A performance test method for quasi-optical mode converter

Abstract

The present invention relates to a performance testing method for a quasi-optical mode converter, and relates to the technical field of vacuum electronics. The method comprises the following steps: obtaining structural parameters of the quasi-optical mode converter, and based on the structural parameters, performing an adjacency simulation test on the quasi-optical mode converter to obtain a virtual test result; when the virtual test result fails, optimizing the structural parameters to obtain recommended structural parameters; when the virtual test result passes, generating high-order electromagnetic waves through a quasi-optical mode exciter to perform a physical test on the quasi-optical mode converter to obtain a physical test result; and when the physical test result passes, marking the quasi-optical mode converter as a qualified performance. The present invention solves the technical problem in the prior art that the direct use of physical testing and the lack of prior analysis result in a complex test process and a long debugging cycle, thereby affecting test efficiency. The method achieves the technical effect of improving the test efficiency of the quasi-optical mode converter.

Description

Technical Field

[0001] The present invention relates to the technical field of vacuum electronics, and in particular to a performance testing method for a quasi-optical mode converter. Background Art

[0002] Quasi-optical mode converters are important components for achieving efficient output of high-power gyrotrons and are widely used in fields such as high-resolution radar and wireless communications.

[0003] Existing performance testing methods for quasi-optical mode converters typically employ physical testing to directly measure the converter and verify its mode conversion capability, efficiency, and loss characteristics. The lack of an initial performance evaluation reference at the start of testing necessitates repeated signal excitation and measurement, complicating the testing process. Furthermore, each parameter adjustment may require a series of retest operations, lengthening the debugging cycle. For example, in some studies, achieving high power transmission efficiency requires multiple designs and experimental verifications of the radiator and mirror system. These shortcomings severely impact the testing efficiency of quasi-optical mode converters, preventing them from meeting the demands for efficient testing in today's rapidly evolving technologies. Summary of the Invention

[0004] The present invention provides a performance testing method for a quasi-optical mode converter to solve the technical problems in the prior art of directly adopting physical testing and lacking prior analysis, which leads to complex testing processes and long debugging cycles, thus affecting testing efficiency.

[0005] The technical solution of the present invention to solve the above technical problems is as follows:

[0006] The present invention provides a performance testing method for a quasi-optical mode converter, the method comprising: obtaining structural parameters of the quasi-optical mode converter, and performing an adjacency simulation test on the quasi-optical mode converter based on the structural parameters to obtain a virtual test result; when the virtual test result is a failure, optimizing the structural parameters to obtain recommended structural parameters; when the virtual test result is a pass, generating high-order electromagnetic waves through a quasi-optical mode exciter to perform a physical test on the quasi-optical mode converter to obtain a physical test result; and when the physical test result is a pass, marking the quasi-optical mode converter as having qualified performance.

[0007] The beneficial effects of the present invention are as follows: first, by obtaining the structural parameters of the quasi-optical mode converter, an adjacent simulation test is performed to obtain a virtual test result. Through virtual testing, potential problems can be discovered and solved in advance during the design stage, reducing the number and time of physical testing, thereby improving the overall test efficiency. When the virtual test result fails, the structural parameters are optimized to obtain recommended structural parameters, thereby improving the performance and conversion efficiency of the quasi-optical mode converter. When the virtual test result passes, a high-order electromagnetic wave exciter is used to perform a physical test, which can more realistically simulate the actual working conditions, thereby more accurately reflecting the performance of the quasi-optical mode converter in actual applications. This method of combining virtual testing with physical testing not only simplifies the test process and shortens the debugging cycle, but also improves the accuracy and reliability of the test results. In addition, a quality tracking system is established through performance qualification identification, which enhances the controllability of product quality and provides a guarantee for the stability and reliability of the quasi-optical mode converter in actual applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A schematic flow chart of a performance testing method for a quasi-optical mode converter provided by the present invention;

[0009] Figure 2 A schematic diagram of a flow chart for obtaining physical test results in a performance test method for a quasi-optical mode converter provided by the present invention;

[0010] Figure 3 A schematic diagram of a flow chart for obtaining recommended structural parameters in a performance testing method for a quasi-optical mode converter provided by the present invention. DETAILED DESCRIPTION

[0011] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0012] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the specified features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0013] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or illustration". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0014] like Figure 1 As shown, an embodiment of the present invention provides a performance testing method for a quasi-optical mode converter, the method comprising:

[0015] Step S1: obtaining structural parameters of a quasi-optical mode converter, and performing an adjacency simulation test on the quasi-optical mode converter based on the structural parameters to obtain a virtual test result.

[0016] Specifically, the design file of the interactive quasi-optical mode converter is used to obtain the structural parameters of the quasi-optical mode converter, including dimensions (such as length and width), material properties (such as dielectric constant and conductivity) and geometric structure. An adjacent simulation test is performed on the quasi-optical mode converter, that is, the historical test data of the quasi-optical mode converter with similar structural parameters to the quasi-optical mode converter is used to determine whether the performance of the quasi-optical mode converter meets the requirements, thereby obtaining a virtual test result. This virtual test result includes two cases: "pass" and "fail". When the virtual test result is "pass", it means that the performance of the quasi-optical mode converter meets the expected requirements under the simulation environment; when the virtual test result is "fail", it means that the quasi-optical mode converter has performance problems under the simulation environment, such as low mode conversion efficiency, incorrect output mode, etc.

[0017] Through step S1, the performance of the quasi-optical mode converter can be quickly evaluated by utilizing previous test data, and potential design problems can be discovered in advance.

