A method and system for adjusting connector assembly parameters

Abstract

This invention relates to the field of industrial automation assembly technology, specifically providing a method and system for controlling connector assembly parameters. The method involves acquiring precision parameter data during the insertion process; evaluating the precision parameter data based on a preset insertion precision evaluation function to calculate the precision deviation score of the insertion process; and triggering the LAHHO optimization algorithm when the precision deviation score exceeds a preset threshold. This includes inputting the current insertion force curve data and the current insertion target parameters into the LAHHO optimization algorithm to optimize the insertion target parameters and output the optimized insertion target parameters. The insertion process is then controlled based on the optimized insertion target parameters. This invention achieves dynamic adaptation and optimization of insertion parameters, effectively correcting insertion precision deviations and ensuring the stability of connector insertion precision.

Description

Technical Field

[0001] This invention relates to the field of industrial automation assembly technology, specifically to a method and system for adjusting connector assembly parameters. Background Technology

[0002] In industrial production scenarios involving automated connector insertion, insertion accuracy is a core indicator for ensuring connector assembly quality and electrical performance. Dynamic factors such as equipment operating status and material tolerances during insertion can easily cause mismatches between insertion parameters and actual operating conditions, leading to excessive insertion accuracy deviations and impacting product yield. Current solutions often employ preset fixed insertion parameter control modes or perform static optimization of insertion parameters based on conventional optimization algorithms in a single dimension. However, fixed insertion parameters cannot adapt to dynamic changes in operating conditions during insertion, making it difficult to maintain stable insertion accuracy. Conventional optimization algorithms lack customized design for insertion process characteristics, resulting in insufficient coordination between global search and local development, limited adaptability and accuracy of parameter optimization, and an inability to quickly output optimal parameters matching the current insertion process, hindering the dynamic and precise control of insertion parameters. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a connector assembly parameter control method and system. By acquiring precision parameter data from the insertion process, and quantifying the precision deviation score based on a preset insertion precision evaluation function, the system determines the insertion precision status. When the precision deviation score exceeds a preset threshold, a LAHHO optimization algorithm specifically configured for the insertion process is triggered. The current insertion force curve data and insertion target parameters are used as inputs to optimize the parameters. Based on the optimized insertion target parameters output by the algorithm, the insertion process is precisely controlled, achieving dynamic adaptation and optimization of insertion parameters. This effectively corrects insertion precision deviations and ensures the stability of connector insertion precision.

[0004] In a first aspect, the technical solution of the present invention provides a connector assembly parameter adjustment method, comprising the following steps: Obtain precision parameter data during the insertion process; The accuracy parameter data is evaluated based on a preset insertion accuracy evaluation function, and the accuracy deviation score of the insertion process is calculated. When the accuracy deviation score exceeds a preset threshold, the LAHHO optimization algorithm is triggered, including: inputting the current insertion force curve data and the current insertion target parameters into the LAHHO optimization algorithm, optimizing the insertion target parameters, and outputting the optimized insertion target parameters; wherein, the LAHHO optimization algorithm is configured to: use an adaptive AVOA strategy for global search, use a Levy flight disturbance enhanced hard encirclement strategy for local development, and perform directional migration to the global optimum for elite individuals every preset number of generations; The insertion process is controlled based on the optimized insertion target parameters.

[0005] Secondly, the technical solution of the present invention provides a connector assembly parameter control system, comprising: The precision parameter data acquisition module is used to acquire precision parameter data during the insertion process; The accuracy deviation score calculation module is used to evaluate the accuracy parameter data based on a preset insertion accuracy evaluation function and calculate the accuracy deviation score of the insertion process. The insertion target parameter optimization module is used to trigger the LAHHO optimization algorithm when the accuracy deviation score exceeds a preset threshold. It includes: inputting the current insertion force curve data and the current insertion target parameters into the LAHHO optimization algorithm, dynamically optimizing the insertion target parameters, and outputting the optimized insertion target parameters. The LAHHO optimization algorithm is configured to: use an adaptive AVOA strategy for global search, use a Levy flight disturbance enhanced hard encirclement strategy for local development, and perform directional migration to the global optimum for elite individuals every preset number of generations. The insertion control module is used to control the insertion process based on the optimized insertion target parameters.

[0006] As can be seen from the above technical solutions, this application has the following advantages: By collecting full-dimensional precision parameter data of the insertion process, and calculating the precision deviation score based on the preset insertion precision evaluation function, the precision status of insertion can be determined. Precision anomalies in the insertion process can be identified, providing a basis for triggering insertion parameter adjustment actions, improving the timeliness and pertinence of parameter adjustment, and avoiding batch assembly defects caused by the failure to identify precision deviations in time. When the accuracy deviation score exceeds the preset threshold, the LAHHO optimization algorithm is triggered, and the current insertion force curve data and insertion target parameters are used as inputs for parameter optimization. This avoids the defects of traditional fixed parameter mode and conventional static optimization in terms of working condition adaptation, so that the optimized insertion parameters can accurately match the state of equipment, materials, tooling, etc., correct the insertion accuracy deviation, and ensure the adaptability of the insertion process to the accuracy requirements. The LAHHO optimization algorithm underwent targeted configuration of the instrumentation process. During the exploration phase, an adaptive AVOA strategy was adopted to enhance the global search capability, which can fully traverse the reasonable value range of instrumentation parameters and avoid missing the optimal parameter interval. During the development phase, a hard encirclement strategy enhanced by Levy flight perturbation was adopted to improve the local development accuracy, which can finely mine the potential optimal parameter interval. At the same time, a directional migration to the global optimum was performed on elite individuals, which can accelerate the convergence speed of the algorithm. The three aspects work together to solve the problems of insufficient global optimization, easy getting trapped in local optima, and low convergence efficiency of conventional optimization algorithms. This achieves global, efficient, and accurate optimization of instrumentation target parameters, ensuring that the output optimized parameters can effectively improve instrumentation accuracy and control the instrumentation deviation within the allowable range. Attached Figure Description

