Selective assembling method of radio frequency coaxial connector based on phase and impedance matching

Abstract

The invention relates to a radio frequency coaxial connector selective assembly method based on phase and impedance matching, and belongs to the technical field of connector assembly, and the method comprises the steps: detecting batch parts corresponding to each part type based on a preset measuring instrument for all part types participating in connector assembly, obtaining key measurement parameters of each part in the batch of parts; inputting the key measurement parameters into a preset electrical performance influence evaluation model, so that the electrical performance influence evaluation model outputs the electrical performance influence quantity of each part; and according to the electrical performance influence quantity correspondingly obtained by each part and a preset matching strategy, selecting a single part from the batch parts of each part type, combining the single part with single parts selected from other part types to obtain an optimal combination, and outputting the optimal combination. According to the invention, the problem of inconsistent phase and impedance between channels caused by manufacturing tolerance can be fundamentally solved.

Description

Technical Field

[0001] This application relates to the field of connector assembly technology, and in particular to a selective assembly method for radio frequency coaxial connectors based on phase and impedance matching. Background Technology

[0002] The inner conductor, outer conductor, and insulating medium of connectors are mass-produced. Although the dimensional tolerances of individual parts are on the order of millimeters, when these parts are randomly assembled, the electrical length (phase) and characteristic impedance of each finished connector will still fluctuate within a small range due to the cumulative effect of tolerances.

[0003] However, in phased array systems, tens of thousands of channels require tens of thousands of connectors. Even minute differences in their electrical lengths can directly lead to beam pointing deviations and gain reductions. Furthermore, accumulated tolerances can cause the characteristic impedance at connection points to deviate from its nominal value (e.g., 50Ω), resulting in signal reflections and increasing insertion loss and voltage standing wave ratio (VSWR). Currently, however, these errors can only be compensated for after the phased array system is integrated through expensive field calibration or complex delay circuits, which is costly and increases system complexity, thus requiring further improvement. Summary of the Invention

[0004] To fundamentally solve the problem of phase and impedance inconsistency between channels caused by manufacturing tolerances, this application provides a selective assembly method for RF coaxial connectors based on phase and impedance matching.

[0005] In a first aspect, this application provides a selective assembly method for radio frequency coaxial connectors based on phase and impedance matching, comprising: For all part types involved in connector assembly, based on preset measuring instruments, the batch of parts corresponding to each part type are inspected to obtain the key measurement parameters of each part in the batch. Key measurement parameters are input into a preset electrical performance impact assessment model, so that the electrical performance impact assessment model outputs the electrical performance impact of each component; wherein, the electrical performance impact refers to the impact of each component on the key electrical performance of the connector when participating in connector assembly; the electrical performance impact includes at least phase impact and impedance impact. Based on the electrical performance impact of each component and the preset matching strategy, a single component is selected from the batch of components of each component type to be combined with a single component selected from other component types to obtain an optimal combination, and the following conditions are met: the key electrical performance of the finished connector assembled from all single components of different component types included in the optimal combination meets the preset application requirements. The preferred combination is output for the processing personnel to know so that the parts in the preferred combination can be assembled to obtain a connector.

[0006] Optionally, the step of selecting a single part from a batch of parts of each part type to combine with single parts selected from other part types to obtain a preferred combination, based on the electrical performance impact value corresponding to each part and a preset matching strategy, includes: Based on a predefined assembly sequence and preset assembly modules, each part type included in the assembly sequence is sequentially taken as a target part type, and the optimal part is matched for the target part type; wherein, the assembly sequence specifies the order in which parts of all part types are assembled into finished connectors; When the optimal part is matched for the target part type, the optimal part corresponding to the target part type is virtually combined with the current assembly module to form a new assembly module, thereby updating the assembly module and synchronously virtually combining the electrical performance impact of the new assembly module; wherein, before matching the optimal part for the first part type in the assembly sequence, the preset electrical performance impact of the assembly module is zero; The process of matching the target part type to obtain the optimal part includes the following steps: All candidate parts in the batch of parts of the target sequence part type are matched one by one with the current assembly module, and the candidate parts that meet the preset matching strategy are selected as the optimal parts; wherein, the preset matching strategy aims to minimize the impact of the electrical performance of the new assembly module formed by the virtual combination of the current assembly module and the corresponding candidate parts.

[0007] Optionally, the preset matching strategy is configured such that when selecting the optimal part for a target part type, the electrical performance impact of the new assembly module must meet the preset application requirement intensity; wherein, the application requirement intensity is dynamically determined based on the importance weight factor predefined for the target part type; the importance weight factor is used to quantify the degree of impact of the target part type on the key electrical performance of the finished connector.

[0008] Optionally, based on the electrical performance impact of each component and a preset matching strategy, a single component is selected from the batch of components of each component type to be combined with single components selected from other component types to obtain a preferred combination, and the following conditions are met: the key electrical performance of the finished connector assembled from all single components belonging to different component types included in the preferred combination meets preset application requirements, including: Obtain the electrical performance impact of batch parts of all part types; Through a preset global optimization process, with preset application requirements as the optimization target, based on the electrical performance impact, several component combinations are generated, and K preferred combinations are selected from all the component combinations to form a batch matching scheme. Wherein, the part combination refers to the set formed by selecting individual parts from the batch parts of all part types, so that a finished connector can be assembled using all the parts contained in the single part combination; K is a positive integer greater than or equal to 1, and K is the number of finished connectors to be produced; the preset application requirement is to maximize the electrical performance consistency of the K preferred combinations and maximize the proportion of the parts contained in the preferred combinations to the total number of all batch parts.

[0009] Optionally, the step of generating several component combinations based on the electrical performance impact quantity through a preset global optimization process, with preset application requirements as the optimization target, and selecting K preferred combinations from all component combinations to form a batch matching scheme, includes: Using a population-based metaheuristic optimization algorithm, an initial population of multiple candidate batch matching schemes is generated. The fitness value of each candidate batch matching scheme is calculated. Based on the fitness value, selection, crossover, and mutation operations are performed on the candidate batch matching schemes in the population to generate new candidate batch matching schemes and update the population. Whenever the population is updated, the fitness value of each candidate batch matching scheme in the updated population is recalculated, and selection, crossover and mutation operations are performed on the candidate batch matching schemes in the population based on the fitness value; until the updated population meets the preset iteration termination condition, the population update is completed. From the population after the population update, the candidate batch matching scheme with the best fitness is selected as the final batch matching scheme.

[0010] Optionally, the global optimization process may include an experience learning and feedback step: Using batch part data of all part types within a historical period and corresponding determined batch matching schemes, a preset strategy recommendation model is trained and optimized; the strategy recommendation model is configured to output at least one recommended matching strategy to guide part combinations to obtain part combinations based on batch part data of all part types. The step of selecting a single part from a batch of parts of each part type based on the electrical performance impact of each part and a preset matching strategy, and combining it with single parts selected from other part types to obtain an optimal combination, further includes: the key electrical performance of the finished connector assembled from all single parts of different part types included in the optimal combination meets the preset application requirements. The strategy recommendation model analyzes batches of parts of all current part types and outputs recommended matching strategies. When generating part combinations through a preset global optimization process, the recommended matching strategy is used to guide the generation process of part combinations, so that among all the generated part types, there are part combinations that conform to the recommended matching strategy.

[0011] Optionally, the recommended matching strategy includes at least: specifying that when generating a part combination, a first part having a preset first electrical performance influence value or range is preferentially added to the same part combination as a second part having a preset second electrical performance influence value or range.

[0012] Secondly, this application provides a selective assembly system for radio frequency coaxial connectors based on phase and impedance matching, comprising, The component data measurement module is used to detect the batch of components corresponding to each component type based on a preset measuring instrument for all component types involved in connector assembly, so as to obtain the key measurement parameters of each component in the batch. The electrical performance evaluation module is used to input key measurement parameters into a preset electrical performance impact evaluation model, so that the electrical performance impact evaluation model outputs the electrical performance impact of each component; wherein, the electrical performance impact refers to the impact of each component on the key electrical performance of the connector when participating in the connector assembly; the electrical performance impact includes at least phase impact and impedance impact. The component pairing and combination module is used to select a single component from the batch of components of each component type and combine it with single components selected from other component types to obtain an optimal combination based on the electrical performance impact of each component and a preset matching strategy. The optimal combination satisfies the following: the key electrical performance of the finished connector formed by assembling all single components of different component types in the optimal combination meets the preset application requirements. A connector assembly module is used to output the preferred combination for processing personnel to know so as to assemble the parts in the preferred combination to obtain a connector.

[0013] Thirdly, this application provides a selective assembly apparatus for a radio frequency coaxial connector based on phase and impedance matching, including a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as described in any of the first aspects.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described in any of the first aspects.

[0015] In summary, this application includes the following beneficial technical effects: This application utilizes predicted quantities and pre-calculated electrical effects of components, followed by selective matching assembly, to compensate for tolerances and improve the electrical consistency of the finished product. It transforms the uncontrollable accumulation of tolerances in traditional manufacturing into predictable, manageable, and usable variables, achieving "precise prediction and control" of connector performance. This fundamentally improves the phase consistency and impedance matching of multi-channel systems, reduces system calibration costs and complexity, and enhances overall performance (such as beamforming quality). This method achieves high performance while avoiding the high processing costs and high scrap rates associated with pursuing extreme tolerances. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart of the selective assembly method for RF coaxial connectors based on phase and impedance matching disclosed in the embodiments of this application.