[0018] Step S2: When the virtual test result is failure, optimizing the structural parameters to obtain recommended structural parameters.

[0019] Specifically, when the virtual test result in step S1 fails, the structural parameters are used as input variables, the performance target is set (such as conversion efficiency>90%), and the optimization algorithm (such as genetic algorithm, particle swarm optimization algorithm, etc.) is used to optimize the structural parameters to obtain new structural parameters. Then, the adjacency simulation test is performed again based on the new structural parameters to obtain new virtual test results. Check whether the new virtual test result meets the design requirements. If it still fails, continue to optimize the structural parameters, and after multiple iterative calculations, until the virtual test result passes. The structural parameters when the virtual test result passes are output as recommended structural parameters. This recommended structural parameter can significantly improve the performance of the quasi-optical mode converter and can be used as a reference for further optimization of the design.

[0020] By optimizing the structural parameters, the structural parameters of the quasi-optical mode converter can be automatically and efficiently adjusted, thereby improving the efficiency of finding suitable structural parameters and avoiding the blindness of manual adjustment.

[0021] Step S3: When the virtual test result is passed, a quasi-optical mode exciter generates high-order electromagnetic waves to perform a physical test on the quasi-optical mode converter to obtain a physical test result.

[0022] Specifically, a quasi-optical mode exciter is a device that can generate specific high-order electromagnetic waves. When the virtual test result is passed, the output of the quasi-optical mode exciter is connected to the input of the quasi-optical mode converter, and the quasi-optical mode exciter is used to generate high-order electromagnetic waves. At the output of the quasi-optical mode converter, instruments such as a power meter and a spectrum analyzer are used to measure the output electromagnetic wave parameters to obtain the physical test result. Similar to the virtual test results, the physical test results are also divided into two cases: "pass" and "fail". When the physical test result is "pass", it means that the performance of the quasi-optical mode converter meets the requirements in the actual environment; when the physical test result is "fail", it means that there are performance problems with the quasi-optical mode converter in the actual environment.

[0023] Through physical testing, the performance of the quasi-optical mode converter can be verified in a real environment to ensure that it can meet the expected performance requirements under real working conditions.

[0024] Step S4: When the physical test result is passed, the quasi-optical mode converter is marked as having qualified performance.

[0025] Specifically, if the physical test result in step S3 is a pass, the quasi-optical mode converter is marked as qualified by marking it, attaching a label (such as a QR code or RFID tag), or noting it in the relevant test records, and recording its test data and tracking information. This qualified performance marking facilitates product quality control during the production and R&D process, easily identifying quasi-optical mode converters that have passed testing and meet performance standards, and helps improve production efficiency and product quality management.

[0026] Further, such as Figure 2 As shown, step S3 includes:

[0027] When the virtual test result is passed, when a physical test is performed on the quasi-optical mode converter by generating high-order electromagnetic waves through the quasi-optical mode exciter:

[0028] Step S31: Using a three-dimensional mobile platform equipped with a vector network analyzer, the horizontal component electric field of the output field is detected at distances of 60 mm, 100 mm, 140 mm, 160 mm, and 180 mm from the axis of the quasi-optical mode converter.

[0029] Step S32: collecting the output field normalized temperature distribution equipotential map by a thermal imager.

[0030] Step S33: performing beam energy stability verification based on the horizontal component electric field of the output field to obtain a first verification result.

[0031] Step S34: performing a Gaussian distribution check based on the output field normalized temperature distribution equipotential map to obtain a second check result.

[0032] Step S35: When both the first verification result and the second verification result are passed, the entity test result is passed.

[0033] Step S36: Otherwise, the entity test result is failed.

[0034] Specifically, if the virtual test result is a pass, a physical test of the quasi-optical mode converter is performed using a quasi-optical mode exciter to generate high-order electromagnetic waves. The specific test steps include: First, a vector network analyzer is mounted on a three-dimensional mobile platform. Then, the three-dimensional mobile platform is controlled to move to specific distances from the axis of the quasi-optical mode converter, namely 60mm, 100mm, 140mm, 160mm, and 180mm. At each position, the vector network analyzer measures the horizontal component electric field of the quasi-optical mode converter's output field. The three-dimensional mobile platform is a device that can precisely control its movement position, supporting fine-tuning along the X, Y, and Z axes in space to enable measurement or testing at different locations. The horizontal component electric field is the component of the electric field in the horizontal plane and is one of the key parameters for electromagnetic wave mode conversion. Next, a thermal imager is used to detect the output field of the quasi-optical mode converter. The thermal imager collects temperature information of the output field, obtains temperature distribution data caused by electromagnetic wave heating, and converts this data into a normalized temperature distribution equipotential map. This normalized temperature distribution equipotential map normalizes the temperature distribution data and then generates an equipotential line map, which can intuitively reflect the distribution of heating effects caused by the electromagnetic field.

[0035] Based on the measured horizontal electric field component of the output field, a beam energy stability check is performed to analyze whether the energy distribution of the electromagnetic wave remains stable within a specified range, obtaining a first verification result. This first verification result is classified as either pass or fail. The beam energy stability check verifies the energy stability of the quasi-optical mode converter's output beam.