[0007] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic flowchart of a connector assembly parameter adjustment method provided in an embodiment of the present invention.

[0009] Figure 2 A flowchart illustrating the LAHHO optimization algorithm.

[0010] Figure 3 Comparison chart for optimizing AVOA strategies.

[0011] Figure 4 This is a schematic block diagram of a connector assembly parameter control system provided in an embodiment of the present invention. Detailed Implementation

[0012] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0014] Figure 1 This is a schematic flowchart of a connector assembly parameter adjustment method provided in an embodiment of the present invention. Figure 1 The executing entity can be a connector assembly parameter control system. The connector assembly parameter control method provided in this embodiment is executed by a computer device, and correspondingly, the connector assembly parameter control system runs within the computer device. Depending on different requirements, the order of the steps in this flowchart can be changed, and some steps can be omitted.

[0015] like Figure 1 As shown, the method includes the following steps.

[0016] S1, obtain the accuracy parameter data of the insertion process.

[0017] In this embodiment, accuracy parameter data refers to various quantitative indicators used to evaluate the connector insertion quality. Based on their physical meaning, they can be divided into three categories: positional deviation parameters, morphological deviation parameters, and stress deviation parameters. Positional deviation parameters may include center distance deviation in the X direction, center distance deviation in the Y direction, insertion height deviation in the Z direction, and angular deviation around the Z-axis. Morphological deviation parameters refer to the geometric parameter deviations of the connector itself, which may include the relative positional deviation between the connector shell and the pins, pin end face flatness deviation, pin perpendicularity deviation, and connector shell end face parallelism deviation. These geometric parameters affect the mating accuracy and contact reliability of the connector and the mating terminals. Stress deviation parameters may include peak stress deviation during insertion, residual stress deviation after insertion, and stress distribution uniformity deviation.

[0018] The control method in this embodiment can be applied to both batch optimization and process optimization scenarios.

[0019] (1) Optimization scenarios between batches After the insertion process is completed, the actual assembly accuracy data of the connector is obtained through offline testing equipment.

[0020] In some optional implementations, the positional deviation parameter can be determined by using a machine vision system to photograph and process the connector after assembly. First, the camera's intrinsic and extrinsic parameters are calibrated to establish a mapping between pixel coordinates and world coordinates. Then, the actual image coordinates of the connector's bottom center point are extracted using an image recognition algorithm and converted to actual spatial coordinates. Simultaneously, the connector's axial direction is extracted, and the actual axial direction vector is calculated. The actual coordinates are then compared with the theoretical design coordinates and the theoretical axial direction vector to calculate the positional deviation parameter.

[0021] In some optional implementations, regarding the topographic deviation parameters, a 3D laser scanner is used to scan the connector after insertion to obtain high-precision point cloud data. The point cloud data is preprocessed, including denoising, smoothing, and registration. A surface reconstruction algorithm is used to reconstruct the 3D model of the connector, extracting actual geometric parameters, including: coaxiality deviation between the connector shell and the pin centerline, height difference between the pin end face and the connector reference surface, coplanarity deviation of the pin array, and flatness deviation of key mating surfaces of the connector shell. The actual geometric parameters are compared with theoretical design parameters to calculate the topographic deviation parameters.

[0022] In some alternative implementations, regarding the stress deviation parameter, strain gauges are attached to key parts of the connector before insertion; during insertion, stress-time series data are collected in real time using a dynamic strain gauge; after insertion, peak stress, residual stress, and stress distribution uniformity are extracted. The actual stress characteristic values ​​are compared with the theoretical design allowable values ​​to calculate the stress deviation parameter.

[0023] (2) Optimize the scenario during the process During the insertion process, the final assembly accuracy needs to be predicted in advance to achieve online control. In some optional implementations, force sensors deployed on the insertion equipment collect insertion force curve data in real time. This data is in time series format, recording the changes in insertion force from the start of insertion to the current moment. The acquisition frequency is set according to actual needs, ranging from 1kHz to 10kHz, to ensure that key feature points in the insertion process can be captured. The collected raw data undergoes preprocessing, including: removing sensor noise, normalizing to eliminate dimensional effects, and truncating a fixed-length effective data segment for model input.

[0024] This scenario employs a Diff-STSI prediction model based on a conditional diffusion model. This model, trained on historical data, can extract deep features from partial insertion force curves, enabling accurate prediction of the final accuracy. The core architecture of the model includes the following modules.