[0018] Figure 2 This is a structural block diagram of the selective assembly system for radio frequency coaxial connectors based on phase and impedance matching disclosed in the embodiments of this application. Detailed Implementation

[0019] The following is in conjunction with the appendix Figure 1-2 This application will be described in further detail.

[0020] This application discloses a selective assembly method for radio frequency coaxial connectors based on phase and impedance matching (hereinafter referred to as the selective assembly method). The subject of the selective assembly method is a selective assembly system for radio frequency coaxial connectors based on phase and impedance matching (hereinafter referred to as the selective assembly system). The following will be combined with... Figure 1 This section elaborates on the execution process of the selective assembly system.

[0021] S101, for all part types involved in connector assembly, based on a preset measuring instrument, performs testing on the batch of parts corresponding to each part type to obtain the key measurement parameters of each part in the batch.

[0022] S102, input the key measurement parameters into the preset electrical performance impact assessment model so that the electrical performance impact assessment model outputs the electrical performance impact of each component; wherein, the electrical performance impact refers to the impact of each component on the key electrical performance of the connector when participating in the connector assembly; the electrical performance impact includes at least the phase impact and the impedance impact.

[0023] S103, based on the electrical performance impact of each component and the preset matching strategy, select a single component from the batch of components of each component type to combine with single components selected from other component types to obtain an optimal combination, and satisfy the following: the key electrical performance of the finished connector assembled from all single components of different component types included in the optimal combination meets the preset application requirements.

[0024] S104, output the preferred combination for the processing personnel to know so that the parts in the preferred combination can be assembled to obtain the connector.

[0025] In implementation, for S101: all component types involved in connector assembly typically include inner conductors (such as pins), outer conductors (such as sockets), and insulating supports. Pre-set measuring instruments can be non-contact high-precision optical measuring instruments (such as laser scanners or image measuring instruments) or pneumatic measuring instruments. The pre-set measuring instruments measure batches of components (such as batch inner conductors and batch outer conductors) produced in advance for each component type. The key measurement parameters differ for different component types. Specifically, for the inner conductor: its outer diameter d_inner and effective length L_inner are measured; for the outer conductor: the inner diameter d_outer of its contact area with the inner conductor is measured; for the insulating support: its relative permittivity ε_r and length L_ins are measured. Each key measurement parameter is bound to a unique ID (such as a QR code) of the corresponding component and entered into a central database to achieve traceability of the component's entire lifecycle data.

[0026] For S102: The specific implementation steps for calculating the electrical performance impact of each component using the electrical performance impact assessment model are as follows: S1021. The specific scheme for calculating the phase influence ΔΦ using the electrical performance impact assessment model is as follows: 1. A predefined unit phase influence coefficient is provided. This coefficient can be pre-determined manually through limited simulations (such as HFSS, CST) or laboratory measurements and stored in the selective assembly system. Specifically, it determines the change in connector electrical length (unit: degrees / GHz / μm) caused by a unit change in each critical dimension (e.g., 1 μm). For example: For the inner conductor: the unit phase influence coefficient of the outer diameter d_inner is denoted as K_Φ_di (unit: ° / (GHz*μm)), and the unit phase influence coefficient of the effective length L_inner is denoted as K_Φ_Li (unit: ° / (GHz*μm)).

[0027] For the outer conductor: the unit phase influence coefficient of the inner diameter d_outer is denoted as K_Φ_do (unit: ° / (GHz*μm)).

[0028] For the insulating support: the unit phase influence coefficient of the dielectric constant ε_r of the insulating support is denoted as K_Φ_ε (unit: ° / (GHz)); the unit phase influence coefficient of the length L_ins is denoted as K_Φ_Ls (unit: ° / (GHz*μm)).

[0029] Specifically, the unit phase influence coefficient K_Φ_di (unit: ° / (GHz·μm)) and the unit impedance influence coefficient K_Z_di (unit: Ω / μm) described below are proportional coefficients characterizing the degree of influence of critical dimension deviations of components on electrical performance, that is, the amount of phase or impedance change caused by each unit change in critical dimension (e.g., 1 μm). An example of their specific determination method is as follows: (1) Simulation-based determination method: In electromagnetic simulation software (such as HFSS), establish a parametric model of the connector containing the target part (such as the inner conductor). Set the target size (such as the outer diameter of the inner conductor d_inner) as a variable, and select a small variation Δd (e.g., Δd = 1 micrometer) near the nominal value d_nom. Simulate and calculate the S-parameters of the connector at a specified operating frequency f (such as the center frequency) in the following two states: State A: d_inner = d_nom; State B: d_inner = d_nom + Δd. Extract the phase value and characteristic impedance value from the S-parameters. Denote them as Φ_A, Z_A and Φ_B, Z_B, respectively. Calculate the unit influence coefficient: K_Φ_di = (Φ_B - Φ_A) / (Δd * f) [unit: ° / (GHz·μm)]; K_Z_di = (Z_B - Z_A) / Δd [unit: Ω / μm]. To ensure accuracy, Δd should be much smaller than the tolerance range of this size. Simulations can be performed by taking +Δd and -Δd on both sides of d_nom, and the average of the two calculation results can be taken as the final coefficient.

[0030] (2) Experiment-based determination method: A set (e.g., more than 5) of part samples with a stepped distribution of target dimensions (e.g., d_inner) within the tolerance range are prepared by machining. The actual dimension value d_k of each sample is accurately measured using a high-precision measuring instrument (e.g., micrometer). Each sample is assembled into a standard connector with the corresponding nominal part. The S-parameters of each connector at a specified operating frequency f are measured using a vector network analyzer (VNA), and its phase value Φ_k and characteristic impedance value Z_k are extracted. Linear least squares fitting is performed on the measured dataset {d_k, Φ_k} to obtain the fitted line Φ = a_Φ * d + b_Φ, and the slope a_Φ is the approximate value of K_Φ_di. Similarly, fitting the dataset {d_k, Z_k} yields the line Z = a_Z * d + b_Z, and the slope a_Z is the approximate value of K_Z_di. The above method is a conventional technique for device tolerance sensitivity analysis and parameter extraction in the field of radio frequency microwave engineering.

[0031] 2. Based on the predefined unit phase influence coefficient, the phase influence ΔΦ of each part in the batch of parts corresponding to each part type is calculated. ΔΦ is specifically the linear superposition of the deviations of all key measurement parameter values ​​for the part type.

[0032] For example, the phase influence of the inner conductor is ΔΦ = (K_Φ_di * Δd_inner) + (K_Φ_Li * ΔL_inner); where Δd_inner = d_inner_actual measured value - d_inner_preset nominal value; ΔL_inner = L_inner_actual measured value - L_inner_preset nominal value.

[0033] The phase influence of the outer conductor is ΔΦ = K_Φ_do * Δd_outer; where Δd_outer = d_outer_actual measured value - d_outer_preset nominal value.

[0034] The phase influence of the insulating support is ΔΦ = (K_Φ_ε * Δε_r) + (K_Φ_Ls * ΔL_ins); Δε_r = ε_r_actual measured value - ε_r_preset nominal value; ΔL_ins = L_ins_actual measured value - L_ins_preset nominal value.

[0035] Furthermore, a reference frequency f (such as the center frequency) must be specified during the calculation, thus the formula implicitly includes a frequency factor. The final result is the phase deviation at that frequency. For example: Assume an insulating support with ΔL_ins = +5μm and Δε_r = -0.01. Its unit phase influence coefficient (obtained through previous simulations) is known as: K_Φ_Ls = 0.02 ° / (GHz*μm), K_Φ_ε = 15 ° / GHz. At a frequency of 10GHz (i.e., the reference frequency f), its phase influence is: ΔΦ = (0.02 * 5) + (15 * (-0.01)) = 0.1 - 0.15 = -0.05°; this means that at 10GHz, using this insulating support will cause the connector's phase to lead by 0.05 degrees compared to the nominal value.

[0036] S1022. The specific scheme for calculating the impedance influence ΔZ using the electrical performance impact assessment model is as follows: 1. A predefined unit impedance influence coefficient is defined. This coefficient is determined manually beforehand through simulation or experimentation, representing the change in connector characteristic impedance caused by a unit change (e.g., 1 μm) in each critical dimension (as disclosed above, unit: Ω / μm). For example: The unit impedance influence coefficient of the outer diameter d_inner of the inner conductor is denoted as K_Z_di (unit: Ω / μm); The unit impedance influence coefficient of the inner diameter d_outer of the outer conductor is denoted as K_Z_do (unit: Ω / μm); The effect coefficient of the dielectric constant ε_r of the insulating support is denoted as K_Z_ε (unit: Ω).

[0037] 2. Based on the predefined unit impedance influence coefficient, the impedance influence ΔZ of each part in the batch corresponding to each part type is calculated, where the impedance influence ΔZ of the inner conductor is K_Z_di*△d_inner; the impedance influence of the outer conductor is K_Z_do*△d_outer; and the impedance influence of the insulating support is K_Z_ε*△ε_r.