[0036] A Gaussian-like distribution check is performed based on the collected output field normalized temperature distribution equipotential maps to check whether the output field temperature distribution conforms to a Gaussian distribution, verify the converter's beamforming capability, and obtain a second verification result. The second verification result is also divided into two categories: pass and fail. The Gaussian-like distribution check verifies whether the temperature distribution of the quasi-optical mode converter output field conforms to the expected physical model, helping to determine whether its internal physical processes are normal. If the temperature distribution does not conform to a Gaussian-like distribution, it may indicate a problem with the energy conversion or propagation mechanism within the quasi-optical mode converter.

[0037] When both the first verification result and the second verification result are passed, the entity test result is determined to be passed. If one of the first verification result and the second verification result is failed or both are failed, the entity test result is determined to be failed.

[0038] Through the above steps, the performance of the quasi-optical mode converter in terms of beam energy stability and temperature distribution characteristics is comprehensively considered to determine the physical test results, which can more comprehensively and accurately evaluate the performance of the quasi-optical mode converter.

[0039] Furthermore, step S32 includes: obtaining an output field horizontal component electric field verification model, processing the output field horizontal component electric field, and obtaining the first verification result; wherein the step of constructing the output field horizontal component electric field verification model includes:

[0040] Step S321: Obtain the horizontal component of the output field detected at 60 mm, 100 mm, 140 mm, 160 mm, and 180 mm from the axis of the quasi-optical mode converter and record the electric field.

[0041] Step S322: When the variance of the field spot areas at multiple cross sections of the output field horizontal component recording the electric field is less than or equal to the variance threshold, the alignment check result is marked as passed; otherwise, the alignment check result is marked as failed.

[0042] Step S323: When the phase fluctuation trends of the output window ranges at multiple cross sections of the output field horizontal component recording the electric field are the same, the phase consistency check result is marked as passed; otherwise, the phase consistency check result is marked as failed.

[0043] Step S324: When both the collimation check result and the phase consistency check result are passed, the check result is marked as passed; otherwise, the check result is marked as failed.

[0044] Step S325: using the verification result identification information as supervision and the output field horizontal component recorded electric field as input, training the convolution classifier to obtain the output field horizontal component electric field verification model.

[0045] Specifically, when performing beam energy stability verification based on the output field horizontal component electric field, first construct an output field horizontal component electric field verification model, and input the output field horizontal component electric field into the output field horizontal component electric field verification model for processing, thereby obtaining a first verification result.

[0046] The output field horizontal component electric field calibration model is constructed based on a convolution classifier. By collecting beam energy stability calibration data of multiple quasi-optical mode converters as sample data, the convolution classifier is trained to obtain the output field horizontal component electric field calibration model. The specific training process is as follows:

[0047] First, for multiple quasi-optical mode converters, a vector network analyzer was used to measure the horizontal component electric field of the output field at 60 mm, 100 mm, 140 mm, 160 mm, and 180 mm from the axis of the quasi-optical mode converter. The electric field data at these locations were recorded to form the recorded electric field of the horizontal component of the output field. These recorded electric fields will serve as input data for subsequent training of the convolutional classifier. After obtaining the recorded electric fields of the horizontal component of the output field of multiple quasi-optical mode converters, this data needs to be labeled and assigned a pass or fail label.

[0048] The electric field is recorded based on the measured horizontal component of the output field, and image processing algorithms (such as edge detection and threshold segmentation) are used to identify field spots on different cross-sections (60mm, 100mm, 140mm, 160mm, and 180mm from the axis of the quasi-optical mode converter). These field spots are areas with higher electric field intensity and can be extracted by setting a threshold. On each cross-section, the area of ​​each field spot is calculated by segmenting the electric field intensity image or performing a region recognition algorithm. For all field spots on the same cross-section, the variance of their areas is calculated to obtain the variance of the field spot areas at multiple cross-sections. These field spot areas reflect the degree of difference in the electric field distribution at different cross-sections. The smaller the variance, the more uniform the electric field distribution. The calculated variance is compared with a pre-set variance threshold. If the variance is less than or equal to the variance threshold, the collimation check result is marked as passed; otherwise, it is marked as failed.

[0049] For each cross section, record the phase data in the output window (obtained by a vector network analyzer) and observe the changes in the phase in the output window. The phase change with position is obtained by calculating the phase difference between adjacent measurement points. Based on the phase change with position, calculate the change trend of the phase data of multiple cross sections, and determine whether the phase fluctuation trend of the output window range at multiple cross sections is the same. The phase fluctuation trend can be determined by comparing the statistical quantities such as the mean and variance of the phase difference of different cross sections. If the phase fluctuation trends of multiple cross sections are similar, for example, they are gradually increasing or decreasing, or remain relatively stable, it means that the beam is relatively stable, and the phase consistency check result is marked as passed; otherwise, it is marked as failed.

[0050] Check the collimation and phase consistency results. If both pass, mark the result as passed; otherwise, mark it as failed. By comprehensively considering the collimation and phase consistency results, you can more comprehensively evaluate the characteristics of the horizontal component of the output field.