[0025] (1) Adaptive Conditional Guidance Module This module uses the KShape clustering algorithm to cluster historical insertion time-series data, extracting typical curve shapes under various operating conditions as cluster centers. The cluster centers, as conditional variables, play a guiding role in the model prediction process, ensuring that the prediction results conform to actual physical laws. The number of cluster centers is determined based on the elbow rule and profile coefficient, and can be set to 6 categories, corresponding to different operating condition types such as normal insertion, incomplete contact, and insufficient insertion stroke.

[0026] For the real-time collected partial insertion force curves, the normalized cross-correlation metric is first used to calculate the similarity between the curves and each cluster center. The cluster center with the highest similarity is selected as the condition variable for the current working condition, so that the model can adopt the corresponding prediction strategy for different working conditions.

[0027] (2) DFDR-UNet network The DFDR-UNet network is an improved network structure based on the U-Net framework, incorporating a deep feature decomposition and reconstruction module. This network contains four encoders and four decoders, with the specific structure as follows.

[0028] Encoder: Each encoder consists of a multi-scale feature fusion layer and a composite convolutional block. The multi-scale feature fusion layer enhances feature representation through cross-layer feature interaction; the composite convolutional block uses 5×1 large kernel convolution to extract wide-domain features, combines channel and spatial attention mechanisms to focus on key features, and expands the receptive field through dilated convolution. The encoder performs layer-by-layer feature extraction on a portion of the input insertion force curve to obtain a multi-scale deep feature representation.

[0029] Bottleneck Layer and Deep Feature Decomposition and Reconstruction Module: In the bottleneck layer following the encoder, intermediate features undergo trend-peak dual-path decomposition. The trend component is extracted using adaptive average pooling to capture the average variation patterns in the data; the peak component is extracted using adaptive max pooling to emphasize key peak points during the insertion process. After concatenating and linearly projecting the trend and peak features, a multi-head attention mechanism is used to enhance the association of the decomposed features. A gating mechanism is then used to fuse the attention output with the original features to obtain the enhanced feature representation.

[0030] Decoder: The decoder consists of transposed convolutional blocks and uses nearest-neighbor interpolation to progressively recover the feature dimensions of the continuous time series. During decoding, skip connections are used to fuse features from the corresponding layer of the encoder with features from the current decoding layer, preserving the fine-grained information extracted during the encoding stage. Finally, a 1×1 convolution maps the features to the target dimension, outputting an estimated noise distribution.

[0031] (3) Accuracy prediction model The deep feature representation output by the DFDR-UNet network is further input into the PP-LSTM-CNN hybrid network to complete the mapping from deep features to the predicted values ​​of accuracy parameters. This hybrid network first extracts local temporal features through two CNN layers, then uses LSTM layers to model long-term time dependencies, and finally outputs the joint prediction results of multi-dimensional accuracy parameters through a linear mapping layer, including positional deviation parameters, topographic deviation parameters, stress deviation parameters, etc.

[0032] The training of the Diff-STSI model is divided into two stages.

[0033] Phase 1: Data Augmentation and Feature Learning. To address the data imbalance problem of having more normal samples and fewer abnormal samples during the instrumentation process, the model first progressively adds Gaussian noise to the original samples through a forward diffusion process, constructing a Markov noise chain. During training, the model learns inverse denoising capabilities, i.e., recovering the original signal from the noisy signal. The training objective is to minimize the error between the estimated noise and the actual added noise, while using cluster centers as conditional variables to guide the denoising direction. After training, the model can generate high-quality augmented samples that conform to the original data distribution, effectively expanding the data volume for abnormal operating conditions.

[0034] Phase Two: Predictive Model Training. Based on the enhanced complete dataset, dynamic masking technology is used to simulate partial observation states during the actual instrumentation process. For each complete time series data point, the last part is randomly masked (simulating future data that has not yet been collected). The masked portion of the data is then input into the model, and the model is trained to predict the complete instrumentation curve and final accuracy parameters from the partial observation data. The training loss function uses a combination of root mean square error and dynamic time warping distance, simultaneously optimizing the accuracy of the predicted curve shape and accuracy parameter values.

[0035] During real-time instrumentation, accuracy prediction is performed according to the following steps: Collect partial insertion force curves from the start of insertion to the current moment. ;calculate Based on the similarity to the KShape cluster centers obtained during training, the cluster center with the highest similarity is selected as the condition variable for the current working condition. A portion of the insertion force curve is associated with the condition variable and input into the DFDR-UNet network. Multi-scale features are extracted by the encoder, and trend-peak dual-path decomposition and reconstruction are performed by the deep feature decomposition and reconstruction module. The DFDR-UNet network outputs a deep feature representation of the reconstructed complete insertion force curve. This deep feature representation is then input into a PP-LSTM-CNN hybrid network, which outputs the predicted accuracy parameters at the final completion of the insertion process. .

[0036] S2, based on the preset insertion accuracy evaluation function, the accuracy parameter data is evaluated to calculate the accuracy deviation score of the insertion process.

[0037] S2.1, standardize the accuracy parameter data to obtain the standardized deviation value of each accuracy parameter.