[0038] For example, suppose an outer conductor has an outer diameter of Δd_outer = -2μm. The known unit impedance effect factor is K_Z_do = -0.1 Ω / μm (the negative sign indicates that a smaller outer diameter increases the impedance). Its impedance effect is: ΔZ = (-0.1)*(-2) = +0.2Ω; this means that using this outer conductor will cause the connector's characteristic impedance to be 0.2Ω higher than its nominal value.

[0039] S1023. Based on the calculated phase influence ΔΦ and impedance influence ΔZ of each component, classify them according to preset thresholds: Phase levels are divided into: P- (negative deviation, such as ΔΦ < -δ), P0 (zero deviation, -δ ≤ ΔΦ ≤ +δ), and P+ (positive deviation, ΔΦ > +δ).

[0040] Impedance levels are divided into: Z- (low, such as ΔZ < -ζ), Z0 (standard, -ζ ≤ ΔZ ≤ +ζ), and Z+ (high, ΔZ > +ζ).

[0041] Here, δ and ζ are fixed threshold values ​​set according to the customer's application requirements.

[0042] For S103: S103 utilizes the aforementioned hierarchical results to execute intelligent matching logic, selecting individual parts from the batch of parts of each part type to combine with individual parts selected from other part types, ensuring that all parts in the combination can be assembled to form a connector that meets the preset application requirements. Specifically: The preset matching strategies include: Strategy A (complementary matching method for systems with extremely low phase difference): The selective assembly system will actively match a P- level outer conductor and / or insulating support for a P+ level inner conductor. This makes the phase influence of each component cancel each other out, i.e., ΣΔΦ≈0, specifically: ΣΔΦ is in the range of [0-ΔΦ', 0+ΔΦ'], where ΔΦ' is a preset fixed value, thus making the total electrical length of the finished connector close to the nominal value. The preset application requirement that can be achieved is: to provide connector assemblies with extremely high phase consistency for phased array systems.

[0043] Strategy B (“Same-Group Matching” Method for Extremely Low Impedance Mismatch Systems): The selective assembly system selects individual parts of grade Z0 (standard) from the batch parts corresponding to all part types for assembly. This avoids the introduction of any parts with positive or negative impedance deviations, ensuring impedance continuity at the assembly node. By default, this makes the characteristic impedance Z0 of the finished connector as close to 50Ω as possible. The preset application requirement achieved is providing connectors for test and measurement equipment or high-frequency communication systems with extremely high voltage standing wave ratio (VSWR) requirements.

[0044] For step S104, this step is used for precise assembly and output. Specifically, after determining the preferred combinations, it outputs a list containing one or more preferred combinations, clearly identifying the unique ID of each individual part included in the preferred combination. This list can be displayed on the operator's workstation to guide manual assembly; or it can be sent directly to a pre-set automated assembly system, allowing the system's pre-set robot or automatic gripping mechanism to identify the part ID from the sorted part type bins and pick up the part with the corresponding ID in the preferred combination for manual or automated assembly. After assembly, a QR code can be laser-engraved on the finished connector. This QR code contains its phase level, impedance level, and the IDs of all its constituent parts, facilitating full traceability and quality management.

[0045] Optionally, S103 specifically includes the following steps: S1031, according to the predefined assembly sequence and preset assembly modules, each part type included in the assembly sequence is sequentially taken as the target part type, and the optimal part is obtained by matching the target part type; wherein, the assembly sequence specifies the order in which parts of all part types are assembled into finished connectors.

[0046] S1032, when the optimal part is matched for the target part type, the optimal part corresponding to the target part type is virtually combined with the current assembly module to form a new assembly module, thereby updating the assembly module and synchronously virtually combining the electrical performance impact of the new assembly module; wherein, before matching the optimal part for the first part type in the assembly sequence, the preset electrical performance impact of the assembly module is zero.

[0047] The "matching the target part type to obtain the optimal part" in S1031 specifically includes the following steps: All candidate parts in the batch of parts of the target sequence part type are matched one by one with the current assembly module, and the candidate parts that meet the preset matching strategy are selected as the optimal parts; wherein, the preset matching strategy aims to minimize the impact of the electrical performance of the new assembly module formed by the virtual combination of the current assembly module and the corresponding candidate parts.

[0048] The preset matching strategy is configured such that when selecting the optimal part for the target part type, the electrical performance impact of the new assembly module must meet the preset application requirement intensity; whereby the application requirement intensity is dynamically determined based on the importance weight factor predefined for the target part type; the importance weight factor is used to quantify the degree of impact of the target part type on the key electrical performance of the finished connector.

[0049] In implementation, this application predefines an importance weighting factor of [0, 1] for each type of part to quantify the importance of each type of part in the finished connector. A higher importance weighting factor value indicates greater importance. The corresponding importance weighting factor is determined as follows: 1. Using Monte Carlo simulation, all part parameters are randomly varied tens of thousands of times within their tolerance range to obtain the phase deviation distribution (standard deviation denoted as σ_Φ_total) and impedance deviation distribution (standard deviation σ_Z_total) of the finished connector.

[0050] 2. Perform Monte Carlo simulation again, but this time only allow one component parameter (such as the inner conductor length) to vary within its tolerance range, while other parameters remain fixed at their nominal values. Obtain the standard deviation of the phase deviation, σ_Φ_inner_length. Then, the importance weighting factor of this component parameter to the phase is: I_Φ_inner_length = σ_Φ_inner_length / σ_Φ_total. Similarly, the importance weighting factor of this component parameter to the impedance can be obtained based on the standard deviation of the impedance deviation.

[0051] 3. Similarly, the weighting factors of all component parameters on all electrical performance can be calculated. Ultimately, the importance weighting factor for a component type can be the maximum or average of the weighting factors of its main parameters (such as the key measurement parameters mentioned above). Finally, each component type will obtain a set of phase importance weighting factors I_Φ and impedance importance weighting factors I_Z, pre-determined through simulation. For example, for the insulating support, I_Φ_ins = 0.6, I_Z_ins = 0.7; for the inner conductor, I_Φ_inner = 0.9, I_Z_inner = 0.2; for the outer conductor, I_Φ_outer = 0.3, I_Z_outer = 0.8.

[0052] Specifically, the process of determining the importance weighting factor through Monte Carlo simulation can be achieved through the following specific steps: S1. Parameter Distribution Setting: For each critical dimension of a part (such as the inner conductor length L_inner and the outer conductor diameter d_outer), determine a probability distribution for random sampling based on its manufacturing tolerances. For example, when only the upper and lower tolerance limits are known, a uniform distribution within that range can be used; when the statistical characteristics of the dimensions are known, a normal distribution can be used.

[0053] S2. Full-parameter stochastic simulation (obtaining total deviation): Set the number of samples N1, for example, N1=10000; Loop i = 1 to N1: Based on the probability distribution of each dimension set in S1, independently and randomly generate a set of dimension values ​​for all parts; input the generated set of dimension values ​​into the electrical performance impact assessment model to calculate a corresponding finished connector phase value Φ_total_i and impedance value Z_total_i; After the loop ends, calculate the sample standard deviation σ_Φ_total for all N1 Φ_total_i values, and the sample standard deviation σ_Z_total for all Z_total_i values.

[0054] S3. Single-parameter stochastic simulation (obtaining the single-parameter contribution bias, taking L_inner as an example): Set the number of samples N2, which can be N2=N1, for example N2=10000; Loop j=1 to N2: For the target dimension L_inner only, randomly generate a value according to its probability distribution; fix the dimensions of all other parts to their nominal values ​​(i.e., the unbiased state); input this set of dimension values ​​into the electrical performance impact assessment model to calculate the phase value Φ_param_j and the impedance value Z_param_j; after the loop ends, calculate the sample standard deviation σ_Φ_param(L_inner) of all N2 Φ_param_j values ​​and the sample standard deviation σ_Z_param(L_inner) of all Z_param_j values.

[0055] S4: Calculate the weighting factor for a single dimension: Calculate the importance weight of L_inner to the phase: I_Φ(L_inner)=σ_Φ_param(L_inner) / σ_Φ_total; Calculate the importance weight of L_inner to the impedance: I_Z(L_inner)=σ_Z_param(L_inner) / σ_Z_total.

[0056] S5: Looping and Integration: For each critical dimension that needs to be evaluated, repeat S3 and S4 to obtain the weighting factors of each critical dimension for phase and impedance. For a part type (such as an inner conductor), its overall phase importance weight I_Φ can be taken as the maximum value or arithmetic mean of the I_Φ values ​​of its constituent dimensions; the impedance importance weighting factor I_Z is similar.

[0057] The above steps can be performed by writing MATLAB or Python programs, or by using engineering software with Monte Carlo analysis capabilities such as ANSYS optiSLang.

[0058] In addition, the selective assembly system also predefines an assembly sequence S, which defines the assembly order of all part types in the process of assembling all part types into a finished connector. This order is consistent with the assembly order in the actual production line. The assembly sequence includes all part types arranged in the aforementioned order, such as: insulating support body → inner conductor → outer conductor. The corresponding assembly order is to first assemble the insulating support body and the inner conductor, and then assemble the outer conductor.