[0051] The convolutional classifier is trained using a deep learning framework (such as TensorFlow, Keras, or PyTorch), using the verification result identification information (pass or fail) as supervisory information and the output field horizontal component recorded electric field as input data. An appropriate convolutional neural network architecture is selected to construct the convolutional classifier. For example, a classic convolutional neural network architecture such as LeNet-5 can be chosen. The input data (the output field horizontal component recorded electric field data) and the supervisory data (the verification result identification information) are divided into a training set, a validation set, and a test set. This division is generally based on a certain ratio, such as 70% of the data as a training set, 15% as a validation set, and 15% as a test set. During the training phase, the input data of the training set is input into the convolutional classifier. After calculations in the convolutional layer, pooling layer, and fully connected layer, a prediction result (pass or fail) is obtained. A loss function, such as the cross-entropy loss function, is calculated between the predicted result and the supervisory data (the actual verification result identification information). Based on the value of the loss function, an optimization algorithm (such as stochastic gradient descent and its variants Adagrad and Adadelta) is used to update the weight parameters in the convolutional classifier, gradually reducing the value of the loss function. After each training cycle, the trained model is validated using the validation set data, and performance indicators such as accuracy and recall on the validation set are observed. If performance indicators no longer improve or overfitting occurs (such as a decrease in accuracy on the validation set), training is terminated. The trained convolutional classifier is evaluated using the test set data, and performance indicators such as accuracy, recall, and F1 score on the test set are calculated to determine the model's generalization ability. If the model's performance indicators are unsatisfactory, the model can be optimized, such as adjusting the convolutional neural network structure (increasing or decreasing the number of convolutional and pooling layers), adjusting training parameters (such as the learning rate and training cycle), or increasing the amount of data. Training can then be repeated until a satisfactory output field horizontal component electric field calibration model is obtained.

[0052] The output horizontal component electric field calibration model enables automatic calibration of the output horizontal component electric field. When new output horizontal component electric field data is available, the model can be used directly for calibration, improving calibration efficiency and accuracy and reducing errors associated with manual analysis.

[0053] Furthermore, step S33 includes: obtaining a Gaussian distribution calibration model, processing the output field normalized temperature distribution equipotential map, and obtaining the second calibration result; wherein the Gaussian distribution calibration model construction step includes:

[0054] Step S331: Obtain output field normalized temperature distribution equipotential map recording data.

[0055] Step S332: constructing a histogram of the output field normalized temperature distribution equipotential map recording data.

[0056] Step S333: When the histogram is high in the middle and low on both sides, and at least 80 percent of the data is symmetrically distributed, the verification result of the output field normalized temperature distribution equipotential map recorded data is marked as passed; otherwise, the verification result of the output field normalized temperature distribution equipotential map recorded data is marked as failed.

[0057] Step S334: using the verification result identifier as supervision and the output field normalized temperature distribution equipotential map record data as input, training the convolution classifier to obtain the Gaussian-like distribution verification model.

[0058] Specifically, when performing quasi-Gaussian distribution verification based on the output field normalized temperature distribution equipotential map, a quasi-Gaussian distribution verification model is first constructed, and the output field normalized temperature distribution equipotential map is input into the quasi-Gaussian distribution verification model for processing, thereby obtaining a second verification result.

[0059] Similar to the output field horizontal component electric field calibration model, the Gaussian-like distribution calibration model is also built based on the convolution classifier. By collecting the Gaussian-like distribution calibration data of multiple quasi-optical mode converters as sample data, the convolution classifier is trained to obtain the Gaussian-like distribution calibration model. The specific training process is as follows:

[0060] First, for multiple quasi-optical mode converters, a thermal imager is used to capture the normalized temperature distribution contour maps of the output field of the quasi-optical mode converter. The corresponding recorded data for these contour maps is then acquired. Using statistical analysis tools or software, the acquired output field normalized temperature distribution contour map data is constructed into a histogram, with the temperature range as the horizontal axis and the frequency of data occurrence within that range as the vertical axis. This histogram construction allows for a more intuitive observation of the distribution characteristics of the output field normalized temperature distribution contour map data. For example, it can be seen whether the temperature data is concentrated in a certain range or is relatively dispersed.

[0061] Observe the shape characteristics of the constructed histogram to determine whether it satisfies a quasi-Gaussian distribution, i.e., whether it is high in the center and low on both sides, and whether at least 80% of the data is symmetrically distributed. If these conditions are met, the verification result of the output field normalized temperature distribution equipotential map recorded data is marked as passed; otherwise, it is marked as failed.

[0062] Using the verification result (pass or fail) as supervisory information and the output field normalized temperature distribution equipotential map data as input data, the convolution classifier is trained to obtain a Gaussian-like distribution verification model. The training process of the Gaussian-like distribution verification model is similar to that of the output field horizontal component electric field verification model. Those skilled in the art will clearly understand the training process of the Gaussian-like distribution verification model by referring to the above description of the output field horizontal component electric field verification model training process.

[0063] The Gaussian-like calibration model automatically verifies new output field normalized temperature distribution and other potential map data. In subsequent testing, manual histogram construction and feature analysis are no longer necessary; the model can be used directly to quickly obtain calibration results, improving both efficiency and accuracy.

[0064] Furthermore, step S1 includes:

[0065] Step S11: configuring a structural parameter threshold, wherein the structural parameter threshold represents a maximum fault tolerance deviation of the structural parameter.

[0066] Step S12: Obtain sample structural parameters and sample performance test record data of the sample quasi-optical mode converter.

[0067] Step S13: When the structural parameter deviation between the sample structural parameter and the structural parameter meets the structural parameter threshold, the sample performance test record data is added to the first data set to be analyzed.

[0068] Step S14: when the first data set to be analyzed meets the first data volume, traverse the sample structure parameters of the first data set to be analyzed, and based on the structure parameter threshold, collect the second data set to be analyzed.

[0069] Step S15: When the second data set to be analyzed meets the second data volume, the passing ratio of the first data set to be analyzed is counted and set as a first passing probability, and the passing ratio of the second data set to be analyzed is counted and set as a second passing probability, wherein the first passing probability has a first weight, the second passing probability has a second weight, and the first weight is greater than the second weight.

