A coaxial open-circuit load design system based on air medium, based on the coaxial open-circuit load design method based on air medium according to any one of claims 1 to 5, characterized in that: Includes a topology generation module, a degradation monitoring module, and a feedback correction module; The topology generation module is configured to generate a candidate set of irregular air cavity cross-section topologies using the running condition diffusion model, call the full-wave electromagnetic simulation engine to extract the high-order mode cutoff frequency and edge capacitance, calculate the shape gain ratio based on a standard circular cavity, and filter and output the optimal cavity structure to the manufacturing execution system. The degradation monitoring module is configured to build an individualized digital twin model, collect geometric deviation data, material parameter data, insertion and removal number data, and temperature cycle data in real time, and output the degradation deviation to the feedback correction module after calculating the degradation deviation. The feedback correction module is configured to receive degradation deviation data, calculate the feedback trigger coefficient, and trigger calibration parameter update instructions and conditional diffusion model incremental retraining instructions according to the numerical range of the feedback trigger coefficient.

7. The coaxial open-circuit load design system based on air medium according to claim 6, characterized in that: The topology generation module consists of a condition generation unit, a simulation filtering unit, and a structure output unit. The condition generation unit connects to the connector interface standard database and design index input interface, reads the outer conductor inner diameter, inner conductor outer diameter, target maximum operating frequency, and target edge capacitance upper limit as condition inputs, runs the condition diffusion model to generate a candidate set of irregular air cavity cross-section topology, and then outputs it to the simulation screening unit. The simulation screening unit is connected to the full-wave electromagnetic simulation engine. For each candidate topology, the high-order mode cutoff frequency and edge capacitance are simulated and extracted sequentially. The shape gain ratio is calculated based on the corresponding parameters of the standard circular cavity. The shape gain ratio calculation results of each candidate topology are output to the structure output unit. The structure output unit receives the shape gain ratio calculation results of each candidate topology, filters out the candidate topologies with shape gain ratios higher than a preset threshold, sorts them from high to low according to shape gain ratio, and outputs the optimal cavity structure to the manufacturing execution system.

8. The coaxial open-circuit load design system based on air medium according to claim 6, characterized in that: The degradation monitoring module consists of a digital twin construction unit, a data acquisition unit, and a degradation accounting unit; The digital twin building block interfaces with CT scanning equipment and optical 3D measurement equipment, reads the geometric deviation data and material parameter data of open circuit load and writes them into the static parameter layer of the digital twin model, and at the same time interfaces with the long short-term memory network to complete the model initialization and deployment; The data acquisition unit interfaces with temperature sensors, humidity sensors, and mechanical counters to collect ambient temperature data, ambient humidity data, cumulative insertion and removal count data, and insertion and removal torque data in real time. After writing these data into the dynamic parameter layer of the digital twin model, the data is output to the degradation calculation unit. The degradation calculation unit receives static parameter layer data and dynamic parameter layer data, selects multiple monitoring frequency points within the target frequency band to complete the measurement of the real part of the input impedance, calculates the relative impedance drift amplitude at each frequency point based on the factory calibration data, calculates the degradation deviation by combining the temperature cycle fatigue amplification factor, and outputs the calculation results to the feedback correction module.

9. The coaxial open-circuit load design system based on air medium according to claim 6, characterized in that: The feedback correction module consists of a trigger determination unit, a calibration compensation unit, and a model iteration unit. The trigger determination unit receives the degradation deviation degree output by the degradation monitoring module, reads the preset trigger threshold, compares the degradation deviation degree with the preset trigger threshold, reads the deviation degree of the posterior mean of the edge capacitance and the factory reference value output by the digital twin model and the deviation degree of the posterior mean of the resonant frequency and the factory reference value, comprehensively calculates the feedback trigger coefficient, and sends trigger commands to the calibration compensation unit and the model iteration unit respectively according to the numerical range of the feedback trigger coefficient. After receiving the trigger command, the calibration compensation unit reads the posterior mean of the edge capacitance and the posterior mean of the resonant frequency output by the digital twin model, and pushes the posterior mean data to the calibration software. The calibration software automatically updates the polynomial fitting coefficients to complete the calibration parameter compensation. After receiving the trigger command, the model iteration unit reads the degenerate data from the digital twin model and encapsulates it into negative samples. It reads the weight parameters of the currently deployed conditional diffusion model from the cloud model repository, mixes the negative sample data with the original training data to construct an incremental training dataset, performs fine-tuning training on the conditional diffusion model, and then pushes the updated model weights to the cloud model repository.

10. A coaxial open-circuit load device based on air medium, obtained based on the coaxial open-circuit load design method based on air medium according to any one of claims 1 to 5, characterized in that: It includes an inner conductor section (1), an outer conductor section (2), a dielectric support structure (3), and an air dielectric cavity (4); The outer conductor part (2) is the outer metal tube wall that constitutes the coaxial transmission line structure, and the inner conductor part (1) is the central metal conductor that constitutes the coaxial transmission line structure. One end of the inner conductor part (1) is connected to the connector interface, and the other end of the inner conductor part (1) extends into the air medium cavity. The air medium cavity (4) is an open cavity structure in which air medium is filled between the end of the inner conductor part (1) and the outer conductor part (2). The radial cross-sectional shape of the air medium cavity (4) is the irregular cross-sectional shape corresponding to the optimal cavity structure selected by the topology generation step. The dielectric support structure (3) is located between the inner conductor part (1) and the outer conductor part (2). The dielectric support structure (3) is made of solid insulating material. The dielectric support structure (3) only undertakes the mechanical support function of the inner conductor part (1) and does not participate in the electromagnetic function realization of the open circuit load.

Images