[0059] Furthermore, this application proposes a progressive matching scheme: when selecting the preferred combination, the assembly steps of the finished connector are simulated according to the aforementioned assembly sequence S. After assembling a part of one type, the assembled sub-components are regarded as a whole (hereinafter referred to as the assembly module). The electrical characteristics of the current assembly module are matched with the next candidate part according to the matching strategy, thereby more accurately simulating and compensating for the tolerance accumulation effect (such as after assembling the insulating support and the inner conductor, it is treated as a whole, and the electrical characteristics of the whole are matched with the outer conductor according to the matching strategy). This process is carried out step by step until the assembly of all types of parts is completed, and finally the finished connector is obtained.

[0060] The specific progressive matching scheme is an iterative process, with the number of iterations equal to the length of the assembly sequence S: Step 1: During initialization, when selecting a single part from the corresponding batch of parts for the first part type in the assembly sequence, the assembly module is empty. At this time, the electrical performance impact of the assembly module is ΔΦ_module = 0 and ΔZ_module = 0. From the batch of parts of the first part type (such as an insulating support), a part A is selected according to the initial strategy (such as selecting the part with ΔΦ closest to 0). The assembly module is then updated to: part A, and the electrical performance impact of the updated assembly module is (ΔΦ_A, ΔZ_A)).

[0061] Step 2: Prepare to match the second part type (e.g., inner conductor) in the assembly sequence for the current assembly module (part A). At this point, the goal of the matching strategy is to select a part B from the batch of parts of the second part type (e.g., inner conductor) such that the electrical performance impact (ΔΦ_AB, ΔZ_AB) of the new assembly module composed of (assembly module A + part B) can meet the application demand intensity dynamically calculated based on the importance weight factor corresponding to the type of part B (inner conductor). The dynamic calculation method for the application demand intensity is as follows: S2.1: Dynamic Tolerance Allocation: Based on the importance weighting factors (I_Φ_inner, I_Z_inner) of the current part type to be assembled (e.g., the current second part type: inner conductor) and the preset total tolerance of the finished product (δ_Φ_total, δ_Z_total), calculate the residual deviation allowed for the newly assembled module after this step. The residual deviation includes phase tolerance and impedance tolerance. Phase tolerance δ_Φ_step = δ_Φ_total * (1 - I_Φ_inner); Impedance tolerance δ_Z_step = δ_Z_total * (1 - I_Z_inner); S2.2 Dynamic Weight Allocation: Simultaneously, based on the importance weight factor of the current part type to be assembled (e.g., the current second part type: inner conductor), the optimized weights in this matching process are determined. The optimized weights include phase optimization weight W_Φ and impedance optimization weight W_Z. Specifically, the weight factor corresponding to the part type to be assembled during this matching process is used as the corresponding optimized weight, i.e., W_Φ = I_Φ; W_Z = I_Z. For example, I_Φ_inner is used as the specific value of W_Φ for the current second part type (inner conductor), and I_Z_inner is used as the specific value of W_Z. In summary, the dynamic calculation of application requirement intensity is completed.

[0062] Specifically, the design principle and function of the phase tolerance formula δ_Φ_step=δ_Φ_total*(1 - I_Φ) are explained below (the impedance tolerance formula is similar and will not be repeated): This formula serves as a core control rule in the progressive matching method of this invention, and its introduction is based on the following systematic considerations.

[0063] 1. Design Background and Objectives: In the step-by-step assembly process of connectors, each step introduces a certain phase deviation. This invention aims to control the total phase deviation ΣΔΦ of the final product within a preset allowable range [-δ_Φ_total, +δ_Φ_total] through selective matching. Since assembly is performed sequentially, a temporary, phased control target, namely the "step tolerance" δ_Φ_step, needs to be set for the cumulative deviation of the assembled parts at each intermediate step to dynamically guide the entire assembly process towards the final target.

[0064] 2. Physical Meaning of Core Parameters: δ_Φ_total is the upper limit of the total phase tolerance set for the finished connector, representing the final performance target. I_Φ is the importance weighting factor, with a value between 0 and 1. It quantitatively characterizes the potential contribution of the currently assembled part to the total phase deviation of the finished product. The larger the I_Φ value, the more critical the part is, and the higher the proportion of the total deviation that its own manufacturing tolerances may cause. δ_Φ_step is the upper limit of the phase deviation allowed for the assembled "sub-assembly" (i.e., assembly module) before assembling the current part. It is a dynamic value that changes when assembling different parts.

[0065] 3. Logical Reasoning and Engineering Explanation of the Formula: The total allowable tolerance δ_Φ_total is considered as a "deviation budget" that needs to be allocated across the entire assembly chain. Based on the weighting factor I_Φ of the part, the proportion of the deviation budget consumed by assembling the current critical part can be estimated to be approximately I_Φ. Therefore, in order to leave sufficient budget space (δ_Φ_total * I_Φ) for the assembly of the current critical part, the deviations accumulated in all previous steps must be strictly controlled before assembling it. The remaining budget allowed for assembled modules is the total budget minus the portion reserved for the current step: Allowable used budget = Total budget - Budget reserved for the current step: δ_Φ_step = δ_Φ_total - (δ_Φ_total * I_Φ) = δ_Φ_total * (1 - I_Φ).

[0066] This rule ensures that the assembly process is forward-looking: for critical parts that will be assembled later (high I_Φ), the system will apply stricter control (low δ_Φ_step) earlier to prevent the tolerance budget from being exhausted too early in the assembly process, which would result in the inability to assemble a qualified product even if the ideal part is found later.

[0067] 4. Application Example: Suppose the required phase deviation of the finished product is no more than δ_Φ_total = ±1.0°. When assembling a critical inner conductor, its weighting factor is evaluated as I_Φ_inner = 0.7. According to the formula, before assembling this inner conductor, the phase deviation of other assembled parts (such as the insulating support) must be controlled within δ_Φ_step = 1.0°*(1 - 0.7) = ±0.3°. This stringent requirement forces the matching algorithm to select parts that minimize the deviation of the sub-assemblies in the previous step (selecting the insulator), thus allowing a sufficient deviation margin of ±0.7° for the assembly of the critical inner conductor.

[0068] In summary, this formula is not a derivation of existing physical laws, but rather an innovative and heuristic process control rule proposed in this application for effectively managing tolerance accumulation and achieving systematic matching optimization. Its effectiveness has been verified through the implementation of the overall technical solution of this application and is an important component of the intelligent matching logic of this invention.

[0069] After completing the dynamic calculation of the application demand intensity, all parts (hereinafter referred to as candidate parts) in the batch of parts of the second part type (inner conductor) are traversed. For each candidate part B, the electrical performance impact (ΔΦ_AB, ΔZ_AB) of the new assembly module AB after combining it with the current assembly module A is calculated. It should be noted here that the electrical performance impact of the assembly module and the candidate parts (as mentioned above, (ΔΦ_AB, ΔZ_AB)) can be obtained through a pre-stored combination lookup table, or by correcting the simple linear superposition result through a combination correction factor that considers the interaction; for example, ΔΦ_AB = ΔΦ_A + ΔΦ_B + C_Φ_AB, where C_Φ_AB is a specific phase interaction correction amount for the combination of type A parts and type B parts determined manually in advance using finite element simulation, and is a preset value. In addition, this application assumes that when assembling according to the predefined assembly sequence, there is a preset combination correction factor corresponding to each assembly module and the candidate parts, so as to obtain the electrical performance impact of the new assembly module formed by the combination of the assembly module and the candidate parts.

[0070] It should be noted that determining the interaction coefficients of RF connector component combinations using finite element simulation (such as ANSYS HFSS, CST Studio Suite) is a conventional technique in this field. In RF engineering practice, designers commonly use such electromagnetic simulation software to build three-dimensional models and study the impact of component tolerances and combination effects on overall electrical performance (S-parameters, phase, impedance) through parametric scanning or comparative analysis. The finite element simulation specified in this application is a direct application of this well-known technique. For example, those skilled in the art can: establish a model simulating "component A and component B combined according to nominal values" to obtain the phase Φ_AB; then establish models "only component A with tolerances" and "only component B with tolerances" to obtain Φ_A and Φ_B respectively; then the combination correction factor C_Φ_AB can be estimated as Φ_AB-(Φ_A +Φ_B). For example, for a combination of inner conductor (type A) and outer conductor (type B), the phase interaction correction C_Φ_AB can be estimated through the following steps: Establish a model of "part A and part B combined according to nominal values" and simulate to obtain the combined phase Φ_AB; then establish models of "only part A with tolerance" and "only part B with tolerance" respectively, and simulate to obtain Φ_A and Φ_B; then C_Φ_AB can be estimated as Φ_AB - (Φ_A + Φ_B). The combined correction factor C_Φ_AB can be an empirical constant in the range of [-0.1°, +0.1°]. It should be noted that the introduction of this correction factor is to improve the accuracy of virtual assembly prediction, and its omission (i.e., setting C_Φ_AB = 0) does not affect the implementation of the basic method of selective matching assembly in this application.

[0071] Then, check whether the performance of the newly assembled module AB meets the dynamic tolerance target, that is, whether it simultaneously satisfies: |ΔΦ_AB|≤δ_Φ_step and |ΔZ_AB|≤δ_Z_step. Candidate parts that do not meet the requirements are eliminated.