[0070] Step S16: performing a weighted mean analysis on the first pass probability and the second pass probability according to the first weight and the second weight to obtain a fusion pass probability.

[0071] Step S17: When the fusion pass probability is greater than or equal to the pass probability threshold, the virtual test result is passed; otherwise, the virtual test result is failed.

[0072] Specifically, the structural parameter threshold refers to the maximum tolerance for the structural parameters of the quasi-optical mode converter, i.e., the upper limit of the permissible error. This threshold is used to determine whether the structural parameter deviations during converter testing are within acceptable limits. Based on the design requirements of the quasi-optical mode converter, the tolerance ranges for various structural parameters (such as dimensions and materials) are set to ensure they remain within reasonable tolerances.

[0073] Sample quasi-optical mode converters are used for testing and analysis. They have specific structural parameters and corresponding sample performance test records. These samples are selected from the overall population of quasi-optical mode converters and used to infer the performance of the entire quasi-optical mode converter population. Sample structural parameters (such as dimensions, angles, and materials) and performance test results (such as test pass / fail status, power loss, and output stability) are collected for the sample quasi-optical mode converters.

[0074] The sample structural parameters are traversed, and the deviation between the sample structural parameters and the structural parameters (the structural parameters of the quasi-optical mode converter to be tested) is calculated. This deviation is then compared with a structural parameter threshold. If the deviation meets the structural parameter threshold (is within the threshold range), the sample structural parameters and the corresponding sample performance test record data are added to the first dataset to be analyzed.

[0075] When the number of samples in the first dataset to be analyzed reaches a first data volume, the sample structure parameters in the first dataset to be analyzed are traversed, and sample structure parameters whose deviations from the sample structure parameters in the first dataset to be analyzed meet a structure parameter threshold are screened from the remaining sample structure parameters and sample performance test results. These sample structure parameters and the corresponding sample performance test records are then added to the second dataset to be analyzed. The first data volume refers to the number of samples in the first dataset to be analyzed, which is a preset threshold.

[0076] When the second data set to be analyzed meets the second data volume, the pass ratios of the first data set to be analyzed and the second data set to be analyzed are calculated respectively. The second data volume is similar to the first data volume, and is a preset quantity threshold, which represents the number of samples in the second data set to be analyzed. For the first data set to be analyzed, the number of samples that pass the performance test is counted and divided by the total number of samples to obtain the first pass probability; similarly, the second data set to be analyzed is calculated to obtain the second pass probability. Corresponding weights are set for the first pass probability and the second pass probability, the first pass probability has a first weight, and the second pass probability has a second weight. Since the sample structure parameters in the first data set to be analyzed are closer to the structure parameters of the quasi-optical mode converter to be detected, a higher weight is assigned to the first pass probability, that is, the first weight is greater than the second weight.

[0077] Based on the first and second weights, a weighted mean analysis is performed on the first and second pass probabilities to obtain the fused pass probability. Weighted mean analysis comprehensively considers the pass probabilities of the two datasets and yields a more comprehensive and reasonable fused pass probability.

[0078] The fused pass probability is compared with the pass probability threshold. This pass probability threshold is a set minimum pass probability used to determine whether the test result passes. If the fused pass probability is greater than or equal to the pass probability threshold, the virtual test result is considered a pass; otherwise, it is considered a fail. This method allows accurate virtual test results for the quasi-optical mode converter based on pre-set criteria, providing a basis for subsequent physical testing.

[0079] The above steps provide a precise determination method for virtual testing of quasi-optical mode converters. This method not only compares structural parameters with sample performance data but also analyzes the pass probability of multiple data sets using a weighted mean, thereby improving the accuracy and stability of virtual testing. Passing samples can be effectively screened during the virtual testing phase, reducing the workload of physical testing, thereby improving overall testing efficiency and reducing development cycles and costs.

[0080] Further, such as Figure 3 As shown, step S2 includes:

[0081] Step S21: Obtaining the structural parameter constraint interval.

[0082] Step S22: constructing a high-dimensional distribution space according to the structural parameter constraint interval, wherein any point in the high-dimensional distribution space represents a group of structural parameters.

[0083] Step S23: uniformly distribute the particles in the high-dimensional distribution space according to a preset distance threshold to obtain an initial structure particle set.

[0084] Step S24: using the fusion success probability as the fitness function, performing optimization based on the initial structure particle set to obtain the recommended structure parameters.

[0085] Specifically, the structural parameter constraint interval is determined based on factors such as the design requirements, manufacturing process limitations, and physical principles of the quasi-optical mode converter. This structural parameter constraint interval is the range of values ​​for the structural parameters of the quasi-optical mode converter and is used to limit the optimization interval for the structural parameters. For example, based on design requirements, a structural parameter may have a restricted range of values ​​to meet certain optical principles. Similarly, based on manufacturing process limitations and considerations of machining accuracy, a reasonable range of values ​​is also determined. The determination of the structural parameter constraint interval provides boundary conditions for constructing a high-dimensional distribution space, ensuring that the considered structural parameter combinations are feasible.

[0086] Use mathematical modeling tools (such as MATLAB, Python) to construct a high-dimensional distribution space based on the structural parameter constraint interval. If there are n structural parameters and each structural parameter has its corresponding constraint interval, then this high-dimensional distribution space is an n-dimensional space, and the points in the space (x1, x2, ..., x n ) represents a set of structural parameters, where x i Satisfy the constraint interval of the i-th structural parameter. Constructing a high-dimensional distribution space can comprehensively represent all possible structural parameter combinations of the quasi-optical mode converter, providing a search space for the subsequent optimization process.