[0038] To eliminate the influence of different dimensions on the evaluation results, the predicted values ​​of each accuracy parameter are first standardized to obtain standardized deviation values. The standardized deviation value is a dimensionless quantity, and its magnitude directly reflects the degree to which the parameter deviates from the ideal state: a deviation value of 0 indicates that it fully meets the standard, and a larger deviation value indicates a more serious deviation.

[0039] S2.2, based on the preset insertion accuracy evaluation function, a hierarchical weighted evaluation strategy is used to comprehensively calculate the standardized deviation value; the hierarchical weighted evaluation strategy includes at least two different evaluation levels, and different evaluation levels correspond to different weight coefficients and different deviation aggregation methods.

[0040] S2.3, based on the preset hierarchical weighting coefficients, performs weighted calculation on the standardized deviation values ​​to obtain the accuracy deviation score.

[0041] Considering that different precision parameters have varying degrees of impact on insertion quality, and that some parameters are inherently correlated, this embodiment adopts a hierarchical weighted evaluation strategy, dividing the precision parameters into multiple evaluation levels.

[0042] In some optional implementations, the hierarchical weighted evaluation strategy includes a first level and a second level. The first level performs a weighted summation of all accuracy parameters to reflect the overall deviation level, while the second level takes the maximum deviation for each dimension's accuracy parameter to reflect the most severe defect in each dimension. Accordingly, the accuracy deviation score is calculated using the following preset instrumentation accuracy evaluation function:

[0043] in, For the first Predicted values ​​for each precision parameter; For the first Standard values ​​for each precision parameter; For the first The allowable fluctuation range of each precision parameter; For the first The first-level evaluation weights of each accuracy parameter satisfy... ; For the first The evaluation weight coefficients for each dimension, ,satisfy ; For the first The set of accuracy parameter indicators corresponding to each dimension, each They do not overlap; For in set The index that maximizes the standardization deviation; This represents the total number of precision parameters; The total number of dimensions is set according to actual process requirements; This represents the total weight of the first level, indicating the proportion of "overall bias" in the precision bias score. The second-level total weight represents the proportion of "maximum dimensional deviation" in the precision deviation score, satisfying... The parameters are preset by process engineers based on the importance of "overall deviation" and "maximum dimensional deviation".

[0044] The aforementioned weighting coefficients can be determined using any one of the expert scoring method, entropy weighting method, or analytic hierarchy process. Furthermore, multiple weighting coefficients can be pre-set in the system for different product models and operating conditions, and can be loaded as needed during actual application.

[0045] It should be noted that the second level divides the evaluation dimensions according to the physical meaning or process impact of the accuracy parameters. The most severe deviation value is extracted separately for each dimension and weighted calculations are performed. This aims to avoid a situation where a severe deviation in a core parameter within a single dimension is masked by the overall parameter average. In this embodiment, the dimensions of the second level include position deviation, stress deviation, and morphology deviation. For example, for the position deviation dimension, the maximum standardized deviation of all position parameters under this dimension is taken, such as X deviation of 0.6 and Z deviation of 0.75. The maximum deviation of this dimension is 0.75, representing the worst deviation state of the position dimension.

[0046] The accuracy deviation score is calculated. Then, compare it with a preset threshold. Comparison: like The system determines that the accuracy of the current insertion process meets the requirements, and there is no need to trigger parameter optimization. like If the current insertion process is found to be out of tolerance, the LAHHO optimization algorithm is triggered to optimize the parameters.

[0047] threshold The product's precision requirements and process capability index are preset in the system, with a value range typically between [0.5, 1.5], and the specific value is determined based on the product model.

[0048] S3, when the accuracy deviation score exceeds the preset threshold, the LAHHO optimization algorithm is triggered, including: inputting the current insertion force curve data and the current insertion target parameters into the LAHHO optimization algorithm, optimizing the insertion target parameters, and outputting the optimized insertion target parameters; wherein, the LAHHO optimization algorithm is configured to: use an adaptive AVOA strategy for global search, use a Levy flight disturbance enhanced hard encirclement strategy for local development, and perform directional migration to the global optimum for elite individuals every preset number of generations.

[0049] When the accuracy deviation score Exceeding the preset threshold At this point, the system triggers the LAHHO optimization algorithm to dynamically optimize the insertion target parameters. The insertion target parameters include the insertion force F and the termination displacement Y. The optimization objective is to find the optimal combination of target parameters in the current insertion state, so as to minimize the accuracy deviation score after insertion.

[0050] Let the partial insertion force curve collected in real time during the current insertion process be... The preset insertion force is The preset termination displacement is The input-output relationship of the LAHHO optimization algorithm is expressed as follows:

[0051] in For the optimized insertion force, This is the optimized termination displacement.

[0052] The mathematical form of the optimization problem can be expressed as:

[0053] In this embodiment, the LAHHO algorithm is an improvement on the traditional HHO algorithm by introducing the Levy flight disturbance mechanism, adaptive AVOA strategy and elite migration strategy. Figure 2 The flowchart for the LAHHO optimization algorithm includes the following steps.

[0054] S3.1, Initialize a population containing several individuals, each representing a set of candidate instrumentation target parameters.

[0055] Initialize a population containing M individuals. Each individual This represents a set of candidate combinations of instrumentation target parameters.