[0072] Next, for all candidate parts that were not eliminated, their comprehensive score is calculated as Score = W_Φ *|ΔΦ_AB|+ W_Z *|ΔΦ_AB|. The candidate part B with the smallest Score value is selected as the best match in this step.

[0073] Finally, the candidate part B with the smallest score is combined with the current assembly module (part A) to obtain a new assembly module AB, and its electrical performance impact is recorded as (ΔΦ_AB, ΔZ_AB).

[0074] Step 3 and subsequent steps: Repeat the logic of step 2 to match the next part type (such as the outer conductor) in the assembly sequence for the current assembly module (AB). At this time, the importance weighting factor of the outer conductor (I_Φ_outer=0.3, I_Z_outer=0.8) will be used to calculate the new application requirement intensity (i.e., residual deviation and optimization weight) and find the unique optimal part from the batch of parts of the outer conductor accordingly; and so on, repeat the logic of step 2 until the optimal part of the last part type in the assembly sequence S is found, and the optimal part of the last part type is combined with the latest assembly module to obtain the final assembly module (i.e., the preferred combination), thus outputting the preferred combination consisting of the optimal parts of each part type selected by all iteration steps, completing the progressive matching process. It should be noted that executing a progressive matching process once yields an optimal combination, which means selecting all the corresponding part types for a connector. Based on this, the batch part quantity for all part types is updated (the updated batch part quantity means removing the parts selected into the optimal combination and adding the latest produced parts). Then, based on the updated batch part quantity for all part types, the progressive matching process is executed again to obtain a second optimal combination. This process is repeated to achieve repeated execution of the progressive matching process, thereby enabling the assembly of multiple connectors based on the batch parts.

[0075] Optionally, S103 specifically includes the following steps: S1131, Obtain the electrical performance impact of batch parts of all part types; S1132, through a preset global optimization process, with preset application requirements as the optimization objective, based on the electrical performance impact, several part combinations are generated, and K preferred combinations are selected from all part combinations to form a batch matching scheme; where, a part combination refers to a set formed by selecting individual parts from the batch parts for all part types, so that a finished connector can be assembled using all the parts contained in a single part combination; K is a positive integer ≥ 1, and K is the number of finished connectors to be produced; the preset application requirements are: to maximize the electrical performance consistency of the K preferred combinations and to maximize the proportion of the parts contained in the preferred combinations to the total number of parts in the batch.

[0076] Before the global optimization process, there are also experience learning and feedback steps: Using batch part data of all part types within a historical period and corresponding determined batch matching schemes, a preset strategy recommendation model is trained and optimized. The strategy recommendation model is configured to output at least one recommended matching strategy to guide the combination of parts to obtain the part combination based on the batch part data of all part types. S1132 specifically includes the following steps: S11321, using a strategy recommendation model, analyzes batch parts of all current part types and outputs recommended matching strategies. When generating part combinations through a preset global optimization process, the recommended matching strategies guide the generation process, ensuring that among all generated part types, there are part combinations that conform to the recommended matching strategies. The recommended matching strategies include at least: specifying that when generating part combinations, a first part with a preset first electrical performance influence value or range, and a second part with a preset second electrical performance influence value or range, should be preferentially added to the same part combination.

[0077] S11322: Using a population-based metaheuristic optimization algorithm, an initial population of multiple candidate batch matching schemes is generated. The fitness value of each candidate batch matching scheme is calculated. Based on the fitness value, selection, crossover, and mutation operations are performed on the candidate batch matching schemes in the population to generate new candidate batch matching schemes and update the population. S11323, Whenever the population is updated, the fitness value of each candidate batch matching scheme in the updated population is recalculated, and the candidate batch matching schemes in the population are selected, crossover and mutated according to the fitness value; until the updated population meets the preset iteration termination condition, the population update is completed. S11324: From the population after the population update, select the candidate batch matching scheme with the best fitness as the final batch matching scheme.

[0078] In implementation, optionally, in conjunction with the progressive matching process described above, when the number of batch parts corresponding to each part type is large, the efficiency of the above-mentioned iterative execution operation of updating the batch part quantity in real time and then performing the progressive matching process based on the updated batch part quantity is relatively low. Furthermore, it is prone to "greedily" selecting local optima, leading to a large number of parts being idle due to mismatch, resulting in low resource utilization. At the same time, fixed matching strategies cannot learn from past successes and are difficult to continuously optimize. To address this, this application further proposes the following batch selective assembly method: The execution sequence of this method includes three main stages: a data preparation stage (S300), a core optimization stage (S400), and an intelligent evolution stage (S500). The core optimization stage generates batch matching schemes, while the intelligent evolution stage acts as a superimposed intelligent layer, continuously optimizing the efficiency of the core stage.

[0079] Specifically: S300 is the data preparation stage, which provides the data foundation for the entire method and sets the operating parameters. In this stage, the selective assembly system obtains the electrical performance impact quantities (ΔΦ, ΔZ) of all batch parts (including inner conductors, outer conductors, insulating supports, etc.) through measurement and calculation, and stores them in a preset database, with each part having a unique ID.

[0080] Simultaneously, the selective assembly system sets a target production quantity K during this stage (the specific value of K can be manually entered into the selective assembly system to represent the number of connectors to be processed and assembled subsequently, such as 1000 connectors). Parameters for the metaheuristic algorithm (taking a genetic algorithm as an example) are also set, including: population size P_size (e.g., 100), maximum number of generations G_max (e.g., 500), crossover rate P_c, and mutation rate P_m. Weights w1, w2, and w3 of the multi-objective fitness function are also set (e.g., w1=0.5, w2=0.4, w3=0.1) to balance phase consistency, impedance consistency, and part utilization rate.

[0081] S400, the core of this stage, specifically uses the iterative logic of the standard genetic algorithm to find a globally optimal batch assembly scheme, which includes the following detailed steps.

[0082] S401, Problem Definition and Solution Coding Steps.

[0083] 1.1 Optimization Objective Definition: From the parts inventory, select a complete set of parts (containing all M required types) for each of the K connectors planned for production, forming K parts combinations to constitute a complete batch matching scheme. The optimization objective of this batch matching scheme is to maximize the phase consistency and impedance consistency of these K finished connectors, while maximizing the proportion (utilization rate) of the total number of parts used to the available parts in the inventory.

[0084] 1.2 Chromosome Structure Design (Scheme Representation): A batch matching scheme is represented using integer encoding. The batch matching scheme is represented as a two-dimensional array with K rows and M columns, called a chromosome. The k-th row corresponds to the k-th connector to be assembled, and the m-th column corresponds to the m-th part type required for that connector. Each element in the array is an integer representing a unique identifier (ID) of the specific part assigned to that position. This ID must be taken from the list of available part IDs for the corresponding part type in the current inventory.

[0085] 1.3 Constraints: In a batch matching scheme represented by a chromosome, any specific part ID can only appear once in the array to ensure that each physical part is assigned to only one connector.

[0086] S402, Initial population generation steps.

[0087] 2.1 Set the population size: Determine the number of individuals (i.e., schemes) in the initial population, P_size, for example, set it to 100.

[0088] 2.2 The process of generating a single chromosome: For each individual in the population, its chromosome is constructed according to the following process: 2.2.1: Initialize an empty set to record the selected part IDs;

[0089] 2.2.2: Traverse each connector position sequentially (from the 1st to the Kth): For each part type required for this connector (from type 1 to type M): randomly select a part ID from the current inventory list of available IDs for that type that is not yet recorded in the "Selected ID Set" mentioned above. Enter this ID into the corresponding position in the chromosome array. Add this ID to the "Selected ID Set";

[0090] This process ensures that each randomly generated initial solution satisfies the constraint that "parts are not reused".

[0091] S403. Fitness value calculation steps (for a given chromosome scheme).

[0092] 3.1 Virtual Assembly and Performance Calculation Sub-steps:

[0093] 3.1.1 Initialize two empty lists, which are used to temporarily store the phase influence and impedance influence of all K finished products, respectively;

[0094] 3.1.2 Process each linker scheme (i.e., each row) in the chromosome sequentially: Based on the M part IDs stored in the row, retrieve the pre-calculated electrical performance impact values ​​for each part from the system database, including the phase impact value ΔΦ and the impedance impact value ΔZ. Following a predefined physical assembly sequence (e.g., assembling the insulator first, then the inner conductor, and finally the outer conductor), simulate and superimpose the electrical impact values ​​of all parts of this connector (using linear superposition or a combination of correction factors) to calculate the total phase impact value ΔΦ_final and the total impedance impact value ΔZ_final of the finished connector. Store the calculated ΔΦ_final and ΔZ_final into the two lists initialized in step (a), respectively.

[0095] 3.2 Performance Consistency Index Calculation Sub-steps: Calculate the sample standard deviation of the list storing K phase influence quantities, which is σ_Φ. Calculate the sample standard deviation of the list storing K impedance influence quantities, which is σ_Z. The smaller the standard deviation, the higher the performance consistency of a batch of connectors produced under this scheme.