[0087] In a high-dimensional distribution space, points are uniformly selected according to a preset distance threshold. These points form the initial set of structured particles. The preset distance threshold is the minimum distance between adjacent particles in the space, ensuring a uniform distribution of particles and avoiding excessive concentration or duplication. This initial set of structured particles provides the starting search point for the optimization process. This uniform distribution ensures broad coverage of the entire high-dimensional distribution space, preventing the omission of potential optimal solutions.

[0088] Using the fusion probability as the fitness function, the fitness value (i.e., fusion probability) is calculated for each particle in the initial set of structured particles (i.e., the point representing a set of structural parameters). An optimization algorithm is then used to optimize the initial set of structured particles. Taking the particle swarm optimization algorithm as an example, each particle updates its position based on its own velocity and position, as well as the position of the optimal particle in the swarm. This process continuously calculates the fitness value (fusion probability) of each new position. After multiple iterations, the particle with the highest fitness value (i.e., the highest fusion probability) is found, and the structural parameters represented by this particle are the recommended structural parameters. By optimizing the fusion probability as the fitness function, it is possible to identify optimal structural parameter combinations that meet performance requirements (high fusion probability), providing a reference for the design of quasi-optical mode converters.

[0089] By establishing a high-dimensional distribution space and applying algorithms such as particle swarm optimization to optimize the structural parameters of the quasi-optical mode converter, the optimal structural parameters are obtained. By gradually optimizing the parameters using a uniformly distributed initial set of particles and using the fusion pass probability as the fitness function, the accuracy of structural parameter selection is effectively improved, ensuring that the selected parameters achieve a higher pass probability in performance testing. This process not only improves test accuracy but also reduces the workload of manual debugging and trial-and-error through intelligent optimization, thereby improving the research and development efficiency of quasi-optical mode converters.

[0090] The above steps, through a uniformly distributed initial particle set, gradually optimize parameters and use the fusion pass probability as the fitness function. This effectively improves the accuracy of structural parameter selection and ensures that the selected parameters achieve a higher pass probability in performance testing. This process not only improves test accuracy but also reduces the workload of manual debugging and trial and error through intelligent optimization, thereby improving the performance testing efficiency of quasi-optical mode converters.

[0091] Furthermore, step S24 includes:

[0092] Step S241: obtaining an initial structure particle fusion passing probability set of the initial structure particle set.

[0093] Step S242: When the number of initial structure particles whose fusion pass probability set satisfies the pass probability threshold is 0, obtain i target distribution positions of initial structure particles with a larger fusion pass probability and j starting distribution positions of initial structure particles with a smaller fusion pass probability, where i and j are integers and i is less than j.

[0094] Step S243: extracting a first starting distribution position according to the j starting distribution positions of the initial structure particles.

[0095] Step S244: extracting a first target distribution position based on the i initial structure particle target distribution positions.

[0096] Step S245: connecting the first starting distribution position and the first target distribution position in the high-dimensional distribution space to obtain the line-adjacent particle fusion passing probability of the line-adjacent initial structure particles whose distribution distance to the connecting line is less than or equal to the distribution distance threshold.

[0097] Step S246: When the line-adjacent particle fusion passing probability is less than the initial distribution position fusion passing probability of the first initial distribution position, the first target distribution position is updated based on the i initial structure particle target distribution positions.

[0098] Step S247: Otherwise, the first target distribution position is used as the target and the first initial distribution position is used as the starting point to search and obtain newly added structure particles.

[0099] Step S248: performing iterative optimization based on the newly added structural particles to obtain the recommended structural parameters.

[0100] Specifically, for each particle in the initial structure particle set, the fusion pass probability is calculated according to the above-mentioned method for calculating the fusion pass probability, thereby obtaining the initial structure particle fusion pass probability set. For example, if there are n particles in the initial structure particle set, the fusion pass probabilities P1, P2, ..., P of these n particles are calculated respectively. n, then the probability set of initial structure particles fusion is {P1, P2, ..., P n The fusion of initial structure particles provides a basis for subsequent judgment of the performance of initial structure particles through probability set, and can understand the initial state of each particle in the initial structure particle set.

[0101] Check the elements in the initial structure particle fusion probability set and count the number of particles that meet the passing probability threshold. If this number is 0, find the distribution positions of i particles with a larger fusion passing probability (recorded as the initial structure particle target distribution position) and the distribution positions of j particles with a smaller fusion passing probability (recorded as the initial structure particle starting distribution position) from the initial structure particle set. For example, the initial structure particle fusion passing probability set can be sorted, and then the particle positions corresponding to the first i larger values ​​and the particle positions corresponding to the last j smaller values ​​can be selected. Among them, i and j are integers, and i is less than j. When no particle meets the passing probability threshold, by distinguishing the target distribution position and the starting distribution position, a direction is provided for the subsequent optimization, that is, exploring from the position with a small fusion passing probability to the position with a large fusion passing probability.

[0102] A position is randomly or according to a certain rule (such as according to the index order, etc.) extracted from the initial distribution positions of the j initial structure particles, and is defined as the first initial distribution position.

[0103] A position is randomly or according to a certain rule (such as according to the index order, etc.) extracted from the i initial structure particle target distribution positions and defined as the first target distribution position.