[0056] Population initialization employs a combination of two strategies, including: using a first preset probability. Individuals are randomly generated within the parameter value range to ensure population diversity; with a second preset probability. By querying the knowledge base, a curve related to the current insertion force was retrieved. and current instrumentation target parameters Matching similar historical success cases, and using the optimized instrumentation target parameters corresponding to the historical success cases as the initial individuals, improves search efficiency.

[0057] First preset probability It is the probability that an individual is generated randomly in the population. The proportion of individuals is randomly generated within the parameter value range to ensure population diversity and prevent the algorithm from getting trapped in local optima.

[0058] With the first preset probability Individuals are randomly generated within the parameter value range. The range of insertion force is [value missing]. The range of values ​​for the termination displacement is: The formula for random generation is:

[0059]

[0060] in, It is a random number uniformly distributed in the interval (0,1).

[0061] Second preset probability It is the probability that an individual is generated through a knowledge base-guided method, that is, the probability that there are individuals in the population. The proportion of individuals is determined by querying the knowledge base and retrieving historical successful cases similar to the current working conditions. These cases are then used as initial individuals with optimized parameters to improve search efficiency and convergence speed.

[0062] Retrieve the current insertion force curve from the knowledge base. Similar historical instrumentation cases are identified. Similarity is calculated using a curve similarity metric, employing either Dynamic Time Warping (DTW) or Normalized Cross-Correlation (NCC). From the K most similar historical cases retrieved, the optimized instrumentation target parameters are extracted. Based on the parameters of these historical success stories, a small random perturbation was added as the initial individual:

[0063]

[0064] in, It is Gaussian noise. The amplitude of the disturbance.

[0065] For example, Take 0.3, A ratio of 0.7 is used, meaning 30% of individuals are generated randomly, and 70% are generated guided by a knowledge base. By combining these two strategies, both population diversity is ensured, and historical experience is utilized to improve the quality of the search starting points.

[0066] After each parameter optimization, the system combines the process parameters for this insertion process. Insertion process data Accuracy assessment results and optimized parameters Feedback is sent to the knowledge base to update the association mapping function. :

[0067] in, This is a knowledge update function that can modify and optimize the original mapping relationship based on the characteristics of new data. As the assembly task continues, the knowledge base gradually accumulates complete experience data containing all assembly scenarios.

[0068] At the same time, initialize the speed of each individual. Each individual independently generates initial escape energy. .

[0069] S3.2, evaluate the fitness value of each individual, the fitness value is the accuracy deviation score of the insertion based on the insertion target parameters corresponding to the individual, and initialize the individual's historical best position and global best position.

[0070] During the fitness assessment phase, for each candidate individual It is necessary to evaluate its suitability as a target parameter for instrumentation, i.e., to calculate its fitness value. Fitness evaluation is implemented based on a knowledge base.

[0071] S3.2.1, retrieve similar historical insertion cases from the knowledge base that match the current insertion force curve and the corresponding insertion target parameters for this individual.

[0072] Retrieve the current insertion force curve from the knowledge base. and candidate parameters Similar historical instrumentation examples. Similarity calculations comprehensively consider both curve shape similarity and parameter similarity:

[0073] in, is the weighting coefficient, CurveSim is the curve similarity (which can be calculated using dynamic time warping or normalized cross-correlation), and ParamSim is the parameter similarity (which can be calculated using Euclidean distance).

[0074] S3.2.2, Based on the retrieved similar historical instrumentation cases, estimate the accuracy achieved by using the corresponding instrumentation target parameters for this individual.

[0075] Based on similar cases retrieved, candidate parameters are estimated. The accuracy that can be achieved by inserting.

[0076] If there are historical cases with similarity higher than the threshold, the average of the accuracy results of these cases is taken as the estimate. , represented as:

[0077] in, For the set of cases with similarity higher than a threshold, This is the insertion accuracy result corresponding to case c.

[0078] If there are not enough similar cases, then the association mapping function in the knowledge base will be used. The estimation function expresses the mapping relationship between process parameters, process data, and accuracy results. It can be learned from a knowledge base using data mining algorithms (such as regression analysis and neural networks), and is represented as:

[0079] S3.2.3, Based on the estimated accuracy results, calculate the fitness value of the individual using the instrumentation accuracy evaluation function, wherein the fitness value is the accuracy deviation score corresponding to the individual.

[0080] Based on the estimated accuracy results Calculate the fitness value of this individual:

[0081] in, For the aforementioned instrumentation accuracy evaluation function, This refers to the standard values ​​and allowable fluctuation ranges of each precision parameter obtained from statistics in the knowledge base. These are the preset weighting coefficients.

[0082] Set each individual's current position as its historical best position. Set the position of the individual with the lowest fitness among all individuals as the global optimum. .

[0083] S3.3, update the escape energy of each individual based on the current iteration number.

[0084] For each individual, based on its initial escape energy Update current escape energy:

[0085] in, This represents the current iteration number. This represents the maximum number of iterations.