[0096] 3.3 Component Utilization Rate Calculation Sub-step: Count the total number of all unique component IDs used in this chromosome scheme, which is U; calculate the utilization rate index: Utilization_Rate = U / (K * M). Where (K * M) is the total number of components theoretically required to assemble K connectors.

[0097] 3.4 Overall Fitness Calculation Sub-step: Pre-set weighting coefficients w1, w2, and w3 to balance the importance of phase consistency, impedance consistency, and utilization (e.g., w1=0.5, w2=0.4, w3=0.1). Calculate the overall fitness value (Fitness) of the chromosome scheme using the following formula: Fitness = w1 * [1 / (σ_Φ+ε)] + w2 * [1 / (σ_Z+ε)] + w3 * Utilization_Rate. Where ε is a very small positive number (e.g., 10 to the power of -6) to prevent calculation errors when the standard deviation is zero. A higher fitness value indicates better overall quality of the batch matching scheme.

[0098] S404, Genetic Evolution Operation Steps.

[0099] 4.1 Selection Operation: Calculate the fitness value of all individuals in the current population; use roulette wheel selection: determine the probability of each individual being selected for the next generation based on the proportion of its fitness value to the total fitness of the population; individuals with higher fitness have a greater probability of being selected. Based on the above probabilities, select P_size individuals from the current population through random sampling (with repeatable selection) to form a mating pool for subsequent crossover operations.

[0100] 4.2 Crossover Operation: Set the crossover probability P_c (e.g., 0.8); randomly pair individuals in the mating pool. For each pair of parent individuals, generate a random number in the interval [0, 1]. If this random number is less than the crossover probability P_c, crossover is performed on this pair of parents; otherwise, the offspring directly copies the parent. The specific crossover method is as follows: randomly select a connector index as the cut point. Exchange all part allocation information in the two parent individuals, starting from the connector corresponding to the cut point and ending at the last connector. This operation exchanges complete connector units, ensuring that the part combination of each connector in the offspring individual remains complete. After the crossover operation, check whether there are duplicate part ID conflicts (i.e., the same ID appears in different positions) in the two newly generated offspring individuals. If a conflict exists, randomly exchange the conflicting part IDs within the connector unit where the conflict occurred until the conflict is resolved.

[0101] 4.3 Mutation Operation: Set the mutation probability P_m, for example, 0.05. For each individual in the new population obtained after selection and crossover, iterate through every gene position (i.e., every part ID) in its chromosome. For each gene position, generate a random number in the range [0,1]. If the random number is less than the mutation probability P_m, then perform mutation on that gene position: randomly select a part ID from the current inventory list of the part type corresponding to that gene position that is different from the current value and has not been used in other positions in the chromosome, and replace the current ID.

[0102] S405. Overall Algorithm Flow and Termination Steps.

[0103] 5.1 Initialization: Generate an initial population containing P_size random feasible solutions.

[0104] 5.2 Set termination conditions: For example, set the maximum number of generations G_max (e.g., 500 generations), or set the population to stop evolving if the optimal fitness value no longer increases for several consecutive generations (e.g., 20 generations).

[0105] 5.3 Iterative Evolution: Repeat the following steps until the termination condition is met: 5.3.1 Population Assessment: Calculate the fitness value of all individuals in the current population; 5.3.2 Record Elites: Retain the individual with the highest fitness in the current population; 5.3.3 Genetic operations: Perform selection → crossover → mutation operations on the current population to generate the next generation population.

[0106] 5.3.4 Elite Preservation: The elite individuals recorded in step 5.3.2 are directly replaced with the individuals with the lowest fitness in the next generation of the population to ensure that the optimal solution is not lost.

[0107] 5.4 Output Results: The batch matching scheme represented by the individual with the highest fitness found during the evolution process is output as the final optimization result.

[0108] The above steps constitute a complete genetic algorithm optimization process. Those skilled in the art can implement it using common programming languages ​​(such as Python and MATLAB) or optimization libraries based on this process description to solve the part selective matching problem described in this application.

[0109] Furthermore, S500 and S400 are performed asynchronously, constituting a continuous learning process. S500 specifically includes the following timing steps: S501: After each successful execution of S400, a task snapshot is generated and stored in a preset matching experience database. The task snapshot includes: Input features: The distribution feature vector of all part types of batch parts (hereinafter referred to as part inventory) on which this S400 is based. For example, ([Count_P+, Count_P0, Count_P-, Count_Z+, Count_Z0, Count_Z-...]), that is, the number of parts for each phase level and each impedance level.

[0110] Output Strategy: High-frequency matching patterns extracted from the final batch assembly scheme of this S400 execution. For example, association rule analysis revealed a significant high frequency of the combination pattern of "P+ grade inner conductor and P- grade outer conductor" under this inventory distribution. The specific extraction scheme is as follows: Extraction Plan 1: The final batch assembly scheme output by this S400 execution is considered a "transaction database". In this batch assembly scheme, each successfully assembled finished connector (i.e., the finished connector virtually assembled using each combination) is considered a "transaction", and the grade identifiers (e.g., [P+, Z0]) of all parts contained in this finished connector are considered "itemsets" within that transaction. For example, assuming this batch assembly scheme includes 4 combinations (i.e., K=4), corresponding to the production of 4 finished connectors, the following data can be obtained: Transaction 1 (Connector 1): [P+_Inner Conductor, P-_Outer Conductor, Z0_Insulator]; Transaction 2 (Connector 2): [P+_Inner Conductor, P-_Outer Conductor, Z+_Insulator]; Transaction 3 (Connector 3): [P0_Inner Conductor, P0_Outer Conductor, Z0_Insulator]; Transaction 4 (Connector 4): [P+_Inner Conductor, P-_Outer Conductor, Z0_Insulator].

[0111] Next, define the support (support of a part combination (itemset) = number of connectors containing that combination / total number of connectors K); set a minimum support threshold: for example, set it to 50% (i.e., 0.5). The selective matching system counts all occurrences of part combinations and filters out combinations with support ≥ 0.5 as high-frequency matching patterns.

[0112] For example, from the sample data, we can see that the support of the itemset {P+_inner conductor, P-_outer conductor} is 3 / 4 = 0.75 (because transactions 1, 2, and 4 all contain it); the support of the itemset {Z0_insulator} is 2 / 4 = 0.5 (because transactions 1 and 4 contain it); the support of other combinations is all below 0.5. Therefore, the high-frequency matching pattern extracted in this study is: {P+_inner conductor, P-_outer conductor}.

[0113] The input features (parts inventory distribution) and output strategies (i.e., the high-frequency matching patterns discovered, such as {P+_inner conductor, P-_outer conductor}) of this task are bound together and stored in the matching experience database.

[0114] Extraction Plan Two: During this S400 execution, the batch assembly schemes with fitness values ​​higher than a preset threshold within the last output scheme set are used as learning cases. These learning cases are then clustered based on their input context (i.e., parts inventory distribution). For example, a group of cases with "Z0 outer conductor scarcity" might be automatically identified (i.e., Z0 outer conductor inventory < 20% of total demand).

[0115] In the group where Z0 outer conductors are scarce, the optimal solution (i.e., the batch assembly solution with the highest fitness value) among all the learning cases is analyzed. The pairing of outer conductors is statistically analyzed. If a pairing pattern with a frequency higher than a preset frequency exists (e.g., the complementary pairing pattern {Z+ outer conductor, Z- outer conductor}), a conditional rule is summarized: IF (condition): The quantity of Z0 outer conductors in inventory < 20% of the total demand; THEN (action): When generating candidate solutions, actively create pairing combinations of {Z+ outer conductor, Z- insulator} and {Z- outer conductor, Z+ insulator}, forming an output strategy: When the inventory ratio of {part type T: grade L} is lower than the threshold X, the complementary pairing {part type T: grade H} and {part type T: grade L} is enabled. This is stored in the matching experience database.

[0116] S502: Once the task snapshots in the matching experience database have accumulated to a certain scale (e.g., more than 1000 task snapshots), the training process of the machine learning model is initiated, specifically using a multi-label classification model (e.g., using RandomForestClassifier from the Scikit-learn library). The training input features of this model are the parts inventory distribution feature vectors recorded in each record of the database.

[0117] The training target labels for this model are the high-frequency matching patterns or conditional matching rules extracted from each record in the database. The model is trained using the content stored in the matching experience database. Through training, the model learns the complex mapping relationship between parts inventory distribution characteristics and the optimal matching strategy. The final trained model is called a strategy recommendation model.

[0118] S503, when the selective matching system obtains a new batch matching task and needs to execute S400, before executing S400, the selective matching system first analyzes the inventory distribution of the parts inventory corresponding to the current batch matching task, generates the corresponding input features (i.e., distribution feature vector), and then inputs the feature vector into the trained strategy recommendation model; The strategy recommendation model outputs one or more recommended matching strategies for the current inventory distribution. These strategies are consistent in form with the training target labels. For example, they may be simple pairing instructions such as "{P+ inner conductor, P- outer conductor}", or conditional instructions such as "when [a certain condition] is met, adopt [a certain strategy]". For example, when the outer conductor Z0 is insufficient, Z+ and Z- can be used to complement each other.