[0104] In the high-dimensional distribution space, the distance from each initial structure particle to the connecting line connecting the first starting distribution position and the first target distribution position is calculated. This distance may be calculated using a high-dimensional space distance formula, such as a generalized form of the Euclidean distance formula in high-dimensional space. Initial structure particles whose distance is less than or equal to the distribution distance threshold are selected as line-adjacent initial structure particles. The fusion pass probability of these particles is then obtained, denoted as the line-adjacent particle fusion pass probability. By analyzing the line-adjacent particle fusion pass probability, the performance of particles near the path connecting the first starting distribution position and the first target distribution position can be understood, providing a basis for subsequent decision-making.

[0105] Compare the probability of line-neighboring particle fusion passing with the probability of the initial distribution position fusion passing at the first initial distribution position. If the line-neighboring particle fusion passing probability is lower than the initial distribution position fusion passing probability, it indicates that the current target direction may not be optimal and a new target distribution position needs to be selected. At this point, the first target distribution position is updated based on the i initial structure particle target distribution positions. For example, the next unselected initial structure particle target distribution position can be selected. This update mechanism avoids falling into a local optimum during the optimization process and allows exploration of more optimal structural parameter combinations by adjusting the target distribution position.

[0106] If the probability of merging adjacent particles is greater than or equal to the probability of merging the first starting distribution position, it indicates that there is value in exploring this direction. Using the first target distribution position as the target and the first starting distribution position as the starting point, a search strategy (such as gradually moving along the connecting line and making fine adjustments) is performed in the high-dimensional distribution space to obtain new structural particles. The introduction of new structural particles provides new structural parameter combinations for the optimization process, helping to find more optimal recommended structural parameters.

[0107] New structural particles are added to the optimization process, and the fusion probability is used as the fitness function. The optimization steps (comparison, update, search, etc.) are repeated until the stopping conditions (such as reaching a certain number of iterations or a satisfactory fusion probability) are met. Finally, the recommended structural parameters are obtained. Through iterative optimization, the structural parameter combination can be continuously optimized, and finally the recommended structural parameters with good fusion probability are obtained.

[0108] The above steps construct a high-dimensional distribution space and perform iterative optimization based on fusion and probability, continuously adjusting and improving the structural parameter combinations based on the initial particle set. This process fully utilizes efficient search and probabilistic evaluation mechanisms to ensure that the recommended structural parameters have optimal performance, significantly improving the design efficiency and test reliability of quasi-optical mode converters.

[0109] Furthermore, the method further includes: when the physical test result is a failure, marking the quasi-optical mode converter as having a performance abnormality.

[0110] Specifically, when the physical test result is a failure, the performance of the quasi-optical mode converter needs to be marked as abnormal in order to subsequently troubleshoot the problem and optimize the design. The performance abnormality mark can guide subsequent analysis operations. For example, technicians can quickly locate the quasi-optical mode converter based on the mark, and then, based on the information attached to the performance abnormality mark, combined with previous virtual tests, structural parameter analysis, etc., find out the cause of the performance abnormality, which may be structural parameter deviation, manufacturing process defects, or other external factors. In terms of quality control, performance abnormality marking helps to count and analyze the overall performance of the quasi-optical mode converter. By classifying and counting the quasi-optical mode converters with performance abnormality marks, it can be understood which batches or types of quasi-optical mode converters are prone to performance problems, thereby improving the production process, design, etc.

[0111] The performance testing method of a quasi-optical mode converter provided by an embodiment of the present invention has at least the following technical effects:

[0112] First, the structural parameters of the quasi-optical mode converter are obtained, and adjacency simulation tests are performed to obtain virtual test results. Virtual testing can identify and resolve potential issues early in the design phase, reducing the number and time of physical testing and thus improving overall testing efficiency. If the virtual test results fail, an optimization algorithm is used to optimize the structural parameters and obtain recommended parameters, thereby improving the performance and conversion efficiency of the quasi-optical mode converter. If the virtual test results pass, physical testing using a high-order electromagnetic wave exciter can more realistically simulate actual operating conditions, thereby more accurately reflecting the performance of the quasi-optical mode converter in actual applications. This combined virtual and physical testing approach not only simplifies the testing process and shortens the debugging cycle, but also improves the accuracy and reliability of the test results. Furthermore, a quality tracking system, established through performance qualification and failure identification, enhances product quality control and ensures the stability and reliability of the quasi-optical mode converter in actual applications.

[0113] Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts become known.

[0114] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A performance testing method for a quasi-optical mode converter, characterized in that: include: Obtaining structural parameters of the quasi-optical mode converter, and performing an adjacency simulation test on the quasi-optical mode converter based on the structural parameters to obtain a virtual test result; When the virtual test result is failure, optimizing the structural parameters to obtain recommended structural parameters; When the virtual test result is passed, a quasi-optical mode converter is physically tested by generating a high-order electromagnetic wave through a quasi-optical mode exciter to obtain a physical test result; When the physical test result is passed, marking the quasi-optical mode converter as having qualified performance; The method of performing an adjacency simulation test on the quasi-optical mode converter based on the structural parameters to obtain a virtual test result includes: Configuring a structural parameter threshold, wherein the structural parameter threshold represents a maximum fault tolerance deviation of the structural parameter; Obtaining sample structural parameters and sample performance test record data of a sample quasi-optical mode converter; When the structural parameter deviation between the sample structural parameter and the structural parameter satisfies the structural parameter threshold, adding the sample performance test record data into the first data set to be analyzed; When the first data set to be analyzed meets the first data volume, traversing the sample structure parameters of the first data set to be analyzed, and collecting the second data set to be analyzed based on the structure parameter threshold; When the second data set to be analyzed meets the second data volume, a pass ratio of the first data set to be analyzed is calculated and set as a first pass probability, and a pass ratio of the second data set to be analyzed is calculated and set as a second pass probability, wherein the first pass probability has a first weight, the second pass probability has a second weight, and the first weight is greater than the second weight; performing a weighted mean analysis on the first pass probability and the second pass probability according to the first weight and the second weight to obtain a fusion pass probability; When the fusion pass probability is greater than or equal to the pass probability threshold, the virtual test result is passed; otherwise, the virtual test result is failed.