[0086] S3.4, Select the corresponding position update strategy according to the escape energy of each individual, including: when the escape energy is greater than or equal to the first energy threshold, execute the exploration phase and update the individual position using the adaptive AVOA strategy; when the escape energy is less than the first energy threshold but greater than or equal to the second energy threshold, execute the soft encirclement phase and update the individual position using the soft encirclement strategy of the standard HHO algorithm; when the escape energy is less than the second energy threshold, execute the Levy enhanced hard encirclement phase and update the individual position by adding Levy flight disturbances with a preset probability on the basis of the standard hard encirclement.

[0087] For each individual According to its own The value is selected based on the appropriate location update strategy.

[0088] Branch 1: Exploration Phase ) A global search is performed using an adaptive AVOA strategy, and the speed update formula is as follows:

[0089] in, For the first The speed of each individual, i.e., the step size and direction of position updates. For the first The historical best position of an individual The optimal position globally. As a learning factor, This represents the current iteration number.

[0090] Based on the updated velocity, the individual location is updated as follows: .

[0091] Figure 3 The graph shows a comparison of optimization using the AVOA strategy. The horizontal axis (X-axis, range -4 to 4) and the vertical axis (Y-axis, range -4 to 4) together form a two-dimensional solution space search plane. Each point (x, y) on the plane represents a candidate solution to the optimization problem (i.e., a set of process parameters to be optimized, such as the normalized values ​​of the insertion force F and the termination displacement Y). The orange dots represent the positions of individuals in the optimization process of the original algorithm (without the AVOA strategy), showing a more dispersed search path. The blue dots represent optimization using the AVOA strategy, showing a faster convergence to the global optimum. Introducing the AVOA strategy can help the original algorithm approach the global optimum in a shorter time.

[0092] Branch 2: Soft Encirclement Phase ( ) A soft encirclement strategy using the standard HHO algorithm:

[0093] in, , It is a random number uniformly distributed in the interval (0,1).

[0094] Branch 3: Levy-enhanced hard enclosure stage ( ) First, perform a standard hard bounding position update:

[0095] Then, a Levy flight perturbation is added with a preset probability of 50%, and the Levy step size is generated by the following formula:

[0096] in, For a random value in the range (0, 2), the parameter... and A random number that follows a normal distribution is represented as:

[0097] in, , The standard deviation of a normal distribution is expressed as:

[0098] in, This is the Gamma function.

[0099] Finally, the individual position update formula using Levy flight perturbation is:

[0100] in, This is the step size factor, controlling the perturbation amplitude. During the Levy-enhanced hard encirclement phase, the aforementioned Levy perturbation is executed with a 50% probability.

[0101] S3.5, every preset number of generations, perform a directed migration to the global optimum for the top-ranked elite individuals in terms of fitness.

[0102] If the current iteration number If the value is a multiple of 10, then perform an elite migration operation: select the top 10% of individuals in terms of fitness from the current population as elite individuals, and perform a directed migration towards the global optimum for each elite individual:

[0103] in, The elite migration factor can be set to 0.5.

[0104] Then, the fitness of the elite individuals after migration is reassessed, and the global optimum position is re-determined based on the updated fitness. .

[0105] Furthermore, boundary checks are performed on the updated individuals to ensure that the insertion force and termination displacement are within the preset physically feasible range. The fitness value of each individual is then recalculated according to the method in step S3.2, and the historical best position of the individual is updated. and global optimal position .

[0106] S3.6 Repeat steps S3.3 to S3.5 until the maximum number of iterations is reached. Output the insertion force and termination displacement corresponding to the global optimal position as the optimized insertion target parameters.

[0107] S4, adjusts the insertion process according to the optimized insertion target parameters.

[0108] For real-time control during the process, when the system predicts that the current parameters may lead to excessive accuracy and completes the optimization calculation, the following real-time control steps are immediately executed.

[0109] Specifically, the control system receives the optimization parameters output by the LAHHO algorithm. The validity of the data is then checked. If the check fails, the data is truncated according to the safety boundary value, and the anomaly is recorded.

[0110] To avoid shocks or oscillations caused by sudden parameter changes, a gradual adjustment strategy is adopted. Smooth adjustment of the insertion force is expressed as follows:

[0111] in, The transition time is set to 50-100ms to ensure a smooth force change.

[0112] The dynamic update of the termination displacement involves writing the new termination displacement into the controller's target position register, which is then automatically tracked by the position loop of the servo system.

[0113] Then, multi-parameter coordinated adjustment is performed, and parameter adjustment is executed in priority order. First, the smoothed force target value is written into the force control algorithm in real time to adjust the output torque of the servo motor. Then, while the force adjustment is stable, the target value of the position controller is updated to ensure that the insertion head stops when it reaches the new target displacement.

[0114] For batch parameter updates, after insertion is completed, the preset parameters for the next batch of insertion are optimized and updated based on the actual accuracy data of the test.

[0115] Specifically, the optimized parameters received from the LAHHO algorithm are based on the actual accuracy output of this insertion. The validity of the optimized parameters is then verified. The verified parameters are stored in the knowledge base and associated with the operating conditions of this insertion.

[0116] Based on the optimization effect and parameter reliability, decide whether to apply the optimized parameters to subsequent batches. If the optimization effect is significant and the parameters are stable, directly apply the optimized parameters. Write the parameters into the process parameter table as preset parameters for the next batch; if the optimization is significant, use a gradual update approach.