[0119] In step S504, during the execution of S400, the recommended matching strategy derived in S503 is applied to step S401 within S400. That is, when executing S401, batch matching schemes are no longer generated completely randomly; instead, a portion of the batch matching schemes are generated based on the recommended matching strategy. For example, some combinations are prioritized based on "Strategy 1". In this way, the global optimization search efficiency of S400 is significantly improved, converging to a high-quality solution more quickly. The output of this S400 step is then recorded again (corresponding to S501) for the next iteration of training the strategy recommendation model, forming a continuously improving intelligent closed loop.

[0120] Based on the foregoing description, in order to more clearly disclose the strategy recommendation model, this application further discloses the specific implementation steps for constructing and using the strategy recommendation model: 1. Model Input Feature Construction Steps: Generation of Inventory Distribution Feature Vector: 1.1 Component Classification: Based on the phase influence ΔΦ and impedance influence ΔZ of each component, all components are pre-classified into six electrical performance levels: positive phase deviation (P+), zero phase deviation (P0), negative phase deviation (P-), positive impedance deviation (Z+), zero impedance deviation (Z0), and negative impedance deviation (Z-).

[0121] 1.2 Feature Statistics: Suppose there are T types of parts in the system (e.g., T=3, corresponding to inner conductor, outer conductor, and insulator). For the t-th type of part, count the number of parts belonging to the above six levels in its inventory;

[0122] 1.3 Vectorization: The above statistical results are arranged in a fixed order to form a one-dimensional numerical vector of length 6×T, which serves as the feature vector X describing the current overall inventory status;

[0123] 1.4 Example: For a three-type system, the eigenvector X is composed as follows: X = [number of inner conductor P+, number of inner conductor P0, number of inner conductor P-, number of inner conductor Z+, number of inner conductor Z0, number of inner conductor Z-, number of outer conductor P+, …, number of insulator Z-];

[0124] 1.5 Feature Enhancement: To further enrich the feature information, the proportion of each level of parts to the total number of parts of that type can be calculated, or its ratio to the estimated demand (sufficiency rate) can be calculated, and these calculated derived features can be appended to the end of the original feature vector X.

[0125] 2. Steps for obtaining model output labels: Extract high-frequency matching patterns from historical data: 2.1 Historical Data Preparation: Collect records of H successfully executed batch matching tasks in history. Each task produced a high-performance (i.e., high fitness value) batch matching scheme, and each batch matching scheme produced K connectors;

[0126] 2.2 Constructing the Transaction Database: Each successfully produced historical connector is considered an independent "transaction". All parts used by the connector are represented by their corresponding part type-level combinations (e.g., "inner conductor-P+", "outer conductor-P-", "insulator-Z0"). The set of these combinations constitutes the "itemset" of the transaction. Thus, a total of H×K transactions can be obtained from H historical schemes, forming the transaction database D;

[0127] 2.3 Perform association rule mining: Use the Apriori algorithm to mine high-frequency part-level combination patterns (i.e., frequent itemsets) from the transaction database D. The specific steps are as follows: 2.3.1 Set a minimum support threshold: For example, setting min_support = 0.5 means that a pattern is considered high-frequency only if it appears in at least 50% of historical transactions;

[0128] 2.3.2 Scan the database and generate frequent 1-itemsets: First, find the support of all individual part levels (such as "inner conductor-P+"), and retain items with support not less than min_support to form a frequent 1-itemset set L1;

[0129] 2.3.3 Iterative generation of frequent k-itemsets (k≥2): Based on the frequent (k-1)-itemset set L_{k-1}, a candidate k-itemset set C_k is generated through a join operation. For each candidate set in C_k, the transaction database D is scanned, and its support (i.e., the proportion of transactions containing that candidate set) is calculated. Candidate sets with a support not less than min_support are retained to form the frequent k-itemset set L_k;

[0130] 2.3.4 Repeat step 2.3.3 until no new frequent itemsets can be generated. The union of all L_k constitutes the final high-frequency matching pattern library F;

[0131] 2.4 Pattern Example: The mining results may include binary patterns such as {inner conductor-P+, outer conductor-P-} with a support of 0.7, indicating that the pairing of "positive phase deviation inner conductor and negative phase deviation outer conductor" was observed in 70% of the successful historical connectors.

[0132] 3. Supervised learning model training steps: Establish the mapping relationship from features to patterns, including the following sub-steps: 3.1 Training Dataset Construction: For the H historical tasks collected in step 2.1, construct a training sample for each task i: Input X_i: The inventory distribution feature vector generated in step 1 before the task is executed. Output Y_i: A multi-label vector. The dimension of the vector is equal to the total number of patterns in the high-frequency matching pattern library F. If the j-th pattern in the pattern library F has appeared in the successful solution generated by this task, then the j-th element of Y_i is 1; otherwise, it is 0.

[0133] 3.2 Model Selection and Configuration: Select a machine learning model suitable for the multi-label classification task, such as a random forest classifier. A "multi-output classifier" approach can be used, combining multiple binary random forest models, with each model responsible for predicting the presence or absence of a pattern label.

[0134] 3.3 Perform model training: Train the selected model using the constructed training dataset {(X_1, Y_1), (X_2, Y_2), ..., (X_H, Y_H)}. The training process aims to enable the model to learn the complex mapping relationship between inventory distribution features X and high-frequency pattern occurrences Y;

[0135] 3.4 Regarding data scale and quality: It is recommended that the number of training samples H reach several hundred or more to ensure that the model has good generalization ability. The core of training data quality lies in the fact that the output label Y must come from high-performance historical matching schemes that have been validated by optimized algorithms, thereby ensuring that the model learns effective assembly strategies.

[0136] 4. Model application and strategy recommendation steps: 4.1 Feature Extraction: When facing a new batch of parts inventory that needs to be matched and planned, firstly, according to the method described in step 1, a feature vector X_new is generated based on the current inventory status;

[0137] 4.2 Model Inference: Input X_new into the trained policy recommendation model. The policy recommendation model will output a prediction vector Y_pred, where the value (or probability) of each element represents the likelihood that the corresponding high-frequency matching pattern will be recommended in the current inventory state;

[0138] 4.3 Strategy Generation: Based on the prediction results, select the top R most likely patterns (e.g., R=5) and convert them into recommended matching strategies described in natural language. For example: "Strategy 1: Prioritize the combination of P+ level inner conductors with P- level outer conductors"; "Strategy 2: When Z0 level insulator inventory is insufficient, consider pairing Z+ level with Z- level insulators";

[0139] 4.4 Strategy Application: When using global optimization methods such as genetic algorithms for part matching, the above-mentioned recommended strategies can be used to guide the optimization process. For example, when generating the initial population for the genetic algorithm, some connectors can be assembled consciously according to the recommended strategies; or in crossover and mutation operations, combinations of parts that conform to the recommended strategies can be given a higher retention probability, thereby guiding the optimization search direction and improving matching efficiency and result quality.

[0140] This application also discloses a selective assembly system for radio frequency coaxial connectors based on phase and impedance matching. (Refer to...) Figure 2 ,include: The part data measurement module 601 is used to detect the batch of parts corresponding to each part type based on a preset measuring instrument for all part types involved in connector assembly, so as to obtain the key measurement parameters of each part in the batch. The electrical performance evaluation module 602 is used to input key measurement parameters into a preset electrical performance impact evaluation model so that the electrical performance impact evaluation model outputs the electrical performance impact of each component; wherein, the electrical performance impact refers to the impact of each component on the key electrical performance of the connector when participating in the connector assembly; the electrical performance impact includes at least phase impact and impedance impact. The component pairing and combination module 603 is used to select a single component from the batch of components of each component type and combine it with a single component selected from other component types to obtain an optimal combination based on the electrical performance impact of each component and the preset matching strategy. The optimal combination satisfies that the key electrical performance of the finished connectors assembled from all the single components of different component types in the optimal combination meets the preset application requirements. Connector assembly module 604 is used to output preferred combinations for processing personnel to know so as to assemble the parts in the preferred combinations to obtain a connector.

[0141] Optionally, the parts matching and combination module 603 is further configured to, according to a predefined assembly sequence and a preset assembly module, sequentially use each part type included in the assembly sequence as a target part type, and match the optimal part for the target part type; wherein, the assembly sequence specifies the order in which parts of all part types are assembled into finished connectors; when the optimal part is matched for the target part type, the optimal part corresponding to the target part type is virtually combined with the current assembly module to form a new assembly module, thereby updating the assembly module and synchronously virtually combining the electrical performance impact of the new assembly module; wherein, before matching the optimal part for the first part type in the assembly sequence, the preset electrical performance impact of the assembly module is zero; The parts matching and combination module 603 is also used to match all candidate parts in the batch of parts of the target sequence part type with the current assembly module one by one, and select the candidate parts that meet the preset matching strategy as the optimal parts; wherein, the preset matching strategy aims to minimize the electrical performance impact of the new assembly module formed after the current assembly module and the corresponding candidate parts are virtually combined.

[0142] Optionally, the parts pairing and combination module 603 is also used to obtain the electrical performance impact of batch parts of all part types; through a preset global optimization process, with preset application requirements as the optimization target, based on the electrical performance impact, several parts combinations are generated, and K preferred combinations are selected from all parts combinations to form a batch matching scheme; wherein, a parts combination refers to a set formed by selecting individual parts from batch parts of all part types, so that a finished connector can be assembled using all the parts contained in a single parts combination; K is a positive integer ≥ 1, and K is the number of finished connectors to be produced; the preset application requirements are: to maximize the electrical performance consistency of the K preferred combinations and to maximize the proportion of the parts contained in the preferred combinations to the total number of batch parts.