2. The method according to claim 1, wherein When the virtual test result is passed, a physical test is performed on the quasi-optical mode converter by generating a high-order electromagnetic wave by a quasi-optical mode exciter to obtain a physical test result, including: When the virtual test result is passed, when a physical test is performed on the quasi-optical mode converter by generating high-order electromagnetic waves through the quasi-optical mode exciter: The horizontal component electric field of the output field is detected by a vector network analyzer mounted on a three-dimensional mobile platform at 60mm, 100mm, 140mm, 160mm, and 180mm from the axis of the quasi-optical mode converter; The output field normalized temperature distribution equipotential map is collected by thermal imager; Performing beam energy stability verification based on the horizontal component electric field of the output field to obtain a first verification result; Performing a Gaussian distribution check based on the output field normalized temperature distribution equipotential map to obtain a second check result; When both the first verification result and the second verification result are passed, the entity test result is passed; Otherwise, the entity test result is failed.

3. The method according to claim 2, wherein Performing a beam energy stability check based on the horizontal component electric field of the output field to obtain a first check result includes: Obtaining an output field horizontal component electric field verification model, processing the output field horizontal component electric field, and obtaining the first verification result; The steps of constructing the electric field verification model of the horizontal component of the output field include: Obtain the horizontal component of the output field recorded electric field detected at 60 mm, 100 mm, 140 mm, 160 mm, and 180 mm from the axis of the quasi-optical mode converter; When the variance of the field spot areas at multiple cross sections of the electric field recorded by the horizontal component of the output field is less than or equal to the variance threshold, the alignment check result is marked as passed; otherwise, the alignment check result is marked as failed; When the phase fluctuation trends of the output window ranges at multiple cross sections of the output field horizontal component recording the electric field are the same, the phase consistency check result is marked as passed; otherwise, the phase consistency check result is marked as failed; If both the collimation check result and the phase consistency check result are passed, the check result is marked as passed; otherwise, the check result is marked as failed; The verification result identification information is used as supervision, and the electric field recorded by the horizontal component of the output field is used as input to train a convolution classifier to obtain the output field horizontal component electric field verification model.

4. The method according to claim 2, wherein Performing a Gaussian distribution check based on the output field normalized temperature distribution equipotential map to obtain a second check result includes: Obtaining a Gaussian distribution calibration model, processing the output field normalized temperature distribution equipotential map, and obtaining the second calibration result; The Gaussian distribution calibration model construction step includes: Obtain output field normalized temperature distribution equipotential map record data; Constructing a histogram of the output field normalized temperature distribution equipotential map recorded data; When the histogram shows a high center and low sides, and at least 80% of the data are symmetrically distributed, the verification result of the output field normalized temperature distribution equipotential map recorded data is marked as passed; otherwise, the verification result of the output field normalized temperature distribution equipotential map recorded data is marked as failed; The verification result identification is used as supervision, and the output field normalized temperature distribution equipotential map record data is used as input to train the convolution classifier to obtain the Gaussian-like distribution verification model.

5. The method according to claim 1, wherein When the virtual test result is a failure, optimizing the structural parameters to obtain recommended structural parameters includes: Obtain structural parameter constraint intervals; constructing a high-dimensional distribution space according to the structural parameter constraint interval, wherein any point in the high-dimensional distribution space represents a set of structural parameters; uniformly distributing the particles in the high-dimensional distribution space according to a preset distance threshold to obtain an initial structure particle set; The fusion success probability is used as the fitness function, and optimization is performed based on the initial structure particle set to obtain the recommended structure parameters.

6. The method according to claim 5, wherein Taking the fusion success probability as the fitness function, optimizing based on the initial structure particle set to obtain the recommended structure parameters includes: Obtaining an initial structure particle fusion passing probability set of the initial structure particle set; When the number of initial structure particles whose fusion pass probability set satisfies the pass probability threshold is 0, i target distribution positions of initial structure particles with larger fusion pass probability and j starting distribution positions of initial structure particles with smaller fusion pass probability are obtained, where i and j are integers and i is less than j; Extracting a first starting distribution position according to the j starting distribution positions of the initial structure particles; Extracting a first target distribution position according to the i initial structure particle target distribution positions; Connecting the first starting distribution position and the first target distribution position in the high-dimensional distribution space to obtain a line-adjacent particle fusion passing probability of a line-adjacent initial structure particle whose distribution distance to the connecting line is less than or equal to a distribution distance threshold; When the line adjacent particle fusion passing probability is less than the initial distribution position fusion passing probability of the first initial distribution position, updating the first target distribution position based on the i initial structure particle target distribution positions; Otherwise, searching is performed with the first target distribution position as the target and the first initial distribution position as the starting point to obtain newly added structure particles; An iterative optimization is performed based on the newly added structural particles to obtain the recommended structural parameters.

7. The method according to claim 1, wherein Also includes: When the physical test result is failure, the quasi-optical mode converter is marked as having a performance abnormality.

Images