[0117] in The learning rate is used to avoid quality fluctuations caused by sudden parameter changes.

[0118] In the next batch of production, collect the insertion force curve and accuracy data under the new parameters, calculate the actual accuracy deviation score, compare it with the historical score, evaluate the optimization effect, and if the effect is not as expected, roll back to the original parameters and re-analyze.

[0119] To test the performance of the optimized algorithm, the IEEE CEC 2021 test suite was used to compare LAHHO with some advanced and classic competitors. Six of the test standard functions were selected for evaluation, as shown in Table 1.

[0120] Table 1: Benchmark Functions

[0121] The population size for each algorithm was set to 100, and the number of algorithm iterations was 1000. The dimension and value range of the benchmark test function solution are shown in Table 1, and the parameters of the other algorithms are shown in Table 2.

[0122] Table 2: Initial Parameter Settings for the Algorithm

[0123] Each algorithm was run independently 30 times on each benchmark function to obtain the optimal solution, average value, and standard deviation of the function. The results are shown in Table 3.

[0124] Table 3: Algorithm Performance Comparison

[0125] As shown in Table 3, the LAHHO algorithm achieves the best results in all evaluation metrics for test functions F1-F6, followed by HHO. Specifically, it finds the optimal solution in each iteration for F2, F3, and F4, with both the mean and variance being 0, demonstrating the stability of the LAHHO algorithm in solving complex problems. Traditional algorithms such as DE, PSO, and GA perform poorly on complex functions, finding optimal solutions with significant deviations. The average values ​​of each group indicate that although the LAHHO algorithm can obtain the global optimum for test functions F1, F5, and F6, local optima may still occur during the iteration process. Overall, LAHHO exhibits strong optimization ability and stability, demonstrating higher performance and fitness when handling various test functions.

[0126] The above text provides a detailed description of an embodiment of a connector assembly parameter control method. Based on the connector assembly parameter control method described in the above embodiment, this invention also provides a connector assembly parameter control system corresponding to the method.

[0127] Figure 4 This is a schematic block diagram of a connector assembly parameter control system provided in an embodiment of the present invention. In this embodiment, the connector assembly parameter control system can be divided into multiple functional modules according to the functions it performs. A module, as referred to in this invention, is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and is stored in memory.

[0128] The precision parameter data acquisition module is used to acquire precision parameter data during the insertion process.

[0129] The accuracy deviation score calculation module is used to evaluate the accuracy parameter data based on a preset insertion accuracy evaluation function and calculate the accuracy deviation score of the insertion process.

[0130] The insertion target parameter optimization module is used to trigger the LAHHO optimization algorithm when the accuracy deviation score exceeds a preset threshold. It includes: inputting the current insertion force curve data and the current insertion target parameters into the LAHHO optimization algorithm, dynamically optimizing the insertion target parameters, and outputting the optimized insertion target parameters. The LAHHO optimization algorithm is configured to: use an adaptive AVOA strategy for global search, use a Levy flight disturbance enhanced hard encirclement strategy for local development, and perform directional migration to the global optimum for elite individuals every preset number of generations.

[0131] The insertion control module is used to control the insertion process based on the optimized insertion target parameters.

[0132] The connector assembly parameter control system of this embodiment is used to implement the aforementioned connector assembly parameter control method. Therefore, the specific implementation of this system can be found in the embodiment section of the connector assembly parameter control method above. Thus, the specific implementation can be referred to the description of the corresponding embodiments, and will not be elaborated here.

[0133] Furthermore, since the connector assembly parameter control system in this embodiment is used to implement the aforementioned connector assembly parameter control method, its function corresponds to the function of the above method, and will not be described again here.

[0134] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for adjusting connector assembly parameters, characterized in that, Includes the following steps: Obtain precision parameter data during the insertion process; The accuracy parameter data is evaluated based on a preset insertion accuracy evaluation function, and the accuracy deviation score of the insertion process is calculated. When the accuracy deviation score exceeds a preset threshold, the LAHHO optimization algorithm is triggered, including: inputting the current insertion force curve data and the current insertion target parameters into the LAHHO optimization algorithm, optimizing the insertion target parameters, and outputting the optimized insertion target parameters; wherein, the LAHHO optimization algorithm is configured to: use an adaptive AVOA strategy for global search, use a Levy flight disturbance enhanced hard encirclement strategy for local development, and perform directional migration to the global optimum for elite individuals every preset number of generations; The insertion process is controlled based on the optimized insertion target parameters.

2. The connector assembly parameter adjustment method according to claim 1, characterized in that, The accuracy parameter data is evaluated based on a preset insertion accuracy evaluation function, and the accuracy deviation score of the insertion process is calculated, specifically including: The accuracy parameter data is standardized to obtain the standardized deviation value of each accuracy parameter; Based on a preset insertion accuracy evaluation function, a hierarchical weighted evaluation strategy is used to comprehensively calculate the standardized deviation value; the hierarchical weighted evaluation strategy includes at least two different evaluation levels, with different weight coefficients and different deviation aggregation methods corresponding to different evaluation levels. Based on preset hierarchical weighting coefficients, the standardized deviation values ​​are weighted and calculated to obtain the accuracy deviation score.