[0143] Optionally, the parts matching and combination module 603 is further configured to use a population-based metaheuristic optimization algorithm to generate an initial population of multiple candidate batch matching schemes, calculate the fitness value of each candidate batch matching scheme, and perform selection, crossover, and mutation operations on the candidate batch matching schemes in the population based on the fitness value to generate new candidate batch matching schemes and update the population; whenever the population is updated, the fitness value of each candidate batch matching scheme contained in the updated population is recalculated, and selection, crossover, and mutation operations are performed on the candidate batch matching schemes in the population based on the fitness value; until the updated population meets the preset iteration termination condition, the population update is completed; from the population after the population update is completed, the candidate batch matching scheme with the best fitness is selected as the final batch matching scheme.

[0144] Optionally, a learning and feedback module is also included, which is used to train and optimize a preset strategy recommendation model using batch part data of all part types within a historical period and corresponding determined batch matching schemes; the strategy recommendation model is configured to output at least one recommended matching strategy to guide part combination to obtain part combination based on batch part data of all part types.

[0145] The parts matching and combination module 603 is also used to analyze the batch of parts of all current part types through a strategy recommendation model and output a recommended matching strategy; when generating parts combinations through a preset global optimization process, the recommended matching strategy is used to guide the generation process of parts combinations so that among all generated part types, there are parts combinations that conform to the recommended matching strategy.

[0146] This application also discloses a selective assembly device for RF coaxial connectors based on phase and impedance matching. The selective assembly device for RF coaxial connectors based on phase and impedance matching includes a memory and a processor. The memory stores a computer program that can be loaded by the processor and executed as described above for the selective assembly method of RF coaxial connectors based on phase and impedance matching.

[0147] This application also discloses a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described above regarding the selective assembly method of a radio frequency coaxial connector based on phase and impedance matching. The computer-readable storage medium includes, for example, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

Claims

1. A selective assembly method for RF coaxial connectors based on phase and impedance matching, characterized in that, include: For all part types involved in connector assembly, based on preset measuring instruments, the batch of parts corresponding to each part type are inspected to obtain the key measurement parameters of each part in the batch. Key measurement parameters are input into a preset electrical performance impact assessment model, so that the electrical performance impact assessment model outputs the electrical performance impact of each component; wherein, the electrical performance impact refers to the impact of each component on the key electrical performance of the connector when participating in connector assembly; the electrical performance impact includes at least phase impact and impedance impact. Based on the electrical performance impact of each component and the preset matching strategy, a single component is selected from the batch of components of each component type to be combined with a single component selected from other component types to obtain an optimal combination, and the following conditions are met: the key electrical performance of the finished connector assembled from all single components of different component types included in the optimal combination meets the preset application requirements. The preferred combination is output for the processing personnel to know so that the parts in the preferred combination can be assembled to obtain a connector.

2. The selective assembly method for RF coaxial connectors based on phase and impedance matching according to claim 1, characterized in that, The step of selecting a single part from a batch of parts of each part type and combining it with single parts selected from other part types to obtain an optimal combination, based on the electrical performance impact of each part and a preset matching strategy, includes: Based on a predefined assembly sequence and preset assembly modules, each part type included in the assembly sequence is sequentially taken as a target part type, and the optimal part is matched for the target part type; wherein, the assembly sequence specifies the order in which parts of all part types are assembled into finished connectors; When the optimal part is matched for the target part type, the optimal part corresponding to the target part type is virtually combined with the current assembly module to form a new assembly module, thereby updating the assembly module and synchronously virtually combining the electrical performance impact of the new assembly module; wherein, before matching the optimal part for the first part type in the assembly sequence, the preset electrical performance impact of the assembly module is zero; The process of matching the target part type to obtain the optimal part includes the following steps: All candidate parts in the batch of parts of the target sequence part type are matched one by one with the current assembly module, and the candidate parts that meet the preset matching strategy are selected as the optimal parts; wherein, the preset matching strategy aims to minimize the impact of the electrical performance of the new assembly module formed by the virtual combination of the current assembly module and the corresponding candidate parts.

3. The selective assembly method for RF coaxial connectors based on phase and impedance matching according to claim 2, characterized in that, The preset matching strategy is configured such that when selecting the optimal part for a target part type, the electrical performance impact of the new assembly module must meet the preset application requirement intensity; wherein, the application requirement intensity is dynamically determined based on the importance weight factor predefined for the target part type; the importance weight factor is used to quantify the degree of impact of the target part type on the key electrical performance of the finished connector.

4. The selective assembly method for RF coaxial connectors based on phase and impedance matching according to claim 1, characterized in that, The process involves selecting individual parts from a batch of parts of each part type based on the electrical performance impact of each component and a preset matching strategy. These individual parts are then combined with selected individual parts from other part types to obtain an optimal combination. This optimal combination satisfies the following: the key electrical performance of the finished connector assembled from all individual parts belonging to different part types in the optimal combination meets preset application requirements, including: Obtain the electrical performance impact of batch parts of all part types; Through a preset global optimization process, with preset application requirements as the optimization target, based on the electrical performance impact, several component combinations are generated, and K preferred combinations are selected from all the component combinations to form a batch matching scheme. Wherein, the part combination refers to the set formed by selecting individual parts from the batch parts of all part types, so that a finished connector can be assembled using all the parts contained in the single part combination; K is a positive integer greater than or equal to 1, and K is the number of finished connectors to be produced; the preset application requirement is to maximize the electrical performance consistency of the K preferred combinations and maximize the proportion of the parts contained in the preferred combinations to the total number of all batch parts.

5. The selective assembly method for RF coaxial connectors based on phase and impedance matching according to claim 4, characterized in that, The process involves a pre-defined global optimization, targeting pre-defined application requirements, generating several component combinations based on the electrical performance impact, and selecting K preferred combinations from all component combinations to form a batch matching scheme, including: Using a population-based metaheuristic optimization algorithm, an initial population of multiple candidate batch matching schemes is generated. The fitness value of each candidate batch matching scheme is calculated. Based on the fitness value, selection, crossover, and mutation operations are performed on the candidate batch matching schemes in the population to generate new candidate batch matching schemes and update the population. Whenever the population is updated, the fitness value of each candidate batch matching scheme in the updated population is recalculated, and selection, crossover and mutation operations are performed on the candidate batch matching schemes in the population based on the fitness value; until the updated population meets the preset iteration termination condition, the population update is completed. From the population after the population update, the candidate batch matching scheme with the best fitness is selected as the final batch matching scheme.

6. The selective assembly method for RF coaxial connectors based on phase and impedance matching according to claim 5, characterized in that, Prior to the global optimization process, there are also experience learning and feedback steps: Using batch part data of all part types within a historical period and corresponding determined batch matching schemes, a preset strategy recommendation model is trained and optimized; the strategy recommendation model is configured to output at least one recommended matching strategy to guide part combinations to obtain part combinations based on batch part data of all part types. The step of selecting a single part from a batch of parts of each part type based on the electrical performance impact of each part and a preset matching strategy, and combining it with single parts selected from other part types to obtain an optimal combination, further includes: the key electrical performance of the finished connector assembled from all single parts of different part types included in the optimal combination meets the preset application requirements. The strategy recommendation model analyzes batches of parts of all current part types and outputs recommended matching strategies. When generating part combinations through a preset global optimization process, the recommended matching strategy is used to guide the generation process of part combinations, so that among all the generated part types, there are part combinations that conform to the recommended matching strategy.

7. The selective assembly method for RF coaxial connectors based on phase and impedance matching according to claim 6, characterized in that, The recommended matching strategy includes at least the following: when generating a part combination, a first part with a preset first electrical performance influence value or range is preferentially added to the same part combination as a second part with a preset second electrical performance influence value or range.

8. A selective assembly system for radio frequency coaxial connectors based on phase and impedance matching, characterized in that, include, The part data measurement module (601) is used to detect the batch of parts corresponding to each part type based on a preset measuring instrument for all part types involved in the connector assembly, so as to obtain the key measurement parameters of each part in the batch. The electrical performance evaluation module (602) is used to input key measurement parameters into a preset electrical performance impact evaluation model so that the electrical performance impact evaluation model outputs the electrical performance impact of each component; wherein, the electrical performance impact refers to the impact of each component on the key electrical performance of the connector when participating in the connector assembly; the electrical performance impact includes at least phase impact and impedance impact. The component pairing and combination module (603) is used to select a single component from the batch of components of each component type and combine it with a single component selected from other component types to obtain an optimal combination based on the electrical performance impact of each component and the preset matching strategy, and to satisfy that: the key electrical performance of the finished connectors formed by assembling all the single components of different component types included in the optimal combination meets the preset application requirements. The connector assembly module (604) outputs the preferred combination for the processing personnel to know so as to assemble the parts in the preferred combination to obtain a connector.

9. A selective assembly device for radio frequency coaxial connectors based on phase and impedance matching, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed as described in any one of claims 1 to 7.

Images