3. The connector assembly parameter adjustment method according to claim 2, characterized in that, The hierarchical weighted evaluation strategy includes a first level and a second level. The first level performs a weighted summation of all accuracy parameters, while the second level takes the maximum deviation for each accuracy parameter in different dimensions. The accuracy deviation score is then calculated using the following preset instrumentation accuracy evaluation function: in, For the first Predicted values ​​for each precision parameter; For the first Standard values ​​for each precision parameter; For the first The allowable fluctuation range of each precision parameter; For the first The first-level evaluation weights of each accuracy parameter satisfy... ; For the first The evaluation weight coefficients for each dimension, ,satisfy ; For the first A set of accuracy parameter indicators corresponding to each dimension; For in set The index that maximizes the standardization deviation; This represents the total number of precision parameters; The total number of dimensions; The total weight of the first level. This represents the total weight of the second level. .

4. The connector assembly parameter adjustment method according to claim 1, characterized in that, The current insertion force curve data and the current insertion target parameters are input into the LAHHO optimization algorithm to dynamically optimize the insertion target parameters and output the optimized insertion target parameters, specifically including: Step 1: Initialize a population containing several individuals, each representing a set of candidate instrumentation target parameters; Step 2: Evaluate the fitness value of each individual. The fitness value is the accuracy deviation score of the instrumentation based on the instrumentation target parameters corresponding to that individual, and initialize the individual's historical best position and global best position. Step 3: Update the escape energy of each individual based on the current iteration number; Step 4: Select the appropriate position update strategy based on the escape energy of each individual, including: when the escape energy is greater than or equal to the first energy threshold, execute the exploration phase and update the individual position using the adaptive AVOA strategy; when the escape energy is less than the first energy threshold but greater than or equal to the second energy threshold, execute the soft encirclement phase and update the individual position using the soft encirclement strategy of the standard HHO algorithm; when the escape energy is less than the second energy threshold, execute the Levy enhanced hard encirclement phase and update the individual position by adding Levy flight disturbances with a preset probability on the basis of the standard hard encirclement. Step 5: Every preset number of generations, perform a directed migration to the global optimum for the top-preset proportion of elite individuals in terms of fitness. Step 6: Repeat steps 3 to 5 until the maximum number of iterations is reached. Output the insertion force and termination displacement corresponding to the global optimal position as the optimized insertion target parameters.

5. The connector assembly parameter adjustment method according to claim 4, characterized in that, During the exploration phase, an adaptive AVOA strategy is used to update individual positions, and its velocity update formula is as follows: in, For the first The speed of each individual For the first The historical best position of an individual The optimal position globally. As a learning factor, This represents the current iteration number; Based on the updated velocity, the individual location is updated as follows: .

6. The connector assembly parameter adjustment method according to claim 4, characterized in that, The Levy enhanced hard bounding phase first performs the standard hard bounding position update: Then, Levy flight perturbations are added with preset probabilities, and the Levy step size is generated using the following formula: in, For a random value in the range (0, 2), the parameter... and A random number that follows a normal distribution is represented as: in, , The standard deviation of a normal distribution is expressed as: The individual position update formula using Levy flight perturbation is then: in, This is the step size factor.

7. The connector assembly parameter adjustment method according to claim 4, characterized in that, The position update formula for elite individuals performing directed migration to the global optimum is: in, It is an elite migration factor.

8. The connector assembly parameter adjustment method according to claim 4, characterized in that, Initialize a population consisting of several individuals, specifically including: Individuals are randomly generated within the parameter value range with a first preset probability; By querying the knowledge base with a second preset probability, similar historical successful cases that match the current insertion force curve and the current insertion target parameters are retrieved, and the optimized insertion target parameters corresponding to the historical successful cases are used as the initial individuals.

9. The connector assembly parameter adjustment method according to claim 4, characterized in that, Assess each individual's fitness score, specifically including: Retrieve historical insertion cases in the knowledge base that are similar to the current insertion force curve and the corresponding insertion target parameters for this individual; Based on the retrieved similar historical instrumentation cases, estimate the accuracy achieved by using the corresponding instrumentation target parameters for this individual. Based on the estimated accuracy results, the fitness value of the individual is calculated using the instrumentation accuracy evaluation function, and the fitness value is the accuracy deviation score corresponding to the individual.

10. A connector assembly parameter control system, characterized in that, include: The precision parameter data acquisition module is used to acquire precision parameter data during the insertion process; The accuracy deviation score calculation module is used to evaluate the accuracy parameter data based on a preset insertion accuracy evaluation function and calculate the accuracy deviation score of the insertion process. The insertion target parameter optimization module is used to trigger the LAHHO optimization algorithm when the accuracy deviation score exceeds a preset threshold. It includes: inputting the current insertion force curve data and the current insertion target parameters into the LAHHO optimization algorithm, dynamically optimizing the insertion target parameters, and outputting the optimized insertion target parameters. The LAHHO optimization algorithm is configured to: use an adaptive AVOA strategy for global search, use a Levy flight disturbance enhanced hard encirclement strategy for local development, and perform directional migration to the global optimum for elite individuals every preset number of generations. The insertion control module is used to control the insertion process based on the optimized insertion target parameters.

Images